projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: introduce btrfs_{start, end}_nocow_write() for each subvolume
[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"
f46b5a66 61
abccd00f
HM		 62#ifdef CONFIG_64BIT
63/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
64 * structures are incorrect, as the timespec structure from userspace
65 * is 4 bytes too small. We define these alternatives here to teach
66 * the kernel about the 32-bit struct packing.
67 */
68struct btrfs_ioctl_timespec_32 {
69 __u64 sec;
70 __u32 nsec;
71} __attribute__ ((__packed__));
72
73struct btrfs_ioctl_received_subvol_args_32 {
74 char uuid[BTRFS_UUID_SIZE]; /* in */
75 __u64 stransid; /* in */
76 __u64 rtransid; /* out */
77 struct btrfs_ioctl_timespec_32 stime; /* in */
78 struct btrfs_ioctl_timespec_32 rtime; /* out */
79 __u64 flags; /* in */
80 __u64 reserved[16]; /* in */
81} __attribute__ ((__packed__));
82
83#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
84 struct btrfs_ioctl_received_subvol_args_32)
85#endif
86
87
416161db
MF		 88static int btrfs_clone(struct inode *src, struct inode *inode,
89 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
90
6cbff00f
CH		 91/* Mask out flags that are inappropriate for the given type of inode. */
92static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
93{
94 if (S_ISDIR(mode))
95 return flags;
96 else if (S_ISREG(mode))
97 return flags & ~FS_DIRSYNC_FL;
98 else
99 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
100}
101
102/*
103 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
104 */
105static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
106{
107 unsigned int iflags = 0;
108
109 if (flags & BTRFS_INODE_SYNC)
110 iflags |= FS_SYNC_FL;
111 if (flags & BTRFS_INODE_IMMUTABLE)
112 iflags |= FS_IMMUTABLE_FL;
113 if (flags & BTRFS_INODE_APPEND)
114 iflags |= FS_APPEND_FL;
115 if (flags & BTRFS_INODE_NODUMP)
116 iflags |= FS_NODUMP_FL;
117 if (flags & BTRFS_INODE_NOATIME)
118 iflags |= FS_NOATIME_FL;
119 if (flags & BTRFS_INODE_DIRSYNC)
120 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 121 if (flags & BTRFS_INODE_NODATACOW)
122 iflags |= FS_NOCOW_FL;
123
124 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
125 iflags |= FS_COMPR_FL;
126 else if (flags & BTRFS_INODE_NOCOMPRESS)
127 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 128
129 return iflags;
130}
131
132/*
133 * Update inode->i_flags based on the btrfs internal flags.
134 */
135void btrfs_update_iflags(struct inode *inode)
136{
137 struct btrfs_inode *ip = BTRFS_I(inode);
138
139 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
140
141 if (ip->flags & BTRFS_INODE_SYNC)
142 inode->i_flags |= S_SYNC;
143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 inode->i_flags |= S_IMMUTABLE;
145 if (ip->flags & BTRFS_INODE_APPEND)
146 inode->i_flags |= S_APPEND;
147 if (ip->flags & BTRFS_INODE_NOATIME)
148 inode->i_flags |= S_NOATIME;
149 if (ip->flags & BTRFS_INODE_DIRSYNC)
150 inode->i_flags |= S_DIRSYNC;
151}
152
153/*
154 * Inherit flags from the parent inode.
155 *
e27425d6 156 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 157 */
158void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
159{
0b4dcea5
CM		 160 unsigned int flags;
161
162 if (!dir)
163 return;
164
165 flags = BTRFS_I(dir)->flags;
6cbff00f 166
e27425d6
JB		 167 if (flags & BTRFS_INODE_NOCOMPRESS) {
168 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
169 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
170 } else if (flags & BTRFS_INODE_COMPRESS) {
171 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
172 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
173 }
174
213490b3 175 if (flags & BTRFS_INODE_NODATACOW) {
e27425d6 176 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
213490b3
LB		 177 if (S_ISREG(inode->i_mode))
178 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
179 }
6cbff00f 180
6cbff00f
CH		 181 btrfs_update_iflags(inode);
182}
183
184static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
185{
496ad9aa 186 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
6cbff00f
CH		 187 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
188
189 if (copy_to_user(arg, &flags, sizeof(flags)))
190 return -EFAULT;
191 return 0;
192}
193
75e7cb7f
LB		 194static int check_flags(unsigned int flags)
195{
196 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
197 FS_NOATIME_FL | FS_NODUMP_FL | \
198 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 199 FS_NOCOMP_FL | FS_COMPR_FL |
200 FS_NOCOW_FL))
75e7cb7f
LB		 201 return -EOPNOTSUPP;
202
203 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
204 return -EINVAL;
205
75e7cb7f
LB		 206 return 0;
207}
208
6cbff00f
CH		 209static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
210{
496ad9aa 211 struct inode *inode = file_inode(file);
6cbff00f
CH		 212 struct btrfs_inode *ip = BTRFS_I(inode);
213 struct btrfs_root *root = ip->root;
214 struct btrfs_trans_handle *trans;
215 unsigned int flags, oldflags;
216 int ret;
f062abf0
LZ		 217 u64 ip_oldflags;
218 unsigned int i_oldflags;
7e97b8da 219 umode_t mode;
6cbff00f 220
bd60ea0f
DS		 221 if (!inode_owner_or_capable(inode))
222 return -EPERM;
223
b83cc969
LZ		 224 if (btrfs_root_readonly(root))
225 return -EROFS;
226
6cbff00f
CH		 227 if (copy_from_user(&flags, arg, sizeof(flags)))
228 return -EFAULT;
229
75e7cb7f
LB		 230 ret = check_flags(flags);
231 if (ret)
232 return ret;
f46b5a66 233
e7848683
JK		 234 ret = mnt_want_write_file(file);
235 if (ret)
236 return ret;
237
6cbff00f
CH		 238 mutex_lock(&inode->i_mutex);
239
f062abf0
LZ		 240 ip_oldflags = ip->flags;
241 i_oldflags = inode->i_flags;
7e97b8da 242 mode = inode->i_mode;
f062abf0 243
6cbff00f
CH		 244 flags = btrfs_mask_flags(inode->i_mode, flags);
245 oldflags = btrfs_flags_to_ioctl(ip->flags);
246 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
247 if (!capable(CAP_LINUX_IMMUTABLE)) {
248 ret = -EPERM;
249 goto out_unlock;
250 }
251 }
252
6cbff00f
CH		 253 if (flags & FS_SYNC_FL)
254 ip->flags |= BTRFS_INODE_SYNC;
255 else
256 ip->flags &= ~BTRFS_INODE_SYNC;
257 if (flags & FS_IMMUTABLE_FL)
258 ip->flags |= BTRFS_INODE_IMMUTABLE;
259 else
260 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
261 if (flags & FS_APPEND_FL)
262 ip->flags |= BTRFS_INODE_APPEND;
263 else
264 ip->flags &= ~BTRFS_INODE_APPEND;
265 if (flags & FS_NODUMP_FL)
266 ip->flags |= BTRFS_INODE_NODUMP;
267 else
268 ip->flags &= ~BTRFS_INODE_NODUMP;
269 if (flags & FS_NOATIME_FL)
270 ip->flags |= BTRFS_INODE_NOATIME;
271 else
272 ip->flags &= ~BTRFS_INODE_NOATIME;
273 if (flags & FS_DIRSYNC_FL)
274 ip->flags |= BTRFS_INODE_DIRSYNC;
275 else
276 ip->flags &= ~BTRFS_INODE_DIRSYNC;
7e97b8da
DS		 277 if (flags & FS_NOCOW_FL) {
278 if (S_ISREG(mode)) {
279 /*
280 * It's safe to turn csums off here, no extents exist.
281 * Otherwise we want the flag to reflect the real COW
282 * status of the file and will not set it.
283 */
284 if (inode->i_size == 0)
285 ip->flags |= BTRFS_INODE_NODATACOW
286 | BTRFS_INODE_NODATASUM;
287 } else {
288 ip->flags |= BTRFS_INODE_NODATACOW;
289 }
290 } else {
291 /*
292 * Revert back under same assuptions as above
293 */
294 if (S_ISREG(mode)) {
295 if (inode->i_size == 0)
296 ip->flags &= ~(BTRFS_INODE_NODATACOW
297 | BTRFS_INODE_NODATASUM);
298 } else {
299 ip->flags &= ~BTRFS_INODE_NODATACOW;
300 }
301 }
6cbff00f 302
75e7cb7f
LB		 303 /*
304 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
305 * flag may be changed automatically if compression code won't make
306 * things smaller.
307 */
308 if (flags & FS_NOCOMP_FL) {
309 ip->flags &= ~BTRFS_INODE_COMPRESS;
310 ip->flags |= BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 311
312 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
313 if (ret && ret != -ENODATA)
314 goto out_drop;
75e7cb7f 315 } else if (flags & FS_COMPR_FL) {
63541927
FDBM		 316 const char *comp;
317
75e7cb7f
LB		 318 ip->flags |= BTRFS_INODE_COMPRESS;
319 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 320
321 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
322 comp = "lzo";
323 else
324 comp = "zlib";
325 ret = btrfs_set_prop(inode, "btrfs.compression",
326 comp, strlen(comp), 0);
327 if (ret)
328 goto out_drop;
329
ebcb904d
LZ		 330 } else {
331 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 332 }
6cbff00f 333
4da6f1a3 334 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 335 if (IS_ERR(trans)) {
336 ret = PTR_ERR(trans);
337 goto out_drop;
338 }
6cbff00f 339
306424cc 340 btrfs_update_iflags(inode);
0c4d2d95 341 inode_inc_iversion(inode);
306424cc 342 inode->i_ctime = CURRENT_TIME;
6cbff00f 343 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 344
6cbff00f 345 btrfs_end_transaction(trans, root);
f062abf0
LZ		 346 out_drop:
347 if (ret) {
348 ip->flags = ip_oldflags;
349 inode->i_flags = i_oldflags;
350 }
6cbff00f 351
6cbff00f
CH		 352 out_unlock:
353 mutex_unlock(&inode->i_mutex);
e7848683 354 mnt_drop_write_file(file);
2d4e6f6a 355 return ret;
6cbff00f
CH		 356}
357
358static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
359{
496ad9aa 360 struct inode *inode = file_inode(file);
6cbff00f
CH		 361
362 return put_user(inode->i_generation, arg);
363}
f46b5a66 364
f7039b1d
LD		 365static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
366{
54563d41 367 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
f7039b1d
LD		 368 struct btrfs_device *device;
369 struct request_queue *q;
370 struct fstrim_range range;
371 u64 minlen = ULLONG_MAX;
372 u64 num_devices = 0;
815745cf 373 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 374 int ret;
375
376 if (!capable(CAP_SYS_ADMIN))
377 return -EPERM;
378
1f78160c
XG		 379 rcu_read_lock();
380 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
381 dev_list) {
f7039b1d
LD		 382 if (!device->bdev)
383 continue;
384 q = bdev_get_queue(device->bdev);
385 if (blk_queue_discard(q)) {
386 num_devices++;
387 minlen = min((u64)q->limits.discard_granularity,
388 minlen);
389 }
390 }
1f78160c 391 rcu_read_unlock();
f4c697e6 392
f7039b1d
LD		 393 if (!num_devices)
394 return -EOPNOTSUPP;
f7039b1d
LD		 395 if (copy_from_user(&range, arg, sizeof(range)))
396 return -EFAULT;
e515c18b
LC		 397 if (range.start > total_bytes ||
398 range.len < fs_info->sb->s_blocksize)
f4c697e6 399 return -EINVAL;
f7039b1d 400
f4c697e6 401 range.len = min(range.len, total_bytes - range.start);
f7039b1d 402 range.minlen = max(range.minlen, minlen);
815745cf 403 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 404 if (ret < 0)
405 return ret;
406
407 if (copy_to_user(arg, &range, sizeof(range)))
408 return -EFAULT;
409
410 return 0;
411}
412
dd5f9615
SB		 413int btrfs_is_empty_uuid(u8 *uuid)
414{
46e0f66a
CM		 415 int i;
416
417 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
418 if (uuid[i])
419 return 0;
420 }
421 return 1;
dd5f9615
SB		 422}
423
d5c12070 424static noinline int create_subvol(struct inode *dir,
cb8e7090 425 struct dentry *dentry,
72fd032e 426 char *name, int namelen,
6f72c7e2 427 u64 *async_transid,
8696c533 428 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 429{
430 struct btrfs_trans_handle *trans;
431 struct btrfs_key key;
432 struct btrfs_root_item root_item;
433 struct btrfs_inode_item *inode_item;
434 struct extent_buffer *leaf;
d5c12070 435 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 436 struct btrfs_root *new_root;
d5c12070 437 struct btrfs_block_rsv block_rsv;
8ea05e3a 438 struct timespec cur_time = CURRENT_TIME;
5662344b 439 struct inode *inode;
f46b5a66
CH		 440 int ret;
441 int err;
442 u64 objectid;
443 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 444 u64 index = 0;
d5c12070 445 u64 qgroup_reserved;
8ea05e3a 446 uuid_le new_uuid;
f46b5a66 447
581bb050 448 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 449 if (ret)
a22285a6 450 return ret;
6a912213 451
d5c12070 452 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 453 /*
d5c12070
MX		 454 * The same as the snapshot creation, please see the comment
455 * of create_snapshot().
9ed74f2d 456 */
d5c12070 457 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 458 8, &qgroup_reserved, false);
d5c12070
MX		 459 if (ret)
460 return ret;
461
462 trans = btrfs_start_transaction(root, 0);
463 if (IS_ERR(trans)) {
464 ret = PTR_ERR(trans);
de6e8200
LB		 465 btrfs_subvolume_release_metadata(root, &block_rsv,
466 qgroup_reserved);
467 return ret;
d5c12070
MX		 468 }
469 trans->block_rsv = &block_rsv;
470 trans->bytes_reserved = block_rsv.size;
f46b5a66 471
8696c533 472 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 473 if (ret)
474 goto fail;
475
5d4f98a2 476 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
5581a51a 477 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 478 if (IS_ERR(leaf)) {
479 ret = PTR_ERR(leaf);
480 goto fail;
481 }
f46b5a66 482
5d4f98a2 483 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 484 btrfs_set_header_bytenr(leaf, leaf->start);
485 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 486 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 487 btrfs_set_header_owner(leaf, objectid);
488
0a4e5586 489 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
f46b5a66 490 BTRFS_FSID_SIZE);
5d4f98a2 491 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
b308bc2f 492 btrfs_header_chunk_tree_uuid(leaf),
5d4f98a2 493 BTRFS_UUID_SIZE);
f46b5a66
CH		 494 btrfs_mark_buffer_dirty(leaf);
495
8ea05e3a
AB		 496 memset(&root_item, 0, sizeof(root_item));
497
f46b5a66 498 inode_item = &root_item.inode;
3cae210f
QW		 499 btrfs_set_stack_inode_generation(inode_item, 1);
500 btrfs_set_stack_inode_size(inode_item, 3);
501 btrfs_set_stack_inode_nlink(inode_item, 1);
502 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
503 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 504
3cae210f
QW		 505 btrfs_set_root_flags(&root_item, 0);
506 btrfs_set_root_limit(&root_item, 0);
507 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 508
f46b5a66 509 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 510 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 511 btrfs_set_root_level(&root_item, 0);
512 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 513 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 514 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 515
8ea05e3a
AB		 516 btrfs_set_root_generation_v2(&root_item,
517 btrfs_root_generation(&root_item));
518 uuid_le_gen(&new_uuid);
519 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 520 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
521 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 522 root_item.ctime = root_item.otime;
523 btrfs_set_root_ctransid(&root_item, trans->transid);
524 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 525
925baedd 526 btrfs_tree_unlock(leaf);
f46b5a66
CH		 527 free_extent_buffer(leaf);
528 leaf = NULL;
529
530 btrfs_set_root_dirid(&root_item, new_dirid);
531
532 key.objectid = objectid;
5d4f98a2 533 key.offset = 0;
f46b5a66
CH		 534 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
535 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
536 &root_item);
537 if (ret)
538 goto fail;
539
76dda93c
YZ		 540 key.offset = (u64)-1;
541 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 542 if (IS_ERR(new_root)) {
543 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
544 ret = PTR_ERR(new_root);
545 goto fail;
546 }
76dda93c
YZ		 547
548 btrfs_record_root_in_trans(trans, new_root);
549
63541927 550 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 551 if (ret) {
552 /* We potentially lose an unused inode item here */
79787eaa 553 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 554 goto fail;
555 }
556
f46b5a66
CH		 557 /*
558 * insert the directory item
559 */
3de4586c 560 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 561 if (ret) {
562 btrfs_abort_transaction(trans, root, ret);
563 goto fail;
564 }
3de4586c
CM		 565
566 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 567 name, namelen, dir, &key,
3de4586c 568 BTRFS_FT_DIR, index);
79787eaa
JM		 569 if (ret) {
570 btrfs_abort_transaction(trans, root, ret);
f46b5a66 571 goto fail;
79787eaa 572 }
0660b5af 573
52c26179
YZ		 574 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
575 ret = btrfs_update_inode(trans, root, dir);
576 BUG_ON(ret);
577
0660b5af 578 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 579 objectid, root->root_key.objectid,
33345d01 580 btrfs_ino(dir), index, name, namelen);
76dda93c 581 BUG_ON(ret);
f46b5a66 582
dd5f9615
SB		 583 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
584 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
585 objectid);
586 if (ret)
587 btrfs_abort_transaction(trans, root, ret);
588
f46b5a66 589fail:
d5c12070
MX		 590 trans->block_rsv = NULL;
591 trans->bytes_reserved = 0;
de6e8200
LB		 592 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
593
72fd032e
SW		 594 if (async_transid) {
595 *async_transid = trans->transid;
596 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 597 if (err)
598 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 599 } else {
600 err = btrfs_commit_transaction(trans, root);
601 }
f46b5a66
CH		 602 if (err && !ret)
603 ret = err;
1a65e24b 604
5662344b
TI		 605 if (!ret) {
606 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 607 if (IS_ERR(inode))
608 return PTR_ERR(inode);
5662344b
TI		 609 d_instantiate(dentry, inode);
610 }
f46b5a66
CH		 611 return ret;
612}
613
8257b2dc
MX		 614static void btrfs_wait_nocow_write(struct btrfs_root *root)
615{
616 s64 writers;
617 DEFINE_WAIT(wait);
618
619 do {
620 prepare_to_wait(&root->subv_writers->wait, &wait,
621 TASK_UNINTERRUPTIBLE);
622
623 writers = percpu_counter_sum(&root->subv_writers->counter);
624 if (writers)
625 schedule();
626
627 finish_wait(&root->subv_writers->wait, &wait);
628 } while (writers);
629}
630
e9662f70
MX		 631static int create_snapshot(struct btrfs_root *root, struct inode *dir,
632 struct dentry *dentry, char *name, int namelen,
633 u64 *async_transid, bool readonly,
634 struct btrfs_qgroup_inherit *inherit)
f46b5a66 635{
2e4bfab9 636 struct inode *inode;
f46b5a66
CH		 637 struct btrfs_pending_snapshot *pending_snapshot;
638 struct btrfs_trans_handle *trans;
2e4bfab9 639 int ret;
f46b5a66
CH		 640
641 if (!root->ref_cows)
642 return -EINVAL;
643
8257b2dc
MX		 644 atomic_inc(&root->will_be_snapshoted);
645 smp_mb__after_atomic_inc();
646 btrfs_wait_nocow_write(root);
647
6a03843d
MX		 648 ret = btrfs_start_delalloc_inodes(root, 0);
649 if (ret)
8257b2dc 650 goto out;
6a03843d 651
b0244199 652 btrfs_wait_ordered_extents(root, -1);
6a03843d 653
3de4586c 654 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
8257b2dc
MX		 655 if (!pending_snapshot) {
656 ret = -ENOMEM;
657 goto out;
658 }
a22285a6 659
66d8f3dd
MX		 660 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
661 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 662 /*
663 * 1 - parent dir inode
664 * 2 - dir entries
665 * 1 - root item
666 * 2 - root ref/backref
667 * 1 - root of snapshot
dd5f9615 668 * 1 - UUID item
d5c12070
MX		 669 */
670 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 671 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 672 &pending_snapshot->qgroup_reserved,
673 false);
d5c12070 674 if (ret)
8257b2dc 675 goto free;
d5c12070 676
3de4586c 677 pending_snapshot->dentry = dentry;
f46b5a66 678 pending_snapshot->root = root;
b83cc969 679 pending_snapshot->readonly = readonly;
e9662f70 680 pending_snapshot->dir = dir;
8696c533 681 pending_snapshot->inherit = inherit;
a22285a6 682
d5c12070 683 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 684 if (IS_ERR(trans)) {
685 ret = PTR_ERR(trans);
686 goto fail;
687 }
688
8351583e 689 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 690 list_add(&pending_snapshot->list,
691 &trans->transaction->pending_snapshots);
8351583e 692 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 693 if (async_transid) {
694 *async_transid = trans->transid;
695 ret = btrfs_commit_transaction_async(trans,
696 root->fs_info->extent_root, 1);
00d71c9c
MX		 697 if (ret)
698 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 699 } else {
700 ret = btrfs_commit_transaction(trans,
701 root->fs_info->extent_root);
702 }
aec8030a 703 if (ret)
c37b2b62 704 goto fail;
a22285a6
YZ		 705
706 ret = pending_snapshot->error;
707 if (ret)
708 goto fail;
709
66b4ffd1
JB		 710 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
711 if (ret)
712 goto fail;
f46b5a66 713
2fbe8c8a 714 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 715 if (IS_ERR(inode)) {
716 ret = PTR_ERR(inode);
717 goto fail;
718 }
5662344b 719
2e4bfab9
YZ		 720 d_instantiate(dentry, inode);
721 ret = 0;
722fail:
d5c12070
MX		 723 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
724 &pending_snapshot->block_rsv,
725 pending_snapshot->qgroup_reserved);
8257b2dc 726free:
a22285a6 727 kfree(pending_snapshot);
8257b2dc
MX		 728out:
729 atomic_dec(&root->will_be_snapshoted);
f46b5a66
CH		 730 return ret;
731}
732
4260f7c7
SW		 733/* copy of check_sticky in fs/namei.c()
734* It's inline, so penalty for filesystems that don't use sticky bit is
735* minimal.
