projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: add missing kfree in btrfs_destroy_workqueue
[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 988 if (!em) {
308d9800
FM		 989 struct extent_state *cached = NULL;
990 u64 end = start + len - 1;
991
6c282eb4 992 /* get the big lock and read metadata off disk */
308d9800 993 lock_extent_bits(io_tree, start, end, 0, &cached);
6c282eb4 994 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
308d9800 995 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
6c282eb4
LZ		 996
997 if (IS_ERR(em))
998 return NULL;
999 }
1000
1001 return em;
1002}
17ce6ef8 1003
6c282eb4
LZ		 1004static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1005{
1006 struct extent_map *next;
1007 bool ret = true;
1008
1009 /* this is the last extent */
1010 if (em->start + em->len >= i_size_read(inode))
1011 return false;
1012
1013 next = defrag_lookup_extent(inode, em->start + em->len);
e2127cf0
FM		 1014 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE ||
1015 (em->block_start + em->block_len == next->block_start))
6c282eb4
LZ		 1016 ret = false;
1017
1018 free_extent_map(next);
17ce6ef8
LB		 1019 return ret;
1020}
1021
6c282eb4 1022static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 1023 u64 *last_len, u64 *skip, u64 *defrag_end,
1024 int compress)
940100a4 1025{
6c282eb4 1026 struct extent_map *em;
940100a4 1027 int ret = 1;
6c282eb4 1028 bool next_mergeable = true;
940100a4
CM		 1029
1030 /*
008873ea 1031 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1032 * defragging it
1033 */
1034 if (start < *defrag_end)
1035 return 1;
1036
1037 *skip = 0;
1038
6c282eb4
LZ		 1039 em = defrag_lookup_extent(inode, start);
1040 if (!em)
1041 return 0;
940100a4
CM		 1042
1043 /* this will cover holes, and inline extents */
17ce6ef8 1044 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1045 ret = 0;
17ce6ef8
LB		 1046 goto out;
1047 }
1048
6c282eb4 1049 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM		 1050
1051 /*
6c282eb4
LZ		 1052 * we hit a real extent, if it is big or the next extent is not a
1053 * real extent, don't bother defragging it
940100a4 1054 */
a43a2111 1055 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 1056 (em->len >= thresh || !next_mergeable))
940100a4 1057 ret = 0;
17ce6ef8 1058out:
940100a4
CM		 1059 /*
1060 * last_len ends up being a counter of how many bytes we've defragged.
1061 * every time we choose not to defrag an extent, we reset *last_len
1062 * so that the next tiny extent will force a defrag.
1063 *
1064 * The end result of this is that tiny extents before a single big
1065 * extent will force at least part of that big extent to be defragged.
1066 */
1067 if (ret) {
940100a4
CM		 1068 *defrag_end = extent_map_end(em);
1069 } else {
1070 *last_len = 0;
1071 *skip = extent_map_end(em);
1072 *defrag_end = 0;
1073 }
1074
1075 free_extent_map(em);
1076 return ret;
1077}
1078
4cb5300b
CM		 1079/*
1080 * it doesn't do much good to defrag one or two pages
1081 * at a time. This pulls in a nice chunk of pages
1082 * to COW and defrag.
1083 *
1084 * It also makes sure the delalloc code has enough
1085 * dirty data to avoid making new small extents as part
1086 * of the defrag
1087 *
1088 * It's a good idea to start RA on this range
1089 * before calling this.
