projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: Fix compile error when CONFIG_SECURITY is not set.
[linux-2.6-block.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
8ea05e3a 44#include <linux/uuid.h>
55e301fd 45#include <linux/btrfs.h>
416161db 46#include <linux/uaccess.h>
f46b5a66
CH		 47#include "ctree.h"
48#include "disk-io.h"
49#include "transaction.h"
50#include "btrfs_inode.h"
f46b5a66
CH		 51#include "print-tree.h"
52#include "volumes.h"
925baedd 53#include "locking.h"
581bb050 54#include "inode-map.h"
d7728c96 55#include "backref.h"
606686ee 56#include "rcu-string.h"
31db9f7c 57#include "send.h"
3f6bcfbd 58#include "dev-replace.h"
63541927 59#include "props.h"
3b02a68a 60#include "sysfs.h"
fcebe456 61#include "qgroup.h"
f46b5a66 62
abccd00f
HM		 63#ifdef CONFIG_64BIT
64/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
68 */
69struct btrfs_ioctl_timespec_32 {
70 __u64 sec;
71 __u32 nsec;
72} __attribute__ ((__packed__));
73
74struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
80 __u64 flags; /* in */
81 __u64 reserved[16]; /* in */
82} __attribute__ ((__packed__));
83
84#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
86#endif
87
88
416161db
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 334 } else {
78a017a2
FM		 335 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
336 if (ret && ret != -ENODATA)
337 goto out_drop;
ebcb904d 338 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 339 }
6cbff00f 340
4da6f1a3 341 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 342 if (IS_ERR(trans)) {
343 ret = PTR_ERR(trans);
344 goto out_drop;
345 }
6cbff00f 346
306424cc 347 btrfs_update_iflags(inode);
0c4d2d95 348 inode_inc_iversion(inode);
306424cc 349 inode->i_ctime = CURRENT_TIME;
6cbff00f 350 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 351
6cbff00f 352 btrfs_end_transaction(trans, root);
f062abf0
LZ		 353 out_drop:
354 if (ret) {
355 ip->flags = ip_oldflags;
356 inode->i_flags = i_oldflags;
357 }
6cbff00f 358
6cbff00f
CH		 359 out_unlock:
360 mutex_unlock(&inode->i_mutex);
e7848683 361 mnt_drop_write_file(file);
2d4e6f6a 362 return ret;
6cbff00f
CH		 363}
364
365static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
366{
496ad9aa 367 struct inode *inode = file_inode(file);
6cbff00f
CH		 368
369 return put_user(inode->i_generation, arg);
370}
f46b5a66 371
f7039b1d
LD		 372static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
373{
54563d41 374 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
f7039b1d
LD		 375 struct btrfs_device *device;
376 struct request_queue *q;
377 struct fstrim_range range;
378 u64 minlen = ULLONG_MAX;
379 u64 num_devices = 0;
815745cf 380 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 381 int ret;
382
383 if (!capable(CAP_SYS_ADMIN))
384 return -EPERM;
385
1f78160c
XG		 386 rcu_read_lock();
387 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
388 dev_list) {
f7039b1d
LD		 389 if (!device->bdev)
390 continue;
391 q = bdev_get_queue(device->bdev);
392 if (blk_queue_discard(q)) {
393 num_devices++;
394 minlen = min((u64)q->limits.discard_granularity,
395 minlen);
396 }
397 }
1f78160c 398 rcu_read_unlock();
f4c697e6 399
f7039b1d
LD		 400 if (!num_devices)
401 return -EOPNOTSUPP;
f7039b1d
LD		 402 if (copy_from_user(&range, arg, sizeof(range)))
403 return -EFAULT;
e515c18b
LC		 404 if (range.start > total_bytes ||
405 range.len < fs_info->sb->s_blocksize)
f4c697e6 406 return -EINVAL;
f7039b1d 407
f4c697e6 408 range.len = min(range.len, total_bytes - range.start);
f7039b1d 409 range.minlen = max(range.minlen, minlen);
815745cf 410 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 411 if (ret < 0)
412 return ret;
413
414 if (copy_to_user(arg, &range, sizeof(range)))
415 return -EFAULT;
416
417 return 0;
418}
419
dd5f9615
SB		 420int btrfs_is_empty_uuid(u8 *uuid)
421{
46e0f66a
CM		 422 int i;
423
424 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
425 if (uuid[i])
426 return 0;
427 }
428 return 1;
dd5f9615
SB		 429}
430
d5c12070 431static noinline int create_subvol(struct inode *dir,
cb8e7090 432 struct dentry *dentry,
72fd032e 433 char *name, int namelen,
6f72c7e2 434 u64 *async_transid,
8696c533 435 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 436{
437 struct btrfs_trans_handle *trans;
438 struct btrfs_key key;
439 struct btrfs_root_item root_item;
440 struct btrfs_inode_item *inode_item;
441 struct extent_buffer *leaf;
d5c12070 442 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 443 struct btrfs_root *new_root;
d5c12070 444 struct btrfs_block_rsv block_rsv;
8ea05e3a 445 struct timespec cur_time = CURRENT_TIME;
5662344b 446 struct inode *inode;
f46b5a66
CH		 447 int ret;
448 int err;
449 u64 objectid;
450 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 451 u64 index = 0;
d5c12070 452 u64 qgroup_reserved;
8ea05e3a 453 uuid_le new_uuid;
f46b5a66 454
581bb050 455 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 456 if (ret)
a22285a6 457 return ret;
6a912213 458
d5c12070 459 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 460 /*
d5c12070
MX		 461 * The same as the snapshot creation, please see the comment
462 * of create_snapshot().
9ed74f2d 463 */
d5c12070 464 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 465 8, &qgroup_reserved, false);
d5c12070
MX		 466 if (ret)
467 return ret;
468
469 trans = btrfs_start_transaction(root, 0);
470 if (IS_ERR(trans)) {
471 ret = PTR_ERR(trans);
de6e8200
LB		 472 btrfs_subvolume_release_metadata(root, &block_rsv,
473 qgroup_reserved);
474 return ret;
d5c12070
MX		 475 }
476 trans->block_rsv = &block_rsv;
477 trans->bytes_reserved = block_rsv.size;
f46b5a66 478
8696c533 479 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 480 if (ret)
481 goto fail;
482
4d75f8a9 483 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 484 if (IS_ERR(leaf)) {
485 ret = PTR_ERR(leaf);
486 goto fail;
487 }
f46b5a66 488
5d4f98a2 489 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 490 btrfs_set_header_bytenr(leaf, leaf->start);
491 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 492 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 493 btrfs_set_header_owner(leaf, objectid);
494
0a4e5586 495 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
f46b5a66 496 BTRFS_FSID_SIZE);
5d4f98a2 497 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
b308bc2f 498 btrfs_header_chunk_tree_uuid(leaf),
5d4f98a2 499 BTRFS_UUID_SIZE);
f46b5a66
CH		 500 btrfs_mark_buffer_dirty(leaf);
501
8ea05e3a
AB		 502 memset(&root_item, 0, sizeof(root_item));
503
f46b5a66 504 inode_item = &root_item.inode;
3cae210f
QW		 505 btrfs_set_stack_inode_generation(inode_item, 1);
506 btrfs_set_stack_inode_size(inode_item, 3);
507 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 508 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 509 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 510
3cae210f
QW		 511 btrfs_set_root_flags(&root_item, 0);
512 btrfs_set_root_limit(&root_item, 0);
513 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 514
f46b5a66 515 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 516 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 517 btrfs_set_root_level(&root_item, 0);
518 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 519 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 520 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 521
8ea05e3a
AB		 522 btrfs_set_root_generation_v2(&root_item,
523 btrfs_root_generation(&root_item));
524 uuid_le_gen(&new_uuid);
525 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 526 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
527 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 528 root_item.ctime = root_item.otime;
529 btrfs_set_root_ctransid(&root_item, trans->transid);
530 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 531
925baedd 532 btrfs_tree_unlock(leaf);
f46b5a66
CH		 533 free_extent_buffer(leaf);
534 leaf = NULL;
535
536 btrfs_set_root_dirid(&root_item, new_dirid);
537
538 key.objectid = objectid;
5d4f98a2 539 key.offset = 0;
962a298f 540 key.type = BTRFS_ROOT_ITEM_KEY;
f46b5a66
CH		 541 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
542 &root_item);
543 if (ret)
544 goto fail;
545
76dda93c
YZ		 546 key.offset = (u64)-1;
547 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 548 if (IS_ERR(new_root)) {
549 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
550 ret = PTR_ERR(new_root);
551 goto fail;
552 }
76dda93c
YZ		 553
554 btrfs_record_root_in_trans(trans, new_root);
555
63541927 556 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 557 if (ret) {
558 /* We potentially lose an unused inode item here */
79787eaa 559 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 560 goto fail;
561 }
562
f46b5a66
CH		 563 /*
564 * insert the directory item
565 */
3de4586c 566 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 567 if (ret) {
568 btrfs_abort_transaction(trans, root, ret);
569 goto fail;
570 }
3de4586c
CM		 571
572 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 573 name, namelen, dir, &key,
3de4586c 574 BTRFS_FT_DIR, index);
79787eaa
JM		 575 if (ret) {
576 btrfs_abort_transaction(trans, root, ret);
f46b5a66 577 goto fail;
79787eaa 578 }
0660b5af 579
52c26179
YZ		 580 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
581 ret = btrfs_update_inode(trans, root, dir);
582 BUG_ON(ret);
583
0660b5af 584 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 585 objectid, root->root_key.objectid,
33345d01 586 btrfs_ino(dir), index, name, namelen);
76dda93c 587 BUG_ON(ret);
f46b5a66 588
dd5f9615
SB		 589 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
590 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
591 objectid);
592 if (ret)
593 btrfs_abort_transaction(trans, root, ret);
594
f46b5a66 595fail:
d5c12070
MX		 596 trans->block_rsv = NULL;
597 trans->bytes_reserved = 0;
de6e8200
LB		 598 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
599
72fd032e
SW		 600 if (async_transid) {
601 *async_transid = trans->transid;
602 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 603 if (err)
604 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 605 } else {
606 err = btrfs_commit_transaction(trans, root);
607 }
f46b5a66
CH		 608 if (err && !ret)
609 ret = err;
1a65e24b 610
5662344b
TI		 611 if (!ret) {
612 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 613 if (IS_ERR(inode))
614 return PTR_ERR(inode);
5662344b
TI		 615 d_instantiate(dentry, inode);
616 }
f46b5a66
CH		 617 return ret;
618}
619
8257b2dc
MX		 620static void btrfs_wait_nocow_write(struct btrfs_root *root)
621{
622 s64 writers;
623 DEFINE_WAIT(wait);
624
625 do {
626 prepare_to_wait(&root->subv_writers->wait, &wait,
627 TASK_UNINTERRUPTIBLE);
628
629 writers = percpu_counter_sum(&root->subv_writers->counter);
630 if (writers)
631 schedule();
632
633 finish_wait(&root->subv_writers->wait, &wait);
634 } while (writers);
635}
636
e9662f70
MX		 637static int create_snapshot(struct btrfs_root *root, struct inode *dir,
638 struct dentry *dentry, char *name, int namelen,
639 u64 *async_transid, bool readonly,
640 struct btrfs_qgroup_inherit *inherit)
f46b5a66 641{
2e4bfab9 642 struct inode *inode;
f46b5a66
CH		 643 struct btrfs_pending_snapshot *pending_snapshot;
644 struct btrfs_trans_handle *trans;
2e4bfab9 645 int ret;
f46b5a66 646
27cdeb70 647 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
f46b5a66
CH		 648 return -EINVAL;
649
8257b2dc 650 atomic_inc(&root->will_be_snapshoted);
4e857c58 651 smp_mb__after_atomic();
8257b2dc
MX		 652 btrfs_wait_nocow_write(root);
653
6a03843d
MX		 654 ret = btrfs_start_delalloc_inodes(root, 0);
655 if (ret)
8257b2dc 656 goto out;
6a03843d 657
b0244199 658 btrfs_wait_ordered_extents(root, -1);
6a03843d 659
3de4586c 660 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
8257b2dc
MX		 661 if (!pending_snapshot) {
662 ret = -ENOMEM;
663 goto out;
664 }
a22285a6 665
66d8f3dd
MX		 666 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
667 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 668 /*
669 * 1 - parent dir inode
670 * 2 - dir entries
671 * 1 - root item
672 * 2 - root ref/backref
673 * 1 - root of snapshot
dd5f9615 674 * 1 - UUID item
d5c12070
MX		 675 */
676 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 677 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 678 &pending_snapshot->qgroup_reserved,
679 false);
d5c12070 680 if (ret)
8257b2dc 681 goto free;
d5c12070 682
3de4586c 683 pending_snapshot->dentry = dentry;
f46b5a66 684 pending_snapshot->root = root;
b83cc969 685 pending_snapshot->readonly = readonly;
e9662f70 686 pending_snapshot->dir = dir;
8696c533 687 pending_snapshot->inherit = inherit;
a22285a6 688
d5c12070 689 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 690 if (IS_ERR(trans)) {
691 ret = PTR_ERR(trans);
692 goto fail;
693 }
694
8351583e 695 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 696 list_add(&pending_snapshot->list,
697 &trans->transaction->pending_snapshots);
8351583e 698 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 699 if (async_transid) {
700 *async_transid = trans->transid;
701 ret = btrfs_commit_transaction_async(trans,
702 root->fs_info->extent_root, 1);
00d71c9c
MX		 703 if (ret)
704 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 705 } else {
706 ret = btrfs_commit_transaction(trans,
707 root->fs_info->extent_root);
708 }
aec8030a 709 if (ret)
c37b2b62 710 goto fail;
a22285a6
YZ		 711
712 ret = pending_snapshot->error;
713 if (ret)
714 goto fail;
715
2fbe8c8a 716 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 717 if (IS_ERR(inode)) {
718 ret = PTR_ERR(inode);
719 goto fail;
720 }
5662344b 721
2e4bfab9
YZ		 722 d_instantiate(dentry, inode);
723 ret = 0;
724fail:
d5c12070
MX		 725 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
726 &pending_snapshot->block_rsv,
727 pending_snapshot->qgroup_reserved);
8257b2dc 728free:
a22285a6 729 kfree(pending_snapshot);
8257b2dc
MX		 730out:
731 atomic_dec(&root->will_be_snapshoted);
f46b5a66
CH		 732 return ret;
733}
734
4260f7c7
SW		 735/* copy of check_sticky in fs/namei.c()
736* It's inline, so penalty for filesystems that don't use sticky bit is
737* minimal.
738*/
739static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
740{
2f2f43d3 741 kuid_t fsuid = current_fsuid();
4260f7c7
SW		 742
743 if (!(dir->i_mode & S_ISVTX))
744 return 0;
2f2f43d3 745 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 746 return 0;
2f2f43d3 747 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW		 748 return 0;
749 return !capable(CAP_FOWNER);
750}
751
752/* copy of may_delete in fs/namei.c()
753 * Check whether we can remove a link victim from directory dir, check
754 * whether the type of victim is right.
755 * 1. We can't do it if dir is read-only (done in permission())
756 * 2. We should have write and exec permissions on dir
757 * 3. We can't remove anything from append-only dir
758 * 4. We can't do anything with immutable dir (done in permission())
759 * 5. If the sticky bit on dir is set we should either
760 * a. be owner of dir, or
761 * b. be owner of victim, or
762 * c. have CAP_FOWNER capability
763 * 6. If the victim is append-only or immutable we can't do antyhing with
764 * links pointing to it.
765 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
766 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
767 * 9. We can't remove a root or mountpoint.
768 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
769 * nfs_async_unlink().
770 */
771
67871254 772static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 773{
774 int error;
775
776 if (!victim->d_inode)
777 return -ENOENT;
778
779 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 780 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 781
782 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
783 if (error)
784 return error;
785 if (IS_APPEND(dir))
786 return -EPERM;
787 if (btrfs_check_sticky(dir, victim->d_inode)||
788 IS_APPEND(victim->d_inode)||
789 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
790 return -EPERM;
791 if (isdir) {
792 if (!S_ISDIR(victim->d_inode->i_mode))
793 return -ENOTDIR;
794 if (IS_ROOT(victim))
795 return -EBUSY;
796 } else if (S_ISDIR(victim->d_inode->i_mode))
797 return -EISDIR;
798 if (IS_DEADDIR(dir))
799 return -ENOENT;
800 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
801 return -EBUSY;
802 return 0;
803}
804
cb8e7090
CH		 805/* copy of may_create in fs/namei.c() */
806static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
807{
808 if (child->d_inode)
809 return -EEXIST;
810 if (IS_DEADDIR(dir))
811 return -ENOENT;
812 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
813}
814
815/*
816 * Create a new subvolume below @parent. This is largely modeled after
817 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
818 * inside this filesystem so it's quite a bit simpler.
819 */
76dda93c
YZ		 820static noinline int btrfs_mksubvol(struct path *parent,
821 char *name, int namelen,
72fd032e 822 struct btrfs_root *snap_src,
6f72c7e2 823 u64 *async_transid, bool readonly,
8696c533 824 struct btrfs_qgroup_inherit *inherit)
cb8e7090 825{
76dda93c 826 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 827 struct dentry *dentry;
828 int error;
829
5c50c9b8
DS		 830 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
831 if (error == -EINTR)
832 return error;
cb8e7090
CH		 833
834 dentry = lookup_one_len(name, parent->dentry, namelen);
835 error = PTR_ERR(dentry);
836 if (IS_ERR(dentry))
837 goto out_unlock;
838
839 error = -EEXIST;
840 if (dentry->d_inode)
841 goto out_dput;
842
76dda93c 843 error = btrfs_may_create(dir, dentry);
cb8e7090 844 if (error)
a874a63e 845 goto out_dput;
cb8e7090 846
9c52057c
CM		 847 /*
848 * even if this name doesn't exist, we may get hash collisions.
849 * check for them now when we can safely fail
850 */
851 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
852 dir->i_ino, name,
853 namelen);
854 if (error)
855 goto out_dput;
856
76dda93c
YZ		 857 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
858
859 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
860 goto out_up_read;
861
3de4586c 862 if (snap_src) {
e9662f70 863 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 864 async_transid, readonly, inherit);
3de4586c 865 } else {
d5c12070
MX		 866 error = create_subvol(dir, dentry, name, namelen,
867 async_transid, inherit);
3de4586c 868 }
76dda93c
YZ		 869 if (!error)
870 fsnotify_mkdir(dir, dentry);
871out_up_read:
872 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 873out_dput:
874 dput(dentry);
875out_unlock:
76dda93c 876 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 877 return error;
878}
879
4cb5300b
CM		 880/*
881 * When we're defragging a range, we don't want to kick it off again
882 * if it is really just waiting for delalloc to send it down.
883 * If we find a nice big extent or delalloc range for the bytes in the
884 * file you want to defrag, we return 0 to let you know to skip this
885 * part of the file
886 */
aab110ab 887static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
4cb5300b
CM		 888{
889 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
890 struct extent_map *em = NULL;
891 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
892 u64 end;
893
894 read_lock(&em_tree->lock);
895 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
896 read_unlock(&em_tree->lock);
897
898 if (em) {
899 end = extent_map_end(em);
900 free_extent_map(em);
901 if (end - offset > thresh)
902 return 0;
903 }
904 /* if we already have a nice delalloc here, just stop */
905 thresh /= 2;
906 end = count_range_bits(io_tree, &offset, offset + thresh,
907 thresh, EXTENT_DELALLOC, 1);
908 if (end >= thresh)
909 return 0;
910 return 1;
911}
912
913/*
914 * helper function to walk through a file and find extents
915 * newer than a specific transid, and smaller than thresh.