736*/
737static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
738{
2f2f43d3 739 kuid_t fsuid = current_fsuid();
4260f7c7
SW		 740
741 if (!(dir->i_mode & S_ISVTX))
742 return 0;
2f2f43d3 743 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 744 return 0;
2f2f43d3 745 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW		 746 return 0;
747 return !capable(CAP_FOWNER);
748}
749
750/* copy of may_delete in fs/namei.c()
751 * Check whether we can remove a link victim from directory dir, check
752 * whether the type of victim is right.
753 * 1. We can't do it if dir is read-only (done in permission())
754 * 2. We should have write and exec permissions on dir
755 * 3. We can't remove anything from append-only dir
756 * 4. We can't do anything with immutable dir (done in permission())
757 * 5. If the sticky bit on dir is set we should either
758 * a. be owner of dir, or
759 * b. be owner of victim, or
760 * c. have CAP_FOWNER capability
761 * 6. If the victim is append-only or immutable we can't do antyhing with
762 * links pointing to it.
763 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
764 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
765 * 9. We can't remove a root or mountpoint.
766 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
767 * nfs_async_unlink().
768 */
769
67871254 770static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 771{
772 int error;
773
774 if (!victim->d_inode)
775 return -ENOENT;
776
777 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 778 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 779
780 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
781 if (error)
782 return error;
783 if (IS_APPEND(dir))
784 return -EPERM;
785 if (btrfs_check_sticky(dir, victim->d_inode)||
786 IS_APPEND(victim->d_inode)||
787 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
788 return -EPERM;
789 if (isdir) {
790 if (!S_ISDIR(victim->d_inode->i_mode))
791 return -ENOTDIR;
792 if (IS_ROOT(victim))
793 return -EBUSY;
794 } else if (S_ISDIR(victim->d_inode->i_mode))
795 return -EISDIR;
796 if (IS_DEADDIR(dir))
797 return -ENOENT;
798 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
799 return -EBUSY;
800 return 0;
801}
802
cb8e7090
CH		 803/* copy of may_create in fs/namei.c() */
804static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
805{
806 if (child->d_inode)
807 return -EEXIST;
808 if (IS_DEADDIR(dir))
809 return -ENOENT;
810 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
811}
812
813/*
814 * Create a new subvolume below @parent. This is largely modeled after
815 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
816 * inside this filesystem so it's quite a bit simpler.
817 */
76dda93c
YZ		 818static noinline int btrfs_mksubvol(struct path *parent,
819 char *name, int namelen,
72fd032e 820 struct btrfs_root *snap_src,
6f72c7e2 821 u64 *async_transid, bool readonly,
8696c533 822 struct btrfs_qgroup_inherit *inherit)
cb8e7090 823{
76dda93c 824 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 825 struct dentry *dentry;
826 int error;
827
5c50c9b8
DS		 828 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
829 if (error == -EINTR)
830 return error;
cb8e7090
CH		 831
832 dentry = lookup_one_len(name, parent->dentry, namelen);
833 error = PTR_ERR(dentry);
834 if (IS_ERR(dentry))
835 goto out_unlock;
836
837 error = -EEXIST;
838 if (dentry->d_inode)
839 goto out_dput;
840
76dda93c 841 error = btrfs_may_create(dir, dentry);
cb8e7090 842 if (error)
a874a63e 843 goto out_dput;
cb8e7090 844
9c52057c
CM		 845 /*
846 * even if this name doesn't exist, we may get hash collisions.
847 * check for them now when we can safely fail
848 */
849 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
850 dir->i_ino, name,
851 namelen);
852 if (error)
853 goto out_dput;
854
76dda93c
YZ		 855 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
856
857 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
858 goto out_up_read;
859
3de4586c 860 if (snap_src) {
e9662f70 861 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 862 async_transid, readonly, inherit);
3de4586c 863 } else {
d5c12070
MX		 864 error = create_subvol(dir, dentry, name, namelen,
865 async_transid, inherit);
3de4586c 866 }
76dda93c
YZ		 867 if (!error)
868 fsnotify_mkdir(dir, dentry);
869out_up_read:
870 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 871out_dput:
872 dput(dentry);
873out_unlock:
76dda93c 874 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 875 return error;
876}
877
4cb5300b
CM		 878/*
879 * When we're defragging a range, we don't want to kick it off again
880 * if it is really just waiting for delalloc to send it down.
881 * If we find a nice big extent or delalloc range for the bytes in the
882 * file you want to defrag, we return 0 to let you know to skip this
883 * part of the file
884 */
885static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
886{
887 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
888 struct extent_map *em = NULL;
889 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
890 u64 end;
891
892 read_lock(&em_tree->lock);
893 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
894 read_unlock(&em_tree->lock);
895
896 if (em) {
897 end = extent_map_end(em);
898 free_extent_map(em);
899 if (end - offset > thresh)
900 return 0;
901 }
902 /* if we already have a nice delalloc here, just stop */
903 thresh /= 2;
904 end = count_range_bits(io_tree, &offset, offset + thresh,
905 thresh, EXTENT_DELALLOC, 1);
906 if (end >= thresh)
907 return 0;
908 return 1;
909}
910
911/*
912 * helper function to walk through a file and find extents
913 * newer than a specific transid, and smaller than thresh.
914 *
915 * This is used by the defragging code to find new and small
916 * extents
917 */
918static int find_new_extents(struct btrfs_root *root,
919 struct inode *inode, u64 newer_than,
920 u64 *off, int thresh)
921{
922 struct btrfs_path *path;
923 struct btrfs_key min_key;
4cb5300b
CM		 924 struct extent_buffer *leaf;
925 struct btrfs_file_extent_item *extent;
926 int type;
927 int ret;
a4689d2b 928 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 929
930 path = btrfs_alloc_path();
931 if (!path)
932 return -ENOMEM;
933
a4689d2b 934 min_key.objectid = ino;
4cb5300b
CM		 935 min_key.type = BTRFS_EXTENT_DATA_KEY;
936 min_key.offset = *off;
937
4cb5300b
CM		 938 path->keep_locks = 1;
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;
a4689d2b 944 if (min_key.objectid != ino)
4cb5300b
CM		 945 goto none;
946 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
947 goto none;
948
949 leaf = path->nodes[0];
950 extent = btrfs_item_ptr(leaf, path->slots[0],
951 struct btrfs_file_extent_item);
952
953 type = btrfs_file_extent_type(leaf, extent);
954 if (type == BTRFS_FILE_EXTENT_REG &&
955 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
956 check_defrag_in_cache(inode, min_key.offset, thresh)) {
957 *off = min_key.offset;
958 btrfs_free_path(path);
959 return 0;
960 }
961
962 if (min_key.offset == (u64)-1)
963 goto none;
964
965 min_key.offset++;
966 btrfs_release_path(path);
967 }
968none:
969 btrfs_free_path(path);
970 return -ENOENT;
971}
972
6c282eb4 973static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 974{
975 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 976 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
977 struct extent_map *em;
978 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 979
6c282eb4
LZ		 980 /*
981 * hopefully we have this extent in the tree already, try without
982 * the full extent lock
983 */
17ce6ef8 984 read_lock(&em_tree->lock);
6c282eb4 985 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 986 read_unlock(&em_tree->lock);
987
6c282eb4
LZ		 988 if (!em) {
989 /* get the big lock and read metadata off disk */
990 lock_extent(io_tree, start, start + len - 1);
991 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
992 unlock_extent(io_tree, start, start + len - 1);
993
994 if (IS_ERR(em))
995 return NULL;
996 }
997
998 return em;
999}
17ce6ef8 1000
6c282eb4
LZ		 1001static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1002{
1003 struct extent_map *next;
1004 bool ret = true;
1005
1006 /* this is the last extent */
1007 if (em->start + em->len >= i_size_read(inode))
1008 return false;
1009
1010 next = defrag_lookup_extent(inode, em->start + em->len);
e2127cf0
FM		 1011 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE ||
1012 (em->block_start + em->block_len == next->block_start))
6c282eb4
LZ		 1013 ret = false;
1014
1015 free_extent_map(next);
17ce6ef8
LB		 1016 return ret;
1017}
1018
6c282eb4 1019static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 1020 u64 *last_len, u64 *skip, u64 *defrag_end,
1021 int compress)
940100a4 1022{
6c282eb4 1023 struct extent_map *em;
940100a4 1024 int ret = 1;
6c282eb4 1025 bool next_mergeable = true;
940100a4
CM		 1026
1027 /*
008873ea 1028 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1029 * defragging it
1030 */
1031 if (start < *defrag_end)
1032 return 1;
1033
1034 *skip = 0;
1035
6c282eb4
LZ		 1036 em = defrag_lookup_extent(inode, start);
1037 if (!em)
1038 return 0;
940100a4
CM		 1039
1040 /* this will cover holes, and inline extents */
17ce6ef8 1041 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1042 ret = 0;
17ce6ef8
LB		 1043 goto out;
1044 }
1045
6c282eb4 1046 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM		 1047
1048 /*
6c282eb4
LZ		 1049 * we hit a real extent, if it is big or the next extent is not a
1050 * real extent, don't bother defragging it
940100a4 1051 */
a43a2111 1052 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 1053 (em->len >= thresh || !next_mergeable))
940100a4 1054 ret = 0;
17ce6ef8 1055out:
940100a4
CM		 1056 /*
1057 * last_len ends up being a counter of how many bytes we've defragged.
1058 * every time we choose not to defrag an extent, we reset *last_len
1059 * so that the next tiny extent will force a defrag.
1060 *
1061 * The end result of this is that tiny extents before a single big
1062 * extent will force at least part of that big extent to be defragged.
1063 */
1064 if (ret) {
940100a4
CM		 1065 *defrag_end = extent_map_end(em);
1066 } else {
1067 *last_len = 0;
1068 *skip = extent_map_end(em);
1069 *defrag_end = 0;
1070 }
1071
1072 free_extent_map(em);
1073 return ret;
1074}
1075
4cb5300b
CM		 1076/*
1077 * it doesn't do much good to defrag one or two pages
1078 * at a time. This pulls in a nice chunk of pages
1079 * to COW and defrag.
1080 *
1081 * It also makes sure the delalloc code has enough
1082 * dirty data to avoid making new small extents as part
1083 * of the defrag
1084 *
1085 * It's a good idea to start RA on this range
1086 * before calling this.
1087 */
1088static int cluster_pages_for_defrag(struct inode *inode,
1089 struct page **pages,
1090 unsigned long start_index,
c41570c9 1091 unsigned long num_pages)
f46b5a66 1092{
4cb5300b
CM		 1093 unsigned long file_end;
1094 u64 isize = i_size_read(inode);
1095 u64 page_start;
1096 u64 page_end;
1f12bd06 1097 u64 page_cnt;
4cb5300b
CM		 1098 int ret;
1099 int i;
1100 int i_done;
3eaa2885 1101 struct btrfs_ordered_extent *ordered;
4cb5300b 1102 struct extent_state *cached_state = NULL;
600a45e1 1103 struct extent_io_tree *tree;
3b16a4e3 1104 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1105
4cb5300b 1106 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1107 if (!isize || start_index > file_end)
1108 return 0;
1109
1110 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1111
1112 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1113 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1114 if (ret)
1115 return ret;
4cb5300b 1116 i_done = 0;
600a45e1 1117 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1118
1119 /* step one, lock all the pages */
1f12bd06 1120 for (i = 0; i < page_cnt; i++) {
4cb5300b 1121 struct page *page;
600a45e1 1122again:
a94733d0 1123 page = find_or_create_page(inode->i_mapping,
600a45e1 1124 start_index + i, mask);
4cb5300b
CM		 1125 if (!page)
1126 break;
1127
600a45e1
MX		 1128 page_start = page_offset(page);
1129 page_end = page_start + PAGE_CACHE_SIZE - 1;
1130 while (1) {
d0082371 1131 lock_extent(tree, page_start, page_end);
600a45e1
MX		 1132 ordered = btrfs_lookup_ordered_extent(inode,
1133 page_start);
d0082371 1134 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 1135 if (!ordered)
1136 break;
1137
1138 unlock_page(page);
1139 btrfs_start_ordered_extent(inode, ordered, 1);
1140 btrfs_put_ordered_extent(ordered);
1141 lock_page(page);
1f12bd06
LB		 1142 /*
1143 * we unlocked the page above, so we need check if
1144 * it was released or not.
1145 */
1146 if (page->mapping != inode->i_mapping) {
1147 unlock_page(page);
1148 page_cache_release(page);
1149 goto again;
1150 }
600a45e1
MX		 1151 }
1152
4cb5300b
CM		 1153 if (!PageUptodate(page)) {
1154 btrfs_readpage(NULL, page);
1155 lock_page(page);
1156 if (!PageUptodate(page)) {
1157 unlock_page(page);
1158 page_cache_release(page);
1159 ret = -EIO;
1160 break;
1161 }
1162 }
600a45e1 1163
600a45e1
MX		 1164 if (page->mapping != inode->i_mapping) {
1165 unlock_page(page);
1166 page_cache_release(page);
1167 goto again;
1168 }
1169
4cb5300b
CM		 1170 pages[i] = page;
1171 i_done++;
1172 }
1173 if (!i_done || ret)
1174 goto out;
1175
1176 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1177 goto out;
1178
1179 /*
1180 * so now we have a nice long stream of locked
1181 * and up to date pages, lets wait on them
1182 */
1183 for (i = 0; i < i_done; i++)
1184 wait_on_page_writeback(pages[i]);
1185
1186 page_start = page_offset(pages[0]);
1187 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1188
1189 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1190 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1191 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1192 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1193 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1194 &cached_state, GFP_NOFS);
4cb5300b 1195
1f12bd06 1196 if (i_done != page_cnt) {
9e0baf60
JB		 1197 spin_lock(&BTRFS_I(inode)->lock);
1198 BTRFS_I(inode)->outstanding_extents++;
1199 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1200 btrfs_delalloc_release_space(inode,
1f12bd06 1201 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1202 }
1203
1204
9e8a4a8b
LB		 1205 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1206 &cached_state, GFP_NOFS);
4cb5300b
CM		 1207
1208 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1209 page_start, page_end - 1, &cached_state,
1210 GFP_NOFS);
1211
1212 for (i = 0; i < i_done; i++) {
1213 clear_page_dirty_for_io(pages[i]);
1214 ClearPageChecked(pages[i]);
1215 set_page_extent_mapped(pages[i]);
1216 set_page_dirty(pages[i]);
1217 unlock_page(pages[i]);
1218 page_cache_release(pages[i]);
1219 }
1220 return i_done;
1221out:
1222 for (i = 0; i < i_done; i++) {
1223 unlock_page(pages[i]);
1224 page_cache_release(pages[i]);
1225 }
1f12bd06 1226 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1227 return ret;
1228
1229}
1230
1231int btrfs_defrag_file(struct inode *inode, struct file *file,
1232 struct btrfs_ioctl_defrag_range_args *range,
1233 u64 newer_than, unsigned long max_to_defrag)
1234{
1235 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1236 struct file_ra_state *ra = NULL;
f46b5a66 1237 unsigned long last_index;
151a31b2 1238 u64 isize = i_size_read(inode);
940100a4
CM		 1239 u64 last_len = 0;
1240 u64 skip = 0;
1241 u64 defrag_end = 0;
4cb5300b 1242 u64 newer_off = range->start;
f46b5a66 1243 unsigned long i;
008873ea 1244 unsigned long ra_index = 0;
f46b5a66 1245 int ret;
4cb5300b 1246 int defrag_count = 0;
1a419d85 1247 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1248 int extent_thresh = range->extent_thresh;
c41570c9
JM		 1249 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1250 unsigned long cluster = max_cluster;
4cb5300b
CM		 1251 u64 new_align = ~((u64)128 * 1024 - 1);
1252 struct page **pages = NULL;
1253
0abd5b17
LB		 1254 if (isize == 0)
1255 return 0;
1256
1257 if (range->start >= isize)
1258 return -EINVAL;
1a419d85
LZ		 1259
1260 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1261 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1262 return -EINVAL;
1263 if (range->compress_type)
1264 compress_type = range->compress_type;
1265 }
f46b5a66 1266
0abd5b17
LB		 1267 if (extent_thresh == 0)
1268 extent_thresh = 256 * 1024;
940100a4 1269
4cb5300b
CM		 1270 /*
1271 * if we were not given a file, allocate a readahead
1272 * context
1273 */
1274 if (!file) {
1275 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1276 if (!ra)
1277 return -ENOMEM;
1278 file_ra_state_init(ra, inode->i_mapping);
1279 } else {
1280 ra = &file->f_ra;
1281 }
1282
d9b0d9ba 1283 pages = kmalloc_array(max_cluster, sizeof(struct page *),
4cb5300b
CM		 1284 GFP_NOFS);
1285 if (!pages) {
1286 ret = -ENOMEM;
1287 goto out_ra;
1288 }
1289
1290 /* find the last page to defrag */
1e701a32 1291 if (range->start + range->len > range->start) {
151a31b2 1292 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1293 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1294 } else {
151a31b2 1295 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1296 }
1297
4cb5300b
CM		 1298 if (newer_than) {
1299 ret = find_new_extents(root, inode, newer_than,
1300 &newer_off, 64 * 1024);
1301 if (!ret) {
1302 range->start = newer_off;
1303 /*
1304 * we always align our defrag to help keep
1305 * the extents in the file evenly spaced
1306 */
1307 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1308 } else
1309 goto out_ra;
1310 } else {
1311 i = range->start >> PAGE_CACHE_SHIFT;
1312 }
1313 if (!max_to_defrag)
7ec31b54 1314 max_to_defrag = last_index + 1;
4cb5300b 1315
2a0f7f57
LZ		 1316 /*
1317 * make writeback starts from i, so the defrag range can be
1318 * written sequentially.