1090 */
1091static int cluster_pages_for_defrag(struct inode *inode,
1092 struct page **pages,
1093 unsigned long start_index,
c41570c9 1094 unsigned long num_pages)
f46b5a66 1095{
4cb5300b
CM		 1096 unsigned long file_end;
1097 u64 isize = i_size_read(inode);
1098 u64 page_start;
1099 u64 page_end;
1f12bd06 1100 u64 page_cnt;
4cb5300b
CM		 1101 int ret;
1102 int i;
1103 int i_done;
3eaa2885 1104 struct btrfs_ordered_extent *ordered;
4cb5300b 1105 struct extent_state *cached_state = NULL;
600a45e1 1106 struct extent_io_tree *tree;
3b16a4e3 1107 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1108
4cb5300b 1109 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1110 if (!isize || start_index > file_end)
1111 return 0;
1112
1113 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1114
1115 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1116 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1117 if (ret)
1118 return ret;
4cb5300b 1119 i_done = 0;
600a45e1 1120 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1121
1122 /* step one, lock all the pages */
1f12bd06 1123 for (i = 0; i < page_cnt; i++) {
4cb5300b 1124 struct page *page;
600a45e1 1125again:
a94733d0 1126 page = find_or_create_page(inode->i_mapping,
600a45e1 1127 start_index + i, mask);
4cb5300b
CM		 1128 if (!page)
1129 break;
1130
600a45e1
MX		 1131 page_start = page_offset(page);
1132 page_end = page_start + PAGE_CACHE_SIZE - 1;
1133 while (1) {
308d9800
FM		 1134 lock_extent_bits(tree, page_start, page_end,
1135 0, &cached_state);
600a45e1
MX		 1136 ordered = btrfs_lookup_ordered_extent(inode,
1137 page_start);
308d9800
FM		 1138 unlock_extent_cached(tree, page_start, page_end,
1139 &cached_state, GFP_NOFS);
600a45e1
MX		 1140 if (!ordered)
1141 break;
1142
1143 unlock_page(page);
1144 btrfs_start_ordered_extent(inode, ordered, 1);
1145 btrfs_put_ordered_extent(ordered);
1146 lock_page(page);
1f12bd06
LB		 1147 /*
1148 * we unlocked the page above, so we need check if
1149 * it was released or not.
1150 */
1151 if (page->mapping != inode->i_mapping) {
1152 unlock_page(page);
1153 page_cache_release(page);
1154 goto again;
1155 }
600a45e1
MX		 1156 }
1157
4cb5300b
CM		 1158 if (!PageUptodate(page)) {
1159 btrfs_readpage(NULL, page);
1160 lock_page(page);
1161 if (!PageUptodate(page)) {
1162 unlock_page(page);
1163 page_cache_release(page);
1164 ret = -EIO;
1165 break;
1166 }
1167 }
600a45e1 1168
600a45e1
MX		 1169 if (page->mapping != inode->i_mapping) {
1170 unlock_page(page);
1171 page_cache_release(page);
1172 goto again;
1173 }
1174
4cb5300b
CM		 1175 pages[i] = page;
1176 i_done++;
1177 }
1178 if (!i_done || ret)
1179 goto out;
1180
1181 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1182 goto out;
1183
1184 /*
1185 * so now we have a nice long stream of locked
1186 * and up to date pages, lets wait on them
1187 */
1188 for (i = 0; i < i_done; i++)
1189 wait_on_page_writeback(pages[i]);
1190
1191 page_start = page_offset(pages[0]);
1192 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1193
1194 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1195 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1196 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1197 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1198 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1199 &cached_state, GFP_NOFS);
4cb5300b 1200
1f12bd06 1201 if (i_done != page_cnt) {
9e0baf60
JB		 1202 spin_lock(&BTRFS_I(inode)->lock);
1203 BTRFS_I(inode)->outstanding_extents++;
1204 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1205 btrfs_delalloc_release_space(inode,
1f12bd06 1206 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1207 }
1208
1209
9e8a4a8b
LB		 1210 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1211 &cached_state, GFP_NOFS);
4cb5300b
CM		 1212
1213 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1214 page_start, page_end - 1, &cached_state,
1215 GFP_NOFS);
1216
1217 for (i = 0; i < i_done; i++) {
1218 clear_page_dirty_for_io(pages[i]);
1219 ClearPageChecked(pages[i]);
1220 set_page_extent_mapped(pages[i]);
1221 set_page_dirty(pages[i]);
