]> git.ipfire.org Git - thirdparty/linux.git/blame - fs/btrfs/ioctl.c
projects / thirdparty / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: fix null pointer dereference in btrfs_show_devname when name is null
[thirdparty/linux.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
MF		 89static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
91
6cbff00f
CH		 92/* Mask out flags that are inappropriate for the given type of inode. */
93static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
94{
95 if (S_ISDIR(mode))
96 return flags;
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
99 else
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
101}
102
103/*
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105 */
106static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
107{
108 unsigned int iflags = 0;
109
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
124
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 129
130 return iflags;
131}
132
133/*
134 * Update inode->i_flags based on the btrfs internal flags.
135 */
136void btrfs_update_iflags(struct inode *inode)
137{
138 struct btrfs_inode *ip = BTRFS_I(inode);
3cc79392 139 unsigned int new_fl = 0;
6cbff00f
CH		 140
141 if (ip->flags & BTRFS_INODE_SYNC)
3cc79392 142 new_fl |= S_SYNC;
6cbff00f 143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
3cc79392 144 new_fl |= S_IMMUTABLE;
6cbff00f 145 if (ip->flags & BTRFS_INODE_APPEND)
3cc79392 146 new_fl |= S_APPEND;
6cbff00f 147 if (ip->flags & BTRFS_INODE_NOATIME)
3cc79392 148 new_fl |= S_NOATIME;
6cbff00f 149 if (ip->flags & BTRFS_INODE_DIRSYNC)
3cc79392
FM		 150 new_fl |= S_DIRSYNC;
151
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
154 new_fl);
6cbff00f
CH		 155}
156
157/*
158 * Inherit flags from the parent inode.
159 *
e27425d6 160 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 161 */
162void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
163{
0b4dcea5
CM		 164 unsigned int flags;
165
166 if (!dir)
167 return;
168
169 flags = BTRFS_I(dir)->flags;
6cbff00f 170
e27425d6
JB		 171 if (flags & BTRFS_INODE_NOCOMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 } else if (flags & BTRFS_INODE_COMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
177 }
178
213490b3 179 if (flags & BTRFS_INODE_NODATACOW) {
e27425d6 180 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
213490b3
LB		 181 if (S_ISREG(inode->i_mode))
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
183 }
6cbff00f 184
6cbff00f
CH		 185 btrfs_update_iflags(inode);
186}
187
188static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
189{
496ad9aa 190 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
6cbff00f
CH		 191 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
192
193 if (copy_to_user(arg, &flags, sizeof(flags)))
194 return -EFAULT;
195 return 0;
196}
197
75e7cb7f
LB		 198static int check_flags(unsigned int flags)
199{
200 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 FS_NOATIME_FL | FS_NODUMP_FL | \
202 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 203 FS_NOCOMP_FL | FS_COMPR_FL |
204 FS_NOCOW_FL))
75e7cb7f
LB		 205 return -EOPNOTSUPP;
206
207 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
208 return -EINVAL;
209
75e7cb7f
LB		 210 return 0;
211}
212
6cbff00f
CH		 213static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
214{
496ad9aa 215 struct inode *inode = file_inode(file);
6cbff00f
CH		 216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
220 int ret;
f062abf0
LZ		 221 u64 ip_oldflags;
222 unsigned int i_oldflags;
7e97b8da 223 umode_t mode;
6cbff00f 224
bd60ea0f
DS		 225 if (!inode_owner_or_capable(inode))
226 return -EPERM;
227
b83cc969
LZ		 228 if (btrfs_root_readonly(root))
229 return -EROFS;
230
6cbff00f
CH		 231 if (copy_from_user(&flags, arg, sizeof(flags)))
232 return -EFAULT;
233
75e7cb7f
LB		 234 ret = check_flags(flags);
235 if (ret)
236 return ret;
f46b5a66 237
e7848683
JK		 238 ret = mnt_want_write_file(file);
239 if (ret)
240 return ret;
241
6cbff00f
CH		 242 mutex_lock(&inode->i_mutex);
243
f062abf0
LZ		 244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
7e97b8da 246 mode = inode->i_mode;
f062abf0 247
6cbff00f
CH		 248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
252 ret = -EPERM;
253 goto out_unlock;
254 }
255 }
256
6cbff00f
CH		 257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
259 else
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
263 else
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
267 else
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
271 else
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
275 else
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
279 else
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
7e97b8da
DS		 281 if (flags & FS_NOCOW_FL) {
282 if (S_ISREG(mode)) {
283 /*
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
287 */
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
291 } else {
292 ip->flags |= BTRFS_INODE_NODATACOW;
293 }
294 } else {
295 /*
296 * Revert back under same assuptions as above
297 */
298 if (S_ISREG(mode)) {
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
302 } else {
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
304 }
305 }
6cbff00f 306
75e7cb7f
LB		 307 /*
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
310 * things smaller.
311 */
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 315
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
318 goto out_drop;
75e7cb7f 319 } else if (flags & FS_COMPR_FL) {
63541927
FDBM		 320 const char *comp;
321
75e7cb7f
LB		 322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 324
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
326 comp = "lzo";
327 else
328 comp = "zlib";
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
331 if (ret)
332 goto out_drop;
333
ebcb904d
LZ		 334 } else {
335 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 336 }
6cbff00f 337
4da6f1a3 338 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 339 if (IS_ERR(trans)) {
340 ret = PTR_ERR(trans);
341 goto out_drop;
342 }
6cbff00f 343
306424cc 344 btrfs_update_iflags(inode);
0c4d2d95 345 inode_inc_iversion(inode);
306424cc 346 inode->i_ctime = CURRENT_TIME;
6cbff00f 347 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 348
6cbff00f 349 btrfs_end_transaction(trans, root);
f062abf0
LZ		 350 out_drop:
351 if (ret) {
352 ip->flags = ip_oldflags;
353 inode->i_flags = i_oldflags;
354 }
6cbff00f 355
6cbff00f
CH		 356 out_unlock:
357 mutex_unlock(&inode->i_mutex);
e7848683 358 mnt_drop_write_file(file);
2d4e6f6a 359 return ret;
6cbff00f
CH		 360}
361
362static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
363{
496ad9aa 364 struct inode *inode = file_inode(file);
6cbff00f
CH		 365
366 return put_user(inode->i_generation, arg);
367}
f46b5a66 368
f7039b1d
LD		 369static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
370{
54563d41 371 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
f7039b1d
LD		 372 struct btrfs_device *device;
373 struct request_queue *q;
374 struct fstrim_range range;
375 u64 minlen = ULLONG_MAX;
376 u64 num_devices = 0;
815745cf 377 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 378 int ret;
379
380 if (!capable(CAP_SYS_ADMIN))
381 return -EPERM;
382
1f78160c
XG		 383 rcu_read_lock();
384 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
385 dev_list) {
f7039b1d
LD		 386 if (!device->bdev)
387 continue;
388 q = bdev_get_queue(device->bdev);
389 if (blk_queue_discard(q)) {
390 num_devices++;
391 minlen = min((u64)q->limits.discard_granularity,
392 minlen);
393 }
394 }
1f78160c 395 rcu_read_unlock();
f4c697e6 396
f7039b1d
LD		 397 if (!num_devices)
398 return -EOPNOTSUPP;
f7039b1d
LD		 399 if (copy_from_user(&range, arg, sizeof(range)))
400 return -EFAULT;
e515c18b
LC		 401 if (range.start > total_bytes ||
402 range.len < fs_info->sb->s_blocksize)
f4c697e6 403 return -EINVAL;
f7039b1d 404
f4c697e6 405 range.len = min(range.len, total_bytes - range.start);
f7039b1d 406 range.minlen = max(range.minlen, minlen);
815745cf 407 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 408 if (ret < 0)
409 return ret;
410
411 if (copy_to_user(arg, &range, sizeof(range)))
412 return -EFAULT;
413
414 return 0;
415}
416
dd5f9615
SB		 417int btrfs_is_empty_uuid(u8 *uuid)
418{
46e0f66a
CM		 419 int i;
420
421 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
422 if (uuid[i])
423 return 0;
424 }
425 return 1;
dd5f9615
SB		 426}
427
d5c12070 428static noinline int create_subvol(struct inode *dir,
cb8e7090 429 struct dentry *dentry,
72fd032e 430 char *name, int namelen,
6f72c7e2 431 u64 *async_transid,
8696c533 432 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 433{
434 struct btrfs_trans_handle *trans;
435 struct btrfs_key key;
436 struct btrfs_root_item root_item;
437 struct btrfs_inode_item *inode_item;
438 struct extent_buffer *leaf;
d5c12070 439 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 440 struct btrfs_root *new_root;
d5c12070 441 struct btrfs_block_rsv block_rsv;
8ea05e3a 442 struct timespec cur_time = CURRENT_TIME;
5662344b 443 struct inode *inode;
f46b5a66
CH		 444 int ret;
445 int err;
446 u64 objectid;
447 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 448 u64 index = 0;
d5c12070 449 u64 qgroup_reserved;
8ea05e3a 450 uuid_le new_uuid;
f46b5a66 451
581bb050 452 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 453 if (ret)
a22285a6 454 return ret;
6a912213 455
d5c12070 456 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 457 /*
d5c12070
MX		 458 * The same as the snapshot creation, please see the comment
459 * of create_snapshot().
9ed74f2d 460 */
d5c12070 461 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 462 8, &qgroup_reserved, false);
d5c12070
MX		 463 if (ret)
464 return ret;
465
466 trans = btrfs_start_transaction(root, 0);
467 if (IS_ERR(trans)) {
468 ret = PTR_ERR(trans);
de6e8200
LB		 469 btrfs_subvolume_release_metadata(root, &block_rsv,
470 qgroup_reserved);
471 return ret;
d5c12070
MX		 472 }
473 trans->block_rsv = &block_rsv;
474 trans->bytes_reserved = block_rsv.size;
f46b5a66 475
8696c533 476 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 477 if (ret)
478 goto fail;
479
5d4f98a2 480 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
5581a51a 481 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 482 if (IS_ERR(leaf)) {
483 ret = PTR_ERR(leaf);
484 goto fail;
485 }
f46b5a66 486
5d4f98a2 487 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 488 btrfs_set_header_bytenr(leaf, leaf->start);
489 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 490 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 491 btrfs_set_header_owner(leaf, objectid);
492
0a4e5586 493 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
f46b5a66 494 BTRFS_FSID_SIZE);
5d4f98a2 495 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
b308bc2f 496 btrfs_header_chunk_tree_uuid(leaf),
5d4f98a2 497 BTRFS_UUID_SIZE);
f46b5a66
CH		 498 btrfs_mark_buffer_dirty(leaf);
499
8ea05e3a
AB		 500 memset(&root_item, 0, sizeof(root_item));
501
f46b5a66 502 inode_item = &root_item.inode;
3cae210f
QW		 503 btrfs_set_stack_inode_generation(inode_item, 1);
504 btrfs_set_stack_inode_size(inode_item, 3);
505 btrfs_set_stack_inode_nlink(inode_item, 1);
506 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
507 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 508
3cae210f
QW		 509 btrfs_set_root_flags(&root_item, 0);
510 btrfs_set_root_limit(&root_item, 0);
511 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 512
f46b5a66 513 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 514 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 515 btrfs_set_root_level(&root_item, 0);
516 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 517 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 518 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 519
8ea05e3a
AB		 520 btrfs_set_root_generation_v2(&root_item,
521 btrfs_root_generation(&root_item));
522 uuid_le_gen(&new_uuid);
523 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 524 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
525 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 526 root_item.ctime = root_item.otime;
527 btrfs_set_root_ctransid(&root_item, trans->transid);
528 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 529
925baedd 530 btrfs_tree_unlock(leaf);
f46b5a66
CH		 531 free_extent_buffer(leaf);
532 leaf = NULL;
533
534 btrfs_set_root_dirid(&root_item, new_dirid);
535
536 key.objectid = objectid;
5d4f98a2 537 key.offset = 0;
f46b5a66
CH		 538 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
539 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
540 &root_item);
541 if (ret)
542 goto fail;
543
76dda93c
YZ		 544 key.offset = (u64)-1;
545 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 546 if (IS_ERR(new_root)) {
547 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
548 ret = PTR_ERR(new_root);
549 goto fail;
550 }
76dda93c
YZ		 551
552 btrfs_record_root_in_trans(trans, new_root);
553
63541927 554 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 555 if (ret) {
556 /* We potentially lose an unused inode item here */
79787eaa 557 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 558 goto fail;
559 }
560
f46b5a66
CH		 561 /*
562 * insert the directory item
563 */
3de4586c 564 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 565 if (ret) {
566 btrfs_abort_transaction(trans, root, ret);
567 goto fail;
568 }
3de4586c
CM		 569
570 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 571 name, namelen, dir, &key,
3de4586c 572 BTRFS_FT_DIR, index);
79787eaa
JM		 573 if (ret) {
574 btrfs_abort_transaction(trans, root, ret);
f46b5a66 575 goto fail;
79787eaa 576 }
0660b5af 577
52c26179
YZ		 578 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
579 ret = btrfs_update_inode(trans, root, dir);
580 BUG_ON(ret);
581
0660b5af 582 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 583 objectid, root->root_key.objectid,
33345d01 584 btrfs_ino(dir), index, name, namelen);
76dda93c 585 BUG_ON(ret);
f46b5a66 586
dd5f9615
SB		 587 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
588 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
589 objectid);
590 if (ret)
591 btrfs_abort_transaction(trans, root, ret);
592
f46b5a66 593fail:
d5c12070
MX		 594 trans->block_rsv = NULL;
595 trans->bytes_reserved = 0;
de6e8200
LB		 596 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
597
72fd032e
SW		 598 if (async_transid) {
599 *async_transid = trans->transid;
600 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 601 if (err)
602 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 603 } else {
604 err = btrfs_commit_transaction(trans, root);
605 }
f46b5a66
CH		 606 if (err && !ret)
607 ret = err;
1a65e24b 608
5662344b
TI		 609 if (!ret) {
610 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 611 if (IS_ERR(inode))
612 return PTR_ERR(inode);
5662344b
TI		 613 d_instantiate(dentry, inode);
614 }
f46b5a66
CH		 615 return ret;
616}
617
8257b2dc
MX		 618static void btrfs_wait_nocow_write(struct btrfs_root *root)
619{
620 s64 writers;
621 DEFINE_WAIT(wait);
622
623 do {
624 prepare_to_wait(&root->subv_writers->wait, &wait,
625 TASK_UNINTERRUPTIBLE);
626
627 writers = percpu_counter_sum(&root->subv_writers->counter);
628 if (writers)
629 schedule();
630
631 finish_wait(&root->subv_writers->wait, &wait);
632 } while (writers);
633}
634
e9662f70
MX		 635static int create_snapshot(struct btrfs_root *root, struct inode *dir,
636 struct dentry *dentry, char *name, int namelen,
637 u64 *async_transid, bool readonly,
638 struct btrfs_qgroup_inherit *inherit)
f46b5a66 639{
2e4bfab9 640 struct inode *inode;
f46b5a66
CH		 641 struct btrfs_pending_snapshot *pending_snapshot;
642 struct btrfs_trans_handle *trans;
2e4bfab9 643 int ret;
f46b5a66 644
27cdeb70 645 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
f46b5a66
CH		 646 return -EINVAL;
647
8257b2dc
MX		 648 atomic_inc(&root->will_be_snapshoted);
649 smp_mb__after_atomic_inc();
650 btrfs_wait_nocow_write(root);
651
6a03843d
MX		 652 ret = btrfs_start_delalloc_inodes(root, 0);
653 if (ret)
8257b2dc 654 goto out;
6a03843d 655
b0244199 656 btrfs_wait_ordered_extents(root, -1);
6a03843d 657
3de4586c 658 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
8257b2dc
MX		 659 if (!pending_snapshot) {
660 ret = -ENOMEM;
661 goto out;
662 }
a22285a6 663
66d8f3dd
MX		 664 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
665 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 666 /*
667 * 1 - parent dir inode
668 * 2 - dir entries
669 * 1 - root item
670 * 2 - root ref/backref
671 * 1 - root of snapshot
dd5f9615 672 * 1 - UUID item
d5c12070
MX		 673 */
674 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 675 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 676 &pending_snapshot->qgroup_reserved,
677 false);
d5c12070 678 if (ret)
8257b2dc 679 goto free;
d5c12070 680
3de4586c 681 pending_snapshot->dentry = dentry;
f46b5a66 682 pending_snapshot->root = root;
b83cc969 683 pending_snapshot->readonly = readonly;
e9662f70 684 pending_snapshot->dir = dir;
8696c533 685 pending_snapshot->inherit = inherit;
a22285a6 686
d5c12070 687 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 688 if (IS_ERR(trans)) {
689 ret = PTR_ERR(trans);
690 goto fail;
691 }
692
8351583e 693 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 694 list_add(&pending_snapshot->list,
695 &trans->transaction->pending_snapshots);
8351583e 696 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 697 if (async_transid) {
698 *async_transid = trans->transid;
699 ret = btrfs_commit_transaction_async(trans,
700 root->fs_info->extent_root, 1);
00d71c9c
MX		 701 if (ret)
702 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 703 } else {
704 ret = btrfs_commit_transaction(trans,
705 root->fs_info->extent_root);
706 }
aec8030a 707 if (ret)
c37b2b62 708 goto fail;
a22285a6
YZ		 709
710 ret = pending_snapshot->error;
711 if (ret)
712 goto fail;
713
66b4ffd1
JB		 714 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
715 if (ret)
716 goto fail;
f46b5a66 717
3821f348
FM		 718 /*
719 * If orphan cleanup did remove any orphans, it means the tree was
720 * modified and therefore the commit root is not the same as the
721 * current root anymore. This is a problem, because send uses the
722 * commit root and therefore can see inode items that don't exist
723 * in the current root anymore, and for example make calls to
724 * btrfs_iget, which will do tree lookups based on the current root
725 * and not on the commit root. Those lookups will fail, returning a
726 * -ESTALE error, and making send fail with that error. So make sure
727 * a send does not see any orphans we have just removed, and that it
728 * will see the same inodes regardless of whether a transaction
729 * commit happened before it started (meaning that the commit root
730 * will be the same as the current root) or not.
731 */
732 if (readonly && pending_snapshot->snap->node !=
733 pending_snapshot->snap->commit_root) {
734 trans = btrfs_join_transaction(pending_snapshot->snap);
735 if (IS_ERR(trans) && PTR_ERR(trans) != -ENOENT) {
736 ret = PTR_ERR(trans);
737 goto fail;
738 }
739 if (!IS_ERR(trans)) {
740 ret = btrfs_commit_transaction(trans,
741 pending_snapshot->snap);
742 if (ret)
743 goto fail;
744 }
745 }
746
2fbe8c8a 747 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 748 if (IS_ERR(inode)) {
749 ret = PTR_ERR(inode);
750 goto fail;
751 }
5662344b 752
2e4bfab9
YZ		 753 d_instantiate(dentry, inode);
754 ret = 0;
755fail:
d5c12070
MX		 756 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
757 &pending_snapshot->block_rsv,
758 pending_snapshot->qgroup_reserved);
8257b2dc 759free:
a22285a6 760 kfree(pending_snapshot);
8257b2dc
MX		 761out:
762 atomic_dec(&root->will_be_snapshoted);
f46b5a66
CH		 763 return ret;
764}
765
4260f7c7
SW		 766/* copy of check_sticky in fs/namei.c()
767* It's inline, so penalty for filesystems that don't use sticky bit is
768* minimal.
769*/
770static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
771{
2f2f43d3 772 kuid_t fsuid = current_fsuid();
4260f7c7
SW		 773
774 if (!(dir->i_mode & S_ISVTX))
775 return 0;
2f2f43d3 776 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 777 return 0;
2f2f43d3 778 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW		 779 return 0;
780 return !capable(CAP_FOWNER);
781}
782
783/* copy of may_delete in fs/namei.c()
784 * Check whether we can remove a link victim from directory dir, check
785 * whether the type of victim is right.
786 * 1. We can't do it if dir is read-only (done in permission())
787 * 2. We should have write and exec permissions on dir
788 * 3. We can't remove anything from append-only dir
789 * 4. We can't do anything with immutable dir (done in permission())
790 * 5. If the sticky bit on dir is set we should either
791 * a. be owner of dir, or
792 * b. be owner of victim, or
793 * c. have CAP_FOWNER capability
794 * 6. If the victim is append-only or immutable we can't do antyhing with
795 * links pointing to it.
796 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
797 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
798 * 9. We can't remove a root or mountpoint.
799 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
800 * nfs_async_unlink().
