]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - fs/btrfs/ioctl.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: fix defrag to merge tail file extent
[mirror_ubuntu-hirsute-kernel.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
8ea05e3a 44#include <linux/uuid.h>
55e301fd 45#include <linux/btrfs.h>
416161db 46#include <linux/uaccess.h>
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;
4a3560c4 1033 bool prev_mergeable = true;
940100a4
CM		 1034
1035 /*
008873ea 1036 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1037 * defragging it
1038 */
1039 if (start < *defrag_end)
1040 return 1;
1041
1042 *skip = 0;
1043
6c282eb4
LZ		 1044 em = defrag_lookup_extent(inode, start);
1045 if (!em)
1046 return 0;
940100a4
CM		 1047
1048 /* this will cover holes, and inline extents */
17ce6ef8 1049 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1050 ret = 0;
17ce6ef8
LB		 1051 goto out;
1052 }
1053
4a3560c4
LB		 1054 if (!*defrag_end)
1055 prev_mergeable = false;
1056
6c282eb4 1057 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1058 /*
6c282eb4
LZ		 1059 * we hit a real extent, if it is big or the next extent is not a
1060 * real extent, don't bother defragging it
940100a4 1061 */
a43a2111 1062 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
4a3560c4 1063 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
940100a4 1064 ret = 0;
17ce6ef8 1065out:
940100a4
CM		 1066 /*
1067 * last_len ends up being a counter of how many bytes we've defragged.
1068 * every time we choose not to defrag an extent, we reset *last_len
1069 * so that the next tiny extent will force a defrag.
1070 *
1071 * The end result of this is that tiny extents before a single big
1072 * extent will force at least part of that big extent to be defragged.
1073 */
1074 if (ret) {
940100a4
CM		 1075 *defrag_end = extent_map_end(em);
1076 } else {
1077 *last_len = 0;
1078 *skip = extent_map_end(em);
1079 *defrag_end = 0;
1080 }
1081
1082 free_extent_map(em);
1083 return ret;
1084}
1085
4cb5300b
CM		 1086/*
1087 * it doesn't do much good to defrag one or two pages
1088 * at a time. This pulls in a nice chunk of pages
1089 * to COW and defrag.
1090 *
1091 * It also makes sure the delalloc code has enough
1092 * dirty data to avoid making new small extents as part
1093 * of the defrag
1094 *
1095 * It's a good idea to start RA on this range
1096 * before calling this.
1097 */
1098static int cluster_pages_for_defrag(struct inode *inode,
1099 struct page **pages,
1100 unsigned long start_index,
c41570c9 1101 unsigned long num_pages)
f46b5a66 1102{
4cb5300b
CM		 1103 unsigned long file_end;
1104 u64 isize = i_size_read(inode);
1105 u64 page_start;
1106 u64 page_end;
1f12bd06 1107 u64 page_cnt;
4cb5300b
CM		 1108 int ret;
1109 int i;
1110 int i_done;
3eaa2885 1111 struct btrfs_ordered_extent *ordered;
4cb5300b 1112 struct extent_state *cached_state = NULL;
600a45e1 1113 struct extent_io_tree *tree;
3b16a4e3 1114 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1115
4cb5300b 1116 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1117 if (!isize || start_index > file_end)
1118 return 0;
1119
1120 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1121
1122 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1123 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1124 if (ret)
1125 return ret;
4cb5300b 1126 i_done = 0;
600a45e1 1127 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1128
1129 /* step one, lock all the pages */
1f12bd06 1130 for (i = 0; i < page_cnt; i++) {
4cb5300b 1131 struct page *page;
600a45e1 1132again:
a94733d0 1133 page = find_or_create_page(inode->i_mapping,
600a45e1 1134 start_index + i, mask);
4cb5300b
CM		 1135 if (!page)
1136 break;
1137
600a45e1
MX		 1138 page_start = page_offset(page);
1139 page_end = page_start + PAGE_CACHE_SIZE - 1;
1140 while (1) {
308d9800
FM		 1141 lock_extent_bits(tree, page_start, page_end,
1142 0, &cached_state);
600a45e1
MX		 1143 ordered = btrfs_lookup_ordered_extent(inode,
1144 page_start);
308d9800
FM		 1145 unlock_extent_cached(tree, page_start, page_end,
1146 &cached_state, GFP_NOFS);
600a45e1
MX		 1147 if (!ordered)
1148 break;
1149
1150 unlock_page(page);
1151 btrfs_start_ordered_extent(inode, ordered, 1);
1152 btrfs_put_ordered_extent(ordered);
1153 lock_page(page);
1f12bd06
LB		 1154 /*
1155 * we unlocked the page above, so we need check if
1156 * it was released or not.
1157 */
1158 if (page->mapping != inode->i_mapping) {
1159 unlock_page(page);
1160 page_cache_release(page);
1161 goto again;
1162 }
600a45e1
MX		 1163 }
1164
4cb5300b
CM		 1165 if (!PageUptodate(page)) {
1166 btrfs_readpage(NULL, page);
1167 lock_page(page);
1168 if (!PageUptodate(page)) {
1169 unlock_page(page);
1170 page_cache_release(page);
1171 ret = -EIO;
1172 break;
1173 }
1174 }
600a45e1 1175
600a45e1
MX		 1176 if (page->mapping != inode->i_mapping) {
1177 unlock_page(page);
1178 page_cache_release(page);
1179 goto again;
1180 }
1181
4cb5300b
CM		 1182 pages[i] = page;
1183 i_done++;
1184 }
1185 if (!i_done || ret)
1186 goto out;
1187
1188 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1189 goto out;
1190
1191 /*
1192 * so now we have a nice long stream of locked
1193 * and up to date pages, lets wait on them
1194 */
1195 for (i = 0; i < i_done; i++)
1196 wait_on_page_writeback(pages[i]);
1197
1198 page_start = page_offset(pages[0]);
1199 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1200
1201 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1202 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1203 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1204 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1205 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1206 &cached_state, GFP_NOFS);
4cb5300b 1207
1f12bd06 1208 if (i_done != page_cnt) {
9e0baf60
JB		 1209 spin_lock(&BTRFS_I(inode)->lock);
1210 BTRFS_I(inode)->outstanding_extents++;
1211 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1212 btrfs_delalloc_release_space(inode,
1f12bd06 1213 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1214 }
1215
1216
9e8a4a8b
LB		 1217 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1218 &cached_state, GFP_NOFS);
4cb5300b
CM		 1219
1220 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1221 page_start, page_end - 1, &cached_state,
1222 GFP_NOFS);
1223
1224 for (i = 0; i < i_done; i++) {
1225 clear_page_dirty_for_io(pages[i]);
1226 ClearPageChecked(pages[i]);
1227 set_page_extent_mapped(pages[i]);
1228 set_page_dirty(pages[i]);
1229 unlock_page(pages[i]);
1230 page_cache_release(pages[i]);
1231 }
1232 return i_done;
1233out:
1234 for (i = 0; i < i_done; i++) {
1235 unlock_page(pages[i]);
1236 page_cache_release(pages[i]);
1237 }
1f12bd06 1238 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1239 return ret;
1240
1241}
1242
1243int btrfs_defrag_file(struct inode *inode, struct file *file,
1244 struct btrfs_ioctl_defrag_range_args *range,
1245 u64 newer_than, unsigned long max_to_defrag)
1246{
1247 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1248 struct file_ra_state *ra = NULL;
f46b5a66 1249 unsigned long last_index;
151a31b2 1250 u64 isize = i_size_read(inode);
940100a4
CM		 1251 u64 last_len = 0;
1252 u64 skip = 0;
1253 u64 defrag_end = 0;
4cb5300b 1254 u64 newer_off = range->start;
f46b5a66 1255 unsigned long i;
008873ea 1256 unsigned long ra_index = 0;
f46b5a66 1257 int ret;
4cb5300b 1258 int defrag_count = 0;
1a419d85 1259 int compress_type = BTRFS_COMPRESS_ZLIB;
aab110ab 1260 u32 extent_thresh = range->extent_thresh;
c41570c9
JM		 1261 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1262 unsigned long cluster = max_cluster;
4cb5300b
CM		 1263 u64 new_align = ~((u64)128 * 1024 - 1);
1264 struct page **pages = NULL;
1265
0abd5b17
LB		 1266 if (isize == 0)
1267 return 0;
1268
1269 if (range->start >= isize)
1270 return -EINVAL;
1a419d85
LZ		 1271
1272 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1273 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1274 return -EINVAL;
1275 if (range->compress_type)
1276 compress_type = range->compress_type;
1277 }
f46b5a66 1278
0abd5b17
LB		 1279 if (extent_thresh == 0)
1280 extent_thresh = 256 * 1024;
940100a4 1281
4cb5300b
CM		 1282 /*
1283 * if we were not given a file, allocate a readahead
1284 * context
1285 */
1286 if (!file) {
1287 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1288 if (!ra)
1289 return -ENOMEM;
1290 file_ra_state_init(ra, inode->i_mapping);
1291 } else {
1292 ra = &file->f_ra;
1293 }
1294
d9b0d9ba 1295 pages = kmalloc_array(max_cluster, sizeof(struct page *),
4cb5300b
CM		 1296 GFP_NOFS);
1297 if (!pages) {
1298 ret = -ENOMEM;
1299 goto out_ra;
1300 }
1301
1302 /* find the last page to defrag */
1e701a32 1303 if (range->start + range->len > range->start) {
151a31b2 1304 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1305 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1306 } else {
151a31b2 1307 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1308 }
1309
4cb5300b
CM		 1310 if (newer_than) {
1311 ret = find_new_extents(root, inode, newer_than,
1312 &newer_off, 64 * 1024);
1313 if (!ret) {
1314 range->start = newer_off;
1315 /*
1316 * we always align our defrag to help keep
1317 * the extents in the file evenly spaced
1318 */
1319 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1320 } else
1321 goto out_ra;
1322 } else {
1323 i = range->start >> PAGE_CACHE_SHIFT;
1324 }
1325 if (!max_to_defrag)
070034bd 1326 max_to_defrag = last_index - i + 1;
4cb5300b 1327
2a0f7f57
LZ		 1328 /*
1329 * make writeback starts from i, so the defrag range can be
1330 * written sequentially.
1331 */
1332 if (i < inode->i_mapping->writeback_index)
1333 inode->i_mapping->writeback_index = i;
1334
f7f43cc8 1335 while (i <= last_index && defrag_count < max_to_defrag &&
ed6078f7 1336 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
4cb5300b
CM		 1337 /*
1338 * make sure we stop running if someone unmounts
1339 * the FS
1340 */
1341 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1342 break;
1343
210549eb 1344 if (btrfs_defrag_cancelled(root->fs_info)) {
efe120a0 1345 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
210549eb
DS		 1346 ret = -EAGAIN;
1347 break;
1348 }
1349
66c26892 1350 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1351 extent_thresh, &last_len, &skip,
a43a2111
AM		 1352 &defrag_end, range->flags &
1353 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1354 unsigned long next;
1355 /*
1356 * the should_defrag function tells us how much to skip
1357 * bump our counter by the suggested amount
1358 */
ed6078f7 1359 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
940100a4
CM		 1360 i = max(i + 1, next);
1361 continue;
1362 }
008873ea
LZ		 1363
1364 if (!newer_than) {
1365 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1366 PAGE_CACHE_SHIFT) - i;
1367 cluster = min(cluster, max_cluster);
1368 } else {
1369 cluster = max_cluster;
1370 }
1371
008873ea
LZ		 1372 if (i + cluster > ra_index) {
1373 ra_index = max(i, ra_index);
1374 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1375 cluster);
e4826a5b 1376 ra_index += cluster;
008873ea 1377 }
940100a4 1378
ecb8bea8 1379 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1380 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1381 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1382 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1383 if (ret < 0) {
1384 mutex_unlock(&inode->i_mutex);
4cb5300b 1385 goto out_ra;
ecb8bea8 1386 }
4cb5300b
CM		 1387
1388 defrag_count += ret;
d0e1d66b 1389 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1390 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1391
1392 if (newer_than) {
1393 if (newer_off == (u64)-1)
1394 break;
1395
e1f041e1
LB		 1396 if (ret > 0)
1397 i += ret;
1398
4cb5300b
CM		 1399 newer_off = max(newer_off + 1,
1400 (u64)i << PAGE_CACHE_SHIFT);
1401
1402 ret = find_new_extents(root, inode,
1403 newer_than, &newer_off,
1404 64 * 1024);
1405 if (!ret) {
1406 range->start = newer_off;
1407 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1408 } else {
1409 break;
f46b5a66 1410 }
4cb5300b 1411 } else {
008873ea 1412 if (ret > 0) {
cbcc8326 1413 i += ret;
008873ea
LZ		 1414 last_len += ret << PAGE_CACHE_SHIFT;
1415 } else {
cbcc8326 1416 i++;
008873ea
LZ		 1417 last_len = 0;
1418 }
f46b5a66 1419 }
f46b5a66
CH		 1420 }
1421
dec8ef90 1422 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1423 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1424 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1425 &BTRFS_I(inode)->runtime_flags))
1426 filemap_flush(inode->i_mapping);
1427 }
1e701a32
CM		 1428
1429 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1430 /* the filemap_flush will queue IO into the worker threads, but
1431 * we have to make sure the IO is actually started and that
1432 * ordered extents get created before we return
1433 */
1434 atomic_inc(&root->fs_info->async_submit_draining);
1435 while (atomic_read(&root->fs_info->nr_async_submits) ||
1436 atomic_read(&root->fs_info->async_delalloc_pages)) {
1437 wait_event(root->fs_info->async_submit_wait,
1438 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1439 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1440 }
1441 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1442 }
1443
1a419d85 1444 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1445 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1446 }
1447
60ccf82f 1448 ret = defrag_count;
940100a4 1449
4cb5300b 1450out_ra:
633085c7
FDBM		 1451 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1452 mutex_lock(&inode->i_mutex);
1453 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1454 mutex_unlock(&inode->i_mutex);
1455 }
4cb5300b
CM		 1456 if (!file)
1457 kfree(ra);
1458 kfree(pages);
940100a4 1459 return ret;
f46b5a66
CH		 1460}
1461
198605a8 1462static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1463 void __user *arg)
f46b5a66
CH		 1464{
1465 u64 new_size;
1466 u64 old_size;
1467 u64 devid = 1;
496ad9aa 1468 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1469 struct btrfs_ioctl_vol_args *vol_args;
1470 struct btrfs_trans_handle *trans;
1471 struct btrfs_device *device = NULL;
1472 char *sizestr;
9a40f122 1473 char *retptr;
f46b5a66
CH		 1474 char *devstr = NULL;
1475 int ret = 0;
f46b5a66
CH		 1476 int mod = 0;
1477
e441d54d
CM		 1478 if (!capable(CAP_SYS_ADMIN))
1479 return -EPERM;
1480
198605a8
MX		 1481 ret = mnt_want_write_file(file);
1482 if (ret)
1483 return ret;
1484
5ac00add
SB		 1485 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1486 1)) {
97547676 1487 mnt_drop_write_file(file);
e57138b3 1488 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1489 }
1490
5ac00add 1491 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1492 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1493 if (IS_ERR(vol_args)) {
1494 ret = PTR_ERR(vol_args);
1495 goto out;
1496 }
5516e595
MF		 1497
1498 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1499
f46b5a66
CH		 1500 sizestr = vol_args->name;
1501 devstr = strchr(sizestr, ':');
1502 if (devstr) {
f46b5a66
CH		 1503 sizestr = devstr + 1;
1504 *devstr = '\0';
1505 devstr = vol_args->name;
58dfae63
Z		 1506 ret = kstrtoull(devstr, 10, &devid);
1507 if (ret)
1508 goto out_free;
dfd79829
MX		 1509 if (!devid) {
1510 ret = -EINVAL;
1511 goto out_free;
1512 }
efe120a0 1513 btrfs_info(root->fs_info, "resizing devid %llu", devid);
f46b5a66 1514 }
dba60f3f 1515
aa1b8cd4 1516 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1517 if (!device) {
efe120a0 1518 btrfs_info(root->fs_info, "resizer unable to find device %llu",
c1c9ff7c 1519 devid);
dfd79829 1520 ret = -ENODEV;
c9e9f97b 1521 goto out_free;
f46b5a66 1522 }
dba60f3f
MX		 1523
1524 if (!device->writeable) {
efe120a0
FH		 1525 btrfs_info(root->fs_info,
1526 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1527 devid);
dfd79829 1528 ret = -EPERM;
4e42ae1b
LB		 1529 goto out_free;
1530 }
1531
f46b5a66
CH		 1532 if (!strcmp(sizestr, "max"))
1533 new_size = device->bdev->bd_inode->i_size;
1534 else {
1535 if (sizestr[0] == '-') {
1536 mod = -1;
1537 sizestr++;
1538 } else if (sizestr[0] == '+') {
1539 mod = 1;
1540 sizestr++;
1541 }
9a40f122
GH		 1542 new_size = memparse(sizestr, &retptr);
1543 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1544 ret = -EINVAL;
c9e9f97b 1545 goto out_free;
f46b5a66
CH		 1546 }
1547 }
1548
63a212ab 1549 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1550 ret = -EPERM;
63a212ab
SB		 1551 goto out_free;
1552 }
1553
7cc8e58d 1554 old_size = btrfs_device_get_total_bytes(device);
f46b5a66
CH		 1555
1556 if (mod < 0) {
1557 if (new_size > old_size) {
1558 ret = -EINVAL;
c9e9f97b 1559 goto out_free;
f46b5a66
CH		 1560 }
1561 new_size = old_size - new_size;
1562 } else if (mod > 0) {
eb8052e0 1563 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1564 ret = -ERANGE;
eb8052e0
WF		 1565 goto out_free;
1566 }
f46b5a66
CH		 1567 new_size = old_size + new_size;
1568 }
1569
1570 if (new_size < 256 * 1024 * 1024) {
1571 ret = -EINVAL;