1319 */
1320 if (i < inode->i_mapping->writeback_index)
1321 inode->i_mapping->writeback_index = i;
1322
f7f43cc8
CM		 1323 while (i <= last_index && defrag_count < max_to_defrag &&
1324 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1325 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1326 /*
1327 * make sure we stop running if someone unmounts
1328 * the FS
1329 */
1330 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1331 break;
1332
210549eb 1333 if (btrfs_defrag_cancelled(root->fs_info)) {
efe120a0 1334 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
210549eb
DS		 1335 ret = -EAGAIN;
1336 break;
1337 }
1338
66c26892 1339 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1340 extent_thresh, &last_len, &skip,
a43a2111
AM		 1341 &defrag_end, range->flags &
1342 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1343 unsigned long next;
1344 /*
1345 * the should_defrag function tells us how much to skip
1346 * bump our counter by the suggested amount
1347 */
1348 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1349 i = max(i + 1, next);
1350 continue;
1351 }
008873ea
LZ		 1352
1353 if (!newer_than) {
1354 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1355 PAGE_CACHE_SHIFT) - i;
1356 cluster = min(cluster, max_cluster);
1357 } else {
1358 cluster = max_cluster;
1359 }
1360
008873ea
LZ		 1361 if (i + cluster > ra_index) {
1362 ra_index = max(i, ra_index);
1363 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1364 cluster);
1365 ra_index += max_cluster;
1366 }
940100a4 1367
ecb8bea8 1368 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1369 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1370 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1371 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1372 if (ret < 0) {
1373 mutex_unlock(&inode->i_mutex);
4cb5300b 1374 goto out_ra;
ecb8bea8 1375 }
4cb5300b
CM		 1376
1377 defrag_count += ret;
d0e1d66b 1378 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1379 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1380
1381 if (newer_than) {
1382 if (newer_off == (u64)-1)
1383 break;
1384
e1f041e1
LB		 1385 if (ret > 0)
1386 i += ret;
1387
4cb5300b
CM		 1388 newer_off = max(newer_off + 1,
1389 (u64)i << PAGE_CACHE_SHIFT);
1390
1391 ret = find_new_extents(root, inode,
1392 newer_than, &newer_off,
1393 64 * 1024);
1394 if (!ret) {
1395 range->start = newer_off;
1396 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1397 } else {
1398 break;
f46b5a66 1399 }
4cb5300b 1400 } else {
008873ea 1401 if (ret > 0) {
cbcc8326 1402 i += ret;
008873ea
LZ		 1403 last_len += ret << PAGE_CACHE_SHIFT;
1404 } else {
cbcc8326 1405 i++;
008873ea
LZ		 1406 last_len = 0;
1407 }
f46b5a66 1408 }
f46b5a66
CH		 1409 }
1410
dec8ef90 1411 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1412 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1413 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1414 &BTRFS_I(inode)->runtime_flags))
1415 filemap_flush(inode->i_mapping);
1416 }
1e701a32
CM		 1417
1418 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1419 /* the filemap_flush will queue IO into the worker threads, but
1420 * we have to make sure the IO is actually started and that
1421 * ordered extents get created before we return
1422 */
1423 atomic_inc(&root->fs_info->async_submit_draining);
1424 while (atomic_read(&root->fs_info->nr_async_submits) ||
1425 atomic_read(&root->fs_info->async_delalloc_pages)) {
1426 wait_event(root->fs_info->async_submit_wait,
1427 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1428 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1429 }
1430 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1431 }
1432
1a419d85 1433 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1434 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1435 }
1436
60ccf82f 1437 ret = defrag_count;
940100a4 1438
4cb5300b 1439out_ra:
633085c7
FDBM		 1440 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1441 mutex_lock(&inode->i_mutex);
1442 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1443 mutex_unlock(&inode->i_mutex);
1444 }
4cb5300b
CM		 1445 if (!file)
1446 kfree(ra);
1447 kfree(pages);
940100a4 1448 return ret;
f46b5a66
CH		 1449}
1450
198605a8 1451static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1452 void __user *arg)
f46b5a66
CH		 1453{
1454 u64 new_size;
1455 u64 old_size;
1456 u64 devid = 1;
496ad9aa 1457 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1458 struct btrfs_ioctl_vol_args *vol_args;
1459 struct btrfs_trans_handle *trans;
1460 struct btrfs_device *device = NULL;
1461 char *sizestr;
1462 char *devstr = NULL;
1463 int ret = 0;
f46b5a66
CH		 1464 int mod = 0;
1465
e441d54d
CM		 1466 if (!capable(CAP_SYS_ADMIN))
1467 return -EPERM;
1468
198605a8
MX		 1469 ret = mnt_want_write_file(file);
1470 if (ret)
1471 return ret;
1472
5ac00add
SB		 1473 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1474 1)) {
97547676 1475 mnt_drop_write_file(file);
e57138b3 1476 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1477 }
1478
5ac00add 1479 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1480 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1481 if (IS_ERR(vol_args)) {
1482 ret = PTR_ERR(vol_args);
1483 goto out;
1484 }
5516e595
MF		 1485
1486 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1487
f46b5a66
CH		 1488 sizestr = vol_args->name;
1489 devstr = strchr(sizestr, ':');
1490 if (devstr) {
1491 char *end;
1492 sizestr = devstr + 1;
1493 *devstr = '\0';
1494 devstr = vol_args->name;
1495 devid = simple_strtoull(devstr, &end, 10);
dfd79829
MX		 1496 if (!devid) {
1497 ret = -EINVAL;
1498 goto out_free;
1499 }
efe120a0 1500 btrfs_info(root->fs_info, "resizing devid %llu", devid);
f46b5a66 1501 }
dba60f3f 1502
aa1b8cd4 1503 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1504 if (!device) {
efe120a0 1505 btrfs_info(root->fs_info, "resizer unable to find device %llu",
c1c9ff7c 1506 devid);
dfd79829 1507 ret = -ENODEV;
c9e9f97b 1508 goto out_free;
f46b5a66 1509 }
dba60f3f
MX		 1510
1511 if (!device->writeable) {
efe120a0
FH		 1512 btrfs_info(root->fs_info,
1513 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1514 devid);
dfd79829 1515 ret = -EPERM;
4e42ae1b
LB		 1516 goto out_free;
1517 }
1518
f46b5a66
CH		 1519 if (!strcmp(sizestr, "max"))
1520 new_size = device->bdev->bd_inode->i_size;
1521 else {
1522 if (sizestr[0] == '-') {
1523 mod = -1;
1524 sizestr++;
1525 } else if (sizestr[0] == '+') {
1526 mod = 1;
1527 sizestr++;
1528 }
91748467 1529 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1530 if (new_size == 0) {
1531 ret = -EINVAL;
c9e9f97b 1532 goto out_free;
f46b5a66
CH		 1533 }
1534 }
1535
63a212ab 1536 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1537 ret = -EPERM;
63a212ab
SB		 1538 goto out_free;
1539 }
1540
f46b5a66
CH		 1541 old_size = device->total_bytes;
1542
1543 if (mod < 0) {
1544 if (new_size > old_size) {
1545 ret = -EINVAL;
c9e9f97b 1546 goto out_free;
f46b5a66
CH		 1547 }
1548 new_size = old_size - new_size;
1549 } else if (mod > 0) {
eb8052e0
WF		 1550 if (new_size > ULLONG_MAX - old_size) {
1551 ret = -EINVAL;
1552 goto out_free;
1553 }
f46b5a66
CH		 1554 new_size = old_size + new_size;
1555 }
1556
1557 if (new_size < 256 * 1024 * 1024) {
1558 ret = -EINVAL;
c9e9f97b 1559 goto out_free;
f46b5a66
CH		 1560 }
1561 if (new_size > device->bdev->bd_inode->i_size) {
1562 ret = -EFBIG;
c9e9f97b 1563 goto out_free;
f46b5a66
CH		 1564 }
1565
1566 do_div(new_size, root->sectorsize);
1567 new_size *= root->sectorsize;
1568
efe120a0 1569 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
c1c9ff7c 1570 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1571
1572 if (new_size > old_size) {
a22285a6 1573 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1574 if (IS_ERR(trans)) {
1575 ret = PTR_ERR(trans);
c9e9f97b 1576 goto out_free;
98d5dc13 1577 }
f46b5a66
CH		 1578 ret = btrfs_grow_device(trans, device, new_size);
1579 btrfs_commit_transaction(trans, root);
ece7d20e 1580 } else if (new_size < old_size) {
f46b5a66 1581 ret = btrfs_shrink_device(device, new_size);
0253f40e 1582 } /* equal, nothing need to do */
f46b5a66 1583
c9e9f97b 1584out_free:
f46b5a66 1585 kfree(vol_args);
c9e9f97b
ID		 1586out:
1587 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1588 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1589 mnt_drop_write_file(file);
f46b5a66
CH		 1590 return ret;
1591}
1592
72fd032e 1593static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1594 char *name, unsigned long fd, int subvol,
1595 u64 *transid, bool readonly,
8696c533 1596 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1597{
f46b5a66 1598 int namelen;
3de4586c 1599 int ret = 0;
f46b5a66 1600
a874a63e
LB		 1601 ret = mnt_want_write_file(file);
1602 if (ret)
1603 goto out;
1604
72fd032e
SW		 1605 namelen = strlen(name);
1606 if (strchr(name, '/')) {
f46b5a66 1607 ret = -EINVAL;
a874a63e 1608 goto out_drop_write;
f46b5a66
CH		 1609 }
1610
16780cab
CM		 1611 if (name[0] == '.' &&
1612 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1613 ret = -EEXIST;
a874a63e 1614 goto out_drop_write;
16780cab
CM		 1615 }
1616
3de4586c 1617 if (subvol) {
72fd032e 1618 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1619 NULL, transid, readonly, inherit);
cb8e7090 1620 } else {
2903ff01 1621 struct fd src = fdget(fd);
3de4586c 1622 struct inode *src_inode;
2903ff01 1623 if (!src.file) {
3de4586c 1624 ret = -EINVAL;
a874a63e 1625 goto out_drop_write;
3de4586c
CM		 1626 }
1627
496ad9aa
AV		 1628 src_inode = file_inode(src.file);
1629 if (src_inode->i_sb != file_inode(file)->i_sb) {
efe120a0
FH		 1630 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1631 "Snapshot src from another FS");
23ad5b17 1632 ret = -EXDEV;
d0242061
DS		 1633 } else if (!inode_owner_or_capable(src_inode)) {
1634 /*
1635 * Subvolume creation is not restricted, but snapshots
1636 * are limited to own subvolumes only
1637 */
1638 ret = -EPERM;
ecd18815
AV		 1639 } else {
1640 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1641 BTRFS_I(src_inode)->root,
1642 transid, readonly, inherit);
3de4586c 1643 }
2903ff01 1644 fdput(src);
cb8e7090 1645 }
a874a63e
LB		 1646out_drop_write:
1647 mnt_drop_write_file(file);
f46b5a66 1648out:
72fd032e
SW		 1649 return ret;
1650}
1651
1652static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1653 void __user *arg, int subvol)
72fd032e 1654{
fa0d2b9b 1655 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1656 int ret;
1657
fa0d2b9b
LZ		 1658 vol_args = memdup_user(arg, sizeof(*vol_args));
1659 if (IS_ERR(vol_args))
1660 return PTR_ERR(vol_args);
1661 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1662
fa0d2b9b 1663 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1664 vol_args->fd, subvol,
6f72c7e2 1665 NULL, false, NULL);
fdfb1e4f 1666
fa0d2b9b
LZ		 1667 kfree(vol_args);
1668 return ret;
1669}
fdfb1e4f 1670
fa0d2b9b
LZ		 1671static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1672 void __user *arg, int subvol)
1673{
1674 struct btrfs_ioctl_vol_args_v2 *vol_args;
1675 int ret;
1676 u64 transid = 0;
1677 u64 *ptr = NULL;
b83cc969 1678 bool readonly = false;
6f72c7e2 1679 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1680
fa0d2b9b
LZ		 1681 vol_args = memdup_user(arg, sizeof(*vol_args));
1682 if (IS_ERR(vol_args))
1683 return PTR_ERR(vol_args);
1684 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1685
b83cc969 1686 if (vol_args->flags &
6f72c7e2
AJ		 1687 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1688 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1689 ret = -EOPNOTSUPP;
fa0d2b9b 1690 goto out;
72fd032e 1691 }
fa0d2b9b
LZ		 1692
1693 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1694 ptr = &transid;
b83cc969
LZ		 1695 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1696 readonly = true;
6f72c7e2
AJ		 1697 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1698 if (vol_args->size > PAGE_CACHE_SIZE) {
1699 ret = -EINVAL;
1700 goto out;
1701 }
1702 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1703 if (IS_ERR(inherit)) {
1704 ret = PTR_ERR(inherit);
1705 goto out;
1706 }
1707 }
fa0d2b9b
LZ		 1708
1709 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1710 vol_args->fd, subvol, ptr,
8696c533 1711 readonly, inherit);
fa0d2b9b
LZ		 1712
1713 if (ret == 0 && ptr &&
1714 copy_to_user(arg +
1715 offsetof(struct btrfs_ioctl_vol_args_v2,
1716 transid), ptr, sizeof(*ptr)))
1717 ret = -EFAULT;
fdfb1e4f 1718out:
f46b5a66 1719 kfree(vol_args);
6f72c7e2 1720 kfree(inherit);
f46b5a66
CH		 1721 return ret;
1722}
1723
0caa102d
LZ		 1724static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1725 void __user *arg)
1726{
496ad9aa 1727 struct inode *inode = file_inode(file);
0caa102d
LZ		 1728 struct btrfs_root *root = BTRFS_I(inode)->root;
1729 int ret = 0;
1730 u64 flags = 0;
1731
33345d01 1732 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1733 return -EINVAL;
1734
1735 down_read(&root->fs_info->subvol_sem);
1736 if (btrfs_root_readonly(root))
1737 flags |= BTRFS_SUBVOL_RDONLY;
1738 up_read(&root->fs_info->subvol_sem);
1739
1740 if (copy_to_user(arg, &flags, sizeof(flags)))
1741 ret = -EFAULT;
1742
1743 return ret;
1744}
1745
1746static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1747 void __user *arg)
1748{
496ad9aa 1749 struct inode *inode = file_inode(file);
0caa102d
LZ		 1750 struct btrfs_root *root = BTRFS_I(inode)->root;
1751 struct btrfs_trans_handle *trans;
1752 u64 root_flags;
1753 u64 flags;
1754 int ret = 0;
1755
bd60ea0f
DS		 1756 if (!inode_owner_or_capable(inode))
1757 return -EPERM;
1758
b9ca0664
LB		 1759 ret = mnt_want_write_file(file);
1760 if (ret)
1761 goto out;
0caa102d 1762
b9ca0664
LB		 1763 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1764 ret = -EINVAL;
1765 goto out_drop_write;
1766 }
0caa102d 1767
b9ca0664
LB		 1768 if (copy_from_user(&flags, arg, sizeof(flags))) {
1769 ret = -EFAULT;
1770 goto out_drop_write;
1771 }
0caa102d 1772
b9ca0664
LB		 1773 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1774 ret = -EINVAL;
1775 goto out_drop_write;
1776 }
0caa102d 1777
b9ca0664
LB		 1778 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1779 ret = -EOPNOTSUPP;
1780 goto out_drop_write;
1781 }
0caa102d
LZ		 1782
1783 down_write(&root->fs_info->subvol_sem);
1784
1785 /* nothing to do */
1786 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1787 goto out_drop_sem;
0caa102d
LZ		 1788