801 */
802
67871254 803static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 804{
805 int error;
806
807 if (!victim->d_inode)
808 return -ENOENT;
809
810 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 811 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 812
813 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
814 if (error)
815 return error;
816 if (IS_APPEND(dir))
817 return -EPERM;
818 if (btrfs_check_sticky(dir, victim->d_inode)||
819 IS_APPEND(victim->d_inode)||
820 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
821 return -EPERM;
822 if (isdir) {
823 if (!S_ISDIR(victim->d_inode->i_mode))
824 return -ENOTDIR;
825 if (IS_ROOT(victim))
826 return -EBUSY;
827 } else if (S_ISDIR(victim->d_inode->i_mode))
828 return -EISDIR;
829 if (IS_DEADDIR(dir))
830 return -ENOENT;
831 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
832 return -EBUSY;
833 return 0;
834}
835
cb8e7090
CH		 836/* copy of may_create in fs/namei.c() */
837static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
838{
839 if (child->d_inode)
840 return -EEXIST;
841 if (IS_DEADDIR(dir))
842 return -ENOENT;
843 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
844}
845
846/*
847 * Create a new subvolume below @parent. This is largely modeled after
848 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
849 * inside this filesystem so it's quite a bit simpler.
850 */
76dda93c
YZ		 851static noinline int btrfs_mksubvol(struct path *parent,
852 char *name, int namelen,
72fd032e 853 struct btrfs_root *snap_src,
6f72c7e2 854 u64 *async_transid, bool readonly,
8696c533 855 struct btrfs_qgroup_inherit *inherit)
cb8e7090 856{
76dda93c 857 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 858 struct dentry *dentry;
859 int error;
860
5c50c9b8
DS		 861 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
862 if (error == -EINTR)
863 return error;
cb8e7090
CH		 864
865 dentry = lookup_one_len(name, parent->dentry, namelen);
866 error = PTR_ERR(dentry);
867 if (IS_ERR(dentry))
868 goto out_unlock;
869
870 error = -EEXIST;
871 if (dentry->d_inode)
872 goto out_dput;
873
76dda93c 874 error = btrfs_may_create(dir, dentry);
cb8e7090 875 if (error)
a874a63e 876 goto out_dput;
cb8e7090 877
9c52057c
CM		 878 /*
879 * even if this name doesn't exist, we may get hash collisions.
880 * check for them now when we can safely fail
881 */
882 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
883 dir->i_ino, name,
884 namelen);
885 if (error)
886 goto out_dput;
887
76dda93c
YZ		 888 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
889
890 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
891 goto out_up_read;
892
3de4586c 893 if (snap_src) {
e9662f70 894 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 895 async_transid, readonly, inherit);
3de4586c 896 } else {
d5c12070
MX		 897 error = create_subvol(dir, dentry, name, namelen,
898 async_transid, inherit);
3de4586c 899 }
76dda93c
YZ		 900 if (!error)
901 fsnotify_mkdir(dir, dentry);
902out_up_read:
903 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 904out_dput:
905 dput(dentry);
906out_unlock:
76dda93c 907 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 908 return error;
909}
910
4cb5300b
CM		 911/*
912 * When we're defragging a range, we don't want to kick it off again
913 * if it is really just waiting for delalloc to send it down.
914 * If we find a nice big extent or delalloc range for the bytes in the
915 * file you want to defrag, we return 0 to let you know to skip this
916 * part of the file
917 */
918static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
919{
920 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
921 struct extent_map *em = NULL;
922 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
923 u64 end;
924
925 read_lock(&em_tree->lock);
926 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
927 read_unlock(&em_tree->lock);
928
929 if (em) {
930 end = extent_map_end(em);
931 free_extent_map(em);
932 if (end - offset > thresh)
933 return 0;
934 }
935 /* if we already have a nice delalloc here, just stop */
936 thresh /= 2;
937 end = count_range_bits(io_tree, &offset, offset + thresh,
938 thresh, EXTENT_DELALLOC, 1);
939 if (end >= thresh)
940 return 0;
941 return 1;
942}
943
944/*
945 * helper function to walk through a file and find extents
946 * newer than a specific transid, and smaller than thresh.
947 *
948 * This is used by the defragging code to find new and small
949 * extents
950 */
951static int find_new_extents(struct btrfs_root *root,
952 struct inode *inode, u64 newer_than,
953 u64 *off, int thresh)
954{
955 struct btrfs_path *path;
956 struct btrfs_key min_key;
4cb5300b
CM		 957 struct extent_buffer *leaf;
958 struct btrfs_file_extent_item *extent;
959 int type;
960 int ret;
a4689d2b 961 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 962
963 path = btrfs_alloc_path();
964 if (!path)
965 return -ENOMEM;
966
a4689d2b 967 min_key.objectid = ino;
4cb5300b
CM		 968 min_key.type = BTRFS_EXTENT_DATA_KEY;
969 min_key.offset = *off;
970
67871254 971 while (1) {
f094c9bd 972 path->keep_locks = 1;
6174d3cb 973 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 974 if (ret != 0)
975 goto none;
f094c9bd
FM		 976 path->keep_locks = 0;
977 btrfs_unlock_up_safe(path, 1);
978process_slot:
a4689d2b 979 if (min_key.objectid != ino)
4cb5300b
CM		 980 goto none;
981 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
982 goto none;
983
984 leaf = path->nodes[0];
985 extent = btrfs_item_ptr(leaf, path->slots[0],
986 struct btrfs_file_extent_item);
987
988 type = btrfs_file_extent_type(leaf, extent);
989 if (type == BTRFS_FILE_EXTENT_REG &&
990 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
991 check_defrag_in_cache(inode, min_key.offset, thresh)) {
992 *off = min_key.offset;
993 btrfs_free_path(path);
994 return 0;
995 }
996
f094c9bd
FM		 997 path->slots[0]++;
998 if (path->slots[0] < btrfs_header_nritems(leaf)) {
999 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1000 goto process_slot;
1001 }
1002
4cb5300b
CM		 1003 if (min_key.offset == (u64)-1)
1004 goto none;
1005
1006 min_key.offset++;
1007 btrfs_release_path(path);
1008 }
1009none:
1010 btrfs_free_path(path);
1011 return -ENOENT;
1012}
1013
6c282eb4 1014static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 1015{
1016 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 1017 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1018 struct extent_map *em;
1019 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 1020
6c282eb4
LZ		 1021 /*
1022 * hopefully we have this extent in the tree already, try without
1023 * the full extent lock
1024 */
17ce6ef8 1025 read_lock(&em_tree->lock);
6c282eb4 1026 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 1027 read_unlock(&em_tree->lock);
1028
6c282eb4 1029 if (!em) {
308d9800
FM		 1030 struct extent_state *cached = NULL;
1031 u64 end = start + len - 1;
1032
6c282eb4 1033 /* get the big lock and read metadata off disk */
308d9800 1034 lock_extent_bits(io_tree, start, end, 0, &cached);
6c282eb4 1035 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
308d9800 1036 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
6c282eb4
LZ		 1037
1038 if (IS_ERR(em))
1039 return NULL;
1040 }
1041
1042 return em;
1043}
17ce6ef8 1044
6c282eb4
LZ		 1045static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1046{
1047 struct extent_map *next;
1048 bool ret = true;
1049
1050 /* this is the last extent */
1051 if (em->start + em->len >= i_size_read(inode))
1052 return false;
1053
1054 next = defrag_lookup_extent(inode, em->start + em->len);
e2127cf0
FM		 1055 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE ||
1056 (em->block_start + em->block_len == next->block_start))
6c282eb4
LZ		 1057 ret = false;
1058
1059 free_extent_map(next);
17ce6ef8
LB		 1060 return ret;
1061}
1062
6c282eb4 1063static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 1064 u64 *last_len, u64 *skip, u64 *defrag_end,
1065 int compress)
940100a4 1066{
6c282eb4 1067 struct extent_map *em;
940100a4 1068 int ret = 1;
6c282eb4 1069 bool next_mergeable = true;
940100a4
CM		 1070
1071 /*
008873ea 1072 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1073 * defragging it
1074 */
1075 if (start < *defrag_end)
1076 return 1;
1077
1078 *skip = 0;
1079
6c282eb4
LZ		 1080 em = defrag_lookup_extent(inode, start);
1081 if (!em)
1082 return 0;
940100a4
CM		 1083
1084 /* this will cover holes, and inline extents */
17ce6ef8 1085 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1086 ret = 0;
17ce6ef8
LB		 1087 goto out;
1088 }
1089
6c282eb4 1090 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM		 1091
1092 /*
6c282eb4
LZ		 1093 * we hit a real extent, if it is big or the next extent is not a
1094 * real extent, don't bother defragging it
940100a4 1095 */
a43a2111 1096 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 1097 (em->len >= thresh || !next_mergeable))
940100a4 1098 ret = 0;
17ce6ef8 1099out:
940100a4
CM		 1100 /*
1101 * last_len ends up being a counter of how many bytes we've defragged.
1102 * every time we choose not to defrag an extent, we reset *last_len
1103 * so that the next tiny extent will force a defrag.
1104 *
1105 * The end result of this is that tiny extents before a single big
1106 * extent will force at least part of that big extent to be defragged.
1107 */
1108 if (ret) {
940100a4
CM		 1109 *defrag_end = extent_map_end(em);
1110 } else {
1111 *last_len = 0;
1112 *skip = extent_map_end(em);
1113 *defrag_end = 0;
1114 }
1115
1116 free_extent_map(em);
1117 return ret;
1118}
1119
4cb5300b
CM		 1120/*
1121 * it doesn't do much good to defrag one or two pages
1122 * at a time. This pulls in a nice chunk of pages
1123 * to COW and defrag.
1124 *
1125 * It also makes sure the delalloc code has enough
1126 * dirty data to avoid making new small extents as part
1127 * of the defrag
1128 *
1129 * It's a good idea to start RA on this range
1130 * before calling this.
1131 */
1132static int cluster_pages_for_defrag(struct inode *inode,
1133 struct page **pages,
1134 unsigned long start_index,
c41570c9 1135 unsigned long num_pages)
f46b5a66 1136{
4cb5300b
CM		 1137 unsigned long file_end;
1138 u64 isize = i_size_read(inode);
1139 u64 page_start;
1140 u64 page_end;
1f12bd06 1141 u64 page_cnt;
4cb5300b
CM		 1142 int ret;
1143 int i;
1144 int i_done;
3eaa2885 1145 struct btrfs_ordered_extent *ordered;
4cb5300b 1146 struct extent_state *cached_state = NULL;
600a45e1 1147 struct extent_io_tree *tree;
3b16a4e3 1148 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1149
4cb5300b 1150 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1151 if (!isize || start_index > file_end)
1152 return 0;
1153
1154 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1155
1156 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1157 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1158 if (ret)
1159 return ret;
4cb5300b 1160 i_done = 0;
600a45e1 1161 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1162
1163 /* step one, lock all the pages */
1f12bd06 1164 for (i = 0; i < page_cnt; i++) {
4cb5300b 1165 struct page *page;
600a45e1 1166again:
a94733d0 1167 page = find_or_create_page(inode->i_mapping,
600a45e1 1168 start_index + i, mask);
4cb5300b
CM		 1169 if (!page)
1170 break;
1171
600a45e1
MX		 1172 page_start = page_offset(page);
1173 page_end = page_start + PAGE_CACHE_SIZE - 1;
1174 while (1) {
308d9800
FM		 1175 lock_extent_bits(tree, page_start, page_end,
1176 0, &cached_state);
600a45e1
MX		 1177 ordered = btrfs_lookup_ordered_extent(inode,
1178 page_start);
308d9800
FM		 1179 unlock_extent_cached(tree, page_start, page_end,
1180 &cached_state, GFP_NOFS);
600a45e1
MX		 1181 if (!ordered)
1182 break;
1183
1184 unlock_page(page);
1185 btrfs_start_ordered_extent(inode, ordered, 1);
1186 btrfs_put_ordered_extent(ordered);
1187 lock_page(page);
1f12bd06
LB		 1188 /*
1189 * we unlocked the page above, so we need check if
1190 * it was released or not.
1191 */
1192 if (page->mapping != inode->i_mapping) {
1193 unlock_page(page);
1194 page_cache_release(page);
1195 goto again;
1196 }
600a45e1
MX		 1197 }
1198
4cb5300b
CM		 1199 if (!PageUptodate(page)) {
1200 btrfs_readpage(NULL, page);
1201 lock_page(page);
1202 if (!PageUptodate(page)) {
1203 unlock_page(page);
1204 page_cache_release(page);
1205 ret = -EIO;
1206 break;
1207 }
1208 }
600a45e1 1209
600a45e1
MX		 1210 if (page->mapping != inode->i_mapping) {
1211 unlock_page(page);
1212 page_cache_release(page);
1213 goto again;
1214 }
1215
4cb5300b
CM		 1216 pages[i] = page;
1217 i_done++;
1218 }
1219 if (!i_done || ret)
1220 goto out;
1221
1222 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1223 goto out;
1224
1225 /*
1226 * so now we have a nice long stream of locked
1227 * and up to date pages, lets wait on them
1228 */
1229 for (i = 0; i < i_done; i++)
1230 wait_on_page_writeback(pages[i]);
1231
1232 page_start = page_offset(pages[0]);
1233 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1234
1235 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1236 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1237 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1238 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1239 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1240 &cached_state, GFP_NOFS);
4cb5300b 1241
1f12bd06 1242 if (i_done != page_cnt) {
9e0baf60
JB		 1243 spin_lock(&BTRFS_I(inode)->lock);
1244 BTRFS_I(inode)->outstanding_extents++;
1245 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1246 btrfs_delalloc_release_space(inode,
1f12bd06 1247 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1248 }
1249
1250
9e8a4a8b
LB		 1251 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1252 &cached_state, GFP_NOFS);
4cb5300b
CM		 1253
1254 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1255 page_start, page_end - 1, &cached_state,
1256 GFP_NOFS);
1257
1258 for (i = 0; i < i_done; i++) {
1259 clear_page_dirty_for_io(pages[i]);
1260 ClearPageChecked(pages[i]);
1261 set_page_extent_mapped(pages[i]);
1262 set_page_dirty(pages[i]);
1263 unlock_page(pages[i]);
1264 page_cache_release(pages[i]);
1265 }
1266 return i_done;
1267out:
1268 for (i = 0; i < i_done; i++) {
1269 unlock_page(pages[i]);
1270 page_cache_release(pages[i]);
1271 }
1f12bd06 1272 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1273 return ret;
1274
1275}
1276
1277int btrfs_defrag_file(struct inode *inode, struct file *file,
1278 struct btrfs_ioctl_defrag_range_args *range,
1279 u64 newer_than, unsigned long max_to_defrag)
1280{
1281 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1282 struct file_ra_state *ra = NULL;
f46b5a66 1283 unsigned long last_index;
151a31b2 1284 u64 isize = i_size_read(inode);
940100a4
CM		 1285 u64 last_len = 0;
1286 u64 skip = 0;
1287 u64 defrag_end = 0;
4cb5300b 1288 u64 newer_off = range->start;
f46b5a66 1289 unsigned long i;
008873ea 1290 unsigned long ra_index = 0;
f46b5a66 1291 int ret;
4cb5300b 1292 int defrag_count = 0;
1a419d85 1293 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1294 int extent_thresh = range->extent_thresh;
c41570c9
JM		 1295 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1296 unsigned long cluster = max_cluster;
4cb5300b
CM		 1297 u64 new_align = ~((u64)128 * 1024 - 1);
1298 struct page **pages = NULL;
1299
0abd5b17
LB		 1300 if (isize == 0)
1301 return 0;
1302
1303 if (range->start >= isize)
1304 return -EINVAL;
1a419d85
LZ		 1305
1306 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1307 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1308 return -EINVAL;
1309 if (range->compress_type)
1310 compress_type = range->compress_type;
1311 }
f46b5a66 1312
0abd5b17
LB		 1313 if (extent_thresh == 0)
1314 extent_thresh = 256 * 1024;
940100a4 1315
4cb5300b
CM		 1316 /*
1317 * if we were not given a file, allocate a readahead
1318 * context
1319 */
1320 if (!file) {
1321 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1322 if (!ra)
1323 return -ENOMEM;
1324 file_ra_state_init(ra, inode->i_mapping);
1325 } else {
1326 ra = &file->f_ra;
1327 }
1328
d9b0d9ba 1329 pages = kmalloc_array(max_cluster, sizeof(struct page *),
4cb5300b
CM		 1330 GFP_NOFS);
1331 if (!pages) {
1332 ret = -ENOMEM;
1333 goto out_ra;
1334 }
1335
1336 /* find the last page to defrag */
1e701a32 1337 if (range->start + range->len > range->start) {
151a31b2 1338 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1339 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1340 } else {
151a31b2 1341 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1342 }
1343
4cb5300b
CM		 1344 if (newer_than) {
1345 ret = find_new_extents(root, inode, newer_than,
1346 &newer_off, 64 * 1024);
1347 if (!ret) {
1348 range->start = newer_off;
1349 /*
1350 * we always align our defrag to help keep
1351 * the extents in the file evenly spaced
1352 */
1353 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1354 } else
1355 goto out_ra;
1356 } else {
1357 i = range->start >> PAGE_CACHE_SHIFT;
1358 }
1359 if (!max_to_defrag)
7ec31b54 1360 max_to_defrag = last_index + 1;
4cb5300b 1361
2a0f7f57
LZ		 1362 /*
1363 * make writeback starts from i, so the defrag range can be
1364 * written sequentially.