c9e9f97b 1572 goto out_free;
f46b5a66
CH		 1573 }
1574 if (new_size > device->bdev->bd_inode->i_size) {
1575 ret = -EFBIG;
c9e9f97b 1576 goto out_free;
f46b5a66
CH		 1577 }
1578
b8b93add 1579 new_size = div_u64(new_size, root->sectorsize);
f46b5a66
CH		 1580 new_size *= root->sectorsize;
1581
efe120a0 1582 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
c1c9ff7c 1583 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1584
1585 if (new_size > old_size) {
a22285a6 1586 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1587 if (IS_ERR(trans)) {
1588 ret = PTR_ERR(trans);
c9e9f97b 1589 goto out_free;
98d5dc13 1590 }
f46b5a66
CH		 1591 ret = btrfs_grow_device(trans, device, new_size);
1592 btrfs_commit_transaction(trans, root);
ece7d20e 1593 } else if (new_size < old_size) {
f46b5a66 1594 ret = btrfs_shrink_device(device, new_size);
0253f40e 1595 } /* equal, nothing need to do */
f46b5a66 1596
c9e9f97b 1597out_free:
f46b5a66 1598 kfree(vol_args);
c9e9f97b
ID		 1599out:
1600 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1601 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1602 mnt_drop_write_file(file);
f46b5a66
CH		 1603 return ret;
1604}
1605
72fd032e 1606static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1607 char *name, unsigned long fd, int subvol,
1608 u64 *transid, bool readonly,
8696c533 1609 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1610{
f46b5a66 1611 int namelen;
3de4586c 1612 int ret = 0;
f46b5a66 1613
a874a63e
LB		 1614 ret = mnt_want_write_file(file);
1615 if (ret)
1616 goto out;
1617
72fd032e
SW		 1618 namelen = strlen(name);
1619 if (strchr(name, '/')) {
f46b5a66 1620 ret = -EINVAL;
a874a63e 1621 goto out_drop_write;
f46b5a66
CH		 1622 }
1623
16780cab
CM		 1624 if (name[0] == '.' &&
1625 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1626 ret = -EEXIST;
a874a63e 1627 goto out_drop_write;
16780cab
CM		 1628 }
1629
3de4586c 1630 if (subvol) {
72fd032e 1631 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1632 NULL, transid, readonly, inherit);
cb8e7090 1633 } else {
2903ff01 1634 struct fd src = fdget(fd);
3de4586c 1635 struct inode *src_inode;
2903ff01 1636 if (!src.file) {
3de4586c 1637 ret = -EINVAL;
a874a63e 1638 goto out_drop_write;
3de4586c
CM		 1639 }
1640
496ad9aa
AV		 1641 src_inode = file_inode(src.file);
1642 if (src_inode->i_sb != file_inode(file)->i_sb) {
efe120a0
FH		 1643 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1644 "Snapshot src from another FS");
23ad5b17 1645 ret = -EXDEV;
d0242061
DS		 1646 } else if (!inode_owner_or_capable(src_inode)) {
1647 /*
1648 * Subvolume creation is not restricted, but snapshots
1649 * are limited to own subvolumes only
1650 */
1651 ret = -EPERM;
ecd18815
AV		 1652 } else {
1653 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1654 BTRFS_I(src_inode)->root,
1655 transid, readonly, inherit);
3de4586c 1656 }
2903ff01 1657 fdput(src);
cb8e7090 1658 }
a874a63e
LB		 1659out_drop_write:
1660 mnt_drop_write_file(file);
f46b5a66 1661out:
72fd032e
SW		 1662 return ret;
1663}
1664
1665static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1666 void __user *arg, int subvol)
72fd032e 1667{
fa0d2b9b 1668 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1669 int ret;
1670
fa0d2b9b
LZ		 1671 vol_args = memdup_user(arg, sizeof(*vol_args));
1672 if (IS_ERR(vol_args))
1673 return PTR_ERR(vol_args);
1674 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1675
fa0d2b9b 1676 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1677 vol_args->fd, subvol,
6f72c7e2 1678 NULL, false, NULL);
fdfb1e4f 1679
fa0d2b9b
LZ		 1680 kfree(vol_args);
1681 return ret;
1682}
fdfb1e4f 1683
fa0d2b9b
LZ		 1684static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1685 void __user *arg, int subvol)
1686{
1687 struct btrfs_ioctl_vol_args_v2 *vol_args;
1688 int ret;
1689 u64 transid = 0;
1690 u64 *ptr = NULL;
b83cc969 1691 bool readonly = false;
6f72c7e2 1692 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1693
fa0d2b9b
LZ		 1694 vol_args = memdup_user(arg, sizeof(*vol_args));
1695 if (IS_ERR(vol_args))
1696 return PTR_ERR(vol_args);
1697 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1698
b83cc969 1699 if (vol_args->flags &
6f72c7e2
AJ		 1700 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1701 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1702 ret = -EOPNOTSUPP;
c47ca32d 1703 goto free_args;
72fd032e 1704 }
fa0d2b9b
LZ		 1705
1706 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1707 ptr = &transid;
b83cc969
LZ		 1708 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1709 readonly = true;
6f72c7e2
AJ		 1710 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1711 if (vol_args->size > PAGE_CACHE_SIZE) {
1712 ret = -EINVAL;
c47ca32d 1713 goto free_args;
6f72c7e2
AJ		 1714 }
1715 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1716 if (IS_ERR(inherit)) {
1717 ret = PTR_ERR(inherit);
c47ca32d 1718 goto free_args;
6f72c7e2
AJ		 1719 }
1720 }
fa0d2b9b
LZ		 1721
1722 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1723 vol_args->fd, subvol, ptr,
8696c533 1724 readonly, inherit);
c47ca32d
DC		 1725 if (ret)
1726 goto free_inherit;
fa0d2b9b 1727
c47ca32d
DC		 1728 if (ptr && copy_to_user(arg +
1729 offsetof(struct btrfs_ioctl_vol_args_v2,
1730 transid),
1731 ptr, sizeof(*ptr)))
fa0d2b9b 1732 ret = -EFAULT;
c47ca32d
DC		 1733
1734free_inherit:
6f72c7e2 1735 kfree(inherit);
c47ca32d
DC		 1736free_args:
1737 kfree(vol_args);
f46b5a66
CH		 1738 return ret;
1739}
1740
0caa102d
LZ		 1741static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1742 void __user *arg)
1743{
496ad9aa 1744 struct inode *inode = file_inode(file);
0caa102d
LZ		 1745 struct btrfs_root *root = BTRFS_I(inode)->root;
1746 int ret = 0;
1747 u64 flags = 0;
1748
33345d01 1749 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1750 return -EINVAL;
1751
1752 down_read(&root->fs_info->subvol_sem);
1753 if (btrfs_root_readonly(root))
1754 flags |= BTRFS_SUBVOL_RDONLY;
1755 up_read(&root->fs_info->subvol_sem);
1756
1757 if (copy_to_user(arg, &flags, sizeof(flags)))
1758 ret = -EFAULT;
1759
1760 return ret;
1761}
1762
1763static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1764 void __user *arg)
1765{
496ad9aa 1766 struct inode *inode = file_inode(file);
0caa102d
LZ		 1767 struct btrfs_root *root = BTRFS_I(inode)->root;
1768 struct btrfs_trans_handle *trans;
1769 u64 root_flags;
1770 u64 flags;
1771 int ret = 0;
1772
bd60ea0f
DS		 1773 if (!inode_owner_or_capable(inode))
1774 return -EPERM;
1775
b9ca0664
LB		 1776 ret = mnt_want_write_file(file);
1777 if (ret)
1778 goto out;
0caa102d 1779
b9ca0664
LB		 1780 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1781 ret = -EINVAL;
1782 goto out_drop_write;
1783 }
0caa102d 1784
b9ca0664
LB		 1785 if (copy_from_user(&flags, arg, sizeof(flags))) {
1786 ret = -EFAULT;
1787 goto out_drop_write;
1788 }
0caa102d 1789
b9ca0664
LB		 1790 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1791 ret = -EINVAL;
1792 goto out_drop_write;
1793 }
0caa102d 1794
b9ca0664
LB		 1795 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1796 ret = -EOPNOTSUPP;
1797 goto out_drop_write;
1798 }
0caa102d
LZ		 1799
1800 down_write(&root->fs_info->subvol_sem);
1801
1802 /* nothing to do */
1803 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1804 goto out_drop_sem;
0caa102d
LZ		 1805
1806 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1807 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1808 btrfs_set_root_flags(&root->root_item,
1809 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1810 } else {
1811 /*
1812 * Block RO -> RW transition if this subvolume is involved in
1813 * send
1814 */
1815 spin_lock(&root->root_item_lock);
1816 if (root->send_in_progress == 0) {
1817 btrfs_set_root_flags(&root->root_item,
0caa102d 1818 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1819 spin_unlock(&root->root_item_lock);
1820 } else {
1821 spin_unlock(&root->root_item_lock);
1822 btrfs_warn(root->fs_info,
1823 "Attempt to set subvolume %llu read-write during send",
1824 root->root_key.objectid);
1825 ret = -EPERM;
1826 goto out_drop_sem;
1827 }
1828 }
0caa102d
LZ		 1829
1830 trans = btrfs_start_transaction(root, 1);
1831 if (IS_ERR(trans)) {
1832 ret = PTR_ERR(trans);
1833 goto out_reset;
1834 }
1835
b4dc2b8c 1836 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1837 &root->root_key, &root->root_item);
1838
1839 btrfs_commit_transaction(trans, root);
1840out_reset:
1841 if (ret)
1842 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1843out_drop_sem:
0caa102d 1844 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1845out_drop_write:
1846 mnt_drop_write_file(file);
1847out:
0caa102d
LZ		 1848 return ret;
1849}
1850
76dda93c
YZ		 1851/*
1852 * helper to check if the subvolume references other subvolumes
1853 */
1854static noinline int may_destroy_subvol(struct btrfs_root *root)
1855{
1856 struct btrfs_path *path;
175a2b87 1857 struct btrfs_dir_item *di;
76dda93c 1858 struct btrfs_key key;
175a2b87 1859 u64 dir_id;
76dda93c
YZ		 1860 int ret;
1861
1862 path = btrfs_alloc_path();
1863 if (!path)
1864 return -ENOMEM;
1865
175a2b87
JB		 1866 /* Make sure this root isn't set as the default subvol */
1867 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1868 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1869 dir_id, "default", 7, 0);
1870 if (di && !IS_ERR(di)) {
1871 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1872 if (key.objectid == root->root_key.objectid) {
72de6b53
GS		 1873 ret = -EPERM;
1874 btrfs_err(root->fs_info, "deleting default subvolume "
1875 "%llu is not allowed", key.objectid);
175a2b87
JB		 1876 goto out;
1877 }
1878 btrfs_release_path(path);
1879 }
1880
76dda93c
YZ		 1881 key.objectid = root->root_key.objectid;
1882 key.type = BTRFS_ROOT_REF_KEY;
1883 key.offset = (u64)-1;
1884
1885 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1886 &key, path, 0, 0);
1887 if (ret < 0)
1888 goto out;
1889 BUG_ON(ret == 0);
1890
1891 ret = 0;
1892 if (path->slots[0] > 0) {
1893 path->slots[0]--;
1894 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1895 if (key.objectid == root->root_key.objectid &&
1896 key.type == BTRFS_ROOT_REF_KEY)
1897 ret = -ENOTEMPTY;
1898 }
1899out:
1900 btrfs_free_path(path);
1901 return ret;
1902}
1903
ac8e9819
CM		 1904static noinline int key_in_sk(struct btrfs_key *key,
1905 struct btrfs_ioctl_search_key *sk)
1906{
abc6e134
CM		 1907 struct btrfs_key test;
1908 int ret;
1909
1910 test.objectid = sk->min_objectid;
1911 test.type = sk->min_type;
1912 test.offset = sk->min_offset;
1913
1914 ret = btrfs_comp_cpu_keys(key, &test);
1915 if (ret < 0)
ac8e9819 1916 return 0;
abc6e134
CM		 1917
1918 test.objectid = sk->max_objectid;
1919 test.type = sk->max_type;
1920 test.offset = sk->max_offset;
1921
1922 ret = btrfs_comp_cpu_keys(key, &test);
1923 if (ret > 0)
ac8e9819
CM		 1924 return 0;
1925 return 1;
1926}
1927
1928static noinline int copy_to_sk(struct btrfs_root *root,
1929 struct btrfs_path *path,
1930 struct btrfs_key *key,
1931 struct btrfs_ioctl_search_key *sk,
9b6e817d 1932 size_t *buf_size,
ba346b35 1933 char __user *ubuf,
ac8e9819
CM		 1934 unsigned long *sk_offset,
1935 int *num_found)
1936{
1937 u64 found_transid;
1938 struct extent_buffer *leaf;
1939 struct btrfs_ioctl_search_header sh;
dd81d459 1940 struct btrfs_key test;
ac8e9819
CM		 1941 unsigned long item_off;
1942 unsigned long item_len;
1943 int nritems;
1944 int i;
1945 int slot;
ac8e9819
CM		 1946 int ret = 0;
1947
1948 leaf = path->nodes[0];
1949 slot = path->slots[0];
1950 nritems = btrfs_header_nritems(leaf);
1951
1952 if (btrfs_header_generation(leaf) > sk->max_transid) {
1953 i = nritems;
1954 goto advance_key;
1955 }
1956 found_transid = btrfs_header_generation(leaf);
1957
1958 for (i = slot; i < nritems; i++) {
1959 item_off = btrfs_item_ptr_offset(leaf, i);
1960 item_len = btrfs_item_size_nr(leaf, i);
1961
03b71c6c
GP		 1962 btrfs_item_key_to_cpu(leaf, key, i);
1963 if (!key_in_sk(key, sk))
1964 continue;
1965
9b6e817d 1966 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 1967 if (*num_found) {
1968 ret = 1;
1969 goto out;
1970 }
1971
1972 /*
1973 * return one empty item back for v1, which does not
1974 * handle -EOVERFLOW
1975 */
1976
9b6e817d 1977 *buf_size = sizeof(sh) + item_len;
ac8e9819 1978 item_len = 0;
8f5f6178
GH		 1979 ret = -EOVERFLOW;
1980 }
ac8e9819 1981
9b6e817d 1982 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 1983 ret = 1;
25c9bc2e 1984 goto out;
ac8e9819
CM		 1985 }
1986
ac8e9819
CM		 1987 sh.objectid = key->objectid;
1988 sh.offset = key->offset;
1989 sh.type = key->type;
1990 sh.len = item_len;
1991 sh.transid = found_transid;
1992
1993 /* copy search result header */
ba346b35
GH		 1994 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1995 ret = -EFAULT;
1996 goto out;
1997 }
1998
ac8e9819
CM		 1999 *sk_offset += sizeof(sh);
2000
2001 if (item_len) {
ba346b35 2002 char __user *up = ubuf + *sk_offset;
ac8e9819 2003 /* copy the item */
ba346b35
GH		 2004 if (read_extent_buffer_to_user(leaf, up,
2005 item_off, item_len)) {
2006 ret = -EFAULT;
2007 goto out;
2008 }
2009
ac8e9819 2010 *sk_offset += item_len;
ac8e9819 2011 }
e2156867 2012 (*num_found)++;
ac8e9819 2013
8f5f6178
GH		 2014 if (ret) /* -EOVERFLOW from above */
2015 goto out;
2016
25c9bc2e
GH		 2017 if (*num_found >= sk->nr_items) {
2018 ret = 1;
2019 goto out;
2020 }
ac8e9819
CM		 2021 }
2022advance_key:
abc6e134 2023 ret = 0;
dd81d459
NA		 2024 test.objectid = sk->max_objectid;
2025 test.type = sk->max_type;
2026 test.offset = sk->max_offset;
2027 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2028 ret = 1;
2029 else if (key->offset < (u64)-1)
ac8e9819 2030 key->offset++;
dd81d459 2031 else if (key->type < (u8)-1) {
abc6e134 2032 key->offset = 0;
ac8e9819 2033 key->type++;
dd81d459 2034 } else if (key->objectid < (u64)-1) {
abc6e134
CM		 2035 key->offset = 0;
2036 key->type = 0;
ac8e9819 2037 key->objectid++;
abc6e134
CM		 2038 } else
2039 ret = 1;
25c9bc2e 2040out:
ba346b35
GH		 2041 /*
2042 * 0: all items from this leaf copied, continue with next
2043 * 1: * more items can be copied, but unused buffer is too small
2044 * * all items were found
2045 * Either way, it will stops the loop which iterates to the next
2046 * leaf
2047 * -EOVERFLOW: item was to large for buffer
2048 * -EFAULT: could not copy extent buffer back to userspace
2049 */
ac8e9819
CM		 2050 return ret;
2051}
2052
2053static noinline int search_ioctl(struct inode *inode,
12544442 2054 struct btrfs_ioctl_search_key *sk,
9b6e817d 2055 size_t *buf_size,
ba346b35 2056 char __user *ubuf)
ac8e9819
CM		 2057{
2058 struct btrfs_root *root;
2059 struct btrfs_key key;
ac8e9819 2060 struct btrfs_path *path;
ac8e9819
CM		 2061 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2062 int ret;
2063 int num_found = 0;
2064 unsigned long sk_offset = 0;
2065
9b6e817d
GH		 2066 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2067 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2068 return -EOVERFLOW;
9b6e817d 2069 }
12544442 2070
ac8e9819
CM		 2071 path = btrfs_alloc_path();
2072 if (!path)
2073 return -ENOMEM;
2074
2075 if (sk->tree_id == 0) {
2076 /* search the root of the inode that was passed */
2077 root = BTRFS_I(inode)->root;
2078 } else {
2079 key.objectid = sk->tree_id;
2080 key.type = BTRFS_ROOT_ITEM_KEY;
2081 key.offset = (u64)-1;
2082 root = btrfs_read_fs_root_no_name(info, &key);
2083 if (IS_ERR(root)) {
efe120a0 2084 printk(KERN_ERR "BTRFS: could not find root %llu\n",
ac8e9819
CM		 2085 sk->tree_id);
2086 btrfs_free_path(path);
2087 return -ENOENT;
2088 }
2089 }
2090
2091 key.objectid = sk->min_objectid;
2092 key.type = sk->min_type;
2093 key.offset = sk->min_offset;
2094
67871254 2095 while (1) {
6174d3cb 2096 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2097 if (ret != 0) {
2098 if (ret > 0)
2099 ret = 0;
2100 goto err;
2101 }
ba346b35 2102 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
ac8e9819 2103 &sk_offset, &num_found);
b3b4aa74 2104 btrfs_release_path(path);
25c9bc2e 2105 if (ret)
ac8e9819
CM		 2106 break;
2107
2108 }
8f5f6178
GH		 2109 if (ret > 0)
2110 ret = 0;
ac8e9819
CM		 2111err:
2112 sk->nr_items = num_found;
2113 btrfs_free_path(path);
2114 return ret;
2115}
2116
2117static noinline int btrfs_ioctl_tree_search(struct file *file,
2118 void __user *argp)
2119{
ba346b35
GH		 2120 struct btrfs_ioctl_search_args __user *uargs;
2121 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2122 struct inode *inode;
2123 int ret;
2124 size_t buf_size;
ac8e9819
CM		 2125
2126 if (!capable(CAP_SYS_ADMIN))
2127 return -EPERM;
2128
ba346b35
GH		 2129 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2130
2131 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2132 return -EFAULT;
ac8e9819 2133
ba346b35 2134 buf_size = sizeof(uargs->buf);
ac8e9819 2135
496ad9aa 2136 inode = file_inode(file);
ba346b35 2137 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2138
2139 /*
2140 * In the origin implementation an overflow is handled by returning a
2141 * search header with a len of zero, so reset ret.
