projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
[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>
f46b5a66
CH		 47#include "ctree.h"
48#include "disk-io.h"
49#include "transaction.h"
50#include "btrfs_inode.h"
f46b5a66
CH		 51#include "print-tree.h"
52#include "volumes.h"
925baedd 53#include "locking.h"
581bb050 54#include "inode-map.h"
d7728c96 55#include "backref.h"
606686ee 56#include "rcu-string.h"
31db9f7c 57#include "send.h"
3f6bcfbd 58#include "dev-replace.h"
63541927 59#include "props.h"
3b02a68a 60#include "sysfs.h"
fcebe456 61#include "qgroup.h"
f46b5a66 62
abccd00f
HM		 63#ifdef CONFIG_64BIT
64/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
68 */
69struct btrfs_ioctl_timespec_32 {
70 __u64 sec;
71 __u32 nsec;
72} __attribute__ ((__packed__));
73
74struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
80 __u64 flags; /* in */
81 __u64 reserved[16]; /* in */
82} __attribute__ ((__packed__));
83
84#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
86#endif
87
88
416161db 89static int btrfs_clone(struct inode *src, struct inode *inode,
1c919a5e
MF		 90 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
91 int no_time_update);
416161db 92
6cbff00f
CH		 93/* Mask out flags that are inappropriate for the given type of inode. */
94static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
95{
96 if (S_ISDIR(mode))
97 return flags;
98 else if (S_ISREG(mode))
99 return flags & ~FS_DIRSYNC_FL;
100 else
101 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
102}
103
104/*
105 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
106 */
107static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
108{
109 unsigned int iflags = 0;
110
111 if (flags & BTRFS_INODE_SYNC)
112 iflags |= FS_SYNC_FL;
113 if (flags & BTRFS_INODE_IMMUTABLE)
114 iflags |= FS_IMMUTABLE_FL;
115 if (flags & BTRFS_INODE_APPEND)
116 iflags |= FS_APPEND_FL;
117 if (flags & BTRFS_INODE_NODUMP)
118 iflags |= FS_NODUMP_FL;
119 if (flags & BTRFS_INODE_NOATIME)
120 iflags |= FS_NOATIME_FL;
121 if (flags & BTRFS_INODE_DIRSYNC)
122 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 123 if (flags & BTRFS_INODE_NODATACOW)
124 iflags |= FS_NOCOW_FL;
125
126 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
127 iflags |= FS_COMPR_FL;
128 else if (flags & BTRFS_INODE_NOCOMPRESS)
129 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 130
131 return iflags;
132}
133
134/*
135 * Update inode->i_flags based on the btrfs internal flags.
136 */
137void btrfs_update_iflags(struct inode *inode)
138{
139 struct btrfs_inode *ip = BTRFS_I(inode);
3cc79392 140 unsigned int new_fl = 0;
6cbff00f
CH		 141
142 if (ip->flags & BTRFS_INODE_SYNC)
3cc79392 143 new_fl |= S_SYNC;
6cbff00f 144 if (ip->flags & BTRFS_INODE_IMMUTABLE)
3cc79392 145 new_fl |= S_IMMUTABLE;
6cbff00f 146 if (ip->flags & BTRFS_INODE_APPEND)
3cc79392 147 new_fl |= S_APPEND;
6cbff00f 148 if (ip->flags & BTRFS_INODE_NOATIME)
3cc79392 149 new_fl |= S_NOATIME;
6cbff00f 150 if (ip->flags & BTRFS_INODE_DIRSYNC)
3cc79392
FM		 151 new_fl |= S_DIRSYNC;
152
153 set_mask_bits(&inode->i_flags,
154 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
155 new_fl);
6cbff00f
CH		 156}
157
158/*
159 * Inherit flags from the parent inode.
160 *
e27425d6 161 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 162 */
163void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
164{
0b4dcea5
CM		 165 unsigned int flags;
166
167 if (!dir)
168 return;
169
170 flags = BTRFS_I(dir)->flags;
6cbff00f 171
e27425d6
JB		 172 if (flags & BTRFS_INODE_NOCOMPRESS) {
173 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
174 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
175 } else if (flags & BTRFS_INODE_COMPRESS) {
176 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
177 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
178 }
179
213490b3 180 if (flags & BTRFS_INODE_NODATACOW) {
e27425d6 181 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
213490b3
LB		 182 if (S_ISREG(inode->i_mode))
183 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
184 }
6cbff00f 185
6cbff00f
CH		 186 btrfs_update_iflags(inode);
187}
188
189static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
190{
496ad9aa 191 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
6cbff00f
CH		 192 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
193
194 if (copy_to_user(arg, &flags, sizeof(flags)))
195 return -EFAULT;
196 return 0;
197}
198
75e7cb7f
LB		 199static int check_flags(unsigned int flags)
200{
201 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
202 FS_NOATIME_FL | FS_NODUMP_FL | \
203 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 204 FS_NOCOMP_FL | FS_COMPR_FL |
205 FS_NOCOW_FL))
75e7cb7f
LB		 206 return -EOPNOTSUPP;
207
208 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
209 return -EINVAL;
210
75e7cb7f
LB		 211 return 0;
212}
213
6cbff00f
CH		 214static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
215{
496ad9aa 216 struct inode *inode = file_inode(file);
6cbff00f
CH		 217 struct btrfs_inode *ip = BTRFS_I(inode);
218 struct btrfs_root *root = ip->root;
219 struct btrfs_trans_handle *trans;
220 unsigned int flags, oldflags;
221 int ret;
f062abf0
LZ		 222 u64 ip_oldflags;
223 unsigned int i_oldflags;
7e97b8da 224 umode_t mode;
6cbff00f 225
bd60ea0f
DS		 226 if (!inode_owner_or_capable(inode))
227 return -EPERM;
228
b83cc969
LZ		 229 if (btrfs_root_readonly(root))
230 return -EROFS;
231
6cbff00f
CH		 232 if (copy_from_user(&flags, arg, sizeof(flags)))
233 return -EFAULT;
234
75e7cb7f
LB		 235 ret = check_flags(flags);
236 if (ret)
237 return ret;
f46b5a66 238
e7848683
JK		 239 ret = mnt_want_write_file(file);
240 if (ret)
241 return ret;
242
6cbff00f
CH		 243 mutex_lock(&inode->i_mutex);
244
f062abf0
LZ		 245 ip_oldflags = ip->flags;
246 i_oldflags = inode->i_flags;
7e97b8da 247 mode = inode->i_mode;
f062abf0 248
6cbff00f
CH		 249 flags = btrfs_mask_flags(inode->i_mode, flags);
250 oldflags = btrfs_flags_to_ioctl(ip->flags);
251 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
252 if (!capable(CAP_LINUX_IMMUTABLE)) {
253 ret = -EPERM;
254 goto out_unlock;
255 }
256 }
257
6cbff00f
CH		 258 if (flags & FS_SYNC_FL)
259 ip->flags |= BTRFS_INODE_SYNC;
260 else
261 ip->flags &= ~BTRFS_INODE_SYNC;
262 if (flags & FS_IMMUTABLE_FL)
263 ip->flags |= BTRFS_INODE_IMMUTABLE;
264 else
265 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
266 if (flags & FS_APPEND_FL)
267 ip->flags |= BTRFS_INODE_APPEND;
268 else
269 ip->flags &= ~BTRFS_INODE_APPEND;
270 if (flags & FS_NODUMP_FL)
271 ip->flags |= BTRFS_INODE_NODUMP;
272 else
273 ip->flags &= ~BTRFS_INODE_NODUMP;
274 if (flags & FS_NOATIME_FL)
275 ip->flags |= BTRFS_INODE_NOATIME;
276 else
277 ip->flags &= ~BTRFS_INODE_NOATIME;
278 if (flags & FS_DIRSYNC_FL)
279 ip->flags |= BTRFS_INODE_DIRSYNC;
280 else
281 ip->flags &= ~BTRFS_INODE_DIRSYNC;
7e97b8da
DS		 282 if (flags & FS_NOCOW_FL) {
283 if (S_ISREG(mode)) {
284 /*
285 * It's safe to turn csums off here, no extents exist.
286 * Otherwise we want the flag to reflect the real COW
287 * status of the file and will not set it.
288 */
289 if (inode->i_size == 0)
290 ip->flags |= BTRFS_INODE_NODATACOW
291 | BTRFS_INODE_NODATASUM;
292 } else {
293 ip->flags |= BTRFS_INODE_NODATACOW;
294 }
295 } else {
296 /*
297 * Revert back under same assuptions as above
298 */
299 if (S_ISREG(mode)) {
300 if (inode->i_size == 0)
301 ip->flags &= ~(BTRFS_INODE_NODATACOW
302 | BTRFS_INODE_NODATASUM);
303 } else {
304 ip->flags &= ~BTRFS_INODE_NODATACOW;
305 }
306 }
6cbff00f 307
75e7cb7f
LB		 308 /*
309 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
310 * flag may be changed automatically if compression code won't make
311 * things smaller.
312 */
313 if (flags & FS_NOCOMP_FL) {
314 ip->flags &= ~BTRFS_INODE_COMPRESS;
315 ip->flags |= BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 316
317 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
318 if (ret && ret != -ENODATA)
319 goto out_drop;
75e7cb7f 320 } else if (flags & FS_COMPR_FL) {
63541927
FDBM		 321 const char *comp;
322
75e7cb7f
LB		 323 ip->flags |= BTRFS_INODE_COMPRESS;
324 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 325
326 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
327 comp = "lzo";
328 else
329 comp = "zlib";
330 ret = btrfs_set_prop(inode, "btrfs.compression",
331 comp, strlen(comp), 0);
332 if (ret)
333 goto out_drop;
334
ebcb904d 335 } else {
78a017a2
FM		 336 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
337 if (ret && ret != -ENODATA)
338 goto out_drop;
ebcb904d 339 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 340 }
6cbff00f 341
4da6f1a3 342 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 343 if (IS_ERR(trans)) {
344 ret = PTR_ERR(trans);
345 goto out_drop;
346 }
6cbff00f 347
306424cc 348 btrfs_update_iflags(inode);
0c4d2d95 349 inode_inc_iversion(inode);
306424cc 350 inode->i_ctime = CURRENT_TIME;
6cbff00f 351 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 352
6cbff00f 353 btrfs_end_transaction(trans, root);
f062abf0
LZ		 354 out_drop:
355 if (ret) {
356 ip->flags = ip_oldflags;
357 inode->i_flags = i_oldflags;
358 }
6cbff00f 359
6cbff00f
CH		 360 out_unlock:
361 mutex_unlock(&inode->i_mutex);
e7848683 362 mnt_drop_write_file(file);
2d4e6f6a 363 return ret;
6cbff00f
CH		 364}
365
366static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
367{
496ad9aa 368 struct inode *inode = file_inode(file);
6cbff00f
CH		 369
370 return put_user(inode->i_generation, arg);
371}
f46b5a66 372
f7039b1d
LD		 373static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
374{
54563d41 375 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
f7039b1d
LD		 376 struct btrfs_device *device;
377 struct request_queue *q;
378 struct fstrim_range range;
379 u64 minlen = ULLONG_MAX;
380 u64 num_devices = 0;
815745cf 381 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 382 int ret;
383
384 if (!capable(CAP_SYS_ADMIN))
385 return -EPERM;
386
1f78160c
XG		 387 rcu_read_lock();
388 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
389 dev_list) {
f7039b1d
LD		 390 if (!device->bdev)
391 continue;
392 q = bdev_get_queue(device->bdev);
393 if (blk_queue_discard(q)) {
394 num_devices++;
395 minlen = min((u64)q->limits.discard_granularity,
396 minlen);
397 }
398 }
1f78160c 399 rcu_read_unlock();
f4c697e6 400
f7039b1d
LD		 401 if (!num_devices)
402 return -EOPNOTSUPP;
f7039b1d
LD		 403 if (copy_from_user(&range, arg, sizeof(range)))
404 return -EFAULT;
e515c18b
LC		 405 if (range.start > total_bytes ||
406 range.len < fs_info->sb->s_blocksize)
f4c697e6 407 return -EINVAL;
f7039b1d 408
f4c697e6 409 range.len = min(range.len, total_bytes - range.start);
f7039b1d 410 range.minlen = max(range.minlen, minlen);
815745cf 411 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 412 if (ret < 0)
413 return ret;
414
415 if (copy_to_user(arg, &range, sizeof(range)))
416 return -EFAULT;
417
418 return 0;
419}
420
dd5f9615
SB		 421int btrfs_is_empty_uuid(u8 *uuid)
422{
46e0f66a
CM		 423 int i;
424
425 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
426 if (uuid[i])
427 return 0;
428 }
429 return 1;
dd5f9615
SB		 430}
431
d5c12070 432static noinline int create_subvol(struct inode *dir,
cb8e7090 433 struct dentry *dentry,
72fd032e 434 char *name, int namelen,
6f72c7e2 435 u64 *async_transid,
8696c533 436 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 437{
438 struct btrfs_trans_handle *trans;
439 struct btrfs_key key;
440 struct btrfs_root_item root_item;
441 struct btrfs_inode_item *inode_item;
442 struct extent_buffer *leaf;
d5c12070 443 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 444 struct btrfs_root *new_root;
d5c12070 445 struct btrfs_block_rsv block_rsv;
8ea05e3a 446 struct timespec cur_time = CURRENT_TIME;
5662344b 447 struct inode *inode;
f46b5a66
CH		 448 int ret;
449 int err;
450 u64 objectid;
451 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 452 u64 index = 0;
d5c12070 453 u64 qgroup_reserved;
8ea05e3a 454 uuid_le new_uuid;
f46b5a66 455
581bb050 456 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 457 if (ret)
a22285a6 458 return ret;
6a912213 459
e09fe2d2
QW		 460 /*
461 * Don't create subvolume whose level is not zero. Or qgroup will be
462 * screwed up since it assume subvolme qgroup's level to be 0.
463 */
464 if (btrfs_qgroup_level(objectid))
465 return -ENOSPC;
466
d5c12070 467 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 468 /*
d5c12070
MX		 469 * The same as the snapshot creation, please see the comment
470 * of create_snapshot().
9ed74f2d 471 */
d5c12070 472 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 473 8, &qgroup_reserved, false);
d5c12070
MX		 474 if (ret)
475 return ret;
476
477 trans = btrfs_start_transaction(root, 0);
478 if (IS_ERR(trans)) {
479 ret = PTR_ERR(trans);
de6e8200
LB		 480 btrfs_subvolume_release_metadata(root, &block_rsv,
481 qgroup_reserved);
482 return ret;
d5c12070
MX		 483 }
484 trans->block_rsv = &block_rsv;
485 trans->bytes_reserved = block_rsv.size;
f46b5a66 486
8696c533 487 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 488 if (ret)
489 goto fail;
490
4d75f8a9 491 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 492 if (IS_ERR(leaf)) {
493 ret = PTR_ERR(leaf);
494 goto fail;
495 }
f46b5a66 496
5d4f98a2 497 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 498 btrfs_set_header_bytenr(leaf, leaf->start);
499 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 500 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 501 btrfs_set_header_owner(leaf, objectid);
502
0a4e5586 503 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
f46b5a66 504 BTRFS_FSID_SIZE);
5d4f98a2 505 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
b308bc2f 506 btrfs_header_chunk_tree_uuid(leaf),
5d4f98a2 507 BTRFS_UUID_SIZE);
f46b5a66
CH		 508 btrfs_mark_buffer_dirty(leaf);
509
8ea05e3a
AB		 510 memset(&root_item, 0, sizeof(root_item));
511
f46b5a66 512 inode_item = &root_item.inode;
3cae210f
QW		 513 btrfs_set_stack_inode_generation(inode_item, 1);
514 btrfs_set_stack_inode_size(inode_item, 3);
515 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 516 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 517 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 518
3cae210f
QW		 519 btrfs_set_root_flags(&root_item, 0);
520 btrfs_set_root_limit(&root_item, 0);
521 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 522
f46b5a66 523 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 524 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 525 btrfs_set_root_level(&root_item, 0);
526 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 527 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 528 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 529
8ea05e3a
AB		 530 btrfs_set_root_generation_v2(&root_item,
531 btrfs_root_generation(&root_item));
532 uuid_le_gen(&new_uuid);
533 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 534 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
535 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 536 root_item.ctime = root_item.otime;
537 btrfs_set_root_ctransid(&root_item, trans->transid);
538 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 539
925baedd 540 btrfs_tree_unlock(leaf);
f46b5a66
CH		 541 free_extent_buffer(leaf);
542 leaf = NULL;
543
544 btrfs_set_root_dirid(&root_item, new_dirid);
545
546 key.objectid = objectid;
5d4f98a2 547 key.offset = 0;
962a298f 548 key.type = BTRFS_ROOT_ITEM_KEY;
f46b5a66
CH		 549 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
550 &root_item);
551 if (ret)
552 goto fail;
553
76dda93c
YZ		 554 key.offset = (u64)-1;
555 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa 556 if (IS_ERR(new_root)) {
79787eaa 557 ret = PTR_ERR(new_root);
6d13f549 558 btrfs_abort_transaction(trans, root, ret);
79787eaa
JM		 559 goto fail;
560 }
76dda93c
YZ		 561
562 btrfs_record_root_in_trans(trans, new_root);
563
63541927 564 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 565 if (ret) {
566 /* We potentially lose an unused inode item here */
79787eaa 567 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 568 goto fail;
569 }
570
f46b5a66
CH		 571 /*
572 * insert the directory item
573 */
3de4586c 574 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 575 if (ret) {
576 btrfs_abort_transaction(trans, root, ret);
577 goto fail;
578 }
3de4586c
CM		 579
580 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 581 name, namelen, dir, &key,
3de4586c 582 BTRFS_FT_DIR, index);
79787eaa
JM		 583 if (ret) {
584 btrfs_abort_transaction(trans, root, ret);
f46b5a66 585 goto fail;
79787eaa 586 }
0660b5af 587
52c26179
YZ		 588 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
589 ret = btrfs_update_inode(trans, root, dir);
590 BUG_ON(ret);
591
0660b5af 592 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 593 objectid, root->root_key.objectid,
33345d01 594 btrfs_ino(dir), index, name, namelen);
76dda93c 595 BUG_ON(ret);
f46b5a66 596
dd5f9615
SB		 597 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
598 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
599 objectid);
600 if (ret)
601 btrfs_abort_transaction(trans, root, ret);
602
f46b5a66 603fail:
d5c12070
MX		 604 trans->block_rsv = NULL;
605 trans->bytes_reserved = 0;
de6e8200
LB		 606 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
607
72fd032e
SW		 608 if (async_transid) {
609 *async_transid = trans->transid;
610 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 611 if (err)
612 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 613 } else {
614 err = btrfs_commit_transaction(trans, root);
615 }
f46b5a66
CH		 616 if (err && !ret)
617 ret = err;
1a65e24b 618
5662344b
TI		 619 if (!ret) {
620 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 621 if (IS_ERR(inode))
622 return PTR_ERR(inode);
5662344b
TI		 623 d_instantiate(dentry, inode);
624 }
f46b5a66
CH		 625 return ret;
626}
627
9ea24bbe 628static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
8257b2dc
MX		 629{
630 s64 writers;
631 DEFINE_WAIT(wait);
632
633 do {
634 prepare_to_wait(&root->subv_writers->wait, &wait,
635 TASK_UNINTERRUPTIBLE);
636
637 writers = percpu_counter_sum(&root->subv_writers->counter);
638 if (writers)
639 schedule();
640
641 finish_wait(&root->subv_writers->wait, &wait);
642 } while (writers);
643}
644
e9662f70
MX		 645static int create_snapshot(struct btrfs_root *root, struct inode *dir,
646 struct dentry *dentry, char *name, int namelen,
647 u64 *async_transid, bool readonly,
648 struct btrfs_qgroup_inherit *inherit)
f46b5a66 649{
2e4bfab9 650 struct inode *inode;
f46b5a66
CH		 651 struct btrfs_pending_snapshot *pending_snapshot;
652 struct btrfs_trans_handle *trans;
2e4bfab9 653 int ret;
f46b5a66 654
27cdeb70 655 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
f46b5a66
CH		 656 return -EINVAL;
657
8257b2dc 658 atomic_inc(&root->will_be_snapshoted);
4e857c58 659 smp_mb__after_atomic();
9ea24bbe 660 btrfs_wait_for_no_snapshoting_writes(root);
8257b2dc 661
6a03843d
MX		 662 ret = btrfs_start_delalloc_inodes(root, 0);
663 if (ret)
8257b2dc 664 goto out;
6a03843d 665
b0244199 666 btrfs_wait_ordered_extents(root, -1);
6a03843d 667
3de4586c 668 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
8257b2dc
MX		 669 if (!pending_snapshot) {
670 ret = -ENOMEM;
671 goto out;
672 }
a22285a6 673
66d8f3dd
MX		 674 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
675 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 676 /*
677 * 1 - parent dir inode
678 * 2 - dir entries
679 * 1 - root item
680 * 2 - root ref/backref
681 * 1 - root of snapshot
dd5f9615 682 * 1 - UUID item
d5c12070
MX		 683 */
684 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 685 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 686 &pending_snapshot->qgroup_reserved,
687 false);
d5c12070 688 if (ret)
8257b2dc 689 goto free;
d5c12070 690
3de4586c 691 pending_snapshot->dentry = dentry;
f46b5a66 692 pending_snapshot->root = root;
b83cc969 693 pending_snapshot->readonly = readonly;
e9662f70 694 pending_snapshot->dir = dir;
8696c533 695 pending_snapshot->inherit = inherit;
a22285a6 696
d5c12070 697 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 698 if (IS_ERR(trans)) {
699 ret = PTR_ERR(trans);
700 goto fail;
701 }
702
8351583e 703 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 704 list_add(&pending_snapshot->list,
705 &trans->transaction->pending_snapshots);
8351583e 706 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 707 if (async_transid) {
708 *async_transid = trans->transid;
709 ret = btrfs_commit_transaction_async(trans,
710 root->fs_info->extent_root, 1);
00d71c9c
MX		 711 if (ret)
712 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 713 } else {
714 ret = btrfs_commit_transaction(trans,
715 root->fs_info->extent_root);
716 }
aec8030a 717 if (ret)
c37b2b62 718 goto fail;
a22285a6
YZ		 719
720 ret = pending_snapshot->error;
721 if (ret)
722 goto fail;
723
d3797308
CM		 724 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
725 if (ret)
726 goto fail;
727
2b0143b5 728 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
2e4bfab9
YZ		 729 if (IS_ERR(inode)) {
730 ret = PTR_ERR(inode);
731 goto fail;
732 }
5662344b 733
2e4bfab9
YZ		 734 d_instantiate(dentry, inode);
735 ret = 0;
736fail:
d5c12070
MX		 737 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
738 &pending_snapshot->block_rsv,
739 pending_snapshot->qgroup_reserved);
8257b2dc 740free:
a22285a6 741 kfree(pending_snapshot);
8257b2dc 742out:
9ea24bbe
FM		 743 if (atomic_dec_and_test(&root->will_be_snapshoted))
744 wake_up_atomic_t(&root->will_be_snapshoted);
f46b5a66
CH		 745 return ret;
746}
747
4260f7c7
SW		 748/* copy of may_delete in fs/namei.c()
749 * Check whether we can remove a link victim from directory dir, check
750 * whether the type of victim is right.