1365 */
1366 if (i < inode->i_mapping->writeback_index)
1367 inode->i_mapping->writeback_index = i;
1368
f7f43cc8
CM		 1369 while (i <= last_index && defrag_count < max_to_defrag &&
1370 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1371 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1372 /*
1373 * make sure we stop running if someone unmounts
1374 * the FS
1375 */
1376 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1377 break;
1378
210549eb 1379 if (btrfs_defrag_cancelled(root->fs_info)) {
efe120a0 1380 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
210549eb
DS		 1381 ret = -EAGAIN;
1382 break;
1383 }
1384
66c26892 1385 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1386 extent_thresh, &last_len, &skip,
a43a2111
AM		 1387 &defrag_end, range->flags &
1388 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1389 unsigned long next;
1390 /*
1391 * the should_defrag function tells us how much to skip
1392 * bump our counter by the suggested amount
1393 */
1394 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1395 i = max(i + 1, next);
1396 continue;
1397 }
008873ea
LZ		 1398
1399 if (!newer_than) {
1400 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1401 PAGE_CACHE_SHIFT) - i;
1402 cluster = min(cluster, max_cluster);
1403 } else {
1404 cluster = max_cluster;
1405 }
1406
008873ea
LZ		 1407 if (i + cluster > ra_index) {
1408 ra_index = max(i, ra_index);
1409 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1410 cluster);
1411 ra_index += max_cluster;
1412 }
940100a4 1413
ecb8bea8 1414 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1415 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1416 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1417 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1418 if (ret < 0) {
1419 mutex_unlock(&inode->i_mutex);
4cb5300b 1420 goto out_ra;
ecb8bea8 1421 }
4cb5300b
CM		 1422
1423 defrag_count += ret;
d0e1d66b 1424 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1425 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1426
1427 if (newer_than) {
1428 if (newer_off == (u64)-1)
1429 break;
1430
e1f041e1
LB		 1431 if (ret > 0)
1432 i += ret;
1433
4cb5300b
CM		 1434 newer_off = max(newer_off + 1,
1435 (u64)i << PAGE_CACHE_SHIFT);
1436
1437 ret = find_new_extents(root, inode,
1438 newer_than, &newer_off,
1439 64 * 1024);
1440 if (!ret) {
1441 range->start = newer_off;
1442 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1443 } else {
1444 break;
f46b5a66 1445 }
4cb5300b 1446 } else {
008873ea 1447 if (ret > 0) {
cbcc8326 1448 i += ret;
008873ea
LZ		 1449 last_len += ret << PAGE_CACHE_SHIFT;
1450 } else {
cbcc8326 1451 i++;
008873ea
LZ		 1452 last_len = 0;
1453 }
f46b5a66 1454 }
f46b5a66
CH		 1455 }
1456
dec8ef90 1457 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1458 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1459 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1460 &BTRFS_I(inode)->runtime_flags))
1461 filemap_flush(inode->i_mapping);
1462 }
1e701a32
CM		 1463
1464 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1465 /* the filemap_flush will queue IO into the worker threads, but
1466 * we have to make sure the IO is actually started and that
1467 * ordered extents get created before we return
1468 */
1469 atomic_inc(&root->fs_info->async_submit_draining);
1470 while (atomic_read(&root->fs_info->nr_async_submits) ||
1471 atomic_read(&root->fs_info->async_delalloc_pages)) {
1472 wait_event(root->fs_info->async_submit_wait,
1473 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1474 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1475 }
1476 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1477 }
1478
1a419d85 1479 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1480 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1481 }
1482
60ccf82f 1483 ret = defrag_count;
940100a4 1484
4cb5300b 1485out_ra:
633085c7
FDBM		 1486 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1487 mutex_lock(&inode->i_mutex);
1488 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1489 mutex_unlock(&inode->i_mutex);
1490 }
4cb5300b
CM		 1491 if (!file)
1492 kfree(ra);
1493 kfree(pages);
940100a4 1494 return ret;
f46b5a66
CH		 1495}
1496
198605a8 1497static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1498 void __user *arg)
f46b5a66
CH		 1499{
1500 u64 new_size;
1501 u64 old_size;
1502 u64 devid = 1;
496ad9aa 1503 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1504 struct btrfs_ioctl_vol_args *vol_args;
1505 struct btrfs_trans_handle *trans;
1506 struct btrfs_device *device = NULL;
1507 char *sizestr;
9a40f122 1508 char *retptr;
f46b5a66
CH		 1509 char *devstr = NULL;
1510 int ret = 0;
f46b5a66
CH		 1511 int mod = 0;
1512
e441d54d
CM		 1513 if (!capable(CAP_SYS_ADMIN))
1514 return -EPERM;
1515
198605a8
MX		 1516 ret = mnt_want_write_file(file);
1517 if (ret)
1518 return ret;
1519
5ac00add
SB		 1520 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1521 1)) {
97547676 1522 mnt_drop_write_file(file);
e57138b3 1523 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1524 }
1525
5ac00add 1526 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1527 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1528 if (IS_ERR(vol_args)) {
1529 ret = PTR_ERR(vol_args);
1530 goto out;
1531 }
5516e595
MF		 1532
1533 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1534
f46b5a66
CH		 1535 sizestr = vol_args->name;
1536 devstr = strchr(sizestr, ':');
1537 if (devstr) {
f46b5a66
CH		 1538 sizestr = devstr + 1;
1539 *devstr = '\0';
1540 devstr = vol_args->name;
58dfae63
Z		 1541 ret = kstrtoull(devstr, 10, &devid);
1542 if (ret)
1543 goto out_free;
dfd79829
MX		 1544 if (!devid) {
1545 ret = -EINVAL;
1546 goto out_free;
1547 }
efe120a0 1548 btrfs_info(root->fs_info, "resizing devid %llu", devid);
f46b5a66 1549 }
dba60f3f 1550
aa1b8cd4 1551 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1552 if (!device) {
efe120a0 1553 btrfs_info(root->fs_info, "resizer unable to find device %llu",
c1c9ff7c 1554 devid);
dfd79829 1555 ret = -ENODEV;
c9e9f97b 1556 goto out_free;
f46b5a66 1557 }
dba60f3f
MX		 1558
1559 if (!device->writeable) {
efe120a0
FH		 1560 btrfs_info(root->fs_info,
1561 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1562 devid);
dfd79829 1563 ret = -EPERM;
4e42ae1b
LB		 1564 goto out_free;
1565 }
1566
f46b5a66
CH		 1567 if (!strcmp(sizestr, "max"))
1568 new_size = device->bdev->bd_inode->i_size;
1569 else {
1570 if (sizestr[0] == '-') {
1571 mod = -1;
1572 sizestr++;
1573 } else if (sizestr[0] == '+') {
1574 mod = 1;
1575 sizestr++;
1576 }
9a40f122
GH		 1577 new_size = memparse(sizestr, &retptr);
1578 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1579 ret = -EINVAL;
c9e9f97b 1580 goto out_free;
f46b5a66
CH		 1581 }
1582 }
1583
63a212ab 1584 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1585 ret = -EPERM;
63a212ab
SB		 1586 goto out_free;
1587 }
1588
f46b5a66
CH		 1589 old_size = device->total_bytes;
1590
1591 if (mod < 0) {
1592 if (new_size > old_size) {
1593 ret = -EINVAL;
c9e9f97b 1594 goto out_free;
f46b5a66
CH		 1595 }
1596 new_size = old_size - new_size;
1597 } else if (mod > 0) {
eb8052e0 1598 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1599 ret = -ERANGE;
eb8052e0
WF		 1600 goto out_free;
1601 }
f46b5a66
CH		 1602 new_size = old_size + new_size;
1603 }
1604
1605 if (new_size < 256 * 1024 * 1024) {
1606 ret = -EINVAL;
c9e9f97b 1607 goto out_free;
f46b5a66
CH		 1608 }
1609 if (new_size > device->bdev->bd_inode->i_size) {
1610 ret = -EFBIG;
c9e9f97b 1611 goto out_free;
f46b5a66
CH		 1612 }
1613
1614 do_div(new_size, root->sectorsize);
1615 new_size *= root->sectorsize;
1616
efe120a0 1617 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
c1c9ff7c 1618 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1619
1620 if (new_size > old_size) {
a22285a6 1621 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1622 if (IS_ERR(trans)) {
1623 ret = PTR_ERR(trans);
c9e9f97b 1624 goto out_free;
98d5dc13 1625 }
f46b5a66
CH		 1626 ret = btrfs_grow_device(trans, device, new_size);
1627 btrfs_commit_transaction(trans, root);
ece7d20e 1628 } else if (new_size < old_size) {
f46b5a66 1629 ret = btrfs_shrink_device(device, new_size);
0253f40e 1630 } /* equal, nothing need to do */
f46b5a66 1631
c9e9f97b 1632out_free:
f46b5a66 1633 kfree(vol_args);
c9e9f97b
ID		 1634out:
1635 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1636 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1637 mnt_drop_write_file(file);
f46b5a66
CH		 1638 return ret;
1639}
1640
72fd032e 1641static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1642 char *name, unsigned long fd, int subvol,
1643 u64 *transid, bool readonly,
8696c533 1644 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1645{
f46b5a66 1646 int namelen;
3de4586c 1647 int ret = 0;
f46b5a66 1648
a874a63e
LB		 1649 ret = mnt_want_write_file(file);
1650 if (ret)
1651 goto out;
1652
72fd032e
SW		 1653 namelen = strlen(name);
1654 if (strchr(name, '/')) {
f46b5a66 1655 ret = -EINVAL;
a874a63e 1656 goto out_drop_write;
f46b5a66
CH		 1657 }
1658
16780cab
CM		 1659 if (name[0] == '.' &&
1660 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1661 ret = -EEXIST;
a874a63e 1662 goto out_drop_write;
16780cab
CM		 1663 }
1664
3de4586c 1665 if (subvol) {
72fd032e 1666 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1667 NULL, transid, readonly, inherit);
cb8e7090 1668 } else {
2903ff01 1669 struct fd src = fdget(fd);
3de4586c 1670 struct inode *src_inode;
2903ff01 1671 if (!src.file) {
3de4586c 1672 ret = -EINVAL;
a874a63e 1673 goto out_drop_write;
3de4586c
CM		 1674 }
1675
496ad9aa
AV		 1676 src_inode = file_inode(src.file);
1677 if (src_inode->i_sb != file_inode(file)->i_sb) {
efe120a0
FH		 1678 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1679 "Snapshot src from another FS");
23ad5b17 1680 ret = -EXDEV;
d0242061
DS		 1681 } else if (!inode_owner_or_capable(src_inode)) {
1682 /*
1683 * Subvolume creation is not restricted, but snapshots
1684 * are limited to own subvolumes only
1685 */
1686 ret = -EPERM;
ecd18815
AV		 1687 } else {
1688 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1689 BTRFS_I(src_inode)->root,
1690 transid, readonly, inherit);
3de4586c 1691 }
2903ff01 1692 fdput(src);
cb8e7090 1693 }
a874a63e
LB		 1694out_drop_write:
1695 mnt_drop_write_file(file);
f46b5a66 1696out:
72fd032e
SW		 1697 return ret;
1698}
1699
1700static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1701 void __user *arg, int subvol)
72fd032e 1702{
fa0d2b9b 1703 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1704 int ret;
1705
fa0d2b9b
LZ		 1706 vol_args = memdup_user(arg, sizeof(*vol_args));
1707 if (IS_ERR(vol_args))
1708 return PTR_ERR(vol_args);
1709 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1710
fa0d2b9b 1711 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1712 vol_args->fd, subvol,
6f72c7e2 1713 NULL, false, NULL);
fdfb1e4f 1714
fa0d2b9b
LZ		 1715 kfree(vol_args);
1716 return ret;
1717}
fdfb1e4f 1718
fa0d2b9b
LZ		 1719static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1720 void __user *arg, int subvol)
1721{
1722 struct btrfs_ioctl_vol_args_v2 *vol_args;
1723 int ret;
1724 u64 transid = 0;
1725 u64 *ptr = NULL;
b83cc969 1726 bool readonly = false;
6f72c7e2 1727 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1728
fa0d2b9b
LZ		 1729 vol_args = memdup_user(arg, sizeof(*vol_args));
1730 if (IS_ERR(vol_args))
1731 return PTR_ERR(vol_args);
1732 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1733
b83cc969 1734 if (vol_args->flags &
6f72c7e2
AJ		 1735 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1736 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1737 ret = -EOPNOTSUPP;
fa0d2b9b 1738 goto out;
72fd032e 1739 }
fa0d2b9b
LZ		 1740
1741 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1742 ptr = &transid;
b83cc969
LZ		 1743 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1744 readonly = true;
6f72c7e2
AJ		 1745 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1746 if (vol_args->size > PAGE_CACHE_SIZE) {
1747 ret = -EINVAL;
1748 goto out;
1749 }
1750 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1751 if (IS_ERR(inherit)) {
1752 ret = PTR_ERR(inherit);
1753 goto out;
1754 }
1755 }
fa0d2b9b
LZ		 1756
1757 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1758 vol_args->fd, subvol, ptr,
8696c533 1759 readonly, inherit);
fa0d2b9b
LZ		 1760
1761 if (ret == 0 && ptr &&
1762 copy_to_user(arg +
1763 offsetof(struct btrfs_ioctl_vol_args_v2,
1764 transid), ptr, sizeof(*ptr)))
1765 ret = -EFAULT;
fdfb1e4f 1766out:
f46b5a66 1767 kfree(vol_args);
6f72c7e2 1768 kfree(inherit);
f46b5a66
CH		 1769 return ret;
1770}
1771
0caa102d
LZ		 1772static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1773 void __user *arg)
1774{
496ad9aa 1775 struct inode *inode = file_inode(file);
0caa102d
LZ		 1776 struct btrfs_root *root = BTRFS_I(inode)->root;
1777 int ret = 0;
1778 u64 flags = 0;
1779
33345d01 1780 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1781 return -EINVAL;
1782
1783 down_read(&root->fs_info->subvol_sem);
1784 if (btrfs_root_readonly(root))
1785 flags |= BTRFS_SUBVOL_RDONLY;
1786 up_read(&root->fs_info->subvol_sem);
1787
1788 if (copy_to_user(arg, &flags, sizeof(flags)))
1789 ret = -EFAULT;
1790
1791 return ret;
1792}
1793
1794static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1795 void __user *arg)
1796{
496ad9aa 1797 struct inode *inode = file_inode(file);
0caa102d
LZ		 1798 struct btrfs_root *root = BTRFS_I(inode)->root;
1799 struct btrfs_trans_handle *trans;
1800 u64 root_flags;
1801 u64 flags;
1802 int ret = 0;
1803
bd60ea0f
DS		 1804 if (!inode_owner_or_capable(inode))
1805 return -EPERM;
1806
b9ca0664
LB		 1807 ret = mnt_want_write_file(file);
1808 if (ret)
1809 goto out;
0caa102d 1810
b9ca0664
LB		 1811 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1812 ret = -EINVAL;
1813 goto out_drop_write;
1814 }
0caa102d 1815
b9ca0664
LB		 1816 if (copy_from_user(&flags, arg, sizeof(flags))) {
1817 ret = -EFAULT;
1818 goto out_drop_write;
1819 }
0caa102d 1820
b9ca0664
LB		 1821 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1822 ret = -EINVAL;
1823 goto out_drop_write;
1824 }
0caa102d 1825
b9ca0664
LB		 1826 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1827 ret = -EOPNOTSUPP;
1828 goto out_drop_write;
1829 }
0caa102d
LZ		 1830
1831 down_write(&root->fs_info->subvol_sem);
1832
1833 /* nothing to do */
1834 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1835 goto out_drop_sem;
0caa102d
LZ		 1836
1837 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1838 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1839 btrfs_set_root_flags(&root->root_item,
1840 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1841 } else {
1842 /*
1843 * Block RO -> RW transition if this subvolume is involved in
1844 * send
1845 */
1846 spin_lock(&root->root_item_lock);
1847 if (root->send_in_progress == 0) {
1848 btrfs_set_root_flags(&root->root_item,
0caa102d 1849 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1850 spin_unlock(&root->root_item_lock);
1851 } else {
1852 spin_unlock(&root->root_item_lock);
1853 btrfs_warn(root->fs_info,
1854 "Attempt to set subvolume %llu read-write during send",
1855 root->root_key.objectid);
1856 ret = -EPERM;
1857 goto out_drop_sem;
1858 }
1859 }
0caa102d
LZ		 1860
1861 trans = btrfs_start_transaction(root, 1);
1862 if (IS_ERR(trans)) {
1863 ret = PTR_ERR(trans);
1864 goto out_reset;
1865 }
1866
b4dc2b8c 1867 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1868 &root->root_key, &root->root_item);
1869
1870 btrfs_commit_transaction(trans, root);
1871out_reset:
1872 if (ret)
1873 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1874out_drop_sem:
0caa102d 1875 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1876out_drop_write:
1877 mnt_drop_write_file(file);
1878out:
0caa102d
LZ		 1879 return ret;
1880}
1881
76dda93c
YZ		 1882/*
1883 * helper to check if the subvolume references other subvolumes
1884 */
1885static noinline int may_destroy_subvol(struct btrfs_root *root)
1886{
1887 struct btrfs_path *path;
175a2b87 1888 struct btrfs_dir_item *di;
76dda93c 1889 struct btrfs_key key;
175a2b87 1890 u64 dir_id;
76dda93c
YZ		 1891 int ret;
1892
1893 path = btrfs_alloc_path();
1894 if (!path)
1895 return -ENOMEM;
1896
175a2b87
JB		 1897 /* Make sure this root isn't set as the default subvol */
1898 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1899 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1900 dir_id, "default", 7, 0);
1901 if (di && !IS_ERR(di)) {
1902 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1903 if (key.objectid == root->root_key.objectid) {
72de6b53
GS		 1904 ret = -EPERM;
1905 btrfs_err(root->fs_info, "deleting default subvolume "
1906 "%llu is not allowed", key.objectid);
175a2b87
JB		 1907 goto out;
1908 }
1909 btrfs_release_path(path);
1910 }
1911
76dda93c
YZ		 1912 key.objectid = root->root_key.objectid;
1913 key.type = BTRFS_ROOT_REF_KEY;
1914 key.offset = (u64)-1;
1915
1916 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1917 &key, path, 0, 0);
1918 if (ret < 0)
1919 goto out;
1920 BUG_ON(ret == 0);
1921
1922 ret = 0;
1923 if (path->slots[0] > 0) {
1924 path->slots[0]--;
1925 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1926 if (key.objectid == root->root_key.objectid &&
1927 key.type == BTRFS_ROOT_REF_KEY)
1928 ret = -ENOTEMPTY;
1929 }
1930out:
1931 btrfs_free_path(path);
1932 return ret;
1933}
1934
ac8e9819
CM		 1935static noinline int key_in_sk(struct btrfs_key *key,
1936 struct btrfs_ioctl_search_key *sk)
1937{
abc6e134
CM		 1938 struct btrfs_key test;
1939 int ret;
1940
1941 test.objectid = sk->min_objectid;
1942 test.type = sk->min_type;
1943 test.offset = sk->min_offset;
1944
1945 ret = btrfs_comp_cpu_keys(key, &test);
1946 if (ret < 0)
ac8e9819 1947 return 0;
abc6e134
CM		 1948
1949 test.objectid = sk->max_objectid;
1950 test.type = sk->max_type;
1951 test.offset = sk->max_offset;
1952
1953 ret = btrfs_comp_cpu_keys(key, &test);
1954 if (ret > 0)
ac8e9819
CM		 1955 return 0;
1956 return 1;
1957}
1958
1959static noinline int copy_to_sk(struct btrfs_root *root,
1960 struct btrfs_path *path,
1961 struct btrfs_key *key,
1962 struct btrfs_ioctl_search_key *sk,
9b6e817d 1963 size_t *buf_size,
ba346b35 1964 char __user *ubuf,
ac8e9819
CM		 1965 unsigned long *sk_offset,
1966 int *num_found)
1967{
1968 u64 found_transid;
1969 struct extent_buffer *leaf;
1970 struct btrfs_ioctl_search_header sh;
1971 unsigned long item_off;
1972 unsigned long item_len;
1973 int nritems;
1974 int i;
1975 int slot;
ac8e9819
CM		 1976 int ret = 0;
1977
1978 leaf = path->nodes[0];
1979 slot = path->slots[0];
1980 nritems = btrfs_header_nritems(leaf);
1981
1982 if (btrfs_header_generation(leaf) > sk->max_transid) {
1983 i = nritems;
1984 goto advance_key;
1985 }
1986 found_transid = btrfs_header_generation(leaf);
1987
1988 for (i = slot; i < nritems; i++) {
1989 item_off = btrfs_item_ptr_offset(leaf, i);
1990 item_len = btrfs_item_size_nr(leaf, i);
1991
03b71c6c
GP		 1992 btrfs_item_key_to_cpu(leaf, key, i);
1993 if (!key_in_sk(key, sk))
1994 continue;
1995
9b6e817d 1996 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 1997 if (*num_found) {
1998 ret = 1;
1999 goto out;
2000 }
2001
2002 /*
2003 * return one empty item back for v1, which does not
2004 * handle -EOVERFLOW
2005 */
2006
9b6e817d 2007 *buf_size = sizeof(sh) + item_len;
ac8e9819 2008 item_len = 0;
8f5f6178
GH		 2009 ret = -EOVERFLOW;
2010 }
ac8e9819 2011
9b6e817d 2012 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 2013 ret = 1;
25c9bc2e 2014 goto out;
ac8e9819
CM		 2015 }
2016
ac8e9819
CM		 2017 sh.objectid = key->objectid;
2018 sh.offset = key->offset;
2019 sh.type = key->type;
2020 sh.len = item_len;
2021 sh.transid = found_transid;
2022
2023 /* copy search result header */
ba346b35
GH		 2024 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2025 ret = -EFAULT;
2026 goto out;
2027 }
2028
ac8e9819
CM		 2029 *sk_offset += sizeof(sh);
2030
2031 if (item_len) {
ba346b35 2032 char __user *up = ubuf + *sk_offset;
ac8e9819 2033 /* copy the item */
ba346b35
GH		 2034 if (read_extent_buffer_to_user(leaf, up,
2035 item_off, item_len)) {
2036 ret = -EFAULT;
2037 goto out;
2038 }
2039
ac8e9819 2040 *sk_offset += item_len;
ac8e9819 2041 }
e2156867 2042 (*num_found)++;
ac8e9819 2043
8f5f6178
GH		 2044 if (ret) /* -EOVERFLOW from above */
2045 goto out;
2046
25c9bc2e
GH		 2047 if (*num_found >= sk->nr_items) {
2048 ret = 1;
2049 goto out;
2050 }
ac8e9819
CM		 2051 }
2052advance_key:
abc6e134
CM		 2053 ret = 0;
2054 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 2055 key->offset++;
abc6e134
CM		 2056 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2057 key->offset = 0;
ac8e9819 2058 key->type++;
abc6e134
CM		 2059 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2060 key->offset = 0;
2061 key->type = 0;
ac8e9819 2062 key->objectid++;
abc6e134
CM		 2063 } else
2064 ret = 1;
25c9bc2e 2065out:
ba346b35
GH		 2066 /*
2067 * 0: all items from this leaf copied, continue with next
2068 * 1: * more items can be copied, but unused buffer is too small
2069 * * all items were found
2070 * Either way, it will stops the loop which iterates to the next
2071 * leaf
2072 * -EOVERFLOW: item was to large for buffer
2073 * -EFAULT: could not copy extent buffer back to userspace
2074 */
ac8e9819
CM		 2075 return ret;
2076}
2077
2078static noinline int search_ioctl(struct inode *inode,
12544442 2079 struct btrfs_ioctl_search_key *sk,
9b6e817d 2080 size_t *buf_size,
ba346b35 2081 char __user *ubuf)
ac8e9819
CM		 2082{
2083 struct btrfs_root *root;
2084 struct btrfs_key key;
ac8e9819 2085 struct btrfs_path *path;
ac8e9819
CM		 2086 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2087 int ret;
2088 int num_found = 0;
2089 unsigned long sk_offset = 0;
2090
9b6e817d
GH		 2091 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2092 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2093 return -EOVERFLOW;
9b6e817d 2094 }
12544442 2095
ac8e9819
CM		 2096 path = btrfs_alloc_path();
2097 if (!path)
2098 return -ENOMEM;
2099
2100 if (sk->tree_id == 0) {
2101 /* search the root of the inode that was passed */
2102 root = BTRFS_I(inode)->root;
2103 } else {
2104 key.objectid = sk->tree_id;
2105 key.type = BTRFS_ROOT_ITEM_KEY;
2106 key.offset = (u64)-1;
2107 root = btrfs_read_fs_root_no_name(info, &key);
2108 if (IS_ERR(root)) {
efe120a0 2109 printk(KERN_ERR "BTRFS: could not find root %llu\n",
ac8e9819
CM		 2110 sk->tree_id);
2111 btrfs_free_path(path);
2112 return -ENOENT;
2113 }
2114 }
2115
2116 key.objectid = sk->min_objectid;
2117 key.type = sk->min_type;
2118 key.offset = sk->min_offset;
2119
ac8e9819
CM		 2120 path->keep_locks = 1;
2121
67871254 2122 while (1) {
6174d3cb 2123 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2124 if (ret != 0) {
2125 if (ret > 0)
2126 ret = 0;
2127 goto err;
2128 }
ba346b35 2129 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
ac8e9819 2130 &sk_offset, &num_found);
b3b4aa74 2131 btrfs_release_path(path);
25c9bc2e 2132 if (ret)
ac8e9819
CM		 2133 break;
2134
2135 }
8f5f6178
GH		 2136 if (ret > 0)
2137 ret = 0;
ac8e9819
CM		 2138err:
2139 sk->nr_items = num_found;
2140 btrfs_free_path(path);
2141 return ret;
2142}
2143
2144static noinline int btrfs_ioctl_tree_search(struct file *file,
2145 void __user *argp)
2146{
ba346b35
GH		 2147 struct btrfs_ioctl_search_args __user *uargs;
2148 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2149 struct inode *inode;
2150 int ret;
2151 size_t buf_size;
ac8e9819
CM		 2152
2153 if (!capable(CAP_SYS_ADMIN))
2154 return -EPERM;
2155
ba346b35
GH		 2156 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2157
2158 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2159 return -EFAULT;
ac8e9819 2160
ba346b35 2161 buf_size = sizeof(uargs->buf);
9b6e817d 2162
496ad9aa 2163 inode = file_inode(file);
ba346b35 2164 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2165
2166 /*
2167 * In the origin implementation an overflow is handled by returning a
2168 * search header with a len of zero, so reset ret.