916 *
917 * This is used by the defragging code to find new and small
918 * extents
919 */
920static int find_new_extents(struct btrfs_root *root,
921 struct inode *inode, u64 newer_than,
aab110ab 922 u64 *off, u32 thresh)
4cb5300b
CM		 923{
924 struct btrfs_path *path;
925 struct btrfs_key min_key;
4cb5300b
CM		 926 struct extent_buffer *leaf;
927 struct btrfs_file_extent_item *extent;
928 int type;
929 int ret;
a4689d2b 930 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 931
932 path = btrfs_alloc_path();
933 if (!path)
934 return -ENOMEM;
935
a4689d2b 936 min_key.objectid = ino;
4cb5300b
CM		 937 min_key.type = BTRFS_EXTENT_DATA_KEY;
938 min_key.offset = *off;
939
67871254 940 while (1) {
6174d3cb 941 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 942 if (ret != 0)
943 goto none;
f094c9bd 944process_slot:
a4689d2b 945 if (min_key.objectid != ino)
4cb5300b
CM		 946 goto none;
947 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
948 goto none;
949
950 leaf = path->nodes[0];
951 extent = btrfs_item_ptr(leaf, path->slots[0],
952 struct btrfs_file_extent_item);
953
954 type = btrfs_file_extent_type(leaf, extent);
955 if (type == BTRFS_FILE_EXTENT_REG &&
956 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
957 check_defrag_in_cache(inode, min_key.offset, thresh)) {
958 *off = min_key.offset;
959 btrfs_free_path(path);
960 return 0;
961 }
962
f094c9bd
FM		 963 path->slots[0]++;
964 if (path->slots[0] < btrfs_header_nritems(leaf)) {
965 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
966 goto process_slot;
967 }
968
4cb5300b
CM		 969 if (min_key.offset == (u64)-1)
970 goto none;
971
972 min_key.offset++;
973 btrfs_release_path(path);
974 }
975none:
976 btrfs_free_path(path);
977 return -ENOENT;
978}
979
6c282eb4 980static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 981{
982 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 983 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
984 struct extent_map *em;
985 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 986
6c282eb4
LZ		 987 /*
988 * hopefully we have this extent in the tree already, try without
989 * the full extent lock
990 */
17ce6ef8 991 read_lock(&em_tree->lock);
6c282eb4 992 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 993 read_unlock(&em_tree->lock);
994
6c282eb4 995 if (!em) {
308d9800
FM		 996 struct extent_state *cached = NULL;
997 u64 end = start + len - 1;
998
6c282eb4 999 /* get the big lock and read metadata off disk */
308d9800 1000 lock_extent_bits(io_tree, start, end, 0, &cached);
6c282eb4 1001 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
308d9800 1002 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
6c282eb4
LZ		 1003
1004 if (IS_ERR(em))
1005 return NULL;
1006 }
1007
1008 return em;
1009}
17ce6ef8 1010
6c282eb4
LZ		 1011static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1012{
1013 struct extent_map *next;
1014 bool ret = true;
1015
1016 /* this is the last extent */
1017 if (em->start + em->len >= i_size_read(inode))
1018 return false;
1019
1020 next = defrag_lookup_extent(inode, em->start + em->len);
e9512d72
CM		 1021 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1022 ret = false;
1023 else if ((em->block_start + em->block_len == next->block_start) &&
1024 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
6c282eb4
LZ		 1025 ret = false;
1026
1027 free_extent_map(next);
17ce6ef8
LB		 1028 return ret;
1029}
1030
aab110ab 1031static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
a43a2111
AM		 1032 u64 *last_len, u64 *skip, u64 *defrag_end,
1033 int compress)
940100a4 1034{
6c282eb4 1035 struct extent_map *em;
940100a4 1036 int ret = 1;
6c282eb4 1037 bool next_mergeable = true;
940100a4
CM		 1038
1039 /*
008873ea 1040 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1041 * defragging it
1042 */
1043 if (start < *defrag_end)
1044 return 1;
1045
1046 *skip = 0;
1047
6c282eb4
LZ		 1048 em = defrag_lookup_extent(inode, start);
1049 if (!em)
1050 return 0;
940100a4
CM		 1051
1052 /* this will cover holes, and inline extents */
17ce6ef8 1053 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1054 ret = 0;
17ce6ef8
LB		 1055 goto out;
1056 }
1057
6c282eb4 1058 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1059 /*
6c282eb4
LZ		 1060 * we hit a real extent, if it is big or the next extent is not a
1061 * real extent, don't bother defragging it
940100a4 1062 */
a43a2111 1063 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 1064 (em->len >= thresh || !next_mergeable))
940100a4 1065 ret = 0;
17ce6ef8 1066out:
940100a4
CM		 1067 /*
1068 * last_len ends up being a counter of how many bytes we've defragged.
1069 * every time we choose not to defrag an extent, we reset *last_len
1070 * so that the next tiny extent will force a defrag.
1071 *
1072 * The end result of this is that tiny extents before a single big
1073 * extent will force at least part of that big extent to be defragged.
1074 */
1075 if (ret) {
940100a4
CM		 1076 *defrag_end = extent_map_end(em);
1077 } else {
1078 *last_len = 0;
1079 *skip = extent_map_end(em);
1080 *defrag_end = 0;
1081 }
1082
1083 free_extent_map(em);
1084 return ret;
1085}
1086
4cb5300b
CM		 1087/*
1088 * it doesn't do much good to defrag one or two pages
1089 * at a time. This pulls in a nice chunk of pages
1090 * to COW and defrag.
1091 *
1092 * It also makes sure the delalloc code has enough
1093 * dirty data to avoid making new small extents as part
1094 * of the defrag
1095 *
1096 * It's a good idea to start RA on this range
1097 * before calling this.
1098 */
1099static int cluster_pages_for_defrag(struct inode *inode,
1100 struct page **pages,
1101 unsigned long start_index,
c41570c9 1102 unsigned long num_pages)
f46b5a66 1103{
4cb5300b
CM		 1104 unsigned long file_end;
1105 u64 isize = i_size_read(inode);
1106 u64 page_start;
1107 u64 page_end;
1f12bd06 1108 u64 page_cnt;
4cb5300b
CM		 1109 int ret;
1110 int i;
1111 int i_done;
3eaa2885 1112 struct btrfs_ordered_extent *ordered;
4cb5300b 1113 struct extent_state *cached_state = NULL;
600a45e1 1114 struct extent_io_tree *tree;
3b16a4e3 1115 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1116
4cb5300b 1117 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1118 if (!isize || start_index > file_end)
1119 return 0;
1120
1121 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1122
1123 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1124 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1125 if (ret)
1126 return ret;
4cb5300b 1127 i_done = 0;
600a45e1 1128 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1129
1130 /* step one, lock all the pages */
1f12bd06 1131 for (i = 0; i < page_cnt; i++) {
4cb5300b 1132 struct page *page;
600a45e1 1133again:
a94733d0 1134 page = find_or_create_page(inode->i_mapping,
600a45e1 1135 start_index + i, mask);
4cb5300b
CM		 1136 if (!page)
1137 break;
1138
600a45e1
MX		 1139 page_start = page_offset(page);
1140 page_end = page_start + PAGE_CACHE_SIZE - 1;
1141 while (1) {
308d9800
FM		 1142 lock_extent_bits(tree, page_start, page_end,
1143 0, &cached_state);
600a45e1
MX		 1144 ordered = btrfs_lookup_ordered_extent(inode,
1145 page_start);
308d9800
FM		 1146 unlock_extent_cached(tree, page_start, page_end,
1147 &cached_state, GFP_NOFS);
600a45e1
MX		 1148 if (!ordered)
1149 break;
1150
1151 unlock_page(page);
1152 btrfs_start_ordered_extent(inode, ordered, 1);
1153 btrfs_put_ordered_extent(ordered);
1154 lock_page(page);
1f12bd06
LB		 1155 /*
1156 * we unlocked the page above, so we need check if
1157 * it was released or not.
1158 */
1159 if (page->mapping != inode->i_mapping) {
1160 unlock_page(page);
1161 page_cache_release(page);
1162 goto again;
1163 }
600a45e1
MX		 1164 }
1165
4cb5300b
CM		 1166 if (!PageUptodate(page)) {
1167 btrfs_readpage(NULL, page);
1168 lock_page(page);
1169 if (!PageUptodate(page)) {
1170 unlock_page(page);
1171 page_cache_release(page);
1172 ret = -EIO;
1173 break;
1174 }
1175 }
600a45e1 1176
600a45e1
MX		 1177 if (page->mapping != inode->i_mapping) {
1178 unlock_page(page);
1179 page_cache_release(page);
1180 goto again;
1181 }
1182
4cb5300b
CM		 1183 pages[i] = page;
1184 i_done++;
1185 }
1186 if (!i_done || ret)
1187 goto out;
1188
1189 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1190 goto out;
1191
1192 /*
1193 * so now we have a nice long stream of locked
1194 * and up to date pages, lets wait on them
1195 */
1196 for (i = 0; i < i_done; i++)
1197 wait_on_page_writeback(pages[i]);
1198
1199 page_start = page_offset(pages[0]);
1200 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1201
1202 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1203 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1204 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1205 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1206 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1207 &cached_state, GFP_NOFS);
4cb5300b 1208
1f12bd06 1209 if (i_done != page_cnt) {
9e0baf60
JB		 1210 spin_lock(&BTRFS_I(inode)->lock);
1211 BTRFS_I(inode)->outstanding_extents++;
1212 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1213 btrfs_delalloc_release_space(inode,
1f12bd06 1214 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1215 }
1216
1217
9e8a4a8b
LB		 1218 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1219 &cached_state, GFP_NOFS);
4cb5300b
CM		 1220
1221 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1222 page_start, page_end - 1, &cached_state,
1223 GFP_NOFS);
1224
1225 for (i = 0; i < i_done; i++) {
1226 clear_page_dirty_for_io(pages[i]);
1227 ClearPageChecked(pages[i]);
1228 set_page_extent_mapped(pages[i]);
1229 set_page_dirty(pages[i]);
1230 unlock_page(pages[i]);
1231 page_cache_release(pages[i]);
1232 }
1233 return i_done;
1234out:
1235 for (i = 0; i < i_done; i++) {
1236 unlock_page(pages[i]);
1237 page_cache_release(pages[i]);
1238 }
1f12bd06 1239 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1240 return ret;
1241
1242}
1243
1244int btrfs_defrag_file(struct inode *inode, struct file *file,
1245 struct btrfs_ioctl_defrag_range_args *range,
1246 u64 newer_than, unsigned long max_to_defrag)
1247{
1248 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1249 struct file_ra_state *ra = NULL;
f46b5a66 1250 unsigned long last_index;
151a31b2 1251 u64 isize = i_size_read(inode);
940100a4
CM		 1252 u64 last_len = 0;
1253 u64 skip = 0;
1254 u64 defrag_end = 0;
4cb5300b 1255 u64 newer_off = range->start;
f46b5a66 1256 unsigned long i;
008873ea 1257 unsigned long ra_index = 0;
f46b5a66 1258 int ret;
4cb5300b 1259 int defrag_count = 0;
1a419d85 1260 int compress_type = BTRFS_COMPRESS_ZLIB;
aab110ab 1261 u32 extent_thresh = range->extent_thresh;
c41570c9
JM		 1262 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1263 unsigned long cluster = max_cluster;
4cb5300b
CM		 1264 u64 new_align = ~((u64)128 * 1024 - 1);
1265 struct page **pages = NULL;
1266
0abd5b17
LB		 1267 if (isize == 0)
1268 return 0;
1269
1270 if (range->start >= isize)
1271 return -EINVAL;
1a419d85
LZ		 1272
1273 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1274 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1275 return -EINVAL;
1276 if (range->compress_type)
1277 compress_type = range->compress_type;
1278 }
f46b5a66 1279
0abd5b17
LB		 1280 if (extent_thresh == 0)
1281 extent_thresh = 256 * 1024;
940100a4 1282
4cb5300b
CM		 1283 /*
1284 * if we were not given a file, allocate a readahead
1285 * context
1286 */
1287 if (!file) {
1288 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1289 if (!ra)
1290 return -ENOMEM;
1291 file_ra_state_init(ra, inode->i_mapping);
1292 } else {
1293 ra = &file->f_ra;
1294 }
1295
d9b0d9ba 1296 pages = kmalloc_array(max_cluster, sizeof(struct page *),
4cb5300b
CM		 1297 GFP_NOFS);
1298 if (!pages) {
1299 ret = -ENOMEM;
1300 goto out_ra;
1301 }
1302
1303 /* find the last page to defrag */
1e701a32 1304 if (range->start + range->len > range->start) {
151a31b2 1305 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1306 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1307 } else {
151a31b2 1308 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1309 }
1310
4cb5300b
CM		 1311 if (newer_than) {
1312 ret = find_new_extents(root, inode, newer_than,
1313 &newer_off, 64 * 1024);
1314 if (!ret) {
1315 range->start = newer_off;
1316 /*
1317 * we always align our defrag to help keep
1318 * the extents in the file evenly spaced
1319 */
1320 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1321 } else
1322 goto out_ra;
1323 } else {
1324 i = range->start >> PAGE_CACHE_SHIFT;
1325 }
1326 if (!max_to_defrag)
7ec31b54 1327 max_to_defrag = last_index + 1;
4cb5300b 1328
2a0f7f57
LZ		 1329 /*
1330 * make writeback starts from i, so the defrag range can be
1331 * written sequentially.
1332 */
1333 if (i < inode->i_mapping->writeback_index)
1334 inode->i_mapping->writeback_index = i;
1335
f7f43cc8 1336 while (i <= last_index && defrag_count < max_to_defrag &&
ed6078f7 1337 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
4cb5300b
CM		 1338 /*
1339 * make sure we stop running if someone unmounts
1340 * the FS
1341 */
1342 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1343 break;
1344
210549eb 1345 if (btrfs_defrag_cancelled(root->fs_info)) {
efe120a0 1346 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
210549eb
DS		 1347 ret = -EAGAIN;
1348 break;
1349 }
1350
66c26892 1351 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1352 extent_thresh, &last_len, &skip,
a43a2111
AM		 1353 &defrag_end, range->flags &
1354 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1355 unsigned long next;
1356 /*
1357 * the should_defrag function tells us how much to skip
1358 * bump our counter by the suggested amount
1359 */
ed6078f7 1360 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
940100a4
CM		 1361 i = max(i + 1, next);
1362 continue;
1363 }
008873ea
LZ		 1364
1365 if (!newer_than) {
1366 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1367 PAGE_CACHE_SHIFT) - i;
1368 cluster = min(cluster, max_cluster);
1369 } else {
1370 cluster = max_cluster;
1371 }
1372
008873ea
LZ		 1373 if (i + cluster > ra_index) {
1374 ra_index = max(i, ra_index);
1375 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1376 cluster);
1377 ra_index += max_cluster;
1378 }
940100a4 1379
ecb8bea8 1380 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1381 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1382 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1383 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1384 if (ret < 0) {
1385 mutex_unlock(&inode->i_mutex);
4cb5300b 1386 goto out_ra;
ecb8bea8 1387 }
4cb5300b
CM		 1388
1389 defrag_count += ret;
d0e1d66b 1390 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1391 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1392
1393 if (newer_than) {
1394 if (newer_off == (u64)-1)
1395 break;
1396
e1f041e1
LB		 1397 if (ret > 0)
1398 i += ret;
1399
4cb5300b
CM		 1400 newer_off = max(newer_off + 1,
1401 (u64)i << PAGE_CACHE_SHIFT);
1402
1403 ret = find_new_extents(root, inode,
1404 newer_than, &newer_off,
1405 64 * 1024);
1406 if (!ret) {
1407 range->start = newer_off;
1408 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1409 } else {
1410 break;
f46b5a66 1411 }
4cb5300b 1412 } else {
008873ea 1413 if (ret > 0) {
cbcc8326 1414 i += ret;
008873ea
LZ		 1415 last_len += ret << PAGE_CACHE_SHIFT;
1416 } else {
cbcc8326 1417 i++;
008873ea
LZ		 1418 last_len = 0;
1419 }
f46b5a66 1420 }
f46b5a66
CH		 1421 }
1422
dec8ef90 1423 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1424 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1425 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1426 &BTRFS_I(inode)->runtime_flags))
1427 filemap_flush(inode->i_mapping);
1428 }
1e701a32
CM		 1429
1430 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1431 /* the filemap_flush will queue IO into the worker threads, but
1432 * we have to make sure the IO is actually started and that
1433 * ordered extents get created before we return
1434 */
1435 atomic_inc(&root->fs_info->async_submit_draining);
1436 while (atomic_read(&root->fs_info->nr_async_submits) ||
1437 atomic_read(&root->fs_info->async_delalloc_pages)) {
1438 wait_event(root->fs_info->async_submit_wait,
1439 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1440 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1441 }
1442 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1443 }
1444
1a419d85 1445 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1446 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1447 }
1448
60ccf82f 1449 ret = defrag_count;
940100a4 1450
4cb5300b 1451out_ra:
633085c7
FDBM		 1452 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1453 mutex_lock(&inode->i_mutex);
1454 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1455 mutex_unlock(&inode->i_mutex);
1456 }
4cb5300b
CM		 1457 if (!file)
1458 kfree(ra);
1459 kfree(pages);
940100a4 1460 return ret;
f46b5a66
CH		 1461}
1462
198605a8 1463static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1464 void __user *arg)
f46b5a66
CH		 1465{
1466 u64 new_size;
1467 u64 old_size;
1468 u64 devid = 1;
496ad9aa 1469 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1470 struct btrfs_ioctl_vol_args *vol_args;
1471 struct btrfs_trans_handle *trans;
1472 struct btrfs_device *device = NULL;
1473 char *sizestr;
9a40f122 1474 char *retptr;
f46b5a66
CH		 1475 char *devstr = NULL;
1476 int ret = 0;
f46b5a66
CH		 1477 int mod = 0;
1478
e441d54d
CM		 1479 if (!capable(CAP_SYS_ADMIN))
1480 return -EPERM;
1481
198605a8
MX		 1482 ret = mnt_want_write_file(file);
1483 if (ret)
1484 return ret;
1485
5ac00add
SB		 1486 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1487 1)) {
97547676 1488 mnt_drop_write_file(file);
e57138b3 1489 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1490 }
1491
5ac00add 1492 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1493 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1494 if (IS_ERR(vol_args)) {
1495 ret = PTR_ERR(vol_args);
1496 goto out;
1497 }
5516e595
MF		 1498
1499 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1500
f46b5a66
CH		 1501 sizestr = vol_args->name;
1502 devstr = strchr(sizestr, ':');
1503 if (devstr) {
f46b5a66
CH		 1504 sizestr = devstr + 1;
1505 *devstr = '\0';
1506 devstr = vol_args->name;
58dfae63
Z		 1507 ret = kstrtoull(devstr, 10, &devid);
1508 if (ret)
1509 goto out_free;
dfd79829
MX		 1510 if (!devid) {
1511 ret = -EINVAL;
1512 goto out_free;
1513 }