1789 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1790 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1791 btrfs_set_root_flags(&root->root_item,
1792 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1793 } else {
1794 /*
1795 * Block RO -> RW transition if this subvolume is involved in
1796 * send
1797 */
1798 spin_lock(&root->root_item_lock);
1799 if (root->send_in_progress == 0) {
1800 btrfs_set_root_flags(&root->root_item,
0caa102d 1801 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1802 spin_unlock(&root->root_item_lock);
1803 } else {
1804 spin_unlock(&root->root_item_lock);
1805 btrfs_warn(root->fs_info,
1806 "Attempt to set subvolume %llu read-write during send",
1807 root->root_key.objectid);
1808 ret = -EPERM;
1809 goto out_drop_sem;
1810 }
1811 }
0caa102d
LZ		 1812
1813 trans = btrfs_start_transaction(root, 1);
1814 if (IS_ERR(trans)) {
1815 ret = PTR_ERR(trans);
1816 goto out_reset;
1817 }
1818
b4dc2b8c 1819 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1820 &root->root_key, &root->root_item);
1821
1822 btrfs_commit_transaction(trans, root);
1823out_reset:
1824 if (ret)
1825 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1826out_drop_sem:
0caa102d 1827 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1828out_drop_write:
1829 mnt_drop_write_file(file);
1830out:
0caa102d
LZ		 1831 return ret;
1832}
1833
76dda93c
YZ		 1834/*
1835 * helper to check if the subvolume references other subvolumes
1836 */
1837static noinline int may_destroy_subvol(struct btrfs_root *root)
1838{
1839 struct btrfs_path *path;
175a2b87 1840 struct btrfs_dir_item *di;
76dda93c 1841 struct btrfs_key key;
175a2b87 1842 u64 dir_id;
76dda93c
YZ		 1843 int ret;
1844
1845 path = btrfs_alloc_path();
1846 if (!path)
1847 return -ENOMEM;
1848
175a2b87
JB		 1849 /* Make sure this root isn't set as the default subvol */
1850 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1851 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1852 dir_id, "default", 7, 0);
1853 if (di && !IS_ERR(di)) {
1854 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1855 if (key.objectid == root->root_key.objectid) {
1856 ret = -ENOTEMPTY;
1857 goto out;
1858 }
1859 btrfs_release_path(path);
1860 }
1861
76dda93c
YZ		 1862 key.objectid = root->root_key.objectid;
1863 key.type = BTRFS_ROOT_REF_KEY;
1864 key.offset = (u64)-1;
1865
1866 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1867 &key, path, 0, 0);
1868 if (ret < 0)
1869 goto out;
1870 BUG_ON(ret == 0);
1871
1872 ret = 0;
1873 if (path->slots[0] > 0) {
1874 path->slots[0]--;
1875 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1876 if (key.objectid == root->root_key.objectid &&
1877 key.type == BTRFS_ROOT_REF_KEY)
1878 ret = -ENOTEMPTY;
1879 }
1880out:
1881 btrfs_free_path(path);
1882 return ret;
1883}
1884
ac8e9819
CM		 1885static noinline int key_in_sk(struct btrfs_key *key,
1886 struct btrfs_ioctl_search_key *sk)
1887{
abc6e134
CM		 1888 struct btrfs_key test;
1889 int ret;
1890
1891 test.objectid = sk->min_objectid;
1892 test.type = sk->min_type;
1893 test.offset = sk->min_offset;
1894
1895 ret = btrfs_comp_cpu_keys(key, &test);
1896 if (ret < 0)
ac8e9819 1897 return 0;
abc6e134
CM		 1898
1899 test.objectid = sk->max_objectid;
1900 test.type = sk->max_type;
1901 test.offset = sk->max_offset;
1902
1903 ret = btrfs_comp_cpu_keys(key, &test);
1904 if (ret > 0)
ac8e9819
CM		 1905 return 0;
1906 return 1;
1907}
1908
1909static noinline int copy_to_sk(struct btrfs_root *root,
1910 struct btrfs_path *path,
1911 struct btrfs_key *key,
1912 struct btrfs_ioctl_search_key *sk,
1913 char *buf,
1914 unsigned long *sk_offset,
1915 int *num_found)
1916{
1917 u64 found_transid;
1918 struct extent_buffer *leaf;
1919 struct btrfs_ioctl_search_header sh;
1920 unsigned long item_off;
1921 unsigned long item_len;
1922 int nritems;
1923 int i;
1924 int slot;
ac8e9819
CM		 1925 int ret = 0;
1926
1927 leaf = path->nodes[0];
1928 slot = path->slots[0];
1929 nritems = btrfs_header_nritems(leaf);
1930
1931 if (btrfs_header_generation(leaf) > sk->max_transid) {
1932 i = nritems;
1933 goto advance_key;
1934 }
1935 found_transid = btrfs_header_generation(leaf);
1936
1937 for (i = slot; i < nritems; i++) {
1938 item_off = btrfs_item_ptr_offset(leaf, i);
1939 item_len = btrfs_item_size_nr(leaf, i);
1940
03b71c6c
GP		 1941 btrfs_item_key_to_cpu(leaf, key, i);
1942 if (!key_in_sk(key, sk))
1943 continue;
1944
1945 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
ac8e9819
CM		 1946 item_len = 0;
1947
1948 if (sizeof(sh) + item_len + *sk_offset >
1949 BTRFS_SEARCH_ARGS_BUFSIZE) {
1950 ret = 1;
1951 goto overflow;
1952 }
1953
ac8e9819
CM		 1954 sh.objectid = key->objectid;
1955 sh.offset = key->offset;
1956 sh.type = key->type;
1957 sh.len = item_len;
1958 sh.transid = found_transid;
1959
1960 /* copy search result header */
1961 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1962 *sk_offset += sizeof(sh);
1963
1964 if (item_len) {
1965 char *p = buf + *sk_offset;
1966 /* copy the item */
1967 read_extent_buffer(leaf, p,
1968 item_off, item_len);
1969 *sk_offset += item_len;
ac8e9819 1970 }
e2156867 1971 (*num_found)++;
ac8e9819
CM		 1972
1973 if (*num_found >= sk->nr_items)
1974 break;
1975 }
1976advance_key:
abc6e134
CM		 1977 ret = 0;
1978 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1979 key->offset++;
abc6e134
CM		 1980 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1981 key->offset = 0;
ac8e9819 1982 key->type++;
abc6e134
CM		 1983 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1984 key->offset = 0;
1985 key->type = 0;
ac8e9819 1986 key->objectid++;
abc6e134
CM		 1987 } else
1988 ret = 1;
ac8e9819 1989overflow:
ac8e9819
CM		 1990 return ret;
1991}
1992
1993static noinline int search_ioctl(struct inode *inode,
1994 struct btrfs_ioctl_search_args *args)
1995{
1996 struct btrfs_root *root;
1997 struct btrfs_key key;
ac8e9819
CM		 1998 struct btrfs_path *path;
1999 struct btrfs_ioctl_search_key *sk = &args->key;
2000 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2001 int ret;
2002 int num_found = 0;
2003 unsigned long sk_offset = 0;
2004
2005 path = btrfs_alloc_path();
2006 if (!path)
2007 return -ENOMEM;
2008
2009 if (sk->tree_id == 0) {
2010 /* search the root of the inode that was passed */
2011 root = BTRFS_I(inode)->root;
2012 } else {
2013 key.objectid = sk->tree_id;
2014 key.type = BTRFS_ROOT_ITEM_KEY;
2015 key.offset = (u64)-1;
2016 root = btrfs_read_fs_root_no_name(info, &key);
2017 if (IS_ERR(root)) {
efe120a0 2018 printk(KERN_ERR "BTRFS: could not find root %llu\n",
ac8e9819
CM		 2019 sk->tree_id);
2020 btrfs_free_path(path);
2021 return -ENOENT;
2022 }
2023 }
2024
2025 key.objectid = sk->min_objectid;
2026 key.type = sk->min_type;
2027 key.offset = sk->min_offset;
2028
ac8e9819
CM		 2029 path->keep_locks = 1;
2030
67871254 2031 while (1) {
6174d3cb 2032 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2033 if (ret != 0) {
2034 if (ret > 0)
2035 ret = 0;
2036 goto err;
2037 }
2038 ret = copy_to_sk(root, path, &key, sk, args->buf,
2039 &sk_offset, &num_found);
b3b4aa74 2040 btrfs_release_path(path);
ac8e9819
CM		 2041 if (ret || num_found >= sk->nr_items)
2042 break;
2043
2044 }
2045 ret = 0;
2046err:
2047 sk->nr_items = num_found;
2048 btrfs_free_path(path);
2049 return ret;
2050}
2051
2052static noinline int btrfs_ioctl_tree_search(struct file *file,
2053 void __user *argp)
2054{
2055 struct btrfs_ioctl_search_args *args;
2056 struct inode *inode;
2057 int ret;
2058
2059 if (!capable(CAP_SYS_ADMIN))
2060 return -EPERM;
2061
2354d08f
JL		 2062 args = memdup_user(argp, sizeof(*args));
2063 if (IS_ERR(args))
2064 return PTR_ERR(args);
ac8e9819 2065
496ad9aa 2066 inode = file_inode(file);
ac8e9819
CM		 2067 ret = search_ioctl(inode, args);
2068 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2069 ret = -EFAULT;
2070 kfree(args);
2071 return ret;
2072}
2073
98d377a0 2074/*
ac8e9819
CM		 2075 * Search INODE_REFs to identify path name of 'dirid' directory
2076 * in a 'tree_id' tree. and sets path name to 'name'.
2077 */
98d377a0
TH		 2078static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2079 u64 tree_id, u64 dirid, char *name)
2080{
2081 struct btrfs_root *root;
2082 struct btrfs_key key;
ac8e9819 2083 char *ptr;
98d377a0
TH		 2084 int ret = -1;
2085 int slot;
2086 int len;
2087 int total_len = 0;
2088 struct btrfs_inode_ref *iref;
2089 struct extent_buffer *l;
2090 struct btrfs_path *path;
2091
2092 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2093 name[0]='\0';
2094 return 0;
2095 }
2096
2097 path = btrfs_alloc_path();
2098 if (!path)
2099 return -ENOMEM;
2100
ac8e9819 2101 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2102
2103 key.objectid = tree_id;
2104 key.type = BTRFS_ROOT_ITEM_KEY;
2105 key.offset = (u64)-1;
2106 root = btrfs_read_fs_root_no_name(info, &key);
2107 if (IS_ERR(root)) {
efe120a0 2108 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
8ad6fcab
CM		 2109 ret = -ENOENT;
2110 goto out;
98d377a0
TH		 2111 }
2112
2113 key.objectid = dirid;
2114 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2115 key.offset = (u64)-1;
98d377a0 2116
67871254 2117 while (1) {
98d377a0
TH		 2118 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2119 if (ret < 0)
2120 goto out;
18674c6c
FDBM		 2121 else if (ret > 0) {
2122 ret = btrfs_previous_item(root, path, dirid,
2123 BTRFS_INODE_REF_KEY);
2124 if (ret < 0)
2125 goto out;
2126 else if (ret > 0) {
2127 ret = -ENOENT;
2128 goto out;
2129 }
2130 }
98d377a0
TH		 2131
2132 l = path->nodes[0];
2133 slot = path->slots[0];
2134 btrfs_item_key_to_cpu(l, &key, slot);
2135
98d377a0
TH		 2136 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2137 len = btrfs_inode_ref_name_len(l, iref);
2138 ptr -= len + 1;
2139 total_len += len + 1;
a696cf35
FDBM		 2140 if (ptr < name) {
2141 ret = -ENAMETOOLONG;
98d377a0 2142 goto out;
a696cf35 2143 }
98d377a0
TH		 2144
2145 *(ptr + len) = '/';
67871254 2146 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2147
2148 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2149 break;
2150
b3b4aa74 2151 btrfs_release_path(path);
98d377a0 2152 key.objectid = key.offset;
8ad6fcab 2153 key.offset = (u64)-1;
98d377a0 2154 dirid = key.objectid;
98d377a0 2155 }
77906a50 2156 memmove(name, ptr, total_len);
67871254 2157 name[total_len] = '\0';
98d377a0
TH		 2158 ret = 0;
2159out:
2160 btrfs_free_path(path);
ac8e9819
CM		 2161 return ret;
2162}
2163
2164static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2165 void __user *argp)
2166{
2167 struct btrfs_ioctl_ino_lookup_args *args;
2168 struct inode *inode;
2169 int ret;
2170
2171 if (!capable(CAP_SYS_ADMIN))
2172 return -EPERM;
2173
2354d08f
JL		 2174 args = memdup_user(argp, sizeof(*args));
2175 if (IS_ERR(args))
2176 return PTR_ERR(args);
c2b96929 2177
496ad9aa 2178 inode = file_inode(file);
ac8e9819 2179
1b53ac4d
CM		 2180 if (args->treeid == 0)
2181 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2182
ac8e9819
CM		 2183 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2184 args->treeid, args->objectid,
2185 args->name);
2186
2187 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2188 ret = -EFAULT;
2189
2190 kfree(args);
98d377a0
TH		 2191 return ret;
2192}
2193
76dda93c
YZ		 2194static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2195 void __user *arg)
2196{
54563d41 2197 struct dentry *parent = file->f_path.dentry;
76dda93c
YZ		 2198 struct dentry *dentry;
2199 struct inode *dir = parent->d_inode;
2200 struct inode *inode;
2201 struct btrfs_root *root = BTRFS_I(dir)->root;
2202 struct btrfs_root *dest = NULL;
2203 struct btrfs_ioctl_vol_args *vol_args;
2204 struct btrfs_trans_handle *trans;
c58aaad2
MX		 2205 struct btrfs_block_rsv block_rsv;
2206 u64 qgroup_reserved;
76dda93c
YZ		 2207 int namelen;
2208 int ret;
2209 int err = 0;
2210
76dda93c
YZ		 2211 vol_args = memdup_user(arg, sizeof(*vol_args));
2212 if (IS_ERR(vol_args))
2213 return PTR_ERR(vol_args);
2214
2215 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2216 namelen = strlen(vol_args->name);
2217 if (strchr(vol_args->name, '/') ||
2218 strncmp(vol_args->name, "..", namelen) == 0) {
2219 err = -EINVAL;
2220 goto out;
2221 }
2222
a561be71 2223 err = mnt_want_write_file(file);
76dda93c
YZ		 2224 if (err)
2225 goto out;
2226
5c50c9b8
DS		 2227 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2228 if (err == -EINTR)
e43f998e 2229 goto out_drop_write;
76dda93c
YZ		 2230 dentry = lookup_one_len(vol_args->name, parent, namelen);
2231 if (IS_ERR(dentry)) {
2232 err = PTR_ERR(dentry);
2233 goto out_unlock_dir;
2234 }
2235
2236 if (!dentry->d_inode) {
2237 err = -ENOENT;
2238 goto out_dput;
2239 }
2240
2241 inode = dentry->d_inode;
4260f7c7 2242 dest = BTRFS_I(inode)->root;
67871254 2243 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2244 /*
2245 * Regular user. Only allow this with a special mount
2246 * option, when the user has write+exec access to the
2247 * subvol root, and when rmdir(2) would have been
2248 * allowed.
2249 *
2250 * Note that this is _not_ check that the subvol is
2251 * empty or doesn't contain data that we wouldn't
2252 * otherwise be able to delete.
2253 *
2254 * Users who want to delete empty subvols should try
2255 * rmdir(2).
2256 */
2257 err = -EPERM;
2258 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2259 goto out_dput;
2260
2261 /*
2262 * Do not allow deletion if the parent dir is the same
2263 * as the dir to be deleted. That means the ioctl
2264 * must be called on the dentry referencing the root
2265 * of the subvol, not a random directory contained
2266 * within it.
2267 */
2268 err = -EINVAL;
2269 if (root == dest)
2270 goto out_dput;
2271
2272 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2273 if (err)
2274 goto out_dput;
4260f7c7
SW		 2275 }
2276
5c39da5b
MX		 2277 /* check if subvolume may be deleted by a user */
2278 err = btrfs_may_delete(dir, dentry, 1);
2279 if (err)
2280 goto out_dput;
2281
33345d01 2282 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2283 err = -EINVAL;
2284 goto out_dput;
2285 }
2286
76dda93c
YZ		 2287 mutex_lock(&inode->i_mutex);
2288 err = d_invalidate(dentry);
2289 if (err)
2290 goto out_unlock;
2291
2292 down_write(&root->fs_info->subvol_sem);
2293
2294 err = may_destroy_subvol(dest);
2295 if (err)
2296 goto out_up_write;
2297
c58aaad2
MX		 2298 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2299 /*
2300 * One for dir inode, two for dir entries, two for root
2301 * ref/backref.