2142 */
2143 if (ret == -EOVERFLOW)
2144 ret = 0;
2145
ba346b35 2146 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2147 ret = -EFAULT;
ac8e9819
CM		 2148 return ret;
2149}
2150
cc68a8a5
GH		 2151static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2152 void __user *argp)
2153{
2154 struct btrfs_ioctl_search_args_v2 __user *uarg;
2155 struct btrfs_ioctl_search_args_v2 args;
2156 struct inode *inode;
2157 int ret;
2158 size_t buf_size;
2159 const size_t buf_limit = 16 * 1024 * 1024;
2160
2161 if (!capable(CAP_SYS_ADMIN))
2162 return -EPERM;
2163
2164 /* copy search header and buffer size */
2165 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2166 if (copy_from_user(&args, uarg, sizeof(args)))
2167 return -EFAULT;
2168
2169 buf_size = args.buf_size;
2170
2171 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2172 return -EOVERFLOW;
2173
2174 /* limit result size to 16MB */
2175 if (buf_size > buf_limit)
2176 buf_size = buf_limit;
2177
2178 inode = file_inode(file);
2179 ret = search_ioctl(inode, &args.key, &buf_size,
2180 (char *)(&uarg->buf[0]));
2181 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2182 ret = -EFAULT;
2183 else if (ret == -EOVERFLOW &&
2184 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2185 ret = -EFAULT;
2186
ac8e9819
CM		 2187 return ret;
2188}
2189
98d377a0 2190/*
ac8e9819
CM		 2191 * Search INODE_REFs to identify path name of 'dirid' directory
2192 * in a 'tree_id' tree. and sets path name to 'name'.
2193 */
98d377a0
TH		 2194static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2195 u64 tree_id, u64 dirid, char *name)
2196{
2197 struct btrfs_root *root;
2198 struct btrfs_key key;
ac8e9819 2199 char *ptr;
98d377a0
TH		 2200 int ret = -1;
2201 int slot;
2202 int len;
2203 int total_len = 0;
2204 struct btrfs_inode_ref *iref;
2205 struct extent_buffer *l;
2206 struct btrfs_path *path;
2207
2208 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2209 name[0]='\0';
2210 return 0;
2211 }
2212
2213 path = btrfs_alloc_path();
2214 if (!path)
2215 return -ENOMEM;
2216
ac8e9819 2217 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2218
2219 key.objectid = tree_id;
2220 key.type = BTRFS_ROOT_ITEM_KEY;
2221 key.offset = (u64)-1;
2222 root = btrfs_read_fs_root_no_name(info, &key);
2223 if (IS_ERR(root)) {
efe120a0 2224 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
8ad6fcab
CM		 2225 ret = -ENOENT;
2226 goto out;
98d377a0
TH		 2227 }
2228
2229 key.objectid = dirid;
2230 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2231 key.offset = (u64)-1;
98d377a0 2232
67871254 2233 while (1) {
98d377a0
TH		 2234 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2235 if (ret < 0)
2236 goto out;
18674c6c
FDBM		 2237 else if (ret > 0) {
2238 ret = btrfs_previous_item(root, path, dirid,
2239 BTRFS_INODE_REF_KEY);
2240 if (ret < 0)
2241 goto out;
2242 else if (ret > 0) {
2243 ret = -ENOENT;
2244 goto out;
2245 }
2246 }
98d377a0
TH		 2247
2248 l = path->nodes[0];
2249 slot = path->slots[0];
2250 btrfs_item_key_to_cpu(l, &key, slot);
2251
98d377a0
TH		 2252 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2253 len = btrfs_inode_ref_name_len(l, iref);
2254 ptr -= len + 1;
2255 total_len += len + 1;
a696cf35
FDBM		 2256 if (ptr < name) {
2257 ret = -ENAMETOOLONG;
98d377a0 2258 goto out;
a696cf35 2259 }
98d377a0
TH		 2260
2261 *(ptr + len) = '/';
67871254 2262 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2263
2264 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2265 break;
2266
b3b4aa74 2267 btrfs_release_path(path);
98d377a0 2268 key.objectid = key.offset;
8ad6fcab 2269 key.offset = (u64)-1;
98d377a0 2270 dirid = key.objectid;
98d377a0 2271 }
77906a50 2272 memmove(name, ptr, total_len);
67871254 2273 name[total_len] = '\0';
98d377a0
TH		 2274 ret = 0;
2275out:
2276 btrfs_free_path(path);
ac8e9819
CM		 2277 return ret;
2278}
2279
2280static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2281 void __user *argp)
2282{
2283 struct btrfs_ioctl_ino_lookup_args *args;
2284 struct inode *inode;
01b810b8 2285 int ret = 0;
ac8e9819 2286
2354d08f
JL		 2287 args = memdup_user(argp, sizeof(*args));
2288 if (IS_ERR(args))
2289 return PTR_ERR(args);
c2b96929 2290
496ad9aa 2291 inode = file_inode(file);
ac8e9819 2292
01b810b8
DS		 2293 /*
2294 * Unprivileged query to obtain the containing subvolume root id. The
2295 * path is reset so it's consistent with btrfs_search_path_in_tree.
2296 */
1b53ac4d
CM		 2297 if (args->treeid == 0)
2298 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2299
01b810b8
DS		 2300 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2301 args->name[0] = 0;
2302 goto out;
2303 }
2304
2305 if (!capable(CAP_SYS_ADMIN)) {
2306 ret = -EPERM;
2307 goto out;
2308 }
2309
ac8e9819
CM		 2310 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2311 args->treeid, args->objectid,
2312 args->name);
2313
01b810b8 2314out:
ac8e9819
CM		 2315 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2316 ret = -EFAULT;
2317
2318 kfree(args);
98d377a0
TH		 2319 return ret;
2320}
2321
76dda93c
YZ		 2322static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2323 void __user *arg)
2324{
54563d41 2325 struct dentry *parent = file->f_path.dentry;
76dda93c 2326 struct dentry *dentry;
2b0143b5 2327 struct inode *dir = d_inode(parent);
76dda93c
YZ		 2328 struct inode *inode;
2329 struct btrfs_root *root = BTRFS_I(dir)->root;
2330 struct btrfs_root *dest = NULL;
2331 struct btrfs_ioctl_vol_args *vol_args;
2332 struct btrfs_trans_handle *trans;
c58aaad2 2333 struct btrfs_block_rsv block_rsv;
521e0546 2334 u64 root_flags;
c58aaad2 2335 u64 qgroup_reserved;
76dda93c
YZ		 2336 int namelen;
2337 int ret;
2338 int err = 0;
2339
76dda93c
YZ		 2340 vol_args = memdup_user(arg, sizeof(*vol_args));
2341 if (IS_ERR(vol_args))
2342 return PTR_ERR(vol_args);
2343
2344 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2345 namelen = strlen(vol_args->name);
2346 if (strchr(vol_args->name, '/') ||
2347 strncmp(vol_args->name, "..", namelen) == 0) {
2348 err = -EINVAL;
2349 goto out;
2350 }
2351
a561be71 2352 err = mnt_want_write_file(file);
76dda93c
YZ		 2353 if (err)
2354 goto out;
2355
521e0546 2356
5c50c9b8
DS		 2357 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2358 if (err == -EINTR)
e43f998e 2359 goto out_drop_write;
76dda93c
YZ		 2360 dentry = lookup_one_len(vol_args->name, parent, namelen);
2361 if (IS_ERR(dentry)) {
2362 err = PTR_ERR(dentry);
2363 goto out_unlock_dir;
2364 }
2365
2b0143b5 2366 if (d_really_is_negative(dentry)) {
76dda93c
YZ		 2367 err = -ENOENT;
2368 goto out_dput;
2369 }
2370
2b0143b5 2371 inode = d_inode(dentry);
4260f7c7 2372 dest = BTRFS_I(inode)->root;
67871254 2373 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2374 /*
2375 * Regular user. Only allow this with a special mount
2376 * option, when the user has write+exec access to the
2377 * subvol root, and when rmdir(2) would have been
2378 * allowed.
2379 *
2380 * Note that this is _not_ check that the subvol is
2381 * empty or doesn't contain data that we wouldn't
2382 * otherwise be able to delete.
2383 *
2384 * Users who want to delete empty subvols should try
2385 * rmdir(2).
2386 */
2387 err = -EPERM;
2388 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2389 goto out_dput;
2390
2391 /*
2392 * Do not allow deletion if the parent dir is the same
2393 * as the dir to be deleted. That means the ioctl
2394 * must be called on the dentry referencing the root
2395 * of the subvol, not a random directory contained
2396 * within it.
2397 */
2398 err = -EINVAL;
2399 if (root == dest)
2400 goto out_dput;
2401
2402 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2403 if (err)
2404 goto out_dput;
4260f7c7
SW		 2405 }
2406
5c39da5b
MX		 2407 /* check if subvolume may be deleted by a user */
2408 err = btrfs_may_delete(dir, dentry, 1);
2409 if (err)
2410 goto out_dput;
2411
33345d01 2412 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2413 err = -EINVAL;
2414 goto out_dput;
2415 }
2416
76dda93c 2417 mutex_lock(&inode->i_mutex);
521e0546
DS		 2418
2419 /*
2420 * Don't allow to delete a subvolume with send in progress. This is
2421 * inside the i_mutex so the error handling that has to drop the bit
2422 * again is not run concurrently.
2423 */
2424 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2425 root_flags = btrfs_root_flags(&dest->root_item);
2426 if (dest->send_in_progress == 0) {
2427 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2428 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2429 spin_unlock(&dest->root_item_lock);
2430 } else {
2431 spin_unlock(&dest->root_item_lock);
2432 btrfs_warn(root->fs_info,
2433 "Attempt to delete subvolume %llu during send",
c55bfa67 2434 dest->root_key.objectid);
521e0546 2435 err = -EPERM;
909e26dc 2436 goto out_unlock_inode;
521e0546
DS		 2437 }
2438
76dda93c
YZ		 2439 down_write(&root->fs_info->subvol_sem);
2440
2441 err = may_destroy_subvol(dest);
2442 if (err)
2443 goto out_up_write;
2444
c58aaad2
MX		 2445 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2446 /*
2447 * One for dir inode, two for dir entries, two for root
2448 * ref/backref.
2449 */
2450 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2451 5, &qgroup_reserved, true);
c58aaad2
MX		 2452 if (err)
2453 goto out_up_write;
2454
a22285a6
YZ		 2455 trans = btrfs_start_transaction(root, 0);
2456 if (IS_ERR(trans)) {
2457 err = PTR_ERR(trans);
c58aaad2 2458 goto out_release;
a22285a6 2459 }
c58aaad2
MX		 2460 trans->block_rsv = &block_rsv;
2461 trans->bytes_reserved = block_rsv.size;
a22285a6 2462
76dda93c
YZ		 2463 ret = btrfs_unlink_subvol(trans, root, dir,
2464 dest->root_key.objectid,
2465 dentry->d_name.name,
2466 dentry->d_name.len);
79787eaa
JM		 2467 if (ret) {
2468 err = ret;
2469 btrfs_abort_transaction(trans, root, ret);
2470 goto out_end_trans;
2471 }
76dda93c
YZ		 2472
2473 btrfs_record_root_in_trans(trans, dest);
2474
2475 memset(&dest->root_item.drop_progress, 0,
2476 sizeof(dest->root_item.drop_progress));
2477 dest->root_item.drop_level = 0;
2478 btrfs_set_root_refs(&dest->root_item, 0);
2479
27cdeb70 2480 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
d68fc57b
YZ		 2481 ret = btrfs_insert_orphan_item(trans,
2482 root->fs_info->tree_root,
2483 dest->root_key.objectid);
79787eaa
JM		 2484 if (ret) {
2485 btrfs_abort_transaction(trans, root, ret);
2486 err = ret;
2487 goto out_end_trans;
2488 }
d68fc57b 2489 }
dd5f9615
SB		 2490
2491 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2492 dest->root_item.uuid, BTRFS_UUID_KEY_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 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2500 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2501 dest->root_item.received_uuid,
2502 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2503 dest->root_key.objectid);
2504 if (ret && ret != -ENOENT) {
2505 btrfs_abort_transaction(trans, root, ret);
2506 err = ret;
2507 goto out_end_trans;
2508 }
2509 }
2510
79787eaa 2511out_end_trans:
c58aaad2
MX		 2512 trans->block_rsv = NULL;
2513 trans->bytes_reserved = 0;
531cb13f 2514 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2515 if (ret && !err)
2516 err = ret;
76dda93c 2517 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2518out_release:
2519 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2520out_up_write:
2521 up_write(&root->fs_info->subvol_sem);
521e0546
DS		 2522 if (err) {
2523 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2524 root_flags = btrfs_root_flags(&dest->root_item);
2525 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2526 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2527 spin_unlock(&dest->root_item_lock);
2528 }
909e26dc 2529out_unlock_inode:
76dda93c
YZ		 2530 mutex_unlock(&inode->i_mutex);
2531 if (!err) {
64ad6c48 2532 d_invalidate(dentry);
76dda93c
YZ		 2533 btrfs_invalidate_inodes(dest);
2534 d_delete(dentry);
61155aa0 2535 ASSERT(dest->send_in_progress == 0);
fa6ac876
LB		 2536
2537 /* the last ref */
57cdc8db
DS		 2538 if (dest->ino_cache_inode) {
2539 iput(dest->ino_cache_inode);
2540 dest->ino_cache_inode = NULL;
fa6ac876 2541 }
76dda93c
YZ		 2542 }
2543out_dput:
2544 dput(dentry);
2545out_unlock_dir:
2546 mutex_unlock(&dir->i_mutex);
e43f998e 2547out_drop_write:
2a79f17e 2548 mnt_drop_write_file(file);
76dda93c
YZ		 2549out:
2550 kfree(vol_args);
2551 return err;
2552}
2553
1e701a32 2554static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2555{
496ad9aa 2556 struct inode *inode = file_inode(file);
f46b5a66 2557 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2558 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2559 int ret;
2560
25122d15
ID		 2561 ret = mnt_want_write_file(file);
2562 if (ret)
2563 return ret;
b83cc969 2564
25122d15
ID		 2565 if (btrfs_root_readonly(root)) {
2566 ret = -EROFS;
2567 goto out;
5ac00add 2568 }
f46b5a66
CH		 2569
2570 switch (inode->i_mode & S_IFMT) {
2571 case S_IFDIR:
e441d54d
CM		 2572 if (!capable(CAP_SYS_ADMIN)) {
2573 ret = -EPERM;
2574 goto out;
2575 }
de78b51a 2576 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2577 if (ret)
2578 goto out;
de78b51a 2579 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2580 break;
2581 case S_IFREG:
e441d54d
CM		 2582 if (!(file->f_mode & FMODE_WRITE)) {
2583 ret = -EINVAL;
2584 goto out;
2585 }
1e701a32
CM		 2586
2587 range = kzalloc(sizeof(*range), GFP_KERNEL);
2588 if (!range) {
2589 ret = -ENOMEM;
2590 goto out;
2591 }
2592
2593 if (argp) {
2594 if (copy_from_user(range, argp,
2595 sizeof(*range))) {
2596 ret = -EFAULT;
2597 kfree(range);
683be16e 2598 goto out;
1e701a32
CM		 2599 }
2600 /* compression requires us to start the IO */
2601 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2602 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2603 range->extent_thresh = (u32)-1;
2604 }
2605 } else {
2606 /* the rest are all set to zero by kzalloc */
2607 range->len = (u64)-1;
2608 }
496ad9aa 2609 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2610 range, 0, 0);
2611 if (ret > 0)
2612 ret = 0;
1e701a32 2613 kfree(range);
f46b5a66 2614 break;
8929ecfa
YZ		 2615 default:
2616 ret = -EINVAL;
f46b5a66 2617 }
e441d54d 2618out:
25122d15 2619 mnt_drop_write_file(file);
e441d54d 2620 return ret;
f46b5a66
CH		 2621}
2622
b2950863 2623static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2624{
2625 struct btrfs_ioctl_vol_args *vol_args;
2626 int ret;
2627
e441d54d
CM		 2628 if (!capable(CAP_SYS_ADMIN))
2629 return -EPERM;
2630
5ac00add
SB		 2631 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2632 1)) {
e57138b3 2633 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 2634 }
2635
5ac00add 2636 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2637 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2638 if (IS_ERR(vol_args)) {
2639 ret = PTR_ERR(vol_args);
2640 goto out;
2641 }
f46b5a66 2642
5516e595 2643 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2644 ret = btrfs_init_new_device(root, vol_args->name);
2645
43d20761
AJ		 2646 if (!ret)
2647 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2648
f46b5a66 2649 kfree(vol_args);
c9e9f97b
ID		 2650out:
2651 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2652 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2653 return ret;
2654}
2655
da24927b 2656static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2657{
496ad9aa 2658 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2659 struct btrfs_ioctl_vol_args *vol_args;
2660 int ret;
2661
e441d54d
CM		 2662 if (!capable(CAP_SYS_ADMIN))
2663 return -EPERM;
2664
da24927b
MX		 2665 ret = mnt_want_write_file(file);
2666 if (ret)
2667 return ret;
c146afad 2668
dae7b665 2669 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2670 if (IS_ERR(vol_args)) {
2671 ret = PTR_ERR(vol_args);
c47ca32d 2672 goto err_drop;
c9e9f97b 2673 }
f46b5a66 2674
5516e595 2675 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2676
183860f6
AJ		 2677 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2678 1)) {
2679 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2680 goto out;
2681 }
2682
2683 mutex_lock(&root->fs_info->volume_mutex);
2684 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2685 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2686 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6 2687
ec95d491
AJ		 2688 if (!ret)
2689 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2690
183860f6
AJ		 2691out:
2692 kfree(vol_args);
c47ca32d 2693err_drop:
4ac20c70 2694 mnt_drop_write_file(file);
f46b5a66
CH		 2695 return ret;
2696}
2697
475f6387
JS		 2698static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2699{
027ed2f0 2700 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2701 struct btrfs_device *device;
2702 struct btrfs_device *next;