efe120a0 1514 btrfs_info(root->fs_info, "resizing devid %llu", devid);
f46b5a66 1515 }
dba60f3f 1516
aa1b8cd4 1517 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1518 if (!device) {
efe120a0 1519 btrfs_info(root->fs_info, "resizer unable to find device %llu",
c1c9ff7c 1520 devid);
dfd79829 1521 ret = -ENODEV;
c9e9f97b 1522 goto out_free;
f46b5a66 1523 }
dba60f3f
MX		 1524
1525 if (!device->writeable) {
efe120a0
FH		 1526 btrfs_info(root->fs_info,
1527 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1528 devid);
dfd79829 1529 ret = -EPERM;
4e42ae1b
LB		 1530 goto out_free;
1531 }
1532
f46b5a66
CH		 1533 if (!strcmp(sizestr, "max"))
1534 new_size = device->bdev->bd_inode->i_size;
1535 else {
1536 if (sizestr[0] == '-') {
1537 mod = -1;
1538 sizestr++;
1539 } else if (sizestr[0] == '+') {
1540 mod = 1;
1541 sizestr++;
1542 }
9a40f122
GH		 1543 new_size = memparse(sizestr, &retptr);
1544 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1545 ret = -EINVAL;
c9e9f97b 1546 goto out_free;
f46b5a66
CH		 1547 }
1548 }
1549
63a212ab 1550 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1551 ret = -EPERM;
63a212ab
SB		 1552 goto out_free;
1553 }
1554
7cc8e58d 1555 old_size = btrfs_device_get_total_bytes(device);
f46b5a66
CH		 1556
1557 if (mod < 0) {
1558 if (new_size > old_size) {
1559 ret = -EINVAL;
c9e9f97b 1560 goto out_free;
f46b5a66
CH		 1561 }
1562 new_size = old_size - new_size;
1563 } else if (mod > 0) {
eb8052e0 1564 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1565 ret = -ERANGE;
eb8052e0
WF		 1566 goto out_free;
1567 }
f46b5a66
CH		 1568 new_size = old_size + new_size;
1569 }
1570
1571 if (new_size < 256 * 1024 * 1024) {
1572 ret = -EINVAL;
c9e9f97b 1573 goto out_free;
f46b5a66
CH		 1574 }
1575 if (new_size > device->bdev->bd_inode->i_size) {
1576 ret = -EFBIG;
c9e9f97b 1577 goto out_free;
f46b5a66
CH		 1578 }
1579
1580 do_div(new_size, root->sectorsize);
1581 new_size *= root->sectorsize;
1582
efe120a0 1583 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
c1c9ff7c 1584 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1585
1586 if (new_size > old_size) {
a22285a6 1587 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1588 if (IS_ERR(trans)) {
1589 ret = PTR_ERR(trans);
c9e9f97b 1590 goto out_free;
98d5dc13 1591 }
f46b5a66
CH		 1592 ret = btrfs_grow_device(trans, device, new_size);
1593 btrfs_commit_transaction(trans, root);
ece7d20e 1594 } else if (new_size < old_size) {
f46b5a66 1595 ret = btrfs_shrink_device(device, new_size);
0253f40e 1596 } /* equal, nothing need to do */
f46b5a66 1597
c9e9f97b 1598out_free:
f46b5a66 1599 kfree(vol_args);
c9e9f97b
ID		 1600out:
1601 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1602 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1603 mnt_drop_write_file(file);
f46b5a66
CH		 1604 return ret;
1605}
1606
72fd032e 1607static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1608 char *name, unsigned long fd, int subvol,
1609 u64 *transid, bool readonly,
8696c533 1610 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1611{
f46b5a66 1612 int namelen;
3de4586c 1613 int ret = 0;
f46b5a66 1614
a874a63e
LB		 1615 ret = mnt_want_write_file(file);
1616 if (ret)
1617 goto out;
1618
72fd032e
SW		 1619 namelen = strlen(name);
1620 if (strchr(name, '/')) {
f46b5a66 1621 ret = -EINVAL;
a874a63e 1622 goto out_drop_write;
f46b5a66
CH		 1623 }
1624
16780cab
CM		 1625 if (name[0] == '.' &&
1626 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1627 ret = -EEXIST;
a874a63e 1628 goto out_drop_write;
16780cab
CM		 1629 }
1630
3de4586c 1631 if (subvol) {
72fd032e 1632 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1633 NULL, transid, readonly, inherit);
cb8e7090 1634 } else {
2903ff01 1635 struct fd src = fdget(fd);
3de4586c 1636 struct inode *src_inode;
2903ff01 1637 if (!src.file) {
3de4586c 1638 ret = -EINVAL;
a874a63e 1639 goto out_drop_write;
3de4586c
CM		 1640 }
1641
496ad9aa
AV		 1642 src_inode = file_inode(src.file);
1643 if (src_inode->i_sb != file_inode(file)->i_sb) {
efe120a0
FH		 1644 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1645 "Snapshot src from another FS");
23ad5b17 1646 ret = -EXDEV;
d0242061
DS		 1647 } else if (!inode_owner_or_capable(src_inode)) {
1648 /*
1649 * Subvolume creation is not restricted, but snapshots
1650 * are limited to own subvolumes only
1651 */
1652 ret = -EPERM;
ecd18815
AV		 1653 } else {
1654 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1655 BTRFS_I(src_inode)->root,
1656 transid, readonly, inherit);
3de4586c 1657 }
2903ff01 1658 fdput(src);
cb8e7090 1659 }
a874a63e
LB		 1660out_drop_write:
1661 mnt_drop_write_file(file);
f46b5a66 1662out:
72fd032e
SW		 1663 return ret;
1664}
1665
1666static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1667 void __user *arg, int subvol)
72fd032e 1668{
fa0d2b9b 1669 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1670 int ret;
1671
fa0d2b9b
LZ		 1672 vol_args = memdup_user(arg, sizeof(*vol_args));
1673 if (IS_ERR(vol_args))
1674 return PTR_ERR(vol_args);
1675 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1676
fa0d2b9b 1677 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1678 vol_args->fd, subvol,
6f72c7e2 1679 NULL, false, NULL);
fdfb1e4f 1680
fa0d2b9b
LZ		 1681 kfree(vol_args);
1682 return ret;
1683}
fdfb1e4f 1684
fa0d2b9b
LZ		 1685static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1686 void __user *arg, int subvol)
1687{
1688 struct btrfs_ioctl_vol_args_v2 *vol_args;
1689 int ret;
1690 u64 transid = 0;
1691 u64 *ptr = NULL;
b83cc969 1692 bool readonly = false;
6f72c7e2 1693 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1694
fa0d2b9b
LZ		 1695 vol_args = memdup_user(arg, sizeof(*vol_args));
1696 if (IS_ERR(vol_args))
1697 return PTR_ERR(vol_args);
1698 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1699
b83cc969 1700 if (vol_args->flags &
6f72c7e2
AJ		 1701 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1702 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1703 ret = -EOPNOTSUPP;
c47ca32d 1704 goto free_args;
72fd032e 1705 }
fa0d2b9b
LZ		 1706
1707 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1708 ptr = &transid;
b83cc969
LZ		 1709 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1710 readonly = true;
6f72c7e2
AJ		 1711 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1712 if (vol_args->size > PAGE_CACHE_SIZE) {
1713 ret = -EINVAL;
c47ca32d 1714 goto free_args;
6f72c7e2
AJ		 1715 }
1716 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1717 if (IS_ERR(inherit)) {
1718 ret = PTR_ERR(inherit);
c47ca32d 1719 goto free_args;
6f72c7e2
AJ		 1720 }
1721 }
fa0d2b9b
LZ		 1722
1723 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1724 vol_args->fd, subvol, ptr,
8696c533 1725 readonly, inherit);
c47ca32d
DC		 1726 if (ret)
1727 goto free_inherit;
fa0d2b9b 1728
c47ca32d
DC		 1729 if (ptr && copy_to_user(arg +
1730 offsetof(struct btrfs_ioctl_vol_args_v2,
1731 transid),
1732 ptr, sizeof(*ptr)))
fa0d2b9b 1733 ret = -EFAULT;
c47ca32d
DC		 1734
1735free_inherit:
6f72c7e2 1736 kfree(inherit);
c47ca32d
DC		 1737free_args:
1738 kfree(vol_args);
f46b5a66
CH		 1739 return ret;
1740}
1741
0caa102d
LZ		 1742static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1743 void __user *arg)
1744{
496ad9aa 1745 struct inode *inode = file_inode(file);
0caa102d
LZ		 1746 struct btrfs_root *root = BTRFS_I(inode)->root;
1747 int ret = 0;
1748 u64 flags = 0;
1749
33345d01 1750 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1751 return -EINVAL;
1752
1753 down_read(&root->fs_info->subvol_sem);
1754 if (btrfs_root_readonly(root))
1755 flags |= BTRFS_SUBVOL_RDONLY;
1756 up_read(&root->fs_info->subvol_sem);
1757
1758 if (copy_to_user(arg, &flags, sizeof(flags)))
1759 ret = -EFAULT;
1760
1761 return ret;
1762}
1763
1764static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1765 void __user *arg)
1766{
496ad9aa 1767 struct inode *inode = file_inode(file);
0caa102d
LZ		 1768 struct btrfs_root *root = BTRFS_I(inode)->root;
1769 struct btrfs_trans_handle *trans;
1770 u64 root_flags;
1771 u64 flags;
1772 int ret = 0;
1773
bd60ea0f
DS		 1774 if (!inode_owner_or_capable(inode))
1775 return -EPERM;
1776
b9ca0664
LB		 1777 ret = mnt_want_write_file(file);
1778 if (ret)
1779 goto out;
0caa102d 1780
b9ca0664
LB		 1781 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1782 ret = -EINVAL;
1783 goto out_drop_write;
1784 }
0caa102d 1785
b9ca0664
LB		 1786 if (copy_from_user(&flags, arg, sizeof(flags))) {
1787 ret = -EFAULT;
1788 goto out_drop_write;
1789 }
0caa102d 1790
b9ca0664
LB		 1791 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1792 ret = -EINVAL;
1793 goto out_drop_write;
1794 }
0caa102d 1795
b9ca0664
LB		 1796 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1797 ret = -EOPNOTSUPP;
1798 goto out_drop_write;
1799 }
0caa102d
LZ		 1800
1801 down_write(&root->fs_info->subvol_sem);
1802
1803 /* nothing to do */
1804 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1805 goto out_drop_sem;
0caa102d
LZ		 1806
1807 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1808 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1809 btrfs_set_root_flags(&root->root_item,
1810 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1811 } else {
1812 /*
1813 * Block RO -> RW transition if this subvolume is involved in
1814 * send
1815 */
1816 spin_lock(&root->root_item_lock);
1817 if (root->send_in_progress == 0) {
1818 btrfs_set_root_flags(&root->root_item,
0caa102d 1819 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1820 spin_unlock(&root->root_item_lock);
1821 } else {
1822 spin_unlock(&root->root_item_lock);
1823 btrfs_warn(root->fs_info,
1824 "Attempt to set subvolume %llu read-write during send",
1825 root->root_key.objectid);
1826 ret = -EPERM;
1827 goto out_drop_sem;
1828 }
1829 }
0caa102d
LZ		 1830
1831 trans = btrfs_start_transaction(root, 1);
1832 if (IS_ERR(trans)) {
1833 ret = PTR_ERR(trans);
1834 goto out_reset;
1835 }
1836
b4dc2b8c 1837 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1838 &root->root_key, &root->root_item);
1839
1840 btrfs_commit_transaction(trans, root);
1841out_reset:
1842 if (ret)
1843 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1844out_drop_sem:
0caa102d 1845 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1846out_drop_write:
1847 mnt_drop_write_file(file);
1848out:
0caa102d
LZ		 1849 return ret;
1850}
1851
76dda93c
YZ		 1852/*
1853 * helper to check if the subvolume references other subvolumes
1854 */
1855static noinline int may_destroy_subvol(struct btrfs_root *root)
1856{
1857 struct btrfs_path *path;
175a2b87 1858 struct btrfs_dir_item *di;
76dda93c 1859 struct btrfs_key key;
175a2b87 1860 u64 dir_id;
76dda93c
YZ		 1861 int ret;
1862
1863 path = btrfs_alloc_path();
1864 if (!path)
1865 return -ENOMEM;
1866
175a2b87
JB		 1867 /* Make sure this root isn't set as the default subvol */
1868 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1869 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1870 dir_id, "default", 7, 0);
1871 if (di && !IS_ERR(di)) {
1872 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1873 if (key.objectid == root->root_key.objectid) {
72de6b53
GS		 1874 ret = -EPERM;
1875 btrfs_err(root->fs_info, "deleting default subvolume "
1876 "%llu is not allowed", key.objectid);
175a2b87
JB		 1877 goto out;
1878 }
1879 btrfs_release_path(path);
1880 }
1881
76dda93c
YZ		 1882 key.objectid = root->root_key.objectid;
1883 key.type = BTRFS_ROOT_REF_KEY;
1884 key.offset = (u64)-1;
1885
1886 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1887 &key, path, 0, 0);
1888 if (ret < 0)
1889 goto out;
1890 BUG_ON(ret == 0);
1891
1892 ret = 0;
1893 if (path->slots[0] > 0) {
1894 path->slots[0]--;
1895 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1896 if (key.objectid == root->root_key.objectid &&
1897 key.type == BTRFS_ROOT_REF_KEY)
1898 ret = -ENOTEMPTY;
1899 }
1900out:
1901 btrfs_free_path(path);
1902 return ret;
1903}
1904
ac8e9819
CM		 1905static noinline int key_in_sk(struct btrfs_key *key,
1906 struct btrfs_ioctl_search_key *sk)
1907{
abc6e134
CM		 1908 struct btrfs_key test;
1909 int ret;
1910
1911 test.objectid = sk->min_objectid;
1912 test.type = sk->min_type;
1913 test.offset = sk->min_offset;
1914
1915 ret = btrfs_comp_cpu_keys(key, &test);
1916 if (ret < 0)
ac8e9819 1917 return 0;
abc6e134
CM		 1918
1919 test.objectid = sk->max_objectid;
1920 test.type = sk->max_type;
1921 test.offset = sk->max_offset;
1922
1923 ret = btrfs_comp_cpu_keys(key, &test);
1924 if (ret > 0)
ac8e9819
CM		 1925 return 0;
1926 return 1;
1927}
1928
1929static noinline int copy_to_sk(struct btrfs_root *root,
1930 struct btrfs_path *path,
1931 struct btrfs_key *key,
1932 struct btrfs_ioctl_search_key *sk,
9b6e817d 1933 size_t *buf_size,
ba346b35 1934 char __user *ubuf,
ac8e9819
CM		 1935 unsigned long *sk_offset,
1936 int *num_found)
1937{
1938 u64 found_transid;
1939 struct extent_buffer *leaf;
1940 struct btrfs_ioctl_search_header sh;
1941 unsigned long item_off;
1942 unsigned long item_len;
1943 int nritems;
1944 int i;
1945 int slot;
ac8e9819
CM		 1946 int ret = 0;
1947
1948 leaf = path->nodes[0];
1949 slot = path->slots[0];
1950 nritems = btrfs_header_nritems(leaf);
1951
1952 if (btrfs_header_generation(leaf) > sk->max_transid) {
1953 i = nritems;
1954 goto advance_key;
1955 }
1956 found_transid = btrfs_header_generation(leaf);
1957
1958 for (i = slot; i < nritems; i++) {
1959 item_off = btrfs_item_ptr_offset(leaf, i);
1960 item_len = btrfs_item_size_nr(leaf, i);
1961
03b71c6c
GP		 1962 btrfs_item_key_to_cpu(leaf, key, i);
1963 if (!key_in_sk(key, sk))
1964 continue;
1965
9b6e817d 1966 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 1967 if (*num_found) {
1968 ret = 1;
1969 goto out;
1970 }
1971
1972 /*
1973 * return one empty item back for v1, which does not
1974 * handle -EOVERFLOW
1975 */
1976
9b6e817d 1977 *buf_size = sizeof(sh) + item_len;
ac8e9819 1978 item_len = 0;
8f5f6178
GH		 1979 ret = -EOVERFLOW;
1980 }
ac8e9819 1981
9b6e817d 1982 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 1983 ret = 1;
25c9bc2e 1984 goto out;
ac8e9819
CM		 1985 }
1986
ac8e9819
CM		 1987 sh.objectid = key->objectid;
1988 sh.offset = key->offset;
1989 sh.type = key->type;
1990 sh.len = item_len;
1991 sh.transid = found_transid;
1992
1993 /* copy search result header */
ba346b35
GH		 1994 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1995 ret = -EFAULT;
1996 goto out;
1997 }
1998
ac8e9819
CM		 1999 *sk_offset += sizeof(sh);
2000
2001 if (item_len) {
ba346b35 2002 char __user *up = ubuf + *sk_offset;
ac8e9819 2003 /* copy the item */
ba346b35
GH		 2004 if (read_extent_buffer_to_user(leaf, up,
2005 item_off, item_len)) {
2006 ret = -EFAULT;
2007 goto out;
2008 }
2009
ac8e9819 2010 *sk_offset += item_len;
ac8e9819 2011 }
e2156867 2012 (*num_found)++;
ac8e9819 2013
8f5f6178
GH		 2014 if (ret) /* -EOVERFLOW from above */
2015 goto out;
2016
25c9bc2e
GH		 2017 if (*num_found >= sk->nr_items) {
2018 ret = 1;
2019 goto out;
2020 }
ac8e9819
CM		 2021 }
2022advance_key:
abc6e134
CM		 2023 ret = 0;
2024 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 2025 key->offset++;
abc6e134
CM		 2026 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2027 key->offset = 0;
ac8e9819 2028 key->type++;
abc6e134
CM		 2029 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2030 key->offset = 0;
2031 key->type = 0;
ac8e9819 2032 key->objectid++;
abc6e134
CM		 2033 } else
2034 ret = 1;
25c9bc2e 2035out:
ba346b35
GH		 2036 /*
2037 * 0: all items from this leaf copied, continue with next
2038 * 1: * more items can be copied, but unused buffer is too small
2039 * * all items were found
2040 * Either way, it will stops the loop which iterates to the next
2041 * leaf
2042 * -EOVERFLOW: item was to large for buffer
2043 * -EFAULT: could not copy extent buffer back to userspace
2044 */
ac8e9819
CM		 2045 return ret;
2046}
2047
2048static noinline int search_ioctl(struct inode *inode,
12544442 2049 struct btrfs_ioctl_search_key *sk,
9b6e817d 2050 size_t *buf_size,
ba346b35 2051 char __user *ubuf)
ac8e9819
CM		 2052{
2053 struct btrfs_root *root;
2054 struct btrfs_key key;
ac8e9819 2055 struct btrfs_path *path;
ac8e9819
CM		 2056 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2057 int ret;
2058 int num_found = 0;
2059 unsigned long sk_offset = 0;
2060
9b6e817d
GH		 2061 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2062 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2063 return -EOVERFLOW;
9b6e817d 2064 }
12544442 2065
ac8e9819
CM		 2066 path = btrfs_alloc_path();
2067 if (!path)
2068 return -ENOMEM;
2069
2070 if (sk->tree_id == 0) {
2071 /* search the root of the inode that was passed */
2072 root = BTRFS_I(inode)->root;
2073 } else {
2074 key.objectid = sk->tree_id;
2075 key.type = BTRFS_ROOT_ITEM_KEY;
2076 key.offset = (u64)-1;
2077 root = btrfs_read_fs_root_no_name(info, &key);
2078 if (IS_ERR(root)) {
efe120a0 2079 printk(KERN_ERR "BTRFS: could not find root %llu\n",
ac8e9819
CM		 2080 sk->tree_id);
2081 btrfs_free_path(path);
2082 return -ENOENT;
2083 }
2084 }
2085
2086 key.objectid = sk->min_objectid;
2087 key.type = sk->min_type;
2088 key.offset = sk->min_offset;
2089
67871254 2090 while (1) {
6174d3cb 2091 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2092 if (ret != 0) {
2093 if (ret > 0)
2094 ret = 0;
2095 goto err;
2096 }
ba346b35 2097 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
ac8e9819 2098 &sk_offset, &num_found);
b3b4aa74 2099 btrfs_release_path(path);
25c9bc2e 2100 if (ret)
ac8e9819
CM		 2101 break;
2102
2103 }
8f5f6178
GH		 2104 if (ret > 0)
2105 ret = 0;
ac8e9819
CM		 2106err:
2107 sk->nr_items = num_found;
2108 btrfs_free_path(path);
2109 return ret;
2110}
2111
2112static noinline int btrfs_ioctl_tree_search(struct file *file,
2113 void __user *argp)
2114{
ba346b35
GH		 2115 struct btrfs_ioctl_search_args __user *uargs;
2116 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2117 struct inode *inode;
2118 int ret;
2119 size_t buf_size;
ac8e9819
CM		 2120
2121 if (!capable(CAP_SYS_ADMIN))
2122 return -EPERM;
2123
ba346b35
GH		 2124 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2125
2126 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2127 return -EFAULT;
ac8e9819 2128
ba346b35 2129 buf_size = sizeof(uargs->buf);
ac8e9819 2130
496ad9aa 2131 inode = file_inode(file);
ba346b35 2132 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2133
2134 /*
2135 * In the origin implementation an overflow is handled by returning a
2136 * search header with a len of zero, so reset ret.