2302 */
2303 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2304 5, &qgroup_reserved, true);
c58aaad2
MX		 2305 if (err)
2306 goto out_up_write;
2307
a22285a6
YZ		 2308 trans = btrfs_start_transaction(root, 0);
2309 if (IS_ERR(trans)) {
2310 err = PTR_ERR(trans);
c58aaad2 2311 goto out_release;
a22285a6 2312 }
c58aaad2
MX		 2313 trans->block_rsv = &block_rsv;
2314 trans->bytes_reserved = block_rsv.size;
a22285a6 2315
76dda93c
YZ		 2316 ret = btrfs_unlink_subvol(trans, root, dir,
2317 dest->root_key.objectid,
2318 dentry->d_name.name,
2319 dentry->d_name.len);
79787eaa
JM		 2320 if (ret) {
2321 err = ret;
2322 btrfs_abort_transaction(trans, root, ret);
2323 goto out_end_trans;
2324 }
76dda93c
YZ		 2325
2326 btrfs_record_root_in_trans(trans, dest);
2327
2328 memset(&dest->root_item.drop_progress, 0,
2329 sizeof(dest->root_item.drop_progress));
2330 dest->root_item.drop_level = 0;
2331 btrfs_set_root_refs(&dest->root_item, 0);
2332
d68fc57b
YZ		 2333 if (!xchg(&dest->orphan_item_inserted, 1)) {
2334 ret = btrfs_insert_orphan_item(trans,
2335 root->fs_info->tree_root,
2336 dest->root_key.objectid);
79787eaa
JM		 2337 if (ret) {
2338 btrfs_abort_transaction(trans, root, ret);
2339 err = ret;
2340 goto out_end_trans;
2341 }
d68fc57b 2342 }
dd5f9615
SB		 2343
2344 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2345 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2346 dest->root_key.objectid);
2347 if (ret && ret != -ENOENT) {
2348 btrfs_abort_transaction(trans, root, ret);
2349 err = ret;
2350 goto out_end_trans;
2351 }
2352 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2353 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2354 dest->root_item.received_uuid,
2355 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2356 dest->root_key.objectid);
2357 if (ret && ret != -ENOENT) {
2358 btrfs_abort_transaction(trans, root, ret);
2359 err = ret;
2360 goto out_end_trans;
2361 }
2362 }
2363
79787eaa 2364out_end_trans:
c58aaad2
MX		 2365 trans->block_rsv = NULL;
2366 trans->bytes_reserved = 0;
531cb13f 2367 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2368 if (ret && !err)
2369 err = ret;
76dda93c 2370 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2371out_release:
2372 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2373out_up_write:
2374 up_write(&root->fs_info->subvol_sem);
2375out_unlock:
2376 mutex_unlock(&inode->i_mutex);
2377 if (!err) {
efefb143 2378 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2379 btrfs_invalidate_inodes(dest);
2380 d_delete(dentry);
fa6ac876
LB		 2381
2382 /* the last ref */
2383 if (dest->cache_inode) {
2384 iput(dest->cache_inode);
2385 dest->cache_inode = NULL;
2386 }
76dda93c
YZ		 2387 }
2388out_dput:
2389 dput(dentry);
2390out_unlock_dir:
2391 mutex_unlock(&dir->i_mutex);
e43f998e 2392out_drop_write:
2a79f17e 2393 mnt_drop_write_file(file);
76dda93c
YZ		 2394out:
2395 kfree(vol_args);
2396 return err;
2397}
2398
1e701a32 2399static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2400{
496ad9aa 2401 struct inode *inode = file_inode(file);
f46b5a66 2402 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2403 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2404 int ret;
2405
25122d15
ID		 2406 ret = mnt_want_write_file(file);
2407 if (ret)
2408 return ret;
b83cc969 2409
25122d15
ID		 2410 if (btrfs_root_readonly(root)) {
2411 ret = -EROFS;
2412 goto out;
5ac00add 2413 }
f46b5a66
CH		 2414
2415 switch (inode->i_mode & S_IFMT) {
2416 case S_IFDIR:
e441d54d
CM		 2417 if (!capable(CAP_SYS_ADMIN)) {
2418 ret = -EPERM;
2419 goto out;
2420 }
de78b51a 2421 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2422 if (ret)
2423 goto out;
de78b51a 2424 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2425 break;
2426 case S_IFREG:
e441d54d
CM		 2427 if (!(file->f_mode & FMODE_WRITE)) {
2428 ret = -EINVAL;
2429 goto out;
2430 }
1e701a32
CM		 2431
2432 range = kzalloc(sizeof(*range), GFP_KERNEL);
2433 if (!range) {
2434 ret = -ENOMEM;
2435 goto out;
2436 }
2437
2438 if (argp) {
2439 if (copy_from_user(range, argp,
2440 sizeof(*range))) {
2441 ret = -EFAULT;
2442 kfree(range);
683be16e 2443 goto out;
1e701a32
CM		 2444 }
2445 /* compression requires us to start the IO */
2446 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2447 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2448 range->extent_thresh = (u32)-1;
2449 }
2450 } else {
2451 /* the rest are all set to zero by kzalloc */
2452 range->len = (u64)-1;
2453 }
496ad9aa 2454 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2455 range, 0, 0);
2456 if (ret > 0)
2457 ret = 0;
1e701a32 2458 kfree(range);
f46b5a66 2459 break;
8929ecfa
YZ		 2460 default:
2461 ret = -EINVAL;
f46b5a66 2462 }
e441d54d 2463out:
25122d15 2464 mnt_drop_write_file(file);
e441d54d 2465 return ret;
f46b5a66
CH		 2466}
2467
b2950863 2468static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2469{
2470 struct btrfs_ioctl_vol_args *vol_args;
2471 int ret;
2472
e441d54d
CM		 2473 if (!capable(CAP_SYS_ADMIN))
2474 return -EPERM;
2475
5ac00add
SB		 2476 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2477 1)) {
e57138b3 2478 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 2479 }
2480
5ac00add 2481 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2482 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2483 if (IS_ERR(vol_args)) {
2484 ret = PTR_ERR(vol_args);
2485 goto out;
2486 }
f46b5a66 2487
5516e595 2488 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2489 ret = btrfs_init_new_device(root, vol_args->name);
2490
f46b5a66 2491 kfree(vol_args);
c9e9f97b
ID		 2492out:
2493 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2494 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2495 return ret;
2496}
2497
da24927b 2498static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2499{
496ad9aa 2500 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2501 struct btrfs_ioctl_vol_args *vol_args;
2502 int ret;
2503
e441d54d
CM		 2504 if (!capable(CAP_SYS_ADMIN))
2505 return -EPERM;
2506
da24927b
MX		 2507 ret = mnt_want_write_file(file);
2508 if (ret)
2509 return ret;
c146afad 2510
dae7b665 2511 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2512 if (IS_ERR(vol_args)) {
2513 ret = PTR_ERR(vol_args);
2514 goto out;
2515 }
f46b5a66 2516
5516e595 2517 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2518
183860f6
AJ		 2519 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2520 1)) {
2521 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2522 goto out;
2523 }
2524
2525 mutex_lock(&root->fs_info->volume_mutex);
2526 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2527 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2528 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6
AJ		 2529
2530out:
2531 kfree(vol_args);
4ac20c70 2532 mnt_drop_write_file(file);
f46b5a66
CH		 2533 return ret;
2534}
2535
475f6387
JS		 2536static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2537{
027ed2f0 2538 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2539 struct btrfs_device *device;
2540 struct btrfs_device *next;
2541 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2542 int ret = 0;
475f6387
JS		 2543
2544 if (!capable(CAP_SYS_ADMIN))
2545 return -EPERM;
2546
027ed2f0
LZ		 2547 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2548 if (!fi_args)
2549 return -ENOMEM;
2550
f7171750 2551 mutex_lock(&fs_devices->device_list_mutex);
027ed2f0
LZ		 2552 fi_args->num_devices = fs_devices->num_devices;
2553 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387 2554
475f6387 2555 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2556 if (device->devid > fi_args->max_id)
2557 fi_args->max_id = device->devid;
475f6387
JS		 2558 }
2559 mutex_unlock(&fs_devices->device_list_mutex);
2560
027ed2f0
LZ		 2561 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2562 ret = -EFAULT;
475f6387 2563
027ed2f0
LZ		 2564 kfree(fi_args);
2565 return ret;
475f6387
JS		 2566}
2567
2568static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2569{
2570 struct btrfs_ioctl_dev_info_args *di_args;
2571 struct btrfs_device *dev;
2572 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2573 int ret = 0;
2574 char *s_uuid = NULL;
475f6387
JS		 2575
2576 if (!capable(CAP_SYS_ADMIN))
2577 return -EPERM;
2578
2579 di_args = memdup_user(arg, sizeof(*di_args));
2580 if (IS_ERR(di_args))
2581 return PTR_ERR(di_args);
2582
dd5f9615 2583 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2584 s_uuid = di_args->uuid;
2585
2586 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2587 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2588
2589 if (!dev) {
2590 ret = -ENODEV;
2591 goto out;
2592 }
2593
2594 di_args->devid = dev->devid;
2595 di_args->bytes_used = dev->bytes_used;
2596 di_args->total_bytes = dev->total_bytes;
2597 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2598 if (dev->name) {
606686ee
JB		 2599 struct rcu_string *name;
2600
2601 rcu_read_lock();
2602 name = rcu_dereference(dev->name);
2603 strncpy(di_args->path, name->str, sizeof(di_args->path));
2604 rcu_read_unlock();
a27202fb
JM		 2605 di_args->path[sizeof(di_args->path) - 1] = 0;
2606 } else {
99ba55ad 2607 di_args->path[0] = '\0';
a27202fb 2608 }
475f6387
JS		 2609
2610out:
55793c0d 2611 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2612 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2613 ret = -EFAULT;
2614
2615 kfree(di_args);
2616 return ret;
2617}
2618
416161db
MF		 2619static struct page *extent_same_get_page(struct inode *inode, u64 off)
2620{
2621 struct page *page;
2622 pgoff_t index;
2623 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2624
2625 index = off >> PAGE_CACHE_SHIFT;
2626
2627 page = grab_cache_page(inode->i_mapping, index);
2628 if (!page)
2629 return NULL;
2630
2631 if (!PageUptodate(page)) {
2632 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2633 0))
2634 return NULL;
2635 lock_page(page);
2636 if (!PageUptodate(page)) {
2637 unlock_page(page);
2638 page_cache_release(page);
2639 return NULL;
2640 }
2641 }
2642 unlock_page(page);
2643
2644 return page;
2645}
2646
77fe20dc
MF		 2647static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2648{
2649 /* do any pending delalloc/csum calc on src, one way or
2650 another, and lock file content */
2651 while (1) {
2652 struct btrfs_ordered_extent *ordered;
2653 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2654 ordered = btrfs_lookup_first_ordered_extent(inode,
2655 off + len - 1);
2656 if (!ordered &&
2657 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2658 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2659 break;
2660 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2661 if (ordered)
2662 btrfs_put_ordered_extent(ordered);
2663 btrfs_wait_ordered_range(inode, off, len);
2664 }
2665}
2666
416161db
MF		 2667static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2668 struct inode *inode2, u64 loff2, u64 len)
2669{
2670 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2671 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2672
2673 mutex_unlock(&inode1->i_mutex);
2674 mutex_unlock(&inode2->i_mutex);
2675}
2676
2677static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2678 struct inode *inode2, u64 loff2, u64 len)
2679{
2680 if (inode1 < inode2) {
2681 swap(inode1, inode2);
2682 swap(loff1, loff2);
2683 }
2684
2685 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2686 lock_extent_range(inode1, loff1, len);
2687 if (inode1 != inode2) {
2688 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2689 lock_extent_range(inode2, loff2, len);
2690 }
2691}
2692
2693static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2694 u64 dst_loff, u64 len)
2695{
2696 int ret = 0;
2697 struct page *src_page, *dst_page;
2698 unsigned int cmp_len = PAGE_CACHE_SIZE;
2699 void *addr, *dst_addr;
2700
2701 while (len) {
2702 if (len < PAGE_CACHE_SIZE)
2703 cmp_len = len;
2704
2705 src_page = extent_same_get_page(src, loff);
2706 if (!src_page)
2707 return -EINVAL;
2708 dst_page = extent_same_get_page(dst, dst_loff);
2709 if (!dst_page) {
2710 page_cache_release(src_page);
2711 return -EINVAL;
2712 }
2713 addr = kmap_atomic(src_page);
2714 dst_addr = kmap_atomic(dst_page);
2715
2716 flush_dcache_page(src_page);
2717 flush_dcache_page(dst_page);
2718
2719 if (memcmp(addr, dst_addr, cmp_len))
2720 ret = BTRFS_SAME_DATA_DIFFERS;
2721
2722 kunmap_atomic(addr);
2723 kunmap_atomic(dst_addr);
2724 page_cache_release(src_page);
2725 page_cache_release(dst_page);
2726
2727 if (ret)
2728 break;
2729
2730 loff += cmp_len;
2731 dst_loff += cmp_len;
2732 len -= cmp_len;
2733 }
2734
2735 return ret;
2736}
2737
2738static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2739{
2740 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2741
2742 if (off + len > inode->i_size || off + len < off)
2743 return -EINVAL;
2744 /* Check that we are block aligned - btrfs_clone() requires this */
2745 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2746 return -EINVAL;
2747
2748 return 0;
2749}
2750
2751static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2752 struct inode *dst, u64 dst_loff)
2753{
2754 int ret;
2755
2756 /*
2757 * btrfs_clone() can't handle extents in the same file
2758 * yet. Once that works, we can drop this check and replace it
2759 * with a check for the same inode, but overlapping extents.
2760 */
2761 if (src == dst)
2762 return -EINVAL;
2763
2764 btrfs_double_lock(src, loff, dst, dst_loff, len);
2765
2766 ret = extent_same_check_offsets(src, loff, len);
2767 if (ret)
2768 goto out_unlock;
2769
2770 ret = extent_same_check_offsets(dst, dst_loff, len);
2771 if (ret)
2772 goto out_unlock;
2773
2774 /* don't make the dst file partly checksummed */
2775 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2776 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2777 ret = -EINVAL;
2778 goto out_unlock;
2779 }
2780
2781 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2782 if (ret == 0)
2783 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2784
2785out_unlock:
2786 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2787
2788 return ret;
2789}
2790
2791#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2792
2793static long btrfs_ioctl_file_extent_same(struct file *file,
2794 void __user *argp)
2795{
cbf8b8ca
MF		 2796 struct btrfs_ioctl_same_args tmp;
2797 struct btrfs_ioctl_same_args *same;
2798 struct btrfs_ioctl_same_extent_info *info;
416161db
MF		 2799 struct inode *src = file->f_dentry->d_inode;
2800 struct file *dst_file = NULL;
2801 struct inode *dst;
2802 u64 off;
2803 u64 len;
2804 int i;
2805 int ret;
cbf8b8ca 2806 unsigned long size;
416161db
MF		 2807 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2808 bool is_admin = capable(CAP_SYS_ADMIN);
2809
2810 if (!(file->f_mode & FMODE_READ))
2811 return -EINVAL;
2812
2813 ret = mnt_want_write_file(file);
2814 if (ret)
2815 return ret;
2816
cbf8b8ca 2817 if (copy_from_user(&tmp,
416161db 2818 (struct btrfs_ioctl_same_args __user *)argp,
cbf8b8ca 2819 sizeof(tmp))) {
416161db
MF		 2820 ret = -EFAULT;
2821 goto out;
2822 }
2823
cbf8b8ca
MF		 2824 size = sizeof(tmp) +
2825 tmp.dest_count * sizeof(struct btrfs_ioctl_same_extent_info);
2826
229eed43 2827 same = memdup_user((struct btrfs_ioctl_same_args __user *)argp, size);
cbf8b8ca 2828
229eed43
GB		 2829 if (IS_ERR(same)) {
2830 ret = PTR_ERR(same);
cbf8b8ca
MF		 2831 goto out;
2832 }
2833
2834 off = same->logical_offset;
2835 len = same->length;
416161db
MF		 2836
2837 /*
2838 * Limit the total length we will dedupe for each operation.
2839 * This is intended to bound the total time spent in this
2840 * ioctl to something sane.
2841 */
2842 if (len > BTRFS_MAX_DEDUPE_LEN)
2843 len = BTRFS_MAX_DEDUPE_LEN;
2844
2845 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2846 /*
2847 * Btrfs does not support blocksize < page_size. As a
2848 * result, btrfs_cmp_data() won't correctly handle
2849 * this situation without an update.
2850 */
2851 ret = -EINVAL;
2852 goto out;
2853 }
2854
2855 ret = -EISDIR;
2856 if (S_ISDIR(src->i_mode))
2857 goto out;
2858
2859 ret = -EACCES;
2860 if (!S_ISREG(src->i_mode))
2861 goto out;
2862
cbf8b8ca
MF		 2863 /* pre-format output fields to sane values */
2864 for (i = 0; i < same->dest_count; i++) {
2865 same->info[i].bytes_deduped = 0ULL;
2866 same->info[i].status = 0;
2867 }
416161db 2868
cbf8b8ca
MF		 2869 ret = 0;
2870 for (i = 0; i < same->dest_count; i++) {
2871 info = &same->info[i];
416161db 2872
cbf8b8ca 2873 dst_file = fget(info->fd);
416161db 2874 if (!dst_file) {
cbf8b8ca 2875 info->status = -EBADF;
416161db
MF		 2876 goto next;
2877 }
2878
2879 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
cbf8b8ca 2880 info->status = -EINVAL;
416161db
MF		 2881 goto next;
2882 }
2883
cbf8b8ca 2884 info->status = -EXDEV;
416161db
MF		 2885 if (file->f_path.mnt != dst_file->f_path.mnt)
2886 goto next;
2887
2888 dst = dst_file->f_dentry->d_inode;
2889 if (src->i_sb != dst->i_sb)
2890 goto next;
2891
2892 if (S_ISDIR(dst->i_mode)) {
cbf8b8ca 2893 info->status = -EISDIR;
416161db
MF		 2894 goto next;
2895 }
2896
2897 if (!S_ISREG(dst->i_mode)) {
cbf8b8ca 2898 info->status = -EACCES;
416161db
MF		 2899 goto next;
2900 }
2901
cbf8b8ca
MF		 2902 info->status = btrfs_extent_same(src, off, len, dst,
2903 info->logical_offset);
2904 if (info->status == 0)
2905 info->bytes_deduped += len;
416161db
MF		 2906
2907next:
2908 if (dst_file)
2909 fput(dst_file);
416161db
MF		 2910 }
2911
cbf8b8ca
MF		 2912 ret = copy_to_user(argp, same, size);
2913 if (ret)
2914 ret = -EFAULT;
2915
416161db
MF		 2916out:
2917 mnt_drop_write_file(file);
2918 return ret;
2919}
2920
32b7c687
MF		 2921/**
2922 * btrfs_clone() - clone a range from inode file to another
2923 *
2924 * @src: Inode to clone from
2925 * @inode: Inode to clone to
2926 * @off: Offset within source to start clone from
2927 * @olen: Original length, passed by user, of range to clone
2928 * @olen_aligned: Block-aligned value of olen, extent_same uses
2929 * identical values here
2930 * @destoff: Offset within @inode to start clone
2931 */
2932static int btrfs_clone(struct inode *src, struct inode *inode,
2933 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
f46b5a66 2934{
f46b5a66 2935 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 2936 struct btrfs_path *path = NULL;
f46b5a66 2937 struct extent_buffer *leaf;
32b7c687
MF		 2938 struct btrfs_trans_handle *trans;
2939 char *buf = NULL;
ae01a0ab 2940 struct btrfs_key key;
f46b5a66
CH		 2941 u32 nritems;
2942 int slot;
ae01a0ab 2943 int ret;
32b7c687 2944 u64 len = olen_aligned;
ae01a0ab
YZ		 2945
2946 ret = -ENOMEM;
2947 buf = vmalloc(btrfs_level_size(root, 0));
2948 if (!buf)
32b7c687 2949 return ret;
ae01a0ab
YZ		 2950
2951 path = btrfs_alloc_path();
2952 if (!path) {
2953 vfree(buf);
32b7c687 2954 return ret;
d525e8ab
LZ		 2955 }
2956
32b7c687 2957 path->reada = 2;
c5c9cd4d 2958 /* clone data */
33345d01 2959 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2960 key.type = BTRFS_EXTENT_DATA_KEY;
2961 key.offset = 0;
f46b5a66
CH		 2962
2963 while (1) {
2964 /*
2965 * note the key will change type as we walk through the
2966 * tree.