2169 */
2170 if (ret == -EOVERFLOW)
2171 ret = 0;
2172
ba346b35 2173 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2174 ret = -EFAULT;
ac8e9819
CM		 2175 return ret;
2176}
2177
cc68a8a5
GH		 2178static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2179 void __user *argp)
2180{
2181 struct btrfs_ioctl_search_args_v2 __user *uarg;
2182 struct btrfs_ioctl_search_args_v2 args;
2183 struct inode *inode;
2184 int ret;
2185 size_t buf_size;
2186 const size_t buf_limit = 16 * 1024 * 1024;
2187
2188 if (!capable(CAP_SYS_ADMIN))
2189 return -EPERM;
2190
2191 /* copy search header and buffer size */
2192 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2193 if (copy_from_user(&args, uarg, sizeof(args)))
2194 return -EFAULT;
2195
2196 buf_size = args.buf_size;
2197
2198 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2199 return -EOVERFLOW;
2200
2201 /* limit result size to 16MB */
2202 if (buf_size > buf_limit)
2203 buf_size = buf_limit;
2204
2205 inode = file_inode(file);
2206 ret = search_ioctl(inode, &args.key, &buf_size,
2207 (char *)(&uarg->buf[0]));
2208 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2209 ret = -EFAULT;
2210 else if (ret == -EOVERFLOW &&
2211 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2212 ret = -EFAULT;
2213
2214 return ret;
2215}
2216
98d377a0 2217/*
ac8e9819
CM		 2218 * Search INODE_REFs to identify path name of 'dirid' directory
2219 * in a 'tree_id' tree. and sets path name to 'name'.
2220 */
98d377a0
TH		 2221static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2222 u64 tree_id, u64 dirid, char *name)
2223{
2224 struct btrfs_root *root;
2225 struct btrfs_key key;
ac8e9819 2226 char *ptr;
98d377a0
TH		 2227 int ret = -1;
2228 int slot;
2229 int len;
2230 int total_len = 0;
2231 struct btrfs_inode_ref *iref;
2232 struct extent_buffer *l;
2233 struct btrfs_path *path;
2234
2235 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2236 name[0]='\0';
2237 return 0;
2238 }
2239
2240 path = btrfs_alloc_path();
2241 if (!path)
2242 return -ENOMEM;
2243
ac8e9819 2244 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2245
2246 key.objectid = tree_id;
2247 key.type = BTRFS_ROOT_ITEM_KEY;
2248 key.offset = (u64)-1;
2249 root = btrfs_read_fs_root_no_name(info, &key);
2250 if (IS_ERR(root)) {
efe120a0 2251 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
8ad6fcab
CM		 2252 ret = -ENOENT;
2253 goto out;
98d377a0
TH		 2254 }
2255
2256 key.objectid = dirid;
2257 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2258 key.offset = (u64)-1;
98d377a0 2259
67871254 2260 while (1) {
98d377a0
TH		 2261 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2262 if (ret < 0)
2263 goto out;
18674c6c
FDBM		 2264 else if (ret > 0) {
2265 ret = btrfs_previous_item(root, path, dirid,
2266 BTRFS_INODE_REF_KEY);
2267 if (ret < 0)
2268 goto out;
2269 else if (ret > 0) {
2270 ret = -ENOENT;
2271 goto out;
2272 }
2273 }
98d377a0
TH		 2274
2275 l = path->nodes[0];
2276 slot = path->slots[0];
2277 btrfs_item_key_to_cpu(l, &key, slot);
2278
98d377a0
TH		 2279 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2280 len = btrfs_inode_ref_name_len(l, iref);
2281 ptr -= len + 1;
2282 total_len += len + 1;
a696cf35
FDBM		 2283 if (ptr < name) {
2284 ret = -ENAMETOOLONG;
98d377a0 2285 goto out;
a696cf35 2286 }
98d377a0
TH		 2287
2288 *(ptr + len) = '/';
67871254 2289 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2290
2291 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2292 break;
2293
b3b4aa74 2294 btrfs_release_path(path);
98d377a0 2295 key.objectid = key.offset;
8ad6fcab 2296 key.offset = (u64)-1;
98d377a0 2297 dirid = key.objectid;
98d377a0 2298 }
77906a50 2299 memmove(name, ptr, total_len);
67871254 2300 name[total_len] = '\0';
98d377a0
TH		 2301 ret = 0;
2302out:
2303 btrfs_free_path(path);
ac8e9819
CM		 2304 return ret;
2305}
2306
2307static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2308 void __user *argp)
2309{
2310 struct btrfs_ioctl_ino_lookup_args *args;
2311 struct inode *inode;
2312 int ret;
2313
2314 if (!capable(CAP_SYS_ADMIN))
2315 return -EPERM;
2316
2354d08f
JL		 2317 args = memdup_user(argp, sizeof(*args));
2318 if (IS_ERR(args))
2319 return PTR_ERR(args);
c2b96929 2320
496ad9aa 2321 inode = file_inode(file);
ac8e9819 2322
1b53ac4d
CM		 2323 if (args->treeid == 0)
2324 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2325
ac8e9819
CM		 2326 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2327 args->treeid, args->objectid,
2328 args->name);
2329
2330 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2331 ret = -EFAULT;
2332
2333 kfree(args);
98d377a0
TH		 2334 return ret;
2335}
2336
76dda93c
YZ		 2337static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2338 void __user *arg)
2339{
54563d41 2340 struct dentry *parent = file->f_path.dentry;
76dda93c
YZ		 2341 struct dentry *dentry;
2342 struct inode *dir = parent->d_inode;
2343 struct inode *inode;
2344 struct btrfs_root *root = BTRFS_I(dir)->root;
2345 struct btrfs_root *dest = NULL;
2346 struct btrfs_ioctl_vol_args *vol_args;
2347 struct btrfs_trans_handle *trans;
c58aaad2 2348 struct btrfs_block_rsv block_rsv;
521e0546 2349 u64 root_flags;
c58aaad2 2350 u64 qgroup_reserved;
76dda93c
YZ		 2351 int namelen;
2352 int ret;
2353 int err = 0;
2354
76dda93c
YZ		 2355 vol_args = memdup_user(arg, sizeof(*vol_args));
2356 if (IS_ERR(vol_args))
2357 return PTR_ERR(vol_args);
2358
2359 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2360 namelen = strlen(vol_args->name);
2361 if (strchr(vol_args->name, '/') ||
2362 strncmp(vol_args->name, "..", namelen) == 0) {
2363 err = -EINVAL;
2364 goto out;
2365 }
2366
a561be71 2367 err = mnt_want_write_file(file);
76dda93c
YZ		 2368 if (err)
2369 goto out;
2370
521e0546 2371
5c50c9b8
DS		 2372 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2373 if (err == -EINTR)
e43f998e 2374 goto out_drop_write;
76dda93c
YZ		 2375 dentry = lookup_one_len(vol_args->name, parent, namelen);
2376 if (IS_ERR(dentry)) {
2377 err = PTR_ERR(dentry);
2378 goto out_unlock_dir;
2379 }
2380
2381 if (!dentry->d_inode) {
2382 err = -ENOENT;
2383 goto out_dput;
2384 }
2385
2386 inode = dentry->d_inode;
4260f7c7 2387 dest = BTRFS_I(inode)->root;
67871254 2388 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2389 /*
2390 * Regular user. Only allow this with a special mount
2391 * option, when the user has write+exec access to the
2392 * subvol root, and when rmdir(2) would have been
2393 * allowed.
2394 *
2395 * Note that this is _not_ check that the subvol is
2396 * empty or doesn't contain data that we wouldn't
2397 * otherwise be able to delete.
2398 *
2399 * Users who want to delete empty subvols should try
2400 * rmdir(2).
2401 */
2402 err = -EPERM;
2403 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2404 goto out_dput;
2405
2406 /*
2407 * Do not allow deletion if the parent dir is the same
2408 * as the dir to be deleted. That means the ioctl
2409 * must be called on the dentry referencing the root
2410 * of the subvol, not a random directory contained
2411 * within it.
2412 */
2413 err = -EINVAL;
2414 if (root == dest)
2415 goto out_dput;
2416
2417 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2418 if (err)
2419 goto out_dput;
4260f7c7
SW		 2420 }
2421
5c39da5b
MX		 2422 /* check if subvolume may be deleted by a user */
2423 err = btrfs_may_delete(dir, dentry, 1);
2424 if (err)
2425 goto out_dput;
2426
33345d01 2427 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2428 err = -EINVAL;
2429 goto out_dput;
2430 }
2431
76dda93c 2432 mutex_lock(&inode->i_mutex);
521e0546
DS		 2433
2434 /*
2435 * Don't allow to delete a subvolume with send in progress. This is
2436 * inside the i_mutex so the error handling that has to drop the bit
2437 * again is not run concurrently.
2438 */
2439 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2440 root_flags = btrfs_root_flags(&dest->root_item);
2441 if (dest->send_in_progress == 0) {
2442 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2443 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2444 spin_unlock(&dest->root_item_lock);
2445 } else {
2446 spin_unlock(&dest->root_item_lock);
2447 btrfs_warn(root->fs_info,
2448 "Attempt to delete subvolume %llu during send",
c55bfa67 2449 dest->root_key.objectid);
521e0546
DS		 2450 err = -EPERM;
2451 goto out_dput;
2452 }
2453
76dda93c
YZ		 2454 err = d_invalidate(dentry);
2455 if (err)
2456 goto out_unlock;
2457
2458 down_write(&root->fs_info->subvol_sem);
2459
2460 err = may_destroy_subvol(dest);
2461 if (err)
2462 goto out_up_write;
2463
c58aaad2
MX		 2464 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2465 /*
2466 * One for dir inode, two for dir entries, two for root
2467 * ref/backref.
2468 */
2469 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2470 5, &qgroup_reserved, true);
c58aaad2
MX		 2471 if (err)
2472 goto out_up_write;
2473
a22285a6
YZ		 2474 trans = btrfs_start_transaction(root, 0);
2475 if (IS_ERR(trans)) {
2476 err = PTR_ERR(trans);
c58aaad2 2477 goto out_release;
a22285a6 2478 }
c58aaad2
MX		 2479 trans->block_rsv = &block_rsv;
2480 trans->bytes_reserved = block_rsv.size;
a22285a6 2481
76dda93c
YZ		 2482 ret = btrfs_unlink_subvol(trans, root, dir,
2483 dest->root_key.objectid,
2484 dentry->d_name.name,
2485 dentry->d_name.len);
79787eaa
JM		 2486 if (ret) {
2487 err = ret;
2488 btrfs_abort_transaction(trans, root, ret);
2489 goto out_end_trans;
2490 }
76dda93c
YZ		 2491
2492 btrfs_record_root_in_trans(trans, dest);
2493
2494 memset(&dest->root_item.drop_progress, 0,
2495 sizeof(dest->root_item.drop_progress));
2496 dest->root_item.drop_level = 0;
2497 btrfs_set_root_refs(&dest->root_item, 0);
2498
27cdeb70 2499 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
d68fc57b
YZ		 2500 ret = btrfs_insert_orphan_item(trans,
2501 root->fs_info->tree_root,
2502 dest->root_key.objectid);
79787eaa
JM		 2503 if (ret) {
2504 btrfs_abort_transaction(trans, root, ret);
2505 err = ret;
2506 goto out_end_trans;
2507 }
d68fc57b 2508 }
dd5f9615
SB		 2509
2510 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2511 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2512 dest->root_key.objectid);
2513 if (ret && ret != -ENOENT) {
2514 btrfs_abort_transaction(trans, root, ret);
2515 err = ret;
2516 goto out_end_trans;
2517 }
2518 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2519 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2520 dest->root_item.received_uuid,
2521 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2522 dest->root_key.objectid);
2523 if (ret && ret != -ENOENT) {
2524 btrfs_abort_transaction(trans, root, ret);
2525 err = ret;
2526 goto out_end_trans;
2527 }
2528 }
2529
79787eaa 2530out_end_trans:
c58aaad2
MX		 2531 trans->block_rsv = NULL;
2532 trans->bytes_reserved = 0;
531cb13f 2533 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2534 if (ret && !err)
2535 err = ret;
76dda93c 2536 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2537out_release:
2538 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2539out_up_write:
2540 up_write(&root->fs_info->subvol_sem);
2541out_unlock:
521e0546
DS		 2542 if (err) {
2543 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2544 root_flags = btrfs_root_flags(&dest->root_item);
2545 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2546 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2547 spin_unlock(&dest->root_item_lock);
2548 }
76dda93c
YZ		 2549 mutex_unlock(&inode->i_mutex);
2550 if (!err) {
efefb143 2551 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2552 btrfs_invalidate_inodes(dest);
2553 d_delete(dentry);
61155aa0 2554 ASSERT(dest->send_in_progress == 0);
fa6ac876
LB		 2555
2556 /* the last ref */
2557 if (dest->cache_inode) {
2558 iput(dest->cache_inode);
2559 dest->cache_inode = NULL;
2560 }
76dda93c
YZ		 2561 }
2562out_dput:
2563 dput(dentry);
2564out_unlock_dir:
2565 mutex_unlock(&dir->i_mutex);
e43f998e 2566out_drop_write:
2a79f17e 2567 mnt_drop_write_file(file);
76dda93c
YZ		 2568out:
2569 kfree(vol_args);
2570 return err;
2571}
2572
1e701a32 2573static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2574{
496ad9aa 2575 struct inode *inode = file_inode(file);
f46b5a66 2576 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2577 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2578 int ret;
2579
25122d15
ID		 2580 ret = mnt_want_write_file(file);
2581 if (ret)
2582 return ret;
b83cc969 2583
25122d15
ID		 2584 if (btrfs_root_readonly(root)) {
2585 ret = -EROFS;
2586 goto out;
5ac00add 2587 }
f46b5a66
CH		 2588
2589 switch (inode->i_mode & S_IFMT) {
2590 case S_IFDIR:
e441d54d
CM		 2591 if (!capable(CAP_SYS_ADMIN)) {
2592 ret = -EPERM;
2593 goto out;
2594 }
de78b51a 2595 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2596 if (ret)
2597 goto out;
de78b51a 2598 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2599 break;
2600 case S_IFREG:
e441d54d
CM		 2601 if (!(file->f_mode & FMODE_WRITE)) {
2602 ret = -EINVAL;
2603 goto out;
2604 }
1e701a32
CM		 2605
2606 range = kzalloc(sizeof(*range), GFP_KERNEL);
2607 if (!range) {
2608 ret = -ENOMEM;
2609 goto out;
2610 }
2611
2612 if (argp) {
2613 if (copy_from_user(range, argp,
2614 sizeof(*range))) {
2615 ret = -EFAULT;
2616 kfree(range);
683be16e 2617 goto out;
1e701a32
CM		 2618 }
2619 /* compression requires us to start the IO */
2620 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2621 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2622 range->extent_thresh = (u32)-1;
2623 }
2624 } else {
2625 /* the rest are all set to zero by kzalloc */
2626 range->len = (u64)-1;
2627 }
496ad9aa 2628 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2629 range, 0, 0);
2630 if (ret > 0)
2631 ret = 0;
1e701a32 2632 kfree(range);
f46b5a66 2633 break;
8929ecfa
YZ		 2634 default:
2635 ret = -EINVAL;
f46b5a66 2636 }
e441d54d 2637out:
25122d15 2638 mnt_drop_write_file(file);
e441d54d 2639 return ret;
f46b5a66
CH		 2640}
2641
b2950863 2642static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2643{
2644 struct btrfs_ioctl_vol_args *vol_args;
2645 int ret;
2646
e441d54d
CM		 2647 if (!capable(CAP_SYS_ADMIN))
2648 return -EPERM;
2649
5ac00add
SB		 2650 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2651 1)) {
e57138b3 2652 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 2653 }
2654
5ac00add 2655 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2656 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2657 if (IS_ERR(vol_args)) {
2658 ret = PTR_ERR(vol_args);
2659 goto out;
2660 }
f46b5a66 2661
5516e595 2662 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2663 ret = btrfs_init_new_device(root, vol_args->name);
2664
f46b5a66 2665 kfree(vol_args);
c9e9f97b
ID		 2666out:
2667 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2668 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2669 return ret;
2670}
2671
da24927b 2672static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2673{
496ad9aa 2674 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2675 struct btrfs_ioctl_vol_args *vol_args;
2676 int ret;
2677
e441d54d
CM		 2678 if (!capable(CAP_SYS_ADMIN))
2679 return -EPERM;
2680
da24927b
MX		 2681 ret = mnt_want_write_file(file);
2682 if (ret)
2683 return ret;
c146afad 2684
dae7b665 2685 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2686 if (IS_ERR(vol_args)) {
2687 ret = PTR_ERR(vol_args);
2688 goto out;
2689 }
f46b5a66 2690
5516e595 2691 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2692
183860f6
AJ		 2693 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2694 1)) {
2695 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2696 goto out;
2697 }
2698
2699 mutex_lock(&root->fs_info->volume_mutex);
2700 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2701 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2702 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6
AJ		 2703
2704out:
2705 kfree(vol_args);
4ac20c70 2706 mnt_drop_write_file(file);
f46b5a66
CH		 2707 return ret;
2708}
2709
475f6387
JS		 2710static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2711{
027ed2f0 2712 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2713 struct btrfs_device *device;
2714 struct btrfs_device *next;
2715 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2716 int ret = 0;
475f6387 2717
027ed2f0
LZ		 2718 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2719 if (!fi_args)
2720 return -ENOMEM;
2721
f7171750 2722 mutex_lock(&fs_devices->device_list_mutex);
027ed2f0
LZ		 2723 fi_args->num_devices = fs_devices->num_devices;
2724 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387 2725
475f6387 2726 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2727 if (device->devid > fi_args->max_id)
2728 fi_args->max_id = device->devid;
475f6387
JS		 2729 }
2730 mutex_unlock(&fs_devices->device_list_mutex);
2731
80a773fb
DS		 2732 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2733 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2734 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2735
027ed2f0
LZ		 2736 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2737 ret = -EFAULT;
475f6387 2738
027ed2f0
LZ		 2739 kfree(fi_args);
2740 return ret;
475f6387
JS		 2741}
2742
2743static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2744{
2745 struct btrfs_ioctl_dev_info_args *di_args;
2746 struct btrfs_device *dev;
2747 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2748 int ret = 0;
2749 char *s_uuid = NULL;
475f6387 2750
475f6387
JS		 2751 di_args = memdup_user(arg, sizeof(*di_args));
2752 if (IS_ERR(di_args))
2753 return PTR_ERR(di_args);
2754
dd5f9615 2755 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2756 s_uuid = di_args->uuid;
2757
2758 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2759 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2760
2761 if (!dev) {
2762 ret = -ENODEV;
2763 goto out;
2764 }
2765
2766 di_args->devid = dev->devid;
2767 di_args->bytes_used = dev->bytes_used;
2768 di_args->total_bytes = dev->total_bytes;
2769 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2770 if (dev->name) {
606686ee
JB		 2771 struct rcu_string *name;
2772
2773 rcu_read_lock();
2774 name = rcu_dereference(dev->name);
2775 strncpy(di_args->path, name->str, sizeof(di_args->path));
2776 rcu_read_unlock();
a27202fb
JM		 2777 di_args->path[sizeof(di_args->path) - 1] = 0;
2778 } else {
99ba55ad 2779 di_args->path[0] = '\0';
a27202fb 2780 }
475f6387
JS		 2781
2782out:
55793c0d 2783 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2784 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2785 ret = -EFAULT;
2786
2787 kfree(di_args);
2788 return ret;
2789}
2790
416161db
MF		 2791static struct page *extent_same_get_page(struct inode *inode, u64 off)
2792{
2793 struct page *page;
2794 pgoff_t index;
2795 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2796
2797 index = off >> PAGE_CACHE_SHIFT;
2798
2799 page = grab_cache_page(inode->i_mapping, index);
2800 if (!page)
2801 return NULL;
2802
2803 if (!PageUptodate(page)) {
2804 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2805 0))
2806 return NULL;
2807 lock_page(page);
2808 if (!PageUptodate(page)) {
2809 unlock_page(page);
2810 page_cache_release(page);
2811 return NULL;
2812 }
2813 }
2814 unlock_page(page);
2815
2816 return page;
2817}
2818
77fe20dc
MF		 2819static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2820{
2821 /* do any pending delalloc/csum calc on src, one way or
2822 another, and lock file content */
2823 while (1) {
2824 struct btrfs_ordered_extent *ordered;
2825 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2826 ordered = btrfs_lookup_first_ordered_extent(inode,
2827 off + len - 1);
ff5df9b8
FM		 2828 if ((!ordered ||
2829 ordered->file_offset + ordered->len <= off ||
2830 ordered->file_offset >= off + len) &&
77fe20dc 2831 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
ff5df9b8
FM		 2832 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2833 if (ordered)
2834 btrfs_put_ordered_extent(ordered);
77fe20dc 2835 break;
ff5df9b8 2836 }
77fe20dc
MF		 2837 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2838 if (ordered)
2839 btrfs_put_ordered_extent(ordered);
2840 btrfs_wait_ordered_range(inode, off, len);
2841 }
2842}
2843
416161db
MF		 2844static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2845 struct inode *inode2, u64 loff2, u64 len)
2846{
2847 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2848 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2849
2850 mutex_unlock(&inode1->i_mutex);
2851 mutex_unlock(&inode2->i_mutex);
2852}
2853
2854static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2855 struct inode *inode2, u64 loff2, u64 len)
2856{
2857 if (inode1 < inode2) {
2858 swap(inode1, inode2);
2859 swap(loff1, loff2);
2860 }
2861
2862 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2863 lock_extent_range(inode1, loff1, len);
2864 if (inode1 != inode2) {
2865 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2866 lock_extent_range(inode2, loff2, len);
2867 }
2868}
2869
2870static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2871 u64 dst_loff, u64 len)
2872{
2873 int ret = 0;
2874 struct page *src_page, *dst_page;
2875 unsigned int cmp_len = PAGE_CACHE_SIZE;
2876 void *addr, *dst_addr;
2877
2878 while (len) {
2879 if (len < PAGE_CACHE_SIZE)
2880 cmp_len = len;
2881
2882 src_page = extent_same_get_page(src, loff);
2883 if (!src_page)
2884 return -EINVAL;
2885 dst_page = extent_same_get_page(dst, dst_loff);
2886 if (!dst_page) {
2887 page_cache_release(src_page);
2888 return -EINVAL;
2889 }
2890 addr = kmap_atomic(src_page);
2891 dst_addr = kmap_atomic(dst_page);
2892
2893 flush_dcache_page(src_page);
2894 flush_dcache_page(dst_page);
2895
2896 if (memcmp(addr, dst_addr, cmp_len))
2897 ret = BTRFS_SAME_DATA_DIFFERS;
2898
2899 kunmap_atomic(addr);
2900 kunmap_atomic(dst_addr);
2901 page_cache_release(src_page);
2902 page_cache_release(dst_page);
2903
2904 if (ret)
2905 break;
2906
2907 loff += cmp_len;
2908 dst_loff += cmp_len;
2909 len -= cmp_len;
2910 }
2911
2912 return ret;
2913}
2914
2915static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2916{
2917 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2918
2919 if (off + len > inode->i_size || off + len < off)
2920 return -EINVAL;
2921 /* Check that we are block aligned - btrfs_clone() requires this */
2922 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2923 return -EINVAL;
2924
2925 return 0;
2926}
2927
2928static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2929 struct inode *dst, u64 dst_loff)
2930{
2931 int ret;
2932
2933 /*
2934 * btrfs_clone() can't handle extents in the same file
2935 * yet. Once that works, we can drop this check and replace it
2936 * with a check for the same inode, but overlapping extents.