2137 */
2138 if (ret == -EOVERFLOW)
2139 ret = 0;
2140
ba346b35 2141 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2142 ret = -EFAULT;
ac8e9819
CM		 2143 return ret;
2144}
2145
cc68a8a5
GH		 2146static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2147 void __user *argp)
2148{
2149 struct btrfs_ioctl_search_args_v2 __user *uarg;
2150 struct btrfs_ioctl_search_args_v2 args;
2151 struct inode *inode;
2152 int ret;
2153 size_t buf_size;
2154 const size_t buf_limit = 16 * 1024 * 1024;
2155
2156 if (!capable(CAP_SYS_ADMIN))
2157 return -EPERM;
2158
2159 /* copy search header and buffer size */
2160 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2161 if (copy_from_user(&args, uarg, sizeof(args)))
2162 return -EFAULT;
2163
2164 buf_size = args.buf_size;
2165
2166 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2167 return -EOVERFLOW;
2168
2169 /* limit result size to 16MB */
2170 if (buf_size > buf_limit)
2171 buf_size = buf_limit;
2172
2173 inode = file_inode(file);
2174 ret = search_ioctl(inode, &args.key, &buf_size,
2175 (char *)(&uarg->buf[0]));
2176 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2177 ret = -EFAULT;
2178 else if (ret == -EOVERFLOW &&
2179 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2180 ret = -EFAULT;
2181
ac8e9819
CM		 2182 return ret;
2183}
2184
98d377a0 2185/*
ac8e9819
CM		 2186 * Search INODE_REFs to identify path name of 'dirid' directory
2187 * in a 'tree_id' tree. and sets path name to 'name'.
2188 */
98d377a0
TH		 2189static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2190 u64 tree_id, u64 dirid, char *name)
2191{
2192 struct btrfs_root *root;
2193 struct btrfs_key key;
ac8e9819 2194 char *ptr;
98d377a0
TH		 2195 int ret = -1;
2196 int slot;
2197 int len;
2198 int total_len = 0;
2199 struct btrfs_inode_ref *iref;
2200 struct extent_buffer *l;
2201 struct btrfs_path *path;
2202
2203 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2204 name[0]='\0';
2205 return 0;
2206 }
2207
2208 path = btrfs_alloc_path();
2209 if (!path)
2210 return -ENOMEM;
2211
ac8e9819 2212 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2213
2214 key.objectid = tree_id;
2215 key.type = BTRFS_ROOT_ITEM_KEY;
2216 key.offset = (u64)-1;
2217 root = btrfs_read_fs_root_no_name(info, &key);
2218 if (IS_ERR(root)) {
efe120a0 2219 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
8ad6fcab
CM		 2220 ret = -ENOENT;
2221 goto out;
98d377a0
TH		 2222 }
2223
2224 key.objectid = dirid;
2225 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2226 key.offset = (u64)-1;
98d377a0 2227
67871254 2228 while (1) {
98d377a0
TH		 2229 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2230 if (ret < 0)
2231 goto out;
18674c6c
FDBM		 2232 else if (ret > 0) {
2233 ret = btrfs_previous_item(root, path, dirid,
2234 BTRFS_INODE_REF_KEY);
2235 if (ret < 0)
2236 goto out;
2237 else if (ret > 0) {
2238 ret = -ENOENT;
2239 goto out;
2240 }
2241 }
98d377a0
TH		 2242
2243 l = path->nodes[0];
2244 slot = path->slots[0];
2245 btrfs_item_key_to_cpu(l, &key, slot);
2246
98d377a0
TH		 2247 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2248 len = btrfs_inode_ref_name_len(l, iref);
2249 ptr -= len + 1;
2250 total_len += len + 1;
a696cf35
FDBM		 2251 if (ptr < name) {
2252 ret = -ENAMETOOLONG;
98d377a0 2253 goto out;
a696cf35 2254 }
98d377a0
TH		 2255
2256 *(ptr + len) = '/';
67871254 2257 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2258
2259 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2260 break;
2261
b3b4aa74 2262 btrfs_release_path(path);
98d377a0 2263 key.objectid = key.offset;
8ad6fcab 2264 key.offset = (u64)-1;
98d377a0 2265 dirid = key.objectid;
98d377a0 2266 }
77906a50 2267 memmove(name, ptr, total_len);
67871254 2268 name[total_len] = '\0';
98d377a0
TH		 2269 ret = 0;
2270out:
2271 btrfs_free_path(path);
ac8e9819
CM		 2272 return ret;
2273}
2274
2275static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2276 void __user *argp)
2277{
2278 struct btrfs_ioctl_ino_lookup_args *args;
2279 struct inode *inode;
2280 int ret;
2281
2282 if (!capable(CAP_SYS_ADMIN))
2283 return -EPERM;
2284
2354d08f
JL		 2285 args = memdup_user(argp, sizeof(*args));
2286 if (IS_ERR(args))
2287 return PTR_ERR(args);
c2b96929 2288
496ad9aa 2289 inode = file_inode(file);
ac8e9819 2290
1b53ac4d
CM		 2291 if (args->treeid == 0)
2292 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2293
ac8e9819
CM		 2294 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2295 args->treeid, args->objectid,
2296 args->name);
2297
2298 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2299 ret = -EFAULT;
2300
2301 kfree(args);
98d377a0
TH		 2302 return ret;
2303}
2304
76dda93c
YZ		 2305static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2306 void __user *arg)
2307{
54563d41 2308 struct dentry *parent = file->f_path.dentry;
76dda93c
YZ		 2309 struct dentry *dentry;
2310 struct inode *dir = parent->d_inode;
2311 struct inode *inode;
2312 struct btrfs_root *root = BTRFS_I(dir)->root;
2313 struct btrfs_root *dest = NULL;
2314 struct btrfs_ioctl_vol_args *vol_args;
2315 struct btrfs_trans_handle *trans;
c58aaad2 2316 struct btrfs_block_rsv block_rsv;
521e0546 2317 u64 root_flags;
c58aaad2 2318 u64 qgroup_reserved;
76dda93c
YZ		 2319 int namelen;
2320 int ret;
2321 int err = 0;
2322
76dda93c
YZ		 2323 vol_args = memdup_user(arg, sizeof(*vol_args));
2324 if (IS_ERR(vol_args))
2325 return PTR_ERR(vol_args);
2326
2327 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2328 namelen = strlen(vol_args->name);
2329 if (strchr(vol_args->name, '/') ||
2330 strncmp(vol_args->name, "..", namelen) == 0) {
2331 err = -EINVAL;
2332 goto out;
2333 }
2334
a561be71 2335 err = mnt_want_write_file(file);
76dda93c
YZ		 2336 if (err)
2337 goto out;
2338
521e0546 2339
5c50c9b8
DS		 2340 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2341 if (err == -EINTR)
e43f998e 2342 goto out_drop_write;
76dda93c
YZ		 2343 dentry = lookup_one_len(vol_args->name, parent, namelen);
2344 if (IS_ERR(dentry)) {
2345 err = PTR_ERR(dentry);
2346 goto out_unlock_dir;
2347 }
2348
2349 if (!dentry->d_inode) {
2350 err = -ENOENT;
2351 goto out_dput;
2352 }
2353
2354 inode = dentry->d_inode;
4260f7c7 2355 dest = BTRFS_I(inode)->root;
67871254 2356 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2357 /*
2358 * Regular user. Only allow this with a special mount
2359 * option, when the user has write+exec access to the
2360 * subvol root, and when rmdir(2) would have been
2361 * allowed.
2362 *
2363 * Note that this is _not_ check that the subvol is
2364 * empty or doesn't contain data that we wouldn't
2365 * otherwise be able to delete.
2366 *
2367 * Users who want to delete empty subvols should try
2368 * rmdir(2).
2369 */
2370 err = -EPERM;
2371 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2372 goto out_dput;
2373
2374 /*
2375 * Do not allow deletion if the parent dir is the same
2376 * as the dir to be deleted. That means the ioctl
2377 * must be called on the dentry referencing the root
2378 * of the subvol, not a random directory contained
2379 * within it.
2380 */
2381 err = -EINVAL;
2382 if (root == dest)
2383 goto out_dput;
2384
2385 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2386 if (err)
2387 goto out_dput;
4260f7c7
SW		 2388 }
2389
5c39da5b
MX		 2390 /* check if subvolume may be deleted by a user */
2391 err = btrfs_may_delete(dir, dentry, 1);
2392 if (err)
2393 goto out_dput;
2394
33345d01 2395 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2396 err = -EINVAL;
2397 goto out_dput;
2398 }
2399
76dda93c 2400 mutex_lock(&inode->i_mutex);
521e0546
DS		 2401
2402 /*
2403 * Don't allow to delete a subvolume with send in progress. This is
2404 * inside the i_mutex so the error handling that has to drop the bit
2405 * again is not run concurrently.
2406 */
2407 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2408 root_flags = btrfs_root_flags(&dest->root_item);
2409 if (dest->send_in_progress == 0) {
2410 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2411 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2412 spin_unlock(&dest->root_item_lock);
2413 } else {
2414 spin_unlock(&dest->root_item_lock);
2415 btrfs_warn(root->fs_info,
2416 "Attempt to delete subvolume %llu during send",
c55bfa67 2417 dest->root_key.objectid);
521e0546
DS		 2418 err = -EPERM;
2419 goto out_dput;
2420 }
2421
76dda93c
YZ		 2422 err = d_invalidate(dentry);
2423 if (err)
2424 goto out_unlock;
2425
2426 down_write(&root->fs_info->subvol_sem);
2427
2428 err = may_destroy_subvol(dest);
2429 if (err)
2430 goto out_up_write;
2431
c58aaad2
MX		 2432 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2433 /*
2434 * One for dir inode, two for dir entries, two for root
2435 * ref/backref.
2436 */
2437 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2438 5, &qgroup_reserved, true);
c58aaad2
MX		 2439 if (err)
2440 goto out_up_write;
2441
a22285a6
YZ		 2442 trans = btrfs_start_transaction(root, 0);
2443 if (IS_ERR(trans)) {
2444 err = PTR_ERR(trans);
c58aaad2 2445 goto out_release;
a22285a6 2446 }
c58aaad2
MX		 2447 trans->block_rsv = &block_rsv;
2448 trans->bytes_reserved = block_rsv.size;
a22285a6 2449
76dda93c
YZ		 2450 ret = btrfs_unlink_subvol(trans, root, dir,
2451 dest->root_key.objectid,
2452 dentry->d_name.name,
2453 dentry->d_name.len);
79787eaa
JM		 2454 if (ret) {
2455 err = ret;
2456 btrfs_abort_transaction(trans, root, ret);
2457 goto out_end_trans;
2458 }
76dda93c
YZ		 2459
2460 btrfs_record_root_in_trans(trans, dest);
2461
2462 memset(&dest->root_item.drop_progress, 0,
2463 sizeof(dest->root_item.drop_progress));
2464 dest->root_item.drop_level = 0;
2465 btrfs_set_root_refs(&dest->root_item, 0);
2466
27cdeb70 2467 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
d68fc57b
YZ		 2468 ret = btrfs_insert_orphan_item(trans,
2469 root->fs_info->tree_root,
2470 dest->root_key.objectid);
79787eaa
JM		 2471 if (ret) {
2472 btrfs_abort_transaction(trans, root, ret);
2473 err = ret;
2474 goto out_end_trans;
2475 }
d68fc57b 2476 }
dd5f9615
SB		 2477
2478 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2479 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2480 dest->root_key.objectid);
2481 if (ret && ret != -ENOENT) {
2482 btrfs_abort_transaction(trans, root, ret);
2483 err = ret;
2484 goto out_end_trans;
2485 }
2486 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2487 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2488 dest->root_item.received_uuid,
2489 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2490 dest->root_key.objectid);
2491 if (ret && ret != -ENOENT) {
2492 btrfs_abort_transaction(trans, root, ret);
2493 err = ret;
2494 goto out_end_trans;
2495 }
2496 }
2497
79787eaa 2498out_end_trans:
c58aaad2
MX		 2499 trans->block_rsv = NULL;
2500 trans->bytes_reserved = 0;
531cb13f 2501 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2502 if (ret && !err)
2503 err = ret;
76dda93c 2504 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2505out_release:
2506 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2507out_up_write:
2508 up_write(&root->fs_info->subvol_sem);
2509out_unlock:
521e0546
DS		 2510 if (err) {
2511 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2512 root_flags = btrfs_root_flags(&dest->root_item);
2513 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2514 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2515 spin_unlock(&dest->root_item_lock);
2516 }
76dda93c
YZ		 2517 mutex_unlock(&inode->i_mutex);
2518 if (!err) {
efefb143 2519 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2520 btrfs_invalidate_inodes(dest);
2521 d_delete(dentry);
61155aa0 2522 ASSERT(dest->send_in_progress == 0);
fa6ac876
LB		 2523
2524 /* the last ref */
57cdc8db
DS		 2525 if (dest->ino_cache_inode) {
2526 iput(dest->ino_cache_inode);
2527 dest->ino_cache_inode = NULL;
fa6ac876 2528 }
76dda93c
YZ		 2529 }
2530out_dput:
2531 dput(dentry);
2532out_unlock_dir:
2533 mutex_unlock(&dir->i_mutex);
e43f998e 2534out_drop_write:
2a79f17e 2535 mnt_drop_write_file(file);
76dda93c
YZ		 2536out:
2537 kfree(vol_args);
2538 return err;
2539}
2540
1e701a32 2541static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2542{
496ad9aa 2543 struct inode *inode = file_inode(file);
f46b5a66 2544 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2545 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2546 int ret;
2547
25122d15
ID		 2548 ret = mnt_want_write_file(file);
2549 if (ret)
2550 return ret;
b83cc969 2551
25122d15
ID		 2552 if (btrfs_root_readonly(root)) {
2553 ret = -EROFS;
2554 goto out;
5ac00add 2555 }
f46b5a66
CH		 2556
2557 switch (inode->i_mode & S_IFMT) {
2558 case S_IFDIR:
e441d54d
CM		 2559 if (!capable(CAP_SYS_ADMIN)) {
2560 ret = -EPERM;
2561 goto out;
2562 }
de78b51a 2563 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2564 if (ret)
2565 goto out;
de78b51a 2566 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2567 break;
2568 case S_IFREG:
e441d54d
CM		 2569 if (!(file->f_mode & FMODE_WRITE)) {
2570 ret = -EINVAL;
2571 goto out;
2572 }
1e701a32
CM		 2573
2574 range = kzalloc(sizeof(*range), GFP_KERNEL);
2575 if (!range) {
2576 ret = -ENOMEM;
2577 goto out;
2578 }
2579
2580 if (argp) {
2581 if (copy_from_user(range, argp,
2582 sizeof(*range))) {
2583 ret = -EFAULT;
2584 kfree(range);
683be16e 2585 goto out;
1e701a32
CM		 2586 }
2587 /* compression requires us to start the IO */
2588 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2589 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2590 range->extent_thresh = (u32)-1;
2591 }
2592 } else {
2593 /* the rest are all set to zero by kzalloc */
2594 range->len = (u64)-1;
2595 }
496ad9aa 2596 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2597 range, 0, 0);
2598 if (ret > 0)
2599 ret = 0;
1e701a32 2600 kfree(range);
f46b5a66 2601 break;
8929ecfa
YZ		 2602 default:
2603 ret = -EINVAL;
f46b5a66 2604 }
e441d54d 2605out:
25122d15 2606 mnt_drop_write_file(file);
e441d54d 2607 return ret;
f46b5a66
CH		 2608}
2609
b2950863 2610static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2611{
2612 struct btrfs_ioctl_vol_args *vol_args;
2613 int ret;
2614
e441d54d
CM		 2615 if (!capable(CAP_SYS_ADMIN))
2616 return -EPERM;
2617
5ac00add
SB		 2618 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2619 1)) {
e57138b3 2620 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 2621 }
2622
5ac00add 2623 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2624 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2625 if (IS_ERR(vol_args)) {
2626 ret = PTR_ERR(vol_args);
2627 goto out;
2628 }
f46b5a66 2629
5516e595 2630 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2631 ret = btrfs_init_new_device(root, vol_args->name);
2632
43d20761
AJ		 2633 if (!ret)
2634 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2635
f46b5a66 2636 kfree(vol_args);
c9e9f97b
ID		 2637out:
2638 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2639 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2640 return ret;
2641}
2642
da24927b 2643static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2644{
496ad9aa 2645 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2646 struct btrfs_ioctl_vol_args *vol_args;
2647 int ret;
2648
e441d54d
CM		 2649 if (!capable(CAP_SYS_ADMIN))
2650 return -EPERM;
2651
da24927b
MX		 2652 ret = mnt_want_write_file(file);
2653 if (ret)
2654 return ret;
c146afad 2655
dae7b665 2656 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2657 if (IS_ERR(vol_args)) {
2658 ret = PTR_ERR(vol_args);
c47ca32d 2659 goto err_drop;
c9e9f97b 2660 }
f46b5a66 2661
5516e595 2662 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2663
183860f6
AJ		 2664 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2665 1)) {
2666 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2667 goto out;
2668 }
2669
2670 mutex_lock(&root->fs_info->volume_mutex);
2671 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2672 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2673 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6 2674
ec95d491
AJ		 2675 if (!ret)
2676 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2677
183860f6
AJ		 2678out:
2679 kfree(vol_args);
c47ca32d 2680err_drop:
4ac20c70 2681 mnt_drop_write_file(file);
f46b5a66
CH		 2682 return ret;
2683}
2684
475f6387
JS		 2685static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2686{
027ed2f0 2687 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2688 struct btrfs_device *device;
2689 struct btrfs_device *next;
2690 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2691 int ret = 0;
475f6387 2692
027ed2f0
LZ		 2693 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2694 if (!fi_args)
2695 return -ENOMEM;
2696
f7171750 2697 mutex_lock(&fs_devices->device_list_mutex);
027ed2f0
LZ		 2698 fi_args->num_devices = fs_devices->num_devices;
2699 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387 2700
475f6387 2701 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2702 if (device->devid > fi_args->max_id)
2703 fi_args->max_id = device->devid;
475f6387
JS		 2704 }
2705 mutex_unlock(&fs_devices->device_list_mutex);
2706
80a773fb
DS		 2707 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2708 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2709 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2710
027ed2f0
LZ		 2711 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2712 ret = -EFAULT;
475f6387 2713
027ed2f0
LZ		 2714 kfree(fi_args);
2715 return ret;
475f6387
JS		 2716}
2717
2718static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2719{
2720 struct btrfs_ioctl_dev_info_args *di_args;
2721 struct btrfs_device *dev;
2722 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2723 int ret = 0;
2724 char *s_uuid = NULL;
475f6387 2725
475f6387
JS		 2726 di_args = memdup_user(arg, sizeof(*di_args));
2727 if (IS_ERR(di_args))
2728 return PTR_ERR(di_args);
2729
dd5f9615 2730 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2731 s_uuid = di_args->uuid;
2732
2733 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2734 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2735
2736 if (!dev) {
2737 ret = -ENODEV;
2738 goto out;
2739 }
2740
2741 di_args->devid = dev->devid;
7cc8e58d
MX		 2742 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2743 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
475f6387 2744 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2745 if (dev->name) {
606686ee
JB		 2746 struct rcu_string *name;
2747
2748 rcu_read_lock();
2749 name = rcu_dereference(dev->name);
2750 strncpy(di_args->path, name->str, sizeof(di_args->path));
2751 rcu_read_unlock();
a27202fb
JM		 2752 di_args->path[sizeof(di_args->path) - 1] = 0;
2753 } else {
99ba55ad 2754 di_args->path[0] = '\0';
a27202fb 2755 }
475f6387
JS		 2756
2757out:
55793c0d 2758 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2759 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2760 ret = -EFAULT;
2761
2762 kfree(di_args);
2763 return ret;
2764}
2765
416161db
MF		 2766static struct page *extent_same_get_page(struct inode *inode, u64 off)
2767{
2768 struct page *page;
2769 pgoff_t index;
2770 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2771
2772 index = off >> PAGE_CACHE_SHIFT;
2773
2774 page = grab_cache_page(inode->i_mapping, index);
2775 if (!page)
2776 return NULL;
2777
2778 if (!PageUptodate(page)) {
2779 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2780 0))
2781 return NULL;
2782 lock_page(page);
2783 if (!PageUptodate(page)) {
2784 unlock_page(page);
2785 page_cache_release(page);
2786 return NULL;
2787 }
2788 }
2789 unlock_page(page);
2790
2791 return page;
2792}
2793
77fe20dc
MF		 2794static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2795{
2796 /* do any pending delalloc/csum calc on src, one way or
2797 another, and lock file content */
2798 while (1) {
2799 struct btrfs_ordered_extent *ordered;
2800 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2801 ordered = btrfs_lookup_first_ordered_extent(inode,
2802 off + len - 1);
ff5df9b8
FM		 2803 if ((!ordered ||
2804 ordered->file_offset + ordered->len <= off ||
2805 ordered->file_offset >= off + len) &&
77fe20dc 2806 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
ff5df9b8
FM		 2807 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2808 if (ordered)
2809 btrfs_put_ordered_extent(ordered);
77fe20dc 2810 break;
ff5df9b8 2811 }
77fe20dc
MF		 2812 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2813 if (ordered)
2814 btrfs_put_ordered_extent(ordered);
2815 btrfs_wait_ordered_range(inode, off, len);
2816 }
2817}
2818
416161db
MF		 2819static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2820 struct inode *inode2, u64 loff2, u64 len)
2821{
2822 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2823 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2824
2825 mutex_unlock(&inode1->i_mutex);
2826 mutex_unlock(&inode2->i_mutex);
2827}
2828
2829static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2830 struct inode *inode2, u64 loff2, u64 len)
2831{
2832 if (inode1 < inode2) {
2833 swap(inode1, inode2);
2834 swap(loff1, loff2);
2835 }
2836
2837 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2838 lock_extent_range(inode1, loff1, len);
2839 if (inode1 != inode2) {
2840 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2841 lock_extent_range(inode2, loff2, len);
2842 }
2843}
2844
2845static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2846 u64 dst_loff, u64 len)
2847{
2848 int ret = 0;
2849 struct page *src_page, *dst_page;
2850 unsigned int cmp_len = PAGE_CACHE_SIZE;
2851 void *addr, *dst_addr;
2852
2853 while (len) {
2854 if (len < PAGE_CACHE_SIZE)
2855 cmp_len = len;
2856
2857 src_page = extent_same_get_page(src, loff);
2858 if (!src_page)
2859 return -EINVAL;
2860 dst_page = extent_same_get_page(dst, dst_loff);
2861 if (!dst_page) {
2862 page_cache_release(src_page);
2863 return -EINVAL;
2864 }
2865 addr = kmap_atomic(src_page);
2866 dst_addr = kmap_atomic(dst_page);
2867
2868 flush_dcache_page(src_page);
2869 flush_dcache_page(dst_page);
2870
2871 if (memcmp(addr, dst_addr, cmp_len))
2872 ret = BTRFS_SAME_DATA_DIFFERS;
2873
2874 kunmap_atomic(addr);
2875 kunmap_atomic(dst_addr);
2876 page_cache_release(src_page);
2877 page_cache_release(dst_page);
2878
2879 if (ret)
2880 break;
2881
2882 loff += cmp_len;
2883 dst_loff += cmp_len;
2884 len -= cmp_len;
2885 }
2886
2887 return ret;
2888}
2889
2890static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2891{
2892 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2893
2894 if (off + len > inode->i_size || off + len < off)
2895 return -EINVAL;
2896 /* Check that we are block aligned - btrfs_clone() requires this */
2897 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2898 return -EINVAL;
2899
2900 return 0;
2901}
2902
2903static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2904 struct inode *dst, u64 dst_loff)
2905{
2906 int ret;
2907
2908 /*
2909 * btrfs_clone() can't handle extents in the same file
2910 * yet. Once that works, we can drop this check and replace it
2911 * with a check for the same inode, but overlapping extents.