751 * 1. We can't do it if dir is read-only (done in permission())
752 * 2. We should have write and exec permissions on dir
753 * 3. We can't remove anything from append-only dir
754 * 4. We can't do anything with immutable dir (done in permission())
755 * 5. If the sticky bit on dir is set we should either
756 * a. be owner of dir, or
757 * b. be owner of victim, or
758 * c. have CAP_FOWNER capability
759 * 6. If the victim is append-only or immutable we can't do antyhing with
760 * links pointing to it.
761 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
762 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
763 * 9. We can't remove a root or mountpoint.
764 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
765 * nfs_async_unlink().
766 */
767
67871254 768static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 769{
770 int error;
771
2b0143b5 772 if (d_really_is_negative(victim))
4260f7c7
SW		 773 return -ENOENT;
774
2b0143b5 775 BUG_ON(d_inode(victim->d_parent) != dir);
4fa6b5ec 776 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 777
778 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
779 if (error)
780 return error;
781 if (IS_APPEND(dir))
782 return -EPERM;
2b0143b5
DH		 783 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
784 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
4260f7c7
SW		 785 return -EPERM;
786 if (isdir) {
e36cb0b8 787 if (!d_is_dir(victim))
4260f7c7
SW		 788 return -ENOTDIR;
789 if (IS_ROOT(victim))
790 return -EBUSY;
e36cb0b8 791 } else if (d_is_dir(victim))
4260f7c7
SW		 792 return -EISDIR;
793 if (IS_DEADDIR(dir))
794 return -ENOENT;
795 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
796 return -EBUSY;
797 return 0;
798}
799
cb8e7090
CH		 800/* copy of may_create in fs/namei.c() */
801static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
802{
2b0143b5 803 if (d_really_is_positive(child))
cb8e7090
CH		 804 return -EEXIST;
805 if (IS_DEADDIR(dir))
806 return -ENOENT;
807 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
808}
809
810/*
811 * Create a new subvolume below @parent. This is largely modeled after
812 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
813 * inside this filesystem so it's quite a bit simpler.
814 */
76dda93c
YZ		 815static noinline int btrfs_mksubvol(struct path *parent,
816 char *name, int namelen,
72fd032e 817 struct btrfs_root *snap_src,
6f72c7e2 818 u64 *async_transid, bool readonly,
8696c533 819 struct btrfs_qgroup_inherit *inherit)
cb8e7090 820{
2b0143b5 821 struct inode *dir = d_inode(parent->dentry);
cb8e7090
CH		 822 struct dentry *dentry;
823 int error;
824
5c50c9b8
DS		 825 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
826 if (error == -EINTR)
827 return error;
cb8e7090
CH		 828
829 dentry = lookup_one_len(name, parent->dentry, namelen);
830 error = PTR_ERR(dentry);
831 if (IS_ERR(dentry))
832 goto out_unlock;
833
834 error = -EEXIST;
2b0143b5 835 if (d_really_is_positive(dentry))
cb8e7090
CH		 836 goto out_dput;
837
76dda93c 838 error = btrfs_may_create(dir, dentry);
cb8e7090 839 if (error)
a874a63e 840 goto out_dput;
cb8e7090 841
9c52057c
CM		 842 /*
843 * even if this name doesn't exist, we may get hash collisions.
844 * check for them now when we can safely fail
845 */
846 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
847 dir->i_ino, name,
848 namelen);
849 if (error)
850 goto out_dput;
851
76dda93c
YZ		 852 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
853
854 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
855 goto out_up_read;
856
3de4586c 857 if (snap_src) {
e9662f70 858 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 859 async_transid, readonly, inherit);
3de4586c 860 } else {
d5c12070
MX		 861 error = create_subvol(dir, dentry, name, namelen,
862 async_transid, inherit);
3de4586c 863 }
76dda93c
YZ		 864 if (!error)
865 fsnotify_mkdir(dir, dentry);
866out_up_read:
867 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 868out_dput:
869 dput(dentry);
870out_unlock:
76dda93c 871 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 872 return error;
873}
874
4cb5300b
CM		 875/*
876 * When we're defragging a range, we don't want to kick it off again
877 * if it is really just waiting for delalloc to send it down.
878 * If we find a nice big extent or delalloc range for the bytes in the
879 * file you want to defrag, we return 0 to let you know to skip this
880 * part of the file
881 */
aab110ab 882static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
4cb5300b
CM		 883{
884 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
885 struct extent_map *em = NULL;
886 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
887 u64 end;
888
889 read_lock(&em_tree->lock);
890 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
891 read_unlock(&em_tree->lock);
892
893 if (em) {
894 end = extent_map_end(em);
895 free_extent_map(em);
896 if (end - offset > thresh)
897 return 0;
898 }
899 /* if we already have a nice delalloc here, just stop */
900 thresh /= 2;
901 end = count_range_bits(io_tree, &offset, offset + thresh,
902 thresh, EXTENT_DELALLOC, 1);
903 if (end >= thresh)
904 return 0;
905 return 1;
906}
907
908/*
909 * helper function to walk through a file and find extents
910 * newer than a specific transid, and smaller than thresh.
911 *
912 * This is used by the defragging code to find new and small
913 * extents
914 */
915static int find_new_extents(struct btrfs_root *root,
916 struct inode *inode, u64 newer_than,
aab110ab 917 u64 *off, u32 thresh)
4cb5300b
CM		 918{
919 struct btrfs_path *path;
920 struct btrfs_key min_key;
4cb5300b
CM		 921 struct extent_buffer *leaf;
922 struct btrfs_file_extent_item *extent;
923 int type;
924 int ret;
a4689d2b 925 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 926
927 path = btrfs_alloc_path();
928 if (!path)
929 return -ENOMEM;
930
a4689d2b 931 min_key.objectid = ino;
4cb5300b
CM		 932 min_key.type = BTRFS_EXTENT_DATA_KEY;
933 min_key.offset = *off;
934
67871254 935 while (1) {
6174d3cb 936 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 937 if (ret != 0)
938 goto none;
f094c9bd 939process_slot:
a4689d2b 940 if (min_key.objectid != ino)
4cb5300b
CM		 941 goto none;
942 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
943 goto none;
944
945 leaf = path->nodes[0];
946 extent = btrfs_item_ptr(leaf, path->slots[0],
947 struct btrfs_file_extent_item);
948
949 type = btrfs_file_extent_type(leaf, extent);
950 if (type == BTRFS_FILE_EXTENT_REG &&
951 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
952 check_defrag_in_cache(inode, min_key.offset, thresh)) {
953 *off = min_key.offset;
954 btrfs_free_path(path);
955 return 0;
956 }
957
f094c9bd
FM		 958 path->slots[0]++;
959 if (path->slots[0] < btrfs_header_nritems(leaf)) {
960 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
961 goto process_slot;
962 }
963
4cb5300b
CM		 964 if (min_key.offset == (u64)-1)
965 goto none;
966
967 min_key.offset++;
968 btrfs_release_path(path);
969 }
970none:
971 btrfs_free_path(path);
972 return -ENOENT;
973}
974
6c282eb4 975static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 976{
977 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 978 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
979 struct extent_map *em;
980 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 981
6c282eb4
LZ		 982 /*
983 * hopefully we have this extent in the tree already, try without
984 * the full extent lock
985 */
17ce6ef8 986 read_lock(&em_tree->lock);
6c282eb4 987 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 988 read_unlock(&em_tree->lock);
989
6c282eb4 990 if (!em) {
308d9800
FM		 991 struct extent_state *cached = NULL;
992 u64 end = start + len - 1;
993
6c282eb4 994 /* get the big lock and read metadata off disk */
308d9800 995 lock_extent_bits(io_tree, start, end, 0, &cached);
6c282eb4 996 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
308d9800 997 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
6c282eb4
LZ		 998
999 if (IS_ERR(em))
1000 return NULL;
1001 }
1002
1003 return em;
1004}
17ce6ef8 1005
6c282eb4
LZ		 1006static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1007{
1008 struct extent_map *next;
1009 bool ret = true;
1010
1011 /* this is the last extent */
1012 if (em->start + em->len >= i_size_read(inode))
1013 return false;
1014
1015 next = defrag_lookup_extent(inode, em->start + em->len);
e9512d72
CM		 1016 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1017 ret = false;
1018 else if ((em->block_start + em->block_len == next->block_start) &&
1019 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
6c282eb4
LZ		 1020 ret = false;
1021
1022 free_extent_map(next);
17ce6ef8
LB		 1023 return ret;
1024}
1025
aab110ab 1026static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
a43a2111
AM		 1027 u64 *last_len, u64 *skip, u64 *defrag_end,
1028 int compress)
940100a4 1029{
6c282eb4 1030 struct extent_map *em;
940100a4 1031 int ret = 1;
6c282eb4 1032 bool next_mergeable = true;
940100a4
CM		 1033
1034 /*
008873ea 1035 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1036 * defragging it
1037 */
1038 if (start < *defrag_end)
1039 return 1;
1040
1041 *skip = 0;
1042
6c282eb4
LZ		 1043 em = defrag_lookup_extent(inode, start);
1044 if (!em)
1045 return 0;
940100a4
CM		 1046
1047 /* this will cover holes, and inline extents */
17ce6ef8 1048 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1049 ret = 0;
17ce6ef8
LB		 1050 goto out;
1051 }
1052
6c282eb4 1053 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1054 /*
6c282eb4
LZ		 1055 * we hit a real extent, if it is big or the next extent is not a
1056 * real extent, don't bother defragging it
940100a4 1057 */
a43a2111 1058 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 1059 (em->len >= thresh || !next_mergeable))
940100a4 1060 ret = 0;
17ce6ef8 1061out:
940100a4
CM		 1062 /*
1063 * last_len ends up being a counter of how many bytes we've defragged.
1064 * every time we choose not to defrag an extent, we reset *last_len
1065 * so that the next tiny extent will force a defrag.
1066 *
1067 * The end result of this is that tiny extents before a single big
1068 * extent will force at least part of that big extent to be defragged.
1069 */
1070 if (ret) {
940100a4
CM		 1071 *defrag_end = extent_map_end(em);
1072 } else {
1073 *last_len = 0;
1074 *skip = extent_map_end(em);
1075 *defrag_end = 0;
1076 }
1077
1078 free_extent_map(em);
1079 return ret;
1080}
1081
4cb5300b
CM		 1082/*
1083 * it doesn't do much good to defrag one or two pages
1084 * at a time. This pulls in a nice chunk of pages
1085 * to COW and defrag.
1086 *
1087 * It also makes sure the delalloc code has enough
1088 * dirty data to avoid making new small extents as part
1089 * of the defrag
1090 *
1091 * It's a good idea to start RA on this range
1092 * before calling this.
1093 */
1094static int cluster_pages_for_defrag(struct inode *inode,
1095 struct page **pages,
1096 unsigned long start_index,
c41570c9 1097 unsigned long num_pages)
f46b5a66 1098{
4cb5300b
CM		 1099 unsigned long file_end;
1100 u64 isize = i_size_read(inode);
1101 u64 page_start;
1102 u64 page_end;
1f12bd06 1103 u64 page_cnt;
4cb5300b
CM		 1104 int ret;
1105 int i;
1106 int i_done;
3eaa2885 1107 struct btrfs_ordered_extent *ordered;
4cb5300b 1108 struct extent_state *cached_state = NULL;
600a45e1 1109 struct extent_io_tree *tree;
3b16a4e3 1110 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1111
4cb5300b 1112 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1113 if (!isize || start_index > file_end)
1114 return 0;
1115
1116 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1117
1118 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1119 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1120 if (ret)
1121 return ret;
4cb5300b 1122 i_done = 0;
600a45e1 1123 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1124
1125 /* step one, lock all the pages */
1f12bd06 1126 for (i = 0; i < page_cnt; i++) {
4cb5300b 1127 struct page *page;
600a45e1 1128again:
a94733d0 1129 page = find_or_create_page(inode->i_mapping,
600a45e1 1130 start_index + i, mask);
4cb5300b
CM		 1131 if (!page)
1132 break;
1133
600a45e1
MX		 1134 page_start = page_offset(page);
1135 page_end = page_start + PAGE_CACHE_SIZE - 1;
1136 while (1) {
308d9800
FM		 1137 lock_extent_bits(tree, page_start, page_end,
1138 0, &cached_state);
600a45e1
MX		 1139 ordered = btrfs_lookup_ordered_extent(inode,
1140 page_start);
308d9800
FM		 1141 unlock_extent_cached(tree, page_start, page_end,
1142 &cached_state, GFP_NOFS);
600a45e1
MX		 1143 if (!ordered)
1144 break;
1145
1146 unlock_page(page);
1147 btrfs_start_ordered_extent(inode, ordered, 1);
1148 btrfs_put_ordered_extent(ordered);
1149 lock_page(page);
1f12bd06
LB		 1150 /*
1151 * we unlocked the page above, so we need check if
1152 * it was released or not.
1153 */
1154 if (page->mapping != inode->i_mapping) {
1155 unlock_page(page);
1156 page_cache_release(page);
1157 goto again;
1158 }
600a45e1
MX		 1159 }
1160
4cb5300b
CM		 1161 if (!PageUptodate(page)) {
1162 btrfs_readpage(NULL, page);
1163 lock_page(page);
1164 if (!PageUptodate(page)) {
1165 unlock_page(page);
1166 page_cache_release(page);
1167 ret = -EIO;
1168 break;
1169 }
1170 }
600a45e1 1171
600a45e1
MX		 1172 if (page->mapping != inode->i_mapping) {
1173 unlock_page(page);
1174 page_cache_release(page);
1175 goto again;
1176 }
1177
4cb5300b
CM		 1178 pages[i] = page;
1179 i_done++;
1180 }
1181 if (!i_done || ret)
1182 goto out;
1183
1184 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1185 goto out;
1186
1187 /*
1188 * so now we have a nice long stream of locked
1189 * and up to date pages, lets wait on them
1190 */
1191 for (i = 0; i < i_done; i++)
1192 wait_on_page_writeback(pages[i]);
1193
1194 page_start = page_offset(pages[0]);
1195 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1196
1197 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1198 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1199 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1200 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1201 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1202 &cached_state, GFP_NOFS);
4cb5300b 1203
1f12bd06 1204 if (i_done != page_cnt) {
9e0baf60
JB		 1205 spin_lock(&BTRFS_I(inode)->lock);
1206 BTRFS_I(inode)->outstanding_extents++;
1207 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1208 btrfs_delalloc_release_space(inode,
1f12bd06 1209 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1210 }
1211
1212
9e8a4a8b
LB		 1213 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1214 &cached_state, GFP_NOFS);
4cb5300b
CM		 1215
1216 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1217 page_start, page_end - 1, &cached_state,
1218 GFP_NOFS);
1219
1220 for (i = 0; i < i_done; i++) {
1221 clear_page_dirty_for_io(pages[i]);
1222 ClearPageChecked(pages[i]);
1223 set_page_extent_mapped(pages[i]);
1224 set_page_dirty(pages[i]);
1225 unlock_page(pages[i]);
1226 page_cache_release(pages[i]);
1227 }
1228 return i_done;
1229out:
1230 for (i = 0; i < i_done; i++) {
1231 unlock_page(pages[i]);
1232 page_cache_release(pages[i]);
1233 }
1f12bd06 1234 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1235 return ret;
1236
1237}
1238
1239int btrfs_defrag_file(struct inode *inode, struct file *file,
1240 struct btrfs_ioctl_defrag_range_args *range,
1241 u64 newer_than, unsigned long max_to_defrag)
1242{
1243 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1244 struct file_ra_state *ra = NULL;
f46b5a66 1245 unsigned long last_index;
151a31b2 1246 u64 isize = i_size_read(inode);
940100a4
CM		 1247 u64 last_len = 0;
1248 u64 skip = 0;
1249 u64 defrag_end = 0;
4cb5300b 1250 u64 newer_off = range->start;
f46b5a66 1251 unsigned long i;
008873ea 1252 unsigned long ra_index = 0;
f46b5a66 1253 int ret;
4cb5300b 1254 int defrag_count = 0;
1a419d85 1255 int compress_type = BTRFS_COMPRESS_ZLIB;
aab110ab 1256 u32 extent_thresh = range->extent_thresh;
c41570c9
JM		 1257 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1258 unsigned long cluster = max_cluster;
4cb5300b
CM		 1259 u64 new_align = ~((u64)128 * 1024 - 1);
1260 struct page **pages = NULL;
1261
0abd5b17
LB		 1262 if (isize == 0)
1263 return 0;
1264
1265 if (range->start >= isize)
1266 return -EINVAL;
1a419d85
LZ		 1267
1268 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1269 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1270 return -EINVAL;
1271 if (range->compress_type)
1272 compress_type = range->compress_type;
1273 }
f46b5a66 1274
0abd5b17
LB		 1275 if (extent_thresh == 0)
1276 extent_thresh = 256 * 1024;
940100a4 1277
4cb5300b
CM		 1278 /*
1279 * if we were not given a file, allocate a readahead
1280 * context
1281 */
1282 if (!file) {
1283 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1284 if (!ra)
1285 return -ENOMEM;
1286 file_ra_state_init(ra, inode->i_mapping);
1287 } else {
1288 ra = &file->f_ra;
1289 }
1290
d9b0d9ba 1291 pages = kmalloc_array(max_cluster, sizeof(struct page *),
4cb5300b
CM		 1292 GFP_NOFS);
1293 if (!pages) {
1294 ret = -ENOMEM;
1295 goto out_ra;
1296 }
1297
1298 /* find the last page to defrag */
1e701a32 1299 if (range->start + range->len > range->start) {
151a31b2 1300 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1301 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1302 } else {
151a31b2 1303 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1304 }
1305
4cb5300b
CM		 1306 if (newer_than) {
1307 ret = find_new_extents(root, inode, newer_than,
1308 &newer_off, 64 * 1024);
1309 if (!ret) {
1310 range->start = newer_off;
1311 /*
1312 * we always align our defrag to help keep
1313 * the extents in the file evenly spaced
1314 */
1315 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1316 } else
1317 goto out_ra;
1318 } else {
1319 i = range->start >> PAGE_CACHE_SHIFT;
1320 }
1321 if (!max_to_defrag)
070034bd 1322 max_to_defrag = last_index - i + 1;
4cb5300b 1323
2a0f7f57
LZ		 1324 /*
1325 * make writeback starts from i, so the defrag range can be
1326 * written sequentially.