2937 */
2938 if (src == dst)
2939 return -EINVAL;
2940
2941 btrfs_double_lock(src, loff, dst, dst_loff, len);
2942
2943 ret = extent_same_check_offsets(src, loff, len);
2944 if (ret)
2945 goto out_unlock;
2946
2947 ret = extent_same_check_offsets(dst, dst_loff, len);
2948 if (ret)
2949 goto out_unlock;
2950
2951 /* don't make the dst file partly checksummed */
2952 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2953 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2954 ret = -EINVAL;
2955 goto out_unlock;
2956 }
2957
2958 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2959 if (ret == 0)
2960 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2961
2962out_unlock:
2963 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2964
2965 return ret;
2966}
2967
2968#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2969
2970static long btrfs_ioctl_file_extent_same(struct file *file,
1c1c8747 2971 struct btrfs_ioctl_same_args __user *argp)
416161db 2972{
cbf8b8ca
MF		 2973 struct btrfs_ioctl_same_args *same;
2974 struct btrfs_ioctl_same_extent_info *info;
1c1c8747 2975 struct inode *src = file_inode(file);
416161db
MF		 2976 u64 off;
2977 u64 len;
2978 int i;
2979 int ret;
cbf8b8ca 2980 unsigned long size;
416161db
MF		 2981 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2982 bool is_admin = capable(CAP_SYS_ADMIN);
1c1c8747 2983 u16 count;
416161db
MF		 2984
2985 if (!(file->f_mode & FMODE_READ))
2986 return -EINVAL;
2987
2988 ret = mnt_want_write_file(file);
2989 if (ret)
2990 return ret;
2991
1c1c8747 2992 if (get_user(count, &argp->dest_count)) {
416161db
MF		 2993 ret = -EFAULT;
2994 goto out;
2995 }
2996
1c1c8747 2997 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
cbf8b8ca 2998
1c1c8747 2999 same = memdup_user(argp, size);
cbf8b8ca 3000
229eed43
GB		 3001 if (IS_ERR(same)) {
3002 ret = PTR_ERR(same);
cbf8b8ca
MF		 3003 goto out;
3004 }
3005
3006 off = same->logical_offset;
3007 len = same->length;
416161db
MF		 3008
3009 /*
3010 * Limit the total length we will dedupe for each operation.
3011 * This is intended to bound the total time spent in this
3012 * ioctl to something sane.
3013 */
3014 if (len > BTRFS_MAX_DEDUPE_LEN)
3015 len = BTRFS_MAX_DEDUPE_LEN;
3016
3017 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3018 /*
3019 * Btrfs does not support blocksize < page_size. As a
3020 * result, btrfs_cmp_data() won't correctly handle
3021 * this situation without an update.
3022 */
3023 ret = -EINVAL;
3024 goto out;
3025 }
3026
3027 ret = -EISDIR;
3028 if (S_ISDIR(src->i_mode))
3029 goto out;
3030
3031 ret = -EACCES;
3032 if (!S_ISREG(src->i_mode))
3033 goto out;
3034
cbf8b8ca 3035 /* pre-format output fields to sane values */
1c1c8747 3036 for (i = 0; i < count; i++) {
cbf8b8ca
MF		 3037 same->info[i].bytes_deduped = 0ULL;
3038 same->info[i].status = 0;
3039 }
416161db 3040
1c1c8747
AV		 3041 for (i = 0, info = same->info; i < count; i++, info++) {
3042 struct inode *dst;
3043 struct fd dst_file = fdget(info->fd);
3044 if (!dst_file.file) {
cbf8b8ca 3045 info->status = -EBADF;
1c1c8747 3046 continue;
416161db 3047 }
1c1c8747 3048 dst = file_inode(dst_file.file);
416161db 3049
1c1c8747 3050 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
cbf8b8ca 3051 info->status = -EINVAL;
1c1c8747
AV		 3052 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3053 info->status = -EXDEV;
3054 } else if (S_ISDIR(dst->i_mode)) {
cbf8b8ca 3055 info->status = -EISDIR;
1c1c8747 3056 } else if (!S_ISREG(dst->i_mode)) {
cbf8b8ca 3057 info->status = -EACCES;
1c1c8747
AV		 3058 } else {
3059 info->status = btrfs_extent_same(src, off, len, dst,
3060 info->logical_offset);
3061 if (info->status == 0)
3062 info->bytes_deduped += len;
416161db 3063 }
1c1c8747 3064 fdput(dst_file);
416161db
MF		 3065 }
3066
cbf8b8ca
MF		 3067 ret = copy_to_user(argp, same, size);
3068 if (ret)
3069 ret = -EFAULT;
3070
416161db
MF		 3071out:
3072 mnt_drop_write_file(file);
3073 return ret;
3074}
3075
fcebe456
JB		 3076/* Helper to check and see if this root currently has a ref on the given disk
3077 * bytenr. If it does then we need to update the quota for this root. This
3078 * doesn't do anything if quotas aren't enabled.
3079 */
3080static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3081 u64 disko)
3082{
3083 struct seq_list tree_mod_seq_elem = {};
3084 struct ulist *roots;
3085 struct ulist_iterator uiter;
3086 struct ulist_node *root_node = NULL;
3087 int ret;
3088
3089 if (!root->fs_info->quota_enabled)
3090 return 1;
3091
3092 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3093 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3094 tree_mod_seq_elem.seq, &roots);
3095 if (ret < 0)
3096 goto out;
3097 ret = 0;
3098 ULIST_ITER_INIT(&uiter);
3099 while ((root_node = ulist_next(roots, &uiter))) {
3100 if (root_node->val == root->objectid) {
3101 ret = 1;
3102 break;
3103 }
3104 }
3105 ulist_free(roots);
3106out:
3107 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3108 return ret;
3109}
3110
f82a9901
FM		 3111static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3112 struct inode *inode,
3113 u64 endoff,
3114 const u64 destoff,
3115 const u64 olen)
3116{
3117 struct btrfs_root *root = BTRFS_I(inode)->root;
3118 int ret;
3119
3120 inode_inc_iversion(inode);
3121 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3122 /*
3123 * We round up to the block size at eof when determining which
3124 * extents to clone above, but shouldn't round up the file size.
3125 */
3126 if (endoff > destoff + olen)
3127 endoff = destoff + olen;
3128 if (endoff > inode->i_size)
3129 btrfs_i_size_write(inode, endoff);
3130
3131 ret = btrfs_update_inode(trans, root, inode);
3132 if (ret) {
3133 btrfs_abort_transaction(trans, root, ret);
3134 btrfs_end_transaction(trans, root);
3135 goto out;
3136 }
3137 ret = btrfs_end_transaction(trans, root);
3138out:
3139 return ret;
3140}
3141
7ffbb598
FM		 3142static void clone_update_extent_map(struct inode *inode,
3143 const struct btrfs_trans_handle *trans,
3144 const struct btrfs_path *path,
3145 struct btrfs_file_extent_item *fi,
3146 const u64 hole_offset,
3147 const u64 hole_len)
3148{
3149 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3150 struct extent_map *em;
3151 int ret;
3152
3153 em = alloc_extent_map();
3154 if (!em) {
3155 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3156 &BTRFS_I(inode)->runtime_flags);
3157 return;
3158 }
3159
3160 if (fi) {
3161 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3162 em->generation = -1;
3163 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3164 BTRFS_FILE_EXTENT_INLINE)
3165 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3166 &BTRFS_I(inode)->runtime_flags);
3167 } else {
3168 em->start = hole_offset;
3169 em->len = hole_len;
3170 em->ram_bytes = em->len;
3171 em->orig_start = hole_offset;
3172 em->block_start = EXTENT_MAP_HOLE;
3173 em->block_len = 0;
3174 em->orig_block_len = 0;
3175 em->compress_type = BTRFS_COMPRESS_NONE;
3176 em->generation = trans->transid;
3177 }
3178
3179 while (1) {
3180 write_lock(&em_tree->lock);
3181 ret = add_extent_mapping(em_tree, em, 1);
3182 write_unlock(&em_tree->lock);
3183 if (ret != -EEXIST) {
3184 free_extent_map(em);
3185 break;
3186 }
3187 btrfs_drop_extent_cache(inode, em->start,
3188 em->start + em->len - 1, 0);
3189 }
3190
3191 if (unlikely(ret))
3192 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3193 &BTRFS_I(inode)->runtime_flags);
3194}
3195
32b7c687
MF		 3196/**
3197 * btrfs_clone() - clone a range from inode file to another
3198 *
3199 * @src: Inode to clone from
3200 * @inode: Inode to clone to
3201 * @off: Offset within source to start clone from
3202 * @olen: Original length, passed by user, of range to clone
3203 * @olen_aligned: Block-aligned value of olen, extent_same uses
3204 * identical values here
3205 * @destoff: Offset within @inode to start clone
3206 */
3207static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901
FM		 3208 const u64 off, const u64 olen, const u64 olen_aligned,
3209 const u64 destoff)
f46b5a66 3210{
f46b5a66 3211 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3212 struct btrfs_path *path = NULL;
f46b5a66 3213 struct extent_buffer *leaf;
32b7c687
MF		 3214 struct btrfs_trans_handle *trans;
3215 char *buf = NULL;
ae01a0ab 3216 struct btrfs_key key;
f46b5a66
CH		 3217 u32 nritems;
3218 int slot;
ae01a0ab 3219 int ret;
fcebe456 3220 int no_quota;
f82a9901 3221 const u64 len = olen_aligned;
fcebe456 3222 u64 last_disko = 0;
f82a9901 3223 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3224
3225 ret = -ENOMEM;
3226 buf = vmalloc(btrfs_level_size(root, 0));
3227 if (!buf)
32b7c687 3228 return ret;
ae01a0ab
YZ		 3229
3230 path = btrfs_alloc_path();
3231 if (!path) {
3232 vfree(buf);
32b7c687 3233 return ret;
d525e8ab
LZ		 3234 }
3235
32b7c687 3236 path->reada = 2;
c5c9cd4d 3237 /* clone data */
33345d01 3238 key.objectid = btrfs_ino(src);
ae01a0ab 3239 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3240 key.offset = off;
f46b5a66
CH		 3241
3242 while (1) {
3243 /*
3244 * note the key will change type as we walk through the
3245 * tree.
3246 */
e4355f34 3247 path->leave_spinning = 1;
362a20c5
DS		 3248 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3249 0, 0);
f46b5a66
CH		 3250 if (ret < 0)
3251 goto out;
2c463823
FM		 3252 /*
3253 * First search, if no extent item that starts at offset off was
3254 * found but the previous item is an extent item, it's possible
3255 * it might overlap our target range, therefore process it.
3256 */
3257 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3258 btrfs_item_key_to_cpu(path->nodes[0], &key,
3259 path->slots[0] - 1);
3260 if (key.type == BTRFS_EXTENT_DATA_KEY)
3261 path->slots[0]--;
3262 }
f46b5a66 3263
ae01a0ab 3264 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3265process_slot:
fcebe456 3266 no_quota = 1;
ae01a0ab 3267 if (path->slots[0] >= nritems) {
362a20c5 3268 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3269 if (ret < 0)
3270 goto out;
3271 if (ret > 0)
3272 break;
ae01a0ab 3273 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3274 }
3275 leaf = path->nodes[0];
3276 slot = path->slots[0];
f46b5a66 3277
ae01a0ab 3278 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 3279 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 3280 key.objectid != btrfs_ino(src))
f46b5a66
CH		 3281 break;
3282
c5c9cd4d
SW		 3283 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
3284 struct btrfs_file_extent_item *extent;
3285 int type;
31840ae1
ZY		 3286 u32 size;
3287 struct btrfs_key new_key;
c5c9cd4d
SW		 3288 u64 disko = 0, diskl = 0;
3289 u64 datao = 0, datal = 0;
3290 u8 comp;
f82a9901 3291 u64 drop_start;
31840ae1 3292
c5c9cd4d
SW		 3293 extent = btrfs_item_ptr(leaf, slot,
3294 struct btrfs_file_extent_item);
3295 comp = btrfs_file_extent_compression(leaf, extent);
3296 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3297 if (type == BTRFS_FILE_EXTENT_REG ||
3298 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3299 disko = btrfs_file_extent_disk_bytenr(leaf,
3300 extent);
3301 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3302 extent);
c5c9cd4d 3303 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3304 datal = btrfs_file_extent_num_bytes(leaf,
3305 extent);
c5c9cd4d
SW		 3306 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3307 /* take upper bound, may be compressed */
3308 datal = btrfs_file_extent_ram_bytes(leaf,
3309 extent);
3310 }
31840ae1 3311
2c463823
FM		 3312 /*
3313 * The first search might have left us at an extent
3314 * item that ends before our target range's start, can
3315 * happen if we have holes and NO_HOLES feature enabled.
3316 */
3317 if (key.offset + datal <= off) {
e4355f34
FDBM		 3318 path->slots[0]++;
3319 goto process_slot;
2c463823
FM		 3320 } else if (key.offset >= off + len) {
3321 break;
e4355f34
FDBM		 3322 }
3323
3324 size = btrfs_item_size_nr(leaf, slot);
3325 read_extent_buffer(leaf, buf,
3326 btrfs_item_ptr_offset(leaf, slot),
3327 size);
3328
3329 btrfs_release_path(path);
3330 path->leave_spinning = 0;
c5c9cd4d 3331
31840ae1 3332 memcpy(&new_key, &key, sizeof(new_key));
33345d01 3333 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 3334 if (off <= key.offset)
3335 new_key.offset = key.offset + destoff - off;
3336 else
3337 new_key.offset = destoff;
31840ae1 3338
f82a9901
FM		 3339 /*
3340 * Deal with a hole that doesn't have an extent item
3341 * that represents it (NO_HOLES feature enabled).
3342 * This hole is either in the middle of the cloning
3343 * range or at the beginning (fully overlaps it or
3344 * partially overlaps it).