2703 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2704 int ret = 0;
475f6387 2705
027ed2f0
LZ		 2706 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2707 if (!fi_args)
2708 return -ENOMEM;
2709
f7171750 2710 mutex_lock(&fs_devices->device_list_mutex);
027ed2f0
LZ		 2711 fi_args->num_devices = fs_devices->num_devices;
2712 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387 2713
475f6387 2714 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2715 if (device->devid > fi_args->max_id)
2716 fi_args->max_id = device->devid;
475f6387
JS		 2717 }
2718 mutex_unlock(&fs_devices->device_list_mutex);
2719
80a773fb
DS		 2720 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2721 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2722 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2723
027ed2f0
LZ		 2724 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2725 ret = -EFAULT;
475f6387 2726
027ed2f0
LZ		 2727 kfree(fi_args);
2728 return ret;
475f6387
JS		 2729}
2730
2731static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2732{
2733 struct btrfs_ioctl_dev_info_args *di_args;
2734 struct btrfs_device *dev;
2735 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2736 int ret = 0;
2737 char *s_uuid = NULL;
475f6387 2738
475f6387
JS		 2739 di_args = memdup_user(arg, sizeof(*di_args));
2740 if (IS_ERR(di_args))
2741 return PTR_ERR(di_args);
2742
dd5f9615 2743 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2744 s_uuid = di_args->uuid;
2745
2746 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2747 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2748
2749 if (!dev) {
2750 ret = -ENODEV;
2751 goto out;
2752 }
2753
2754 di_args->devid = dev->devid;
7cc8e58d
MX		 2755 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2756 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
475f6387 2757 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2758 if (dev->name) {
606686ee
JB		 2759 struct rcu_string *name;
2760
2761 rcu_read_lock();
2762 name = rcu_dereference(dev->name);
2763 strncpy(di_args->path, name->str, sizeof(di_args->path));
2764 rcu_read_unlock();
a27202fb
JM		 2765 di_args->path[sizeof(di_args->path) - 1] = 0;
2766 } else {
99ba55ad 2767 di_args->path[0] = '\0';
a27202fb 2768 }
475f6387
JS		 2769
2770out:
55793c0d 2771 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2772 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2773 ret = -EFAULT;
2774
2775 kfree(di_args);
2776 return ret;
2777}
2778
f4414602 2779static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
416161db
MF		 2780{
2781 struct page *page;
416161db
MF		 2782 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2783
416161db
MF		 2784 page = grab_cache_page(inode->i_mapping, index);
2785 if (!page)
2786 return NULL;
2787
2788 if (!PageUptodate(page)) {
2789 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2790 0))
2791 return NULL;
2792 lock_page(page);
2793 if (!PageUptodate(page)) {
2794 unlock_page(page);
2795 page_cache_release(page);
2796 return NULL;
2797 }
2798 }
2799 unlock_page(page);
2800
2801 return page;
2802}
2803
f4414602
MF		 2804static int gather_extent_pages(struct inode *inode, struct page **pages,
2805 int num_pages, u64 off)
2806{
2807 int i;
2808 pgoff_t index = off >> PAGE_CACHE_SHIFT;
2809
2810 for (i = 0; i < num_pages; i++) {
2811 pages[i] = extent_same_get_page(inode, index + i);
2812 if (!pages[i])
2813 return -ENOMEM;
2814 }
2815 return 0;
2816}
2817
77fe20dc
MF		 2818static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2819{
2820 /* do any pending delalloc/csum calc on src, one way or
2821 another, and lock file content */
2822 while (1) {
2823 struct btrfs_ordered_extent *ordered;
2824 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2825 ordered = btrfs_lookup_first_ordered_extent(inode,
2826 off + len - 1);
ff5df9b8
FM		 2827 if ((!ordered ||
2828 ordered->file_offset + ordered->len <= off ||
2829 ordered->file_offset >= off + len) &&
77fe20dc 2830 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
ff5df9b8
FM		 2831 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2832 if (ordered)
2833 btrfs_put_ordered_extent(ordered);
77fe20dc 2834 break;
ff5df9b8 2835 }
77fe20dc
MF		 2836 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2837 if (ordered)
2838 btrfs_put_ordered_extent(ordered);
2839 btrfs_wait_ordered_range(inode, off, len);
2840 }
2841}
2842
f4414602 2843static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
416161db 2844{
416161db
MF		 2845 mutex_unlock(&inode1->i_mutex);
2846 mutex_unlock(&inode2->i_mutex);
2847}
2848
f4414602
MF		 2849static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2850{
2851 if (inode1 < inode2)
2852 swap(inode1, inode2);
2853
2854 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2855 if (inode1 != inode2)
2856 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2857}
2858
2859static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2860 struct inode *inode2, u64 loff2, u64 len)
2861{
2862 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2863 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2864}
2865
2866static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2867 struct inode *inode2, u64 loff2, u64 len)
416161db
MF		 2868{
2869 if (inode1 < inode2) {
2870 swap(inode1, inode2);
2871 swap(loff1, loff2);
2872 }
416161db 2873 lock_extent_range(inode1, loff1, len);
f4414602 2874 if (inode1 != inode2)
416161db 2875 lock_extent_range(inode2, loff2, len);
f4414602
MF		 2876}
2877
2878struct cmp_pages {
2879 int num_pages;
2880 struct page **src_pages;
2881 struct page **dst_pages;
2882};
2883
2884static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2885{
2886 int i;
2887 struct page *pg;
2888
2889 for (i = 0; i < cmp->num_pages; i++) {
2890 pg = cmp->src_pages[i];
2891 if (pg)
2892 page_cache_release(pg);
2893 pg = cmp->dst_pages[i];
2894 if (pg)
2895 page_cache_release(pg);
2896 }
2897 kfree(cmp->src_pages);
2898 kfree(cmp->dst_pages);
2899}
2900
2901static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2902 struct inode *dst, u64 dst_loff,
2903 u64 len, struct cmp_pages *cmp)
2904{
2905 int ret;
2906 int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
2907 struct page **src_pgarr, **dst_pgarr;
2908
2909 /*
2910 * We must gather up all the pages before we initiate our
2911 * extent locking. We use an array for the page pointers. Size
2912 * of the array is bounded by len, which is in turn bounded by
2913 * BTRFS_MAX_DEDUPE_LEN.
2914 */
2915 src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2916 dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2917 if (!src_pgarr || !dst_pgarr) {
2918 kfree(src_pgarr);
2919 kfree(dst_pgarr);
2920 return -ENOMEM;
416161db 2921 }
f4414602
MF		 2922 cmp->num_pages = num_pages;
2923 cmp->src_pages = src_pgarr;
2924 cmp->dst_pages = dst_pgarr;
2925
2926 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2927 if (ret)
2928 goto out;
2929
2930 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2931
2932out:
2933 if (ret)
2934 btrfs_cmp_data_free(cmp);
2935 return 0;
416161db
MF		 2936}
2937
2938static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
f4414602 2939 u64 dst_loff, u64 len, struct cmp_pages *cmp)
416161db
MF		 2940{
2941 int ret = 0;
f4414602 2942 int i;
416161db
MF		 2943 struct page *src_page, *dst_page;
2944 unsigned int cmp_len = PAGE_CACHE_SIZE;
2945 void *addr, *dst_addr;
2946
f4414602 2947 i = 0;
416161db
MF		 2948 while (len) {
2949 if (len < PAGE_CACHE_SIZE)
2950 cmp_len = len;
2951
f4414602
MF		 2952 BUG_ON(i >= cmp->num_pages);
2953
2954 src_page = cmp->src_pages[i];
2955 dst_page = cmp->dst_pages[i];
2956
416161db
MF		 2957 addr = kmap_atomic(src_page);
2958 dst_addr = kmap_atomic(dst_page);
2959
2960 flush_dcache_page(src_page);
2961 flush_dcache_page(dst_page);
2962
2963 if (memcmp(addr, dst_addr, cmp_len))
2964 ret = BTRFS_SAME_DATA_DIFFERS;
2965
2966 kunmap_atomic(addr);
2967 kunmap_atomic(dst_addr);
416161db
MF		 2968
2969 if (ret)
2970 break;
2971
416161db 2972 len -= cmp_len;
f4414602 2973 i++;
416161db
MF		 2974 }
2975
2976 return ret;
2977}
2978
e1d227a4
MF		 2979static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
2980 u64 olen)
416161db 2981{
e1d227a4 2982 u64 len = *plen;
416161db
MF		 2983 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2984
e1d227a4 2985 if (off + olen > inode->i_size || off + olen < off)
416161db 2986 return -EINVAL;
e1d227a4
MF		 2987
2988 /* if we extend to eof, continue to block boundary */
2989 if (off + len == inode->i_size)
2990 *plen = len = ALIGN(inode->i_size, bs) - off;
2991
416161db
MF		 2992 /* Check that we are block aligned - btrfs_clone() requires this */
2993 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2994 return -EINVAL;
2995
2996 return 0;
2997}
2998
e1d227a4 2999static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
416161db
MF		 3000 struct inode *dst, u64 dst_loff)
3001{
3002 int ret;
e1d227a4 3003 u64 len = olen;
f4414602 3004 struct cmp_pages cmp;
0efa9f48
MF		 3005 int same_inode = 0;
3006 u64 same_lock_start = 0;
3007 u64 same_lock_len = 0;
416161db 3008
416161db 3009 if (src == dst)
0efa9f48 3010 same_inode = 1;
416161db 3011
113e8283
FM		 3012 if (len == 0)
3013 return 0;
3014
0efa9f48
MF		 3015 if (same_inode) {
3016 mutex_lock(&src->i_mutex);
416161db 3017
0efa9f48
MF		 3018 ret = extent_same_check_offsets(src, loff, &len, olen);
3019 if (ret)
3020 goto out_unlock;
416161db 3021
0efa9f48
MF		 3022 /*
3023 * Single inode case wants the same checks, except we
3024 * don't want our length pushed out past i_size as
3025 * comparing that data range makes no sense.
3026 *
3027 * extent_same_check_offsets() will do this for an
3028 * unaligned length at i_size, so catch it here and
3029 * reject the request.
3030 *
3031 * This effectively means we require aligned extents
3032 * for the single-inode case, whereas the other cases
3033 * allow an unaligned length so long as it ends at
3034 * i_size.
3035 */
3036 if (len != olen) {
3037 ret = -EINVAL;
3038 goto out_unlock;
3039 }
3040
3041 /* Check for overlapping ranges */
3042 if (dst_loff + len > loff && dst_loff < loff + len) {
3043 ret = -EINVAL;
3044 goto out_unlock;
3045 }
3046
3047 same_lock_start = min_t(u64, loff, dst_loff);
3048 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3049 } else {
3050 btrfs_double_inode_lock(src, dst);
3051
3052 ret = extent_same_check_offsets(src, loff, &len, olen);
3053 if (ret)
3054 goto out_unlock;
3055
3056 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3057 if (ret)
3058 goto out_unlock;
3059 }
416161db
MF		 3060
3061 /* don't make the dst file partly checksummed */
3062 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3063 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3064 ret = -EINVAL;
3065 goto out_unlock;
3066 }
3067
f4414602
MF		 3068 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3069 if (ret)
3070 goto out_unlock;
3071
0efa9f48
MF		 3072 if (same_inode)
3073 lock_extent_range(src, same_lock_start, same_lock_len);
3074 else
3075 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
f4414602 3076
207910dd 3077 /* pass original length for comparison so we stay within i_size */
f4414602 3078 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
416161db 3079 if (ret == 0)
1c919a5e 3080 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
416161db 3081
0efa9f48
MF		 3082 if (same_inode)
3083 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3084 same_lock_start + same_lock_len - 1);
3085 else
3086 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
f4414602
MF		 3087
3088 btrfs_cmp_data_free(&cmp);
416161db 3089out_unlock:
0efa9f48
MF		 3090 if (same_inode)
3091 mutex_unlock(&src->i_mutex);
3092 else
3093 btrfs_double_inode_unlock(src, dst);
416161db
MF		 3094
3095 return ret;
3096}
3097
3098#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
3099
3100static long btrfs_ioctl_file_extent_same(struct file *file,
1c1c8747 3101 struct btrfs_ioctl_same_args __user *argp)
416161db 3102{
497b4050 3103 struct btrfs_ioctl_same_args *same = NULL;
cbf8b8ca 3104 struct btrfs_ioctl_same_extent_info *info;
1c1c8747 3105 struct inode *src = file_inode(file);
416161db
MF		 3106 u64 off;
3107 u64 len;
3108 int i;
3109 int ret;
cbf8b8ca 3110 unsigned long size;
416161db
MF		 3111 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3112 bool is_admin = capable(CAP_SYS_ADMIN);
1c1c8747 3113 u16 count;
416161db
MF		 3114
3115 if (!(file->f_mode & FMODE_READ))
3116 return -EINVAL;
3117
3118 ret = mnt_want_write_file(file);
3119 if (ret)
3120 return ret;
3121
1c1c8747 3122 if (get_user(count, &argp->dest_count)) {
416161db
MF		 3123 ret = -EFAULT;
3124 goto out;
3125 }
3126
1c1c8747 3127 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
cbf8b8ca 3128
1c1c8747 3129 same = memdup_user(argp, size);
cbf8b8ca 3130
229eed43
GB		 3131 if (IS_ERR(same)) {
3132 ret = PTR_ERR(same);
497b4050 3133 same = NULL;
cbf8b8ca
MF		 3134 goto out;
3135 }
3136
3137 off = same->logical_offset;
3138 len = same->length;
416161db
MF		 3139
3140 /*
3141 * Limit the total length we will dedupe for each operation.
3142 * This is intended to bound the total time spent in this
3143 * ioctl to something sane.
3144 */
3145 if (len > BTRFS_MAX_DEDUPE_LEN)
3146 len = BTRFS_MAX_DEDUPE_LEN;
3147
3148 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3149 /*
3150 * Btrfs does not support blocksize < page_size. As a
3151 * result, btrfs_cmp_data() won't correctly handle
3152 * this situation without an update.
3153 */
3154 ret = -EINVAL;
3155 goto out;
3156 }
3157
3158 ret = -EISDIR;
3159 if (S_ISDIR(src->i_mode))
3160 goto out;
3161
3162 ret = -EACCES;
3163 if (!S_ISREG(src->i_mode))
3164 goto out;
3165
cbf8b8ca 3166 /* pre-format output fields to sane values */
1c1c8747 3167 for (i = 0; i < count; i++) {
cbf8b8ca
MF		 3168 same->info[i].bytes_deduped = 0ULL;
3169 same->info[i].status = 0;
3170 }
416161db 3171
1c1c8747
AV		 3172 for (i = 0, info = same->info; i < count; i++, info++) {
3173 struct inode *dst;
3174 struct fd dst_file = fdget(info->fd);
3175 if (!dst_file.file) {
cbf8b8ca 3176 info->status = -EBADF;
1c1c8747 3177 continue;
416161db 3178 }
1c1c8747 3179 dst = file_inode(dst_file.file);
416161db 3180
1c1c8747 3181 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
cbf8b8ca 3182 info->status = -EINVAL;
1c1c8747
AV		 3183 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3184 info->status = -EXDEV;
3185 } else if (S_ISDIR(dst->i_mode)) {
cbf8b8ca 3186 info->status = -EISDIR;
1c1c8747 3187 } else if (!S_ISREG(dst->i_mode)) {
cbf8b8ca 3188 info->status = -EACCES;
1c1c8747
AV		 3189 } else {
3190 info->status = btrfs_extent_same(src, off, len, dst,
3191 info->logical_offset);
3192 if (info->status == 0)
3193 info->bytes_deduped += len;
416161db 3194 }
1c1c8747 3195 fdput(dst_file);
416161db
MF		 3196 }
3197
cbf8b8ca
MF		 3198 ret = copy_to_user(argp, same, size);
3199 if (ret)
3200 ret = -EFAULT;
3201
416161db
MF		 3202out:
3203 mnt_drop_write_file(file);
497b4050 3204 kfree(same);
416161db
MF		 3205 return ret;
3206}
3207
fcebe456
JB		 3208/* Helper to check and see if this root currently has a ref on the given disk
3209 * bytenr. If it does then we need to update the quota for this root. This
3210 * doesn't do anything if quotas aren't enabled.
3211 */
3212static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3213 u64 disko)
3214{
3284da7b 3215 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
fcebe456
JB		 3216 struct ulist *roots;
3217 struct ulist_iterator uiter;
3218 struct ulist_node *root_node = NULL;
3219 int ret;
3220
3221 if (!root->fs_info->quota_enabled)
3222 return 1;
3223
3224 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3225 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3226 tree_mod_seq_elem.seq, &roots);
3227 if (ret < 0)
3228 goto out;
3229 ret = 0;
3230 ULIST_ITER_INIT(&uiter);
3231 while ((root_node = ulist_next(roots, &uiter))) {
3232 if (root_node->val == root->objectid) {
3233 ret = 1;
3234 break;
3235 }
3236 }
3237 ulist_free(roots);
3238out:
3239 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3240 return ret;
3241}
3242
f82a9901
FM		 3243static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3244 struct inode *inode,
3245 u64 endoff,
3246 const u64 destoff,
1c919a5e
MF		 3247 const u64 olen,
3248 int no_time_update)
f82a9901
FM		 3249{
3250 struct btrfs_root *root = BTRFS_I(inode)->root;
3251 int ret;
3252
3253 inode_inc_iversion(inode);
1c919a5e
MF		 3254 if (!no_time_update)
3255 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
f82a9901
FM		 3256 /*
3257 * We round up to the block size at eof when determining which
3258 * extents to clone above, but shouldn't round up the file size.