1327 */
1328 if (i < inode->i_mapping->writeback_index)
1329 inode->i_mapping->writeback_index = i;
1330
f7f43cc8 1331 while (i <= last_index && defrag_count < max_to_defrag &&
ed6078f7 1332 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
4cb5300b
CM		 1333 /*
1334 * make sure we stop running if someone unmounts
1335 * the FS
1336 */
1337 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1338 break;
1339
210549eb 1340 if (btrfs_defrag_cancelled(root->fs_info)) {
efe120a0 1341 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
210549eb
DS		 1342 ret = -EAGAIN;
1343 break;
1344 }
1345
66c26892 1346 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1347 extent_thresh, &last_len, &skip,
a43a2111
AM		 1348 &defrag_end, range->flags &
1349 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1350 unsigned long next;
1351 /*
1352 * the should_defrag function tells us how much to skip
1353 * bump our counter by the suggested amount
1354 */
ed6078f7 1355 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
940100a4
CM		 1356 i = max(i + 1, next);
1357 continue;
1358 }
008873ea
LZ		 1359
1360 if (!newer_than) {
1361 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1362 PAGE_CACHE_SHIFT) - i;
1363 cluster = min(cluster, max_cluster);
1364 } else {
1365 cluster = max_cluster;
1366 }
1367
008873ea
LZ		 1368 if (i + cluster > ra_index) {
1369 ra_index = max(i, ra_index);
1370 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1371 cluster);
e4826a5b 1372 ra_index += cluster;
008873ea 1373 }
940100a4 1374
ecb8bea8 1375 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1376 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1377 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1378 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1379 if (ret < 0) {
1380 mutex_unlock(&inode->i_mutex);
4cb5300b 1381 goto out_ra;
ecb8bea8 1382 }
4cb5300b
CM		 1383
1384 defrag_count += ret;
d0e1d66b 1385 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1386 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1387
1388 if (newer_than) {
1389 if (newer_off == (u64)-1)
1390 break;
1391
e1f041e1
LB		 1392 if (ret > 0)
1393 i += ret;
1394
4cb5300b
CM		 1395 newer_off = max(newer_off + 1,
1396 (u64)i << PAGE_CACHE_SHIFT);
1397
1398 ret = find_new_extents(root, inode,
1399 newer_than, &newer_off,
1400 64 * 1024);
1401 if (!ret) {
1402 range->start = newer_off;
1403 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1404 } else {
1405 break;
f46b5a66 1406 }
4cb5300b 1407 } else {
008873ea 1408 if (ret > 0) {
cbcc8326 1409 i += ret;
008873ea
LZ		 1410 last_len += ret << PAGE_CACHE_SHIFT;
1411 } else {
cbcc8326 1412 i++;
008873ea
LZ		 1413 last_len = 0;
1414 }
f46b5a66 1415 }
f46b5a66
CH		 1416 }
1417
dec8ef90 1418 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1419 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1420 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1421 &BTRFS_I(inode)->runtime_flags))
1422 filemap_flush(inode->i_mapping);
1423 }
1e701a32
CM		 1424
1425 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1426 /* the filemap_flush will queue IO into the worker threads, but
1427 * we have to make sure the IO is actually started and that
1428 * ordered extents get created before we return
1429 */
1430 atomic_inc(&root->fs_info->async_submit_draining);
1431 while (atomic_read(&root->fs_info->nr_async_submits) ||
1432 atomic_read(&root->fs_info->async_delalloc_pages)) {
1433 wait_event(root->fs_info->async_submit_wait,
1434 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1435 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1436 }
1437 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1438 }
1439
1a419d85 1440 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1441 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1442 }
1443
60ccf82f 1444 ret = defrag_count;
940100a4 1445
4cb5300b 1446out_ra:
633085c7
FDBM		 1447 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1448 mutex_lock(&inode->i_mutex);
1449 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1450 mutex_unlock(&inode->i_mutex);
1451 }
4cb5300b
CM		 1452 if (!file)
1453 kfree(ra);
1454 kfree(pages);
940100a4 1455 return ret;
f46b5a66
CH		 1456}
1457
198605a8 1458static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1459 void __user *arg)
f46b5a66
CH		 1460{
1461 u64 new_size;
1462 u64 old_size;
1463 u64 devid = 1;
496ad9aa 1464 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1465 struct btrfs_ioctl_vol_args *vol_args;
1466 struct btrfs_trans_handle *trans;
1467 struct btrfs_device *device = NULL;
1468 char *sizestr;
9a40f122 1469 char *retptr;
f46b5a66
CH		 1470 char *devstr = NULL;
1471 int ret = 0;
f46b5a66
CH		 1472 int mod = 0;
1473
e441d54d
CM		 1474 if (!capable(CAP_SYS_ADMIN))
1475 return -EPERM;
1476
198605a8
MX		 1477 ret = mnt_want_write_file(file);
1478 if (ret)
1479 return ret;
1480
5ac00add
SB		 1481 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1482 1)) {
97547676 1483 mnt_drop_write_file(file);
e57138b3 1484 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1485 }
1486
5ac00add 1487 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1488 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1489 if (IS_ERR(vol_args)) {
1490 ret = PTR_ERR(vol_args);
1491 goto out;
1492 }
5516e595
MF		 1493
1494 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1495
f46b5a66
CH		 1496 sizestr = vol_args->name;
1497 devstr = strchr(sizestr, ':');
1498 if (devstr) {
f46b5a66
CH		 1499 sizestr = devstr + 1;
1500 *devstr = '\0';
1501 devstr = vol_args->name;
58dfae63
Z		 1502 ret = kstrtoull(devstr, 10, &devid);
1503 if (ret)
1504 goto out_free;
dfd79829
MX		 1505 if (!devid) {
1506 ret = -EINVAL;
1507 goto out_free;
1508 }
efe120a0 1509 btrfs_info(root->fs_info, "resizing devid %llu", devid);
f46b5a66 1510 }
dba60f3f 1511
aa1b8cd4 1512 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1513 if (!device) {
efe120a0 1514 btrfs_info(root->fs_info, "resizer unable to find device %llu",
c1c9ff7c 1515 devid);
dfd79829 1516 ret = -ENODEV;
c9e9f97b 1517 goto out_free;
f46b5a66 1518 }
dba60f3f
MX		 1519
1520 if (!device->writeable) {
efe120a0
FH		 1521 btrfs_info(root->fs_info,
1522 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1523 devid);
dfd79829 1524 ret = -EPERM;
4e42ae1b
LB		 1525 goto out_free;
1526 }
1527
f46b5a66
CH		 1528 if (!strcmp(sizestr, "max"))
1529 new_size = device->bdev->bd_inode->i_size;
1530 else {
1531 if (sizestr[0] == '-') {
1532 mod = -1;
1533 sizestr++;
1534 } else if (sizestr[0] == '+') {
1535 mod = 1;
1536 sizestr++;
1537 }
9a40f122
GH		 1538 new_size = memparse(sizestr, &retptr);
1539 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1540 ret = -EINVAL;
c9e9f97b 1541 goto out_free;
f46b5a66
CH		 1542 }
1543 }
1544
63a212ab 1545 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1546 ret = -EPERM;
63a212ab
SB		 1547 goto out_free;
1548 }
1549
7cc8e58d 1550 old_size = btrfs_device_get_total_bytes(device);
f46b5a66
CH		 1551
1552 if (mod < 0) {
1553 if (new_size > old_size) {
1554 ret = -EINVAL;
c9e9f97b 1555 goto out_free;
f46b5a66
CH		 1556 }
1557 new_size = old_size - new_size;
1558 } else if (mod > 0) {
eb8052e0 1559 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1560 ret = -ERANGE;
eb8052e0
WF		 1561 goto out_free;
1562 }
f46b5a66
CH		 1563 new_size = old_size + new_size;
1564 }
1565
1566 if (new_size < 256 * 1024 * 1024) {
1567 ret = -EINVAL;
c9e9f97b 1568 goto out_free;
f46b5a66
CH		 1569 }
1570 if (new_size > device->bdev->bd_inode->i_size) {
1571 ret = -EFBIG;
c9e9f97b 1572 goto out_free;
f46b5a66
CH		 1573 }
1574
b8b93add 1575 new_size = div_u64(new_size, root->sectorsize);
f46b5a66
CH		 1576 new_size *= root->sectorsize;
1577
efe120a0 1578 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
c1c9ff7c 1579 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1580
1581 if (new_size > old_size) {
a22285a6 1582 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1583 if (IS_ERR(trans)) {
1584 ret = PTR_ERR(trans);
c9e9f97b 1585 goto out_free;
98d5dc13 1586 }
f46b5a66
CH		 1587 ret = btrfs_grow_device(trans, device, new_size);
1588 btrfs_commit_transaction(trans, root);
ece7d20e 1589 } else if (new_size < old_size) {
f46b5a66 1590 ret = btrfs_shrink_device(device, new_size);
0253f40e 1591 } /* equal, nothing need to do */
f46b5a66 1592
c9e9f97b 1593out_free:
f46b5a66 1594 kfree(vol_args);
c9e9f97b
ID		 1595out:
1596 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1597 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1598 mnt_drop_write_file(file);
f46b5a66
CH		 1599 return ret;
1600}
1601
72fd032e 1602static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1603 char *name, unsigned long fd, int subvol,
1604 u64 *transid, bool readonly,
8696c533 1605 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1606{
f46b5a66 1607 int namelen;
3de4586c 1608 int ret = 0;
f46b5a66 1609
a874a63e
LB		 1610 ret = mnt_want_write_file(file);
1611 if (ret)
1612 goto out;
1613
72fd032e
SW		 1614 namelen = strlen(name);
1615 if (strchr(name, '/')) {
f46b5a66 1616 ret = -EINVAL;
a874a63e 1617 goto out_drop_write;
f46b5a66
CH		 1618 }
1619
16780cab
CM		 1620 if (name[0] == '.' &&
1621 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1622 ret = -EEXIST;
a874a63e 1623 goto out_drop_write;
16780cab
CM		 1624 }
1625
3de4586c 1626 if (subvol) {
72fd032e 1627 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1628 NULL, transid, readonly, inherit);
cb8e7090 1629 } else {
2903ff01 1630 struct fd src = fdget(fd);
3de4586c 1631 struct inode *src_inode;
2903ff01 1632 if (!src.file) {
3de4586c 1633 ret = -EINVAL;
a874a63e 1634 goto out_drop_write;
3de4586c
CM		 1635 }
1636
496ad9aa
AV		 1637 src_inode = file_inode(src.file);
1638 if (src_inode->i_sb != file_inode(file)->i_sb) {
efe120a0
FH		 1639 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1640 "Snapshot src from another FS");
23ad5b17 1641 ret = -EXDEV;
d0242061
DS		 1642 } else if (!inode_owner_or_capable(src_inode)) {
1643 /*
1644 * Subvolume creation is not restricted, but snapshots
1645 * are limited to own subvolumes only
1646 */
1647 ret = -EPERM;
ecd18815
AV		 1648 } else {
1649 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1650 BTRFS_I(src_inode)->root,
1651 transid, readonly, inherit);
3de4586c 1652 }
2903ff01 1653 fdput(src);
cb8e7090 1654 }
a874a63e
LB		 1655out_drop_write:
1656 mnt_drop_write_file(file);
f46b5a66 1657out:
72fd032e
SW		 1658 return ret;
1659}
1660
1661static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1662 void __user *arg, int subvol)
72fd032e 1663{
fa0d2b9b 1664 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1665 int ret;
1666
fa0d2b9b
LZ		 1667 vol_args = memdup_user(arg, sizeof(*vol_args));
1668 if (IS_ERR(vol_args))
1669 return PTR_ERR(vol_args);
1670 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1671
fa0d2b9b 1672 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1673 vol_args->fd, subvol,
6f72c7e2 1674 NULL, false, NULL);
fdfb1e4f 1675
fa0d2b9b
LZ		 1676 kfree(vol_args);
1677 return ret;
1678}
fdfb1e4f 1679
fa0d2b9b
LZ		 1680static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1681 void __user *arg, int subvol)
1682{
1683 struct btrfs_ioctl_vol_args_v2 *vol_args;
1684 int ret;
1685 u64 transid = 0;
1686 u64 *ptr = NULL;
b83cc969 1687 bool readonly = false;
6f72c7e2 1688 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1689
fa0d2b9b
LZ		 1690 vol_args = memdup_user(arg, sizeof(*vol_args));
1691 if (IS_ERR(vol_args))
1692 return PTR_ERR(vol_args);
1693 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1694
b83cc969 1695 if (vol_args->flags &
6f72c7e2
AJ		 1696 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1697 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1698 ret = -EOPNOTSUPP;
c47ca32d 1699 goto free_args;
72fd032e 1700 }
fa0d2b9b
LZ		 1701
1702 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1703 ptr = &transid;
b83cc969
LZ		 1704 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1705 readonly = true;
6f72c7e2
AJ		 1706 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1707 if (vol_args->size > PAGE_CACHE_SIZE) {
1708 ret = -EINVAL;
c47ca32d 1709 goto free_args;
6f72c7e2
AJ		 1710 }
1711 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1712 if (IS_ERR(inherit)) {
1713 ret = PTR_ERR(inherit);
c47ca32d 1714 goto free_args;
6f72c7e2
AJ		 1715 }
1716 }
fa0d2b9b
LZ		 1717
1718 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1719 vol_args->fd, subvol, ptr,
8696c533 1720 readonly, inherit);
c47ca32d
DC		 1721 if (ret)
1722 goto free_inherit;
fa0d2b9b 1723
c47ca32d
DC		 1724 if (ptr && copy_to_user(arg +
1725 offsetof(struct btrfs_ioctl_vol_args_v2,
1726 transid),
1727 ptr, sizeof(*ptr)))
fa0d2b9b 1728 ret = -EFAULT;
c47ca32d
DC		 1729
1730free_inherit:
6f72c7e2 1731 kfree(inherit);
c47ca32d
DC		 1732free_args:
1733 kfree(vol_args);
f46b5a66
CH		 1734 return ret;
1735}
1736
0caa102d
LZ		 1737static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1738 void __user *arg)
1739{
496ad9aa 1740 struct inode *inode = file_inode(file);
0caa102d
LZ		 1741 struct btrfs_root *root = BTRFS_I(inode)->root;
1742 int ret = 0;
1743 u64 flags = 0;
1744
33345d01 1745 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1746 return -EINVAL;
1747
1748 down_read(&root->fs_info->subvol_sem);
1749 if (btrfs_root_readonly(root))
1750 flags |= BTRFS_SUBVOL_RDONLY;
1751 up_read(&root->fs_info->subvol_sem);
1752
1753 if (copy_to_user(arg, &flags, sizeof(flags)))
1754 ret = -EFAULT;
1755
1756 return ret;
1757}
1758
1759static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1760 void __user *arg)
1761{
496ad9aa 1762 struct inode *inode = file_inode(file);
0caa102d
LZ		 1763 struct btrfs_root *root = BTRFS_I(inode)->root;
1764 struct btrfs_trans_handle *trans;
1765 u64 root_flags;
1766 u64 flags;
1767 int ret = 0;
1768
bd60ea0f
DS		 1769 if (!inode_owner_or_capable(inode))
1770 return -EPERM;
1771
b9ca0664
LB		 1772 ret = mnt_want_write_file(file);
1773 if (ret)
1774 goto out;
0caa102d 1775
b9ca0664
LB		 1776 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1777 ret = -EINVAL;
1778 goto out_drop_write;
1779 }
0caa102d 1780
b9ca0664
LB		 1781 if (copy_from_user(&flags, arg, sizeof(flags))) {
1782 ret = -EFAULT;
1783 goto out_drop_write;
1784 }
0caa102d 1785
b9ca0664
LB		 1786 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1787 ret = -EINVAL;
1788 goto out_drop_write;
1789 }
0caa102d 1790
b9ca0664
LB		 1791 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1792 ret = -EOPNOTSUPP;
1793 goto out_drop_write;
1794 }
0caa102d
LZ		 1795
1796 down_write(&root->fs_info->subvol_sem);
1797
1798 /* nothing to do */
1799 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1800 goto out_drop_sem;
0caa102d
LZ		 1801
1802 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1803 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1804 btrfs_set_root_flags(&root->root_item,
1805 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1806 } else {
1807 /*
1808 * Block RO -> RW transition if this subvolume is involved in
1809 * send
1810 */
1811 spin_lock(&root->root_item_lock);
1812 if (root->send_in_progress == 0) {
1813 btrfs_set_root_flags(&root->root_item,
0caa102d 1814 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1815 spin_unlock(&root->root_item_lock);
1816 } else {
1817 spin_unlock(&root->root_item_lock);
1818 btrfs_warn(root->fs_info,
1819 "Attempt to set subvolume %llu read-write during send",
1820 root->root_key.objectid);
1821 ret = -EPERM;
1822 goto out_drop_sem;
1823 }
1824 }
0caa102d
LZ		 1825
1826 trans = btrfs_start_transaction(root, 1);
1827 if (IS_ERR(trans)) {
1828 ret = PTR_ERR(trans);
1829 goto out_reset;
1830 }
1831
b4dc2b8c 1832 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1833 &root->root_key, &root->root_item);
1834
1835 btrfs_commit_transaction(trans, root);
1836out_reset:
1837 if (ret)
1838 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1839out_drop_sem:
0caa102d 1840 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1841out_drop_write:
1842 mnt_drop_write_file(file);
1843out:
0caa102d
LZ		 1844 return ret;
1845}
1846
76dda93c
YZ		 1847/*
1848 * helper to check if the subvolume references other subvolumes
1849 */
1850static noinline int may_destroy_subvol(struct btrfs_root *root)
1851{
1852 struct btrfs_path *path;
175a2b87 1853 struct btrfs_dir_item *di;
76dda93c 1854 struct btrfs_key key;
175a2b87 1855 u64 dir_id;
76dda93c
YZ		 1856 int ret;
1857
1858 path = btrfs_alloc_path();
1859 if (!path)
1860 return -ENOMEM;
1861
175a2b87
JB		 1862 /* Make sure this root isn't set as the default subvol */
1863 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1864 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1865 dir_id, "default", 7, 0);
1866 if (di && !IS_ERR(di)) {
1867 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1868 if (key.objectid == root->root_key.objectid) {
72de6b53
GS		 1869 ret = -EPERM;
1870 btrfs_err(root->fs_info, "deleting default subvolume "
1871 "%llu is not allowed", key.objectid);
175a2b87
JB		 1872 goto out;
1873 }
1874 btrfs_release_path(path);
1875 }
1876
76dda93c
YZ		 1877 key.objectid = root->root_key.objectid;
1878 key.type = BTRFS_ROOT_REF_KEY;
1879 key.offset = (u64)-1;
1880
1881 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1882 &key, path, 0, 0);
1883 if (ret < 0)
1884 goto out;
1885 BUG_ON(ret == 0);
1886
1887 ret = 0;
1888 if (path->slots[0] > 0) {
1889 path->slots[0]--;
1890 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1891 if (key.objectid == root->root_key.objectid &&
1892 key.type == BTRFS_ROOT_REF_KEY)
1893 ret = -ENOTEMPTY;
1894 }
1895out:
1896 btrfs_free_path(path);
1897 return ret;
1898}
1899
ac8e9819
CM		 1900static noinline int key_in_sk(struct btrfs_key *key,
1901 struct btrfs_ioctl_search_key *sk)
1902{
abc6e134
CM		 1903 struct btrfs_key test;
1904 int ret;
1905
1906 test.objectid = sk->min_objectid;
1907 test.type = sk->min_type;
1908 test.offset = sk->min_offset;
1909
1910 ret = btrfs_comp_cpu_keys(key, &test);
1911 if (ret < 0)
ac8e9819 1912 return 0;
abc6e134
CM		 1913
1914 test.objectid = sk->max_objectid;
1915 test.type = sk->max_type;
1916 test.offset = sk->max_offset;
1917
1918 ret = btrfs_comp_cpu_keys(key, &test);
1919 if (ret > 0)
ac8e9819
CM		 1920 return 0;
1921 return 1;
1922}
1923
1924static noinline int copy_to_sk(struct btrfs_root *root,
1925 struct btrfs_path *path,
1926 struct btrfs_key *key,
1927 struct btrfs_ioctl_search_key *sk,
9b6e817d 1928 size_t *buf_size,
ba346b35 1929 char __user *ubuf,
ac8e9819
CM		 1930 unsigned long *sk_offset,
1931 int *num_found)
1932{
1933 u64 found_transid;
1934 struct extent_buffer *leaf;
1935 struct btrfs_ioctl_search_header sh;
1936 unsigned long item_off;
1937 unsigned long item_len;
1938 int nritems;
1939 int i;
1940 int slot;
ac8e9819
CM		 1941 int ret = 0;
1942
1943 leaf = path->nodes[0];
1944 slot = path->slots[0];
1945 nritems = btrfs_header_nritems(leaf);
1946
1947 if (btrfs_header_generation(leaf) > sk->max_transid) {
1948 i = nritems;
1949 goto advance_key;
1950 }
1951 found_transid = btrfs_header_generation(leaf);
1952
1953 for (i = slot; i < nritems; i++) {
1954 item_off = btrfs_item_ptr_offset(leaf, i);
1955 item_len = btrfs_item_size_nr(leaf, i);
1956
03b71c6c
GP		 1957 btrfs_item_key_to_cpu(leaf, key, i);
1958 if (!key_in_sk(key, sk))
1959 continue;
1960
9b6e817d 1961 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 1962 if (*num_found) {
1963 ret = 1;
1964 goto out;
1965 }
1966
1967 /*
1968 * return one empty item back for v1, which does not
1969 * handle -EOVERFLOW
1970 */
1971
9b6e817d 1972 *buf_size = sizeof(sh) + item_len;
ac8e9819 1973 item_len = 0;
8f5f6178
GH		 1974 ret = -EOVERFLOW;
1975 }
ac8e9819 1976
9b6e817d 1977 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 1978 ret = 1;
25c9bc2e 1979 goto out;
ac8e9819
CM		 1980 }
1981
ac8e9819
CM		 1982 sh.objectid = key->objectid;
1983 sh.offset = key->offset;
1984 sh.type = key->type;
1985 sh.len = item_len;
1986 sh.transid = found_transid;
1987
1988 /* copy search result header */
ba346b35
GH		 1989 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1990 ret = -EFAULT;
1991 goto out;
1992 }
1993
ac8e9819
CM		 1994 *sk_offset += sizeof(sh);
1995
1996 if (item_len) {
ba346b35 1997 char __user *up = ubuf + *sk_offset;
ac8e9819 1998 /* copy the item */
ba346b35
GH		 1999 if (read_extent_buffer_to_user(leaf, up,
2000 item_off, item_len)) {
2001 ret = -EFAULT;
2002 goto out;
2003 }
2004
ac8e9819 2005 *sk_offset += item_len;
ac8e9819 2006 }
e2156867 2007 (*num_found)++;
ac8e9819 2008
8f5f6178
GH		 2009 if (ret) /* -EOVERFLOW from above */
2010 goto out;
2011
25c9bc2e
GH		 2012 if (*num_found >= sk->nr_items) {
2013 ret = 1;
2014 goto out;
2015 }
ac8e9819
CM		 2016 }
2017advance_key:
abc6e134
CM		 2018 ret = 0;
2019 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 2020 key->offset++;
abc6e134
CM		 2021 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2022 key->offset = 0;
ac8e9819 2023 key->type++;
abc6e134
CM		 2024 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2025 key->offset = 0;
2026 key->type = 0;
ac8e9819 2027 key->objectid++;
abc6e134
CM		 2028 } else
2029 ret = 1;
25c9bc2e 2030out:
ba346b35
GH		 2031 /*
2032 * 0: all items from this leaf copied, continue with next
2033 * 1: * more items can be copied, but unused buffer is too small
2034 * * all items were found
2035 * Either way, it will stops the loop which iterates to the next
2036 * leaf
2037 * -EOVERFLOW: item was to large for buffer
2038 * -EFAULT: could not copy extent buffer back to userspace
2039 */
ac8e9819
CM		 2040 return ret;
2041}
2042
2043static noinline int search_ioctl(struct inode *inode,
12544442 2044 struct btrfs_ioctl_search_key *sk,
9b6e817d 2045 size_t *buf_size,
ba346b35 2046 char __user *ubuf)
ac8e9819
CM		 2047{
2048 struct btrfs_root *root;
2049 struct btrfs_key key;
ac8e9819 2050 struct btrfs_path *path;
ac8e9819
CM		 2051 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2052 int ret;
2053 int num_found = 0;
2054 unsigned long sk_offset = 0;
2055
9b6e817d
GH		 2056 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2057 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2058 return -EOVERFLOW;
9b6e817d 2059 }
12544442 2060
ac8e9819
CM		 2061 path = btrfs_alloc_path();
2062 if (!path)
2063 return -ENOMEM;
2064
2065 if (sk->tree_id == 0) {
2066 /* search the root of the inode that was passed */
2067 root = BTRFS_I(inode)->root;
2068 } else {
2069 key.objectid = sk->tree_id;
2070 key.type = BTRFS_ROOT_ITEM_KEY;
2071 key.offset = (u64)-1;
2072 root = btrfs_read_fs_root_no_name(info, &key);
2073 if (IS_ERR(root)) {
efe120a0 2074 printk(KERN_ERR "BTRFS: could not find root %llu\n",
ac8e9819
CM		 2075 sk->tree_id);
2076 btrfs_free_path(path);
2077 return -ENOENT;
2078 }
2079 }
2080
2081 key.objectid = sk->min_objectid;
2082 key.type = sk->min_type;
2083 key.offset = sk->min_offset;
2084
67871254 2085 while (1) {
6174d3cb 2086 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2087 if (ret != 0) {
2088 if (ret > 0)
2089 ret = 0;
2090 goto err;
2091 }
ba346b35 2092 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
ac8e9819 2093 &sk_offset, &num_found);
b3b4aa74 2094 btrfs_release_path(path);
25c9bc2e 2095 if (ret)
ac8e9819
CM		 2096 break;
2097
2098 }
8f5f6178
GH		 2099 if (ret > 0)
2100 ret = 0;
ac8e9819
CM		 2101err:
2102 sk->nr_items = num_found;
2103 btrfs_free_path(path);
2104 return ret;
2105}
2106
2107static noinline int btrfs_ioctl_tree_search(struct file *file,
2108 void __user *argp)
2109{
ba346b35
GH		 2110 struct btrfs_ioctl_search_args __user *uargs;
2111 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2112 struct inode *inode;
2113 int ret;
2114 size_t buf_size;
ac8e9819
CM		 2115
2116 if (!capable(CAP_SYS_ADMIN))
2117 return -EPERM;
2118
ba346b35
GH		 2119 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2120
2121 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2122 return -EFAULT;
ac8e9819 2123
ba346b35 2124 buf_size = sizeof(uargs->buf);
ac8e9819 2125
496ad9aa 2126 inode = file_inode(file);
ba346b35 2127 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2128
2129 /*
2130 * In the origin implementation an overflow is handled by returning a
2131 * search header with a len of zero, so reset ret.