1222 unlock_page(pages[i]);
1223 page_cache_release(pages[i]);
1224 }
1225 return i_done;
1226out:
1227 for (i = 0; i < i_done; i++) {
1228 unlock_page(pages[i]);
1229 page_cache_release(pages[i]);
1230 }
1f12bd06 1231 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1232 return ret;
1233
1234}
1235
1236int btrfs_defrag_file(struct inode *inode, struct file *file,
1237 struct btrfs_ioctl_defrag_range_args *range,
1238 u64 newer_than, unsigned long max_to_defrag)
1239{
1240 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1241 struct file_ra_state *ra = NULL;
f46b5a66 1242 unsigned long last_index;
151a31b2 1243 u64 isize = i_size_read(inode);
940100a4
CM		 1244 u64 last_len = 0;
1245 u64 skip = 0;
1246 u64 defrag_end = 0;
4cb5300b 1247 u64 newer_off = range->start;
f46b5a66 1248 unsigned long i;
008873ea 1249 unsigned long ra_index = 0;
f46b5a66 1250 int ret;
4cb5300b 1251 int defrag_count = 0;
1a419d85 1252 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1253 int extent_thresh = range->extent_thresh;
c41570c9
JM		 1254 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1255 unsigned long cluster = max_cluster;
4cb5300b
CM		 1256 u64 new_align = ~((u64)128 * 1024 - 1);
1257 struct page **pages = NULL;
1258
0abd5b17
LB		 1259 if (isize == 0)
1260 return 0;
1261
1262 if (range->start >= isize)
1263 return -EINVAL;
1a419d85
LZ		 1264
1265 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1266 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1267 return -EINVAL;
1268 if (range->compress_type)
1269 compress_type = range->compress_type;
1270 }
f46b5a66 1271
0abd5b17
LB		 1272 if (extent_thresh == 0)
1273 extent_thresh = 256 * 1024;
940100a4 1274
4cb5300b
CM		 1275 /*
1276 * if we were not given a file, allocate a readahead
1277 * context
1278 */
1279 if (!file) {
1280 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1281 if (!ra)
1282 return -ENOMEM;
1283 file_ra_state_init(ra, inode->i_mapping);
1284 } else {
1285 ra = &file->f_ra;
1286 }
1287
d9b0d9ba 1288 pages = kmalloc_array(max_cluster, sizeof(struct page *),
4cb5300b
CM		 1289 GFP_NOFS);
1290 if (!pages) {
1291 ret = -ENOMEM;
1292 goto out_ra;
1293 }
1294
1295 /* find the last page to defrag */
1e701a32 1296 if (range->start + range->len > range->start) {
151a31b2 1297 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1298 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1299 } else {
151a31b2 1300 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1301 }
1302
4cb5300b
CM		 1303 if (newer_than) {
1304 ret = find_new_extents(root, inode, newer_than,
1305 &newer_off, 64 * 1024);
1306 if (!ret) {
1307 range->start = newer_off;
1308 /*
1309 * we always align our defrag to help keep
1310 * the extents in the file evenly spaced
1311 */
1312 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1313 } else
1314 goto out_ra;
1315 } else {
1316 i = range->start >> PAGE_CACHE_SHIFT;
1317 }
1318 if (!max_to_defrag)
7ec31b54 1319 max_to_defrag = last_index + 1;
4cb5300b 1320
2a0f7f57
LZ		 1321 /*
1322 * make writeback starts from i, so the defrag range can be
1323 * written sequentially.
1324 */
1325 if (i < inode->i_mapping->writeback_index)
1326 inode->i_mapping->writeback_index = i;
1327
f7f43cc8
CM		 1328 while (i <= last_index && defrag_count < max_to_defrag &&
1329 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1330 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1331 /*
1332 * make sure we stop running if someone unmounts
1333 * the FS
1334 */
1335 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1336 break;
1337
210549eb 1338 if (btrfs_defrag_cancelled(root->fs_info)) {
efe120a0 1339 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
210549eb
DS		 1340 ret = -EAGAIN;
1341 break;
1342 }
1343
66c26892 1344 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1345 extent_thresh, &last_len, &skip,
a43a2111
AM		 1346 &defrag_end, range->flags &
1347 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1348 unsigned long next;
1349 /*
1350 * the should_defrag function tells us how much to skip
1351 * bump our counter by the suggested amount