2912 */
2913 if (src == dst)
2914 return -EINVAL;
2915
2916 btrfs_double_lock(src, loff, dst, dst_loff, len);
2917
2918 ret = extent_same_check_offsets(src, loff, len);
2919 if (ret)
2920 goto out_unlock;
2921
2922 ret = extent_same_check_offsets(dst, dst_loff, len);
2923 if (ret)
2924 goto out_unlock;
2925
2926 /* don't make the dst file partly checksummed */
2927 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2928 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2929 ret = -EINVAL;
2930 goto out_unlock;
2931 }
2932
2933 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2934 if (ret == 0)
2935 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2936
2937out_unlock:
2938 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2939
2940 return ret;
2941}
2942
2943#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2944
2945static long btrfs_ioctl_file_extent_same(struct file *file,
1c1c8747 2946 struct btrfs_ioctl_same_args __user *argp)
416161db 2947{
cbf8b8ca
MF		 2948 struct btrfs_ioctl_same_args *same;
2949 struct btrfs_ioctl_same_extent_info *info;
1c1c8747 2950 struct inode *src = file_inode(file);
416161db
MF		 2951 u64 off;
2952 u64 len;
2953 int i;
2954 int ret;
cbf8b8ca 2955 unsigned long size;
416161db
MF		 2956 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2957 bool is_admin = capable(CAP_SYS_ADMIN);
1c1c8747 2958 u16 count;
416161db
MF		 2959
2960 if (!(file->f_mode & FMODE_READ))
2961 return -EINVAL;
2962
2963 ret = mnt_want_write_file(file);
2964 if (ret)
2965 return ret;
2966
1c1c8747 2967 if (get_user(count, &argp->dest_count)) {
416161db
MF		 2968 ret = -EFAULT;
2969 goto out;
2970 }
2971
1c1c8747 2972 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
cbf8b8ca 2973
1c1c8747 2974 same = memdup_user(argp, size);
cbf8b8ca 2975
229eed43
GB		 2976 if (IS_ERR(same)) {
2977 ret = PTR_ERR(same);
cbf8b8ca
MF		 2978 goto out;
2979 }
2980
2981 off = same->logical_offset;
2982 len = same->length;
416161db
MF		 2983
2984 /*
2985 * Limit the total length we will dedupe for each operation.
2986 * This is intended to bound the total time spent in this
2987 * ioctl to something sane.
2988 */
2989 if (len > BTRFS_MAX_DEDUPE_LEN)
2990 len = BTRFS_MAX_DEDUPE_LEN;
2991
2992 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2993 /*
2994 * Btrfs does not support blocksize < page_size. As a
2995 * result, btrfs_cmp_data() won't correctly handle
2996 * this situation without an update.
2997 */
2998 ret = -EINVAL;
2999 goto out;
3000 }
3001
3002 ret = -EISDIR;
3003 if (S_ISDIR(src->i_mode))
3004 goto out;
3005
3006 ret = -EACCES;
3007 if (!S_ISREG(src->i_mode))
3008 goto out;
3009
cbf8b8ca 3010 /* pre-format output fields to sane values */
1c1c8747 3011 for (i = 0; i < count; i++) {
cbf8b8ca
MF		 3012 same->info[i].bytes_deduped = 0ULL;
3013 same->info[i].status = 0;
3014 }
416161db 3015
1c1c8747
AV		 3016 for (i = 0, info = same->info; i < count; i++, info++) {
3017 struct inode *dst;
3018 struct fd dst_file = fdget(info->fd);
3019 if (!dst_file.file) {
cbf8b8ca 3020 info->status = -EBADF;
1c1c8747 3021 continue;
416161db 3022 }
1c1c8747 3023 dst = file_inode(dst_file.file);
416161db 3024
1c1c8747 3025 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
cbf8b8ca 3026 info->status = -EINVAL;
1c1c8747
AV		 3027 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3028 info->status = -EXDEV;
3029 } else if (S_ISDIR(dst->i_mode)) {
cbf8b8ca 3030 info->status = -EISDIR;
1c1c8747 3031 } else if (!S_ISREG(dst->i_mode)) {
cbf8b8ca 3032 info->status = -EACCES;
1c1c8747
AV		 3033 } else {
3034 info->status = btrfs_extent_same(src, off, len, dst,
3035 info->logical_offset);
3036 if (info->status == 0)
3037 info->bytes_deduped += len;
416161db 3038 }
1c1c8747 3039 fdput(dst_file);
416161db
MF		 3040 }
3041
cbf8b8ca
MF		 3042 ret = copy_to_user(argp, same, size);
3043 if (ret)
3044 ret = -EFAULT;
3045
416161db
MF		 3046out:
3047 mnt_drop_write_file(file);
3048 return ret;
3049}
3050
fcebe456
JB		 3051/* Helper to check and see if this root currently has a ref on the given disk
3052 * bytenr. If it does then we need to update the quota for this root. This
3053 * doesn't do anything if quotas aren't enabled.
3054 */
3055static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3056 u64 disko)
3057{
3058 struct seq_list tree_mod_seq_elem = {};
3059 struct ulist *roots;
3060 struct ulist_iterator uiter;
3061 struct ulist_node *root_node = NULL;
3062 int ret;
3063
3064 if (!root->fs_info->quota_enabled)
3065 return 1;
3066
3067 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3068 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3069 tree_mod_seq_elem.seq, &roots);
3070 if (ret < 0)
3071 goto out;
3072 ret = 0;
3073 ULIST_ITER_INIT(&uiter);
3074 while ((root_node = ulist_next(roots, &uiter))) {
3075 if (root_node->val == root->objectid) {
3076 ret = 1;
3077 break;
3078 }
3079 }
3080 ulist_free(roots);
3081out:
3082 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3083 return ret;
3084}
3085
f82a9901
FM		 3086static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3087 struct inode *inode,
3088 u64 endoff,
3089 const u64 destoff,
3090 const u64 olen)
3091{
3092 struct btrfs_root *root = BTRFS_I(inode)->root;
3093 int ret;
3094
3095 inode_inc_iversion(inode);
3096 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3097 /*
3098 * We round up to the block size at eof when determining which
3099 * extents to clone above, but shouldn't round up the file size.
3100 */
3101 if (endoff > destoff + olen)
3102 endoff = destoff + olen;
3103 if (endoff > inode->i_size)
3104 btrfs_i_size_write(inode, endoff);
3105
3106 ret = btrfs_update_inode(trans, root, inode);
3107 if (ret) {
3108 btrfs_abort_transaction(trans, root, ret);
3109 btrfs_end_transaction(trans, root);
3110 goto out;
3111 }
3112 ret = btrfs_end_transaction(trans, root);
3113out:
3114 return ret;
3115}
3116
7ffbb598
FM		 3117static void clone_update_extent_map(struct inode *inode,
3118 const struct btrfs_trans_handle *trans,
3119 const struct btrfs_path *path,
7ffbb598
FM		 3120 const u64 hole_offset,
3121 const u64 hole_len)
3122{
3123 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3124 struct extent_map *em;
3125 int ret;
3126
3127 em = alloc_extent_map();
3128 if (!em) {
3129 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3130 &BTRFS_I(inode)->runtime_flags);
3131 return;
3132 }
3133
14f59796
FM		 3134 if (path) {
3135 struct btrfs_file_extent_item *fi;
3136
3137 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3138 struct btrfs_file_extent_item);
7ffbb598
FM		 3139 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3140 em->generation = -1;
3141 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3142 BTRFS_FILE_EXTENT_INLINE)
3143 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3144 &BTRFS_I(inode)->runtime_flags);
3145 } else {
3146 em->start = hole_offset;
3147 em->len = hole_len;
3148 em->ram_bytes = em->len;
3149 em->orig_start = hole_offset;
3150 em->block_start = EXTENT_MAP_HOLE;
3151 em->block_len = 0;
3152 em->orig_block_len = 0;
3153 em->compress_type = BTRFS_COMPRESS_NONE;
3154 em->generation = trans->transid;
3155 }
3156
3157 while (1) {
3158 write_lock(&em_tree->lock);
3159 ret = add_extent_mapping(em_tree, em, 1);
3160 write_unlock(&em_tree->lock);
3161 if (ret != -EEXIST) {
3162 free_extent_map(em);
3163 break;
3164 }
3165 btrfs_drop_extent_cache(inode, em->start,
3166 em->start + em->len - 1, 0);
3167 }
3168
ee39b432 3169 if (ret)
7ffbb598
FM		 3170 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3171 &BTRFS_I(inode)->runtime_flags);
3172}
3173
32b7c687
MF		 3174/**
3175 * btrfs_clone() - clone a range from inode file to another
3176 *
3177 * @src: Inode to clone from
3178 * @inode: Inode to clone to
3179 * @off: Offset within source to start clone from
3180 * @olen: Original length, passed by user, of range to clone
3181 * @olen_aligned: Block-aligned value of olen, extent_same uses
3182 * identical values here
3183 * @destoff: Offset within @inode to start clone
3184 */
3185static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901
FM		 3186 const u64 off, const u64 olen, const u64 olen_aligned,
3187 const u64 destoff)
f46b5a66 3188{
f46b5a66 3189 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3190 struct btrfs_path *path = NULL;
f46b5a66 3191 struct extent_buffer *leaf;
32b7c687
MF		 3192 struct btrfs_trans_handle *trans;
3193 char *buf = NULL;
ae01a0ab 3194 struct btrfs_key key;
f46b5a66
CH		 3195 u32 nritems;
3196 int slot;
ae01a0ab 3197 int ret;
fcebe456 3198 int no_quota;
f82a9901 3199 const u64 len = olen_aligned;
fcebe456 3200 u64 last_disko = 0;
f82a9901 3201 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3202
3203 ret = -ENOMEM;
707e8a07 3204 buf = vmalloc(root->nodesize);
ae01a0ab 3205 if (!buf)
32b7c687 3206 return ret;
ae01a0ab
YZ		 3207
3208 path = btrfs_alloc_path();
3209 if (!path) {
3210 vfree(buf);
32b7c687 3211 return ret;
d525e8ab
LZ		 3212 }
3213
32b7c687 3214 path->reada = 2;
c5c9cd4d 3215 /* clone data */
33345d01 3216 key.objectid = btrfs_ino(src);
ae01a0ab 3217 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3218 key.offset = off;
f46b5a66
CH		 3219
3220 while (1) {
3221 /*
3222 * note the key will change type as we walk through the
3223 * tree.
3224 */
e4355f34 3225 path->leave_spinning = 1;
362a20c5
DS		 3226 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3227 0, 0);
f46b5a66
CH		 3228 if (ret < 0)
3229 goto out;
2c463823
FM		 3230 /*
3231 * First search, if no extent item that starts at offset off was
3232 * found but the previous item is an extent item, it's possible
3233 * it might overlap our target range, therefore process it.
3234 */
3235 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3236 btrfs_item_key_to_cpu(path->nodes[0], &key,
3237 path->slots[0] - 1);
3238 if (key.type == BTRFS_EXTENT_DATA_KEY)
3239 path->slots[0]--;
3240 }
f46b5a66 3241
ae01a0ab 3242 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3243process_slot:
fcebe456 3244 no_quota = 1;
ae01a0ab 3245 if (path->slots[0] >= nritems) {
362a20c5 3246 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3247 if (ret < 0)
3248 goto out;
3249 if (ret > 0)
3250 break;
ae01a0ab 3251 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3252 }
3253 leaf = path->nodes[0];
3254 slot = path->slots[0];
f46b5a66 3255
ae01a0ab 3256 btrfs_item_key_to_cpu(leaf, &key, slot);
962a298f 3257 if (key.type > BTRFS_EXTENT_DATA_KEY ||
33345d01 3258 key.objectid != btrfs_ino(src))
f46b5a66
CH		 3259 break;
3260
962a298f 3261 if (key.type == BTRFS_EXTENT_DATA_KEY) {
c5c9cd4d
SW		 3262 struct btrfs_file_extent_item *extent;
3263 int type;
31840ae1
ZY		 3264 u32 size;
3265 struct btrfs_key new_key;
c5c9cd4d
SW		 3266 u64 disko = 0, diskl = 0;
3267 u64 datao = 0, datal = 0;
3268 u8 comp;
f82a9901 3269 u64 drop_start;
31840ae1 3270
c5c9cd4d
SW		 3271 extent = btrfs_item_ptr(leaf, slot,
3272 struct btrfs_file_extent_item);
3273 comp = btrfs_file_extent_compression(leaf, extent);
3274 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3275 if (type == BTRFS_FILE_EXTENT_REG ||
3276 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3277 disko = btrfs_file_extent_disk_bytenr(leaf,
3278 extent);
3279 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3280 extent);
c5c9cd4d 3281 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3282 datal = btrfs_file_extent_num_bytes(leaf,
3283 extent);
c5c9cd4d
SW		 3284 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3285 /* take upper bound, may be compressed */
3286 datal = btrfs_file_extent_ram_bytes(leaf,
3287 extent);
3288 }
31840ae1 3289
2c463823
FM		 3290 /*
3291 * The first search might have left us at an extent
3292 * item that ends before our target range's start, can
3293 * happen if we have holes and NO_HOLES feature enabled.
3294 */
3295 if (key.offset + datal <= off) {
e4355f34
FDBM		 3296 path->slots[0]++;
3297 goto process_slot;
2c463823
FM		 3298 } else if (key.offset >= off + len) {
3299 break;
e4355f34
FDBM		 3300 }
3301
3302 size = btrfs_item_size_nr(leaf, slot);
3303 read_extent_buffer(leaf, buf,
3304 btrfs_item_ptr_offset(leaf, slot),
3305 size);
3306
3307 btrfs_release_path(path);
3308 path->leave_spinning = 0;
c5c9cd4d 3309
31840ae1 3310 memcpy(&new_key, &key, sizeof(new_key));
33345d01 3311 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 3312 if (off <= key.offset)
3313 new_key.offset = key.offset + destoff - off;
3314 else
3315 new_key.offset = destoff;
31840ae1 3316
f82a9901
FM		 3317 /*
3318 * Deal with a hole that doesn't have an extent item
3319 * that represents it (NO_HOLES feature enabled).
3320 * This hole is either in the middle of the cloning
3321 * range or at the beginning (fully overlaps it or
3322 * partially overlaps it).
3323 */
3324 if (new_key.offset != last_dest_end)
3325 drop_start = last_dest_end;
3326 else
3327 drop_start = new_key.offset;
3328
b6f3409b
SW		 3329 /*
3330 * 1 - adjusting old extent (we may have to split it)
3331 * 1 - add new extent
3332 * 1 - inode update
3333 */
3334 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3335 if (IS_ERR(trans)) {
3336 ret = PTR_ERR(trans);
3337 goto out;
3338 }
3339
c8a894d7
CM		 3340 if (type == BTRFS_FILE_EXTENT_REG ||
3341 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3342 /*
3343 * a | --- range to clone ---| b
3344 * | ------------- extent ------------- |
3345 */
3346
93915584 3347 /* subtract range b */
d72c0842
LZ		 3348 if (key.offset + datal > off + len)
3349 datal = off + len - key.offset;
3350
93915584 3351 /* subtract range a */
a22285a6
YZ		 3352 if (off > key.offset) {
3353 datao += off - key.offset;
3354 datal -= off - key.offset;
3355 }
3356
5dc562c5 3357 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3358 drop_start,
a22285a6 3359 new_key.offset + datal,
2671485d 3360 1);
79787eaa 3361 if (ret) {
3f9e3df8 3362 if (ret != -EOPNOTSUPP)
00fdf13a
LB		 3363 btrfs_abort_transaction(trans,
3364 root, ret);
79787eaa
JM		 3365 btrfs_end_transaction(trans, root);
3366 goto out;
3367 }
a22285a6 3368
c5c9cd4d
SW		 3369 ret = btrfs_insert_empty_item(trans, root, path,
3370 &new_key, size);
79787eaa
JM		 3371 if (ret) {
3372 btrfs_abort_transaction(trans, root,
3373 ret);
3374 btrfs_end_transaction(trans, root);
3375 goto out;
3376 }
c5c9cd4d
SW		 3377
3378 leaf = path->nodes[0];
3379 slot = path->slots[0];
3380 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3381 btrfs_item_ptr_offset(leaf, slot),
3382 size);
ae01a0ab 3383
c5c9cd4d 3384 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3385 struct btrfs_file_extent_item);
c5c9cd4d 3386
c5c9cd4d
SW		 3387 /* disko == 0 means it's a hole */
3388 if (!disko)
3389 datao = 0;
c5c9cd4d
SW		 3390
3391 btrfs_set_file_extent_offset(leaf, extent,
3392 datao);
3393 btrfs_set_file_extent_num_bytes(leaf, extent,
3394 datal);
fcebe456
JB		 3395
3396 /*
3397 * We need to look up the roots that point at
3398 * this bytenr and see if the new root does. If
3399 * it does not we need to make sure we update
3400 * quotas appropriately.