2132 */
2133 if (ret == -EOVERFLOW)
2134 ret = 0;
2135
ba346b35 2136 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2137 ret = -EFAULT;
ac8e9819
CM		 2138 return ret;
2139}
2140
cc68a8a5
GH		 2141static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2142 void __user *argp)
2143{
2144 struct btrfs_ioctl_search_args_v2 __user *uarg;
2145 struct btrfs_ioctl_search_args_v2 args;
2146 struct inode *inode;
2147 int ret;
2148 size_t buf_size;
2149 const size_t buf_limit = 16 * 1024 * 1024;
2150
2151 if (!capable(CAP_SYS_ADMIN))
2152 return -EPERM;
2153
2154 /* copy search header and buffer size */
2155 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2156 if (copy_from_user(&args, uarg, sizeof(args)))
2157 return -EFAULT;
2158
2159 buf_size = args.buf_size;
2160
2161 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2162 return -EOVERFLOW;
2163
2164 /* limit result size to 16MB */
2165 if (buf_size > buf_limit)
2166 buf_size = buf_limit;
2167
2168 inode = file_inode(file);
2169 ret = search_ioctl(inode, &args.key, &buf_size,
2170 (char *)(&uarg->buf[0]));
2171 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2172 ret = -EFAULT;
2173 else if (ret == -EOVERFLOW &&
2174 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2175 ret = -EFAULT;
2176
ac8e9819
CM		 2177 return ret;
2178}
2179
98d377a0 2180/*
ac8e9819
CM		 2181 * Search INODE_REFs to identify path name of 'dirid' directory
2182 * in a 'tree_id' tree. and sets path name to 'name'.
2183 */
98d377a0
TH		 2184static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2185 u64 tree_id, u64 dirid, char *name)
2186{
2187 struct btrfs_root *root;
2188 struct btrfs_key key;
ac8e9819 2189 char *ptr;
98d377a0
TH		 2190 int ret = -1;
2191 int slot;
2192 int len;
2193 int total_len = 0;
2194 struct btrfs_inode_ref *iref;
2195 struct extent_buffer *l;
2196 struct btrfs_path *path;
2197
2198 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2199 name[0]='\0';
2200 return 0;
2201 }
2202
2203 path = btrfs_alloc_path();
2204 if (!path)
2205 return -ENOMEM;
2206
ac8e9819 2207 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2208
2209 key.objectid = tree_id;
2210 key.type = BTRFS_ROOT_ITEM_KEY;
2211 key.offset = (u64)-1;
2212 root = btrfs_read_fs_root_no_name(info, &key);
2213 if (IS_ERR(root)) {
efe120a0 2214 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
8ad6fcab
CM		 2215 ret = -ENOENT;
2216 goto out;
98d377a0
TH		 2217 }
2218
2219 key.objectid = dirid;
2220 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2221 key.offset = (u64)-1;
98d377a0 2222
67871254 2223 while (1) {
98d377a0
TH		 2224 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2225 if (ret < 0)
2226 goto out;
18674c6c
FDBM		 2227 else if (ret > 0) {
2228 ret = btrfs_previous_item(root, path, dirid,
2229 BTRFS_INODE_REF_KEY);
2230 if (ret < 0)
2231 goto out;
2232 else if (ret > 0) {
2233 ret = -ENOENT;
2234 goto out;
2235 }
2236 }
98d377a0
TH		 2237
2238 l = path->nodes[0];
2239 slot = path->slots[0];
2240 btrfs_item_key_to_cpu(l, &key, slot);
2241
98d377a0
TH		 2242 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2243 len = btrfs_inode_ref_name_len(l, iref);
2244 ptr -= len + 1;
2245 total_len += len + 1;
a696cf35
FDBM		 2246 if (ptr < name) {
2247 ret = -ENAMETOOLONG;
98d377a0 2248 goto out;
a696cf35 2249 }
98d377a0
TH		 2250
2251 *(ptr + len) = '/';
67871254 2252 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2253
2254 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2255 break;
2256
b3b4aa74 2257 btrfs_release_path(path);
98d377a0 2258 key.objectid = key.offset;
8ad6fcab 2259 key.offset = (u64)-1;
98d377a0 2260 dirid = key.objectid;
98d377a0 2261 }
77906a50 2262 memmove(name, ptr, total_len);
67871254 2263 name[total_len] = '\0';
98d377a0
TH		 2264 ret = 0;
2265out:
2266 btrfs_free_path(path);
ac8e9819
CM		 2267 return ret;
2268}
2269
2270static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2271 void __user *argp)
2272{
2273 struct btrfs_ioctl_ino_lookup_args *args;
2274 struct inode *inode;
01b810b8 2275 int ret = 0;
ac8e9819 2276
2354d08f
JL		 2277 args = memdup_user(argp, sizeof(*args));
2278 if (IS_ERR(args))
2279 return PTR_ERR(args);
c2b96929 2280
496ad9aa 2281 inode = file_inode(file);
ac8e9819 2282
01b810b8
DS		 2283 /*
2284 * Unprivileged query to obtain the containing subvolume root id. The
2285 * path is reset so it's consistent with btrfs_search_path_in_tree.
2286 */
1b53ac4d
CM		 2287 if (args->treeid == 0)
2288 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2289
01b810b8
DS		 2290 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2291 args->name[0] = 0;
2292 goto out;
2293 }
2294
2295 if (!capable(CAP_SYS_ADMIN)) {
2296 ret = -EPERM;
2297 goto out;
2298 }
2299
ac8e9819
CM		 2300 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2301 args->treeid, args->objectid,
2302 args->name);
2303
01b810b8 2304out:
ac8e9819
CM		 2305 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2306 ret = -EFAULT;
2307
2308 kfree(args);
98d377a0
TH		 2309 return ret;
2310}
2311
76dda93c
YZ		 2312static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2313 void __user *arg)
2314{
54563d41 2315 struct dentry *parent = file->f_path.dentry;
76dda93c 2316 struct dentry *dentry;
2b0143b5 2317 struct inode *dir = d_inode(parent);
76dda93c
YZ		 2318 struct inode *inode;
2319 struct btrfs_root *root = BTRFS_I(dir)->root;
2320 struct btrfs_root *dest = NULL;
2321 struct btrfs_ioctl_vol_args *vol_args;
2322 struct btrfs_trans_handle *trans;
c58aaad2 2323 struct btrfs_block_rsv block_rsv;
521e0546 2324 u64 root_flags;
c58aaad2 2325 u64 qgroup_reserved;
76dda93c
YZ		 2326 int namelen;
2327 int ret;
2328 int err = 0;
2329
76dda93c
YZ		 2330 vol_args = memdup_user(arg, sizeof(*vol_args));
2331 if (IS_ERR(vol_args))
2332 return PTR_ERR(vol_args);
2333
2334 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2335 namelen = strlen(vol_args->name);
2336 if (strchr(vol_args->name, '/') ||
2337 strncmp(vol_args->name, "..", namelen) == 0) {
2338 err = -EINVAL;
2339 goto out;
2340 }
2341
a561be71 2342 err = mnt_want_write_file(file);
76dda93c
YZ		 2343 if (err)
2344 goto out;
2345
521e0546 2346
5c50c9b8
DS		 2347 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2348 if (err == -EINTR)
e43f998e 2349 goto out_drop_write;
76dda93c
YZ		 2350 dentry = lookup_one_len(vol_args->name, parent, namelen);
2351 if (IS_ERR(dentry)) {
2352 err = PTR_ERR(dentry);
2353 goto out_unlock_dir;
2354 }
2355
2b0143b5 2356 if (d_really_is_negative(dentry)) {
76dda93c
YZ		 2357 err = -ENOENT;
2358 goto out_dput;
2359 }
2360
2b0143b5 2361 inode = d_inode(dentry);
4260f7c7 2362 dest = BTRFS_I(inode)->root;
67871254 2363 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2364 /*
2365 * Regular user. Only allow this with a special mount
2366 * option, when the user has write+exec access to the
2367 * subvol root, and when rmdir(2) would have been
2368 * allowed.
2369 *
2370 * Note that this is _not_ check that the subvol is
2371 * empty or doesn't contain data that we wouldn't
2372 * otherwise be able to delete.
2373 *
2374 * Users who want to delete empty subvols should try
2375 * rmdir(2).
2376 */
2377 err = -EPERM;
2378 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2379 goto out_dput;
2380
2381 /*
2382 * Do not allow deletion if the parent dir is the same
2383 * as the dir to be deleted. That means the ioctl
2384 * must be called on the dentry referencing the root
2385 * of the subvol, not a random directory contained
2386 * within it.
2387 */
2388 err = -EINVAL;
2389 if (root == dest)
2390 goto out_dput;
2391
2392 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2393 if (err)
2394 goto out_dput;
4260f7c7
SW		 2395 }
2396
5c39da5b
MX		 2397 /* check if subvolume may be deleted by a user */
2398 err = btrfs_may_delete(dir, dentry, 1);
2399 if (err)
2400 goto out_dput;
2401
33345d01 2402 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2403 err = -EINVAL;
2404 goto out_dput;
2405 }
2406
76dda93c 2407 mutex_lock(&inode->i_mutex);
521e0546
DS		 2408
2409 /*
2410 * Don't allow to delete a subvolume with send in progress. This is
2411 * inside the i_mutex so the error handling that has to drop the bit
2412 * again is not run concurrently.
2413 */
2414 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2415 root_flags = btrfs_root_flags(&dest->root_item);
2416 if (dest->send_in_progress == 0) {
2417 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2418 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2419 spin_unlock(&dest->root_item_lock);
2420 } else {
2421 spin_unlock(&dest->root_item_lock);
2422 btrfs_warn(root->fs_info,
2423 "Attempt to delete subvolume %llu during send",
c55bfa67 2424 dest->root_key.objectid);
521e0546 2425 err = -EPERM;
909e26dc 2426 goto out_unlock_inode;
521e0546
DS		 2427 }
2428
76dda93c
YZ		 2429 down_write(&root->fs_info->subvol_sem);
2430
2431 err = may_destroy_subvol(dest);
2432 if (err)
2433 goto out_up_write;
2434
c58aaad2
MX		 2435 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2436 /*
2437 * One for dir inode, two for dir entries, two for root
2438 * ref/backref.
2439 */
2440 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2441 5, &qgroup_reserved, true);
c58aaad2
MX		 2442 if (err)
2443 goto out_up_write;
2444
a22285a6
YZ		 2445 trans = btrfs_start_transaction(root, 0);
2446 if (IS_ERR(trans)) {
2447 err = PTR_ERR(trans);
c58aaad2 2448 goto out_release;
a22285a6 2449 }
c58aaad2
MX		 2450 trans->block_rsv = &block_rsv;
2451 trans->bytes_reserved = block_rsv.size;
a22285a6 2452
76dda93c
YZ		 2453 ret = btrfs_unlink_subvol(trans, root, dir,
2454 dest->root_key.objectid,
2455 dentry->d_name.name,
2456 dentry->d_name.len);
79787eaa
JM		 2457 if (ret) {
2458 err = ret;
2459 btrfs_abort_transaction(trans, root, ret);
2460 goto out_end_trans;
2461 }
76dda93c
YZ		 2462
2463 btrfs_record_root_in_trans(trans, dest);
2464
2465 memset(&dest->root_item.drop_progress, 0,
2466 sizeof(dest->root_item.drop_progress));
2467 dest->root_item.drop_level = 0;
2468 btrfs_set_root_refs(&dest->root_item, 0);
2469
27cdeb70 2470 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
d68fc57b
YZ		 2471 ret = btrfs_insert_orphan_item(trans,
2472 root->fs_info->tree_root,
2473 dest->root_key.objectid);
79787eaa
JM		 2474 if (ret) {
2475 btrfs_abort_transaction(trans, root, ret);
2476 err = ret;
2477 goto out_end_trans;
2478 }
d68fc57b 2479 }
dd5f9615
SB		 2480
2481 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2482 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2483 dest->root_key.objectid);
2484 if (ret && ret != -ENOENT) {
2485 btrfs_abort_transaction(trans, root, ret);
2486 err = ret;
2487 goto out_end_trans;
2488 }
2489 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2490 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2491 dest->root_item.received_uuid,
2492 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2493 dest->root_key.objectid);
2494 if (ret && ret != -ENOENT) {
2495 btrfs_abort_transaction(trans, root, ret);
2496 err = ret;
2497 goto out_end_trans;
2498 }
2499 }
2500
79787eaa 2501out_end_trans:
c58aaad2
MX		 2502 trans->block_rsv = NULL;
2503 trans->bytes_reserved = 0;
531cb13f 2504 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2505 if (ret && !err)
2506 err = ret;
76dda93c 2507 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2508out_release:
2509 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2510out_up_write:
2511 up_write(&root->fs_info->subvol_sem);
521e0546
DS		 2512 if (err) {
2513 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2514 root_flags = btrfs_root_flags(&dest->root_item);
2515 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2516 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2517 spin_unlock(&dest->root_item_lock);
2518 }
909e26dc 2519out_unlock_inode:
76dda93c
YZ		 2520 mutex_unlock(&inode->i_mutex);
2521 if (!err) {
64ad6c48 2522 d_invalidate(dentry);
76dda93c
YZ		 2523 btrfs_invalidate_inodes(dest);
2524 d_delete(dentry);
61155aa0 2525 ASSERT(dest->send_in_progress == 0);
fa6ac876
LB		 2526
2527 /* the last ref */
57cdc8db
DS		 2528 if (dest->ino_cache_inode) {
2529 iput(dest->ino_cache_inode);
2530 dest->ino_cache_inode = NULL;
fa6ac876 2531 }
76dda93c
YZ		 2532 }
2533out_dput:
2534 dput(dentry);
2535out_unlock_dir:
2536 mutex_unlock(&dir->i_mutex);
e43f998e 2537out_drop_write:
2a79f17e 2538 mnt_drop_write_file(file);
76dda93c
YZ		 2539out:
2540 kfree(vol_args);
2541 return err;
2542}
2543
1e701a32 2544static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2545{
496ad9aa 2546 struct inode *inode = file_inode(file);
f46b5a66 2547 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2548 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2549 int ret;
2550
25122d15
ID		 2551 ret = mnt_want_write_file(file);
2552 if (ret)
2553 return ret;
b83cc969 2554
25122d15
ID		 2555 if (btrfs_root_readonly(root)) {
2556 ret = -EROFS;
2557 goto out;
5ac00add 2558 }
f46b5a66
CH		 2559
2560 switch (inode->i_mode & S_IFMT) {
2561 case S_IFDIR:
e441d54d
CM		 2562 if (!capable(CAP_SYS_ADMIN)) {
2563 ret = -EPERM;
2564 goto out;
2565 }
de78b51a 2566 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2567 if (ret)
2568 goto out;
de78b51a 2569 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2570 break;
2571 case S_IFREG:
e441d54d
CM		 2572 if (!(file->f_mode & FMODE_WRITE)) {
2573 ret = -EINVAL;
2574 goto out;
2575 }
1e701a32
CM		 2576
2577 range = kzalloc(sizeof(*range), GFP_KERNEL);
2578 if (!range) {
2579 ret = -ENOMEM;
2580 goto out;
2581 }
2582
2583 if (argp) {
2584 if (copy_from_user(range, argp,
2585 sizeof(*range))) {
2586 ret = -EFAULT;
2587 kfree(range);
683be16e 2588 goto out;
1e701a32
CM		 2589 }
2590 /* compression requires us to start the IO */
2591 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2592 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2593 range->extent_thresh = (u32)-1;
2594 }
2595 } else {
2596 /* the rest are all set to zero by kzalloc */
2597 range->len = (u64)-1;
2598 }
496ad9aa 2599 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2600 range, 0, 0);
2601 if (ret > 0)
2602 ret = 0;
1e701a32 2603 kfree(range);
f46b5a66 2604 break;
8929ecfa
YZ		 2605 default:
2606 ret = -EINVAL;
f46b5a66 2607 }
e441d54d 2608out:
25122d15 2609 mnt_drop_write_file(file);
e441d54d 2610 return ret;
f46b5a66
CH		 2611}
2612
b2950863 2613static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2614{
2615 struct btrfs_ioctl_vol_args *vol_args;
2616 int ret;
2617
e441d54d
CM		 2618 if (!capable(CAP_SYS_ADMIN))
2619 return -EPERM;
2620
5ac00add
SB		 2621 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2622 1)) {
e57138b3 2623 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 2624 }
2625
5ac00add 2626 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2627 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2628 if (IS_ERR(vol_args)) {
2629 ret = PTR_ERR(vol_args);
2630 goto out;
2631 }
f46b5a66 2632
5516e595 2633 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2634 ret = btrfs_init_new_device(root, vol_args->name);
2635
43d20761
AJ		 2636 if (!ret)
2637 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2638
f46b5a66 2639 kfree(vol_args);
c9e9f97b
ID		 2640out:
2641 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2642 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2643 return ret;
2644}
2645
da24927b 2646static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2647{
496ad9aa 2648 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2649 struct btrfs_ioctl_vol_args *vol_args;
2650 int ret;
2651
e441d54d
CM		 2652 if (!capable(CAP_SYS_ADMIN))
2653 return -EPERM;
2654
da24927b
MX		 2655 ret = mnt_want_write_file(file);
2656 if (ret)
2657 return ret;
c146afad 2658
dae7b665 2659 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2660 if (IS_ERR(vol_args)) {
2661 ret = PTR_ERR(vol_args);
c47ca32d 2662 goto err_drop;
c9e9f97b 2663 }
f46b5a66 2664
5516e595 2665 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2666
183860f6
AJ		 2667 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2668 1)) {
2669 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2670 goto out;
2671 }
2672
2673 mutex_lock(&root->fs_info->volume_mutex);
2674 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2675 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2676 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6 2677
ec95d491
AJ		 2678 if (!ret)
2679 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2680
183860f6
AJ		 2681out:
2682 kfree(vol_args);
c47ca32d 2683err_drop:
4ac20c70 2684 mnt_drop_write_file(file);
f46b5a66
CH		 2685 return ret;
2686}
2687
475f6387
JS		 2688static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2689{
027ed2f0 2690 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2691 struct btrfs_device *device;
2692 struct btrfs_device *next;
2693 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2694 int ret = 0;
475f6387 2695
027ed2f0
LZ		 2696 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2697 if (!fi_args)
2698 return -ENOMEM;
2699
f7171750 2700 mutex_lock(&fs_devices->device_list_mutex);
027ed2f0
LZ		 2701 fi_args->num_devices = fs_devices->num_devices;
2702 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387 2703
475f6387 2704 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2705 if (device->devid > fi_args->max_id)
2706 fi_args->max_id = device->devid;
475f6387
JS		 2707 }
2708 mutex_unlock(&fs_devices->device_list_mutex);
2709
80a773fb
DS		 2710 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2711 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2712 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2713
027ed2f0
LZ		 2714 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2715 ret = -EFAULT;
475f6387 2716
027ed2f0
LZ		 2717 kfree(fi_args);
2718 return ret;
475f6387
JS		 2719}
2720
2721static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2722{
2723 struct btrfs_ioctl_dev_info_args *di_args;
2724 struct btrfs_device *dev;
2725 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2726 int ret = 0;
2727 char *s_uuid = NULL;
475f6387 2728
475f6387
JS		 2729 di_args = memdup_user(arg, sizeof(*di_args));
2730 if (IS_ERR(di_args))
2731 return PTR_ERR(di_args);
2732
dd5f9615 2733 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2734 s_uuid = di_args->uuid;
2735
2736 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2737 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2738
2739 if (!dev) {
2740 ret = -ENODEV;
2741 goto out;
2742 }
2743
2744 di_args->devid = dev->devid;
7cc8e58d
MX		 2745 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2746 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
475f6387 2747 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2748 if (dev->name) {
606686ee
JB		 2749 struct rcu_string *name;
2750
2751 rcu_read_lock();
2752 name = rcu_dereference(dev->name);
2753 strncpy(di_args->path, name->str, sizeof(di_args->path));
2754 rcu_read_unlock();
a27202fb
JM		 2755 di_args->path[sizeof(di_args->path) - 1] = 0;
2756 } else {
99ba55ad 2757 di_args->path[0] = '\0';
a27202fb 2758 }
475f6387
JS		 2759
2760out:
55793c0d 2761 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2762 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2763 ret = -EFAULT;
2764
2765 kfree(di_args);
2766 return ret;
2767}
2768
f4414602 2769static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
416161db
MF		 2770{
2771 struct page *page;
416161db
MF		 2772 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2773
416161db
MF		 2774 page = grab_cache_page(inode->i_mapping, index);
2775 if (!page)
2776 return NULL;
2777
2778 if (!PageUptodate(page)) {
2779 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2780 0))
2781 return NULL;
2782 lock_page(page);
2783 if (!PageUptodate(page)) {
2784 unlock_page(page);
2785 page_cache_release(page);
2786 return NULL;
2787 }
2788 }
2789 unlock_page(page);
2790
2791 return page;
2792}
2793
f4414602
MF		 2794static int gather_extent_pages(struct inode *inode, struct page **pages,
2795 int num_pages, u64 off)
2796{
2797 int i;
2798 pgoff_t index = off >> PAGE_CACHE_SHIFT;
2799
2800 for (i = 0; i < num_pages; i++) {
2801 pages[i] = extent_same_get_page(inode, index + i);
2802 if (!pages[i])
2803 return -ENOMEM;
2804 }
2805 return 0;
2806}
2807
77fe20dc
MF		 2808static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2809{
2810 /* do any pending delalloc/csum calc on src, one way or
2811 another, and lock file content */
2812 while (1) {
2813 struct btrfs_ordered_extent *ordered;
2814 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2815 ordered = btrfs_lookup_first_ordered_extent(inode,
2816 off + len - 1);
ff5df9b8
FM		 2817 if ((!ordered ||
2818 ordered->file_offset + ordered->len <= off ||
2819 ordered->file_offset >= off + len) &&
77fe20dc 2820 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
ff5df9b8
FM		 2821 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2822 if (ordered)
2823 btrfs_put_ordered_extent(ordered);
77fe20dc 2824 break;
ff5df9b8 2825 }
77fe20dc
MF		 2826 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2827 if (ordered)
2828 btrfs_put_ordered_extent(ordered);
2829 btrfs_wait_ordered_range(inode, off, len);
2830 }
2831}
2832
f4414602 2833static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
416161db 2834{
416161db
MF		 2835 mutex_unlock(&inode1->i_mutex);
2836 mutex_unlock(&inode2->i_mutex);
2837}
2838
f4414602
MF		 2839static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2840{
2841 if (inode1 < inode2)
2842 swap(inode1, inode2);
2843
2844 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2845 if (inode1 != inode2)
2846 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2847}
2848
2849static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2850 struct inode *inode2, u64 loff2, u64 len)
2851{
2852 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2853 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2854}
2855
2856static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2857 struct inode *inode2, u64 loff2, u64 len)
416161db
MF		 2858{
2859 if (inode1 < inode2) {
2860 swap(inode1, inode2);
2861 swap(loff1, loff2);
2862 }
416161db 2863 lock_extent_range(inode1, loff1, len);
f4414602 2864 if (inode1 != inode2)
416161db 2865 lock_extent_range(inode2, loff2, len);
f4414602
MF		 2866}
2867
2868struct cmp_pages {
2869 int num_pages;
2870 struct page **src_pages;
2871 struct page **dst_pages;
2872};
2873
2874static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2875{
2876 int i;
2877 struct page *pg;
2878
2879 for (i = 0; i < cmp->num_pages; i++) {
2880 pg = cmp->src_pages[i];
2881 if (pg)
2882 page_cache_release(pg);
2883 pg = cmp->dst_pages[i];
2884 if (pg)
2885 page_cache_release(pg);
2886 }
2887 kfree(cmp->src_pages);
2888 kfree(cmp->dst_pages);
2889}
2890
2891static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2892 struct inode *dst, u64 dst_loff,
2893 u64 len, struct cmp_pages *cmp)
2894{
2895 int ret;
2896 int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
2897 struct page **src_pgarr, **dst_pgarr;
2898
2899 /*
2900 * We must gather up all the pages before we initiate our
2901 * extent locking. We use an array for the page pointers. Size
2902 * of the array is bounded by len, which is in turn bounded by
2903 * BTRFS_MAX_DEDUPE_LEN.