1352 */
1353 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1354 i = max(i + 1, next);
1355 continue;
1356 }
008873ea
LZ		 1357
1358 if (!newer_than) {
1359 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1360 PAGE_CACHE_SHIFT) - i;
1361 cluster = min(cluster, max_cluster);
1362 } else {
1363 cluster = max_cluster;
1364 }
1365
008873ea
LZ		 1366 if (i + cluster > ra_index) {
1367 ra_index = max(i, ra_index);
1368 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1369 cluster);
1370 ra_index += max_cluster;
1371 }
940100a4 1372
ecb8bea8 1373 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1374 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1375 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1376 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1377 if (ret < 0) {
1378 mutex_unlock(&inode->i_mutex);
4cb5300b 1379 goto out_ra;
ecb8bea8 1380 }
4cb5300b
CM		 1381
1382 defrag_count += ret;
d0e1d66b 1383 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1384 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1385
1386 if (newer_than) {
1387 if (newer_off == (u64)-1)
1388 break;
1389
e1f041e1
LB		 1390 if (ret > 0)
1391 i += ret;
1392
4cb5300b
CM		 1393 newer_off = max(newer_off + 1,
1394 (u64)i << PAGE_CACHE_SHIFT);
1395
1396 ret = find_new_extents(root, inode,
1397 newer_than, &newer_off,
1398 64 * 1024);
1399 if (!ret) {
1400 range->start = newer_off;
1401 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1402 } else {
1403 break;
f46b5a66 1404 }
4cb5300b 1405 } else {
008873ea 1406 if (ret > 0) {
cbcc8326 1407 i += ret;
008873ea
LZ		 1408 last_len += ret << PAGE_CACHE_SHIFT;
1409 } else {
cbcc8326 1410 i++;
008873ea
LZ		 1411 last_len = 0;
1412 }
f46b5a66 1413 }
f46b5a66
CH		 1414 }
1415
dec8ef90 1416 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1417 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1418 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1419 &BTRFS_I(inode)->runtime_flags))
1420 filemap_flush(inode->i_mapping);
1421 }
1e701a32
CM		 1422
1423 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1424 /* the filemap_flush will queue IO into the worker threads, but
1425 * we have to make sure the IO is actually started and that
1426 * ordered extents get created before we return
1427 */
1428 atomic_inc(&root->fs_info->async_submit_draining);
1429 while (atomic_read(&root->fs_info->nr_async_submits) ||
1430 atomic_read(&root->fs_info->async_delalloc_pages)) {
1431 wait_event(root->fs_info->async_submit_wait,
1432 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1433 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1434 }
1435 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1436 }
1437
1a419d85 1438 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1439 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1440 }
1441
60ccf82f 1442 ret = defrag_count;
940100a4 1443
4cb5300b 1444out_ra:
633085c7
FDBM		 1445 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1446 mutex_lock(&inode->i_mutex);
1447 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1448 mutex_unlock(&inode->i_mutex);
1449 }
4cb5300b
CM		 1450 if (!file)
1451 kfree(ra);
1452 kfree(pages);
940100a4 1453 return ret;
f46b5a66
CH		 1454}
1455
198605a8 1456static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1457 void __user *arg)
f46b5a66
CH		 1458{
1459 u64 new_size;
1460 u64 old_size;
1461 u64 devid = 1;
496ad9aa 1462 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1463 struct btrfs_ioctl_vol_args *vol_args;
1464 struct btrfs_trans_handle *trans;
1465 struct btrfs_device *device = NULL;
1466 char *sizestr;
1467 char *devstr = NULL;
1468 int ret = 0;
f46b5a66
CH		 1469 int mod = 0;
1470
e441d54d
CM		 1471 if (!capable(CAP_SYS_ADMIN))
1472 return -EPERM;
1473
198605a8
MX		 1474 ret = mnt_want_write_file(file);
1475 if (ret)
1476 return ret;
1477
5ac00add
SB		 1478 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1479 1)) {
97547676 1480 mnt_drop_write_file(file);
e57138b3 1481 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1482 }
1483