3259 */
3260 if (endoff > destoff + olen)
3261 endoff = destoff + olen;
3262 if (endoff > inode->i_size)
3263 btrfs_i_size_write(inode, endoff);
3264
3265 ret = btrfs_update_inode(trans, root, inode);
3266 if (ret) {
3267 btrfs_abort_transaction(trans, root, ret);
3268 btrfs_end_transaction(trans, root);
3269 goto out;
3270 }
3271 ret = btrfs_end_transaction(trans, root);
3272out:
3273 return ret;
3274}
3275
7ffbb598
FM		 3276static void clone_update_extent_map(struct inode *inode,
3277 const struct btrfs_trans_handle *trans,
3278 const struct btrfs_path *path,
7ffbb598
FM		 3279 const u64 hole_offset,
3280 const u64 hole_len)
3281{
3282 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3283 struct extent_map *em;
3284 int ret;
3285
3286 em = alloc_extent_map();
3287 if (!em) {
3288 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3289 &BTRFS_I(inode)->runtime_flags);
3290 return;
3291 }
3292
14f59796
FM		 3293 if (path) {
3294 struct btrfs_file_extent_item *fi;
3295
3296 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3297 struct btrfs_file_extent_item);
7ffbb598
FM		 3298 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3299 em->generation = -1;
3300 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3301 BTRFS_FILE_EXTENT_INLINE)
3302 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3303 &BTRFS_I(inode)->runtime_flags);
3304 } else {
3305 em->start = hole_offset;
3306 em->len = hole_len;
3307 em->ram_bytes = em->len;
3308 em->orig_start = hole_offset;
3309 em->block_start = EXTENT_MAP_HOLE;
3310 em->block_len = 0;
3311 em->orig_block_len = 0;
3312 em->compress_type = BTRFS_COMPRESS_NONE;
3313 em->generation = trans->transid;
3314 }
3315
3316 while (1) {
3317 write_lock(&em_tree->lock);
3318 ret = add_extent_mapping(em_tree, em, 1);
3319 write_unlock(&em_tree->lock);
3320 if (ret != -EEXIST) {
3321 free_extent_map(em);
3322 break;
3323 }
3324 btrfs_drop_extent_cache(inode, em->start,
3325 em->start + em->len - 1, 0);
3326 }
3327
ee39b432 3328 if (ret)
7ffbb598
FM		 3329 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3330 &BTRFS_I(inode)->runtime_flags);
3331}
3332
32b7c687
MF		 3333/**
3334 * btrfs_clone() - clone a range from inode file to another
3335 *
3336 * @src: Inode to clone from
3337 * @inode: Inode to clone to
3338 * @off: Offset within source to start clone from
3339 * @olen: Original length, passed by user, of range to clone
1c919a5e 3340 * @olen_aligned: Block-aligned value of olen
32b7c687 3341 * @destoff: Offset within @inode to start clone
1c919a5e 3342 * @no_time_update: Whether to update mtime/ctime on the target inode
32b7c687
MF		 3343 */
3344static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901 3345 const u64 off, const u64 olen, const u64 olen_aligned,
1c919a5e 3346 const u64 destoff, int no_time_update)
f46b5a66 3347{
f46b5a66 3348 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3349 struct btrfs_path *path = NULL;
f46b5a66 3350 struct extent_buffer *leaf;
32b7c687
MF		 3351 struct btrfs_trans_handle *trans;
3352 char *buf = NULL;
ae01a0ab 3353 struct btrfs_key key;
f46b5a66
CH		 3354 u32 nritems;
3355 int slot;
ae01a0ab 3356 int ret;
fcebe456 3357 int no_quota;
f82a9901 3358 const u64 len = olen_aligned;
fcebe456 3359 u64 last_disko = 0;
f82a9901 3360 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3361
3362 ret = -ENOMEM;
707e8a07 3363 buf = vmalloc(root->nodesize);
ae01a0ab 3364 if (!buf)
32b7c687 3365 return ret;
ae01a0ab
YZ		 3366
3367 path = btrfs_alloc_path();
3368 if (!path) {
3369 vfree(buf);
32b7c687 3370 return ret;
d525e8ab
LZ		 3371 }
3372
32b7c687 3373 path->reada = 2;
c5c9cd4d 3374 /* clone data */
33345d01 3375 key.objectid = btrfs_ino(src);
ae01a0ab 3376 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3377 key.offset = off;
f46b5a66
CH		 3378
3379 while (1) {
de249e66 3380 u64 next_key_min_offset = key.offset + 1;
df858e76 3381
f46b5a66
CH		 3382 /*
3383 * note the key will change type as we walk through the
3384 * tree.
3385 */
e4355f34 3386 path->leave_spinning = 1;
362a20c5
DS		 3387 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3388 0, 0);
f46b5a66
CH		 3389 if (ret < 0)
3390 goto out;
2c463823
FM		 3391 /*
3392 * First search, if no extent item that starts at offset off was
3393 * found but the previous item is an extent item, it's possible
3394 * it might overlap our target range, therefore process it.
3395 */
3396 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3397 btrfs_item_key_to_cpu(path->nodes[0], &key,
3398 path->slots[0] - 1);
3399 if (key.type == BTRFS_EXTENT_DATA_KEY)
3400 path->slots[0]--;
3401 }
f46b5a66 3402
ae01a0ab 3403 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3404process_slot:
fcebe456 3405 no_quota = 1;
ae01a0ab 3406 if (path->slots[0] >= nritems) {
362a20c5 3407 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3408 if (ret < 0)
3409 goto out;
3410 if (ret > 0)
3411 break;
ae01a0ab 3412 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3413 }
3414 leaf = path->nodes[0];
3415 slot = path->slots[0];
f46b5a66 3416
ae01a0ab 3417 btrfs_item_key_to_cpu(leaf, &key, slot);
962a298f 3418 if (key.type > BTRFS_EXTENT_DATA_KEY ||
33345d01 3419 key.objectid != btrfs_ino(src))
f46b5a66
CH		 3420 break;
3421
962a298f 3422 if (key.type == BTRFS_EXTENT_DATA_KEY) {
c5c9cd4d
SW		 3423 struct btrfs_file_extent_item *extent;
3424 int type;
31840ae1
ZY		 3425 u32 size;
3426 struct btrfs_key new_key;
c5c9cd4d
SW		 3427 u64 disko = 0, diskl = 0;
3428 u64 datao = 0, datal = 0;
3429 u8 comp;
f82a9901 3430 u64 drop_start;
31840ae1 3431
c5c9cd4d
SW		 3432 extent = btrfs_item_ptr(leaf, slot,
3433 struct btrfs_file_extent_item);
3434 comp = btrfs_file_extent_compression(leaf, extent);
3435 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3436 if (type == BTRFS_FILE_EXTENT_REG ||
3437 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3438 disko = btrfs_file_extent_disk_bytenr(leaf,
3439 extent);
3440 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3441 extent);
c5c9cd4d 3442 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3443 datal = btrfs_file_extent_num_bytes(leaf,
3444 extent);
c5c9cd4d
SW		 3445 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3446 /* take upper bound, may be compressed */
3447 datal = btrfs_file_extent_ram_bytes(leaf,
3448 extent);
3449 }
31840ae1 3450
2c463823
FM		 3451 /*
3452 * The first search might have left us at an extent
3453 * item that ends before our target range's start, can
3454 * happen if we have holes and NO_HOLES feature enabled.
3455 */
3456 if (key.offset + datal <= off) {
e4355f34
FDBM		 3457 path->slots[0]++;
3458 goto process_slot;
2c463823
FM		 3459 } else if (key.offset >= off + len) {
3460 break;
e4355f34 3461 }
df858e76 3462 next_key_min_offset = key.offset + datal;
e4355f34
FDBM		 3463 size = btrfs_item_size_nr(leaf, slot);
3464 read_extent_buffer(leaf, buf,
3465 btrfs_item_ptr_offset(leaf, slot),
3466 size);
3467
3468 btrfs_release_path(path);
3469 path->leave_spinning = 0;
c5c9cd4d 3470
31840ae1 3471 memcpy(&new_key, &key, sizeof(new_key));
33345d01 3472 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 3473 if (off <= key.offset)
3474 new_key.offset = key.offset + destoff - off;
3475 else
3476 new_key.offset = destoff;
31840ae1 3477
f82a9901
FM		 3478 /*
3479 * Deal with a hole that doesn't have an extent item
3480 * that represents it (NO_HOLES feature enabled).
3481 * This hole is either in the middle of the cloning
3482 * range or at the beginning (fully overlaps it or
3483 * partially overlaps it).
3484 */
3485 if (new_key.offset != last_dest_end)
3486 drop_start = last_dest_end;
3487 else
3488 drop_start = new_key.offset;
3489
b6f3409b
SW		 3490 /*
3491 * 1 - adjusting old extent (we may have to split it)
3492 * 1 - add new extent
3493 * 1 - inode update
3494 */
3495 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3496 if (IS_ERR(trans)) {
3497 ret = PTR_ERR(trans);
3498 goto out;
3499 }
3500
c8a894d7
CM		 3501 if (type == BTRFS_FILE_EXTENT_REG ||
3502 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3503 /*
3504 * a | --- range to clone ---| b
3505 * | ------------- extent ------------- |
3506 */
3507
93915584 3508 /* subtract range b */
d72c0842
LZ		 3509 if (key.offset + datal > off + len)
3510 datal = off + len - key.offset;
3511
93915584 3512 /* subtract range a */
a22285a6
YZ		 3513 if (off > key.offset) {
3514 datao += off - key.offset;
3515 datal -= off - key.offset;
3516 }
3517
5dc562c5 3518 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3519 drop_start,
a22285a6 3520 new_key.offset + datal,
2671485d 3521 1);
79787eaa 3522 if (ret) {
3f9e3df8 3523 if (ret != -EOPNOTSUPP)
00fdf13a
LB		 3524 btrfs_abort_transaction(trans,
3525 root, ret);
79787eaa
JM		 3526 btrfs_end_transaction(trans, root);
3527 goto out;
3528 }
a22285a6 3529
c5c9cd4d
SW		 3530 ret = btrfs_insert_empty_item(trans, root, path,
3531 &new_key, size);
79787eaa
JM		 3532 if (ret) {
3533 btrfs_abort_transaction(trans, root,
3534 ret);
3535 btrfs_end_transaction(trans, root);
3536 goto out;
3537 }
c5c9cd4d
SW		 3538
3539 leaf = path->nodes[0];
3540 slot = path->slots[0];
3541 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3542 btrfs_item_ptr_offset(leaf, slot),
3543 size);
ae01a0ab 3544
c5c9cd4d 3545 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3546 struct btrfs_file_extent_item);
c5c9cd4d 3547
c5c9cd4d
SW		 3548 /* disko == 0 means it's a hole */
3549 if (!disko)
3550 datao = 0;
c5c9cd4d
SW		 3551
3552 btrfs_set_file_extent_offset(leaf, extent,
3553 datao);
3554 btrfs_set_file_extent_num_bytes(leaf, extent,
3555 datal);
fcebe456
JB		 3556
3557 /*
3558 * We need to look up the roots that point at
3559 * this bytenr and see if the new root does. If
3560 * it does not we need to make sure we update
3561 * quotas appropriately.
3562 */
3563 if (disko && root != BTRFS_I(src)->root &&
3564 disko != last_disko) {
3565 no_quota = check_ref(trans, root,
3566 disko);
3567 if (no_quota < 0) {
3568 btrfs_abort_transaction(trans,
3569 root,
3570 ret);
3571 btrfs_end_transaction(trans,
3572 root);
3573 ret = no_quota;
3574 goto out;
3575 }
3576 }
3577
c5c9cd4d
SW		 3578 if (disko) {
3579 inode_add_bytes(inode, datal);
ae01a0ab 3580 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3581 disko, diskl, 0,
3582 root->root_key.objectid,
33345d01 3583 btrfs_ino(inode),
66d7e7f0 3584 new_key.offset - datao,
fcebe456 3585 no_quota);
79787eaa
JM		 3586 if (ret) {
3587 btrfs_abort_transaction(trans,
3588 root,
3589 ret);
3590 btrfs_end_transaction(trans,
3591 root);
3592 goto out;
3593
3594 }
f46b5a66 3595 }
c5c9cd4d
SW		 3596 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3597 u64 skip = 0;
3598 u64 trim = 0;
d3ecfcdf
LB		 3599 u64 aligned_end = 0;
3600
ed958762
FM		 3601 /*
3602 * Don't copy an inline extent into an offset
3603 * greater than zero. Having an inline extent
3604 * at such an offset results in chaos as btrfs
3605 * isn't prepared for such cases. Just skip
3606 * this case for the same reasons as commented
3607 * at btrfs_ioctl_clone().
3608 */
3609 if (last_dest_end > 0) {
3610 ret = -EOPNOTSUPP;
3611 btrfs_end_transaction(trans, root);
3612 goto out;
3613 }
3614
c5c9cd4d
SW		 3615 if (off > key.offset) {
3616 skip = off - key.offset;
3617 new_key.offset += skip;
3618 }
d397712b 3619
aa42ffd9
LB		 3620 if (key.offset + datal > off + len)
3621 trim = key.offset + datal - (off + len);
d397712b 3622
c5c9cd4d 3623 if (comp && (skip || trim)) {
c5c9cd4d 3624 ret = -EINVAL;
a22285a6 3625 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3626 goto out;
3627 }
3628 size -= skip + trim;
3629 datal -= skip + trim;
a22285a6 3630
d3ecfcdf
LB		 3631 aligned_end = ALIGN(new_key.offset + datal,
3632 root->sectorsize);
5dc562c5 3633 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3634 drop_start,
d3ecfcdf 3635 aligned_end,
2671485d 3636 1);
79787eaa 3637 if (ret) {
3f9e3df8 3638 if (ret != -EOPNOTSUPP)
3a29bc09
CM		 3639 btrfs_abort_transaction(trans,
3640 root, ret);
79787eaa
JM		 3641 btrfs_end_transaction(trans, root);
3642 goto out;
3643 }
a22285a6 3644
c5c9cd4d
SW		 3645 ret = btrfs_insert_empty_item(trans, root, path,
3646 &new_key, size);
79787eaa
JM		 3647 if (ret) {
3648 btrfs_abort_transaction(trans, root,
3649 ret);
3650 btrfs_end_transaction(trans, root);
3651 goto out;
3652 }
c5c9cd4d
SW		 3653
3654 if (skip) {
d397712b
CM		 3655 u32 start =
3656 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3657 memmove(buf+start, buf+start+skip,
3658 datal);
3659 }
3660
3661 leaf = path->nodes[0];
3662 slot = path->slots[0];
3663 write_extent_buffer(leaf, buf,
3664 btrfs_item_ptr_offset(leaf, slot),
3665 size);
3666 inode_add_bytes(inode, datal);
f46b5a66 3667 }
c5c9cd4d 3668
7ffbb598
FM		 3669 /* If we have an implicit hole (NO_HOLES feature). */
3670 if (drop_start < new_key.offset)
3671 clone_update_extent_map(inode, trans,
14f59796 3672 NULL, drop_start,
7ffbb598
FM		 3673 new_key.offset - drop_start);
3674
14f59796 3675 clone_update_extent_map(inode, trans, path, 0, 0);
7ffbb598 3676
c5c9cd4d 3677 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3678 btrfs_release_path(path);
c5c9cd4d 3679
62e2390e
FM		 3680 last_dest_end = ALIGN(new_key.offset + datal,
3681 root->sectorsize);
f82a9901
FM		 3682 ret = clone_finish_inode_update(trans, inode,
3683 last_dest_end,
1c919a5e
MF		 3684 destoff, olen,
3685 no_time_update);
f82a9901 3686 if (ret)
79787eaa 3687 goto out;
2c463823
FM		 3688 if (new_key.offset + datal >= destoff + len)
3689 break;
a22285a6 3690 }
b3b4aa74 3691 btrfs_release_path(path);
df858e76 3692 key.offset = next_key_min_offset;
f46b5a66 3693 }
f46b5a66 3694 ret = 0;
32b7c687 3695
f82a9901
FM		 3696 if (last_dest_end < destoff + len) {
3697 /*
3698 * We have an implicit hole (NO_HOLES feature is enabled) that
3699 * fully or partially overlaps our cloning range at its end.
3700 */
3701 btrfs_release_path(path);
3702
3703 /*
3704 * 1 - remove extent(s)
3705 * 1 - inode update
3706 */
3707 trans = btrfs_start_transaction(root, 2);
3708 if (IS_ERR(trans)) {
3709 ret = PTR_ERR(trans);
3710 goto out;
3711 }
3712 ret = btrfs_drop_extents(trans, root, inode,
3713 last_dest_end, destoff + len, 1);
3714 if (ret) {
3715 if (ret != -EOPNOTSUPP)
3716 btrfs_abort_transaction(trans, root, ret);
3717 btrfs_end_transaction(trans, root);
3718 goto out;
3719 }
14f59796
FM		 3720 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3721 destoff + len - last_dest_end);
f82a9901 3722 ret = clone_finish_inode_update(trans, inode, destoff + len,
1c919a5e 3723 destoff, olen, no_time_update);
f82a9901
FM		 3724 }
3725
f46b5a66 3726out:
32b7c687
MF		 3727 btrfs_free_path(path);
3728 vfree(buf);
3729 return ret;
3730}
3731
3732static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3733 u64 off, u64 olen, u64 destoff)
3734{
54563d41 3735 struct inode *inode = file_inode(file);
32b7c687
MF		 3736 struct btrfs_root *root = BTRFS_I(inode)->root;
3737 struct fd src_file;
3738 struct inode *src;
3739 int ret;
3740 u64 len = olen;
3741 u64 bs = root->fs_info->sb->s_blocksize;
3742 int same_inode = 0;
3743
3744 /*
3745 * TODO:
3746 * - split compressed inline extents. annoying: we need to
3747 * decompress into destination's address_space (the file offset
3748 * may change, so source mapping won't do), then recompress (or
3749 * otherwise reinsert) a subrange.
00fdf13a
LB		 3750 *
3751 * - split destination inode's inline extents. The inline extents can
3752 * be either compressed or non-compressed.