2904 */
2905 src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2906 dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2907 if (!src_pgarr || !dst_pgarr) {
2908 kfree(src_pgarr);
2909 kfree(dst_pgarr);
2910 return -ENOMEM;
416161db 2911 }
f4414602
MF		 2912 cmp->num_pages = num_pages;
2913 cmp->src_pages = src_pgarr;
2914 cmp->dst_pages = dst_pgarr;
2915
2916 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2917 if (ret)
2918 goto out;
2919
2920 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2921
2922out:
2923 if (ret)
2924 btrfs_cmp_data_free(cmp);
2925 return 0;
416161db
MF		 2926}
2927
2928static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
f4414602 2929 u64 dst_loff, u64 len, struct cmp_pages *cmp)
416161db
MF		 2930{
2931 int ret = 0;
f4414602 2932 int i;
416161db
MF		 2933 struct page *src_page, *dst_page;
2934 unsigned int cmp_len = PAGE_CACHE_SIZE;
2935 void *addr, *dst_addr;
2936
f4414602 2937 i = 0;
416161db
MF		 2938 while (len) {
2939 if (len < PAGE_CACHE_SIZE)
2940 cmp_len = len;
2941
f4414602
MF		 2942 BUG_ON(i >= cmp->num_pages);
2943
2944 src_page = cmp->src_pages[i];
2945 dst_page = cmp->dst_pages[i];
2946
416161db
MF		 2947 addr = kmap_atomic(src_page);
2948 dst_addr = kmap_atomic(dst_page);
2949
2950 flush_dcache_page(src_page);
2951 flush_dcache_page(dst_page);
2952
2953 if (memcmp(addr, dst_addr, cmp_len))
2954 ret = BTRFS_SAME_DATA_DIFFERS;
2955
2956 kunmap_atomic(addr);
2957 kunmap_atomic(dst_addr);
416161db
MF		 2958
2959 if (ret)
2960 break;
2961
416161db 2962 len -= cmp_len;
f4414602 2963 i++;
416161db
MF		 2964 }
2965
2966 return ret;
2967}
2968
e1d227a4
MF		 2969static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
2970 u64 olen)
416161db 2971{
e1d227a4 2972 u64 len = *plen;
416161db
MF		 2973 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2974
e1d227a4 2975 if (off + olen > inode->i_size || off + olen < off)
416161db 2976 return -EINVAL;
e1d227a4
MF		 2977
2978 /* if we extend to eof, continue to block boundary */
2979 if (off + len == inode->i_size)
2980 *plen = len = ALIGN(inode->i_size, bs) - off;
2981
416161db
MF		 2982 /* Check that we are block aligned - btrfs_clone() requires this */
2983 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2984 return -EINVAL;
2985
2986 return 0;
2987}
2988
e1d227a4 2989static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
416161db
MF		 2990 struct inode *dst, u64 dst_loff)
2991{
2992 int ret;
e1d227a4 2993 u64 len = olen;
f4414602 2994 struct cmp_pages cmp;
0efa9f48
MF		 2995 int same_inode = 0;
2996 u64 same_lock_start = 0;
2997 u64 same_lock_len = 0;
416161db 2998
416161db 2999 if (src == dst)
0efa9f48 3000 same_inode = 1;
416161db 3001
113e8283
FM		 3002 if (len == 0)
3003 return 0;
3004
0efa9f48
MF		 3005 if (same_inode) {
3006 mutex_lock(&src->i_mutex);
416161db 3007
0efa9f48
MF		 3008 ret = extent_same_check_offsets(src, loff, &len, olen);
3009 if (ret)
3010 goto out_unlock;
416161db 3011
0efa9f48
MF		 3012 /*
3013 * Single inode case wants the same checks, except we
3014 * don't want our length pushed out past i_size as
3015 * comparing that data range makes no sense.
3016 *
3017 * extent_same_check_offsets() will do this for an
3018 * unaligned length at i_size, so catch it here and
3019 * reject the request.
3020 *
3021 * This effectively means we require aligned extents
3022 * for the single-inode case, whereas the other cases
3023 * allow an unaligned length so long as it ends at
3024 * i_size.
3025 */
3026 if (len != olen) {
3027 ret = -EINVAL;
3028 goto out_unlock;
3029 }
3030
3031 /* Check for overlapping ranges */
3032 if (dst_loff + len > loff && dst_loff < loff + len) {
3033 ret = -EINVAL;
3034 goto out_unlock;
3035 }
3036
3037 same_lock_start = min_t(u64, loff, dst_loff);
3038 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3039 } else {
3040 btrfs_double_inode_lock(src, dst);
3041
3042 ret = extent_same_check_offsets(src, loff, &len, olen);
3043 if (ret)
3044 goto out_unlock;
3045
3046 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3047 if (ret)
3048 goto out_unlock;
3049 }
416161db
MF		 3050
3051 /* don't make the dst file partly checksummed */
3052 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3053 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3054 ret = -EINVAL;
3055 goto out_unlock;
3056 }
3057
f4414602
MF		 3058 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3059 if (ret)
3060 goto out_unlock;
3061
0efa9f48
MF		 3062 if (same_inode)
3063 lock_extent_range(src, same_lock_start, same_lock_len);
3064 else
3065 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
f4414602 3066
207910dd 3067 /* pass original length for comparison so we stay within i_size */
f4414602 3068 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
416161db 3069 if (ret == 0)
1c919a5e 3070 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
416161db 3071
0efa9f48
MF		 3072 if (same_inode)
3073 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3074 same_lock_start + same_lock_len - 1);
3075 else
3076 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
f4414602
MF		 3077
3078 btrfs_cmp_data_free(&cmp);
416161db 3079out_unlock:
0efa9f48
MF		 3080 if (same_inode)
3081 mutex_unlock(&src->i_mutex);
3082 else
3083 btrfs_double_inode_unlock(src, dst);
416161db
MF		 3084
3085 return ret;
3086}
3087
3088#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
3089
3090static long btrfs_ioctl_file_extent_same(struct file *file,
1c1c8747 3091 struct btrfs_ioctl_same_args __user *argp)
416161db 3092{
497b4050 3093 struct btrfs_ioctl_same_args *same = NULL;
cbf8b8ca 3094 struct btrfs_ioctl_same_extent_info *info;
1c1c8747 3095 struct inode *src = file_inode(file);
416161db
MF		 3096 u64 off;
3097 u64 len;
3098 int i;
3099 int ret;
cbf8b8ca 3100 unsigned long size;
416161db
MF		 3101 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3102 bool is_admin = capable(CAP_SYS_ADMIN);
1c1c8747 3103 u16 count;
416161db
MF		 3104
3105 if (!(file->f_mode & FMODE_READ))
3106 return -EINVAL;
3107
3108 ret = mnt_want_write_file(file);
3109 if (ret)
3110 return ret;
3111
1c1c8747 3112 if (get_user(count, &argp->dest_count)) {
416161db
MF		 3113 ret = -EFAULT;
3114 goto out;
3115 }
3116
1c1c8747 3117 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
cbf8b8ca 3118
1c1c8747 3119 same = memdup_user(argp, size);
cbf8b8ca 3120
229eed43
GB		 3121 if (IS_ERR(same)) {
3122 ret = PTR_ERR(same);
497b4050 3123 same = NULL;
cbf8b8ca
MF		 3124 goto out;
3125 }
3126
3127 off = same->logical_offset;
3128 len = same->length;
416161db
MF		 3129
3130 /*
3131 * Limit the total length we will dedupe for each operation.
3132 * This is intended to bound the total time spent in this
3133 * ioctl to something sane.
3134 */
3135 if (len > BTRFS_MAX_DEDUPE_LEN)
3136 len = BTRFS_MAX_DEDUPE_LEN;
3137
3138 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3139 /*
3140 * Btrfs does not support blocksize < page_size. As a
3141 * result, btrfs_cmp_data() won't correctly handle
3142 * this situation without an update.
3143 */
3144 ret = -EINVAL;
3145 goto out;
3146 }
3147
3148 ret = -EISDIR;
3149 if (S_ISDIR(src->i_mode))
3150 goto out;
3151
3152 ret = -EACCES;
3153 if (!S_ISREG(src->i_mode))
3154 goto out;
3155
cbf8b8ca 3156 /* pre-format output fields to sane values */
1c1c8747 3157 for (i = 0; i < count; i++) {
cbf8b8ca
MF		 3158 same->info[i].bytes_deduped = 0ULL;
3159 same->info[i].status = 0;
3160 }
416161db 3161
1c1c8747
AV		 3162 for (i = 0, info = same->info; i < count; i++, info++) {
3163 struct inode *dst;
3164 struct fd dst_file = fdget(info->fd);
3165 if (!dst_file.file) {
cbf8b8ca 3166 info->status = -EBADF;
1c1c8747 3167 continue;
416161db 3168 }
1c1c8747 3169 dst = file_inode(dst_file.file);
416161db 3170
1c1c8747 3171 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
cbf8b8ca 3172 info->status = -EINVAL;
1c1c8747
AV		 3173 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3174 info->status = -EXDEV;
3175 } else if (S_ISDIR(dst->i_mode)) {
cbf8b8ca 3176 info->status = -EISDIR;
1c1c8747 3177 } else if (!S_ISREG(dst->i_mode)) {
cbf8b8ca 3178 info->status = -EACCES;
1c1c8747
AV		 3179 } else {
3180 info->status = btrfs_extent_same(src, off, len, dst,
3181 info->logical_offset);
3182 if (info->status == 0)
3183 info->bytes_deduped += len;
416161db 3184 }
1c1c8747 3185 fdput(dst_file);
416161db
MF		 3186 }
3187
cbf8b8ca
MF		 3188 ret = copy_to_user(argp, same, size);
3189 if (ret)
3190 ret = -EFAULT;
3191
416161db
MF		 3192out:
3193 mnt_drop_write_file(file);
497b4050 3194 kfree(same);
416161db
MF		 3195 return ret;
3196}
3197
fcebe456
JB		 3198/* Helper to check and see if this root currently has a ref on the given disk
3199 * bytenr. If it does then we need to update the quota for this root. This
3200 * doesn't do anything if quotas aren't enabled.
3201 */
3202static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3203 u64 disko)
3204{
3284da7b 3205 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
fcebe456
JB		 3206 struct ulist *roots;
3207 struct ulist_iterator uiter;
3208 struct ulist_node *root_node = NULL;
3209 int ret;
3210
3211 if (!root->fs_info->quota_enabled)
3212 return 1;
3213
3214 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3215 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3216 tree_mod_seq_elem.seq, &roots);
3217 if (ret < 0)
3218 goto out;
3219 ret = 0;
3220 ULIST_ITER_INIT(&uiter);
3221 while ((root_node = ulist_next(roots, &uiter))) {
3222 if (root_node->val == root->objectid) {
3223 ret = 1;
3224 break;
3225 }
3226 }
3227 ulist_free(roots);
3228out:
3229 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3230 return ret;
3231}
3232
f82a9901
FM		 3233static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3234 struct inode *inode,
3235 u64 endoff,
3236 const u64 destoff,
1c919a5e
MF		 3237 const u64 olen,
3238 int no_time_update)
f82a9901
FM		 3239{
3240 struct btrfs_root *root = BTRFS_I(inode)->root;
3241 int ret;
3242
3243 inode_inc_iversion(inode);
1c919a5e
MF		 3244 if (!no_time_update)
3245 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
f82a9901
FM		 3246 /*
3247 * We round up to the block size at eof when determining which
3248 * extents to clone above, but shouldn't round up the file size.
3249 */
3250 if (endoff > destoff + olen)
3251 endoff = destoff + olen;
3252 if (endoff > inode->i_size)
3253 btrfs_i_size_write(inode, endoff);
3254
3255 ret = btrfs_update_inode(trans, root, inode);
3256 if (ret) {
3257 btrfs_abort_transaction(trans, root, ret);
3258 btrfs_end_transaction(trans, root);
3259 goto out;
3260 }
3261 ret = btrfs_end_transaction(trans, root);
3262out:
3263 return ret;
3264}
3265
7ffbb598
FM		 3266static void clone_update_extent_map(struct inode *inode,
3267 const struct btrfs_trans_handle *trans,
3268 const struct btrfs_path *path,
7ffbb598
FM		 3269 const u64 hole_offset,
3270 const u64 hole_len)
3271{
3272 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3273 struct extent_map *em;
3274 int ret;
3275
3276 em = alloc_extent_map();
3277 if (!em) {
3278 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3279 &BTRFS_I(inode)->runtime_flags);
3280 return;
3281 }
3282
14f59796
FM		 3283 if (path) {
3284 struct btrfs_file_extent_item *fi;
3285
3286 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3287 struct btrfs_file_extent_item);
7ffbb598
FM		 3288 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3289 em->generation = -1;
3290 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3291 BTRFS_FILE_EXTENT_INLINE)
3292 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3293 &BTRFS_I(inode)->runtime_flags);
3294 } else {
3295 em->start = hole_offset;
3296 em->len = hole_len;
3297 em->ram_bytes = em->len;
3298 em->orig_start = hole_offset;
3299 em->block_start = EXTENT_MAP_HOLE;
3300 em->block_len = 0;
3301 em->orig_block_len = 0;
3302 em->compress_type = BTRFS_COMPRESS_NONE;
3303 em->generation = trans->transid;
3304 }
3305
3306 while (1) {
3307 write_lock(&em_tree->lock);
3308 ret = add_extent_mapping(em_tree, em, 1);
3309 write_unlock(&em_tree->lock);
3310 if (ret != -EEXIST) {
3311 free_extent_map(em);
3312 break;
3313 }
3314 btrfs_drop_extent_cache(inode, em->start,
3315 em->start + em->len - 1, 0);
3316 }
3317
ee39b432 3318 if (ret)
7ffbb598
FM		 3319 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3320 &BTRFS_I(inode)->runtime_flags);
3321}
3322
32b7c687
MF		 3323/**
3324 * btrfs_clone() - clone a range from inode file to another
3325 *
3326 * @src: Inode to clone from
3327 * @inode: Inode to clone to
3328 * @off: Offset within source to start clone from
3329 * @olen: Original length, passed by user, of range to clone
1c919a5e 3330 * @olen_aligned: Block-aligned value of olen
32b7c687 3331 * @destoff: Offset within @inode to start clone
1c919a5e 3332 * @no_time_update: Whether to update mtime/ctime on the target inode
32b7c687
MF		 3333 */
3334static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901 3335 const u64 off, const u64 olen, const u64 olen_aligned,
1c919a5e 3336 const u64 destoff, int no_time_update)
f46b5a66 3337{
f46b5a66 3338 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3339 struct btrfs_path *path = NULL;
f46b5a66 3340 struct extent_buffer *leaf;
32b7c687
MF		 3341 struct btrfs_trans_handle *trans;
3342 char *buf = NULL;
ae01a0ab 3343 struct btrfs_key key;
f46b5a66
CH		 3344 u32 nritems;
3345 int slot;
ae01a0ab 3346 int ret;
fcebe456 3347 int no_quota;
f82a9901 3348 const u64 len = olen_aligned;
fcebe456 3349 u64 last_disko = 0;
f82a9901 3350 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3351
3352 ret = -ENOMEM;
707e8a07 3353 buf = vmalloc(root->nodesize);
ae01a0ab 3354 if (!buf)
32b7c687 3355 return ret;
ae01a0ab
YZ		 3356
3357 path = btrfs_alloc_path();
3358 if (!path) {
3359 vfree(buf);
32b7c687 3360 return ret;
d525e8ab
LZ		 3361 }
3362
32b7c687 3363 path->reada = 2;
c5c9cd4d 3364 /* clone data */
33345d01 3365 key.objectid = btrfs_ino(src);
ae01a0ab 3366 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3367 key.offset = off;
f46b5a66
CH		 3368
3369 while (1) {
de249e66 3370 u64 next_key_min_offset = key.offset + 1;
df858e76 3371
f46b5a66
CH		 3372 /*
3373 * note the key will change type as we walk through the
3374 * tree.
3375 */
e4355f34 3376 path->leave_spinning = 1;
362a20c5
DS		 3377 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3378 0, 0);
f46b5a66
CH		 3379 if (ret < 0)
3380 goto out;
2c463823
FM		 3381 /*
3382 * First search, if no extent item that starts at offset off was
3383 * found but the previous item is an extent item, it's possible
3384 * it might overlap our target range, therefore process it.
3385 */
3386 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3387 btrfs_item_key_to_cpu(path->nodes[0], &key,
3388 path->slots[0] - 1);
3389 if (key.type == BTRFS_EXTENT_DATA_KEY)
3390 path->slots[0]--;
3391 }
f46b5a66 3392
ae01a0ab 3393 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3394process_slot:
fcebe456 3395 no_quota = 1;
ae01a0ab 3396 if (path->slots[0] >= nritems) {
362a20c5 3397 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3398 if (ret < 0)
3399 goto out;
3400 if (ret > 0)
3401 break;
ae01a0ab 3402 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3403 }
3404 leaf = path->nodes[0];
3405 slot = path->slots[0];
f46b5a66 3406
ae01a0ab 3407 btrfs_item_key_to_cpu(leaf, &key, slot);
962a298f 3408 if (key.type > BTRFS_EXTENT_DATA_KEY ||
33345d01 3409 key.objectid != btrfs_ino(src))
f46b5a66
CH		 3410 break;
3411
962a298f 3412 if (key.type == BTRFS_EXTENT_DATA_KEY) {
c5c9cd4d
SW		 3413 struct btrfs_file_extent_item *extent;
3414 int type;
31840ae1
ZY		 3415 u32 size;
3416 struct btrfs_key new_key;
c5c9cd4d
SW		 3417 u64 disko = 0, diskl = 0;
3418 u64 datao = 0, datal = 0;
3419 u8 comp;
f82a9901 3420 u64 drop_start;
31840ae1 3421
c5c9cd4d
SW		 3422 extent = btrfs_item_ptr(leaf, slot,
3423 struct btrfs_file_extent_item);
3424 comp = btrfs_file_extent_compression(leaf, extent);
3425 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3426 if (type == BTRFS_FILE_EXTENT_REG ||
3427 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3428 disko = btrfs_file_extent_disk_bytenr(leaf,
3429 extent);
3430 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3431 extent);
c5c9cd4d 3432 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3433 datal = btrfs_file_extent_num_bytes(leaf,
3434 extent);
c5c9cd4d
SW		 3435 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3436 /* take upper bound, may be compressed */
3437 datal = btrfs_file_extent_ram_bytes(leaf,
3438 extent);
3439 }
31840ae1 3440
2c463823
FM		 3441 /*
3442 * The first search might have left us at an extent
3443 * item that ends before our target range's start, can
3444 * happen if we have holes and NO_HOLES feature enabled.