5ac00add 1484 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1485 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1486 if (IS_ERR(vol_args)) {
1487 ret = PTR_ERR(vol_args);
1488 goto out;
1489 }
5516e595
MF		 1490
1491 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1492
f46b5a66
CH		 1493 sizestr = vol_args->name;
1494 devstr = strchr(sizestr, ':');
1495 if (devstr) {
1496 char *end;
1497 sizestr = devstr + 1;
1498 *devstr = '\0';
1499 devstr = vol_args->name;
1500 devid = simple_strtoull(devstr, &end, 10);
dfd79829
MX		 1501 if (!devid) {
1502 ret = -EINVAL;
1503 goto out_free;
1504 }
efe120a0 1505 btrfs_info(root->fs_info, "resizing devid %llu", devid);
f46b5a66 1506 }
dba60f3f 1507
aa1b8cd4 1508 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1509 if (!device) {
efe120a0 1510 btrfs_info(root->fs_info, "resizer unable to find device %llu",
c1c9ff7c 1511 devid);
dfd79829 1512 ret = -ENODEV;
c9e9f97b 1513 goto out_free;
f46b5a66 1514 }
dba60f3f
MX		 1515
1516 if (!device->writeable) {
efe120a0
FH		 1517 btrfs_info(root->fs_info,
1518 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1519 devid);
dfd79829 1520 ret = -EPERM;
4e42ae1b
LB		 1521 goto out_free;
1522 }
1523
f46b5a66
CH		 1524 if (!strcmp(sizestr, "max"))
1525 new_size = device->bdev->bd_inode->i_size;
1526 else {
1527 if (sizestr[0] == '-') {
1528 mod = -1;
1529 sizestr++;
1530 } else if (sizestr[0] == '+') {
1531 mod = 1;
1532 sizestr++;
1533 }
91748467 1534 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1535 if (new_size == 0) {
1536 ret = -EINVAL;
c9e9f97b 1537 goto out_free;
f46b5a66
CH		 1538 }
1539 }
1540
63a212ab 1541 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1542 ret = -EPERM;
63a212ab
SB		 1543 goto out_free;
1544 }
1545
f46b5a66
CH		 1546 old_size = device->total_bytes;
1547
1548 if (mod < 0) {
1549 if (new_size > old_size) {
1550 ret = -EINVAL;
c9e9f97b 1551 goto out_free;
f46b5a66
CH		 1552 }
1553 new_size = old_size - new_size;
1554 } else if (mod > 0) {
eb8052e0
WF		 1555 if (new_size > ULLONG_MAX - old_size) {
1556 ret = -EINVAL;
1557 goto out_free;
1558 }
f46b5a66
CH		 1559 new_size = old_size + new_size;
1560 }
1561
1562 if (new_size < 256 * 1024 * 1024) {
1563 ret = -EINVAL;
c9e9f97b 1564 goto out_free;
f46b5a66
CH		 1565 }
1566 if (new_size > device->bdev->bd_inode->i_size) {
1567 ret = -EFBIG;
c9e9f97b 1568 goto out_free;
f46b5a66
CH		 1569 }
1570
1571 do_div(new_size, root->sectorsize);
1572 new_size *= root->sectorsize;
1573
efe120a0 1574 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
c1c9ff7c 1575 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1576
1577 if (new_size > old_size) {
a22285a6 1578 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1579 if (IS_ERR(trans)) {
1580 ret = PTR_ERR(trans);
c9e9f97b 1581 goto out_free;
98d5dc13 1582 }
f46b5a66
CH		 1583 ret = btrfs_grow_device(trans, device, new_size);
1584 btrfs_commit_transaction(trans, root);
ece7d20e 1585 } else if (new_size < old_size) {
f46b5a66 1586 ret = btrfs_shrink_device(device, new_size);
0253f40e 1587 } /* equal, nothing need to do */
f46b5a66 1588
c9e9f97b 1589out_free:
f46b5a66 1590 kfree(vol_args);
c9e9f97b
ID		 1591out:
1592 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1593 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1594 mnt_drop_write_file(file);
f46b5a66
CH		 1595 return ret;
1596}
1597
72fd032e 1598static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1599 char *name, unsigned long fd, int subvol,
1600 u64 *transid, bool readonly,
8696c533 1601 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1602{
f46b5a66 1603 int namelen;
3de4586c 1604 int ret = 0;
f46b5a66 1605
a874a63e
LB		 1606 ret = mnt_want_write_file(file);
1607 if (ret)
1608 goto out;
1609
72fd032e
SW		 1610 namelen = strlen(name);
1611 if (strchr(name, '/')) {
f46b5a66 1612 ret = -EINVAL;
a874a63e 1613 goto out_drop_write;
f46b5a66
CH		 1614 }
1615
16780cab
CM		 1616 if (name[0] == '.' &&
1617 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1618 ret = -EEXIST;
a874a63e 1619 goto out_drop_write;
16780cab
CM		 1620 }
1621
3de4586c 1622 if (subvol) {