32b7c687
MF		 3753 */
3754
3755 /* the destination must be opened for writing */
3756 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3757 return -EINVAL;
3758
3759 if (btrfs_root_readonly(root))
3760 return -EROFS;
3761
3762 ret = mnt_want_write_file(file);
3763 if (ret)
3764 return ret;
3765
3766 src_file = fdget(srcfd);
3767 if (!src_file.file) {
3768 ret = -EBADF;
3769 goto out_drop_write;
3770 }
3771
3772 ret = -EXDEV;
3773 if (src_file.file->f_path.mnt != file->f_path.mnt)
3774 goto out_fput;
3775
3776 src = file_inode(src_file.file);
3777
3778 ret = -EINVAL;
3779 if (src == inode)
3780 same_inode = 1;
3781
3782 /* the src must be open for reading */
3783 if (!(src_file.file->f_mode & FMODE_READ))
3784 goto out_fput;
3785
3786 /* don't make the dst file partly checksummed */
3787 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3788 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3789 goto out_fput;
3790
3791 ret = -EISDIR;
3792 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3793 goto out_fput;
3794
3795 ret = -EXDEV;
3796 if (src->i_sb != inode->i_sb)
3797 goto out_fput;
3798
3799 if (!same_inode) {
3800 if (inode < src) {
3801 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3802 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3803 } else {
3804 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3805 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3806 }
3807 } else {
3808 mutex_lock(&src->i_mutex);
3809 }
3810
3811 /* determine range to clone */
3812 ret = -EINVAL;
3813 if (off + len > src->i_size || off + len < off)
3814 goto out_unlock;
3815 if (len == 0)
3816 olen = len = src->i_size - off;
3817 /* if we extend to eof, continue to block boundary */
3818 if (off + len == src->i_size)
3819 len = ALIGN(src->i_size, bs) - off;
3820
ccccf3d6
FM		 3821 if (len == 0) {
3822 ret = 0;
3823 goto out_unlock;
3824 }
3825
32b7c687
MF		 3826 /* verify the end result is block aligned */
3827 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3828 !IS_ALIGNED(destoff, bs))
3829 goto out_unlock;
3830
3831 /* verify if ranges are overlapped within the same file */
3832 if (same_inode) {
3833 if (destoff + len > off && destoff < off + len)
3834 goto out_unlock;
3835 }
3836
3837 if (destoff > inode->i_size) {
3838 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3839 if (ret)
3840 goto out_unlock;
3841 }
3842
c125b8bf
FM		 3843 /*
3844 * Lock the target range too. Right after we replace the file extent
3845 * items in the fs tree (which now point to the cloned data), we might
3846 * have a worker replace them with extent items relative to a write
3847 * operation that was issued before this clone operation (i.e. confront
3848 * with inode.c:btrfs_finish_ordered_io).
3849 */
3850 if (same_inode) {
3851 u64 lock_start = min_t(u64, off, destoff);
3852 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
32b7c687 3853
c125b8bf
FM		 3854 lock_extent_range(src, lock_start, lock_len);
3855 } else {
3856 lock_extent_range(src, off, len);
3857 lock_extent_range(inode, destoff, len);
3858 }
32b7c687 3859
1c919a5e 3860 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
32b7c687 3861
c125b8bf
FM		 3862 if (same_inode) {
3863 u64 lock_start = min_t(u64, off, destoff);
3864 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3865
3866 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3867 } else {
3868 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3869 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3870 destoff + len - 1);
3871 }
3872 /*
3873 * Truncate page cache pages so that future reads will see the cloned
3874 * data immediately and not the previous data.
3875 */
3876 truncate_inode_pages_range(&inode->i_data, destoff,
3877 PAGE_CACHE_ALIGN(destoff + len) - 1);
f46b5a66 3878out_unlock:
c57c2b3e
FDBM		 3879 if (!same_inode) {
3880 if (inode < src) {
3881 mutex_unlock(&src->i_mutex);
3882 mutex_unlock(&inode->i_mutex);
3883 } else {
3884 mutex_unlock(&inode->i_mutex);
3885 mutex_unlock(&src->i_mutex);
3886 }
3887 } else {
3888 mutex_unlock(&src->i_mutex);
3889 }
f46b5a66 3890out_fput:
2903ff01 3891 fdput(src_file);
ab67b7c1 3892out_drop_write:
2a79f17e 3893 mnt_drop_write_file(file);
f46b5a66
CH		 3894 return ret;
3895}
3896
7a865e8a 3897static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3898{
3899 struct btrfs_ioctl_clone_range_args args;
3900
7a865e8a 3901 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3902 return -EFAULT;
3903 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3904 args.src_length, args.dest_offset);
3905}
3906
f46b5a66
CH		 3907/*
3908 * there are many ways the trans_start and trans_end ioctls can lead
3909 * to deadlocks. They should only be used by applications that
3910 * basically own the machine, and have a very in depth understanding
3911 * of all the possible deadlocks and enospc problems.
3912 */
b2950863 3913static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3914{
496ad9aa 3915 struct inode *inode = file_inode(file);
f46b5a66
CH		 3916 struct btrfs_root *root = BTRFS_I(inode)->root;
3917 struct btrfs_trans_handle *trans;
1ab86aed 3918 int ret;
f46b5a66 3919
1ab86aed 3920 ret = -EPERM;
df5b5520 3921 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3922 goto out;
df5b5520 3923
1ab86aed
SW		 3924 ret = -EINPROGRESS;
3925 if (file->private_data)
f46b5a66 3926 goto out;
9ca9ee09 3927
b83cc969
LZ		 3928 ret = -EROFS;
3929 if (btrfs_root_readonly(root))
3930 goto out;
3931
a561be71 3932 ret = mnt_want_write_file(file);
c146afad
YZ		 3933 if (ret)
3934 goto out;
3935
a4abeea4 3936 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3937
1ab86aed 3938 ret = -ENOMEM;
7a7eaa40 3939 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3940 if (IS_ERR(trans))
1ab86aed
SW		 3941 goto out_drop;
3942
3943 file->private_data = trans;
3944 return 0;
3945
3946out_drop:
a4abeea4 3947 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3948 mnt_drop_write_file(file);
f46b5a66 3949out:
f46b5a66
CH		 3950 return ret;
3951}
3952
6ef5ed0d
JB		 3953static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3954{
496ad9aa 3955 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3956 struct btrfs_root *root = BTRFS_I(inode)->root;
3957 struct btrfs_root *new_root;
3958 struct btrfs_dir_item *di;
3959 struct btrfs_trans_handle *trans;
3960 struct btrfs_path *path;
3961 struct btrfs_key location;
3962 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3963 u64 objectid = 0;
3964 u64 dir_id;
3c04ce01 3965 int ret;
6ef5ed0d
JB		 3966
3967 if (!capable(CAP_SYS_ADMIN))
3968 return -EPERM;
3969
3c04ce01
MX		 3970 ret = mnt_want_write_file(file);
3971 if (ret)
3972 return ret;
3973
3974 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3975 ret = -EFAULT;
3976 goto out;
3977 }
6ef5ed0d
JB		 3978
3979 if (!objectid)
1cecf579 3980 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3981
3982 location.objectid = objectid;
3983 location.type = BTRFS_ROOT_ITEM_KEY;
3984 location.offset = (u64)-1;
3985
3986 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3987 if (IS_ERR(new_root)) {
3988 ret = PTR_ERR(new_root);
3989 goto out;
3990 }
6ef5ed0d 3991
6ef5ed0d 3992 path = btrfs_alloc_path();
3c04ce01
MX		 3993 if (!path) {
3994 ret = -ENOMEM;
3995 goto out;
3996 }
6ef5ed0d
JB		 3997 path->leave_spinning = 1;
3998
3999 trans = btrfs_start_transaction(root, 1);
98d5dc13 4000 if (IS_ERR(trans)) {
6ef5ed0d 4001 btrfs_free_path(path);
3c04ce01
MX		 4002 ret = PTR_ERR(trans);
4003 goto out;
6ef5ed0d
JB		 4004 }
4005
6c41761f 4006 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 4007 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4008 dir_id, "default", 7, 1);
cf1e99a4 4009 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 4010 btrfs_free_path(path);
4011 btrfs_end_transaction(trans, root);
efe120a0
FH		 4012 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
4013 "item, this isn't going to work");
3c04ce01
MX		 4014 ret = -ENOENT;
4015 goto out;
6ef5ed0d
JB		 4016 }
4017
4018 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4019 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4020 btrfs_mark_buffer_dirty(path->nodes[0]);
4021 btrfs_free_path(path);
4022
2b0ce2c2 4023 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 4024 btrfs_end_transaction(trans, root);
3c04ce01
MX		 4025out:
4026 mnt_drop_write_file(file);
4027 return ret;
6ef5ed0d
JB		 4028}
4029
5af3e8cc
SB		 4030void btrfs_get_block_group_info(struct list_head *groups_list,
4031 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 4032{
4033 struct btrfs_block_group_cache *block_group;
4034
4035 space->total_bytes = 0;
4036 space->used_bytes = 0;
4037 space->flags = 0;
4038 list_for_each_entry(block_group, groups_list, list) {
4039 space->flags = block_group->flags;
4040 space->total_bytes += block_group->key.offset;
4041 space->used_bytes +=
4042 btrfs_block_group_used(&block_group->item);
4043 }
4044}
4045
48a3b636 4046static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 4047{
4048 struct btrfs_ioctl_space_args space_args;
4049 struct btrfs_ioctl_space_info space;
4050 struct btrfs_ioctl_space_info *dest;
7fde62bf 4051 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 4052 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 4053 struct btrfs_space_info *info;
bf5fc093
JB		 4054 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4055 BTRFS_BLOCK_GROUP_SYSTEM,
4056 BTRFS_BLOCK_GROUP_METADATA,
4057 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4058 int num_types = 4;
7fde62bf 4059 int alloc_size;
1406e432 4060 int ret = 0;
51788b1b 4061 u64 slot_count = 0;
bf5fc093 4062 int i, c;
1406e432
JB		 4063
4064 if (copy_from_user(&space_args,
4065 (struct btrfs_ioctl_space_args __user *)arg,
4066 sizeof(space_args)))
4067 return -EFAULT;
4068
bf5fc093
JB		 4069 for (i = 0; i < num_types; i++) {
4070 struct btrfs_space_info *tmp;
4071
4072 info = NULL;
4073 rcu_read_lock();
4074 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4075 list) {
4076 if (tmp->flags == types[i]) {
4077 info = tmp;
4078 break;
4079 }
4080 }
4081 rcu_read_unlock();
4082
4083 if (!info)
4084 continue;
4085
4086 down_read(&info->groups_sem);
4087 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4088 if (!list_empty(&info->block_groups[c]))
4089 slot_count++;
4090 }
4091 up_read(&info->groups_sem);
4092 }
7fde62bf 4093
36523e95
DS		 4094 /*
4095 * Global block reserve, exported as a space_info
4096 */
4097 slot_count++;
4098
7fde62bf
CM		 4099 /* space_slots == 0 means they are asking for a count */
4100 if (space_args.space_slots == 0) {
4101 space_args.total_spaces = slot_count;
4102 goto out;
4103 }
bf5fc093 4104
51788b1b 4105 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 4106
7fde62bf 4107 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 4108
7fde62bf
CM		 4109 /* we generally have at most 6 or so space infos, one for each raid
4110 * level. So, a whole page should be more than enough for everyone
4111 */
4112 if (alloc_size > PAGE_CACHE_SIZE)
4113 return -ENOMEM;
4114
1406e432 4115 space_args.total_spaces = 0;
7fde62bf
CM		 4116 dest = kmalloc(alloc_size, GFP_NOFS);
4117 if (!dest)
4118 return -ENOMEM;
4119 dest_orig = dest;
1406e432 4120
7fde62bf 4121 /* now we have a buffer to copy into */
bf5fc093
JB		 4122 for (i = 0; i < num_types; i++) {
4123 struct btrfs_space_info *tmp;
4124
51788b1b
DR		 4125 if (!slot_count)
4126 break;
4127
bf5fc093
JB		 4128 info = NULL;
4129 rcu_read_lock();
4130 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4131 list) {
4132 if (tmp->flags == types[i]) {
4133 info = tmp;
4134 break;
4135 }
4136 }
4137 rcu_read_unlock();
7fde62bf 4138
bf5fc093
JB		 4139 if (!info)
4140 continue;
4141 down_read(&info->groups_sem);
4142 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4143 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 4144 btrfs_get_block_group_info(
4145 &info->block_groups[c], &space);
bf5fc093
JB		 4146 memcpy(dest, &space, sizeof(space));
4147 dest++;
4148 space_args.total_spaces++;
51788b1b 4149 slot_count--;
bf5fc093 4150 }
51788b1b
DR		 4151 if (!slot_count)
4152 break;
bf5fc093
JB		 4153 }
4154 up_read(&info->groups_sem);
1406e432 4155 }
1406e432 4156
36523e95
DS		 4157 /*
4158 * Add global block reserve
4159 */
4160 if (slot_count) {
4161 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4162
4163 spin_lock(&block_rsv->lock);
4164 space.total_bytes = block_rsv->size;
4165 space.used_bytes = block_rsv->size - block_rsv->reserved;
4166 spin_unlock(&block_rsv->lock);
4167 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4168 memcpy(dest, &space, sizeof(space));
4169 space_args.total_spaces++;
4170 }
4171
2eec6c81 4172 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 4173 (arg + sizeof(struct btrfs_ioctl_space_args));
4174
4175 if (copy_to_user(user_dest, dest_orig, alloc_size))
4176 ret = -EFAULT;
4177
4178 kfree(dest_orig);
4179out:
4180 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 4181 ret = -EFAULT;
4182
4183 return ret;
4184}
4185
f46b5a66
CH		 4186/*
4187 * there are many ways the trans_start and trans_end ioctls can lead
4188 * to deadlocks. They should only be used by applications that
4189 * basically own the machine, and have a very in depth understanding
4190 * of all the possible deadlocks and enospc problems.