3401 */
3402 if (disko && root != BTRFS_I(src)->root &&
3403 disko != last_disko) {
3404 no_quota = check_ref(trans, root,
3405 disko);
3406 if (no_quota < 0) {
3407 btrfs_abort_transaction(trans,
3408 root,
3409 ret);
3410 btrfs_end_transaction(trans,
3411 root);
3412 ret = no_quota;
3413 goto out;
3414 }
3415 }
3416
c5c9cd4d
SW		 3417 if (disko) {
3418 inode_add_bytes(inode, datal);
ae01a0ab 3419 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3420 disko, diskl, 0,
3421 root->root_key.objectid,
33345d01 3422 btrfs_ino(inode),
66d7e7f0 3423 new_key.offset - datao,
fcebe456 3424 no_quota);
79787eaa
JM		 3425 if (ret) {
3426 btrfs_abort_transaction(trans,
3427 root,
3428 ret);
3429 btrfs_end_transaction(trans,
3430 root);
3431 goto out;
3432
3433 }
f46b5a66 3434 }
c5c9cd4d
SW		 3435 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3436 u64 skip = 0;
3437 u64 trim = 0;
d3ecfcdf
LB		 3438 u64 aligned_end = 0;
3439
c5c9cd4d
SW		 3440 if (off > key.offset) {
3441 skip = off - key.offset;
3442 new_key.offset += skip;
3443 }
d397712b 3444
aa42ffd9
LB		 3445 if (key.offset + datal > off + len)
3446 trim = key.offset + datal - (off + len);
d397712b 3447
c5c9cd4d 3448 if (comp && (skip || trim)) {
c5c9cd4d 3449 ret = -EINVAL;
a22285a6 3450 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3451 goto out;
3452 }
3453 size -= skip + trim;
3454 datal -= skip + trim;
a22285a6 3455
d3ecfcdf
LB		 3456 aligned_end = ALIGN(new_key.offset + datal,
3457 root->sectorsize);
5dc562c5 3458 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3459 drop_start,
d3ecfcdf 3460 aligned_end,
2671485d 3461 1);
79787eaa 3462 if (ret) {
3f9e3df8 3463 if (ret != -EOPNOTSUPP)
3a29bc09
CM		 3464 btrfs_abort_transaction(trans,
3465 root, ret);
79787eaa
JM		 3466 btrfs_end_transaction(trans, root);
3467 goto out;
3468 }
a22285a6 3469
c5c9cd4d
SW		 3470 ret = btrfs_insert_empty_item(trans, root, path,
3471 &new_key, size);
79787eaa
JM		 3472 if (ret) {
3473 btrfs_abort_transaction(trans, root,
3474 ret);
3475 btrfs_end_transaction(trans, root);
3476 goto out;
3477 }
c5c9cd4d
SW		 3478
3479 if (skip) {
d397712b
CM		 3480 u32 start =
3481 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3482 memmove(buf+start, buf+start+skip,
3483 datal);
3484 }
3485
3486 leaf = path->nodes[0];
3487 slot = path->slots[0];
3488 write_extent_buffer(leaf, buf,
3489 btrfs_item_ptr_offset(leaf, slot),
3490 size);
3491 inode_add_bytes(inode, datal);
f46b5a66 3492 }
c5c9cd4d 3493
7ffbb598
FM		 3494 /* If we have an implicit hole (NO_HOLES feature). */
3495 if (drop_start < new_key.offset)
3496 clone_update_extent_map(inode, trans,
14f59796 3497 NULL, drop_start,
7ffbb598
FM		 3498 new_key.offset - drop_start);
3499
14f59796 3500 clone_update_extent_map(inode, trans, path, 0, 0);
7ffbb598 3501
c5c9cd4d 3502 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3503 btrfs_release_path(path);
c5c9cd4d 3504
62e2390e
FM		 3505 last_dest_end = ALIGN(new_key.offset + datal,
3506 root->sectorsize);
f82a9901
FM		 3507 ret = clone_finish_inode_update(trans, inode,
3508 last_dest_end,
3509 destoff, olen);
3510 if (ret)
79787eaa 3511 goto out;
2c463823
FM		 3512 if (new_key.offset + datal >= destoff + len)
3513 break;
a22285a6 3514 }
b3b4aa74 3515 btrfs_release_path(path);
f46b5a66 3516 key.offset++;
f46b5a66 3517 }
f46b5a66 3518 ret = 0;
32b7c687 3519
f82a9901
FM		 3520 if (last_dest_end < destoff + len) {
3521 /*
3522 * We have an implicit hole (NO_HOLES feature is enabled) that
3523 * fully or partially overlaps our cloning range at its end.
3524 */
3525 btrfs_release_path(path);
3526
3527 /*
3528 * 1 - remove extent(s)
3529 * 1 - inode update
3530 */
3531 trans = btrfs_start_transaction(root, 2);
3532 if (IS_ERR(trans)) {
3533 ret = PTR_ERR(trans);
3534 goto out;
3535 }
3536 ret = btrfs_drop_extents(trans, root, inode,
3537 last_dest_end, destoff + len, 1);
3538 if (ret) {
3539 if (ret != -EOPNOTSUPP)
3540 btrfs_abort_transaction(trans, root, ret);
3541 btrfs_end_transaction(trans, root);
3542 goto out;
3543 }
14f59796
FM		 3544 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3545 destoff + len - last_dest_end);
f82a9901
FM		 3546 ret = clone_finish_inode_update(trans, inode, destoff + len,
3547 destoff, olen);
3548 }
3549
f46b5a66 3550out:
32b7c687
MF		 3551 btrfs_free_path(path);
3552 vfree(buf);
3553 return ret;
3554}
3555
3556static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3557 u64 off, u64 olen, u64 destoff)
3558{
54563d41 3559 struct inode *inode = file_inode(file);
32b7c687
MF		 3560 struct btrfs_root *root = BTRFS_I(inode)->root;
3561 struct fd src_file;
3562 struct inode *src;
3563 int ret;
3564 u64 len = olen;
3565 u64 bs = root->fs_info->sb->s_blocksize;
3566 int same_inode = 0;
3567
3568 /*
3569 * TODO:
3570 * - split compressed inline extents. annoying: we need to
3571 * decompress into destination's address_space (the file offset
3572 * may change, so source mapping won't do), then recompress (or
3573 * otherwise reinsert) a subrange.
00fdf13a
LB		 3574 *
3575 * - split destination inode's inline extents. The inline extents can
3576 * be either compressed or non-compressed.
32b7c687
MF		 3577 */
3578
3579 /* the destination must be opened for writing */
3580 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3581 return -EINVAL;
3582
3583 if (btrfs_root_readonly(root))
3584 return -EROFS;
3585
3586 ret = mnt_want_write_file(file);
3587 if (ret)
3588 return ret;
3589
3590 src_file = fdget(srcfd);
3591 if (!src_file.file) {
3592 ret = -EBADF;
3593 goto out_drop_write;
3594 }
3595
3596 ret = -EXDEV;
3597 if (src_file.file->f_path.mnt != file->f_path.mnt)
3598 goto out_fput;
3599
3600 src = file_inode(src_file.file);
3601
3602 ret = -EINVAL;
3603 if (src == inode)
3604 same_inode = 1;
3605
3606 /* the src must be open for reading */
3607 if (!(src_file.file->f_mode & FMODE_READ))
3608 goto out_fput;
3609
3610 /* don't make the dst file partly checksummed */
3611 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3612 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3613 goto out_fput;
3614
3615 ret = -EISDIR;
3616 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3617 goto out_fput;
3618
3619 ret = -EXDEV;
3620 if (src->i_sb != inode->i_sb)
3621 goto out_fput;
3622
3623 if (!same_inode) {
3624 if (inode < src) {
3625 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3626 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3627 } else {
3628 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3629 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3630 }
3631 } else {
3632 mutex_lock(&src->i_mutex);
3633 }
3634
3635 /* determine range to clone */
3636 ret = -EINVAL;
3637 if (off + len > src->i_size || off + len < off)
3638 goto out_unlock;
3639 if (len == 0)
3640 olen = len = src->i_size - off;
3641 /* if we extend to eof, continue to block boundary */
3642 if (off + len == src->i_size)
3643 len = ALIGN(src->i_size, bs) - off;
3644
3645 /* verify the end result is block aligned */
3646 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3647 !IS_ALIGNED(destoff, bs))
3648 goto out_unlock;
3649
3650 /* verify if ranges are overlapped within the same file */
3651 if (same_inode) {
3652 if (destoff + len > off && destoff < off + len)
3653 goto out_unlock;
3654 }
3655
3656 if (destoff > inode->i_size) {
3657 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3658 if (ret)
3659 goto out_unlock;
3660 }
3661
c125b8bf
FM		 3662 /*
3663 * Lock the target range too. Right after we replace the file extent
3664 * items in the fs tree (which now point to the cloned data), we might
3665 * have a worker replace them with extent items relative to a write
3666 * operation that was issued before this clone operation (i.e. confront
3667 * with inode.c:btrfs_finish_ordered_io).
3668 */
3669 if (same_inode) {
3670 u64 lock_start = min_t(u64, off, destoff);
3671 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
32b7c687 3672
c125b8bf
FM		 3673 lock_extent_range(src, lock_start, lock_len);
3674 } else {
3675 lock_extent_range(src, off, len);
3676 lock_extent_range(inode, destoff, len);
3677 }
32b7c687
MF		 3678
3679 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3680
c125b8bf
FM		 3681 if (same_inode) {
3682 u64 lock_start = min_t(u64, off, destoff);
3683 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3684
3685 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3686 } else {
3687 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3688 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3689 destoff + len - 1);
3690 }
3691 /*
3692 * Truncate page cache pages so that future reads will see the cloned
3693 * data immediately and not the previous data.
3694 */
3695 truncate_inode_pages_range(&inode->i_data, destoff,
3696 PAGE_CACHE_ALIGN(destoff + len) - 1);
f46b5a66 3697out_unlock:
c57c2b3e
FDBM		 3698 if (!same_inode) {
3699 if (inode < src) {
3700 mutex_unlock(&src->i_mutex);
3701 mutex_unlock(&inode->i_mutex);
3702 } else {
3703 mutex_unlock(&inode->i_mutex);
3704 mutex_unlock(&src->i_mutex);
3705 }
3706 } else {
3707 mutex_unlock(&src->i_mutex);
3708 }
f46b5a66 3709out_fput:
2903ff01 3710 fdput(src_file);
ab67b7c1 3711out_drop_write:
2a79f17e 3712 mnt_drop_write_file(file);
f46b5a66
CH		 3713 return ret;
3714}
3715
7a865e8a 3716static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3717{
3718 struct btrfs_ioctl_clone_range_args args;
3719
7a865e8a 3720 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3721 return -EFAULT;
3722 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3723 args.src_length, args.dest_offset);
3724}
3725
f46b5a66
CH		 3726/*
3727 * there are many ways the trans_start and trans_end ioctls can lead
3728 * to deadlocks. They should only be used by applications that
3729 * basically own the machine, and have a very in depth understanding
3730 * of all the possible deadlocks and enospc problems.
3731 */
b2950863 3732static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3733{
496ad9aa 3734 struct inode *inode = file_inode(file);
f46b5a66
CH		 3735 struct btrfs_root *root = BTRFS_I(inode)->root;
3736 struct btrfs_trans_handle *trans;
1ab86aed 3737 int ret;
f46b5a66 3738
1ab86aed 3739 ret = -EPERM;
df5b5520 3740 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3741 goto out;
df5b5520 3742
1ab86aed
SW		 3743 ret = -EINPROGRESS;
3744 if (file->private_data)
f46b5a66 3745 goto out;
9ca9ee09 3746
b83cc969
LZ		 3747 ret = -EROFS;
3748 if (btrfs_root_readonly(root))
3749 goto out;
3750
a561be71 3751 ret = mnt_want_write_file(file);
c146afad
YZ		 3752 if (ret)
3753 goto out;
3754
a4abeea4 3755 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3756
1ab86aed 3757 ret = -ENOMEM;
7a7eaa40 3758 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3759 if (IS_ERR(trans))
1ab86aed
SW		 3760 goto out_drop;
3761
3762 file->private_data = trans;
3763 return 0;
3764
3765out_drop:
a4abeea4 3766 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3767 mnt_drop_write_file(file);
f46b5a66 3768out:
f46b5a66
CH		 3769 return ret;
3770}
3771
6ef5ed0d
JB		 3772static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3773{
496ad9aa 3774 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3775 struct btrfs_root *root = BTRFS_I(inode)->root;
3776 struct btrfs_root *new_root;
3777 struct btrfs_dir_item *di;
3778 struct btrfs_trans_handle *trans;
3779 struct btrfs_path *path;
3780 struct btrfs_key location;
3781 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3782 u64 objectid = 0;
3783 u64 dir_id;
3c04ce01 3784 int ret;
6ef5ed0d
JB		 3785
3786 if (!capable(CAP_SYS_ADMIN))
3787 return -EPERM;
3788
3c04ce01
MX		 3789 ret = mnt_want_write_file(file);
3790 if (ret)
3791 return ret;
3792
3793 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3794 ret = -EFAULT;
3795 goto out;
3796 }
6ef5ed0d
JB		 3797
3798 if (!objectid)
1cecf579 3799 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3800
3801 location.objectid = objectid;
3802 location.type = BTRFS_ROOT_ITEM_KEY;
3803 location.offset = (u64)-1;
3804
3805 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3806 if (IS_ERR(new_root)) {
3807 ret = PTR_ERR(new_root);
3808 goto out;
3809 }
6ef5ed0d 3810
6ef5ed0d 3811 path = btrfs_alloc_path();
3c04ce01
MX		 3812 if (!path) {
3813 ret = -ENOMEM;
3814 goto out;
3815 }
6ef5ed0d
JB		 3816 path->leave_spinning = 1;
3817
3818 trans = btrfs_start_transaction(root, 1);
98d5dc13 3819 if (IS_ERR(trans)) {
6ef5ed0d 3820 btrfs_free_path(path);
3c04ce01
MX		 3821 ret = PTR_ERR(trans);
3822 goto out;
6ef5ed0d
JB		 3823 }
3824
6c41761f 3825 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3826 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3827 dir_id, "default", 7, 1);
cf1e99a4 3828 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3829 btrfs_free_path(path);
3830 btrfs_end_transaction(trans, root);
efe120a0
FH		 3831 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3832 "item, this isn't going to work");
3c04ce01
MX		 3833 ret = -ENOENT;
3834 goto out;
6ef5ed0d
JB		 3835 }
3836
3837 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3838 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3839 btrfs_mark_buffer_dirty(path->nodes[0]);
3840 btrfs_free_path(path);
3841
2b0ce2c2 3842 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3843 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3844out:
3845 mnt_drop_write_file(file);
3846 return ret;
6ef5ed0d
JB		 3847}
3848
5af3e8cc
SB		 3849void btrfs_get_block_group_info(struct list_head *groups_list,
3850 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3851{
3852 struct btrfs_block_group_cache *block_group;
3853
3854 space->total_bytes = 0;
3855 space->used_bytes = 0;
3856 space->flags = 0;
3857 list_for_each_entry(block_group, groups_list, list) {
3858 space->flags = block_group->flags;
3859 space->total_bytes += block_group->key.offset;
3860 space->used_bytes +=
3861 btrfs_block_group_used(&block_group->item);
3862 }
3863}
3864
48a3b636 3865static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3866{
3867 struct btrfs_ioctl_space_args space_args;
3868 struct btrfs_ioctl_space_info space;
3869 struct btrfs_ioctl_space_info *dest;
7fde62bf 3870 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3871 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3872 struct btrfs_space_info *info;
bf5fc093
JB		 3873 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3874 BTRFS_BLOCK_GROUP_SYSTEM,
3875 BTRFS_BLOCK_GROUP_METADATA,
3876 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3877 int num_types = 4;
7fde62bf 3878 int alloc_size;
1406e432 3879 int ret = 0;
51788b1b 3880 u64 slot_count = 0;
bf5fc093 3881 int i, c;
1406e432
JB		 3882
3883 if (copy_from_user(&space_args,
3884 (struct btrfs_ioctl_space_args __user *)arg,
3885 sizeof(space_args)))
3886 return -EFAULT;
3887
bf5fc093
JB		 3888 for (i = 0; i < num_types; i++) {
3889 struct btrfs_space_info *tmp;
3890
3891 info = NULL;
3892 rcu_read_lock();
3893 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3894 list) {
3895 if (tmp->flags == types[i]) {
3896 info = tmp;
3897 break;
3898 }
3899 }
3900 rcu_read_unlock();
3901
3902 if (!info)
3903 continue;
3904
3905 down_read(&info->groups_sem);
3906 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3907 if (!list_empty(&info->block_groups[c]))
3908 slot_count++;
3909 }
3910 up_read(&info->groups_sem);
3911 }
7fde62bf 3912
36523e95
DS		 3913 /*
3914 * Global block reserve, exported as a space_info
3915 */
3916 slot_count++;
3917
7fde62bf
CM		 3918 /* space_slots == 0 means they are asking for a count */
3919 if (space_args.space_slots == 0) {
3920 space_args.total_spaces = slot_count;
3921 goto out;
3922 }
bf5fc093 3923
51788b1b 3924 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3925
7fde62bf 3926 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3927
7fde62bf
CM		 3928 /* we generally have at most 6 or so space infos, one for each raid
3929 * level. So, a whole page should be more than enough for everyone
3930 */
3931 if (alloc_size > PAGE_CACHE_SIZE)
3932 return -ENOMEM;
3933
1406e432 3934 space_args.total_spaces = 0;
7fde62bf
CM		 3935 dest = kmalloc(alloc_size, GFP_NOFS);
3936 if (!dest)
3937 return -ENOMEM;
3938 dest_orig = dest;
1406e432 3939
7fde62bf 3940 /* now we have a buffer to copy into */
bf5fc093
JB		 3941 for (i = 0; i < num_types; i++) {
3942 struct btrfs_space_info *tmp;
3943
51788b1b
DR		 3944 if (!slot_count)
3945 break;
3946
bf5fc093
JB		 3947 info = NULL;
3948 rcu_read_lock();
3949 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3950 list) {
3951 if (tmp->flags == types[i]) {
3952 info = tmp;
3953 break;
3954 }
3955 }
3956 rcu_read_unlock();
7fde62bf 3957
bf5fc093
JB		 3958 if (!info)
3959 continue;
3960 down_read(&info->groups_sem);
3961 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3962 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3963 btrfs_get_block_group_info(
3964 &info->block_groups[c], &space);
bf5fc093
JB		 3965 memcpy(dest, &space, sizeof(space));
3966 dest++;
3967 space_args.total_spaces++;
51788b1b 3968 slot_count--;
bf5fc093 3969 }
51788b1b
DR		 3970 if (!slot_count)
3971 break;
bf5fc093
JB		 3972 }
3973 up_read(&info->groups_sem);
1406e432 3974 }
1406e432 3975
36523e95
DS		 3976 /*
3977 * Add global block reserve
3978 */
3979 if (slot_count) {
3980 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3981
3982 spin_lock(&block_rsv->lock);
3983 space.total_bytes = block_rsv->size;
3984 space.used_bytes = block_rsv->size - block_rsv->reserved;
3985 spin_unlock(&block_rsv->lock);
3986 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3987 memcpy(dest, &space, sizeof(space));
3988 space_args.total_spaces++;
3989 }
3990
2eec6c81 3991 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3992 (arg + sizeof(struct btrfs_ioctl_space_args));
3993
3994 if (copy_to_user(user_dest, dest_orig, alloc_size))
3995 ret = -EFAULT;
3996
3997 kfree(dest_orig);
3998out:
3999 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 4000 ret = -EFAULT;
4001
4002 return ret;
4003}
4004
f46b5a66
CH		 4005/*
4006 * there are many ways the trans_start and trans_end ioctls can lead
4007 * to deadlocks. They should only be used by applications that
4008 * basically own the machine, and have a very in depth understanding
4009 * of all the possible deadlocks and enospc problems.