2967 */
e4355f34 2968 path->leave_spinning = 1;
362a20c5
DS		 2969 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2970 0, 0);
f46b5a66
CH		 2971 if (ret < 0)
2972 goto out;
2973
ae01a0ab 2974 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 2975process_slot:
ae01a0ab 2976 if (path->slots[0] >= nritems) {
362a20c5 2977 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 2978 if (ret < 0)
2979 goto out;
2980 if (ret > 0)
2981 break;
ae01a0ab 2982 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2983 }
2984 leaf = path->nodes[0];
2985 slot = path->slots[0];
f46b5a66 2986
ae01a0ab 2987 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2988 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2989 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2990 break;
2991
c5c9cd4d
SW		 2992 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2993 struct btrfs_file_extent_item *extent;
2994 int type;
31840ae1
ZY		 2995 u32 size;
2996 struct btrfs_key new_key;
c5c9cd4d
SW		 2997 u64 disko = 0, diskl = 0;
2998 u64 datao = 0, datal = 0;
2999 u8 comp;
b5384d48 3000 u64 endoff;
31840ae1 3001
c5c9cd4d
SW		 3002 extent = btrfs_item_ptr(leaf, slot,
3003 struct btrfs_file_extent_item);
3004 comp = btrfs_file_extent_compression(leaf, extent);
3005 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3006 if (type == BTRFS_FILE_EXTENT_REG ||
3007 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3008 disko = btrfs_file_extent_disk_bytenr(leaf,
3009 extent);
3010 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3011 extent);
c5c9cd4d 3012 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3013 datal = btrfs_file_extent_num_bytes(leaf,
3014 extent);
c5c9cd4d
SW		 3015 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3016 /* take upper bound, may be compressed */
3017 datal = btrfs_file_extent_ram_bytes(leaf,
3018 extent);
3019 }
31840ae1 3020
050006a7 3021 if (key.offset + datal <= off ||
e4355f34
FDBM		 3022 key.offset >= off + len - 1) {
3023 path->slots[0]++;
3024 goto process_slot;
3025 }
3026
3027 size = btrfs_item_size_nr(leaf, slot);
3028 read_extent_buffer(leaf, buf,
3029 btrfs_item_ptr_offset(leaf, slot),
3030 size);
3031
3032 btrfs_release_path(path);
3033 path->leave_spinning = 0;
c5c9cd4d 3034
31840ae1 3035 memcpy(&new_key, &key, sizeof(new_key));
33345d01 3036 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 3037 if (off <= key.offset)
3038 new_key.offset = key.offset + destoff - off;
3039 else
3040 new_key.offset = destoff;
31840ae1 3041
b6f3409b
SW		 3042 /*
3043 * 1 - adjusting old extent (we may have to split it)
3044 * 1 - add new extent
3045 * 1 - inode update
3046 */
3047 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3048 if (IS_ERR(trans)) {
3049 ret = PTR_ERR(trans);
3050 goto out;
3051 }
3052
c8a894d7
CM		 3053 if (type == BTRFS_FILE_EXTENT_REG ||
3054 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3055 /*
3056 * a | --- range to clone ---| b
3057 * | ------------- extent ------------- |
3058 */
3059
3060 /* substract range b */
3061 if (key.offset + datal > off + len)
3062 datal = off + len - key.offset;
3063
3064 /* substract range a */
a22285a6
YZ		 3065 if (off > key.offset) {
3066 datao += off - key.offset;
3067 datal -= off - key.offset;
3068 }
3069
5dc562c5 3070 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 3071 new_key.offset,
3072 new_key.offset + datal,
2671485d 3073 1);
79787eaa
JM		 3074 if (ret) {
3075 btrfs_abort_transaction(trans, root,
3076 ret);
3077 btrfs_end_transaction(trans, root);
3078 goto out;
3079 }
a22285a6 3080
c5c9cd4d
SW		 3081 ret = btrfs_insert_empty_item(trans, root, path,
3082 &new_key, size);
79787eaa
JM		 3083 if (ret) {
3084 btrfs_abort_transaction(trans, root,
3085 ret);
3086 btrfs_end_transaction(trans, root);
3087 goto out;
3088 }
c5c9cd4d
SW		 3089
3090 leaf = path->nodes[0];
3091 slot = path->slots[0];
3092 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3093 btrfs_item_ptr_offset(leaf, slot),
3094 size);
ae01a0ab 3095
c5c9cd4d 3096 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3097 struct btrfs_file_extent_item);
c5c9cd4d 3098
c5c9cd4d
SW		 3099 /* disko == 0 means it's a hole */
3100 if (!disko)
3101 datao = 0;
c5c9cd4d
SW		 3102
3103 btrfs_set_file_extent_offset(leaf, extent,
3104 datao);
3105 btrfs_set_file_extent_num_bytes(leaf, extent,
3106 datal);
3107 if (disko) {
3108 inode_add_bytes(inode, datal);
ae01a0ab 3109 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3110 disko, diskl, 0,
3111 root->root_key.objectid,
33345d01 3112 btrfs_ino(inode),
66d7e7f0
AJ		 3113 new_key.offset - datao,
3114 0);
79787eaa
JM		 3115 if (ret) {
3116 btrfs_abort_transaction(trans,
3117 root,
3118 ret);
3119 btrfs_end_transaction(trans,
3120 root);
3121 goto out;
3122
3123 }
f46b5a66 3124 }
c5c9cd4d
SW		 3125 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3126 u64 skip = 0;
3127 u64 trim = 0;
3128 if (off > key.offset) {
3129 skip = off - key.offset;
3130 new_key.offset += skip;
3131 }
d397712b 3132
aa42ffd9
LB		 3133 if (key.offset + datal > off + len)
3134 trim = key.offset + datal - (off + len);
d397712b 3135
c5c9cd4d 3136 if (comp && (skip || trim)) {
c5c9cd4d 3137 ret = -EINVAL;
a22285a6 3138 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3139 goto out;
3140 }
3141 size -= skip + trim;
3142 datal -= skip + trim;
a22285a6 3143
5dc562c5 3144 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 3145 new_key.offset,
3146 new_key.offset + datal,
2671485d 3147 1);
79787eaa
JM		 3148 if (ret) {
3149 btrfs_abort_transaction(trans, root,
3150 ret);
3151 btrfs_end_transaction(trans, root);
3152 goto out;
3153 }
a22285a6 3154
c5c9cd4d
SW		 3155 ret = btrfs_insert_empty_item(trans, root, path,
3156 &new_key, size);
79787eaa
JM		 3157 if (ret) {
3158 btrfs_abort_transaction(trans, root,
3159 ret);
3160 btrfs_end_transaction(trans, root);
3161 goto out;
3162 }
c5c9cd4d
SW		 3163
3164 if (skip) {
d397712b
CM		 3165 u32 start =
3166 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3167 memmove(buf+start, buf+start+skip,
3168 datal);
3169 }
3170
3171 leaf = path->nodes[0];
3172 slot = path->slots[0];
3173 write_extent_buffer(leaf, buf,
3174 btrfs_item_ptr_offset(leaf, slot),
3175 size);
3176 inode_add_bytes(inode, datal);
f46b5a66 3177 }
c5c9cd4d
SW		 3178
3179 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3180 btrfs_release_path(path);
c5c9cd4d 3181
0c4d2d95 3182 inode_inc_iversion(inode);
a22285a6 3183 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 3184
3185 /*
3186 * we round up to the block size at eof when
3187 * determining which extents to clone above,
3188 * but shouldn't round up the file size
3189 */
3190 endoff = new_key.offset + datal;
5f3888ff
LZ		 3191 if (endoff > destoff+olen)
3192 endoff = destoff+olen;
b5384d48
SW		 3193 if (endoff > inode->i_size)
3194 btrfs_i_size_write(inode, endoff);
3195
a22285a6 3196 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 3197 if (ret) {
3198 btrfs_abort_transaction(trans, root, ret);
3199 btrfs_end_transaction(trans, root);
3200 goto out;
3201 }
3202 ret = btrfs_end_transaction(trans, root);
a22285a6 3203 }
b3b4aa74 3204 btrfs_release_path(path);
f46b5a66 3205 key.offset++;
f46b5a66 3206 }
f46b5a66 3207 ret = 0;
32b7c687 3208
f46b5a66 3209out:
b3b4aa74 3210 btrfs_release_path(path);
32b7c687
MF		 3211 btrfs_free_path(path);
3212 vfree(buf);
3213 return ret;
3214}
3215
3216static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3217 u64 off, u64 olen, u64 destoff)
3218{
54563d41 3219 struct inode *inode = file_inode(file);
32b7c687
MF		 3220 struct btrfs_root *root = BTRFS_I(inode)->root;
3221 struct fd src_file;
3222 struct inode *src;
3223 int ret;
3224 u64 len = olen;
3225 u64 bs = root->fs_info->sb->s_blocksize;
3226 int same_inode = 0;
3227
3228 /*
3229 * TODO:
3230 * - split compressed inline extents. annoying: we need to
3231 * decompress into destination's address_space (the file offset
3232 * may change, so source mapping won't do), then recompress (or
3233 * otherwise reinsert) a subrange.
3234 * - allow ranges within the same file to be cloned (provided
3235 * they don't overlap)?
3236 */
3237
3238 /* the destination must be opened for writing */
3239 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3240 return -EINVAL;
3241
3242 if (btrfs_root_readonly(root))
3243 return -EROFS;
3244
3245 ret = mnt_want_write_file(file);
3246 if (ret)
3247 return ret;
3248
3249 src_file = fdget(srcfd);
3250 if (!src_file.file) {
3251 ret = -EBADF;
3252 goto out_drop_write;
3253 }
3254
3255 ret = -EXDEV;
3256 if (src_file.file->f_path.mnt != file->f_path.mnt)
3257 goto out_fput;
3258
3259 src = file_inode(src_file.file);
3260
3261 ret = -EINVAL;
3262 if (src == inode)
3263 same_inode = 1;
3264
3265 /* the src must be open for reading */
3266 if (!(src_file.file->f_mode & FMODE_READ))
3267 goto out_fput;
3268
3269 /* don't make the dst file partly checksummed */
3270 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3271 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3272 goto out_fput;
3273
3274 ret = -EISDIR;
3275 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3276 goto out_fput;
3277
3278 ret = -EXDEV;
3279 if (src->i_sb != inode->i_sb)
3280 goto out_fput;
3281
3282 if (!same_inode) {
3283 if (inode < src) {
3284 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3285 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3286 } else {
3287 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3288 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3289 }
3290 } else {
3291 mutex_lock(&src->i_mutex);
3292 }
3293
3294 /* determine range to clone */
3295 ret = -EINVAL;
3296 if (off + len > src->i_size || off + len < off)
3297 goto out_unlock;
3298 if (len == 0)
3299 olen = len = src->i_size - off;
3300 /* if we extend to eof, continue to block boundary */
3301 if (off + len == src->i_size)
3302 len = ALIGN(src->i_size, bs) - off;
3303
3304 /* verify the end result is block aligned */
3305 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3306 !IS_ALIGNED(destoff, bs))
3307 goto out_unlock;
3308
3309 /* verify if ranges are overlapped within the same file */
3310 if (same_inode) {
3311 if (destoff + len > off && destoff < off + len)
3312 goto out_unlock;
3313 }
3314
3315 if (destoff > inode->i_size) {
3316 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3317 if (ret)
3318 goto out_unlock;
3319 }
3320
3321 /* truncate page cache pages from target inode range */
3322 truncate_inode_pages_range(&inode->i_data, destoff,
3323 PAGE_CACHE_ALIGN(destoff + len) - 1);
3324
3325 lock_extent_range(src, off, len);
3326
3327 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3328
aa42ffd9 3329 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
f46b5a66 3330out_unlock:
c57c2b3e
FDBM		 3331 if (!same_inode) {
3332 if (inode < src) {
3333 mutex_unlock(&src->i_mutex);
3334 mutex_unlock(&inode->i_mutex);
3335 } else {
3336 mutex_unlock(&inode->i_mutex);
3337 mutex_unlock(&src->i_mutex);
3338 }
3339 } else {
3340 mutex_unlock(&src->i_mutex);
3341 }
f46b5a66 3342out_fput:
2903ff01 3343 fdput(src_file);
ab67b7c1 3344out_drop_write:
2a79f17e 3345 mnt_drop_write_file(file);
f46b5a66
CH		 3346 return ret;
3347}
3348
7a865e8a 3349static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3350{
3351 struct btrfs_ioctl_clone_range_args args;
3352
7a865e8a 3353 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3354 return -EFAULT;
3355 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3356 args.src_length, args.dest_offset);
3357}
3358
f46b5a66
CH		 3359/*
3360 * there are many ways the trans_start and trans_end ioctls can lead
3361 * to deadlocks. They should only be used by applications that
3362 * basically own the machine, and have a very in depth understanding
3363 * of all the possible deadlocks and enospc problems.
3364 */
b2950863 3365static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3366{
496ad9aa 3367 struct inode *inode = file_inode(file);
f46b5a66
CH		 3368 struct btrfs_root *root = BTRFS_I(inode)->root;
3369 struct btrfs_trans_handle *trans;
1ab86aed 3370 int ret;
f46b5a66 3371
1ab86aed 3372 ret = -EPERM;
df5b5520 3373 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3374 goto out;
df5b5520 3375
1ab86aed
SW		 3376 ret = -EINPROGRESS;
3377 if (file->private_data)
f46b5a66 3378 goto out;
9ca9ee09 3379
b83cc969
LZ		 3380 ret = -EROFS;
3381 if (btrfs_root_readonly(root))
3382 goto out;
3383
a561be71 3384 ret = mnt_want_write_file(file);
c146afad
YZ		 3385 if (ret)
3386 goto out;
3387
a4abeea4 3388 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3389
1ab86aed 3390 ret = -ENOMEM;
7a7eaa40 3391 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3392 if (IS_ERR(trans))
1ab86aed
SW		 3393 goto out_drop;
3394
3395 file->private_data = trans;
3396 return 0;
3397
3398out_drop:
a4abeea4 3399 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3400 mnt_drop_write_file(file);
f46b5a66 3401out:
f46b5a66
CH		 3402 return ret;
3403}
3404
6ef5ed0d
JB		 3405static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3406{
496ad9aa 3407 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3408 struct btrfs_root *root = BTRFS_I(inode)->root;
3409 struct btrfs_root *new_root;
3410 struct btrfs_dir_item *di;
3411 struct btrfs_trans_handle *trans;
3412 struct btrfs_path *path;
3413 struct btrfs_key location;
3414 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3415 u64 objectid = 0;
3416 u64 dir_id;
3c04ce01 3417 int ret;
6ef5ed0d
JB		 3418
3419 if (!capable(CAP_SYS_ADMIN))
3420 return -EPERM;
3421
3c04ce01
MX		 3422 ret = mnt_want_write_file(file);
3423 if (ret)
3424 return ret;
3425
3426 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3427 ret = -EFAULT;
3428 goto out;
3429 }
6ef5ed0d
JB		 3430
3431 if (!objectid)
1cecf579 3432 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3433
3434 location.objectid = objectid;
3435 location.type = BTRFS_ROOT_ITEM_KEY;
3436 location.offset = (u64)-1;
3437
3438 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3439 if (IS_ERR(new_root)) {
3440 ret = PTR_ERR(new_root);
3441 goto out;
3442 }
6ef5ed0d 3443
6ef5ed0d 3444 path = btrfs_alloc_path();
3c04ce01
MX		 3445 if (!path) {
3446 ret = -ENOMEM;
3447 goto out;
3448 }
6ef5ed0d
JB		 3449 path->leave_spinning = 1;
3450
3451 trans = btrfs_start_transaction(root, 1);
98d5dc13 3452 if (IS_ERR(trans)) {
6ef5ed0d 3453 btrfs_free_path(path);
3c04ce01
MX		 3454 ret = PTR_ERR(trans);
3455 goto out;
6ef5ed0d
JB		 3456 }
3457
6c41761f 3458 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3459 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3460 dir_id, "default", 7, 1);
cf1e99a4 3461 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3462 btrfs_free_path(path);
3463 btrfs_end_transaction(trans, root);
efe120a0
FH		 3464 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3465 "item, this isn't going to work");
3c04ce01
MX		 3466 ret = -ENOENT;
3467 goto out;
6ef5ed0d
JB		 3468 }
3469
3470 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3471 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3472 btrfs_mark_buffer_dirty(path->nodes[0]);
3473 btrfs_free_path(path);
3474
2b0ce2c2 3475 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3476 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3477out:
3478 mnt_drop_write_file(file);
3479 return ret;
6ef5ed0d
JB		 3480}
3481
5af3e8cc
SB		 3482void btrfs_get_block_group_info(struct list_head *groups_list,
3483 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3484{
3485 struct btrfs_block_group_cache *block_group;
3486
3487 space->total_bytes = 0;
3488 space->used_bytes = 0;
3489 space->flags = 0;
3490 list_for_each_entry(block_group, groups_list, list) {
3491 space->flags = block_group->flags;
3492 space->total_bytes += block_group->key.offset;
3493 space->used_bytes +=
3494 btrfs_block_group_used(&block_group->item);
3495 }
3496}
3497
48a3b636 3498static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3499{
3500 struct btrfs_ioctl_space_args space_args;
3501 struct btrfs_ioctl_space_info space;
3502 struct btrfs_ioctl_space_info *dest;
7fde62bf 3503 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3504 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3505 struct btrfs_space_info *info;
bf5fc093
JB		 3506 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3507 BTRFS_BLOCK_GROUP_SYSTEM,
3508 BTRFS_BLOCK_GROUP_METADATA,
3509 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3510 int num_types = 4;
7fde62bf 3511 int alloc_size;
1406e432 3512 int ret = 0;
51788b1b 3513 u64 slot_count = 0;
bf5fc093 3514 int i, c;
1406e432
JB		 3515
3516 if (copy_from_user(&space_args,
3517 (struct btrfs_ioctl_space_args __user *)arg,
3518 sizeof(space_args)))
3519 return -EFAULT;
3520
bf5fc093
JB		 3521 for (i = 0; i < num_types; i++) {
3522 struct btrfs_space_info *tmp;
3523
3524 info = NULL;
3525 rcu_read_lock();
3526 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3527 list) {
3528 if (tmp->flags == types[i]) {
3529 info = tmp;
3530 break;
3531 }
3532 }
3533 rcu_read_unlock();
3534
3535 if (!info)
3536 continue;
3537
3538 down_read(&info->groups_sem);
3539 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3540 if (!list_empty(&info->block_groups[c]))
3541 slot_count++;
3542 }
3543 up_read(&info->groups_sem);
3544 }
7fde62bf
CM		 3545
3546 /* space_slots == 0 means they are asking for a count */
3547 if (space_args.space_slots == 0) {
3548 space_args.total_spaces = slot_count;
3549 goto out;
3550 }
bf5fc093 3551
51788b1b 3552 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3553
7fde62bf 3554 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3555
7fde62bf
CM		 3556 /* we generally have at most 6 or so space infos, one for each raid
3557 * level. So, a whole page should be more than enough for everyone
3558 */
3559 if (alloc_size > PAGE_CACHE_SIZE)
3560 return -ENOMEM;
3561
1406e432 3562 space_args.total_spaces = 0;
7fde62bf
CM		 3563 dest = kmalloc(alloc_size, GFP_NOFS);
3564 if (!dest)
3565 return -ENOMEM;
3566 dest_orig = dest;
1406e432 3567
7fde62bf 3568 /* now we have a buffer to copy into */
bf5fc093
JB		 3569 for (i = 0; i < num_types; i++) {
3570 struct btrfs_space_info *tmp;
3571
51788b1b
DR		 3572 if (!slot_count)
3573 break;
3574
bf5fc093
JB		 3575 info = NULL;
3576 rcu_read_lock();
3577 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3578 list) {
3579 if (tmp->flags == types[i]) {
3580 info = tmp;
3581 break;
3582 }
3583 }
3584 rcu_read_unlock();
7fde62bf 3585
bf5fc093
JB		 3586 if (!info)
3587 continue;
3588 down_read(&info->groups_sem);
3589 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3590 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3591 btrfs_get_block_group_info(
3592 &info->block_groups[c], &space);
bf5fc093
JB		 3593 memcpy(dest, &space, sizeof(space));
3594 dest++;
3595 space_args.total_spaces++;
51788b1b 3596 slot_count--;
bf5fc093 3597 }
51788b1b
DR		 3598 if (!slot_count)
3599 break;
bf5fc093
JB		 3600 }
3601 up_read(&info->groups_sem);
1406e432 3602 }
1406e432 3603
2eec6c81 3604 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3605 (arg + sizeof(struct btrfs_ioctl_space_args));
3606
3607 if (copy_to_user(user_dest, dest_orig, alloc_size))
3608 ret = -EFAULT;
3609
3610 kfree(dest_orig);
3611out:
3612 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 3613 ret = -EFAULT;
3614
3615 return ret;
3616}
3617
f46b5a66
CH		 3618/*
3619 * there are many ways the trans_start and trans_end ioctls can lead
3620 * to deadlocks. They should only be used by applications that
3621 * basically own the machine, and have a very in depth understanding
3622 * of all the possible deadlocks and enospc problems.