4191 */
4192long btrfs_ioctl_trans_end(struct file *file)
4193{
496ad9aa 4194 struct inode *inode = file_inode(file);
f46b5a66
CH		 4195 struct btrfs_root *root = BTRFS_I(inode)->root;
4196 struct btrfs_trans_handle *trans;
f46b5a66 4197
f46b5a66 4198 trans = file->private_data;
1ab86aed
SW		 4199 if (!trans)
4200 return -EINVAL;
b214107e 4201 file->private_data = NULL;
9ca9ee09 4202
1ab86aed
SW		 4203 btrfs_end_transaction(trans, root);
4204
a4abeea4 4205 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 4206
2a79f17e 4207 mnt_drop_write_file(file);
1ab86aed 4208 return 0;
f46b5a66
CH		 4209}
4210
9a8c28be
MX		 4211static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4212 void __user *argp)
46204592 4213{
46204592
SW		 4214 struct btrfs_trans_handle *trans;
4215 u64 transid;
db5b493a 4216 int ret;
46204592 4217
d4edf39b 4218 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4219 if (IS_ERR(trans)) {
4220 if (PTR_ERR(trans) != -ENOENT)
4221 return PTR_ERR(trans);
4222
4223 /* No running transaction, don't bother */
4224 transid = root->fs_info->last_trans_committed;
4225 goto out;
4226 }
46204592 4227 transid = trans->transid;
db5b493a 4228 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 4229 if (ret) {
4230 btrfs_end_transaction(trans, root);
db5b493a 4231 return ret;
8b2b2d3c 4232 }
ff7c1d33 4233out:
46204592
SW		 4234 if (argp)
4235 if (copy_to_user(argp, &transid, sizeof(transid)))
4236 return -EFAULT;
4237 return 0;
4238}
4239
9a8c28be
MX		 4240static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4241 void __user *argp)
46204592 4242{
46204592
SW		 4243 u64 transid;
4244
4245 if (argp) {
4246 if (copy_from_user(&transid, argp, sizeof(transid)))
4247 return -EFAULT;
4248 } else {
4249 transid = 0; /* current trans */
4250 }
4251 return btrfs_wait_for_commit(root, transid);
4252}
4253
b8e95489 4254static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4255{
496ad9aa 4256 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 4257 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4258 int ret;
475f6387
JS		 4259
4260 if (!capable(CAP_SYS_ADMIN))
4261 return -EPERM;
4262
4263 sa = memdup_user(arg, sizeof(*sa));
4264 if (IS_ERR(sa))
4265 return PTR_ERR(sa);
4266
b8e95489
MX		 4267 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4268 ret = mnt_want_write_file(file);
4269 if (ret)
4270 goto out;
4271 }
4272
aa1b8cd4 4273 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4274 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4275 0);
475f6387
JS		 4276
4277 if (copy_to_user(arg, sa, sizeof(*sa)))
4278 ret = -EFAULT;
4279
b8e95489
MX		 4280 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4281 mnt_drop_write_file(file);
4282out:
475f6387
JS		 4283 kfree(sa);
4284 return ret;
4285}
4286
4287static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4288{
4289 if (!capable(CAP_SYS_ADMIN))
4290 return -EPERM;
4291
aa1b8cd4 4292 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 4293}
4294
4295static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4296 void __user *arg)
4297{
4298 struct btrfs_ioctl_scrub_args *sa;
4299 int ret;
4300
4301 if (!capable(CAP_SYS_ADMIN))
4302 return -EPERM;
4303
4304 sa = memdup_user(arg, sizeof(*sa));
4305 if (IS_ERR(sa))
4306 return PTR_ERR(sa);
4307
4308 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4309
4310 if (copy_to_user(arg, sa, sizeof(*sa)))
4311 ret = -EFAULT;
4312
4313 kfree(sa);
4314 return ret;
4315}
4316
c11d2c23 4317static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 4318 void __user *arg)
c11d2c23
SB		 4319{
4320 struct btrfs_ioctl_get_dev_stats *sa;
4321 int ret;
4322
c11d2c23
SB		 4323 sa = memdup_user(arg, sizeof(*sa));
4324 if (IS_ERR(sa))
4325 return PTR_ERR(sa);
4326
b27f7c0c
DS		 4327 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4328 kfree(sa);
4329 return -EPERM;
4330 }
4331
4332 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 4333
4334 if (copy_to_user(arg, sa, sizeof(*sa)))
4335 ret = -EFAULT;
4336
4337 kfree(sa);
4338 return ret;
4339}
4340
3f6bcfbd
SB		 4341static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4342{
4343 struct btrfs_ioctl_dev_replace_args *p;
4344 int ret;
4345
4346 if (!capable(CAP_SYS_ADMIN))
4347 return -EPERM;
4348
4349 p = memdup_user(arg, sizeof(*p));
4350 if (IS_ERR(p))
4351 return PTR_ERR(p);
4352
4353 switch (p->cmd) {
4354 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
adfa97cb
ID		 4355 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4356 ret = -EROFS;
4357 goto out;
4358 }
3f6bcfbd
SB		 4359 if (atomic_xchg(
4360 &root->fs_info->mutually_exclusive_operation_running,
4361 1)) {
e57138b3 4362 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd
SB		 4363 } else {
4364 ret = btrfs_dev_replace_start(root, p);
4365 atomic_set(
4366 &root->fs_info->mutually_exclusive_operation_running,
4367 0);
4368 }
4369 break;
4370 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4371 btrfs_dev_replace_status(root->fs_info, p);
4372 ret = 0;
4373 break;
4374 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4375 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4376 break;
4377 default:
4378 ret = -EINVAL;
4379 break;
4380 }
4381
4382 if (copy_to_user(arg, p, sizeof(*p)))
4383 ret = -EFAULT;
adfa97cb 4384out:
3f6bcfbd
SB		 4385 kfree(p);
4386 return ret;
4387}
4388
d7728c96
JS		 4389static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4390{
4391 int ret = 0;
4392 int i;
740c3d22 4393 u64 rel_ptr;
d7728c96 4394 int size;
806468f8 4395 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4396 struct inode_fs_paths *ipath = NULL;
4397 struct btrfs_path *path;
4398
82b22ac8 4399 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4400 return -EPERM;
4401
4402 path = btrfs_alloc_path();
4403 if (!path) {
4404 ret = -ENOMEM;
4405 goto out;
4406 }
4407
4408 ipa = memdup_user(arg, sizeof(*ipa));
4409 if (IS_ERR(ipa)) {
4410 ret = PTR_ERR(ipa);
4411 ipa = NULL;
4412 goto out;
4413 }
4414
4415 size = min_t(u32, ipa->size, 4096);
4416 ipath = init_ipath(size, root, path);
4417 if (IS_ERR(ipath)) {
4418 ret = PTR_ERR(ipath);
4419 ipath = NULL;
4420 goto out;
4421 }
4422
4423 ret = paths_from_inode(ipa->inum, ipath);
4424 if (ret < 0)
4425 goto out;
4426
4427 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4428 rel_ptr = ipath->fspath->val[i] -
4429 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4430 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4431 }
4432
745c4d8e
JM		 4433 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4434 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 4435 if (ret) {
4436 ret = -EFAULT;
4437 goto out;
4438 }
4439
4440out:
4441 btrfs_free_path(path);
4442 free_ipath(ipath);
4443 kfree(ipa);
4444
4445 return ret;
4446}
4447
4448static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4449{
4450 struct btrfs_data_container *inodes = ctx;
4451 const size_t c = 3 * sizeof(u64);
4452
4453 if (inodes->bytes_left >= c) {
4454 inodes->bytes_left -= c;
4455 inodes->val[inodes->elem_cnt] = inum;
4456 inodes->val[inodes->elem_cnt + 1] = offset;
4457 inodes->val[inodes->elem_cnt + 2] = root;
4458 inodes->elem_cnt += 3;
4459 } else {
4460 inodes->bytes_missing += c - inodes->bytes_left;
4461 inodes->bytes_left = 0;
4462 inodes->elem_missed += 3;
4463 }
4464
4465 return 0;
4466}
4467
4468static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4469 void __user *arg)
4470{
4471 int ret = 0;
4472 int size;
d7728c96
JS		 4473 struct btrfs_ioctl_logical_ino_args *loi;
4474 struct btrfs_data_container *inodes = NULL;
4475 struct btrfs_path *path = NULL;
d7728c96
JS		 4476
4477 if (!capable(CAP_SYS_ADMIN))
4478 return -EPERM;
4479
4480 loi = memdup_user(arg, sizeof(*loi));
4481 if (IS_ERR(loi)) {
4482 ret = PTR_ERR(loi);
4483 loi = NULL;
4484 goto out;
4485 }
4486
4487 path = btrfs_alloc_path();
4488 if (!path) {
4489 ret = -ENOMEM;
4490 goto out;
4491 }
4492
425d17a2 4493 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 4494 inodes = init_data_container(size);
4495 if (IS_ERR(inodes)) {
4496 ret = PTR_ERR(inodes);
4497 inodes = NULL;
4498 goto out;
4499 }
4500
df031f07
LB		 4501 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4502 build_ino_list, inodes);
4503 if (ret == -EINVAL)
d7728c96
JS		 4504 ret = -ENOENT;
4505 if (ret < 0)
4506 goto out;
4507
745c4d8e
JM		 4508 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4509 (void *)(unsigned long)inodes, size);
d7728c96
JS		 4510 if (ret)
4511 ret = -EFAULT;
4512
4513out:
4514 btrfs_free_path(path);
425d17a2 4515 vfree(inodes);
d7728c96
JS		 4516 kfree(loi);
4517
4518 return ret;
4519}
4520
19a39dce 4521void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 4522 struct btrfs_ioctl_balance_args *bargs)
4523{
4524 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4525
4526 bargs->flags = bctl->flags;
4527
837d5b6e
ID		 4528 if (atomic_read(&fs_info->balance_running))
4529 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4530 if (atomic_read(&fs_info->balance_pause_req))
4531 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4532 if (atomic_read(&fs_info->balance_cancel_req))
4533 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4534
c9e9f97b
ID		 4535 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4536 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4537 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 4538
4539 if (lock) {
4540 spin_lock(&fs_info->balance_lock);
4541 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4542 spin_unlock(&fs_info->balance_lock);
4543 } else {
4544 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4545 }
c9e9f97b
ID		 4546}
4547
9ba1f6e4 4548static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4549{
496ad9aa 4550 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4551 struct btrfs_fs_info *fs_info = root->fs_info;
4552 struct btrfs_ioctl_balance_args *bargs;
4553 struct btrfs_balance_control *bctl;
ed0fb78f 4554 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4555 int ret;
4556
4557 if (!capable(CAP_SYS_ADMIN))
4558 return -EPERM;
4559
e54bfa31 4560 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4561 if (ret)
4562 return ret;
4563
ed0fb78f
ID		 4564again:
4565 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4566 mutex_lock(&fs_info->volume_mutex);
4567 mutex_lock(&fs_info->balance_mutex);
4568 need_unlock = true;
4569 goto locked;
4570 }
4571
4572 /*
4573 * mut. excl. ops lock is locked. Three possibilites:
4574 * (1) some other op is running
4575 * (2) balance is running
4576 * (3) balance is paused -- special case (think resume)
4577 */
c9e9f97b 4578 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4579 if (fs_info->balance_ctl) {
4580 /* this is either (2) or (3) */
4581 if (!atomic_read(&fs_info->balance_running)) {
4582 mutex_unlock(&fs_info->balance_mutex);
4583 if (!mutex_trylock(&fs_info->volume_mutex))
4584 goto again;
4585 mutex_lock(&fs_info->balance_mutex);
4586
4587 if (fs_info->balance_ctl &&
4588 !atomic_read(&fs_info->balance_running)) {
4589 /* this is (3) */
4590 need_unlock = false;
4591 goto locked;
4592 }
4593
4594 mutex_unlock(&fs_info->balance_mutex);
4595 mutex_unlock(&fs_info->volume_mutex);
4596 goto again;
4597 } else {
4598 /* this is (2) */
4599 mutex_unlock(&fs_info->balance_mutex);
4600 ret = -EINPROGRESS;
4601 goto out;
4602 }
4603 } else {
4604 /* this is (1) */
4605 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4606 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4607 goto out;
4608 }
4609
4610locked:
4611 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 4612
4613 if (arg) {
4614 bargs = memdup_user(arg, sizeof(*bargs));
4615 if (IS_ERR(bargs)) {
4616 ret = PTR_ERR(bargs);
ed0fb78f 4617 goto out_unlock;
c9e9f97b 4618 }
de322263
ID		 4619
4620 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4621 if (!fs_info->balance_ctl) {
4622 ret = -ENOTCONN;
4623 goto out_bargs;
4624 }
4625
4626 bctl = fs_info->balance_ctl;
4627 spin_lock(&fs_info->balance_lock);
4628 bctl->flags |= BTRFS_BALANCE_RESUME;
4629 spin_unlock(&fs_info->balance_lock);
4630
4631 goto do_balance;
4632 }
c9e9f97b
ID		 4633 } else {
4634 bargs = NULL;
4635 }
4636
ed0fb78f 4637 if (fs_info->balance_ctl) {
837d5b6e
ID		 4638 ret = -EINPROGRESS;
4639 goto out_bargs;
4640 }
4641
c9e9f97b
ID		 4642 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4643 if (!bctl) {
4644 ret = -ENOMEM;
4645 goto out_bargs;
4646 }
4647
4648 bctl->fs_info = fs_info;
4649 if (arg) {
4650 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4651 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4652 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4653
4654 bctl->flags = bargs->flags;
f43ffb60
ID		 4655 } else {
4656 /* balance everything - no filters */
4657 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4658 }
4659
de322263 4660do_balance:
c9e9f97b 4661 /*
ed0fb78f
ID		 4662 * Ownership of bctl and mutually_exclusive_operation_running
4663 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4664 * or, if restriper was paused all the way until unmount, in
4665 * free_fs_info. mutually_exclusive_operation_running is
4666 * cleared in __cancel_balance.
c9e9f97b 4667 */
ed0fb78f
ID		 4668 need_unlock = false;
4669
4670 ret = btrfs_balance(bctl, bargs);
4671
c9e9f97b
ID		 4672 if (arg) {
4673 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4674 ret = -EFAULT;
4675 }
4676
4677out_bargs:
4678 kfree(bargs);
ed0fb78f 4679out_unlock:
c9e9f97b
ID		 4680 mutex_unlock(&fs_info->balance_mutex);
4681 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 4682 if (need_unlock)
4683 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4684out:
e54bfa31 4685 mnt_drop_write_file(file);
c9e9f97b
ID		 4686 return ret;
4687}
4688
837d5b6e
ID		 4689static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4690{
4691 if (!capable(CAP_SYS_ADMIN))
4692 return -EPERM;
4693
4694 switch (cmd) {
4695 case BTRFS_BALANCE_CTL_PAUSE:
4696 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4697 case BTRFS_BALANCE_CTL_CANCEL:
4698 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4699 }
4700
4701 return -EINVAL;
4702}
4703
19a39dce
ID		 4704static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4705 void __user *arg)
4706{
4707 struct btrfs_fs_info *fs_info = root->fs_info;
4708 struct btrfs_ioctl_balance_args *bargs;
4709 int ret = 0;
4710
4711 if (!capable(CAP_SYS_ADMIN))
4712 return -EPERM;
4713
4714 mutex_lock(&fs_info->balance_mutex);
4715 if (!fs_info->balance_ctl) {
4716 ret = -ENOTCONN;
4717 goto out;
4718 }
4719
4720 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4721 if (!bargs) {
4722 ret = -ENOMEM;
4723 goto out;
4724 }
4725
4726 update_ioctl_balance_args(fs_info, 1, bargs);
4727
4728 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4729 ret = -EFAULT;
4730
4731 kfree(bargs);
4732out:
4733 mutex_unlock(&fs_info->balance_mutex);
4734 return ret;
4735}
4736
905b0dda 4737static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4738{
496ad9aa 4739 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4740 struct btrfs_ioctl_quota_ctl_args *sa;
4741 struct btrfs_trans_handle *trans = NULL;
4742 int ret;
4743 int err;
4744
4745 if (!capable(CAP_SYS_ADMIN))
4746 return -EPERM;
4747
905b0dda
MX		 4748 ret = mnt_want_write_file(file);
4749 if (ret)
4750 return ret;
5d13a37b
AJ		 4751
4752 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4753 if (IS_ERR(sa)) {
4754 ret = PTR_ERR(sa);
4755 goto drop_write;
4756 }
5d13a37b 4757
7708f029 4758 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4759 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4760 if (IS_ERR(trans)) {
4761 ret = PTR_ERR(trans);
4762 goto out;
5d13a37b
AJ		 4763 }
4764
4765 switch (sa->cmd) {
4766 case BTRFS_QUOTA_CTL_ENABLE:
4767 ret = btrfs_quota_enable(trans, root->fs_info);
4768 break;
4769 case BTRFS_QUOTA_CTL_DISABLE:
4770 ret = btrfs_quota_disable(trans, root->fs_info);
4771 break;
5d13a37b
AJ		 4772 default:
4773 ret = -EINVAL;
4774 break;
4775 }
4776
2f232036
JS		 4777 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4778 if (err && !ret)
4779 ret = err;
5d13a37b
AJ		 4780out:
4781 kfree(sa);
7708f029 4782 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4783drop_write:
4784 mnt_drop_write_file(file);
5d13a37b
AJ		 4785 return ret;
4786}
4787
905b0dda 4788static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4789{
496ad9aa 4790 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4791 struct btrfs_ioctl_qgroup_assign_args *sa;
4792 struct btrfs_trans_handle *trans;
4793 int ret;
4794 int err;
4795
4796 if (!capable(CAP_SYS_ADMIN))
4797 return -EPERM;
4798
905b0dda
MX		 4799 ret = mnt_want_write_file(file);
4800 if (ret)
4801 return ret;
5d13a37b
AJ		 4802
4803 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4804 if (IS_ERR(sa)) {
4805 ret = PTR_ERR(sa);
4806 goto drop_write;
4807 }
5d13a37b
AJ		 4808
4809 trans = btrfs_join_transaction(root);
4810 if (IS_ERR(trans)) {
4811 ret = PTR_ERR(trans);
4812 goto out;
4813 }
4814
4815 /* FIXME: check if the IDs really exist */
4816 if (sa->assign) {
4817 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4818 sa->src, sa->dst);
4819 } else {
4820 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4821 sa->src, sa->dst);
4822 }
4823
e082f563
QW		 4824 /* update qgroup status and info */
4825 err = btrfs_run_qgroups(trans, root->fs_info);
4826 if (err < 0)
4827 btrfs_error(root->fs_info, ret,
4828 "failed to update qgroup status and info\n");
5d13a37b
AJ		 4829 err = btrfs_end_transaction(trans, root);
4830 if (err && !ret)
4831 ret = err;
4832
4833out:
4834 kfree(sa);
905b0dda
MX		 4835drop_write:
4836 mnt_drop_write_file(file);
5d13a37b
AJ		 4837 return ret;
4838}
4839
905b0dda 4840static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4841{
496ad9aa 4842 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4843 struct btrfs_ioctl_qgroup_create_args *sa;
4844 struct btrfs_trans_handle *trans;
4845 int ret;
4846 int err;
4847
4848 if (!capable(CAP_SYS_ADMIN))
4849 return -EPERM;
4850
905b0dda
MX		 4851 ret = mnt_want_write_file(file);
4852 if (ret)
4853 return ret;
5d13a37b
AJ		 4854
4855 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4856 if (IS_ERR(sa)) {
4857 ret = PTR_ERR(sa);
4858 goto drop_write;
4859 }
5d13a37b 4860
d86e56cf
MX		 4861 if (!sa->qgroupid) {
4862 ret = -EINVAL;
4863 goto out;
4864 }
4865
5d13a37b
AJ		 4866 trans = btrfs_join_transaction(root);
4867 if (IS_ERR(trans)) {
4868 ret = PTR_ERR(trans);
4869 goto out;
4870 }
4871
4872 /* FIXME: check if the IDs really exist */
4873 if (sa->create) {
4087cf24 4874 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
5d13a37b
AJ		 4875 } else {
4876 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4877 }
4878
4879 err = btrfs_end_transaction(trans, root);
4880 if (err && !ret)
4881 ret = err;
4882
4883out:
4884 kfree(sa);
905b0dda
MX		 4885drop_write:
4886 mnt_drop_write_file(file);
5d13a37b
AJ		 4887 return ret;
4888}
4889
905b0dda 4890static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4891{
496ad9aa 4892 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4893 struct btrfs_ioctl_qgroup_limit_args *sa;
4894 struct btrfs_trans_handle *trans;
4895 int ret;
4896 int err;
4897 u64 qgroupid;
4898
4899 if (!capable(CAP_SYS_ADMIN))
4900 return -EPERM;
4901
905b0dda
MX		 4902 ret = mnt_want_write_file(file);
4903 if (ret)
4904 return ret;
5d13a37b
AJ		 4905
4906 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4907 if (IS_ERR(sa)) {
4908 ret = PTR_ERR(sa);
4909 goto drop_write;
4910 }
5d13a37b
AJ		 4911
4912 trans = btrfs_join_transaction(root);
4913 if (IS_ERR(trans)) {
4914 ret = PTR_ERR(trans);
4915 goto out;
4916 }
4917
4918 qgroupid = sa->qgroupid;
4919 if (!qgroupid) {
4920 /* take the current subvol as qgroup */
4921 qgroupid = root->root_key.objectid;
4922 }
4923
4924 /* FIXME: check if the IDs really exist */
4925 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4926
4927 err = btrfs_end_transaction(trans, root);
4928 if (err && !ret)
4929 ret = err;
4930
4931out:
4932 kfree(sa);
905b0dda
MX		 4933drop_write:
4934 mnt_drop_write_file(file);
5d13a37b
AJ		 4935 return ret;
4936}
4937
2f232036
JS		 4938static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4939{