3445 */
3446 if (key.offset + datal <= off) {
e4355f34
FDBM		 3447 path->slots[0]++;
3448 goto process_slot;
2c463823
FM		 3449 } else if (key.offset >= off + len) {
3450 break;
e4355f34 3451 }
df858e76 3452 next_key_min_offset = key.offset + datal;
e4355f34
FDBM		 3453 size = btrfs_item_size_nr(leaf, slot);
3454 read_extent_buffer(leaf, buf,
3455 btrfs_item_ptr_offset(leaf, slot),
3456 size);
3457
3458 btrfs_release_path(path);
3459 path->leave_spinning = 0;
c5c9cd4d 3460
31840ae1 3461 memcpy(&new_key, &key, sizeof(new_key));
33345d01 3462 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 3463 if (off <= key.offset)
3464 new_key.offset = key.offset + destoff - off;
3465 else
3466 new_key.offset = destoff;
31840ae1 3467
f82a9901
FM		 3468 /*
3469 * Deal with a hole that doesn't have an extent item
3470 * that represents it (NO_HOLES feature enabled).
3471 * This hole is either in the middle of the cloning
3472 * range or at the beginning (fully overlaps it or
3473 * partially overlaps it).
3474 */
3475 if (new_key.offset != last_dest_end)
3476 drop_start = last_dest_end;
3477 else
3478 drop_start = new_key.offset;
3479
b6f3409b
SW		 3480 /*
3481 * 1 - adjusting old extent (we may have to split it)
3482 * 1 - add new extent
3483 * 1 - inode update
3484 */
3485 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3486 if (IS_ERR(trans)) {
3487 ret = PTR_ERR(trans);
3488 goto out;
3489 }
3490
c8a894d7
CM		 3491 if (type == BTRFS_FILE_EXTENT_REG ||
3492 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3493 /*
3494 * a | --- range to clone ---| b
3495 * | ------------- extent ------------- |
3496 */
3497
93915584 3498 /* subtract range b */
d72c0842
LZ		 3499 if (key.offset + datal > off + len)
3500 datal = off + len - key.offset;
3501
93915584 3502 /* subtract range a */
a22285a6
YZ		 3503 if (off > key.offset) {
3504 datao += off - key.offset;
3505 datal -= off - key.offset;
3506 }
3507
5dc562c5 3508 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3509 drop_start,
a22285a6 3510 new_key.offset + datal,
2671485d 3511 1);
79787eaa 3512 if (ret) {
3f9e3df8 3513 if (ret != -EOPNOTSUPP)
00fdf13a
LB		 3514 btrfs_abort_transaction(trans,
3515 root, ret);
79787eaa
JM		 3516 btrfs_end_transaction(trans, root);
3517 goto out;
3518 }
a22285a6 3519
c5c9cd4d
SW		 3520 ret = btrfs_insert_empty_item(trans, root, path,
3521 &new_key, size);
79787eaa
JM		 3522 if (ret) {
3523 btrfs_abort_transaction(trans, root,
3524 ret);
3525 btrfs_end_transaction(trans, root);
3526 goto out;
3527 }
c5c9cd4d
SW		 3528
3529 leaf = path->nodes[0];
3530 slot = path->slots[0];
3531 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3532 btrfs_item_ptr_offset(leaf, slot),
3533 size);
ae01a0ab 3534
c5c9cd4d 3535 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3536 struct btrfs_file_extent_item);
c5c9cd4d 3537
c5c9cd4d
SW		 3538 /* disko == 0 means it's a hole */
3539 if (!disko)
3540 datao = 0;
c5c9cd4d
SW		 3541
3542 btrfs_set_file_extent_offset(leaf, extent,
3543 datao);
3544 btrfs_set_file_extent_num_bytes(leaf, extent,
3545 datal);
fcebe456
JB		 3546
3547 /*
3548 * We need to look up the roots that point at
3549 * this bytenr and see if the new root does. If
3550 * it does not we need to make sure we update
3551 * quotas appropriately.
3552 */
3553 if (disko && root != BTRFS_I(src)->root &&
3554 disko != last_disko) {
3555 no_quota = check_ref(trans, root,
3556 disko);
3557 if (no_quota < 0) {
3558 btrfs_abort_transaction(trans,
3559 root,
3560 ret);
3561 btrfs_end_transaction(trans,
3562 root);
3563 ret = no_quota;
3564 goto out;
3565 }
3566 }
3567
c5c9cd4d
SW		 3568 if (disko) {
3569 inode_add_bytes(inode, datal);
ae01a0ab 3570 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3571 disko, diskl, 0,
3572 root->root_key.objectid,
33345d01 3573 btrfs_ino(inode),
66d7e7f0 3574 new_key.offset - datao,
fcebe456 3575 no_quota);
79787eaa
JM		 3576 if (ret) {
3577 btrfs_abort_transaction(trans,
3578 root,
3579 ret);
3580 btrfs_end_transaction(trans,
3581 root);
3582 goto out;
3583
3584 }
f46b5a66 3585 }
c5c9cd4d
SW		 3586 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3587 u64 skip = 0;
3588 u64 trim = 0;
d3ecfcdf
LB		 3589 u64 aligned_end = 0;
3590
ed958762
FM		 3591 /*
3592 * Don't copy an inline extent into an offset
3593 * greater than zero. Having an inline extent
3594 * at such an offset results in chaos as btrfs
3595 * isn't prepared for such cases. Just skip
3596 * this case for the same reasons as commented
3597 * at btrfs_ioctl_clone().
3598 */
3599 if (last_dest_end > 0) {
3600 ret = -EOPNOTSUPP;
3601 btrfs_end_transaction(trans, root);
3602 goto out;
3603 }
3604
c5c9cd4d
SW		 3605 if (off > key.offset) {
3606 skip = off - key.offset;
3607 new_key.offset += skip;
3608 }
d397712b 3609
aa42ffd9
LB		 3610 if (key.offset + datal > off + len)
3611 trim = key.offset + datal - (off + len);
d397712b 3612
c5c9cd4d 3613 if (comp && (skip || trim)) {
c5c9cd4d 3614 ret = -EINVAL;
a22285a6 3615 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3616 goto out;
3617 }
3618 size -= skip + trim;
3619 datal -= skip + trim;
a22285a6 3620
d3ecfcdf
LB		 3621 aligned_end = ALIGN(new_key.offset + datal,
3622 root->sectorsize);
5dc562c5 3623 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3624 drop_start,
d3ecfcdf 3625 aligned_end,
2671485d 3626 1);
79787eaa 3627 if (ret) {
3f9e3df8 3628 if (ret != -EOPNOTSUPP)
3a29bc09
CM		 3629 btrfs_abort_transaction(trans,
3630 root, ret);
79787eaa
JM		 3631 btrfs_end_transaction(trans, root);
3632 goto out;
3633 }
a22285a6 3634
c5c9cd4d
SW		 3635 ret = btrfs_insert_empty_item(trans, root, path,
3636 &new_key, size);
79787eaa
JM		 3637 if (ret) {
3638 btrfs_abort_transaction(trans, root,
3639 ret);
3640 btrfs_end_transaction(trans, root);
3641 goto out;
3642 }
c5c9cd4d
SW		 3643
3644 if (skip) {
d397712b
CM		 3645 u32 start =
3646 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3647 memmove(buf+start, buf+start+skip,
3648 datal);
3649 }
3650
3651 leaf = path->nodes[0];
3652 slot = path->slots[0];
3653 write_extent_buffer(leaf, buf,
3654 btrfs_item_ptr_offset(leaf, slot),
3655 size);
3656 inode_add_bytes(inode, datal);
f46b5a66 3657 }
c5c9cd4d 3658
7ffbb598
FM		 3659 /* If we have an implicit hole (NO_HOLES feature). */
3660 if (drop_start < new_key.offset)
3661 clone_update_extent_map(inode, trans,
14f59796 3662 NULL, drop_start,
7ffbb598
FM		 3663 new_key.offset - drop_start);
3664
14f59796 3665 clone_update_extent_map(inode, trans, path, 0, 0);
7ffbb598 3666
c5c9cd4d 3667 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3668 btrfs_release_path(path);
c5c9cd4d 3669
62e2390e
FM		 3670 last_dest_end = ALIGN(new_key.offset + datal,
3671 root->sectorsize);
f82a9901
FM		 3672 ret = clone_finish_inode_update(trans, inode,
3673 last_dest_end,
1c919a5e
MF		 3674 destoff, olen,
3675 no_time_update);
f82a9901 3676 if (ret)
79787eaa 3677 goto out;
2c463823
FM		 3678 if (new_key.offset + datal >= destoff + len)
3679 break;
a22285a6 3680 }
b3b4aa74 3681 btrfs_release_path(path);
df858e76 3682 key.offset = next_key_min_offset;
f46b5a66 3683 }
f46b5a66 3684 ret = 0;
32b7c687 3685
f82a9901
FM		 3686 if (last_dest_end < destoff + len) {
3687 /*
3688 * We have an implicit hole (NO_HOLES feature is enabled) that
3689 * fully or partially overlaps our cloning range at its end.
3690 */
3691 btrfs_release_path(path);
3692
3693 /*
3694 * 1 - remove extent(s)
3695 * 1 - inode update
3696 */
3697 trans = btrfs_start_transaction(root, 2);
3698 if (IS_ERR(trans)) {
3699 ret = PTR_ERR(trans);
3700 goto out;
3701 }
3702 ret = btrfs_drop_extents(trans, root, inode,
3703 last_dest_end, destoff + len, 1);
3704 if (ret) {
3705 if (ret != -EOPNOTSUPP)
3706 btrfs_abort_transaction(trans, root, ret);
3707 btrfs_end_transaction(trans, root);
3708 goto out;
3709 }
14f59796
FM		 3710 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3711 destoff + len - last_dest_end);
f82a9901 3712 ret = clone_finish_inode_update(trans, inode, destoff + len,
1c919a5e 3713 destoff, olen, no_time_update);
f82a9901
FM		 3714 }
3715
f46b5a66 3716out:
32b7c687
MF		 3717 btrfs_free_path(path);
3718 vfree(buf);
3719 return ret;
3720}
3721
3722static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3723 u64 off, u64 olen, u64 destoff)
3724{
54563d41 3725 struct inode *inode = file_inode(file);
32b7c687
MF		 3726 struct btrfs_root *root = BTRFS_I(inode)->root;
3727 struct fd src_file;
3728 struct inode *src;
3729 int ret;
3730 u64 len = olen;
3731 u64 bs = root->fs_info->sb->s_blocksize;
3732 int same_inode = 0;
3733
3734 /*
3735 * TODO:
3736 * - split compressed inline extents. annoying: we need to
3737 * decompress into destination's address_space (the file offset
3738 * may change, so source mapping won't do), then recompress (or
3739 * otherwise reinsert) a subrange.
00fdf13a
LB		 3740 *
3741 * - split destination inode's inline extents. The inline extents can
3742 * be either compressed or non-compressed.
32b7c687
MF		 3743 */
3744
3745 /* the destination must be opened for writing */
3746 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3747 return -EINVAL;
3748
3749 if (btrfs_root_readonly(root))
3750 return -EROFS;
3751
3752 ret = mnt_want_write_file(file);
3753 if (ret)
3754 return ret;
3755
3756 src_file = fdget(srcfd);
3757 if (!src_file.file) {
3758 ret = -EBADF;
3759 goto out_drop_write;
3760 }
3761
3762 ret = -EXDEV;
3763 if (src_file.file->f_path.mnt != file->f_path.mnt)
3764 goto out_fput;
3765
3766 src = file_inode(src_file.file);
3767
3768 ret = -EINVAL;
3769 if (src == inode)
3770 same_inode = 1;
3771
3772 /* the src must be open for reading */
3773 if (!(src_file.file->f_mode & FMODE_READ))
3774 goto out_fput;
3775
3776 /* don't make the dst file partly checksummed */
3777 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3778 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3779 goto out_fput;
3780
3781 ret = -EISDIR;
3782 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3783 goto out_fput;
3784
3785 ret = -EXDEV;
3786 if (src->i_sb != inode->i_sb)
3787 goto out_fput;
3788
3789 if (!same_inode) {
3790 if (inode < src) {
3791 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3792 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3793 } else {
3794 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3795 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3796 }
3797 } else {
3798 mutex_lock(&src->i_mutex);
3799 }
3800
3801 /* determine range to clone */
3802 ret = -EINVAL;
3803 if (off + len > src->i_size || off + len < off)
3804 goto out_unlock;
3805 if (len == 0)
3806 olen = len = src->i_size - off;
3807 /* if we extend to eof, continue to block boundary */
3808 if (off + len == src->i_size)
3809 len = ALIGN(src->i_size, bs) - off;
3810
ccccf3d6
FM		 3811 if (len == 0) {
3812 ret = 0;
3813 goto out_unlock;
3814 }
3815
32b7c687
MF		 3816 /* verify the end result is block aligned */
3817 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3818 !IS_ALIGNED(destoff, bs))
3819 goto out_unlock;
3820
3821 /* verify if ranges are overlapped within the same file */
3822 if (same_inode) {
3823 if (destoff + len > off && destoff < off + len)
3824 goto out_unlock;
3825 }
3826
3827 if (destoff > inode->i_size) {
3828 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3829 if (ret)
3830 goto out_unlock;
3831 }
3832
c125b8bf
FM		 3833 /*
3834 * Lock the target range too. Right after we replace the file extent
3835 * items in the fs tree (which now point to the cloned data), we might
3836 * have a worker replace them with extent items relative to a write
3837 * operation that was issued before this clone operation (i.e. confront
3838 * with inode.c:btrfs_finish_ordered_io).
3839 */
3840 if (same_inode) {
3841 u64 lock_start = min_t(u64, off, destoff);
3842 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
32b7c687 3843
c125b8bf
FM		 3844 lock_extent_range(src, lock_start, lock_len);
3845 } else {
3846 lock_extent_range(src, off, len);
3847 lock_extent_range(inode, destoff, len);
3848 }
32b7c687 3849
1c919a5e 3850 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
32b7c687 3851
c125b8bf
FM		 3852 if (same_inode) {
3853 u64 lock_start = min_t(u64, off, destoff);
3854 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3855
3856 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3857 } else {
3858 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3859 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3860 destoff + len - 1);
3861 }
3862 /*
3863 * Truncate page cache pages so that future reads will see the cloned
3864 * data immediately and not the previous data.
3865 */
3866 truncate_inode_pages_range(&inode->i_data, destoff,
3867 PAGE_CACHE_ALIGN(destoff + len) - 1);
f46b5a66 3868out_unlock:
c57c2b3e
FDBM		 3869 if (!same_inode) {
3870 if (inode < src) {
3871 mutex_unlock(&src->i_mutex);
3872 mutex_unlock(&inode->i_mutex);
3873 } else {
3874 mutex_unlock(&inode->i_mutex);
3875 mutex_unlock(&src->i_mutex);
3876 }
3877 } else {
3878 mutex_unlock(&src->i_mutex);
3879 }
f46b5a66 3880out_fput:
2903ff01 3881 fdput(src_file);
ab67b7c1 3882out_drop_write:
2a79f17e 3883 mnt_drop_write_file(file);
f46b5a66
CH		 3884 return ret;
3885}
3886
7a865e8a 3887static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3888{
3889 struct btrfs_ioctl_clone_range_args args;
3890
7a865e8a 3891 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3892 return -EFAULT;
3893 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3894 args.src_length, args.dest_offset);
3895}
3896
f46b5a66
CH		 3897/*
3898 * there are many ways the trans_start and trans_end ioctls can lead
3899 * to deadlocks. They should only be used by applications that
3900 * basically own the machine, and have a very in depth understanding
3901 * of all the possible deadlocks and enospc problems.
3902 */
b2950863 3903static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3904{
496ad9aa 3905 struct inode *inode = file_inode(file);
f46b5a66
CH		 3906 struct btrfs_root *root = BTRFS_I(inode)->root;
3907 struct btrfs_trans_handle *trans;
1ab86aed 3908 int ret;
f46b5a66 3909
1ab86aed 3910 ret = -EPERM;
df5b5520 3911 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3912 goto out;
df5b5520 3913
1ab86aed
SW		 3914 ret = -EINPROGRESS;
3915 if (file->private_data)
f46b5a66 3916 goto out;
9ca9ee09 3917
b83cc969
LZ		 3918 ret = -EROFS;
3919 if (btrfs_root_readonly(root))
3920 goto out;
3921
a561be71 3922 ret = mnt_want_write_file(file);
c146afad
YZ		 3923 if (ret)
3924 goto out;
3925
a4abeea4 3926 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3927
1ab86aed 3928 ret = -ENOMEM;
7a7eaa40 3929 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3930 if (IS_ERR(trans))
1ab86aed
SW		 3931 goto out_drop;
3932
3933 file->private_data = trans;
3934 return 0;
3935
3936out_drop:
a4abeea4 3937 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3938 mnt_drop_write_file(file);
f46b5a66 3939out:
f46b5a66
CH		 3940 return ret;
3941}
3942
6ef5ed0d
JB		 3943static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3944{
496ad9aa 3945 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3946 struct btrfs_root *root = BTRFS_I(inode)->root;
3947 struct btrfs_root *new_root;
3948 struct btrfs_dir_item *di;
3949 struct btrfs_trans_handle *trans;
3950 struct btrfs_path *path;
3951 struct btrfs_key location;
3952 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3953 u64 objectid = 0;
3954 u64 dir_id;
3c04ce01 3955 int ret;
6ef5ed0d
JB		 3956
3957 if (!capable(CAP_SYS_ADMIN))
3958 return -EPERM;
3959
3c04ce01
MX		 3960 ret = mnt_want_write_file(file);
3961 if (ret)
3962 return ret;
3963
3964 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3965 ret = -EFAULT;
3966 goto out;
3967 }
6ef5ed0d
JB		 3968
3969 if (!objectid)
1cecf579 3970 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3971
3972 location.objectid = objectid;
3973 location.type = BTRFS_ROOT_ITEM_KEY;
3974 location.offset = (u64)-1;
3975
3976 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3977 if (IS_ERR(new_root)) {
3978 ret = PTR_ERR(new_root);
3979 goto out;
3980 }
6ef5ed0d 3981
6ef5ed0d 3982 path = btrfs_alloc_path();
3c04ce01
MX		 3983 if (!path) {
3984 ret = -ENOMEM;
3985 goto out;
3986 }
6ef5ed0d
JB		 3987 path->leave_spinning = 1;
3988
3989 trans = btrfs_start_transaction(root, 1);
98d5dc13 3990 if (IS_ERR(trans)) {
6ef5ed0d 3991 btrfs_free_path(path);
3c04ce01
MX		 3992 ret = PTR_ERR(trans);
3993 goto out;
6ef5ed0d
JB		 3994 }
3995
6c41761f 3996 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3997 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3998 dir_id, "default", 7, 1);
cf1e99a4 3999 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 4000 btrfs_free_path(path);
4001 btrfs_end_transaction(trans, root);
efe120a0
FH		 4002 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
4003 "item, this isn't going to work");
3c04ce01
MX		 4004 ret = -ENOENT;
4005 goto out;
6ef5ed0d
JB		 4006 }
4007
4008 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4009 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4010 btrfs_mark_buffer_dirty(path->nodes[0]);
4011 btrfs_free_path(path);
4012
2b0ce2c2 4013 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 4014 btrfs_end_transaction(trans, root);
3c04ce01
MX		 4015out:
4016 mnt_drop_write_file(file);
4017 return ret;
6ef5ed0d
JB		 4018}
4019
5af3e8cc
SB		 4020void btrfs_get_block_group_info(struct list_head *groups_list,
4021 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 4022{
4023 struct btrfs_block_group_cache *block_group;
4024
4025 space->total_bytes = 0;
4026 space->used_bytes = 0;
4027 space->flags = 0;
4028 list_for_each_entry(block_group, groups_list, list) {
4029 space->flags = block_group->flags;
4030 space->total_bytes += block_group->key.offset;
4031 space->used_bytes +=
4032 btrfs_block_group_used(&block_group->item);
4033 }
4034}
4035
48a3b636 4036static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 4037{
4038 struct btrfs_ioctl_space_args space_args;
4039 struct btrfs_ioctl_space_info space;
4040 struct btrfs_ioctl_space_info *dest;
7fde62bf 4041 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 4042 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 4043 struct btrfs_space_info *info;
bf5fc093
JB		 4044 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4045 BTRFS_BLOCK_GROUP_SYSTEM,
4046 BTRFS_BLOCK_GROUP_METADATA,
4047 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4048 int num_types = 4;
7fde62bf 4049 int alloc_size;
1406e432 4050 int ret = 0;
51788b1b 4051 u64 slot_count = 0;
bf5fc093 4052 int i, c;
1406e432
JB		 4053
4054 if (copy_from_user(&space_args,
4055 (struct btrfs_ioctl_space_args __user *)arg,
4056 sizeof(space_args)))
4057 return -EFAULT;
4058
bf5fc093
JB		 4059 for (i = 0; i < num_types; i++) {
4060 struct btrfs_space_info *tmp;
4061
4062 info = NULL;
4063 rcu_read_lock();
4064 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4065 list) {
4066 if (tmp->flags == types[i]) {
4067 info = tmp;
4068 break;
4069 }
4070 }
4071 rcu_read_unlock();
4072
4073 if (!info)
4074 continue;
4075
4076 down_read(&info->groups_sem);
4077 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4078 if (!list_empty(&info->block_groups[c]))
4079 slot_count++;
4080 }
4081 up_read(&info->groups_sem);
4082 }
7fde62bf 4083
36523e95
DS		 4084 /*
4085 * Global block reserve, exported as a space_info
4086 */
4087 slot_count++;
4088
7fde62bf
CM		 4089 /* space_slots == 0 means they are asking for a count */
4090 if (space_args.space_slots == 0) {
4091 space_args.total_spaces = slot_count;
4092 goto out;
4093 }
bf5fc093 4094
51788b1b 4095 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 4096
7fde62bf 4097 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 4098
7fde62bf
CM		 4099 /* we generally have at most 6 or so space infos, one for each raid
4100 * level. So, a whole page should be more than enough for everyone
4101 */
4102 if (alloc_size > PAGE_CACHE_SIZE)
4103 return -ENOMEM;
4104
1406e432 4105 space_args.total_spaces = 0;
7fde62bf
CM		 4106 dest = kmalloc(alloc_size, GFP_NOFS);
4107 if (!dest)
4108 return -ENOMEM;
4109 dest_orig = dest;
1406e432 4110
7fde62bf 4111 /* now we have a buffer to copy into */
bf5fc093
JB		 4112 for (i = 0; i < num_types; i++) {
4113 struct btrfs_space_info *tmp;
4114
51788b1b
DR		 4115 if (!slot_count)
4116 break;
4117
bf5fc093
JB		 4118 info = NULL;
4119 rcu_read_lock();
4120 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4121 list) {
4122 if (tmp->flags == types[i]) {
4123 info = tmp;
4124 break;
4125 }
4126 }
4127 rcu_read_unlock();
7fde62bf 4128
bf5fc093
JB		 4129 if (!info)
4130 continue;
4131 down_read(&info->groups_sem);
4132 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4133 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 4134 btrfs_get_block_group_info(
4135 &info->block_groups[c], &space);
bf5fc093
JB		 4136 memcpy(dest, &space, sizeof(space));
4137 dest++;
4138 space_args.total_spaces++;
51788b1b 4139 slot_count--;
bf5fc093 4140 }
51788b1b
DR		 4141 if (!slot_count)
4142 break;
bf5fc093
JB		 4143 }
4144 up_read(&info->groups_sem);
1406e432 4145 }
1406e432 4146
36523e95
DS		 4147 /*
4148 * Add global block reserve
4149 */
4150 if (slot_count) {
4151 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4152
4153 spin_lock(&block_rsv->lock);
4154 space.total_bytes = block_rsv->size;
4155 space.used_bytes = block_rsv->size - block_rsv->reserved;
4156 spin_unlock(&block_rsv->lock);
4157 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4158 memcpy(dest, &space, sizeof(space));
4159 space_args.total_spaces++;
4160 }
4161
2eec6c81 4162 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 4163 (arg + sizeof(struct btrfs_ioctl_space_args));
4164
4165 if (copy_to_user(user_dest, dest_orig, alloc_size))
4166 ret = -EFAULT;
4167
4168 kfree(dest_orig);
4169out:
4170 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 4171 ret = -EFAULT;
4172
4173 return ret;
4174}
4175
f46b5a66
CH		 4176/*
4177 * there are many ways the trans_start and trans_end ioctls can lead
4178 * to deadlocks. They should only be used by applications that
4179 * basically own the machine, and have a very in depth understanding
4180 * of all the possible deadlocks and enospc problems.