3623 */
3624long btrfs_ioctl_trans_end(struct file *file)
3625{
496ad9aa 3626 struct inode *inode = file_inode(file);
f46b5a66
CH		 3627 struct btrfs_root *root = BTRFS_I(inode)->root;
3628 struct btrfs_trans_handle *trans;
f46b5a66 3629
f46b5a66 3630 trans = file->private_data;
1ab86aed
SW		 3631 if (!trans)
3632 return -EINVAL;
b214107e 3633 file->private_data = NULL;
9ca9ee09 3634
1ab86aed
SW		 3635 btrfs_end_transaction(trans, root);
3636
a4abeea4 3637 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 3638
2a79f17e 3639 mnt_drop_write_file(file);
1ab86aed 3640 return 0;
f46b5a66
CH		 3641}
3642
9a8c28be
MX		 3643static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3644 void __user *argp)
46204592 3645{
46204592
SW		 3646 struct btrfs_trans_handle *trans;
3647 u64 transid;
db5b493a 3648 int ret;
46204592 3649
d4edf39b 3650 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 3651 if (IS_ERR(trans)) {
3652 if (PTR_ERR(trans) != -ENOENT)
3653 return PTR_ERR(trans);
3654
3655 /* No running transaction, don't bother */
3656 transid = root->fs_info->last_trans_committed;
3657 goto out;
3658 }
46204592 3659 transid = trans->transid;
db5b493a 3660 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 3661 if (ret) {
3662 btrfs_end_transaction(trans, root);
db5b493a 3663 return ret;
8b2b2d3c 3664 }
ff7c1d33 3665out:
46204592
SW		 3666 if (argp)
3667 if (copy_to_user(argp, &transid, sizeof(transid)))
3668 return -EFAULT;
3669 return 0;
3670}
3671
9a8c28be
MX		 3672static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3673 void __user *argp)
46204592 3674{
46204592
SW		 3675 u64 transid;
3676
3677 if (argp) {
3678 if (copy_from_user(&transid, argp, sizeof(transid)))
3679 return -EFAULT;
3680 } else {
3681 transid = 0; /* current trans */
3682 }
3683 return btrfs_wait_for_commit(root, transid);
3684}
3685
b8e95489 3686static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 3687{
496ad9aa 3688 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 3689 struct btrfs_ioctl_scrub_args *sa;
b8e95489 3690 int ret;
475f6387
JS		 3691
3692 if (!capable(CAP_SYS_ADMIN))
3693 return -EPERM;
3694
3695 sa = memdup_user(arg, sizeof(*sa));
3696 if (IS_ERR(sa))
3697 return PTR_ERR(sa);
3698
b8e95489
MX		 3699 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3700 ret = mnt_want_write_file(file);
3701 if (ret)
3702 goto out;
3703 }
3704
aa1b8cd4 3705 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 3706 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3707 0);
475f6387
JS		 3708
3709 if (copy_to_user(arg, sa, sizeof(*sa)))
3710 ret = -EFAULT;
3711
b8e95489
MX		 3712 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3713 mnt_drop_write_file(file);
3714out:
475f6387
JS		 3715 kfree(sa);
3716 return ret;
3717}
3718
3719static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3720{
3721 if (!capable(CAP_SYS_ADMIN))
3722 return -EPERM;
3723
aa1b8cd4 3724 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 3725}
3726
3727static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3728 void __user *arg)
3729{
3730 struct btrfs_ioctl_scrub_args *sa;
3731 int ret;
3732
3733 if (!capable(CAP_SYS_ADMIN))
3734 return -EPERM;
3735
3736 sa = memdup_user(arg, sizeof(*sa));
3737 if (IS_ERR(sa))
3738 return PTR_ERR(sa);
3739
3740 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3741
3742 if (copy_to_user(arg, sa, sizeof(*sa)))
3743 ret = -EFAULT;
3744
3745 kfree(sa);
3746 return ret;
3747}
3748
c11d2c23 3749static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 3750 void __user *arg)
c11d2c23
SB		 3751{
3752 struct btrfs_ioctl_get_dev_stats *sa;
3753 int ret;
3754
c11d2c23
SB		 3755 sa = memdup_user(arg, sizeof(*sa));
3756 if (IS_ERR(sa))
3757 return PTR_ERR(sa);
3758
b27f7c0c
DS		 3759 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3760 kfree(sa);
3761 return -EPERM;
3762 }
3763
3764 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 3765
3766 if (copy_to_user(arg, sa, sizeof(*sa)))
3767 ret = -EFAULT;
3768
3769 kfree(sa);
3770 return ret;
3771}
3772
3f6bcfbd
SB		 3773static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3774{
3775 struct btrfs_ioctl_dev_replace_args *p;
3776 int ret;
3777
3778 if (!capable(CAP_SYS_ADMIN))
3779 return -EPERM;
3780
3781 p = memdup_user(arg, sizeof(*p));
3782 if (IS_ERR(p))
3783 return PTR_ERR(p);
3784
3785 switch (p->cmd) {
3786 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
adfa97cb
ID		 3787 if (root->fs_info->sb->s_flags & MS_RDONLY) {
3788 ret = -EROFS;
3789 goto out;
3790 }
3f6bcfbd
SB		 3791 if (atomic_xchg(
3792 &root->fs_info->mutually_exclusive_operation_running,
3793 1)) {
e57138b3 3794 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd
SB		 3795 } else {
3796 ret = btrfs_dev_replace_start(root, p);
3797 atomic_set(
3798 &root->fs_info->mutually_exclusive_operation_running,
3799 0);
3800 }
3801 break;
3802 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3803 btrfs_dev_replace_status(root->fs_info, p);
3804 ret = 0;
3805 break;
3806 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3807 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3808 break;
3809 default:
3810 ret = -EINVAL;
3811 break;
3812 }
3813
3814 if (copy_to_user(arg, p, sizeof(*p)))
3815 ret = -EFAULT;
adfa97cb 3816out:
3f6bcfbd
SB		 3817 kfree(p);
3818 return ret;
3819}
3820
d7728c96
JS		 3821static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3822{
3823 int ret = 0;
3824 int i;
740c3d22 3825 u64 rel_ptr;
d7728c96 3826 int size;
806468f8 3827 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3828 struct inode_fs_paths *ipath = NULL;
3829 struct btrfs_path *path;
3830
82b22ac8 3831 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 3832 return -EPERM;
3833
3834 path = btrfs_alloc_path();
3835 if (!path) {
3836 ret = -ENOMEM;
3837 goto out;
3838 }
3839
3840 ipa = memdup_user(arg, sizeof(*ipa));
3841 if (IS_ERR(ipa)) {
3842 ret = PTR_ERR(ipa);
3843 ipa = NULL;
3844 goto out;
3845 }
3846
3847 size = min_t(u32, ipa->size, 4096);
3848 ipath = init_ipath(size, root, path);
3849 if (IS_ERR(ipath)) {
3850 ret = PTR_ERR(ipath);
3851 ipath = NULL;
3852 goto out;
3853 }
3854
3855 ret = paths_from_inode(ipa->inum, ipath);
3856 if (ret < 0)
3857 goto out;
3858
3859 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3860 rel_ptr = ipath->fspath->val[i] -
3861 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3862 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3863 }
3864
745c4d8e
JM		 3865 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3866 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3867 if (ret) {
3868 ret = -EFAULT;
3869 goto out;
3870 }
3871
3872out:
3873 btrfs_free_path(path);
3874 free_ipath(ipath);
3875 kfree(ipa);
3876
3877 return ret;
3878}
3879
3880static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3881{
3882 struct btrfs_data_container *inodes = ctx;
3883 const size_t c = 3 * sizeof(u64);
3884
3885 if (inodes->bytes_left >= c) {
3886 inodes->bytes_left -= c;
3887 inodes->val[inodes->elem_cnt] = inum;
3888 inodes->val[inodes->elem_cnt + 1] = offset;
3889 inodes->val[inodes->elem_cnt + 2] = root;
3890 inodes->elem_cnt += 3;
3891 } else {
3892 inodes->bytes_missing += c - inodes->bytes_left;
3893 inodes->bytes_left = 0;
3894 inodes->elem_missed += 3;
3895 }
3896
3897 return 0;
3898}
3899
3900static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3901 void __user *arg)
3902{
3903 int ret = 0;
3904 int size;
d7728c96
JS		 3905 struct btrfs_ioctl_logical_ino_args *loi;
3906 struct btrfs_data_container *inodes = NULL;
3907 struct btrfs_path *path = NULL;
d7728c96
JS		 3908
3909 if (!capable(CAP_SYS_ADMIN))
3910 return -EPERM;
3911
3912 loi = memdup_user(arg, sizeof(*loi));
3913 if (IS_ERR(loi)) {
3914 ret = PTR_ERR(loi);
3915 loi = NULL;
3916 goto out;
3917 }
3918
3919 path = btrfs_alloc_path();
3920 if (!path) {
3921 ret = -ENOMEM;
3922 goto out;
3923 }
3924
425d17a2 3925 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 3926 inodes = init_data_container(size);
3927 if (IS_ERR(inodes)) {
3928 ret = PTR_ERR(inodes);
3929 inodes = NULL;
3930 goto out;
3931 }
3932
df031f07
LB		 3933 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3934 build_ino_list, inodes);
3935 if (ret == -EINVAL)
d7728c96
JS		 3936 ret = -ENOENT;
3937 if (ret < 0)
3938 goto out;
3939
745c4d8e
JM		 3940 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3941 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3942 if (ret)
3943 ret = -EFAULT;
3944
3945out:
3946 btrfs_free_path(path);
425d17a2 3947 vfree(inodes);
d7728c96
JS		 3948 kfree(loi);
3949
3950 return ret;
3951}
3952
19a39dce 3953void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3954 struct btrfs_ioctl_balance_args *bargs)
3955{
3956 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3957
3958 bargs->flags = bctl->flags;
3959
837d5b6e
ID		 3960 if (atomic_read(&fs_info->balance_running))
3961 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3962 if (atomic_read(&fs_info->balance_pause_req))
3963 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3964 if (atomic_read(&fs_info->balance_cancel_req))
3965 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3966
c9e9f97b
ID		 3967 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3968 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3969 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3970
3971 if (lock) {
3972 spin_lock(&fs_info->balance_lock);
3973 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3974 spin_unlock(&fs_info->balance_lock);
3975 } else {
3976 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3977 }
c9e9f97b
ID		 3978}
3979
9ba1f6e4 3980static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3981{
496ad9aa 3982 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 3983 struct btrfs_fs_info *fs_info = root->fs_info;
3984 struct btrfs_ioctl_balance_args *bargs;
3985 struct btrfs_balance_control *bctl;
ed0fb78f 3986 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 3987 int ret;
3988
3989 if (!capable(CAP_SYS_ADMIN))
3990 return -EPERM;
3991
e54bfa31 3992 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 3993 if (ret)
3994 return ret;
3995
ed0fb78f
ID		 3996again:
3997 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3998 mutex_lock(&fs_info->volume_mutex);
3999 mutex_lock(&fs_info->balance_mutex);
4000 need_unlock = true;
4001 goto locked;
4002 }
4003
4004 /*
4005 * mut. excl. ops lock is locked. Three possibilites:
4006 * (1) some other op is running
4007 * (2) balance is running
4008 * (3) balance is paused -- special case (think resume)
4009 */
c9e9f97b 4010 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4011 if (fs_info->balance_ctl) {
4012 /* this is either (2) or (3) */
4013 if (!atomic_read(&fs_info->balance_running)) {
4014 mutex_unlock(&fs_info->balance_mutex);
4015 if (!mutex_trylock(&fs_info->volume_mutex))
4016 goto again;
4017 mutex_lock(&fs_info->balance_mutex);
4018
4019 if (fs_info->balance_ctl &&
4020 !atomic_read(&fs_info->balance_running)) {
4021 /* this is (3) */
4022 need_unlock = false;
4023 goto locked;
4024 }
4025
4026 mutex_unlock(&fs_info->balance_mutex);
4027 mutex_unlock(&fs_info->volume_mutex);
4028 goto again;
4029 } else {
4030 /* this is (2) */
4031 mutex_unlock(&fs_info->balance_mutex);
4032 ret = -EINPROGRESS;
4033 goto out;
4034 }
4035 } else {
4036 /* this is (1) */
4037 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4038 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4039 goto out;
4040 }
4041
4042locked:
4043 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 4044
4045 if (arg) {
4046 bargs = memdup_user(arg, sizeof(*bargs));
4047 if (IS_ERR(bargs)) {
4048 ret = PTR_ERR(bargs);
ed0fb78f 4049 goto out_unlock;
c9e9f97b 4050 }
de322263
ID		 4051
4052 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4053 if (!fs_info->balance_ctl) {
4054 ret = -ENOTCONN;
4055 goto out_bargs;
4056 }
4057
4058 bctl = fs_info->balance_ctl;
4059 spin_lock(&fs_info->balance_lock);
4060 bctl->flags |= BTRFS_BALANCE_RESUME;
4061 spin_unlock(&fs_info->balance_lock);
4062
4063 goto do_balance;
4064 }
c9e9f97b
ID		 4065 } else {
4066 bargs = NULL;
4067 }
4068
ed0fb78f 4069 if (fs_info->balance_ctl) {
837d5b6e
ID		 4070 ret = -EINPROGRESS;
4071 goto out_bargs;
4072 }
4073
c9e9f97b
ID		 4074 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4075 if (!bctl) {
4076 ret = -ENOMEM;
4077 goto out_bargs;
4078 }
4079
4080 bctl->fs_info = fs_info;
4081 if (arg) {
4082 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4083 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4084 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4085
4086 bctl->flags = bargs->flags;
f43ffb60
ID		 4087 } else {
4088 /* balance everything - no filters */
4089 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4090 }
4091
de322263 4092do_balance:
c9e9f97b 4093 /*
ed0fb78f
ID		 4094 * Ownership of bctl and mutually_exclusive_operation_running
4095 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4096 * or, if restriper was paused all the way until unmount, in
4097 * free_fs_info. mutually_exclusive_operation_running is
4098 * cleared in __cancel_balance.
c9e9f97b 4099 */
ed0fb78f
ID		 4100 need_unlock = false;
4101
4102 ret = btrfs_balance(bctl, bargs);
4103
c9e9f97b
ID		 4104 if (arg) {
4105 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4106 ret = -EFAULT;
4107 }
4108
4109out_bargs:
4110 kfree(bargs);
ed0fb78f 4111out_unlock:
c9e9f97b
ID		 4112 mutex_unlock(&fs_info->balance_mutex);
4113 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 4114 if (need_unlock)
4115 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4116out:
e54bfa31 4117 mnt_drop_write_file(file);
c9e9f97b
ID		 4118 return ret;
4119}
4120
837d5b6e
ID		 4121static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4122{
4123 if (!capable(CAP_SYS_ADMIN))
4124 return -EPERM;
4125
4126 switch (cmd) {
4127 case BTRFS_BALANCE_CTL_PAUSE:
4128 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4129 case BTRFS_BALANCE_CTL_CANCEL:
4130 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4131 }
4132
4133 return -EINVAL;
4134}
4135
19a39dce
ID		 4136static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4137 void __user *arg)
4138{
4139 struct btrfs_fs_info *fs_info = root->fs_info;
4140 struct btrfs_ioctl_balance_args *bargs;
4141 int ret = 0;
4142
4143 if (!capable(CAP_SYS_ADMIN))
4144 return -EPERM;
4145
4146 mutex_lock(&fs_info->balance_mutex);
4147 if (!fs_info->balance_ctl) {
4148 ret = -ENOTCONN;
4149 goto out;
4150 }
4151
4152 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4153 if (!bargs) {
4154 ret = -ENOMEM;
4155 goto out;
4156 }
4157
4158 update_ioctl_balance_args(fs_info, 1, bargs);
4159
4160 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4161 ret = -EFAULT;
4162
4163 kfree(bargs);
4164out:
4165 mutex_unlock(&fs_info->balance_mutex);
4166 return ret;
4167}
4168
905b0dda 4169static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4170{
496ad9aa 4171 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4172 struct btrfs_ioctl_quota_ctl_args *sa;
4173 struct btrfs_trans_handle *trans = NULL;
4174 int ret;
4175 int err;
4176
4177 if (!capable(CAP_SYS_ADMIN))
4178 return -EPERM;
4179
905b0dda
MX		 4180 ret = mnt_want_write_file(file);
4181 if (ret)
4182 return ret;
5d13a37b
AJ		 4183
4184 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4185 if (IS_ERR(sa)) {
4186 ret = PTR_ERR(sa);
4187 goto drop_write;
4188 }
5d13a37b 4189
7708f029 4190 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4191 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4192 if (IS_ERR(trans)) {
4193 ret = PTR_ERR(trans);
4194 goto out;
5d13a37b
AJ		 4195 }
4196
4197 switch (sa->cmd) {
4198 case BTRFS_QUOTA_CTL_ENABLE:
4199 ret = btrfs_quota_enable(trans, root->fs_info);
4200 break;
4201 case BTRFS_QUOTA_CTL_DISABLE:
4202 ret = btrfs_quota_disable(trans, root->fs_info);
4203 break;
5d13a37b
AJ		 4204 default:
4205 ret = -EINVAL;
4206 break;
4207 }
4208
2f232036
JS		 4209 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4210 if (err && !ret)
4211 ret = err;
5d13a37b
AJ		 4212out:
4213 kfree(sa);
7708f029 4214 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4215drop_write:
4216 mnt_drop_write_file(file);
5d13a37b
AJ		 4217 return ret;
4218}
4219
905b0dda 4220static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4221{
496ad9aa 4222 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4223 struct btrfs_ioctl_qgroup_assign_args *sa;
4224 struct btrfs_trans_handle *trans;
4225 int ret;
4226 int err;
4227
4228 if (!capable(CAP_SYS_ADMIN))
4229 return -EPERM;
4230
905b0dda
MX		 4231 ret = mnt_want_write_file(file);
4232 if (ret)
4233 return ret;
5d13a37b
AJ		 4234
4235 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4236 if (IS_ERR(sa)) {
4237 ret = PTR_ERR(sa);
4238 goto drop_write;
4239 }
5d13a37b
AJ		 4240
4241 trans = btrfs_join_transaction(root);
4242 if (IS_ERR(trans)) {
4243 ret = PTR_ERR(trans);
4244 goto out;
4245 }
4246
4247 /* FIXME: check if the IDs really exist */
4248 if (sa->assign) {
4249 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4250 sa->src, sa->dst);
4251 } else {
4252 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4253 sa->src, sa->dst);
4254 }
4255
4256 err = btrfs_end_transaction(trans, root);
4257 if (err && !ret)
4258 ret = err;
4259
4260out:
4261 kfree(sa);
905b0dda
MX		 4262drop_write:
4263 mnt_drop_write_file(file);
5d13a37b
AJ		 4264 return ret;
4265}
4266
905b0dda 4267static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4268{
496ad9aa 4269 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4270 struct btrfs_ioctl_qgroup_create_args *sa;
4271 struct btrfs_trans_handle *trans;
4272 int ret;
4273 int err;
4274
4275 if (!capable(CAP_SYS_ADMIN))
4276 return -EPERM;
4277
905b0dda
MX		 4278 ret = mnt_want_write_file(file);
4279 if (ret)
4280 return ret;
5d13a37b
AJ		 4281
4282 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4283 if (IS_ERR(sa)) {
4284 ret = PTR_ERR(sa);
4285 goto drop_write;
4286 }
5d13a37b 4287
d86e56cf
MX		 4288 if (!sa->qgroupid) {
4289 ret = -EINVAL;
4290 goto out;
4291 }
4292
5d13a37b
AJ		 4293 trans = btrfs_join_transaction(root);
4294 if (IS_ERR(trans)) {
4295 ret = PTR_ERR(trans);
4296 goto out;
4297 }
4298
4299 /* FIXME: check if the IDs really exist */
4300 if (sa->create) {
4301 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4302 NULL);
4303 } else {
4304 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4305 }
4306
4307 err = btrfs_end_transaction(trans, root);
4308 if (err && !ret)
4309 ret = err;
4310
4311out:
4312 kfree(sa);
905b0dda
MX		 4313drop_write:
4314 mnt_drop_write_file(file);
5d13a37b
AJ		 4315 return ret;
4316}
4317
905b0dda 4318static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4319{
496ad9aa 4320 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4321 struct btrfs_ioctl_qgroup_limit_args *sa;
4322 struct btrfs_trans_handle *trans;
4323 int ret;
4324 int err;
4325 u64 qgroupid;
4326
4327 if (!capable(CAP_SYS_ADMIN))
4328 return -EPERM;
4329
905b0dda
MX		 4330 ret = mnt_want_write_file(file);
4331 if (ret)
4332 return ret;
5d13a37b
AJ		 4333
4334 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4335 if (IS_ERR(sa)) {
4336 ret = PTR_ERR(sa);
4337 goto drop_write;
4338 }
5d13a37b
AJ		 4339
4340 trans = btrfs_join_transaction(root);
4341 if (IS_ERR(trans)) {
4342 ret = PTR_ERR(trans);
4343 goto out;
4344 }
4345
4346 qgroupid = sa->qgroupid;
4347 if (!qgroupid) {
4348 /* take the current subvol as qgroup */
4349 qgroupid = root->root_key.objectid;
4350 }
4351
4352 /* FIXME: check if the IDs really exist */
4353 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4354
4355 err = btrfs_end_transaction(trans, root);
4356 if (err && !ret)
4357 ret = err;
4358
4359out:
4360 kfree(sa);
905b0dda
MX		 4361drop_write:
4362 mnt_drop_write_file(file);
5d13a37b
AJ		 4363 return ret;
4364}
4365
2f232036