72fd032e 1623 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1624 NULL, transid, readonly, inherit);
cb8e7090 1625 } else {
2903ff01 1626 struct fd src = fdget(fd);
3de4586c 1627 struct inode *src_inode;
2903ff01 1628 if (!src.file) {
3de4586c 1629 ret = -EINVAL;
a874a63e 1630 goto out_drop_write;
3de4586c
CM		 1631 }
1632
496ad9aa
AV		 1633 src_inode = file_inode(src.file);
1634 if (src_inode->i_sb != file_inode(file)->i_sb) {
efe120a0
FH		 1635 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1636 "Snapshot src from another FS");
23ad5b17 1637 ret = -EXDEV;
d0242061
DS		 1638 } else if (!inode_owner_or_capable(src_inode)) {
1639 /*
1640 * Subvolume creation is not restricted, but snapshots
1641 * are limited to own subvolumes only
1642 */
1643 ret = -EPERM;
ecd18815
AV		 1644 } else {
1645 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1646 BTRFS_I(src_inode)->root,
1647 transid, readonly, inherit);
3de4586c 1648 }
2903ff01 1649 fdput(src);
cb8e7090 1650 }
a874a63e
LB		 1651out_drop_write:
1652 mnt_drop_write_file(file);
f46b5a66 1653out:
72fd032e
SW		 1654 return ret;
1655}
1656
1657static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1658 void __user *arg, int subvol)
72fd032e 1659{
fa0d2b9b 1660 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1661 int ret;
1662
fa0d2b9b
LZ		 1663 vol_args = memdup_user(arg, sizeof(*vol_args));
1664 if (IS_ERR(vol_args))
1665 return PTR_ERR(vol_args);
1666 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1667
fa0d2b9b 1668 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1669 vol_args->fd, subvol,
6f72c7e2 1670 NULL, false, NULL);
fdfb1e4f 1671
fa0d2b9b
LZ		 1672 kfree(vol_args);
1673 return ret;
1674}
fdfb1e4f 1675
fa0d2b9b
LZ		 1676static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1677 void __user *arg, int subvol)
1678{
1679 struct btrfs_ioctl_vol_args_v2 *vol_args;
1680 int ret;
1681 u64 transid = 0;
1682 u64 *ptr = NULL;
b83cc969 1683 bool readonly = false;
6f72c7e2 1684 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1685
fa0d2b9b
LZ		 1686 vol_args = memdup_user(arg, sizeof(*vol_args));
1687 if (IS_ERR(vol_args))
1688 return PTR_ERR(vol_args);
1689 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1690
b83cc969 1691 if (vol_args->flags &
6f72c7e2
AJ		 1692 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1693 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1694 ret = -EOPNOTSUPP;
fa0d2b9b 1695 goto out;
72fd032e 1696 }
fa0d2b9b
LZ		 1697
1698 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1699 ptr = &transid;
b83cc969
LZ		 1700 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1701 readonly = true;
6f72c7e2
AJ		 1702 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1703 if (vol_args->size > PAGE_CACHE_SIZE) {
1704 ret = -EINVAL;
1705 goto out;
1706 }
1707 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1708 if (IS_ERR(inherit)) {
1709 ret = PTR_ERR(inherit);
1710 goto out;
1711 }
1712 }
fa0d2b9b
LZ		 1713
1714 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1715 vol_args->fd, subvol, ptr,
8696c533 1716 readonly, inherit);
fa0d2b9b
LZ		 1717
1718 if (ret == 0 && ptr &&
1719 copy_to_user(arg +
1720 offsetof(struct btrfs_ioctl_vol_args_v2,
1721 transid), ptr, sizeof(*ptr)))
1722 ret = -EFAULT;
fdfb1e4f 1723out:
f46b5a66 1724 kfree(vol_args);
6f72c7e2 1725 kfree(inherit);
f46b5a66
CH		 1726 return ret;
1727}
1728
0caa102d
LZ		 1729static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1730 void __user *arg)
1731{
496ad9aa 1732 struct inode *inode = file_inode(file);
0caa102d
LZ		 1733 struct btrfs_root *root = BTRFS_I(inode)->root;
1734 int ret = 0;
1735 u64 flags = 0;
1736
33345d01 1737 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1738 return -EINVAL;
1739
1740 down_read(&root->fs_info->subvol_sem);
1741 if (btrfs_root_readonly(root))
1742 flags |= BTRFS_SUBVOL_RDONLY;
1743 up_read(&root->fs_info->subvol_sem);
1744
1745 if (copy_to_user(arg, &flags, sizeof(flags)))
1746 ret = -EFAULT;
1747
1748 return ret;
1749}
1750
1751static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1752 void __user *arg)
1753{