4010 */
4011long btrfs_ioctl_trans_end(struct file *file)
4012{
496ad9aa 4013 struct inode *inode = file_inode(file);
f46b5a66
CH		 4014 struct btrfs_root *root = BTRFS_I(inode)->root;
4015 struct btrfs_trans_handle *trans;
f46b5a66 4016
f46b5a66 4017 trans = file->private_data;
1ab86aed
SW		 4018 if (!trans)
4019 return -EINVAL;
b214107e 4020 file->private_data = NULL;
9ca9ee09 4021
1ab86aed
SW		 4022 btrfs_end_transaction(trans, root);
4023
a4abeea4 4024 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 4025
2a79f17e 4026 mnt_drop_write_file(file);
1ab86aed 4027 return 0;
f46b5a66
CH		 4028}
4029
9a8c28be
MX		 4030static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4031 void __user *argp)
46204592 4032{
46204592
SW		 4033 struct btrfs_trans_handle *trans;
4034 u64 transid;
db5b493a 4035 int ret;
46204592 4036
d4edf39b 4037 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4038 if (IS_ERR(trans)) {
4039 if (PTR_ERR(trans) != -ENOENT)
4040 return PTR_ERR(trans);
4041
4042 /* No running transaction, don't bother */
4043 transid = root->fs_info->last_trans_committed;
4044 goto out;
4045 }
46204592 4046 transid = trans->transid;
db5b493a 4047 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 4048 if (ret) {
4049 btrfs_end_transaction(trans, root);
db5b493a 4050 return ret;
8b2b2d3c 4051 }
ff7c1d33 4052out:
46204592
SW		 4053 if (argp)
4054 if (copy_to_user(argp, &transid, sizeof(transid)))
4055 return -EFAULT;
4056 return 0;
4057}
4058
9a8c28be
MX		 4059static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4060 void __user *argp)
46204592 4061{
46204592
SW		 4062 u64 transid;
4063
4064 if (argp) {
4065 if (copy_from_user(&transid, argp, sizeof(transid)))
4066 return -EFAULT;
4067 } else {
4068 transid = 0; /* current trans */
4069 }
4070 return btrfs_wait_for_commit(root, transid);
4071}
4072
b8e95489 4073static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4074{
496ad9aa 4075 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 4076 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4077 int ret;
475f6387
JS		 4078
4079 if (!capable(CAP_SYS_ADMIN))
4080 return -EPERM;
4081
4082 sa = memdup_user(arg, sizeof(*sa));
4083 if (IS_ERR(sa))
4084 return PTR_ERR(sa);
4085
b8e95489
MX		 4086 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4087 ret = mnt_want_write_file(file);
4088 if (ret)
4089 goto out;
4090 }
4091
aa1b8cd4 4092 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4093 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4094 0);
475f6387
JS		 4095
4096 if (copy_to_user(arg, sa, sizeof(*sa)))
4097 ret = -EFAULT;
4098
b8e95489
MX		 4099 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4100 mnt_drop_write_file(file);
4101out:
475f6387
JS		 4102 kfree(sa);
4103 return ret;
4104}
4105
4106static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4107{
4108 if (!capable(CAP_SYS_ADMIN))
4109 return -EPERM;
4110
aa1b8cd4 4111 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 4112}
4113
4114static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4115 void __user *arg)
4116{
4117 struct btrfs_ioctl_scrub_args *sa;
4118 int ret;
4119
4120 if (!capable(CAP_SYS_ADMIN))
4121 return -EPERM;
4122
4123 sa = memdup_user(arg, sizeof(*sa));
4124 if (IS_ERR(sa))
4125 return PTR_ERR(sa);
4126
4127 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4128
4129 if (copy_to_user(arg, sa, sizeof(*sa)))
4130 ret = -EFAULT;
4131
4132 kfree(sa);
4133 return ret;
4134}
4135
c11d2c23 4136static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 4137 void __user *arg)
c11d2c23
SB		 4138{
4139 struct btrfs_ioctl_get_dev_stats *sa;
4140 int ret;
4141
c11d2c23
SB		 4142 sa = memdup_user(arg, sizeof(*sa));
4143 if (IS_ERR(sa))
4144 return PTR_ERR(sa);
4145
b27f7c0c
DS		 4146 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4147 kfree(sa);
4148 return -EPERM;
4149 }
4150
4151 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 4152
4153 if (copy_to_user(arg, sa, sizeof(*sa)))
4154 ret = -EFAULT;
4155
4156 kfree(sa);
4157 return ret;
4158}
4159
3f6bcfbd
SB		 4160static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4161{
4162 struct btrfs_ioctl_dev_replace_args *p;
4163 int ret;
4164
4165 if (!capable(CAP_SYS_ADMIN))
4166 return -EPERM;
4167
4168 p = memdup_user(arg, sizeof(*p));
4169 if (IS_ERR(p))
4170 return PTR_ERR(p);
4171
4172 switch (p->cmd) {
4173 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
adfa97cb
ID		 4174 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4175 ret = -EROFS;
4176 goto out;
4177 }
3f6bcfbd
SB		 4178 if (atomic_xchg(
4179 &root->fs_info->mutually_exclusive_operation_running,
4180 1)) {
e57138b3 4181 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd
SB		 4182 } else {
4183 ret = btrfs_dev_replace_start(root, p);
4184 atomic_set(
4185 &root->fs_info->mutually_exclusive_operation_running,
4186 0);
4187 }
4188 break;
4189 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4190 btrfs_dev_replace_status(root->fs_info, p);
4191 ret = 0;
4192 break;
4193 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4194 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4195 break;
4196 default:
4197 ret = -EINVAL;
4198 break;
4199 }
4200
4201 if (copy_to_user(arg, p, sizeof(*p)))
4202 ret = -EFAULT;
adfa97cb 4203out:
3f6bcfbd
SB		 4204 kfree(p);
4205 return ret;
4206}
4207
d7728c96
JS		 4208static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4209{
4210 int ret = 0;
4211 int i;
740c3d22 4212 u64 rel_ptr;
d7728c96 4213 int size;
806468f8 4214 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4215 struct inode_fs_paths *ipath = NULL;
4216 struct btrfs_path *path;
4217
82b22ac8 4218 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4219 return -EPERM;
4220
4221 path = btrfs_alloc_path();
4222 if (!path) {
4223 ret = -ENOMEM;
4224 goto out;
4225 }
4226
4227 ipa = memdup_user(arg, sizeof(*ipa));
4228 if (IS_ERR(ipa)) {
4229 ret = PTR_ERR(ipa);
4230 ipa = NULL;
4231 goto out;
4232 }
4233
4234 size = min_t(u32, ipa->size, 4096);
4235 ipath = init_ipath(size, root, path);
4236 if (IS_ERR(ipath)) {
4237 ret = PTR_ERR(ipath);
4238 ipath = NULL;
4239 goto out;
4240 }
4241
4242 ret = paths_from_inode(ipa->inum, ipath);
4243 if (ret < 0)
4244 goto out;
4245
4246 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4247 rel_ptr = ipath->fspath->val[i] -
4248 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4249 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4250 }
4251
745c4d8e
JM		 4252 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4253 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 4254 if (ret) {
4255 ret = -EFAULT;
4256 goto out;
4257 }
4258
4259out:
4260 btrfs_free_path(path);
4261 free_ipath(ipath);
4262 kfree(ipa);
4263
4264 return ret;
4265}
4266
4267static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4268{
4269 struct btrfs_data_container *inodes = ctx;
4270 const size_t c = 3 * sizeof(u64);
4271
4272 if (inodes->bytes_left >= c) {
4273 inodes->bytes_left -= c;
4274 inodes->val[inodes->elem_cnt] = inum;
4275 inodes->val[inodes->elem_cnt + 1] = offset;
4276 inodes->val[inodes->elem_cnt + 2] = root;
4277 inodes->elem_cnt += 3;
4278 } else {
4279 inodes->bytes_missing += c - inodes->bytes_left;
4280 inodes->bytes_left = 0;
4281 inodes->elem_missed += 3;
4282 }
4283
4284 return 0;
4285}
4286
4287static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4288 void __user *arg)
4289{
4290 int ret = 0;
4291 int size;
d7728c96
JS		 4292 struct btrfs_ioctl_logical_ino_args *loi;
4293 struct btrfs_data_container *inodes = NULL;
4294 struct btrfs_path *path = NULL;
d7728c96
JS		 4295
4296 if (!capable(CAP_SYS_ADMIN))
4297 return -EPERM;
4298
4299 loi = memdup_user(arg, sizeof(*loi));
4300 if (IS_ERR(loi)) {
4301 ret = PTR_ERR(loi);
4302 loi = NULL;
4303 goto out;
4304 }
4305
4306 path = btrfs_alloc_path();
4307 if (!path) {
4308 ret = -ENOMEM;
4309 goto out;
4310 }
4311
425d17a2 4312 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 4313 inodes = init_data_container(size);
4314 if (IS_ERR(inodes)) {
4315 ret = PTR_ERR(inodes);
4316 inodes = NULL;
4317 goto out;
4318 }
4319
df031f07
LB		 4320 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4321 build_ino_list, inodes);
4322 if (ret == -EINVAL)
d7728c96
JS		 4323 ret = -ENOENT;
4324 if (ret < 0)
4325 goto out;
4326
745c4d8e
JM		 4327 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4328 (void *)(unsigned long)inodes, size);
d7728c96
JS		 4329 if (ret)
4330 ret = -EFAULT;
4331
4332out:
4333 btrfs_free_path(path);
425d17a2 4334 vfree(inodes);
d7728c96
JS		 4335 kfree(loi);
4336
4337 return ret;
4338}
4339
19a39dce 4340void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 4341 struct btrfs_ioctl_balance_args *bargs)
4342{
4343 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4344
4345 bargs->flags = bctl->flags;
4346
837d5b6e
ID		 4347 if (atomic_read(&fs_info->balance_running))
4348 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4349 if (atomic_read(&fs_info->balance_pause_req))
4350 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4351 if (atomic_read(&fs_info->balance_cancel_req))
4352 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4353
c9e9f97b
ID		 4354 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4355 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4356 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 4357
4358 if (lock) {
4359 spin_lock(&fs_info->balance_lock);
4360 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4361 spin_unlock(&fs_info->balance_lock);
4362 } else {
4363 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4364 }
c9e9f97b
ID		 4365}
4366
9ba1f6e4 4367static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4368{
496ad9aa 4369 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4370 struct btrfs_fs_info *fs_info = root->fs_info;
4371 struct btrfs_ioctl_balance_args *bargs;
4372 struct btrfs_balance_control *bctl;
ed0fb78f 4373 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4374 int ret;
4375
4376 if (!capable(CAP_SYS_ADMIN))
4377 return -EPERM;
4378
e54bfa31 4379 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4380 if (ret)
4381 return ret;
4382
ed0fb78f
ID		 4383again:
4384 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4385 mutex_lock(&fs_info->volume_mutex);
4386 mutex_lock(&fs_info->balance_mutex);
4387 need_unlock = true;
4388 goto locked;
4389 }
4390
4391 /*
4392 * mut. excl. ops lock is locked. Three possibilites:
4393 * (1) some other op is running
4394 * (2) balance is running
4395 * (3) balance is paused -- special case (think resume)
4396 */
c9e9f97b 4397 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4398 if (fs_info->balance_ctl) {
4399 /* this is either (2) or (3) */
4400 if (!atomic_read(&fs_info->balance_running)) {
4401 mutex_unlock(&fs_info->balance_mutex);
4402 if (!mutex_trylock(&fs_info->volume_mutex))
4403 goto again;
4404 mutex_lock(&fs_info->balance_mutex);
4405
4406 if (fs_info->balance_ctl &&
4407 !atomic_read(&fs_info->balance_running)) {
4408 /* this is (3) */
4409 need_unlock = false;
4410 goto locked;
4411 }
4412
4413 mutex_unlock(&fs_info->balance_mutex);
4414 mutex_unlock(&fs_info->volume_mutex);
4415 goto again;
4416 } else {
4417 /* this is (2) */
4418 mutex_unlock(&fs_info->balance_mutex);
4419 ret = -EINPROGRESS;
4420 goto out;
4421 }
4422 } else {
4423 /* this is (1) */
4424 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4425 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4426 goto out;
4427 }
4428
4429locked:
4430 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 4431
4432 if (arg) {
4433 bargs = memdup_user(arg, sizeof(*bargs));
4434 if (IS_ERR(bargs)) {
4435 ret = PTR_ERR(bargs);
ed0fb78f 4436 goto out_unlock;
c9e9f97b 4437 }
de322263
ID		 4438
4439 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4440 if (!fs_info->balance_ctl) {
4441 ret = -ENOTCONN;
4442 goto out_bargs;
4443 }
4444
4445 bctl = fs_info->balance_ctl;
4446 spin_lock(&fs_info->balance_lock);
4447 bctl->flags |= BTRFS_BALANCE_RESUME;
4448 spin_unlock(&fs_info->balance_lock);
4449
4450 goto do_balance;
4451 }
c9e9f97b
ID		 4452 } else {
4453 bargs = NULL;
4454 }
4455
ed0fb78f 4456 if (fs_info->balance_ctl) {
837d5b6e
ID		 4457 ret = -EINPROGRESS;
4458 goto out_bargs;
4459 }
4460
c9e9f97b
ID		 4461 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4462 if (!bctl) {
4463 ret = -ENOMEM;
4464 goto out_bargs;
4465 }
4466
4467 bctl->fs_info = fs_info;
4468 if (arg) {
4469 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4470 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4471 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4472
4473 bctl->flags = bargs->flags;
f43ffb60
ID		 4474 } else {
4475 /* balance everything - no filters */
4476 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4477 }
4478
de322263 4479do_balance:
c9e9f97b 4480 /*
ed0fb78f
ID		 4481 * Ownership of bctl and mutually_exclusive_operation_running
4482 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4483 * or, if restriper was paused all the way until unmount, in
4484 * free_fs_info. mutually_exclusive_operation_running is
4485 * cleared in __cancel_balance.
c9e9f97b 4486 */
ed0fb78f
ID		 4487 need_unlock = false;
4488
4489 ret = btrfs_balance(bctl, bargs);
4490
c9e9f97b
ID		 4491 if (arg) {
4492 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4493 ret = -EFAULT;
4494 }
4495
4496out_bargs:
4497 kfree(bargs);
ed0fb78f 4498out_unlock:
c9e9f97b
ID		 4499 mutex_unlock(&fs_info->balance_mutex);
4500 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 4501 if (need_unlock)
4502 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4503out:
e54bfa31 4504 mnt_drop_write_file(file);
c9e9f97b
ID		 4505 return ret;
4506}
4507
837d5b6e
ID		 4508static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4509{
4510 if (!capable(CAP_SYS_ADMIN))
4511 return -EPERM;
4512
4513 switch (cmd) {
4514 case BTRFS_BALANCE_CTL_PAUSE:
4515 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4516 case BTRFS_BALANCE_CTL_CANCEL:
4517 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4518 }
4519
4520 return -EINVAL;
4521}
4522
19a39dce
ID		 4523static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4524 void __user *arg)
4525{
4526 struct btrfs_fs_info *fs_info = root->fs_info;
4527 struct btrfs_ioctl_balance_args *bargs;
4528 int ret = 0;
4529
4530 if (!capable(CAP_SYS_ADMIN))
4531 return -EPERM;
4532
4533 mutex_lock(&fs_info->balance_mutex);
4534 if (!fs_info->balance_ctl) {
4535 ret = -ENOTCONN;
4536 goto out;
4537 }
4538
4539 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4540 if (!bargs) {
4541 ret = -ENOMEM;
4542 goto out;
4543 }
4544
4545 update_ioctl_balance_args(fs_info, 1, bargs);
4546
4547 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4548 ret = -EFAULT;
4549
4550 kfree(bargs);
4551out:
4552 mutex_unlock(&fs_info->balance_mutex);
4553 return ret;
4554}
4555
905b0dda 4556static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4557{
496ad9aa 4558 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4559 struct btrfs_ioctl_quota_ctl_args *sa;
4560 struct btrfs_trans_handle *trans = NULL;
4561 int ret;
4562 int err;
4563
4564 if (!capable(CAP_SYS_ADMIN))
4565 return -EPERM;
4566
905b0dda
MX		 4567 ret = mnt_want_write_file(file);
4568 if (ret)
4569 return ret;
5d13a37b
AJ		 4570
4571 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4572 if (IS_ERR(sa)) {
4573 ret = PTR_ERR(sa);
4574 goto drop_write;
4575 }
5d13a37b 4576
7708f029 4577 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4578 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4579 if (IS_ERR(trans)) {
4580 ret = PTR_ERR(trans);
4581 goto out;
5d13a37b
AJ		 4582 }
4583
4584 switch (sa->cmd) {
4585 case BTRFS_QUOTA_CTL_ENABLE:
4586 ret = btrfs_quota_enable(trans, root->fs_info);
4587 break;
4588 case BTRFS_QUOTA_CTL_DISABLE:
4589 ret = btrfs_quota_disable(trans, root->fs_info);
4590 break;
5d13a37b
AJ		 4591 default:
4592 ret = -EINVAL;
4593 break;
4594 }
4595
2f232036
JS		 4596 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4597 if (err && !ret)
4598 ret = err;
5d13a37b
AJ		 4599out:
4600 kfree(sa);
7708f029 4601 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4602drop_write:
4603 mnt_drop_write_file(file);
5d13a37b
AJ		 4604 return ret;
4605}
4606
905b0dda 4607static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4608{
496ad9aa 4609 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4610 struct btrfs_ioctl_qgroup_assign_args *sa;
4611 struct btrfs_trans_handle *trans;
4612 int ret;
4613 int err;
4614
4615 if (!capable(CAP_SYS_ADMIN))
4616 return -EPERM;
4617
905b0dda
MX		 4618 ret = mnt_want_write_file(file);
4619 if (ret)
4620 return ret;
5d13a37b
AJ		 4621
4622 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4623 if (IS_ERR(sa)) {
4624 ret = PTR_ERR(sa);
4625 goto drop_write;
4626 }
5d13a37b
AJ		 4627
4628 trans = btrfs_join_transaction(root);
4629 if (IS_ERR(trans)) {
4630 ret = PTR_ERR(trans);
4631 goto out;
4632 }
4633
4634 /* FIXME: check if the IDs really exist */
4635 if (sa->assign) {
4636 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4637 sa->src, sa->dst);
4638 } else {
4639 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4640 sa->src, sa->dst);
4641 }
4642
4643 err = btrfs_end_transaction(trans, root);
4644 if (err && !ret)
4645 ret = err;
4646
4647out:
4648 kfree(sa);
905b0dda
MX		 4649drop_write:
4650 mnt_drop_write_file(file);
5d13a37b
AJ		 4651 return ret;
4652}
4653
905b0dda 4654static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4655{
496ad9aa 4656 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4657 struct btrfs_ioctl_qgroup_create_args *sa;
4658 struct btrfs_trans_handle *trans;
4659 int ret;
4660 int err;
4661
4662 if (!capable(CAP_SYS_ADMIN))
4663 return -EPERM;
4664
905b0dda
MX		 4665 ret = mnt_want_write_file(file);
4666 if (ret)
4667 return ret;
5d13a37b
AJ		 4668
4669 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4670 if (IS_ERR(sa)) {
4671 ret = PTR_ERR(sa);
4672 goto drop_write;
4673 }
5d13a37b 4674
d86e56cf
MX		 4675 if (!sa->qgroupid) {
4676 ret = -EINVAL;
4677 goto out;
4678 }
4679
5d13a37b
AJ		 4680 trans = btrfs_join_transaction(root);
4681 if (IS_ERR(trans)) {
4682 ret = PTR_ERR(trans);
4683 goto out;
4684 }
4685
4686 /* FIXME: check if the IDs really exist */
4687 if (sa->create) {
4688 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4689 NULL);
4690 } else {
4691 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4692 }
4693
4694 err = btrfs_end_transaction(trans, root);
4695 if (err && !ret)
4696 ret = err;
4697
4698out:
4699 kfree(sa);
905b0dda
MX		 4700drop_write:
4701 mnt_drop_write_file(file);
5d13a37b
AJ		 4702 return ret;
4703}
4704
905b0dda 4705static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4706{
496ad9aa 4707 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4708 struct btrfs_ioctl_qgroup_limit_args *sa;
4709 struct btrfs_trans_handle *trans;
4710 int ret;
4711 int err;
4712 u64 qgroupid;
4713
4714 if (!capable(CAP_SYS_ADMIN))
4715 return -EPERM;
4716
905b0dda
MX		 4717 ret = mnt_want_write_file(file);
4718 if (ret)
4719 return ret;
5d13a37b
AJ		 4720
4721 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4722 if (IS_ERR(sa)) {
4723 ret = PTR_ERR(sa);
4724 goto drop_write;
4725 }
5d13a37b
AJ		 4726
4727 trans = btrfs_join_transaction(root);
4728 if (IS_ERR(trans)) {
4729 ret = PTR_ERR(trans);
4730 goto out;
4731 }
4732
4733 qgroupid = sa->qgroupid;
4734 if (!qgroupid) {
4735 /* take the current subvol as qgroup */
4736 qgroupid = root->root_key.objectid;
4737 }
4738
4739 /* FIXME: check if the IDs really exist */
4740 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4741
4742 err = btrfs_end_transaction(trans, root);
4743 if (err && !ret)
4744 ret = err;
4745
4746out:
4747 kfree(sa);
905b0dda
MX		 4748drop_write:
4749 mnt_drop_write_file(file);
5d13a37b
AJ		 4750 return ret;
4751}
4752
2f232036
JS		 4753static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4754{
6d0379ec 4755 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4756 struct btrfs_ioctl_quota_rescan_args *qsa;