4181 */
4182long btrfs_ioctl_trans_end(struct file *file)
4183{
496ad9aa 4184 struct inode *inode = file_inode(file);
f46b5a66
CH		 4185 struct btrfs_root *root = BTRFS_I(inode)->root;
4186 struct btrfs_trans_handle *trans;
f46b5a66 4187
f46b5a66 4188 trans = file->private_data;
1ab86aed
SW		 4189 if (!trans)
4190 return -EINVAL;
b214107e 4191 file->private_data = NULL;
9ca9ee09 4192
1ab86aed
SW		 4193 btrfs_end_transaction(trans, root);
4194
a4abeea4 4195 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 4196
2a79f17e 4197 mnt_drop_write_file(file);
1ab86aed 4198 return 0;
f46b5a66
CH		 4199}
4200
9a8c28be
MX		 4201static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4202 void __user *argp)
46204592 4203{
46204592
SW		 4204 struct btrfs_trans_handle *trans;
4205 u64 transid;
db5b493a 4206 int ret;
46204592 4207
d4edf39b 4208 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4209 if (IS_ERR(trans)) {
4210 if (PTR_ERR(trans) != -ENOENT)
4211 return PTR_ERR(trans);
4212
4213 /* No running transaction, don't bother */
4214 transid = root->fs_info->last_trans_committed;
4215 goto out;
4216 }
46204592 4217 transid = trans->transid;
db5b493a 4218 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 4219 if (ret) {
4220 btrfs_end_transaction(trans, root);
db5b493a 4221 return ret;
8b2b2d3c 4222 }
ff7c1d33 4223out:
46204592
SW		 4224 if (argp)
4225 if (copy_to_user(argp, &transid, sizeof(transid)))
4226 return -EFAULT;
4227 return 0;
4228}
4229
9a8c28be
MX		 4230static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4231 void __user *argp)
46204592 4232{
46204592
SW		 4233 u64 transid;
4234
4235 if (argp) {
4236 if (copy_from_user(&transid, argp, sizeof(transid)))
4237 return -EFAULT;
4238 } else {
4239 transid = 0; /* current trans */
4240 }
4241 return btrfs_wait_for_commit(root, transid);
4242}
4243
b8e95489 4244static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4245{
496ad9aa 4246 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 4247 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4248 int ret;
475f6387
JS		 4249
4250 if (!capable(CAP_SYS_ADMIN))
4251 return -EPERM;
4252
4253 sa = memdup_user(arg, sizeof(*sa));
4254 if (IS_ERR(sa))
4255 return PTR_ERR(sa);
4256
b8e95489
MX		 4257 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4258 ret = mnt_want_write_file(file);
4259 if (ret)
4260 goto out;
4261 }
4262
aa1b8cd4 4263 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4264 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4265 0);
475f6387
JS		 4266
4267 if (copy_to_user(arg, sa, sizeof(*sa)))
4268 ret = -EFAULT;
4269
b8e95489
MX		 4270 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4271 mnt_drop_write_file(file);
4272out:
475f6387
JS		 4273 kfree(sa);
4274 return ret;
4275}
4276
4277static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4278{
4279 if (!capable(CAP_SYS_ADMIN))
4280 return -EPERM;
4281
aa1b8cd4 4282 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 4283}
4284
4285static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4286 void __user *arg)
4287{
4288 struct btrfs_ioctl_scrub_args *sa;
4289 int ret;
4290
4291 if (!capable(CAP_SYS_ADMIN))
4292 return -EPERM;
4293
4294 sa = memdup_user(arg, sizeof(*sa));
4295 if (IS_ERR(sa))
4296 return PTR_ERR(sa);
4297
4298 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4299
4300 if (copy_to_user(arg, sa, sizeof(*sa)))
4301 ret = -EFAULT;
4302
4303 kfree(sa);
4304 return ret;
4305}
4306
c11d2c23 4307static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 4308 void __user *arg)
c11d2c23
SB		 4309{
4310 struct btrfs_ioctl_get_dev_stats *sa;
4311 int ret;
4312
c11d2c23
SB		 4313 sa = memdup_user(arg, sizeof(*sa));
4314 if (IS_ERR(sa))
4315 return PTR_ERR(sa);
4316
b27f7c0c
DS		 4317 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4318 kfree(sa);
4319 return -EPERM;
4320 }
4321
4322 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 4323
4324 if (copy_to_user(arg, sa, sizeof(*sa)))
4325 ret = -EFAULT;
4326
4327 kfree(sa);
4328 return ret;
4329}
4330
3f6bcfbd
SB		 4331static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4332{
4333 struct btrfs_ioctl_dev_replace_args *p;
4334 int ret;
4335
4336 if (!capable(CAP_SYS_ADMIN))
4337 return -EPERM;
4338
4339 p = memdup_user(arg, sizeof(*p));
4340 if (IS_ERR(p))
4341 return PTR_ERR(p);
4342
4343 switch (p->cmd) {
4344 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
adfa97cb
ID		 4345 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4346 ret = -EROFS;
4347 goto out;
4348 }
3f6bcfbd
SB		 4349 if (atomic_xchg(
4350 &root->fs_info->mutually_exclusive_operation_running,
4351 1)) {
e57138b3 4352 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd
SB		 4353 } else {
4354 ret = btrfs_dev_replace_start(root, p);
4355 atomic_set(
4356 &root->fs_info->mutually_exclusive_operation_running,
4357 0);
4358 }
4359 break;
4360 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4361 btrfs_dev_replace_status(root->fs_info, p);
4362 ret = 0;
4363 break;
4364 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4365 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4366 break;
4367 default:
4368 ret = -EINVAL;
4369 break;
4370 }
4371
4372 if (copy_to_user(arg, p, sizeof(*p)))
4373 ret = -EFAULT;
adfa97cb 4374out:
3f6bcfbd
SB		 4375 kfree(p);
4376 return ret;
4377}
4378
d7728c96
JS		 4379static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4380{
4381 int ret = 0;
4382 int i;
740c3d22 4383 u64 rel_ptr;
d7728c96 4384 int size;
806468f8 4385 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4386 struct inode_fs_paths *ipath = NULL;
4387 struct btrfs_path *path;
4388
82b22ac8 4389 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4390 return -EPERM;
4391
4392 path = btrfs_alloc_path();
4393 if (!path) {
4394 ret = -ENOMEM;
4395 goto out;
4396 }
4397
4398 ipa = memdup_user(arg, sizeof(*ipa));
4399 if (IS_ERR(ipa)) {
4400 ret = PTR_ERR(ipa);
4401 ipa = NULL;
4402 goto out;
4403 }
4404
4405 size = min_t(u32, ipa->size, 4096);
4406 ipath = init_ipath(size, root, path);
4407 if (IS_ERR(ipath)) {
4408 ret = PTR_ERR(ipath);
4409 ipath = NULL;
4410 goto out;
4411 }
4412
4413 ret = paths_from_inode(ipa->inum, ipath);
4414 if (ret < 0)
4415 goto out;
4416
4417 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4418 rel_ptr = ipath->fspath->val[i] -
4419 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4420 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4421 }
4422
745c4d8e
JM		 4423 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4424 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 4425 if (ret) {
4426 ret = -EFAULT;
4427 goto out;
4428 }
4429
4430out:
4431 btrfs_free_path(path);
4432 free_ipath(ipath);
4433 kfree(ipa);
4434
4435 return ret;
4436}
4437
4438static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4439{
4440 struct btrfs_data_container *inodes = ctx;
4441 const size_t c = 3 * sizeof(u64);
4442
4443 if (inodes->bytes_left >= c) {
4444 inodes->bytes_left -= c;
4445 inodes->val[inodes->elem_cnt] = inum;
4446 inodes->val[inodes->elem_cnt + 1] = offset;
4447 inodes->val[inodes->elem_cnt + 2] = root;
4448 inodes->elem_cnt += 3;
4449 } else {
4450 inodes->bytes_missing += c - inodes->bytes_left;
4451 inodes->bytes_left = 0;
4452 inodes->elem_missed += 3;
4453 }
4454
4455 return 0;
4456}
4457
4458static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4459 void __user *arg)
4460{
4461 int ret = 0;
4462 int size;
d7728c96
JS		 4463 struct btrfs_ioctl_logical_ino_args *loi;
4464 struct btrfs_data_container *inodes = NULL;
4465 struct btrfs_path *path = NULL;
d7728c96
JS		 4466
4467 if (!capable(CAP_SYS_ADMIN))
4468 return -EPERM;
4469
4470 loi = memdup_user(arg, sizeof(*loi));
4471 if (IS_ERR(loi)) {
4472 ret = PTR_ERR(loi);
4473 loi = NULL;
4474 goto out;
4475 }
4476
4477 path = btrfs_alloc_path();
4478 if (!path) {
4479 ret = -ENOMEM;
4480 goto out;
4481 }
4482
425d17a2 4483 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 4484 inodes = init_data_container(size);
4485 if (IS_ERR(inodes)) {
4486 ret = PTR_ERR(inodes);
4487 inodes = NULL;
4488 goto out;
4489 }
4490
df031f07
LB		 4491 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4492 build_ino_list, inodes);
4493 if (ret == -EINVAL)
d7728c96
JS		 4494 ret = -ENOENT;
4495 if (ret < 0)
4496 goto out;
4497
745c4d8e
JM		 4498 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4499 (void *)(unsigned long)inodes, size);
d7728c96
JS		 4500 if (ret)
4501 ret = -EFAULT;
4502
4503out:
4504 btrfs_free_path(path);
425d17a2 4505 vfree(inodes);
d7728c96
JS		 4506 kfree(loi);
4507
4508 return ret;
4509}
4510
19a39dce 4511void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 4512 struct btrfs_ioctl_balance_args *bargs)
4513{
4514 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4515
4516 bargs->flags = bctl->flags;
4517
837d5b6e
ID		 4518 if (atomic_read(&fs_info->balance_running))
4519 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4520 if (atomic_read(&fs_info->balance_pause_req))
4521 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4522 if (atomic_read(&fs_info->balance_cancel_req))
4523 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4524
c9e9f97b
ID		 4525 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4526 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4527 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 4528
4529 if (lock) {
4530 spin_lock(&fs_info->balance_lock);
4531 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4532 spin_unlock(&fs_info->balance_lock);
4533 } else {
4534 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4535 }
c9e9f97b
ID		 4536}
4537
9ba1f6e4 4538static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4539{
496ad9aa 4540 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4541 struct btrfs_fs_info *fs_info = root->fs_info;
4542 struct btrfs_ioctl_balance_args *bargs;
4543 struct btrfs_balance_control *bctl;
ed0fb78f 4544 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4545 int ret;
4546
4547 if (!capable(CAP_SYS_ADMIN))
4548 return -EPERM;
4549
e54bfa31 4550 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4551 if (ret)
4552 return ret;
4553
ed0fb78f
ID		 4554again:
4555 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4556 mutex_lock(&fs_info->volume_mutex);
4557 mutex_lock(&fs_info->balance_mutex);
4558 need_unlock = true;
4559 goto locked;
4560 }
4561
4562 /*
4563 * mut. excl. ops lock is locked. Three possibilites:
4564 * (1) some other op is running
4565 * (2) balance is running
4566 * (3) balance is paused -- special case (think resume)
4567 */
c9e9f97b 4568 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4569 if (fs_info->balance_ctl) {
4570 /* this is either (2) or (3) */
4571 if (!atomic_read(&fs_info->balance_running)) {
4572 mutex_unlock(&fs_info->balance_mutex);
4573 if (!mutex_trylock(&fs_info->volume_mutex))
4574 goto again;
4575 mutex_lock(&fs_info->balance_mutex);
4576
4577 if (fs_info->balance_ctl &&
4578 !atomic_read(&fs_info->balance_running)) {
4579 /* this is (3) */
4580 need_unlock = false;
4581 goto locked;
4582 }
4583
4584 mutex_unlock(&fs_info->balance_mutex);
4585 mutex_unlock(&fs_info->volume_mutex);
4586 goto again;
4587 } else {
4588 /* this is (2) */
4589 mutex_unlock(&fs_info->balance_mutex);
4590 ret = -EINPROGRESS;
4591 goto out;
4592 }
4593 } else {
4594 /* this is (1) */
4595 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4596 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4597 goto out;
4598 }
4599
4600locked:
4601 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 4602
4603 if (arg) {
4604 bargs = memdup_user(arg, sizeof(*bargs));
4605 if (IS_ERR(bargs)) {
4606 ret = PTR_ERR(bargs);
ed0fb78f 4607 goto out_unlock;
c9e9f97b 4608 }
de322263
ID		 4609
4610 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4611 if (!fs_info->balance_ctl) {
4612 ret = -ENOTCONN;
4613 goto out_bargs;
4614 }
4615
4616 bctl = fs_info->balance_ctl;
4617 spin_lock(&fs_info->balance_lock);
4618 bctl->flags |= BTRFS_BALANCE_RESUME;
4619 spin_unlock(&fs_info->balance_lock);
4620
4621 goto do_balance;
4622 }
c9e9f97b
ID		 4623 } else {
4624 bargs = NULL;
4625 }
4626
ed0fb78f 4627 if (fs_info->balance_ctl) {
837d5b6e
ID		 4628 ret = -EINPROGRESS;
4629 goto out_bargs;
4630 }
4631
c9e9f97b
ID		 4632 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4633 if (!bctl) {
4634 ret = -ENOMEM;
4635 goto out_bargs;
4636 }
4637
4638 bctl->fs_info = fs_info;
4639 if (arg) {
4640 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4641 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4642 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4643
4644 bctl->flags = bargs->flags;
f43ffb60
ID		 4645 } else {
4646 /* balance everything - no filters */
4647 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4648 }
4649
de322263 4650do_balance:
c9e9f97b 4651 /*
ed0fb78f
ID		 4652 * Ownership of bctl and mutually_exclusive_operation_running
4653 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4654 * or, if restriper was paused all the way until unmount, in
4655 * free_fs_info. mutually_exclusive_operation_running is
4656 * cleared in __cancel_balance.