496ad9aa 1754 struct inode *inode = file_inode(file);
0caa102d
LZ		 1755 struct btrfs_root *root = BTRFS_I(inode)->root;
1756 struct btrfs_trans_handle *trans;
1757 u64 root_flags;
1758 u64 flags;
1759 int ret = 0;
1760
bd60ea0f
DS		 1761 if (!inode_owner_or_capable(inode))
1762 return -EPERM;
1763
b9ca0664
LB		 1764 ret = mnt_want_write_file(file);
1765 if (ret)
1766 goto out;
0caa102d 1767
b9ca0664
LB		 1768 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1769 ret = -EINVAL;
1770 goto out_drop_write;
1771 }
0caa102d 1772
b9ca0664
LB		 1773 if (copy_from_user(&flags, arg, sizeof(flags))) {
1774 ret = -EFAULT;
1775 goto out_drop_write;
1776 }
0caa102d 1777
b9ca0664
LB		 1778 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1779 ret = -EINVAL;
1780 goto out_drop_write;
1781 }
0caa102d 1782
b9ca0664
LB		 1783 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1784 ret = -EOPNOTSUPP;
1785 goto out_drop_write;
1786 }
0caa102d
LZ		 1787
1788 down_write(&root->fs_info->subvol_sem);
1789
1790 /* nothing to do */
1791 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1792 goto out_drop_sem;
0caa102d
LZ		 1793
1794 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1795 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1796 btrfs_set_root_flags(&root->root_item,
1797 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1798 } else {
1799 /*
1800 * Block RO -> RW transition if this subvolume is involved in
1801 * send
1802 */
1803 spin_lock(&root->root_item_lock);
1804 if (root->send_in_progress == 0) {
1805 btrfs_set_root_flags(&root->root_item,
0caa102d 1806 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1807 spin_unlock(&root->root_item_lock);
1808 } else {
1809 spin_unlock(&root->root_item_lock);
1810 btrfs_warn(root->fs_info,
1811 "Attempt to set subvolume %llu read-write during send",
1812 root->root_key.objectid);
1813 ret = -EPERM;
1814 goto out_drop_sem;
1815 }
1816 }
0caa102d
LZ		 1817
1818 trans = btrfs_start_transaction(root, 1);
1819 if (IS_ERR(trans)) {
1820 ret = PTR_ERR(trans);
1821 goto out_reset;
1822 }
1823
b4dc2b8c 1824 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1825 &root->root_key, &root->root_item);
1826
1827 btrfs_commit_transaction(trans, root);
1828out_reset:
1829 if (ret)
1830 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1831out_drop_sem:
0caa102d 1832 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1833out_drop_write:
1834 mnt_drop_write_file(file);
1835out:
0caa102d
LZ		 1836 return ret;
1837}
1838
76dda93c
YZ		 1839/*
1840 * helper to check if the subvolume references other subvolumes
1841 */
1842static noinline int may_destroy_subvol(struct btrfs_root *root)
1843{
1844 struct btrfs_path *path;
175a2b87 1845 struct btrfs_dir_item *di;
76dda93c 1846 struct btrfs_key key;
175a2b87 1847 u64 dir_id;
76dda93c
YZ		 1848 int ret;
1849
1850 path = btrfs_alloc_path();
1851 if (!path)
1852 return -ENOMEM;
1853
175a2b87
JB		 1854 /* Make sure this root isn't set as the default subvol */
1855 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1856 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1857 dir_id, "default", 7, 0);
1858 if (di && !IS_ERR(di)) {
1859 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1860 if (key.objectid == root->root_key.objectid) {
1861 ret = -ENOTEMPTY;
1862 goto out;
1863 }
1864 btrfs_release_path(path);
1865 }
1866
76dda93c
YZ		 1867 key.objectid = root->root_key.objectid;
1868 key.type = BTRFS_ROOT_REF_KEY;
1869 key.offset = (u64)-1;
1870
1871 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1872 &key, path, 0, 0);
1873 if (ret < 0)
1874 goto out;
1875 BUG_ON(ret == 0);
1876
1877 ret = 0;
1878 if (path->slots[0] > 0) {
1879 path->slots[0]--;
1880 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1881 if (key.objectid == root->root_key.objectid &&
1882 key.type == BTRFS_ROOT_REF_KEY)
1883 ret = -ENOTEMPTY;
1884 }
1885out:
1886 btrfs_free_path(path);
1887 return ret;
1888}
1889
ac8e9819
CM		 1890static noinline int key_in_sk(struct btrfs_key *key,
1891 struct btrfs_ioctl_search_key *sk)
1892{
abc6e134
CM		 1893 struct btrfs_key test;
1894 int ret;
1895