4757 int ret;
4758
4759 if (!capable(CAP_SYS_ADMIN))
4760 return -EPERM;
4761
4762 ret = mnt_want_write_file(file);
4763 if (ret)
4764 return ret;
4765
4766 qsa = memdup_user(arg, sizeof(*qsa));
4767 if (IS_ERR(qsa)) {
4768 ret = PTR_ERR(qsa);
4769 goto drop_write;
4770 }
4771
4772 if (qsa->flags) {
4773 ret = -EINVAL;
4774 goto out;
4775 }
4776
4777 ret = btrfs_qgroup_rescan(root->fs_info);
4778
4779out:
4780 kfree(qsa);
4781drop_write:
4782 mnt_drop_write_file(file);
4783 return ret;
4784}
4785
4786static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4787{
6d0379ec 4788 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4789 struct btrfs_ioctl_quota_rescan_args *qsa;
4790 int ret = 0;
4791
4792 if (!capable(CAP_SYS_ADMIN))
4793 return -EPERM;
4794
4795 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4796 if (!qsa)
4797 return -ENOMEM;
4798
4799 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4800 qsa->flags = 1;
4801 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4802 }
4803
4804 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4805 ret = -EFAULT;
4806
4807 kfree(qsa);
4808 return ret;
4809}
4810
57254b6e
JS		 4811static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4812{
54563d41 4813 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
57254b6e
JS		 4814
4815 if (!capable(CAP_SYS_ADMIN))
4816 return -EPERM;
4817
4818 return btrfs_qgroup_wait_for_completion(root->fs_info);
4819}
4820
abccd00f
HM		 4821static long _btrfs_ioctl_set_received_subvol(struct file *file,
4822 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4823{
496ad9aa 4824 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4825 struct btrfs_root *root = BTRFS_I(inode)->root;
4826 struct btrfs_root_item *root_item = &root->root_item;
4827 struct btrfs_trans_handle *trans;
4828 struct timespec ct = CURRENT_TIME;
4829 int ret = 0;
dd5f9615 4830 int received_uuid_changed;
8ea05e3a 4831
bd60ea0f
DS		 4832 if (!inode_owner_or_capable(inode))
4833 return -EPERM;
4834
8ea05e3a
AB		 4835 ret = mnt_want_write_file(file);
4836 if (ret < 0)
4837 return ret;
4838
4839 down_write(&root->fs_info->subvol_sem);
4840
4841 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4842 ret = -EINVAL;
4843 goto out;
4844 }
4845
4846 if (btrfs_root_readonly(root)) {
4847 ret = -EROFS;
4848 goto out;
4849 }
4850
dd5f9615
SB		 4851 /*
4852 * 1 - root item
4853 * 2 - uuid items (received uuid + subvol uuid)
4854 */
4855 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4856 if (IS_ERR(trans)) {
4857 ret = PTR_ERR(trans);
4858 trans = NULL;
4859 goto out;
4860 }
4861
4862 sa->rtransid = trans->transid;
4863 sa->rtime.sec = ct.tv_sec;
4864 sa->rtime.nsec = ct.tv_nsec;
4865
dd5f9615
SB		 4866 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4867 BTRFS_UUID_SIZE);
4868 if (received_uuid_changed &&
4869 !btrfs_is_empty_uuid(root_item->received_uuid))
4870 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4871 root_item->received_uuid,
4872 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4873 root->root_key.objectid);
8ea05e3a
AB		 4874 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4875 btrfs_set_root_stransid(root_item, sa->stransid);
4876 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4877 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4878 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4879 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4880 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4881
4882 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4883 &root->root_key, &root->root_item);
4884 if (ret < 0) {
4885 btrfs_end_transaction(trans, root);
8ea05e3a 4886 goto out;
dd5f9615
SB		 4887 }
4888 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4889 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4890 sa->uuid,
4891 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4892 root->root_key.objectid);
4893 if (ret < 0 && ret != -EEXIST) {
4894 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 4895 goto out;
dd5f9615
SB		 4896 }
4897 }
4898 ret = btrfs_commit_transaction(trans, root);
4899 if (ret < 0) {
4900 btrfs_abort_transaction(trans, root, ret);
4901 goto out;
8ea05e3a
AB		 4902 }
4903
abccd00f
HM		 4904out:
4905 up_write(&root->fs_info->subvol_sem);
4906 mnt_drop_write_file(file);
4907 return ret;
4908}
4909
4910#ifdef CONFIG_64BIT
4911static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4912 void __user *arg)
4913{
4914 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4915 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4916 int ret = 0;
4917
4918 args32 = memdup_user(arg, sizeof(*args32));
4919 if (IS_ERR(args32)) {
4920 ret = PTR_ERR(args32);
4921 args32 = NULL;
4922 goto out;
4923 }
4924
4925 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
84dbeb87
DC		 4926 if (!args64) {
4927 ret = -ENOMEM;
abccd00f
HM		 4928 goto out;
4929 }
4930
4931 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4932 args64->stransid = args32->stransid;
4933 args64->rtransid = args32->rtransid;
4934 args64->stime.sec = args32->stime.sec;
4935 args64->stime.nsec = args32->stime.nsec;
4936 args64->rtime.sec = args32->rtime.sec;
4937 args64->rtime.nsec = args32->rtime.nsec;
4938 args64->flags = args32->flags;
4939
4940 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4941 if (ret)
4942 goto out;
4943
4944 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4945 args32->stransid = args64->stransid;
4946 args32->rtransid = args64->rtransid;
4947 args32->stime.sec = args64->stime.sec;
4948 args32->stime.nsec = args64->stime.nsec;
4949 args32->rtime.sec = args64->rtime.sec;
4950 args32->rtime.nsec = args64->rtime.nsec;
4951 args32->flags = args64->flags;
4952
4953 ret = copy_to_user(arg, args32, sizeof(*args32));
4954 if (ret)
4955 ret = -EFAULT;
4956
4957out:
4958 kfree(args32);
4959 kfree(args64);
4960 return ret;
4961}
4962#endif
4963
4964static long btrfs_ioctl_set_received_subvol(struct file *file,
4965 void __user *arg)
4966{
4967 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4968 int ret = 0;
4969
4970 sa = memdup_user(arg, sizeof(*sa));
4971 if (IS_ERR(sa)) {
4972 ret = PTR_ERR(sa);
4973 sa = NULL;
4974 goto out;
4975 }
4976
4977 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4978
4979 if (ret)
4980 goto out;
4981
8ea05e3a
AB		 4982 ret = copy_to_user(arg, sa, sizeof(*sa));
4983 if (ret)
4984 ret = -EFAULT;
4985
4986out:
4987 kfree(sa);
8ea05e3a
AB		 4988 return ret;
4989}
4990
867ab667 4991static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4992{
6d0379ec 4993 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 4994 size_t len;
867ab667 4995 int ret;
a1b83ac5
AJ		 4996 char label[BTRFS_LABEL_SIZE];
4997
4998 spin_lock(&root->fs_info->super_lock);
4999 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5000 spin_unlock(&root->fs_info->super_lock);
5001
5002 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5003
5004 if (len == BTRFS_LABEL_SIZE) {
efe120a0
FH		 5005 btrfs_warn(root->fs_info,
5006 "label is too long, return the first %zu bytes", --len);
867ab667 5007 }
5008
867ab667 5009 ret = copy_to_user(arg, label, len);
867ab667 5010
5011 return ret ? -EFAULT : 0;
5012}
5013
a8bfd4ab 5014static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5015{
6d0379ec 5016 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 5017 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5018 struct btrfs_trans_handle *trans;
5019 char label[BTRFS_LABEL_SIZE];
5020 int ret;
5021
5022 if (!capable(CAP_SYS_ADMIN))
5023 return -EPERM;
5024
5025 if (copy_from_user(label, arg, sizeof(label)))
5026 return -EFAULT;
5027
5028 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
efe120a0 5029 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
a8bfd4ab 5030 BTRFS_LABEL_SIZE - 1);
5031 return -EINVAL;
5032 }
5033
5034 ret = mnt_want_write_file(file);
5035 if (ret)
5036 return ret;
5037
a8bfd4ab 5038 trans = btrfs_start_transaction(root, 0);
5039 if (IS_ERR(trans)) {
5040 ret = PTR_ERR(trans);
5041 goto out_unlock;
5042 }
5043
a1b83ac5 5044 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 5045 strcpy(super_block->label, label);
a1b83ac5 5046 spin_unlock(&root->fs_info->super_lock);
d0270aca 5047 ret = btrfs_commit_transaction(trans, root);
a8bfd4ab 5048
5049out_unlock:
a8bfd4ab 5050 mnt_drop_write_file(file);
5051 return ret;
5052}
5053
2eaa055f
JM		 5054#define INIT_FEATURE_FLAGS(suffix) \
5055 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5056 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5057 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5058
5059static int btrfs_ioctl_get_supported_features(struct file *file,
5060 void __user *arg)
5061{
5062 static struct btrfs_ioctl_feature_flags features[3] = {
5063 INIT_FEATURE_FLAGS(SUPP),
5064 INIT_FEATURE_FLAGS(SAFE_SET),
5065 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5066 };
5067
5068 if (copy_to_user(arg, &features, sizeof(features)))
5069 return -EFAULT;
5070
5071 return 0;
5072}
5073
5074static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5075{
5076 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5077 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5078 struct btrfs_ioctl_feature_flags features;
5079
5080 features.compat_flags = btrfs_super_compat_flags(super_block);
5081 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5082 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5083
5084 if (copy_to_user(arg, &features, sizeof(features)))
5085 return -EFAULT;
5086
5087 return 0;
5088}
5089
3b02a68a
JM		 5090static int check_feature_bits(struct btrfs_root *root,
5091 enum btrfs_feature_set set,
2eaa055f
JM		 5092 u64 change_mask, u64 flags, u64 supported_flags,
5093 u64 safe_set, u64 safe_clear)
5094{
3b02a68a
JM		 5095 const char *type = btrfs_feature_set_names[set];
5096 char *names;
2eaa055f
JM		 5097 u64 disallowed, unsupported;
5098 u64 set_mask = flags & change_mask;
5099 u64 clear_mask = ~flags & change_mask;
5100
5101 unsupported = set_mask & ~supported_flags;
5102 if (unsupported) {
3b02a68a
JM		 5103 names = btrfs_printable_features(set, unsupported);
5104 if (names) {
5105 btrfs_warn(root->fs_info,
5106 "this kernel does not support the %s feature bit%s",
5107 names, strchr(names, ',') ? "s" : "");
5108 kfree(names);
5109 } else
5110 btrfs_warn(root->fs_info,
2eaa055f
JM		 5111 "this kernel does not support %s bits 0x%llx",
5112 type, unsupported);
5113 return -EOPNOTSUPP;
5114 }
5115
5116 disallowed = set_mask & ~safe_set;
5117 if (disallowed) {
3b02a68a
JM		 5118 names = btrfs_printable_features(set, disallowed);
5119 if (names) {
5120 btrfs_warn(root->fs_info,
5121 "can't set the %s feature bit%s while mounted",
5122 names, strchr(names, ',') ? "s" : "");
5123 kfree(names);
5124 } else
5125 btrfs_warn(root->fs_info,
2eaa055f
JM		 5126 "can't set %s bits 0x%llx while mounted",
5127 type, disallowed);
5128 return -EPERM;
5129 }
5130
5131 disallowed = clear_mask & ~safe_clear;
5132 if (disallowed) {
3b02a68a
JM		 5133 names = btrfs_printable_features(set, disallowed);
5134 if (names) {
5135 btrfs_warn(root->fs_info,
5136 "can't clear the %s feature bit%s while mounted",
5137 names, strchr(names, ',') ? "s" : "");
5138 kfree(names);
5139 } else
5140 btrfs_warn(root->fs_info,
2eaa055f
JM		 5141 "can't clear %s bits 0x%llx while mounted",
5142 type, disallowed);
5143 return -EPERM;
5144 }
5145
5146 return 0;
5147}
5148
5149#define check_feature(root, change_mask, flags, mask_base) \
3b02a68a 5150check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5151 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5152 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5153 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5154
5155static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5156{
5157 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5158 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5159 struct btrfs_ioctl_feature_flags flags[2];
5160 struct btrfs_trans_handle *trans;
5161 u64 newflags;
5162 int ret;
5163
5164 if (!capable(CAP_SYS_ADMIN))
5165 return -EPERM;
5166
5167 if (copy_from_user(flags, arg, sizeof(flags)))
5168 return -EFAULT;
5169
5170 /* Nothing to do */
5171 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5172 !flags[0].incompat_flags)
5173 return 0;
5174
5175 ret = check_feature(root, flags[0].compat_flags,
5176 flags[1].compat_flags, COMPAT);
5177 if (ret)
5178 return ret;
5179
5180 ret = check_feature(root, flags[0].compat_ro_flags,
5181 flags[1].compat_ro_flags, COMPAT_RO);
5182 if (ret)
5183 return ret;
5184
5185 ret = check_feature(root, flags[0].incompat_flags,
5186 flags[1].incompat_flags, INCOMPAT);
5187 if (ret)
5188 return ret;
5189
8051aa1a 5190 trans = btrfs_start_transaction(root, 0);
2eaa055f
JM		 5191 if (IS_ERR(trans))
5192 return PTR_ERR(trans);
5193
5194 spin_lock(&root->fs_info->super_lock);
5195 newflags = btrfs_super_compat_flags(super_block);
5196 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5197 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5198 btrfs_set_super_compat_flags(super_block, newflags);
5199
5200 newflags = btrfs_super_compat_ro_flags(super_block);
5201 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5202 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5203 btrfs_set_super_compat_ro_flags(super_block, newflags);
5204
5205 newflags = btrfs_super_incompat_flags(super_block);
5206 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5207 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5208 btrfs_set_super_incompat_flags(super_block, newflags);
5209 spin_unlock(&root->fs_info->super_lock);
5210
d0270aca 5211 return btrfs_commit_transaction(trans, root);
2eaa055f
JM		 5212}
5213
f46b5a66
CH		 5214long btrfs_ioctl(struct file *file, unsigned int
5215 cmd, unsigned long arg)
5216{
496ad9aa 5217 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 5218 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5219
5220 switch (cmd) {
6cbff00f
CH		 5221 case FS_IOC_GETFLAGS:
5222 return btrfs_ioctl_getflags(file, argp);
5223 case FS_IOC_SETFLAGS:
5224 return btrfs_ioctl_setflags(file, argp);
5225 case FS_IOC_GETVERSION:
5226 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5227 case FITRIM:
5228 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5229 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5230 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5231 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5232 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5233 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5234 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5235 case BTRFS_IOC_SUBVOL_CREATE_V2:
5236 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5237 case BTRFS_IOC_SNAP_DESTROY:
5238 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5239 case BTRFS_IOC_SUBVOL_GETFLAGS:
5240 return btrfs_ioctl_subvol_getflags(file, argp);
5241 case BTRFS_IOC_SUBVOL_SETFLAGS:
5242 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5243 case BTRFS_IOC_DEFAULT_SUBVOL:
5244 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5245 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5246 return btrfs_ioctl_defrag(file, NULL);
5247 case BTRFS_IOC_DEFRAG_RANGE:
5248 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5249 case BTRFS_IOC_RESIZE:
198605a8 5250 return btrfs_ioctl_resize(file, argp);
f46b5a66 5251 case BTRFS_IOC_ADD_DEV:
4bcabaa3 5252 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 5253 case BTRFS_IOC_RM_DEV:
da24927b 5254 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 5255 case BTRFS_IOC_FS_INFO:
5256 return btrfs_ioctl_fs_info(root, argp);
5257 case BTRFS_IOC_DEV_INFO:
5258 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 5259 case BTRFS_IOC_BALANCE:
9ba1f6e4 5260 return btrfs_ioctl_balance(file, NULL);
f46b5a66 5261 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 5262 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5263 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 5264 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 5265 case BTRFS_IOC_TRANS_START:
5266 return btrfs_ioctl_trans_start(file);
5267 case BTRFS_IOC_TRANS_END:
5268 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 5269 case BTRFS_IOC_TREE_SEARCH:
5270 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5271 case BTRFS_IOC_TREE_SEARCH_V2:
5272 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5273 case BTRFS_IOC_INO_LOOKUP:
5274 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5275 case BTRFS_IOC_INO_PATHS:
5276 return btrfs_ioctl_ino_to_path(root, argp);
5277 case BTRFS_IOC_LOGICAL_INO:
5278 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 5279 case BTRFS_IOC_SPACE_INFO:
5280 return btrfs_ioctl_space_info(root, argp);
9b199859
FDBM		 5281 case BTRFS_IOC_SYNC: {
5282 int ret;
5283
6c255e67 5284 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
9b199859
FDBM		 5285 if (ret)
5286 return ret;
5287 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
2fad4e83
DS		 5288 /*
5289 * The transaction thread may want to do more work,
5290 * namely it pokes the cleaner ktread that will start
5291 * processing uncleaned subvols.
5292 */
5293 wake_up_process(root->fs_info->transaction_kthread);
9b199859
FDBM		 5294 return ret;
5295 }
46204592 5296 case BTRFS_IOC_START_SYNC:
9a8c28be 5297 return btrfs_ioctl_start_sync(root, argp);
46204592 5298 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 5299 return btrfs_ioctl_wait_sync(root, argp);
475f6387 5300 case BTRFS_IOC_SCRUB:
b8e95489 5301 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 5302 case BTRFS_IOC_SCRUB_CANCEL:
5303 return btrfs_ioctl_scrub_cancel(root, argp);
5304 case BTRFS_IOC_SCRUB_PROGRESS:
5305 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 5306 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5307 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 5308 case BTRFS_IOC_BALANCE_CTL:
5309 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 5310 case BTRFS_IOC_BALANCE_PROGRESS:
5311 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 5312 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5313 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5314#ifdef CONFIG_64BIT
5315 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5316 return btrfs_ioctl_set_received_subvol_32(file, argp);
5317#endif
31db9f7c
AB		 5318 case BTRFS_IOC_SEND:
5319 return btrfs_ioctl_send(file, argp);
c11d2c23 5320 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 5321 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 5322 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5323 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5324 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5325 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5326 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5327 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5328 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5329 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5330 case BTRFS_IOC_QUOTA_RESCAN:
5331 return btrfs_ioctl_quota_rescan(file, argp);
5332 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5333 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5334 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5335 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 5336 case BTRFS_IOC_DEV_REPLACE:
5337 return btrfs_ioctl_dev_replace(root, argp);
867ab667 5338 case BTRFS_IOC_GET_FSLABEL:
5339 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5340 case BTRFS_IOC_SET_FSLABEL:
5341 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 5342 case BTRFS_IOC_FILE_EXTENT_SAME:
5343 return btrfs_ioctl_file_extent_same(file, argp);
2eaa055f
JM		 5344 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5345 return btrfs_ioctl_get_supported_features(file, argp);
5346 case BTRFS_IOC_GET_FEATURES:
5347 return btrfs_ioctl_get_features(file, argp);
5348 case BTRFS_IOC_SET_FEATURES:
5349 return btrfs_ioctl_set_features(file, argp);
f46b5a66
CH		 5350 }
5351
5352 return -ENOTTY;
5353}
Linux 6.x block layer and io_uring tree(s)