3345 */
3346 if (new_key.offset != last_dest_end)
3347 drop_start = last_dest_end;
3348 else
3349 drop_start = new_key.offset;
3350
b6f3409b
SW		 3351 /*
3352 * 1 - adjusting old extent (we may have to split it)
3353 * 1 - add new extent
3354 * 1 - inode update
3355 */
3356 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3357 if (IS_ERR(trans)) {
3358 ret = PTR_ERR(trans);
3359 goto out;
3360 }
3361
c8a894d7
CM		 3362 if (type == BTRFS_FILE_EXTENT_REG ||
3363 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3364 /*
3365 * a | --- range to clone ---| b
3366 * | ------------- extent ------------- |
3367 */
3368
93915584 3369 /* subtract range b */
d72c0842
LZ		 3370 if (key.offset + datal > off + len)
3371 datal = off + len - key.offset;
3372
93915584 3373 /* subtract range a */
a22285a6
YZ		 3374 if (off > key.offset) {
3375 datao += off - key.offset;
3376 datal -= off - key.offset;
3377 }
3378
5dc562c5 3379 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3380 drop_start,
a22285a6 3381 new_key.offset + datal,
2671485d 3382 1);
79787eaa 3383 if (ret) {
3f9e3df8 3384 if (ret != -EOPNOTSUPP)
00fdf13a
LB		 3385 btrfs_abort_transaction(trans,
3386 root, ret);
79787eaa
JM		 3387 btrfs_end_transaction(trans, root);
3388 goto out;
3389 }
a22285a6 3390
c5c9cd4d
SW		 3391 ret = btrfs_insert_empty_item(trans, root, path,
3392 &new_key, size);
79787eaa
JM		 3393 if (ret) {
3394 btrfs_abort_transaction(trans, root,
3395 ret);
3396 btrfs_end_transaction(trans, root);
3397 goto out;
3398 }
c5c9cd4d
SW		 3399
3400 leaf = path->nodes[0];
3401 slot = path->slots[0];
3402 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3403 btrfs_item_ptr_offset(leaf, slot),
3404 size);
ae01a0ab 3405
c5c9cd4d 3406 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3407 struct btrfs_file_extent_item);
c5c9cd4d 3408
c5c9cd4d
SW		 3409 /* disko == 0 means it's a hole */
3410 if (!disko)
3411 datao = 0;
c5c9cd4d
SW		 3412
3413 btrfs_set_file_extent_offset(leaf, extent,
3414 datao);
3415 btrfs_set_file_extent_num_bytes(leaf, extent,
3416 datal);
fcebe456
JB		 3417
3418 /*
3419 * We need to look up the roots that point at
3420 * this bytenr and see if the new root does. If
3421 * it does not we need to make sure we update
3422 * quotas appropriately.
3423 */
3424 if (disko && root != BTRFS_I(src)->root &&
3425 disko != last_disko) {
3426 no_quota = check_ref(trans, root,
3427 disko);
3428 if (no_quota < 0) {
3429 btrfs_abort_transaction(trans,
3430 root,
3431 ret);
3432 btrfs_end_transaction(trans,
3433 root);
3434 ret = no_quota;
3435 goto out;
3436 }
3437 }
3438
c5c9cd4d
SW		 3439 if (disko) {
3440 inode_add_bytes(inode, datal);
ae01a0ab 3441 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3442 disko, diskl, 0,
3443 root->root_key.objectid,
33345d01 3444 btrfs_ino(inode),
66d7e7f0 3445 new_key.offset - datao,
fcebe456 3446 no_quota);
79787eaa
JM		 3447 if (ret) {
3448 btrfs_abort_transaction(trans,
3449 root,
3450 ret);
3451 btrfs_end_transaction(trans,
3452 root);
3453 goto out;
3454
3455 }
f46b5a66 3456 }
c5c9cd4d
SW		 3457 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3458 u64 skip = 0;
3459 u64 trim = 0;
d3ecfcdf
LB		 3460 u64 aligned_end = 0;
3461
c5c9cd4d
SW		 3462 if (off > key.offset) {
3463 skip = off - key.offset;
3464 new_key.offset += skip;
3465 }
d397712b 3466
aa42ffd9
LB		 3467 if (key.offset + datal > off + len)
3468 trim = key.offset + datal - (off + len);
d397712b 3469
c5c9cd4d 3470 if (comp && (skip || trim)) {
c5c9cd4d 3471 ret = -EINVAL;
a22285a6 3472 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3473 goto out;
3474 }
3475 size -= skip + trim;
3476 datal -= skip + trim;
a22285a6 3477
d3ecfcdf
LB		 3478 aligned_end = ALIGN(new_key.offset + datal,
3479 root->sectorsize);
5dc562c5 3480 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3481 drop_start,
d3ecfcdf 3482 aligned_end,
2671485d 3483 1);
79787eaa 3484 if (ret) {
3f9e3df8 3485 if (ret != -EOPNOTSUPP)
3a29bc09
CM		 3486 btrfs_abort_transaction(trans,
3487 root, ret);
79787eaa
JM		 3488 btrfs_end_transaction(trans, root);
3489 goto out;
3490 }
a22285a6 3491
c5c9cd4d
SW		 3492 ret = btrfs_insert_empty_item(trans, root, path,
3493 &new_key, size);
79787eaa
JM		 3494 if (ret) {
3495 btrfs_abort_transaction(trans, root,
3496 ret);
3497 btrfs_end_transaction(trans, root);
3498 goto out;
3499 }
c5c9cd4d
SW		 3500
3501 if (skip) {
d397712b
CM		 3502 u32 start =
3503 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3504 memmove(buf+start, buf+start+skip,
3505 datal);
3506 }
3507
3508 leaf = path->nodes[0];
3509 slot = path->slots[0];
3510 write_extent_buffer(leaf, buf,
3511 btrfs_item_ptr_offset(leaf, slot),
3512 size);
3513 inode_add_bytes(inode, datal);
7ffbb598
FM		 3514 extent = btrfs_item_ptr(leaf, slot,
3515 struct btrfs_file_extent_item);
f46b5a66 3516 }
c5c9cd4d 3517
7ffbb598
FM		 3518 /* If we have an implicit hole (NO_HOLES feature). */
3519 if (drop_start < new_key.offset)
3520 clone_update_extent_map(inode, trans,
3521 path, NULL, drop_start,
3522 new_key.offset - drop_start);
3523
3524 clone_update_extent_map(inode, trans, path,
3525 extent, 0, 0);
3526
c5c9cd4d 3527 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3528 btrfs_release_path(path);
c5c9cd4d 3529
f82a9901
FM		 3530 last_dest_end = new_key.offset + datal;
3531 ret = clone_finish_inode_update(trans, inode,
3532 last_dest_end,
3533 destoff, olen);
3534 if (ret)
79787eaa 3535 goto out;
2c463823
FM		 3536 if (new_key.offset + datal >= destoff + len)
3537 break;
a22285a6 3538 }
b3b4aa74 3539 btrfs_release_path(path);
f46b5a66 3540 key.offset++;
f46b5a66 3541 }
f46b5a66 3542 ret = 0;
32b7c687 3543
f82a9901
FM		 3544 if (last_dest_end < destoff + len) {
3545 /*
3546 * We have an implicit hole (NO_HOLES feature is enabled) that
3547 * fully or partially overlaps our cloning range at its end.
3548 */
3549 btrfs_release_path(path);
3550
3551 /*
3552 * 1 - remove extent(s)
3553 * 1 - inode update
3554 */
3555 trans = btrfs_start_transaction(root, 2);
3556 if (IS_ERR(trans)) {
3557 ret = PTR_ERR(trans);
3558 goto out;
3559 }
3560 ret = btrfs_drop_extents(trans, root, inode,
3561 last_dest_end, destoff + len, 1);
3562 if (ret) {
3563 if (ret != -EOPNOTSUPP)
3564 btrfs_abort_transaction(trans, root, ret);
3565 btrfs_end_transaction(trans, root);
3566 goto out;
3567 }
3568 ret = clone_finish_inode_update(trans, inode, destoff + len,
3569 destoff, olen);
7ffbb598
FM		 3570 if (ret)
3571 goto out;
3572 clone_update_extent_map(inode, trans, path, NULL, last_dest_end,
3573 destoff + len - last_dest_end);
f82a9901
FM		 3574 }
3575
f46b5a66 3576out:
32b7c687
MF		 3577 btrfs_free_path(path);
3578 vfree(buf);
3579 return ret;
3580}
3581
3582static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3583 u64 off, u64 olen, u64 destoff)
3584{
54563d41 3585 struct inode *inode = file_inode(file);
32b7c687
MF		 3586 struct btrfs_root *root = BTRFS_I(inode)->root;
3587 struct fd src_file;
3588 struct inode *src;
3589 int ret;
3590 u64 len = olen;
3591 u64 bs = root->fs_info->sb->s_blocksize;
3592 int same_inode = 0;
3593
3594 /*
3595 * TODO:
3596 * - split compressed inline extents. annoying: we need to
3597 * decompress into destination's address_space (the file offset
3598 * may change, so source mapping won't do), then recompress (or
3599 * otherwise reinsert) a subrange.
00fdf13a
LB		 3600 *
3601 * - split destination inode's inline extents. The inline extents can
3602 * be either compressed or non-compressed.
32b7c687
MF		 3603 */
3604
3605 /* the destination must be opened for writing */
3606 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3607 return -EINVAL;
3608
3609 if (btrfs_root_readonly(root))
3610 return -EROFS;
3611
3612 ret = mnt_want_write_file(file);
3613 if (ret)
3614 return ret;
3615
3616 src_file = fdget(srcfd);
3617 if (!src_file.file) {
3618 ret = -EBADF;
3619 goto out_drop_write;
3620 }
3621
3622 ret = -EXDEV;
3623 if (src_file.file->f_path.mnt != file->f_path.mnt)
3624 goto out_fput;
3625
3626 src = file_inode(src_file.file);
3627
3628 ret = -EINVAL;
3629 if (src == inode)
3630 same_inode = 1;
3631
3632 /* the src must be open for reading */
3633 if (!(src_file.file->f_mode & FMODE_READ))
3634 goto out_fput;
3635
3636 /* don't make the dst file partly checksummed */
3637 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3638 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3639 goto out_fput;
3640
3641 ret = -EISDIR;
3642 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3643 goto out_fput;
3644
3645 ret = -EXDEV;
3646 if (src->i_sb != inode->i_sb)
3647 goto out_fput;
3648
3649 if (!same_inode) {
3650 if (inode < src) {
3651 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3652 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3653 } else {
3654 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3655 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3656 }
3657 } else {
3658 mutex_lock(&src->i_mutex);
3659 }
3660
3661 /* determine range to clone */
3662 ret = -EINVAL;
3663 if (off + len > src->i_size || off + len < off)
3664 goto out_unlock;
3665 if (len == 0)
3666 olen = len = src->i_size - off;
3667 /* if we extend to eof, continue to block boundary */
3668 if (off + len == src->i_size)
3669 len = ALIGN(src->i_size, bs) - off;
3670
3671 /* verify the end result is block aligned */
3672 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3673 !IS_ALIGNED(destoff, bs))
3674 goto out_unlock;
3675
3676 /* verify if ranges are overlapped within the same file */
3677 if (same_inode) {
3678 if (destoff + len > off && destoff < off + len)
3679 goto out_unlock;
3680 }
3681
3682 if (destoff > inode->i_size) {
3683 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3684 if (ret)
3685 goto out_unlock;
3686 }
3687
c125b8bf
FM		 3688 /*
3689 * Lock the target range too. Right after we replace the file extent
3690 * items in the fs tree (which now point to the cloned data), we might
3691 * have a worker replace them with extent items relative to a write
3692 * operation that was issued before this clone operation (i.e. confront
3693 * with inode.c:btrfs_finish_ordered_io).
3694 */
3695 if (same_inode) {
3696 u64 lock_start = min_t(u64, off, destoff);
3697 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
32b7c687 3698
c125b8bf
FM		 3699 lock_extent_range(src, lock_start, lock_len);
3700 } else {
3701 lock_extent_range(src, off, len);
3702 lock_extent_range(inode, destoff, len);
3703 }
32b7c687
MF		 3704
3705 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3706
c125b8bf
FM		 3707 if (same_inode) {
3708 u64 lock_start = min_t(u64, off, destoff);
3709 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3710
3711 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3712 } else {
3713 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3714 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3715 destoff + len - 1);
3716 }
3717 /*
3718 * Truncate page cache pages so that future reads will see the cloned
3719 * data immediately and not the previous data.
3720 */
3721 truncate_inode_pages_range(&inode->i_data, destoff,
3722 PAGE_CACHE_ALIGN(destoff + len) - 1);
f46b5a66 3723out_unlock:
c57c2b3e
FDBM		 3724 if (!same_inode) {
3725 if (inode < src) {
3726 mutex_unlock(&src->i_mutex);
3727 mutex_unlock(&inode->i_mutex);
3728 } else {
3729 mutex_unlock(&inode->i_mutex);
3730 mutex_unlock(&src->i_mutex);
3731 }
3732 } else {
3733 mutex_unlock(&src->i_mutex);
3734 }
f46b5a66 3735out_fput:
2903ff01 3736 fdput(src_file);
ab67b7c1 3737out_drop_write:
2a79f17e 3738 mnt_drop_write_file(file);
f46b5a66
CH		 3739 return ret;
3740}
3741
7a865e8a 3742static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3743{
3744 struct btrfs_ioctl_clone_range_args args;
3745
7a865e8a 3746 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3747 return -EFAULT;
3748 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3749 args.src_length, args.dest_offset);
3750}
3751
f46b5a66
CH		 3752/*
3753 * there are many ways the trans_start and trans_end ioctls can lead
3754 * to deadlocks. They should only be used by applications that
3755 * basically own the machine, and have a very in depth understanding
3756 * of all the possible deadlocks and enospc problems.
3757 */
b2950863 3758static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3759{
496ad9aa 3760 struct inode *inode = file_inode(file);
f46b5a66
CH		 3761 struct btrfs_root *root = BTRFS_I(inode)->root;
3762 struct btrfs_trans_handle *trans;
1ab86aed 3763 int ret;
f46b5a66 3764
1ab86aed 3765 ret = -EPERM;
df5b5520 3766 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3767 goto out;
df5b5520 3768
1ab86aed
SW		 3769 ret = -EINPROGRESS;
3770 if (file->private_data)
f46b5a66 3771 goto out;
9ca9ee09 3772
b83cc969
LZ		 3773 ret = -EROFS;
3774 if (btrfs_root_readonly(root))
3775 goto out;
3776
a561be71 3777 ret = mnt_want_write_file(file);
c146afad
YZ		 3778 if (ret)
3779 goto out;
3780
a4abeea4 3781 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3782
1ab86aed 3783 ret = -ENOMEM;
7a7eaa40 3784 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3785 if (IS_ERR(trans))
1ab86aed
SW		 3786 goto out_drop;
3787
3788 file->private_data = trans;
3789 return 0;
3790
3791out_drop:
a4abeea4 3792 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3793 mnt_drop_write_file(file);
f46b5a66 3794out:
f46b5a66
CH		 3795 return ret;
3796}
3797
6ef5ed0d
JB		 3798static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3799{
496ad9aa 3800 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3801 struct btrfs_root *root = BTRFS_I(inode)->root;
3802 struct btrfs_root *new_root;
3803 struct btrfs_dir_item *di;
3804 struct btrfs_trans_handle *trans;
3805 struct btrfs_path *path;
3806 struct btrfs_key location;
3807 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3808 u64 objectid = 0;
3809 u64 dir_id;
3c04ce01 3810 int ret;
6ef5ed0d
JB		 3811
3812 if (!capable(CAP_SYS_ADMIN))
3813 return -EPERM;
3814
3c04ce01
MX		 3815 ret = mnt_want_write_file(file);
3816 if (ret)
3817 return ret;
3818
3819 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3820 ret = -EFAULT;
3821 goto out;
3822 }
6ef5ed0d
JB		 3823
3824 if (!objectid)
1cecf579 3825 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3826
3827 location.objectid = objectid;
3828 location.type = BTRFS_ROOT_ITEM_KEY;
3829 location.offset = (u64)-1;
3830
3831 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3832 if (IS_ERR(new_root)) {
3833 ret = PTR_ERR(new_root);
3834 goto out;
3835 }
6ef5ed0d 3836
6ef5ed0d 3837 path = btrfs_alloc_path();
3c04ce01
MX		 3838 if (!path) {
3839 ret = -ENOMEM;
3840 goto out;
3841 }
6ef5ed0d
JB		 3842 path->leave_spinning = 1;
3843
3844 trans = btrfs_start_transaction(root, 1);
98d5dc13 3845 if (IS_ERR(trans)) {
6ef5ed0d 3846 btrfs_free_path(path);
3c04ce01
MX		 3847 ret = PTR_ERR(trans);
3848 goto out;
6ef5ed0d
JB		 3849 }
3850
6c41761f 3851 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3852 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3853 dir_id, "default", 7, 1);
cf1e99a4 3854 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3855 btrfs_free_path(path);
3856 btrfs_end_transaction(trans, root);
efe120a0
FH		 3857 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3858 "item, this isn't going to work");
3c04ce01
MX		 3859 ret = -ENOENT;
3860 goto out;
6ef5ed0d
JB		 3861 }
3862
3863 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3864 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3865 btrfs_mark_buffer_dirty(path->nodes[0]);
3866 btrfs_free_path(path);
3867
2b0ce2c2 3868 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3869 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3870out:
3871 mnt_drop_write_file(file);
3872 return ret;
6ef5ed0d
JB		 3873}
3874
5af3e8cc
SB		 3875void btrfs_get_block_group_info(struct list_head *groups_list,
3876 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3877{
3878 struct btrfs_block_group_cache *block_group;
3879
3880 space->total_bytes = 0;
3881 space->used_bytes = 0;
3882 space->flags = 0;
3883 list_for_each_entry(block_group, groups_list, list) {
3884 space->flags = block_group->flags;
3885 space->total_bytes += block_group->key.offset;
3886 space->used_bytes +=
3887 btrfs_block_group_used(&block_group->item);
3888 }
3889}
3890
48a3b636 3891static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3892{
3893 struct btrfs_ioctl_space_args space_args;
3894 struct btrfs_ioctl_space_info space;
3895 struct btrfs_ioctl_space_info *dest;
7fde62bf 3896 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3897 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3898 struct btrfs_space_info *info;
bf5fc093
JB		 3899 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3900 BTRFS_BLOCK_GROUP_SYSTEM,
3901 BTRFS_BLOCK_GROUP_METADATA,
3902 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3903 int num_types = 4;
7fde62bf 3904 int alloc_size;
1406e432 3905 int ret = 0;
51788b1b 3906 u64 slot_count = 0;
bf5fc093 3907 int i, c;
1406e432
JB		 3908
3909 if (copy_from_user(&space_args,
3910 (struct btrfs_ioctl_space_args __user *)arg,
3911 sizeof(space_args)))
3912 return -EFAULT;
3913
bf5fc093
JB		 3914 for (i = 0; i < num_types; i++) {
3915 struct btrfs_space_info *tmp;
3916
3917 info = NULL;
3918 rcu_read_lock();
3919 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3920 list) {
3921 if (tmp->flags == types[i]) {
3922 info = tmp;
3923 break;
3924 }
3925 }
3926 rcu_read_unlock();
3927
3928 if (!info)
3929 continue;
3930
3931 down_read(&info->groups_sem);
3932 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3933 if (!list_empty(&info->block_groups[c]))
3934 slot_count++;
3935 }
3936 up_read(&info->groups_sem);
3937 }
7fde62bf 3938
36523e95
DS		 3939 /*
3940 * Global block reserve, exported as a space_info
3941 */
3942 slot_count++;
3943
7fde62bf
CM		 3944 /* space_slots == 0 means they are asking for a count */
3945 if (space_args.space_slots == 0) {
3946 space_args.total_spaces = slot_count;
3947 goto out;
3948 }
bf5fc093 3949
51788b1b 3950 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3951
7fde62bf 3952 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3953
7fde62bf
CM		 3954 /* we generally have at most 6 or so space infos, one for each raid
3955 * level. So, a whole page should be more than enough for everyone
3956 */
3957 if (alloc_size > PAGE_CACHE_SIZE)
3958 return -ENOMEM;
3959
1406e432 3960 space_args.total_spaces = 0;
7fde62bf
CM		 3961 dest = kmalloc(alloc_size, GFP_NOFS);
3962 if (!dest)
3963 return -ENOMEM;
3964 dest_orig = dest;
1406e432 3965
7fde62bf 3966 /* now we have a buffer to copy into */
bf5fc093
JB		 3967 for (i = 0; i < num_types; i++) {
3968 struct btrfs_space_info *tmp;
3969
51788b1b
DR		 3970 if (!slot_count)
3971 break;
3972
bf5fc093
JB		 3973 info = NULL;
3974 rcu_read_lock();
3975 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3976 list) {
3977 if (tmp->flags == types[i]) {
3978 info = tmp;
3979 break;
3980 }
3981 }
3982 rcu_read_unlock();
7fde62bf 3983
bf5fc093
JB		 3984 if (!info)
3985 continue;
3986 down_read(&info->groups_sem);
3987 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3988 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3989 btrfs_get_block_group_info(
3990 &info->block_groups[c], &space);
bf5fc093
JB		 3991 memcpy(dest, &space, sizeof(space));
3992 dest++;
3993 space_args.total_spaces++;
51788b1b 3994 slot_count--;
bf5fc093 3995 }
51788b1b
DR		 3996 if (!slot_count)
3997 break;
bf5fc093
JB		 3998 }
3999 up_read(&info->groups_sem);
1406e432 4000 }
1406e432 4001
36523e95
DS		 4002 /*
4003 * Add global block reserve
4004 */
4005 if (slot_count) {
4006 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4007
4008 spin_lock(&block_rsv->lock);
4009 space.total_bytes = block_rsv->size;
4010 space.used_bytes = block_rsv->size - block_rsv->reserved;
4011 spin_unlock(&block_rsv->lock);
4012 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4013 memcpy(dest, &space, sizeof(space));
4014 space_args.total_spaces++;
4015 }
4016
2eec6c81 4017 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 4018 (arg + sizeof(struct btrfs_ioctl_space_args));
4019
4020 if (copy_to_user(user_dest, dest_orig, alloc_size))
4021 ret = -EFAULT;
4022
4023 kfree(dest_orig);
4024out:
4025 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 4026 ret = -EFAULT;
4027
4028 return ret;
4029}
4030
f46b5a66
CH		 4031/*
4032 * there are many ways the trans_start and trans_end ioctls can lead
4033 * to deadlocks. They should only be used by applications that
4034 * basically own the machine, and have a very in depth understanding
4035 * of all the possible deadlocks and enospc problems.