c9e9f97b 4657 */
ed0fb78f
ID		 4658 need_unlock = false;
4659
4660 ret = btrfs_balance(bctl, bargs);
4661
c9e9f97b
ID		 4662 if (arg) {
4663 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4664 ret = -EFAULT;
4665 }
4666
4667out_bargs:
4668 kfree(bargs);
ed0fb78f 4669out_unlock:
c9e9f97b
ID		 4670 mutex_unlock(&fs_info->balance_mutex);
4671 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 4672 if (need_unlock)
4673 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4674out:
e54bfa31 4675 mnt_drop_write_file(file);
c9e9f97b
ID		 4676 return ret;
4677}
4678
837d5b6e
ID		 4679static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4680{
4681 if (!capable(CAP_SYS_ADMIN))
4682 return -EPERM;
4683
4684 switch (cmd) {
4685 case BTRFS_BALANCE_CTL_PAUSE:
4686 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4687 case BTRFS_BALANCE_CTL_CANCEL:
4688 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4689 }
4690
4691 return -EINVAL;
4692}
4693
19a39dce
ID		 4694static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4695 void __user *arg)
4696{
4697 struct btrfs_fs_info *fs_info = root->fs_info;
4698 struct btrfs_ioctl_balance_args *bargs;
4699 int ret = 0;
4700
4701 if (!capable(CAP_SYS_ADMIN))
4702 return -EPERM;
4703
4704 mutex_lock(&fs_info->balance_mutex);
4705 if (!fs_info->balance_ctl) {
4706 ret = -ENOTCONN;
4707 goto out;
4708 }
4709
4710 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4711 if (!bargs) {
4712 ret = -ENOMEM;
4713 goto out;
4714 }
4715
4716 update_ioctl_balance_args(fs_info, 1, bargs);
4717
4718 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4719 ret = -EFAULT;
4720
4721 kfree(bargs);
4722out:
4723 mutex_unlock(&fs_info->balance_mutex);
4724 return ret;
4725}
4726
905b0dda 4727static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4728{
496ad9aa 4729 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4730 struct btrfs_ioctl_quota_ctl_args *sa;
4731 struct btrfs_trans_handle *trans = NULL;
4732 int ret;
4733 int err;
4734
4735 if (!capable(CAP_SYS_ADMIN))
4736 return -EPERM;
4737
905b0dda
MX		 4738 ret = mnt_want_write_file(file);
4739 if (ret)
4740 return ret;
5d13a37b
AJ		 4741
4742 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4743 if (IS_ERR(sa)) {
4744 ret = PTR_ERR(sa);
4745 goto drop_write;
4746 }
5d13a37b 4747
7708f029 4748 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4749 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4750 if (IS_ERR(trans)) {
4751 ret = PTR_ERR(trans);
4752 goto out;
5d13a37b
AJ		 4753 }
4754
4755 switch (sa->cmd) {
4756 case BTRFS_QUOTA_CTL_ENABLE:
4757 ret = btrfs_quota_enable(trans, root->fs_info);
4758 break;
4759 case BTRFS_QUOTA_CTL_DISABLE:
4760 ret = btrfs_quota_disable(trans, root->fs_info);
4761 break;
5d13a37b
AJ		 4762 default:
4763 ret = -EINVAL;
4764 break;
4765 }
4766
2f232036
JS		 4767 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4768 if (err && !ret)
4769 ret = err;
5d13a37b
AJ		 4770out:
4771 kfree(sa);
7708f029 4772 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4773drop_write:
4774 mnt_drop_write_file(file);
5d13a37b
AJ		 4775 return ret;
4776}
4777
905b0dda 4778static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4779{
496ad9aa 4780 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4781 struct btrfs_ioctl_qgroup_assign_args *sa;
4782 struct btrfs_trans_handle *trans;
4783 int ret;
4784 int err;
4785
4786 if (!capable(CAP_SYS_ADMIN))
4787 return -EPERM;
4788
905b0dda
MX		 4789 ret = mnt_want_write_file(file);
4790 if (ret)
4791 return ret;
5d13a37b
AJ		 4792
4793 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4794 if (IS_ERR(sa)) {
4795 ret = PTR_ERR(sa);
4796 goto drop_write;
4797 }
5d13a37b
AJ		 4798
4799 trans = btrfs_join_transaction(root);
4800 if (IS_ERR(trans)) {
4801 ret = PTR_ERR(trans);
4802 goto out;
4803 }
4804
4805 /* FIXME: check if the IDs really exist */
4806 if (sa->assign) {
4807 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4808 sa->src, sa->dst);
4809 } else {
4810 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4811 sa->src, sa->dst);
4812 }
4813
e082f563
QW		 4814 /* update qgroup status and info */
4815 err = btrfs_run_qgroups(trans, root->fs_info);
4816 if (err < 0)
4817 btrfs_error(root->fs_info, ret,
4818 "failed to update qgroup status and info\n");
5d13a37b
AJ		 4819 err = btrfs_end_transaction(trans, root);
4820 if (err && !ret)
4821 ret = err;
4822
4823out:
4824 kfree(sa);
905b0dda
MX		 4825drop_write:
4826 mnt_drop_write_file(file);
5d13a37b
AJ		 4827 return ret;
4828}
4829
905b0dda 4830static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4831{
496ad9aa 4832 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4833 struct btrfs_ioctl_qgroup_create_args *sa;
4834 struct btrfs_trans_handle *trans;
4835 int ret;
4836 int err;
4837
4838 if (!capable(CAP_SYS_ADMIN))
4839 return -EPERM;
4840
905b0dda
MX		 4841 ret = mnt_want_write_file(file);
4842 if (ret)
4843 return ret;
5d13a37b
AJ		 4844
4845 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4846 if (IS_ERR(sa)) {
4847 ret = PTR_ERR(sa);
4848 goto drop_write;
4849 }
5d13a37b 4850
d86e56cf
MX		 4851 if (!sa->qgroupid) {
4852 ret = -EINVAL;
4853 goto out;
4854 }
4855
5d13a37b
AJ		 4856 trans = btrfs_join_transaction(root);
4857 if (IS_ERR(trans)) {
4858 ret = PTR_ERR(trans);
4859 goto out;
4860 }
4861
4862 /* FIXME: check if the IDs really exist */
4863 if (sa->create) {
4087cf24 4864 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
5d13a37b
AJ		 4865 } else {
4866 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4867 }
4868
4869 err = btrfs_end_transaction(trans, root);
4870 if (err && !ret)
4871 ret = err;
4872
4873out:
4874 kfree(sa);
905b0dda
MX		 4875drop_write:
4876 mnt_drop_write_file(file);
5d13a37b
AJ		 4877 return ret;
4878}
4879
905b0dda 4880static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4881{
496ad9aa 4882 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4883 struct btrfs_ioctl_qgroup_limit_args *sa;
4884 struct btrfs_trans_handle *trans;
4885 int ret;
4886 int err;
4887 u64 qgroupid;
4888
4889 if (!capable(CAP_SYS_ADMIN))
4890 return -EPERM;
4891
905b0dda
MX		 4892 ret = mnt_want_write_file(file);
4893 if (ret)
4894 return ret;
5d13a37b
AJ		 4895
4896 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4897 if (IS_ERR(sa)) {
4898 ret = PTR_ERR(sa);
4899 goto drop_write;
4900 }
5d13a37b
AJ		 4901
4902 trans = btrfs_join_transaction(root);
4903 if (IS_ERR(trans)) {
4904 ret = PTR_ERR(trans);
4905 goto out;
4906 }
4907
4908 qgroupid = sa->qgroupid;
4909 if (!qgroupid) {
4910 /* take the current subvol as qgroup */
4911 qgroupid = root->root_key.objectid;
4912 }
4913
4914 /* FIXME: check if the IDs really exist */
4915 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4916
4917 err = btrfs_end_transaction(trans, root);
4918 if (err && !ret)
4919 ret = err;
4920
4921out:
4922 kfree(sa);
905b0dda
MX		 4923drop_write:
4924 mnt_drop_write_file(file);
5d13a37b
AJ		 4925 return ret;
4926}
4927
2f232036
JS		 4928static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4929{
6d0379ec 4930 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4931 struct btrfs_ioctl_quota_rescan_args *qsa;
4932 int ret;
4933
4934 if (!capable(CAP_SYS_ADMIN))
4935 return -EPERM;
4936
4937 ret = mnt_want_write_file(file);
4938 if (ret)
4939 return ret;
4940
4941 qsa = memdup_user(arg, sizeof(*qsa));
4942 if (IS_ERR(qsa)) {
4943 ret = PTR_ERR(qsa);
4944 goto drop_write;
4945 }
4946
4947 if (qsa->flags) {
4948 ret = -EINVAL;
4949 goto out;
4950 }
4951
4952 ret = btrfs_qgroup_rescan(root->fs_info);
4953
4954out:
4955 kfree(qsa);
4956drop_write:
4957 mnt_drop_write_file(file);
4958 return ret;
4959}
4960
4961static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4962{
6d0379ec 4963 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4964 struct btrfs_ioctl_quota_rescan_args *qsa;
4965 int ret = 0;
4966
4967 if (!capable(CAP_SYS_ADMIN))
4968 return -EPERM;
4969
4970 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4971 if (!qsa)
4972 return -ENOMEM;
4973
4974 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4975 qsa->flags = 1;
4976 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4977 }
4978
4979 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4980 ret = -EFAULT;
4981
4982 kfree(qsa);
4983 return ret;
4984}
4985
57254b6e
JS		 4986static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4987{
54563d41 4988 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
57254b6e
JS		 4989
4990 if (!capable(CAP_SYS_ADMIN))
4991 return -EPERM;
4992
4993 return btrfs_qgroup_wait_for_completion(root->fs_info);
4994}
4995
abccd00f
HM		 4996static long _btrfs_ioctl_set_received_subvol(struct file *file,
4997 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4998{
496ad9aa 4999 struct inode *inode = file_inode(file);
8ea05e3a
AB		 5000 struct btrfs_root *root = BTRFS_I(inode)->root;
5001 struct btrfs_root_item *root_item = &root->root_item;
5002 struct btrfs_trans_handle *trans;
5003 struct timespec ct = CURRENT_TIME;
5004 int ret = 0;
dd5f9615 5005 int received_uuid_changed;
8ea05e3a 5006
bd60ea0f
DS		 5007 if (!inode_owner_or_capable(inode))
5008 return -EPERM;
5009
8ea05e3a
AB		 5010 ret = mnt_want_write_file(file);
5011 if (ret < 0)
5012 return ret;
5013
5014 down_write(&root->fs_info->subvol_sem);
5015
5016 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5017 ret = -EINVAL;
5018 goto out;
5019 }
5020
5021 if (btrfs_root_readonly(root)) {
5022 ret = -EROFS;
5023 goto out;
5024 }
5025
dd5f9615
SB		 5026 /*
5027 * 1 - root item
5028 * 2 - uuid items (received uuid + subvol uuid)
5029 */
5030 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 5031 if (IS_ERR(trans)) {
5032 ret = PTR_ERR(trans);
5033 trans = NULL;
5034 goto out;
5035 }
5036
5037 sa->rtransid = trans->transid;
5038 sa->rtime.sec = ct.tv_sec;
5039 sa->rtime.nsec = ct.tv_nsec;
5040
dd5f9615
SB		 5041 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5042 BTRFS_UUID_SIZE);
5043 if (received_uuid_changed &&
5044 !btrfs_is_empty_uuid(root_item->received_uuid))
5045 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5046 root_item->received_uuid,
5047 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5048 root->root_key.objectid);
8ea05e3a
AB		 5049 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5050 btrfs_set_root_stransid(root_item, sa->stransid);
5051 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 5052 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5053 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5054 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5055 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 5056
5057 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5058 &root->root_key, &root->root_item);
5059 if (ret < 0) {
5060 btrfs_end_transaction(trans, root);
8ea05e3a 5061 goto out;
dd5f9615
SB		 5062 }
5063 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5064 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5065 sa->uuid,
5066 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5067 root->root_key.objectid);
5068 if (ret < 0 && ret != -EEXIST) {
5069 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 5070 goto out;
dd5f9615
SB		 5071 }
5072 }
5073 ret = btrfs_commit_transaction(trans, root);
5074 if (ret < 0) {
5075 btrfs_abort_transaction(trans, root, ret);
5076 goto out;
8ea05e3a
AB		 5077 }
5078
abccd00f
HM		 5079out:
5080 up_write(&root->fs_info->subvol_sem);
5081 mnt_drop_write_file(file);
5082 return ret;
5083}
5084
5085#ifdef CONFIG_64BIT
5086static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5087 void __user *arg)
5088{
5089 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5090 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5091 int ret = 0;
5092
5093 args32 = memdup_user(arg, sizeof(*args32));
5094 if (IS_ERR(args32)) {
5095 ret = PTR_ERR(args32);
5096 args32 = NULL;
5097 goto out;
5098 }
5099
5100 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
84dbeb87
DC		 5101 if (!args64) {
5102 ret = -ENOMEM;
abccd00f
HM		 5103 goto out;
5104 }
5105
5106 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5107 args64->stransid = args32->stransid;
5108 args64->rtransid = args32->rtransid;
5109 args64->stime.sec = args32->stime.sec;
5110 args64->stime.nsec = args32->stime.nsec;
5111 args64->rtime.sec = args32->rtime.sec;
5112 args64->rtime.nsec = args32->rtime.nsec;
5113 args64->flags = args32->flags;
5114
5115 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5116 if (ret)
5117 goto out;
5118
5119 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5120 args32->stransid = args64->stransid;
5121 args32->rtransid = args64->rtransid;
5122 args32->stime.sec = args64->stime.sec;
5123 args32->stime.nsec = args64->stime.nsec;
5124 args32->rtime.sec = args64->rtime.sec;
5125 args32->rtime.nsec = args64->rtime.nsec;
5126 args32->flags = args64->flags;
5127
5128 ret = copy_to_user(arg, args32, sizeof(*args32));
5129 if (ret)
5130 ret = -EFAULT;
5131
5132out:
5133 kfree(args32);
5134 kfree(args64);
5135 return ret;
5136}
5137#endif
5138
5139static long btrfs_ioctl_set_received_subvol(struct file *file,
5140 void __user *arg)
5141{
5142 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5143 int ret = 0;
5144
5145 sa = memdup_user(arg, sizeof(*sa));
5146 if (IS_ERR(sa)) {
5147 ret = PTR_ERR(sa);
5148 sa = NULL;
5149 goto out;
5150 }
5151
5152 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5153
5154 if (ret)
5155 goto out;
5156
8ea05e3a
AB		 5157 ret = copy_to_user(arg, sa, sizeof(*sa));
5158 if (ret)
5159 ret = -EFAULT;
5160
5161out:
5162 kfree(sa);
8ea05e3a
AB		 5163 return ret;
5164}
5165
867ab667 5166static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5167{
6d0379ec 5168 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 5169 size_t len;
867ab667 5170 int ret;
a1b83ac5
AJ		 5171 char label[BTRFS_LABEL_SIZE];
5172
5173 spin_lock(&root->fs_info->super_lock);
5174 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5175 spin_unlock(&root->fs_info->super_lock);
5176
5177 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5178
5179 if (len == BTRFS_LABEL_SIZE) {
efe120a0
FH		 5180 btrfs_warn(root->fs_info,
5181 "label is too long, return the first %zu bytes", --len);
867ab667 5182 }
5183
867ab667 5184 ret = copy_to_user(arg, label, len);
867ab667 5185
5186 return ret ? -EFAULT : 0;
5187}
5188
a8bfd4ab 5189static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5190{
6d0379ec 5191 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 5192 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5193 struct btrfs_trans_handle *trans;
5194 char label[BTRFS_LABEL_SIZE];
5195 int ret;
5196
5197 if (!capable(CAP_SYS_ADMIN))
5198 return -EPERM;
5199
5200 if (copy_from_user(label, arg, sizeof(label)))
5201 return -EFAULT;
5202
5203 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
efe120a0 5204 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
a8bfd4ab 5205 BTRFS_LABEL_SIZE - 1);
5206 return -EINVAL;
5207 }
5208
5209 ret = mnt_want_write_file(file);
5210 if (ret)
5211 return ret;
5212
a8bfd4ab 5213 trans = btrfs_start_transaction(root, 0);
5214 if (IS_ERR(trans)) {
5215 ret = PTR_ERR(trans);
5216 goto out_unlock;
5217 }
5218
a1b83ac5 5219 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 5220 strcpy(super_block->label, label);
a1b83ac5 5221 spin_unlock(&root->fs_info->super_lock);
d0270aca 5222 ret = btrfs_commit_transaction(trans, root);
a8bfd4ab 5223
5224out_unlock:
a8bfd4ab 5225 mnt_drop_write_file(file);
5226 return ret;
5227}
5228
2eaa055f
JM		 5229#define INIT_FEATURE_FLAGS(suffix) \
5230 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5231 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5232 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5233
5234static int btrfs_ioctl_get_supported_features(struct file *file,
5235 void __user *arg)
5236{
5237 static struct btrfs_ioctl_feature_flags features[3] = {
5238 INIT_FEATURE_FLAGS(SUPP),
5239 INIT_FEATURE_FLAGS(SAFE_SET),
5240 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5241 };
5242
5243 if (copy_to_user(arg, &features, sizeof(features)))
5244 return -EFAULT;
5245
5246 return 0;
5247}
5248
5249static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5250{
5251 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5252 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5253 struct btrfs_ioctl_feature_flags features;
5254
5255 features.compat_flags = btrfs_super_compat_flags(super_block);
5256 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5257 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5258
5259 if (copy_to_user(arg, &features, sizeof(features)))
5260 return -EFAULT;
5261
5262 return 0;
5263}
5264
3b02a68a
JM		 5265static int check_feature_bits(struct btrfs_root *root,
5266 enum btrfs_feature_set set,
2eaa055f
JM		 5267 u64 change_mask, u64 flags, u64 supported_flags,
5268 u64 safe_set, u64 safe_clear)
5269{
3b02a68a
JM		 5270 const char *type = btrfs_feature_set_names[set];
5271 char *names;
2eaa055f
JM		 5272 u64 disallowed, unsupported;
5273 u64 set_mask = flags & change_mask;
5274 u64 clear_mask = ~flags & change_mask;
5275
5276 unsupported = set_mask & ~supported_flags;
5277 if (unsupported) {
3b02a68a
JM		 5278 names = btrfs_printable_features(set, unsupported);
5279 if (names) {
5280 btrfs_warn(root->fs_info,
5281 "this kernel does not support the %s feature bit%s",
5282 names, strchr(names, ',') ? "s" : "");
5283 kfree(names);
5284 } else
5285 btrfs_warn(root->fs_info,
2eaa055f
JM		 5286 "this kernel does not support %s bits 0x%llx",
5287 type, unsupported);
5288 return -EOPNOTSUPP;
5289 }
5290
5291 disallowed = set_mask & ~safe_set;
5292 if (disallowed) {
3b02a68a
JM		 5293 names = btrfs_printable_features(set, disallowed);
5294 if (names) {
5295 btrfs_warn(root->fs_info,
5296 "can't set the %s feature bit%s while mounted",
5297 names, strchr(names, ',') ? "s" : "");
5298 kfree(names);
5299 } else
5300 btrfs_warn(root->fs_info,
2eaa055f
JM		 5301 "can't set %s bits 0x%llx while mounted",
5302 type, disallowed);
5303 return -EPERM;
5304 }
5305
5306 disallowed = clear_mask & ~safe_clear;
5307 if (disallowed) {
3b02a68a
JM		 5308 names = btrfs_printable_features(set, disallowed);
5309 if (names) {
5310 btrfs_warn(root->fs_info,
5311 "can't clear the %s feature bit%s while mounted",
5312 names, strchr(names, ',') ? "s" : "");
5313 kfree(names);
5314 } else
5315 btrfs_warn(root->fs_info,
2eaa055f
JM		 5316 "can't clear %s bits 0x%llx while mounted",
5317 type, disallowed);
5318 return -EPERM;
5319 }
5320
5321 return 0;
5322}
5323
5324#define check_feature(root, change_mask, flags, mask_base) \
3b02a68a 5325check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5326 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5327 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5328 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5329
5330static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5331{
5332 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5333 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5334 struct btrfs_ioctl_feature_flags flags[2];
5335 struct btrfs_trans_handle *trans;
5336 u64 newflags;
5337 int ret;
5338
5339 if (!capable(CAP_SYS_ADMIN))
5340 return -EPERM;
5341
5342 if (copy_from_user(flags, arg, sizeof(flags)))
5343 return -EFAULT;
5344
5345 /* Nothing to do */
5346 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5347 !flags[0].incompat_flags)
5348 return 0;
5349
5350 ret = check_feature(root, flags[0].compat_flags,
5351 flags[1].compat_flags, COMPAT);
5352 if (ret)
5353 return ret;
5354
5355 ret = check_feature(root, flags[0].compat_ro_flags,
5356 flags[1].compat_ro_flags, COMPAT_RO);
5357 if (ret)
5358 return ret;
5359
5360 ret = check_feature(root, flags[0].incompat_flags,
5361 flags[1].incompat_flags, INCOMPAT);
5362 if (ret)
5363 return ret;
5364
8051aa1a 5365 trans = btrfs_start_transaction(root, 0);
2eaa055f
JM		 5366 if (IS_ERR(trans))
5367 return PTR_ERR(trans);
5368
5369 spin_lock(&root->fs_info->super_lock);
5370 newflags = btrfs_super_compat_flags(super_block);
5371 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5372 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5373 btrfs_set_super_compat_flags(super_block, newflags);
5374
5375 newflags = btrfs_super_compat_ro_flags(super_block);
5376 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5377 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5378 btrfs_set_super_compat_ro_flags(super_block, newflags);
5379
5380 newflags = btrfs_super_incompat_flags(super_block);
5381 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5382 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5383 btrfs_set_super_incompat_flags(super_block, newflags);
5384 spin_unlock(&root->fs_info->super_lock);
5385
d0270aca 5386 return btrfs_commit_transaction(trans, root);
2eaa055f
JM		 5387}
5388
f46b5a66
CH		 5389long btrfs_ioctl(struct file *file, unsigned int
5390 cmd, unsigned long arg)
5391{
496ad9aa 5392 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 5393 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5394
5395 switch (cmd) {
6cbff00f
CH		 5396 case FS_IOC_GETFLAGS:
5397 return btrfs_ioctl_getflags(file, argp);
5398 case FS_IOC_SETFLAGS:
5399 return btrfs_ioctl_setflags(file, argp);
5400 case FS_IOC_GETVERSION:
5401 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5402 case FITRIM:
5403 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5404 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5405 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5406 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5407 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5408 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5409 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5410 case BTRFS_IOC_SUBVOL_CREATE_V2:
5411 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5412 case BTRFS_IOC_SNAP_DESTROY:
5413 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5414 case BTRFS_IOC_SUBVOL_GETFLAGS:
5415 return btrfs_ioctl_subvol_getflags(file, argp);
5416 case BTRFS_IOC_SUBVOL_SETFLAGS:
5417 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5418 case BTRFS_IOC_DEFAULT_SUBVOL:
5419 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5420 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5421 return btrfs_ioctl_defrag(file, NULL);
5422 case BTRFS_IOC_DEFRAG_RANGE:
5423 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5424 case BTRFS_IOC_RESIZE:
198605a8 5425 return btrfs_ioctl_resize(file, argp);
f46b5a66 5426 case BTRFS_IOC_ADD_DEV:
4bcabaa3 5427 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 5428 case BTRFS_IOC_RM_DEV:
da24927b 5429 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 5430 case BTRFS_IOC_FS_INFO:
5431 return btrfs_ioctl_fs_info(root, argp);
5432 case BTRFS_IOC_DEV_INFO:
5433 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 5434 case BTRFS_IOC_BALANCE:
9ba1f6e4 5435 return btrfs_ioctl_balance(file, NULL);
f46b5a66 5436 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 5437 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5438 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 5439 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 5440 case BTRFS_IOC_TRANS_START:
5441 return btrfs_ioctl_trans_start(file);
5442 case BTRFS_IOC_TRANS_END:
5443 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 5444 case BTRFS_IOC_TREE_SEARCH:
5445 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5446 case BTRFS_IOC_TREE_SEARCH_V2:
5447 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5448 case BTRFS_IOC_INO_LOOKUP:
5449 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5450 case BTRFS_IOC_INO_PATHS:
5451 return btrfs_ioctl_ino_to_path(root, argp);
5452 case BTRFS_IOC_LOGICAL_INO:
5453 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 5454 case BTRFS_IOC_SPACE_INFO:
5455 return btrfs_ioctl_space_info(root, argp);
9b199859
FDBM		 5456 case BTRFS_IOC_SYNC: {
5457 int ret;
5458
6c255e67 5459 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
9b199859
FDBM		 5460 if (ret)
5461 return ret;
ddb52f4f 5462 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
2fad4e83
DS		 5463 /*
5464 * The transaction thread may want to do more work,
5465 * namely it pokes the cleaner ktread that will start
5466 * processing uncleaned subvols.
5467 */
5468 wake_up_process(root->fs_info->transaction_kthread);
9b199859
FDBM		 5469 return ret;
5470 }
46204592 5471 case BTRFS_IOC_START_SYNC:
9a8c28be 5472 return btrfs_ioctl_start_sync(root, argp);
46204592 5473 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 5474 return btrfs_ioctl_wait_sync(root, argp);
475f6387 5475 case BTRFS_IOC_SCRUB:
b8e95489 5476 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 5477 case BTRFS_IOC_SCRUB_CANCEL:
5478 return btrfs_ioctl_scrub_cancel(root, argp);
5479 case BTRFS_IOC_SCRUB_PROGRESS:
5480 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 5481 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5482 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 5483 case BTRFS_IOC_BALANCE_CTL:
5484 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 5485 case BTRFS_IOC_BALANCE_PROGRESS:
5486 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 5487 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5488 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5489#ifdef CONFIG_64BIT
5490 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5491 return btrfs_ioctl_set_received_subvol_32(file, argp);
5492#endif
31db9f7c
AB		 5493 case BTRFS_IOC_SEND:
5494 return btrfs_ioctl_send(file, argp);
c11d2c23 5495 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 5496 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 5497 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5498 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5499 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5500 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5501 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5502 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5503 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5504 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5505 case BTRFS_IOC_QUOTA_RESCAN:
5506 return btrfs_ioctl_quota_rescan(file, argp);
5507 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5508 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5509 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5510 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 5511 case BTRFS_IOC_DEV_REPLACE:
5512 return btrfs_ioctl_dev_replace(root, argp);
867ab667 5513 case BTRFS_IOC_GET_FSLABEL:
5514 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5515 case BTRFS_IOC_SET_FSLABEL:
5516 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 5517 case BTRFS_IOC_FILE_EXTENT_SAME:
5518 return btrfs_ioctl_file_extent_same(file, argp);
2eaa055f
JM		 5519 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5520 return btrfs_ioctl_get_supported_features(file, argp);
5521 case BTRFS_IOC_GET_FEATURES:
5522 return btrfs_ioctl_get_features(file, argp);
5523 case BTRFS_IOC_SET_FEATURES:
5524 return btrfs_ioctl_set_features(file, argp);
f46b5a66
CH		 5525 }
5526
5527 return -ENOTTY;
5528}
Linux 6.x block layer and io_uring tree(s)