JS		 4366static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4367{
6d0379ec 4368 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4369 struct btrfs_ioctl_quota_rescan_args *qsa;
4370 int ret;
4371
4372 if (!capable(CAP_SYS_ADMIN))
4373 return -EPERM;
4374
4375 ret = mnt_want_write_file(file);
4376 if (ret)
4377 return ret;
4378
4379 qsa = memdup_user(arg, sizeof(*qsa));
4380 if (IS_ERR(qsa)) {
4381 ret = PTR_ERR(qsa);
4382 goto drop_write;
4383 }
4384
4385 if (qsa->flags) {
4386 ret = -EINVAL;
4387 goto out;
4388 }
4389
4390 ret = btrfs_qgroup_rescan(root->fs_info);
4391
4392out:
4393 kfree(qsa);
4394drop_write:
4395 mnt_drop_write_file(file);
4396 return ret;
4397}
4398
4399static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4400{
6d0379ec 4401 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4402 struct btrfs_ioctl_quota_rescan_args *qsa;
4403 int ret = 0;
4404
4405 if (!capable(CAP_SYS_ADMIN))
4406 return -EPERM;
4407
4408 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4409 if (!qsa)
4410 return -ENOMEM;
4411
4412 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4413 qsa->flags = 1;
4414 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4415 }
4416
4417 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4418 ret = -EFAULT;
4419
4420 kfree(qsa);
4421 return ret;
4422}
4423
57254b6e
JS		 4424static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4425{
54563d41 4426 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
57254b6e
JS		 4427
4428 if (!capable(CAP_SYS_ADMIN))
4429 return -EPERM;
4430
4431 return btrfs_qgroup_wait_for_completion(root->fs_info);
4432}
4433
abccd00f
HM		 4434static long _btrfs_ioctl_set_received_subvol(struct file *file,
4435 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4436{
496ad9aa 4437 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4438 struct btrfs_root *root = BTRFS_I(inode)->root;
4439 struct btrfs_root_item *root_item = &root->root_item;
4440 struct btrfs_trans_handle *trans;
4441 struct timespec ct = CURRENT_TIME;
4442 int ret = 0;
dd5f9615 4443 int received_uuid_changed;
8ea05e3a 4444
bd60ea0f
DS		 4445 if (!inode_owner_or_capable(inode))
4446 return -EPERM;
4447
8ea05e3a
AB		 4448 ret = mnt_want_write_file(file);
4449 if (ret < 0)
4450 return ret;
4451
4452 down_write(&root->fs_info->subvol_sem);
4453
4454 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4455 ret = -EINVAL;
4456 goto out;
4457 }
4458
4459 if (btrfs_root_readonly(root)) {
4460 ret = -EROFS;
4461 goto out;
4462 }
4463
dd5f9615
SB		 4464 /*
4465 * 1 - root item
4466 * 2 - uuid items (received uuid + subvol uuid)
4467 */
4468 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4469 if (IS_ERR(trans)) {
4470 ret = PTR_ERR(trans);
4471 trans = NULL;
4472 goto out;
4473 }
4474
4475 sa->rtransid = trans->transid;
4476 sa->rtime.sec = ct.tv_sec;
4477 sa->rtime.nsec = ct.tv_nsec;
4478
dd5f9615
SB		 4479 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4480 BTRFS_UUID_SIZE);
4481 if (received_uuid_changed &&
4482 !btrfs_is_empty_uuid(root_item->received_uuid))
4483 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4484 root_item->received_uuid,
4485 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4486 root->root_key.objectid);
8ea05e3a
AB		 4487 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4488 btrfs_set_root_stransid(root_item, sa->stransid);
4489 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4490 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4491 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4492 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4493 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4494
4495 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4496 &root->root_key, &root->root_item);
4497 if (ret < 0) {
4498 btrfs_end_transaction(trans, root);
8ea05e3a 4499 goto out;
dd5f9615
SB		 4500 }
4501 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4502 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4503 sa->uuid,
4504 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4505 root->root_key.objectid);
4506 if (ret < 0 && ret != -EEXIST) {
4507 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 4508 goto out;
dd5f9615
SB		 4509 }
4510 }
4511 ret = btrfs_commit_transaction(trans, root);
4512 if (ret < 0) {
4513 btrfs_abort_transaction(trans, root, ret);
4514 goto out;
8ea05e3a
AB		 4515 }
4516
abccd00f
HM		 4517out:
4518 up_write(&root->fs_info->subvol_sem);
4519 mnt_drop_write_file(file);
4520 return ret;
4521}
4522
4523#ifdef CONFIG_64BIT
4524static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4525 void __user *arg)
4526{
4527 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4528 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4529 int ret = 0;
4530
4531 args32 = memdup_user(arg, sizeof(*args32));
4532 if (IS_ERR(args32)) {
4533 ret = PTR_ERR(args32);
4534 args32 = NULL;
4535 goto out;
4536 }
4537
4538 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4539 if (IS_ERR(args64)) {
4540 ret = PTR_ERR(args64);
4541 args64 = NULL;
4542 goto out;
4543 }
4544
4545 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4546 args64->stransid = args32->stransid;
4547 args64->rtransid = args32->rtransid;
4548 args64->stime.sec = args32->stime.sec;
4549 args64->stime.nsec = args32->stime.nsec;
4550 args64->rtime.sec = args32->rtime.sec;
4551 args64->rtime.nsec = args32->rtime.nsec;
4552 args64->flags = args32->flags;
4553
4554 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4555 if (ret)
4556 goto out;
4557
4558 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4559 args32->stransid = args64->stransid;
4560 args32->rtransid = args64->rtransid;
4561 args32->stime.sec = args64->stime.sec;
4562 args32->stime.nsec = args64->stime.nsec;
4563 args32->rtime.sec = args64->rtime.sec;
4564 args32->rtime.nsec = args64->rtime.nsec;
4565 args32->flags = args64->flags;
4566
4567 ret = copy_to_user(arg, args32, sizeof(*args32));
4568 if (ret)
4569 ret = -EFAULT;
4570
4571out:
4572 kfree(args32);
4573 kfree(args64);
4574 return ret;
4575}
4576#endif
4577
4578static long btrfs_ioctl_set_received_subvol(struct file *file,
4579 void __user *arg)
4580{
4581 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4582 int ret = 0;
4583
4584 sa = memdup_user(arg, sizeof(*sa));
4585 if (IS_ERR(sa)) {
4586 ret = PTR_ERR(sa);
4587 sa = NULL;
4588 goto out;
4589 }
4590
4591 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4592
4593 if (ret)
4594 goto out;
4595
8ea05e3a
AB		 4596 ret = copy_to_user(arg, sa, sizeof(*sa));
4597 if (ret)
4598 ret = -EFAULT;
4599
4600out:
4601 kfree(sa);
8ea05e3a
AB		 4602 return ret;
4603}
4604
867ab667 4605static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4606{
6d0379ec 4607 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 4608 size_t len;
867ab667 4609 int ret;
a1b83ac5
AJ		 4610 char label[BTRFS_LABEL_SIZE];
4611
4612 spin_lock(&root->fs_info->super_lock);
4613 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4614 spin_unlock(&root->fs_info->super_lock);
4615
4616 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 4617
4618 if (len == BTRFS_LABEL_SIZE) {
efe120a0
FH		 4619 btrfs_warn(root->fs_info,
4620 "label is too long, return the first %zu bytes", --len);
867ab667 4621 }
4622
867ab667 4623 ret = copy_to_user(arg, label, len);
867ab667 4624
4625 return ret ? -EFAULT : 0;
4626}
4627
a8bfd4ab 4628static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4629{
6d0379ec 4630 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 4631 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4632 struct btrfs_trans_handle *trans;
4633 char label[BTRFS_LABEL_SIZE];
4634 int ret;
4635
4636 if (!capable(CAP_SYS_ADMIN))
4637 return -EPERM;
4638
4639 if (copy_from_user(label, arg, sizeof(label)))
4640 return -EFAULT;
4641
4642 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
efe120a0 4643 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
a8bfd4ab 4644 BTRFS_LABEL_SIZE - 1);
4645 return -EINVAL;
4646 }
4647
4648 ret = mnt_want_write_file(file);
4649 if (ret)
4650 return ret;
4651
a8bfd4ab 4652 trans = btrfs_start_transaction(root, 0);
4653 if (IS_ERR(trans)) {
4654 ret = PTR_ERR(trans);
4655 goto out_unlock;
4656 }
4657
a1b83ac5 4658 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 4659 strcpy(super_block->label, label);
a1b83ac5 4660 spin_unlock(&root->fs_info->super_lock);
d0270aca 4661 ret = btrfs_commit_transaction(trans, root);
a8bfd4ab 4662
4663out_unlock:
a8bfd4ab 4664 mnt_drop_write_file(file);
4665 return ret;
4666}
4667
2eaa055f
JM		 4668#define INIT_FEATURE_FLAGS(suffix) \
4669 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4670 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4671 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4672
4673static int btrfs_ioctl_get_supported_features(struct file *file,
4674 void __user *arg)
4675{
4676 static struct btrfs_ioctl_feature_flags features[3] = {
4677 INIT_FEATURE_FLAGS(SUPP),
4678 INIT_FEATURE_FLAGS(SAFE_SET),
4679 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4680 };
4681
4682 if (copy_to_user(arg, &features, sizeof(features)))
4683 return -EFAULT;
4684
4685 return 0;
4686}
4687
4688static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
4689{
4690 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4691 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4692 struct btrfs_ioctl_feature_flags features;
4693
4694 features.compat_flags = btrfs_super_compat_flags(super_block);
4695 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4696 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4697
4698 if (copy_to_user(arg, &features, sizeof(features)))
4699 return -EFAULT;
4700
4701 return 0;
4702}
4703
3b02a68a
JM		 4704static int check_feature_bits(struct btrfs_root *root,
4705 enum btrfs_feature_set set,
2eaa055f
JM		 4706 u64 change_mask, u64 flags, u64 supported_flags,
4707 u64 safe_set, u64 safe_clear)
4708{
3b02a68a
JM		 4709 const char *type = btrfs_feature_set_names[set];
4710 char *names;
2eaa055f
JM		 4711 u64 disallowed, unsupported;
4712 u64 set_mask = flags & change_mask;
4713 u64 clear_mask = ~flags & change_mask;
4714
4715 unsupported = set_mask & ~supported_flags;
4716 if (unsupported) {
3b02a68a
JM		 4717 names = btrfs_printable_features(set, unsupported);
4718 if (names) {
4719 btrfs_warn(root->fs_info,
4720 "this kernel does not support the %s feature bit%s",
4721 names, strchr(names, ',') ? "s" : "");
4722 kfree(names);
4723 } else
4724 btrfs_warn(root->fs_info,
2eaa055f
JM		 4725 "this kernel does not support %s bits 0x%llx",
4726 type, unsupported);
4727 return -EOPNOTSUPP;
4728 }
4729
4730 disallowed = set_mask & ~safe_set;
4731 if (disallowed) {
3b02a68a
JM		 4732 names = btrfs_printable_features(set, disallowed);
4733 if (names) {
4734 btrfs_warn(root->fs_info,
4735 "can't set the %s feature bit%s while mounted",
4736 names, strchr(names, ',') ? "s" : "");
4737 kfree(names);
4738 } else
4739 btrfs_warn(root->fs_info,
2eaa055f
JM		 4740 "can't set %s bits 0x%llx while mounted",
4741 type, disallowed);
4742 return -EPERM;
4743 }
4744
4745 disallowed = clear_mask & ~safe_clear;
4746 if (disallowed) {
3b02a68a
JM		 4747 names = btrfs_printable_features(set, disallowed);
4748 if (names) {
4749 btrfs_warn(root->fs_info,
4750 "can't clear the %s feature bit%s while mounted",
4751 names, strchr(names, ',') ? "s" : "");
4752 kfree(names);
4753 } else
4754 btrfs_warn(root->fs_info,
2eaa055f
JM		 4755 "can't clear %s bits 0x%llx while mounted",
4756 type, disallowed);
4757 return -EPERM;
4758 }
4759
4760 return 0;
4761}
4762
4763#define check_feature(root, change_mask, flags, mask_base) \
3b02a68a 4764check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 4765 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4766 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4767 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4768
4769static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4770{
4771 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4772 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4773 struct btrfs_ioctl_feature_flags flags[2];
4774 struct btrfs_trans_handle *trans;
4775 u64 newflags;
4776 int ret;
4777
4778 if (!capable(CAP_SYS_ADMIN))
4779 return -EPERM;
4780
4781 if (copy_from_user(flags, arg, sizeof(flags)))
4782 return -EFAULT;
4783
4784 /* Nothing to do */
4785 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4786 !flags[0].incompat_flags)
4787 return 0;
4788
4789 ret = check_feature(root, flags[0].compat_flags,
4790 flags[1].compat_flags, COMPAT);
4791 if (ret)
4792 return ret;
4793
4794 ret = check_feature(root, flags[0].compat_ro_flags,
4795 flags[1].compat_ro_flags, COMPAT_RO);
4796 if (ret)
4797 return ret;
4798
4799 ret = check_feature(root, flags[0].incompat_flags,
4800 flags[1].incompat_flags, INCOMPAT);
4801 if (ret)
4802 return ret;
4803
8051aa1a 4804 trans = btrfs_start_transaction(root, 0);
2eaa055f
JM		 4805 if (IS_ERR(trans))
4806 return PTR_ERR(trans);
4807
4808 spin_lock(&root->fs_info->super_lock);
4809 newflags = btrfs_super_compat_flags(super_block);
4810 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4811 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4812 btrfs_set_super_compat_flags(super_block, newflags);
4813
4814 newflags = btrfs_super_compat_ro_flags(super_block);
4815 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4816 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4817 btrfs_set_super_compat_ro_flags(super_block, newflags);
4818
4819 newflags = btrfs_super_incompat_flags(super_block);
4820 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4821 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4822 btrfs_set_super_incompat_flags(super_block, newflags);
4823 spin_unlock(&root->fs_info->super_lock);
4824
d0270aca 4825 return btrfs_commit_transaction(trans, root);
2eaa055f
JM		 4826}
4827
f46b5a66
CH		 4828long btrfs_ioctl(struct file *file, unsigned int
4829 cmd, unsigned long arg)
4830{
496ad9aa 4831 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 4832 void __user *argp = (void __user *)arg;
f46b5a66
CH		 4833
4834 switch (cmd) {
6cbff00f
CH		 4835 case FS_IOC_GETFLAGS:
4836 return btrfs_ioctl_getflags(file, argp);
4837 case FS_IOC_SETFLAGS:
4838 return btrfs_ioctl_setflags(file, argp);
4839 case FS_IOC_GETVERSION:
4840 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 4841 case FITRIM:
4842 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 4843 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 4844 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 4845 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 4846 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 4847 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 4848 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 4849 case BTRFS_IOC_SUBVOL_CREATE_V2:
4850 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 4851 case BTRFS_IOC_SNAP_DESTROY:
4852 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 4853 case BTRFS_IOC_SUBVOL_GETFLAGS:
4854 return btrfs_ioctl_subvol_getflags(file, argp);
4855 case BTRFS_IOC_SUBVOL_SETFLAGS:
4856 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 4857 case BTRFS_IOC_DEFAULT_SUBVOL:
4858 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 4859 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 4860 return btrfs_ioctl_defrag(file, NULL);
4861 case BTRFS_IOC_DEFRAG_RANGE:
4862 return btrfs_ioctl_defrag(file, argp);
f46b5a66 4863 case BTRFS_IOC_RESIZE:
198605a8 4864 return btrfs_ioctl_resize(file, argp);
f46b5a66 4865 case BTRFS_IOC_ADD_DEV:
4bcabaa3 4866 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 4867 case BTRFS_IOC_RM_DEV:
da24927b 4868 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 4869 case BTRFS_IOC_FS_INFO:
4870 return btrfs_ioctl_fs_info(root, argp);
4871 case BTRFS_IOC_DEV_INFO:
4872 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 4873 case BTRFS_IOC_BALANCE:
9ba1f6e4 4874 return btrfs_ioctl_balance(file, NULL);
f46b5a66 4875 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 4876 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4877 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 4878 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 4879 case BTRFS_IOC_TRANS_START:
4880 return btrfs_ioctl_trans_start(file);
4881 case BTRFS_IOC_TRANS_END:
4882 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 4883 case BTRFS_IOC_TREE_SEARCH:
4884 return btrfs_ioctl_tree_search(file, argp);
4885 case BTRFS_IOC_INO_LOOKUP:
4886 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 4887 case BTRFS_IOC_INO_PATHS:
4888 return btrfs_ioctl_ino_to_path(root, argp);
4889 case BTRFS_IOC_LOGICAL_INO:
4890 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 4891 case BTRFS_IOC_SPACE_INFO:
4892 return btrfs_ioctl_space_info(root, argp);
9b199859
FDBM		 4893 case BTRFS_IOC_SYNC: {
4894 int ret;
4895
91aef86f 4896 ret = btrfs_start_delalloc_roots(root->fs_info, 0);
9b199859
FDBM		 4897 if (ret)
4898 return ret;
4899 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
4900 return ret;
4901 }
46204592 4902 case BTRFS_IOC_START_SYNC:
9a8c28be 4903 return btrfs_ioctl_start_sync(root, argp);
46204592 4904 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 4905 return btrfs_ioctl_wait_sync(root, argp);
475f6387 4906 case BTRFS_IOC_SCRUB:
b8e95489 4907 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 4908 case BTRFS_IOC_SCRUB_CANCEL:
4909 return btrfs_ioctl_scrub_cancel(root, argp);
4910 case BTRFS_IOC_SCRUB_PROGRESS:
4911 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 4912 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 4913 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 4914 case BTRFS_IOC_BALANCE_CTL:
4915 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 4916 case BTRFS_IOC_BALANCE_PROGRESS:
4917 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 4918 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4919 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 4920#ifdef CONFIG_64BIT
4921 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4922 return btrfs_ioctl_set_received_subvol_32(file, argp);
4923#endif
31db9f7c
AB		 4924 case BTRFS_IOC_SEND:
4925 return btrfs_ioctl_send(file, argp);
c11d2c23 4926 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 4927 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 4928 case BTRFS_IOC_QUOTA_CTL:
905b0dda 4929 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 4930 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 4931 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 4932 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 4933 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 4934 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 4935 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 4936 case BTRFS_IOC_QUOTA_RESCAN:
4937 return btrfs_ioctl_quota_rescan(file, argp);
4938 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4939 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 4940 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4941 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 4942 case BTRFS_IOC_DEV_REPLACE:
4943 return btrfs_ioctl_dev_replace(root, argp);
867ab667 4944 case BTRFS_IOC_GET_FSLABEL:
4945 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 4946 case BTRFS_IOC_SET_FSLABEL:
4947 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 4948 case BTRFS_IOC_FILE_EXTENT_SAME:
4949 return btrfs_ioctl_file_extent_same(file, argp);
2eaa055f
JM		 4950 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4951 return btrfs_ioctl_get_supported_features(file, argp);
4952 case BTRFS_IOC_GET_FEATURES:
4953 return btrfs_ioctl_get_features(file, argp);
4954 case BTRFS_IOC_SET_FEATURES:
4955 return btrfs_ioctl_set_features(file, argp);
f46b5a66
CH		 4956 }
4957
4958 return -ENOTTY;
4959}
Linux 6.x block layer and io_uring tree(s)