6d0379ec 4940 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4941 struct btrfs_ioctl_quota_rescan_args *qsa;
4942 int ret;
4943
4944 if (!capable(CAP_SYS_ADMIN))
4945 return -EPERM;
4946
4947 ret = mnt_want_write_file(file);
4948 if (ret)
4949 return ret;
4950
4951 qsa = memdup_user(arg, sizeof(*qsa));
4952 if (IS_ERR(qsa)) {
4953 ret = PTR_ERR(qsa);
4954 goto drop_write;
4955 }
4956
4957 if (qsa->flags) {
4958 ret = -EINVAL;
4959 goto out;
4960 }
4961
4962 ret = btrfs_qgroup_rescan(root->fs_info);
4963
4964out:
4965 kfree(qsa);
4966drop_write:
4967 mnt_drop_write_file(file);
4968 return ret;
4969}
4970
4971static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4972{
6d0379ec 4973 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4974 struct btrfs_ioctl_quota_rescan_args *qsa;
4975 int ret = 0;
4976
4977 if (!capable(CAP_SYS_ADMIN))
4978 return -EPERM;
4979
4980 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4981 if (!qsa)
4982 return -ENOMEM;
4983
4984 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4985 qsa->flags = 1;
4986 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4987 }
4988
4989 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4990 ret = -EFAULT;
4991
4992 kfree(qsa);
4993 return ret;
4994}
4995
57254b6e
JS		 4996static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4997{
54563d41 4998 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
57254b6e
JS		 4999
5000 if (!capable(CAP_SYS_ADMIN))
5001 return -EPERM;
5002
5003 return btrfs_qgroup_wait_for_completion(root->fs_info);
5004}
5005
abccd00f
HM		 5006static long _btrfs_ioctl_set_received_subvol(struct file *file,
5007 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 5008{
496ad9aa 5009 struct inode *inode = file_inode(file);
8ea05e3a
AB		 5010 struct btrfs_root *root = BTRFS_I(inode)->root;
5011 struct btrfs_root_item *root_item = &root->root_item;
5012 struct btrfs_trans_handle *trans;
5013 struct timespec ct = CURRENT_TIME;
5014 int ret = 0;
dd5f9615 5015 int received_uuid_changed;
8ea05e3a 5016
bd60ea0f
DS		 5017 if (!inode_owner_or_capable(inode))
5018 return -EPERM;
5019
8ea05e3a
AB		 5020 ret = mnt_want_write_file(file);
5021 if (ret < 0)
5022 return ret;
5023
5024 down_write(&root->fs_info->subvol_sem);
5025
5026 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5027 ret = -EINVAL;
5028 goto out;
5029 }
5030
5031 if (btrfs_root_readonly(root)) {
5032 ret = -EROFS;
5033 goto out;
5034 }
5035
dd5f9615
SB		 5036 /*
5037 * 1 - root item
5038 * 2 - uuid items (received uuid + subvol uuid)
5039 */
5040 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 5041 if (IS_ERR(trans)) {
5042 ret = PTR_ERR(trans);
5043 trans = NULL;
5044 goto out;
5045 }
5046
5047 sa->rtransid = trans->transid;
5048 sa->rtime.sec = ct.tv_sec;
5049 sa->rtime.nsec = ct.tv_nsec;
5050
dd5f9615
SB		 5051 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5052 BTRFS_UUID_SIZE);
5053 if (received_uuid_changed &&
5054 !btrfs_is_empty_uuid(root_item->received_uuid))
5055 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5056 root_item->received_uuid,
5057 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5058 root->root_key.objectid);
8ea05e3a
AB		 5059 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5060 btrfs_set_root_stransid(root_item, sa->stransid);
5061 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 5062 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5063 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5064 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5065 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 5066
5067 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5068 &root->root_key, &root->root_item);
5069 if (ret < 0) {
5070 btrfs_end_transaction(trans, root);
8ea05e3a 5071 goto out;
dd5f9615
SB		 5072 }
5073 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5074 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5075 sa->uuid,
5076 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5077 root->root_key.objectid);
5078 if (ret < 0 && ret != -EEXIST) {
5079 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 5080 goto out;
dd5f9615
SB		 5081 }
5082 }
5083 ret = btrfs_commit_transaction(trans, root);
5084 if (ret < 0) {
5085 btrfs_abort_transaction(trans, root, ret);
5086 goto out;
8ea05e3a
AB		 5087 }
5088
abccd00f
HM		 5089out:
5090 up_write(&root->fs_info->subvol_sem);
5091 mnt_drop_write_file(file);
5092 return ret;
5093}
5094
5095#ifdef CONFIG_64BIT
5096static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5097 void __user *arg)
5098{
5099 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5100 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5101 int ret = 0;
5102
5103 args32 = memdup_user(arg, sizeof(*args32));
5104 if (IS_ERR(args32)) {
5105 ret = PTR_ERR(args32);
5106 args32 = NULL;
5107 goto out;
5108 }
5109
5110 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
84dbeb87
DC		 5111 if (!args64) {
5112 ret = -ENOMEM;
abccd00f
HM		 5113 goto out;
5114 }
5115
5116 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5117 args64->stransid = args32->stransid;
5118 args64->rtransid = args32->rtransid;
5119 args64->stime.sec = args32->stime.sec;
5120 args64->stime.nsec = args32->stime.nsec;
5121 args64->rtime.sec = args32->rtime.sec;
5122 args64->rtime.nsec = args32->rtime.nsec;
5123 args64->flags = args32->flags;
5124
5125 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5126 if (ret)
5127 goto out;
5128
5129 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5130 args32->stransid = args64->stransid;
5131 args32->rtransid = args64->rtransid;
5132 args32->stime.sec = args64->stime.sec;
5133 args32->stime.nsec = args64->stime.nsec;
5134 args32->rtime.sec = args64->rtime.sec;
5135 args32->rtime.nsec = args64->rtime.nsec;
5136 args32->flags = args64->flags;
5137
5138 ret = copy_to_user(arg, args32, sizeof(*args32));
5139 if (ret)
5140 ret = -EFAULT;
5141
5142out:
5143 kfree(args32);
5144 kfree(args64);
5145 return ret;
5146}
5147#endif
5148
5149static long btrfs_ioctl_set_received_subvol(struct file *file,
5150 void __user *arg)
5151{
5152 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5153 int ret = 0;
5154
5155 sa = memdup_user(arg, sizeof(*sa));
5156 if (IS_ERR(sa)) {
5157 ret = PTR_ERR(sa);
5158 sa = NULL;
5159 goto out;
5160 }
5161
5162 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5163
5164 if (ret)
5165 goto out;
5166
8ea05e3a
AB		 5167 ret = copy_to_user(arg, sa, sizeof(*sa));
5168 if (ret)
5169 ret = -EFAULT;
5170
5171out:
5172 kfree(sa);
8ea05e3a
AB		 5173 return ret;
5174}
5175
867ab667 5176static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5177{
6d0379ec 5178 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 5179 size_t len;
867ab667 5180 int ret;
a1b83ac5
AJ		 5181 char label[BTRFS_LABEL_SIZE];
5182
5183 spin_lock(&root->fs_info->super_lock);
5184 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5185 spin_unlock(&root->fs_info->super_lock);
5186
5187 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5188
5189 if (len == BTRFS_LABEL_SIZE) {
efe120a0
FH		 5190 btrfs_warn(root->fs_info,
5191 "label is too long, return the first %zu bytes", --len);
867ab667 5192 }
5193
867ab667 5194 ret = copy_to_user(arg, label, len);
867ab667 5195
5196 return ret ? -EFAULT : 0;
5197}
5198
a8bfd4ab 5199static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5200{
6d0379ec 5201 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 5202 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5203 struct btrfs_trans_handle *trans;
5204 char label[BTRFS_LABEL_SIZE];
5205 int ret;
5206
5207 if (!capable(CAP_SYS_ADMIN))
5208 return -EPERM;
5209
5210 if (copy_from_user(label, arg, sizeof(label)))
5211 return -EFAULT;
5212
5213 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
efe120a0 5214 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
a8bfd4ab 5215 BTRFS_LABEL_SIZE - 1);
5216 return -EINVAL;
5217 }
5218
5219 ret = mnt_want_write_file(file);
5220 if (ret)
5221 return ret;
5222
a8bfd4ab 5223 trans = btrfs_start_transaction(root, 0);
5224 if (IS_ERR(trans)) {
5225 ret = PTR_ERR(trans);
5226 goto out_unlock;
5227 }
5228
a1b83ac5 5229 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 5230 strcpy(super_block->label, label);
a1b83ac5 5231 spin_unlock(&root->fs_info->super_lock);
d0270aca 5232 ret = btrfs_commit_transaction(trans, root);
a8bfd4ab 5233
5234out_unlock:
a8bfd4ab 5235 mnt_drop_write_file(file);
5236 return ret;
5237}
5238
2eaa055f
JM		 5239#define INIT_FEATURE_FLAGS(suffix) \
5240 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5241 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5242 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5243
5244static int btrfs_ioctl_get_supported_features(struct file *file,
5245 void __user *arg)
5246{
5247 static struct btrfs_ioctl_feature_flags features[3] = {
5248 INIT_FEATURE_FLAGS(SUPP),
5249 INIT_FEATURE_FLAGS(SAFE_SET),
5250 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5251 };
5252
5253 if (copy_to_user(arg, &features, sizeof(features)))
5254 return -EFAULT;
5255
5256 return 0;
5257}
5258
5259static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5260{
5261 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5262 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5263 struct btrfs_ioctl_feature_flags features;
5264
5265 features.compat_flags = btrfs_super_compat_flags(super_block);
5266 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5267 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5268
5269 if (copy_to_user(arg, &features, sizeof(features)))
5270 return -EFAULT;
5271
5272 return 0;
5273}
5274
3b02a68a
JM		 5275static int check_feature_bits(struct btrfs_root *root,
5276 enum btrfs_feature_set set,
2eaa055f
JM		 5277 u64 change_mask, u64 flags, u64 supported_flags,
5278 u64 safe_set, u64 safe_clear)
5279{
3b02a68a
JM		 5280 const char *type = btrfs_feature_set_names[set];
5281 char *names;
2eaa055f
JM		 5282 u64 disallowed, unsupported;
5283 u64 set_mask = flags & change_mask;
5284 u64 clear_mask = ~flags & change_mask;
5285
5286 unsupported = set_mask & ~supported_flags;
5287 if (unsupported) {
3b02a68a
JM		 5288 names = btrfs_printable_features(set, unsupported);
5289 if (names) {
5290 btrfs_warn(root->fs_info,
5291 "this kernel does not support the %s feature bit%s",
5292 names, strchr(names, ',') ? "s" : "");
5293 kfree(names);
5294 } else
5295 btrfs_warn(root->fs_info,
2eaa055f
JM		 5296 "this kernel does not support %s bits 0x%llx",
5297 type, unsupported);
5298 return -EOPNOTSUPP;
5299 }
5300
5301 disallowed = set_mask & ~safe_set;
5302 if (disallowed) {
3b02a68a
JM		 5303 names = btrfs_printable_features(set, disallowed);
5304 if (names) {
5305 btrfs_warn(root->fs_info,
5306 "can't set the %s feature bit%s while mounted",
5307 names, strchr(names, ',') ? "s" : "");
5308 kfree(names);
5309 } else
5310 btrfs_warn(root->fs_info,
2eaa055f
JM		 5311 "can't set %s bits 0x%llx while mounted",
5312 type, disallowed);
5313 return -EPERM;
5314 }
5315
5316 disallowed = clear_mask & ~safe_clear;
5317 if (disallowed) {
3b02a68a
JM		 5318 names = btrfs_printable_features(set, disallowed);
5319 if (names) {
5320 btrfs_warn(root->fs_info,
5321 "can't clear the %s feature bit%s while mounted",
5322 names, strchr(names, ',') ? "s" : "");
5323 kfree(names);
5324 } else
5325 btrfs_warn(root->fs_info,
2eaa055f
JM		 5326 "can't clear %s bits 0x%llx while mounted",
5327 type, disallowed);
5328 return -EPERM;
5329 }
5330
5331 return 0;
5332}
5333
5334#define check_feature(root, change_mask, flags, mask_base) \
3b02a68a 5335check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5336 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5337 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5338 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5339
5340static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5341{
5342 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5343 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5344 struct btrfs_ioctl_feature_flags flags[2];
5345 struct btrfs_trans_handle *trans;
5346 u64 newflags;
5347 int ret;
5348
5349 if (!capable(CAP_SYS_ADMIN))
5350 return -EPERM;
5351
5352 if (copy_from_user(flags, arg, sizeof(flags)))
5353 return -EFAULT;
5354
5355 /* Nothing to do */
5356 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5357 !flags[0].incompat_flags)
5358 return 0;
5359
5360 ret = check_feature(root, flags[0].compat_flags,
5361 flags[1].compat_flags, COMPAT);
5362 if (ret)
5363 return ret;
5364
5365 ret = check_feature(root, flags[0].compat_ro_flags,
5366 flags[1].compat_ro_flags, COMPAT_RO);
5367 if (ret)
5368 return ret;
5369
5370 ret = check_feature(root, flags[0].incompat_flags,
5371 flags[1].incompat_flags, INCOMPAT);
5372 if (ret)
5373 return ret;
5374
8051aa1a 5375 trans = btrfs_start_transaction(root, 0);
2eaa055f
JM		 5376 if (IS_ERR(trans))
5377 return PTR_ERR(trans);
5378
5379 spin_lock(&root->fs_info->super_lock);
5380 newflags = btrfs_super_compat_flags(super_block);
5381 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5382 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5383 btrfs_set_super_compat_flags(super_block, newflags);
5384
5385 newflags = btrfs_super_compat_ro_flags(super_block);
5386 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5387 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5388 btrfs_set_super_compat_ro_flags(super_block, newflags);
5389
5390 newflags = btrfs_super_incompat_flags(super_block);
5391 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5392 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5393 btrfs_set_super_incompat_flags(super_block, newflags);
5394 spin_unlock(&root->fs_info->super_lock);
5395
d0270aca 5396 return btrfs_commit_transaction(trans, root);
2eaa055f
JM		 5397}
5398
f46b5a66
CH		 5399long btrfs_ioctl(struct file *file, unsigned int
5400 cmd, unsigned long arg)
5401{
496ad9aa 5402 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 5403 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5404
5405 switch (cmd) {
6cbff00f
CH		 5406 case FS_IOC_GETFLAGS:
5407 return btrfs_ioctl_getflags(file, argp);
5408 case FS_IOC_SETFLAGS:
5409 return btrfs_ioctl_setflags(file, argp);
5410 case FS_IOC_GETVERSION:
5411 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5412 case FITRIM:
5413 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5414 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5415 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5416 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5417 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5418 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5419 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5420 case BTRFS_IOC_SUBVOL_CREATE_V2:
5421 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5422 case BTRFS_IOC_SNAP_DESTROY:
5423 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5424 case BTRFS_IOC_SUBVOL_GETFLAGS:
5425 return btrfs_ioctl_subvol_getflags(file, argp);
5426 case BTRFS_IOC_SUBVOL_SETFLAGS:
5427 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5428 case BTRFS_IOC_DEFAULT_SUBVOL:
5429 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5430 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5431 return btrfs_ioctl_defrag(file, NULL);
5432 case BTRFS_IOC_DEFRAG_RANGE:
5433 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5434 case BTRFS_IOC_RESIZE:
198605a8 5435 return btrfs_ioctl_resize(file, argp);
f46b5a66 5436 case BTRFS_IOC_ADD_DEV:
4bcabaa3 5437 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 5438 case BTRFS_IOC_RM_DEV:
da24927b 5439 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 5440 case BTRFS_IOC_FS_INFO:
5441 return btrfs_ioctl_fs_info(root, argp);
5442 case BTRFS_IOC_DEV_INFO:
5443 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 5444 case BTRFS_IOC_BALANCE:
9ba1f6e4 5445 return btrfs_ioctl_balance(file, NULL);
f46b5a66 5446 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 5447 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5448 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 5449 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 5450 case BTRFS_IOC_TRANS_START:
5451 return btrfs_ioctl_trans_start(file);
5452 case BTRFS_IOC_TRANS_END:
5453 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 5454 case BTRFS_IOC_TREE_SEARCH:
5455 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5456 case BTRFS_IOC_TREE_SEARCH_V2:
5457 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5458 case BTRFS_IOC_INO_LOOKUP:
5459 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5460 case BTRFS_IOC_INO_PATHS:
5461 return btrfs_ioctl_ino_to_path(root, argp);
5462 case BTRFS_IOC_LOGICAL_INO:
5463 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 5464 case BTRFS_IOC_SPACE_INFO:
5465 return btrfs_ioctl_space_info(root, argp);
9b199859
FDBM		 5466 case BTRFS_IOC_SYNC: {
5467 int ret;
5468
6c255e67 5469 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
9b199859
FDBM		 5470 if (ret)
5471 return ret;
ddb52f4f 5472 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
2fad4e83
DS		 5473 /*
5474 * The transaction thread may want to do more work,
5475 * namely it pokes the cleaner ktread that will start
5476 * processing uncleaned subvols.
5477 */
5478 wake_up_process(root->fs_info->transaction_kthread);
9b199859
FDBM		 5479 return ret;
5480 }
46204592 5481 case BTRFS_IOC_START_SYNC:
9a8c28be 5482 return btrfs_ioctl_start_sync(root, argp);
46204592 5483 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 5484 return btrfs_ioctl_wait_sync(root, argp);
475f6387 5485 case BTRFS_IOC_SCRUB:
b8e95489 5486 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 5487 case BTRFS_IOC_SCRUB_CANCEL:
5488 return btrfs_ioctl_scrub_cancel(root, argp);
5489 case BTRFS_IOC_SCRUB_PROGRESS:
5490 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 5491 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5492 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 5493 case BTRFS_IOC_BALANCE_CTL:
5494 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 5495 case BTRFS_IOC_BALANCE_PROGRESS:
5496 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 5497 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5498 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5499#ifdef CONFIG_64BIT
5500 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5501 return btrfs_ioctl_set_received_subvol_32(file, argp);
5502#endif
31db9f7c
AB		 5503 case BTRFS_IOC_SEND:
5504 return btrfs_ioctl_send(file, argp);
c11d2c23 5505 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 5506 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 5507 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5508 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5509 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5510 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5511 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5512 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5513 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5514 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5515 case BTRFS_IOC_QUOTA_RESCAN:
5516 return btrfs_ioctl_quota_rescan(file, argp);
5517 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5518 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5519 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5520 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 5521 case BTRFS_IOC_DEV_REPLACE:
5522 return btrfs_ioctl_dev_replace(root, argp);
867ab667 5523 case BTRFS_IOC_GET_FSLABEL:
5524 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5525 case BTRFS_IOC_SET_FSLABEL:
5526 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 5527 case BTRFS_IOC_FILE_EXTENT_SAME:
5528 return btrfs_ioctl_file_extent_same(file, argp);
2eaa055f
JM		 5529 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5530 return btrfs_ioctl_get_supported_features(file, argp);
5531 case BTRFS_IOC_GET_FEATURES:
5532 return btrfs_ioctl_get_features(file, argp);
5533 case BTRFS_IOC_SET_FEATURES:
5534 return btrfs_ioctl_set_features(file, argp);
f46b5a66
CH		 5535 }
5536
5537 return -ENOTTY;
5538}
Ubuntu Hirsute Linux kernel mirror