4036 */
4037long btrfs_ioctl_trans_end(struct file *file)
4038{
496ad9aa 4039 struct inode *inode = file_inode(file);
f46b5a66
CH		 4040 struct btrfs_root *root = BTRFS_I(inode)->root;
4041 struct btrfs_trans_handle *trans;
f46b5a66 4042
f46b5a66 4043 trans = file->private_data;
1ab86aed
SW		 4044 if (!trans)
4045 return -EINVAL;
b214107e 4046 file->private_data = NULL;
9ca9ee09 4047
1ab86aed
SW		 4048 btrfs_end_transaction(trans, root);
4049
a4abeea4 4050 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 4051
2a79f17e 4052 mnt_drop_write_file(file);
1ab86aed 4053 return 0;
f46b5a66
CH		 4054}
4055
9a8c28be
MX		 4056static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4057 void __user *argp)
46204592 4058{
46204592
SW		 4059 struct btrfs_trans_handle *trans;
4060 u64 transid;
db5b493a 4061 int ret;
46204592 4062
d4edf39b 4063 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4064 if (IS_ERR(trans)) {
4065 if (PTR_ERR(trans) != -ENOENT)
4066 return PTR_ERR(trans);
4067
4068 /* No running transaction, don't bother */
4069 transid = root->fs_info->last_trans_committed;
4070 goto out;
4071 }
46204592 4072 transid = trans->transid;
db5b493a 4073 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 4074 if (ret) {
4075 btrfs_end_transaction(trans, root);
db5b493a 4076 return ret;
8b2b2d3c 4077 }
ff7c1d33 4078out:
46204592
SW		 4079 if (argp)
4080 if (copy_to_user(argp, &transid, sizeof(transid)))
4081 return -EFAULT;
4082 return 0;
4083}
4084
9a8c28be
MX		 4085static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4086 void __user *argp)
46204592 4087{
46204592
SW		 4088 u64 transid;
4089
4090 if (argp) {
4091 if (copy_from_user(&transid, argp, sizeof(transid)))
4092 return -EFAULT;
4093 } else {
4094 transid = 0; /* current trans */
4095 }
4096 return btrfs_wait_for_commit(root, transid);
4097}
4098
b8e95489 4099static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4100{
496ad9aa 4101 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 4102 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4103 int ret;
475f6387
JS		 4104
4105 if (!capable(CAP_SYS_ADMIN))
4106 return -EPERM;
4107
4108 sa = memdup_user(arg, sizeof(*sa));
4109 if (IS_ERR(sa))
4110 return PTR_ERR(sa);
4111
b8e95489
MX		 4112 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4113 ret = mnt_want_write_file(file);
4114 if (ret)
4115 goto out;
4116 }
4117
aa1b8cd4 4118 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4119 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4120 0);
475f6387
JS		 4121
4122 if (copy_to_user(arg, sa, sizeof(*sa)))
4123 ret = -EFAULT;
4124
b8e95489
MX		 4125 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4126 mnt_drop_write_file(file);
4127out:
475f6387
JS		 4128 kfree(sa);
4129 return ret;
4130}
4131
4132static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4133{
4134 if (!capable(CAP_SYS_ADMIN))
4135 return -EPERM;
4136
aa1b8cd4 4137 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 4138}
4139
4140static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4141 void __user *arg)
4142{
4143 struct btrfs_ioctl_scrub_args *sa;
4144 int ret;
4145
4146 if (!capable(CAP_SYS_ADMIN))
4147 return -EPERM;
4148
4149 sa = memdup_user(arg, sizeof(*sa));
4150 if (IS_ERR(sa))
4151 return PTR_ERR(sa);
4152
4153 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4154
4155 if (copy_to_user(arg, sa, sizeof(*sa)))
4156 ret = -EFAULT;
4157
4158 kfree(sa);
4159 return ret;
4160}
4161
c11d2c23 4162static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 4163 void __user *arg)
c11d2c23
SB		 4164{
4165 struct btrfs_ioctl_get_dev_stats *sa;
4166 int ret;
4167
c11d2c23
SB		 4168 sa = memdup_user(arg, sizeof(*sa));
4169 if (IS_ERR(sa))
4170 return PTR_ERR(sa);
4171
b27f7c0c
DS		 4172 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4173 kfree(sa);
4174 return -EPERM;
4175 }
4176
4177 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 4178
4179 if (copy_to_user(arg, sa, sizeof(*sa)))
4180 ret = -EFAULT;
4181
4182 kfree(sa);
4183 return ret;
4184}
4185
3f6bcfbd
SB		 4186static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4187{
4188 struct btrfs_ioctl_dev_replace_args *p;
4189 int ret;
4190
4191 if (!capable(CAP_SYS_ADMIN))
4192 return -EPERM;
4193
4194 p = memdup_user(arg, sizeof(*p));
4195 if (IS_ERR(p))
4196 return PTR_ERR(p);
4197
4198 switch (p->cmd) {
4199 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
adfa97cb
ID		 4200 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4201 ret = -EROFS;
4202 goto out;
4203 }
3f6bcfbd
SB		 4204 if (atomic_xchg(
4205 &root->fs_info->mutually_exclusive_operation_running,
4206 1)) {
e57138b3 4207 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd
SB		 4208 } else {
4209 ret = btrfs_dev_replace_start(root, p);
4210 atomic_set(
4211 &root->fs_info->mutually_exclusive_operation_running,
4212 0);
4213 }
4214 break;
4215 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4216 btrfs_dev_replace_status(root->fs_info, p);
4217 ret = 0;
4218 break;
4219 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4220 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4221 break;
4222 default:
4223 ret = -EINVAL;
4224 break;
4225 }
4226
4227 if (copy_to_user(arg, p, sizeof(*p)))
4228 ret = -EFAULT;
adfa97cb 4229out:
3f6bcfbd
SB		 4230 kfree(p);
4231 return ret;
4232}
4233
d7728c96
JS		 4234static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4235{
4236 int ret = 0;
4237 int i;
740c3d22 4238 u64 rel_ptr;
d7728c96 4239 int size;
806468f8 4240 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4241 struct inode_fs_paths *ipath = NULL;
4242 struct btrfs_path *path;
4243
82b22ac8 4244 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4245 return -EPERM;
4246
4247 path = btrfs_alloc_path();
4248 if (!path) {
4249 ret = -ENOMEM;
4250 goto out;
4251 }
4252
4253 ipa = memdup_user(arg, sizeof(*ipa));
4254 if (IS_ERR(ipa)) {
4255 ret = PTR_ERR(ipa);
4256 ipa = NULL;
4257 goto out;
4258 }
4259
4260 size = min_t(u32, ipa->size, 4096);
4261 ipath = init_ipath(size, root, path);
4262 if (IS_ERR(ipath)) {
4263 ret = PTR_ERR(ipath);
4264 ipath = NULL;
4265 goto out;
4266 }
4267
4268 ret = paths_from_inode(ipa->inum, ipath);
4269 if (ret < 0)
4270 goto out;
4271
4272 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4273 rel_ptr = ipath->fspath->val[i] -
4274 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4275 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4276 }
4277
745c4d8e
JM		 4278 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4279 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 4280 if (ret) {
4281 ret = -EFAULT;
4282 goto out;
4283 }
4284
4285out:
4286 btrfs_free_path(path);
4287 free_ipath(ipath);
4288 kfree(ipa);
4289
4290 return ret;
4291}
4292
4293static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4294{
4295 struct btrfs_data_container *inodes = ctx;
4296 const size_t c = 3 * sizeof(u64);
4297
4298 if (inodes->bytes_left >= c) {
4299 inodes->bytes_left -= c;
4300 inodes->val[inodes->elem_cnt] = inum;
4301 inodes->val[inodes->elem_cnt + 1] = offset;
4302 inodes->val[inodes->elem_cnt + 2] = root;
4303 inodes->elem_cnt += 3;
4304 } else {
4305 inodes->bytes_missing += c - inodes->bytes_left;
4306 inodes->bytes_left = 0;
4307 inodes->elem_missed += 3;
4308 }
4309
4310 return 0;
4311}
4312
4313static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4314 void __user *arg)
4315{
4316 int ret = 0;
4317 int size;
d7728c96
JS		 4318 struct btrfs_ioctl_logical_ino_args *loi;
4319 struct btrfs_data_container *inodes = NULL;
4320 struct btrfs_path *path = NULL;
d7728c96
JS		 4321
4322 if (!capable(CAP_SYS_ADMIN))
4323 return -EPERM;
4324
4325 loi = memdup_user(arg, sizeof(*loi));
4326 if (IS_ERR(loi)) {
4327 ret = PTR_ERR(loi);
4328 loi = NULL;
4329 goto out;
4330 }
4331
4332 path = btrfs_alloc_path();
4333 if (!path) {
4334 ret = -ENOMEM;
4335 goto out;
4336 }
4337
425d17a2 4338 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 4339 inodes = init_data_container(size);
4340 if (IS_ERR(inodes)) {
4341 ret = PTR_ERR(inodes);
4342 inodes = NULL;
4343 goto out;
4344 }
4345
df031f07
LB		 4346 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4347 build_ino_list, inodes);
4348 if (ret == -EINVAL)
d7728c96
JS		 4349 ret = -ENOENT;
4350 if (ret < 0)
4351 goto out;
4352
745c4d8e
JM		 4353 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4354 (void *)(unsigned long)inodes, size);
d7728c96
JS		 4355 if (ret)
4356 ret = -EFAULT;
4357
4358out:
4359 btrfs_free_path(path);
425d17a2 4360 vfree(inodes);
d7728c96
JS		 4361 kfree(loi);
4362
4363 return ret;
4364}
4365
19a39dce 4366void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 4367 struct btrfs_ioctl_balance_args *bargs)
4368{
4369 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4370
4371 bargs->flags = bctl->flags;
4372
837d5b6e
ID		 4373 if (atomic_read(&fs_info->balance_running))
4374 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4375 if (atomic_read(&fs_info->balance_pause_req))
4376 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4377 if (atomic_read(&fs_info->balance_cancel_req))
4378 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4379
c9e9f97b
ID		 4380 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4381 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4382 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 4383
4384 if (lock) {
4385 spin_lock(&fs_info->balance_lock);
4386 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4387 spin_unlock(&fs_info->balance_lock);
4388 } else {
4389 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4390 }
c9e9f97b
ID		 4391}
4392
9ba1f6e4 4393static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4394{
496ad9aa 4395 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4396 struct btrfs_fs_info *fs_info = root->fs_info;
4397 struct btrfs_ioctl_balance_args *bargs;
4398 struct btrfs_balance_control *bctl;
ed0fb78f 4399 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4400 int ret;
4401
4402 if (!capable(CAP_SYS_ADMIN))
4403 return -EPERM;
4404
e54bfa31 4405 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4406 if (ret)
4407 return ret;
4408
ed0fb78f
ID		 4409again:
4410 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4411 mutex_lock(&fs_info->volume_mutex);
4412 mutex_lock(&fs_info->balance_mutex);
4413 need_unlock = true;
4414 goto locked;
4415 }
4416
4417 /*
4418 * mut. excl. ops lock is locked. Three possibilites:
4419 * (1) some other op is running
4420 * (2) balance is running
4421 * (3) balance is paused -- special case (think resume)
4422 */
c9e9f97b 4423 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4424 if (fs_info->balance_ctl) {
4425 /* this is either (2) or (3) */
4426 if (!atomic_read(&fs_info->balance_running)) {
4427 mutex_unlock(&fs_info->balance_mutex);
4428 if (!mutex_trylock(&fs_info->volume_mutex))
4429 goto again;
4430 mutex_lock(&fs_info->balance_mutex);
4431
4432 if (fs_info->balance_ctl &&
4433 !atomic_read(&fs_info->balance_running)) {
4434 /* this is (3) */
4435 need_unlock = false;
4436 goto locked;
4437 }
4438
4439 mutex_unlock(&fs_info->balance_mutex);
4440 mutex_unlock(&fs_info->volume_mutex);
4441 goto again;
4442 } else {
4443 /* this is (2) */
4444 mutex_unlock(&fs_info->balance_mutex);
4445 ret = -EINPROGRESS;
4446 goto out;
4447 }
4448 } else {
4449 /* this is (1) */
4450 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4451 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4452 goto out;
4453 }
4454
4455locked:
4456 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 4457
4458 if (arg) {
4459 bargs = memdup_user(arg, sizeof(*bargs));
4460 if (IS_ERR(bargs)) {
4461 ret = PTR_ERR(bargs);
ed0fb78f 4462 goto out_unlock;
c9e9f97b 4463 }
de322263
ID		 4464
4465 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4466 if (!fs_info->balance_ctl) {
4467 ret = -ENOTCONN;
4468 goto out_bargs;
4469 }
4470
4471 bctl = fs_info->balance_ctl;
4472 spin_lock(&fs_info->balance_lock);
4473 bctl->flags |= BTRFS_BALANCE_RESUME;
4474 spin_unlock(&fs_info->balance_lock);
4475
4476 goto do_balance;
4477 }
c9e9f97b
ID		 4478 } else {
4479 bargs = NULL;
4480 }
4481
ed0fb78f 4482 if (fs_info->balance_ctl) {
837d5b6e
ID		 4483 ret = -EINPROGRESS;
4484 goto out_bargs;
4485 }
4486
c9e9f97b
ID		 4487 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4488 if (!bctl) {
4489 ret = -ENOMEM;
4490 goto out_bargs;
4491 }
4492
4493 bctl->fs_info = fs_info;
4494 if (arg) {
4495 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4496 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4497 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4498
4499 bctl->flags = bargs->flags;
f43ffb60
ID		 4500 } else {
4501 /* balance everything - no filters */
4502 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4503 }
4504
de322263 4505do_balance:
c9e9f97b 4506 /*
ed0fb78f
ID		 4507 * Ownership of bctl and mutually_exclusive_operation_running
4508 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4509 * or, if restriper was paused all the way until unmount, in
4510 * free_fs_info. mutually_exclusive_operation_running is
4511 * cleared in __cancel_balance.
c9e9f97b 4512 */
ed0fb78f
ID		 4513 need_unlock = false;
4514
4515 ret = btrfs_balance(bctl, bargs);
4516
c9e9f97b
ID		 4517 if (arg) {
4518 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4519 ret = -EFAULT;
4520 }
4521
4522out_bargs:
4523 kfree(bargs);
ed0fb78f 4524out_unlock:
c9e9f97b
ID		 4525 mutex_unlock(&fs_info->balance_mutex);
4526 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 4527 if (need_unlock)
4528 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4529out:
e54bfa31 4530 mnt_drop_write_file(file);
c9e9f97b
ID		 4531 return ret;
4532}
4533
837d5b6e
ID		 4534static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4535{
4536 if (!capable(CAP_SYS_ADMIN))
4537 return -EPERM;
4538
4539 switch (cmd) {
4540 case BTRFS_BALANCE_CTL_PAUSE:
4541 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4542 case BTRFS_BALANCE_CTL_CANCEL:
4543 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4544 }
4545
4546 return -EINVAL;
4547}
4548
19a39dce
ID		 4549static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4550 void __user *arg)
4551{
4552 struct btrfs_fs_info *fs_info = root->fs_info;
4553 struct btrfs_ioctl_balance_args *bargs;
4554 int ret = 0;
4555
4556 if (!capable(CAP_SYS_ADMIN))
4557 return -EPERM;
4558
4559 mutex_lock(&fs_info->balance_mutex);
4560 if (!fs_info->balance_ctl) {
4561 ret = -ENOTCONN;
4562 goto out;
4563 }
4564
4565 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4566 if (!bargs) {
4567 ret = -ENOMEM;
4568 goto out;
4569 }
4570
4571 update_ioctl_balance_args(fs_info, 1, bargs);
4572
4573 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4574 ret = -EFAULT;
4575
4576 kfree(bargs);
4577out:
4578 mutex_unlock(&fs_info->balance_mutex);
4579 return ret;
4580}
4581
905b0dda 4582static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4583{
496ad9aa 4584 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4585 struct btrfs_ioctl_quota_ctl_args *sa;
4586 struct btrfs_trans_handle *trans = NULL;
4587 int ret;
4588 int err;
4589
4590 if (!capable(CAP_SYS_ADMIN))
4591 return -EPERM;
4592
905b0dda
MX		 4593 ret = mnt_want_write_file(file);
4594 if (ret)
4595 return ret;
5d13a37b
AJ		 4596
4597 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4598 if (IS_ERR(sa)) {
4599 ret = PTR_ERR(sa);
4600 goto drop_write;
4601 }
5d13a37b 4602
7708f029 4603 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4604 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4605 if (IS_ERR(trans)) {
4606 ret = PTR_ERR(trans);
4607 goto out;
5d13a37b
AJ		 4608 }
4609
4610 switch (sa->cmd) {
4611 case BTRFS_QUOTA_CTL_ENABLE:
4612 ret = btrfs_quota_enable(trans, root->fs_info);
4613 break;
4614 case BTRFS_QUOTA_CTL_DISABLE:
4615 ret = btrfs_quota_disable(trans, root->fs_info);
4616 break;
5d13a37b
AJ		 4617 default:
4618 ret = -EINVAL;
4619 break;
4620 }
4621
2f232036
JS		 4622 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4623 if (err && !ret)
4624 ret = err;
5d13a37b
AJ		 4625out:
4626 kfree(sa);
7708f029 4627 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4628drop_write:
4629 mnt_drop_write_file(file);
5d13a37b
AJ		 4630 return ret;
4631}
4632
905b0dda 4633static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4634{
496ad9aa 4635 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4636 struct btrfs_ioctl_qgroup_assign_args *sa;
4637 struct btrfs_trans_handle *trans;
4638 int ret;
4639 int err;
4640
4641 if (!capable(CAP_SYS_ADMIN))
4642 return -EPERM;
4643
905b0dda
MX		 4644 ret = mnt_want_write_file(file);
4645 if (ret)
4646 return ret;
5d13a37b
AJ		 4647
4648 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4649 if (IS_ERR(sa)) {
4650 ret = PTR_ERR(sa);
4651 goto drop_write;
4652 }
5d13a37b
AJ		 4653
4654 trans = btrfs_join_transaction(root);
4655 if (IS_ERR(trans)) {
4656 ret = PTR_ERR(trans);
4657 goto out;
4658 }
4659
4660 /* FIXME: check if the IDs really exist */
4661 if (sa->assign) {
4662 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4663 sa->src, sa->dst);
4664 } else {
4665 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4666 sa->src, sa->dst);
4667 }
4668
4669 err = btrfs_end_transaction(trans, root);
4670 if (err && !ret)
4671 ret = err;
4672
4673out:
4674 kfree(sa);
905b0dda
MX		 4675drop_write:
4676 mnt_drop_write_file(file);
5d13a37b
AJ		 4677 return ret;
4678}
4679
905b0dda 4680static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4681{
496ad9aa 4682 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4683 struct btrfs_ioctl_qgroup_create_args *sa;
4684 struct btrfs_trans_handle *trans;
4685 int ret;
4686 int err;
4687
4688 if (!capable(CAP_SYS_ADMIN))
4689 return -EPERM;
4690
905b0dda
MX		 4691 ret = mnt_want_write_file(file);
4692 if (ret)
4693 return ret;
5d13a37b
AJ		 4694
4695 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4696 if (IS_ERR(sa)) {
4697 ret = PTR_ERR(sa);
4698 goto drop_write;
4699 }
5d13a37b 4700
d86e56cf
MX		 4701 if (!sa->qgroupid) {
4702 ret = -EINVAL;
4703 goto out;
4704 }
4705
5d13a37b
AJ		 4706 trans = btrfs_join_transaction(root);
4707 if (IS_ERR(trans)) {
4708 ret = PTR_ERR(trans);
4709 goto out;
4710 }
4711
4712 /* FIXME: check if the IDs really exist */
4713 if (sa->create) {
4714 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4715 NULL);
4716 } else {
4717 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4718 }
4719
4720 err = btrfs_end_transaction(trans, root);
4721 if (err && !ret)
4722 ret = err;
4723
4724out:
4725 kfree(sa);
905b0dda
MX		 4726drop_write:
4727 mnt_drop_write_file(file);
5d13a37b
AJ		 4728 return ret;
4729}
4730
905b0dda 4731static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4732{
496ad9aa 4733 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4734 struct btrfs_ioctl_qgroup_limit_args *sa;
4735 struct btrfs_trans_handle *trans;
4736 int ret;
4737 int err;
4738 u64 qgroupid;
4739
4740 if (!capable(CAP_SYS_ADMIN))
4741 return -EPERM;
4742
905b0dda
MX		 4743 ret = mnt_want_write_file(file);
4744 if (ret)
4745 return ret;
5d13a37b
AJ		 4746
4747 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4748 if (IS_ERR(sa)) {
4749 ret = PTR_ERR(sa);
4750 goto drop_write;
4751 }
5d13a37b
AJ		 4752
4753 trans = btrfs_join_transaction(root);
4754 if (IS_ERR(trans)) {
4755 ret = PTR_ERR(trans);
4756 goto out;
4757 }
4758
4759 qgroupid = sa->qgroupid;
4760 if (!qgroupid) {
4761 /* take the current subvol as qgroup */
4762 qgroupid = root->root_key.objectid;
4763 }
4764
4765 /* FIXME: check if the IDs really exist */
4766 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4767
4768 err = btrfs_end_transaction(trans, root);
4769 if (err && !ret)
4770 ret = err;
4771
4772out:
4773 kfree(sa);
905b0dda
MX		 4774drop_write:
4775 mnt_drop_write_file(file);
5d13a37b
AJ		 4776 return ret;
4777}
4778
2f232036
JS		 4779static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4780{
6d0379ec 4781 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4782 struct btrfs_ioctl_quota_rescan_args *qsa;
4783 int ret;
4784
4785 if (!capable(CAP_SYS_ADMIN))
4786 return -EPERM;
4787
4788 ret = mnt_want_write_file(file);
4789 if (ret)
4790 return ret;
4791
4792 qsa = memdup_user(arg, sizeof(*qsa));
4793 if (IS_ERR(qsa)) {
4794 ret = PTR_ERR(qsa);
4795 goto drop_write;
4796 }
4797
4798 if (qsa->flags) {
4799 ret = -EINVAL;
4800 goto out;
4801 }
4802
4803 ret = btrfs_qgroup_rescan(root->fs_info);
4804
4805out:
4806 kfree(qsa);
4807drop_write:
4808 mnt_drop_write_file(file);
4809 return ret;
4810}
4811
4812static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4813{
6d0379ec 4814 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4815 struct btrfs_ioctl_quota_rescan_args *qsa;
4816 int ret = 0;
4817
4818 if (!capable(CAP_SYS_ADMIN))
4819 return -EPERM;
4820
4821 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4822 if (!qsa)
4823 return -ENOMEM;
4824
4825 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4826 qsa->flags = 1;
4827 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4828 }
4829
4830 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4831 ret = -EFAULT;
4832
4833 kfree(qsa);
4834 return ret;
4835}
4836
57254b6e
JS		 4837static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4838{
54563d41 4839 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
57254b6e
JS		 4840
4841 if (!capable(CAP_SYS_ADMIN))
4842 return -EPERM;
4843
4844 return btrfs_qgroup_wait_for_completion(root->fs_info);
4845}
4846
abccd00f
HM		 4847static long _btrfs_ioctl_set_received_subvol(struct file *file,
4848 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4849{
496ad9aa 4850 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4851 struct btrfs_root *root = BTRFS_I(inode)->root;
4852 struct btrfs_root_item *root_item = &root->root_item;
4853 struct btrfs_trans_handle *trans;
4854 struct timespec ct = CURRENT_TIME;
4855 int ret = 0;
dd5f9615 4856 int received_uuid_changed;
8ea05e3a 4857
bd60ea0f
DS		 4858 if (!inode_owner_or_capable(inode))
4859 return -EPERM;
4860
8ea05e3a
AB		 4861 ret = mnt_want_write_file(file);
4862 if (ret < 0)
4863 return ret;
4864
4865 down_write(&root->fs_info->subvol_sem);
4866
4867 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4868 ret = -EINVAL;
4869 goto out;
4870 }
4871
4872 if (btrfs_root_readonly(root)) {
4873 ret = -EROFS;
4874 goto out;
4875 }
4876
dd5f9615
SB		 4877 /*
4878 * 1 - root item
4879 * 2 - uuid items (received uuid + subvol uuid)
4880 */
4881 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4882 if (IS_ERR(trans)) {
4883 ret = PTR_ERR(trans);
4884 trans = NULL;
4885 goto out;
4886 }
4887
4888 sa->rtransid = trans->transid;
4889 sa->rtime.sec = ct.tv_sec;
4890 sa->rtime.nsec = ct.tv_nsec;
4891
dd5f9615
SB		 4892 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4893 BTRFS_UUID_SIZE);
4894 if (received_uuid_changed &&
4895 !btrfs_is_empty_uuid(root_item->received_uuid))
4896 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4897 root_item->received_uuid,
4898 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4899 root->root_key.objectid);
8ea05e3a
AB		 4900 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4901 btrfs_set_root_stransid(root_item, sa->stransid);
4902 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4903 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4904 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4905 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4906 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4907
4908 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4909 &root->root_key, &root->root_item);
4910 if (ret < 0) {
4911 btrfs_end_transaction(trans, root);
8ea05e3a 4912 goto out;
dd5f9615
SB		 4913 }
4914 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4915 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4916 sa->uuid,
4917 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4918 root->root_key.objectid);
4919 if (ret < 0 && ret != -EEXIST) {
4920 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 4921 goto out;
dd5f9615
SB		 4922 }
4923 }
4924 ret = btrfs_commit_transaction(trans, root);
4925 if (ret < 0) {
4926 btrfs_abort_transaction(trans, root, ret);
4927 goto out;
8ea05e3a
AB		 4928 }
4929
abccd00f
HM		 4930out:
4931 up_write(&root->fs_info->subvol_sem);
4932 mnt_drop_write_file(file);
4933 return ret;
4934}
4935
4936#ifdef CONFIG_64BIT
4937static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4938 void __user *arg)
4939{
4940 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4941 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4942 int ret = 0;
4943
4944 args32 = memdup_user(arg, sizeof(*args32));
4945 if (IS_ERR(args32)) {
4946 ret = PTR_ERR(args32);
4947 args32 = NULL;
4948 goto out;
4949 }
4950
4951 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
84dbeb87
DC		 4952 if (!args64) {
4953 ret = -ENOMEM;
abccd00f
HM		 4954 goto out;
4955 }
4956
4957 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4958 args64->stransid = args32->stransid;
4959 args64->rtransid = args32->rtransid;
4960 args64->stime.sec = args32->stime.sec;
4961 args64->stime.nsec = args32->stime.nsec;
4962 args64->rtime.sec = args32->rtime.sec;
4963 args64->rtime.nsec = args32->rtime.nsec;
4964 args64->flags = args32->flags;
4965
4966 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4967 if (ret)
4968 goto out;
4969
4970 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4971 args32->stransid = args64->stransid;
4972 args32->rtransid = args64->rtransid;
4973 args32->stime.sec = args64->stime.sec;
4974 args32->stime.nsec = args64->stime.nsec;
4975 args32->rtime.sec = args64->rtime.sec;
4976 args32->rtime.nsec = args64->rtime.nsec;
4977 args32->flags = args64->flags;
4978
4979 ret = copy_to_user(arg, args32, sizeof(*args32));
4980 if (ret)
4981 ret = -EFAULT;
4982
4983out:
4984 kfree(args32);
4985 kfree(args64);
4986 return ret;
4987}
4988#endif
4989
4990static long btrfs_ioctl_set_received_subvol(struct file *file,
4991 void __user *arg)
4992{
4993 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4994 int ret = 0;
4995
4996 sa = memdup_user(arg, sizeof(*sa));
4997 if (IS_ERR(sa)) {
4998 ret = PTR_ERR(sa);
4999 sa = NULL;
5000 goto out;
5001 }
5002
5003 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5004
5005 if (ret)
5006 goto out;
5007
8ea05e3a
AB		 5008 ret = copy_to_user(arg, sa, sizeof(*sa));
5009 if (ret)
5010 ret = -EFAULT;
5011
5012out:
5013 kfree(sa);
8ea05e3a
AB		 5014 return ret;
5015}
5016
867ab667 5017static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5018{
6d0379ec 5019 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 5020 size_t len;
867ab667 5021 int ret;
a1b83ac5
AJ		 5022 char label[BTRFS_LABEL_SIZE];
5023
5024 spin_lock(&root->fs_info->super_lock);
5025 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5026 spin_unlock(&root->fs_info->super_lock);
5027
5028 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5029
5030 if (len == BTRFS_LABEL_SIZE) {
efe120a0
FH		 5031 btrfs_warn(root->fs_info,
5032 "label is too long, return the first %zu bytes", --len);
867ab667 5033 }
5034
867ab667 5035 ret = copy_to_user(arg, label, len);
867ab667 5036
5037 return ret ? -EFAULT : 0;
5038}
5039
a8bfd4ab 5040static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5041{
6d0379ec 5042 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 5043 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5044 struct btrfs_trans_handle *trans;
5045 char label[BTRFS_LABEL_SIZE];
5046 int ret;
5047
5048 if (!capable(CAP_SYS_ADMIN))
5049 return -EPERM;
5050
5051 if (copy_from_user(label, arg, sizeof(label)))
5052 return -EFAULT;
5053
5054 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
efe120a0 5055 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
a8bfd4ab 5056 BTRFS_LABEL_SIZE - 1);
5057 return -EINVAL;
5058 }
5059
5060 ret = mnt_want_write_file(file);
5061 if (ret)
5062 return ret;
5063
a8bfd4ab 5064 trans = btrfs_start_transaction(root, 0);
5065 if (IS_ERR(trans)) {
5066 ret = PTR_ERR(trans);
5067 goto out_unlock;
5068 }
5069
a1b83ac5 5070 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 5071 strcpy(super_block->label, label);
a1b83ac5 5072 spin_unlock(&root->fs_info->super_lock);
d0270aca 5073 ret = btrfs_commit_transaction(trans, root);
a8bfd4ab 5074
5075out_unlock:
a8bfd4ab 5076 mnt_drop_write_file(file);
5077 return ret;
5078}
5079
2eaa055f
JM		 5080#define INIT_FEATURE_FLAGS(suffix) \
5081 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5082 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5083 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5084
5085static int btrfs_ioctl_get_supported_features(struct file *file,
5086 void __user *arg)
5087{
5088 static struct btrfs_ioctl_feature_flags features[3] = {
5089 INIT_FEATURE_FLAGS(SUPP),
5090 INIT_FEATURE_FLAGS(SAFE_SET),
5091 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5092 };
5093
5094 if (copy_to_user(arg, &features, sizeof(features)))
5095 return -EFAULT;
5096
5097 return 0;
5098}
5099
5100static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5101{
5102 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5103 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5104 struct btrfs_ioctl_feature_flags features;
5105
5106 features.compat_flags = btrfs_super_compat_flags(super_block);
5107 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5108 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5109
5110 if (copy_to_user(arg, &features, sizeof(features)))
5111 return -EFAULT;
5112
5113 return 0;
5114}
5115
3b02a68a
JM		 5116static int check_feature_bits(struct btrfs_root *root,
5117 enum btrfs_feature_set set,
2eaa055f
JM		 5118 u64 change_mask, u64 flags, u64 supported_flags,
5119 u64 safe_set, u64 safe_clear)
5120{
3b02a68a
JM		 5121 const char *type = btrfs_feature_set_names[set];
5122 char *names;
2eaa055f
JM		 5123 u64 disallowed, unsupported;
5124 u64 set_mask = flags & change_mask;
5125 u64 clear_mask = ~flags & change_mask;
5126
5127 unsupported = set_mask & ~supported_flags;
5128 if (unsupported) {
3b02a68a
JM		 5129 names = btrfs_printable_features(set, unsupported);
5130 if (names) {
5131 btrfs_warn(root->fs_info,
5132 "this kernel does not support the %s feature bit%s",
5133 names, strchr(names, ',') ? "s" : "");
5134 kfree(names);
5135 } else
5136 btrfs_warn(root->fs_info,
2eaa055f
JM		 5137 "this kernel does not support %s bits 0x%llx",
5138 type, unsupported);
5139 return -EOPNOTSUPP;
5140 }
5141
5142 disallowed = set_mask & ~safe_set;
5143 if (disallowed) {
3b02a68a
JM		 5144 names = btrfs_printable_features(set, disallowed);
5145 if (names) {
5146 btrfs_warn(root->fs_info,
5147 "can't set the %s feature bit%s while mounted",
5148 names, strchr(names, ',') ? "s" : "");
5149 kfree(names);
5150 } else
5151 btrfs_warn(root->fs_info,
2eaa055f
JM		 5152 "can't set %s bits 0x%llx while mounted",
5153 type, disallowed);
5154 return -EPERM;
5155 }
5156
5157 disallowed = clear_mask & ~safe_clear;
5158 if (disallowed) {
3b02a68a
JM		 5159 names = btrfs_printable_features(set, disallowed);
5160 if (names) {
5161 btrfs_warn(root->fs_info,
5162 "can't clear the %s feature bit%s while mounted",
5163 names, strchr(names, ',') ? "s" : "");
5164 kfree(names);
5165 } else
5166 btrfs_warn(root->fs_info,
2eaa055f
JM		 5167 "can't clear %s bits 0x%llx while mounted",
5168 type, disallowed);
5169 return -EPERM;
5170 }
5171
5172 return 0;
5173}
5174
5175#define check_feature(root, change_mask, flags, mask_base) \
3b02a68a 5176check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5177 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5178 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5179 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5180
5181static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5182{
5183 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5184 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5185 struct btrfs_ioctl_feature_flags flags[2];
5186 struct btrfs_trans_handle *trans;
5187 u64 newflags;
5188 int ret;
5189
5190 if (!capable(CAP_SYS_ADMIN))
5191 return -EPERM;
5192
5193 if (copy_from_user(flags, arg, sizeof(flags)))
5194 return -EFAULT;
5195
5196 /* Nothing to do */
5197 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5198 !flags[0].incompat_flags)
5199 return 0;
5200
5201 ret = check_feature(root, flags[0].compat_flags,
5202 flags[1].compat_flags, COMPAT);
5203 if (ret)
5204 return ret;
5205
5206 ret = check_feature(root, flags[0].compat_ro_flags,
5207 flags[1].compat_ro_flags, COMPAT_RO);
5208 if (ret)
5209 return ret;
5210
5211 ret = check_feature(root, flags[0].incompat_flags,
5212 flags[1].incompat_flags, INCOMPAT);
5213 if (ret)
5214 return ret;
5215
8051aa1a 5216 trans = btrfs_start_transaction(root, 0);
2eaa055f
JM		 5217 if (IS_ERR(trans))
5218 return PTR_ERR(trans);
5219
5220 spin_lock(&root->fs_info->super_lock);
5221 newflags = btrfs_super_compat_flags(super_block);
5222 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5223 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5224 btrfs_set_super_compat_flags(super_block, newflags);
5225
5226 newflags = btrfs_super_compat_ro_flags(super_block);
5227 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5228 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5229 btrfs_set_super_compat_ro_flags(super_block, newflags);
5230
5231 newflags = btrfs_super_incompat_flags(super_block);
5232 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5233 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5234 btrfs_set_super_incompat_flags(super_block, newflags);
5235 spin_unlock(&root->fs_info->super_lock);
5236
d0270aca 5237 return btrfs_commit_transaction(trans, root);
2eaa055f
JM		 5238}
5239
f46b5a66
CH		 5240long btrfs_ioctl(struct file *file, unsigned int
5241 cmd, unsigned long arg)
5242{
496ad9aa 5243 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 5244 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5245
5246 switch (cmd) {
6cbff00f
CH		 5247 case FS_IOC_GETFLAGS:
5248 return btrfs_ioctl_getflags(file, argp);
5249 case FS_IOC_SETFLAGS:
5250 return btrfs_ioctl_setflags(file, argp);
5251 case FS_IOC_GETVERSION:
5252 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5253 case FITRIM:
5254 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5255 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5256 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5257 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5258 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5259 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5260 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5261 case BTRFS_IOC_SUBVOL_CREATE_V2:
5262 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5263 case BTRFS_IOC_SNAP_DESTROY:
5264 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5265 case BTRFS_IOC_SUBVOL_GETFLAGS:
5266 return btrfs_ioctl_subvol_getflags(file, argp);
5267 case BTRFS_IOC_SUBVOL_SETFLAGS:
5268 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5269 case BTRFS_IOC_DEFAULT_SUBVOL:
5270 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5271 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5272 return btrfs_ioctl_defrag(file, NULL);
5273 case BTRFS_IOC_DEFRAG_RANGE:
5274 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5275 case BTRFS_IOC_RESIZE:
198605a8 5276 return btrfs_ioctl_resize(file, argp);
f46b5a66 5277 case BTRFS_IOC_ADD_DEV:
4bcabaa3 5278 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 5279 case BTRFS_IOC_RM_DEV:
da24927b 5280 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 5281 case BTRFS_IOC_FS_INFO:
5282 return btrfs_ioctl_fs_info(root, argp);
5283 case BTRFS_IOC_DEV_INFO:
5284 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 5285 case BTRFS_IOC_BALANCE:
9ba1f6e4 5286 return btrfs_ioctl_balance(file, NULL);
f46b5a66 5287 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 5288 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5289 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 5290 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 5291 case BTRFS_IOC_TRANS_START:
5292 return btrfs_ioctl_trans_start(file);
5293 case BTRFS_IOC_TRANS_END:
5294 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 5295 case BTRFS_IOC_TREE_SEARCH:
5296 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5297 case BTRFS_IOC_TREE_SEARCH_V2:
5298 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5299 case BTRFS_IOC_INO_LOOKUP:
5300 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5301 case BTRFS_IOC_INO_PATHS:
5302 return btrfs_ioctl_ino_to_path(root, argp);
5303 case BTRFS_IOC_LOGICAL_INO:
5304 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 5305 case BTRFS_IOC_SPACE_INFO:
5306 return btrfs_ioctl_space_info(root, argp);
9b199859
FDBM		 5307 case BTRFS_IOC_SYNC: {
5308 int ret;
5309
6c255e67 5310 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
9b199859
FDBM		 5311 if (ret)
5312 return ret;
5313 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5314 return ret;
5315 }
46204592 5316 case BTRFS_IOC_START_SYNC:
9a8c28be 5317 return btrfs_ioctl_start_sync(root, argp);
46204592 5318 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 5319 return btrfs_ioctl_wait_sync(root, argp);
475f6387 5320 case BTRFS_IOC_SCRUB:
b8e95489 5321 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 5322 case BTRFS_IOC_SCRUB_CANCEL:
5323 return btrfs_ioctl_scrub_cancel(root, argp);
5324 case BTRFS_IOC_SCRUB_PROGRESS:
5325 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 5326 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5327 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 5328 case BTRFS_IOC_BALANCE_CTL:
5329 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 5330 case BTRFS_IOC_BALANCE_PROGRESS:
5331 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 5332 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5333 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5334#ifdef CONFIG_64BIT
5335 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5336 return btrfs_ioctl_set_received_subvol_32(file, argp);
5337#endif
31db9f7c
AB		 5338 case BTRFS_IOC_SEND:
5339 return btrfs_ioctl_send(file, argp);
c11d2c23 5340 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 5341 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 5342 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5343 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5344 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5345 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5346 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5347 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5348 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5349 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5350 case BTRFS_IOC_QUOTA_RESCAN:
5351 return btrfs_ioctl_quota_rescan(file, argp);
5352 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5353 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5354 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5355 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 5356 case BTRFS_IOC_DEV_REPLACE:
5357 return btrfs_ioctl_dev_replace(root, argp);
867ab667 5358 case BTRFS_IOC_GET_FSLABEL:
5359 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5360 case BTRFS_IOC_SET_FSLABEL:
5361 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 5362 case BTRFS_IOC_FILE_EXTENT_SAME:
5363 return btrfs_ioctl_file_extent_same(file, argp);
2eaa055f
JM		 5364 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5365 return btrfs_ioctl_get_supported_features(file, argp);
5366 case BTRFS_IOC_GET_FEATURES:
5367 return btrfs_ioctl_get_features(file, argp);
5368 case BTRFS_IOC_SET_FEATURES:
5369 return btrfs_ioctl_set_features(file, argp);
f46b5a66
CH		 5370 }
5371
5372 return -ENOTTY;
5373}
A mirror of Linus' kernel repository