1896 test.objectid = sk->min_objectid;
1897 test.type = sk->min_type;
1898 test.offset = sk->min_offset;
1899
1900 ret = btrfs_comp_cpu_keys(key, &test);
1901 if (ret < 0)
ac8e9819 1902 return 0;
abc6e134
CM		 1903
1904 test.objectid = sk->max_objectid;
1905 test.type = sk->max_type;
1906 test.offset = sk->max_offset;
1907
1908 ret = btrfs_comp_cpu_keys(key, &test);
1909 if (ret > 0)
ac8e9819
CM		 1910 return 0;
1911 return 1;
1912}
1913
1914static noinline int copy_to_sk(struct btrfs_root *root,
1915 struct btrfs_path *path,
1916 struct btrfs_key *key,
1917 struct btrfs_ioctl_search_key *sk,
1918 char *buf,
1919 unsigned long *sk_offset,
1920 int *num_found)
1921{
1922 u64 found_transid;
1923 struct extent_buffer *leaf;
1924 struct btrfs_ioctl_search_header sh;
1925 unsigned long item_off;
1926 unsigned long item_len;
1927 int nritems;
1928 int i;
1929 int slot;
ac8e9819
CM		 1930 int ret = 0;
1931
1932 leaf = path->nodes[0];
1933 slot = path->slots[0];
1934 nritems = btrfs_header_nritems(leaf);
1935
1936 if (btrfs_header_generation(leaf) > sk->max_transid) {
1937 i = nritems;
1938 goto advance_key;
1939 }
1940 found_transid = btrfs_header_generation(leaf);
1941
1942 for (i = slot; i < nritems; i++) {
1943 item_off = btrfs_item_ptr_offset(leaf, i);
1944 item_len = btrfs_item_size_nr(leaf, i);
1945
03b71c6c
GP		 1946 btrfs_item_key_to_cpu(leaf, key, i);
1947 if (!key_in_sk(key, sk))
1948 continue;
1949
1950 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
ac8e9819
CM		 1951 item_len = 0;
1952
1953 if (sizeof(sh) + item_len + *sk_offset >
1954 BTRFS_SEARCH_ARGS_BUFSIZE) {
1955 ret = 1;
1956 goto overflow;
1957 }
1958
ac8e9819
CM		 1959 sh.objectid = key->objectid;
1960 sh.offset = key->offset;
1961 sh.type = key->type;
1962 sh.len = item_len;
1963 sh.transid = found_transid;
1964
1965 /* copy search result header */
1966 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1967 *sk_offset += sizeof(sh);
1968
1969 if (item_len) {
1970 char *p = buf + *sk_offset;
1971 /* copy the item */
1972 read_extent_buffer(leaf, p,
1973 item_off, item_len);
1974 *sk_offset += item_len;
ac8e9819 1975 }
e2156867 1976 (*num_found)++;
ac8e9819
CM		 1977
1978 if (*num_found >= sk->nr_items)
1979 break;
1980 }
1981advance_key:
abc6e134
CM		 1982 ret = 0;
1983 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1984 key->offset++;
abc6e134
CM		 1985 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1986 key->offset = 0;
ac8e9819 1987 key->type++;
abc6e134
CM		 1988 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1989 key->offset = 0;
1990 key->type = 0;
ac8e9819 1991 key->objectid++;
abc6e134
CM		 1992 } else
1993 ret = 1;
ac8e9819 1994overflow:
ac8e9819
CM		 1995 return ret;
1996}
1997
1998static noinline int search_ioctl(struct inode *inode,
1999 struct btrfs_ioctl_search_args *args)
2000{
2001 struct btrfs_root *root;
2002 struct btrfs_key key;
ac8e9819
CM		 2003 struct btrfs_path *path;
2004 struct btrfs_ioctl_search_key *sk = &args->key;
2005 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2006 int ret;
2007 int num_found = 0;
2008 unsigned long sk_offset = 0;
2009
2010 path = btrfs_alloc_path();
2011 if (!path)
2012 return -ENOMEM;
2013
2014 if (sk->tree_id == 0) {
2015 /* search the root of the inode that was passed */
2016 root = BTRFS_I(inode)->root;
2017 } else {
2018 key.objectid = sk->tree_id;
2019 key.type = BTRFS_ROOT_ITEM_KEY;
2020 key.offset = (u64)-1;
2021 root = btrfs_read_fs_root_no_name(info, &key);
2022 if (IS_ERR(root)) {
efe120a0 2023 printk(KERN_ERR "BTRFS: could not find root %llu\n",
ac8e9819
CM		 2024 sk->tree_id);
2025 btrfs_free_path(path);
2026 return -ENOENT;
2027 }
2028 }
2029
2030 key.objectid = sk->min_objectid;
2031 key.type = sk->min_type;
2032 key.offset = sk->min_offset;
2033
ac8e9819
CM		 2034 path->keep_locks = 1;
2035
67871254 2036 while (1) {
6174d3cb 2037 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2038 if (ret != 0) {
2039 if (ret > 0)
2040 ret = 0;
2041 goto err;
2042 }
2043 ret = copy_to_sk(root, path, &key, sk, args->buf,
2044 &sk_offset, &num_found);
b3b4aa74 2045 btrfs_release_path(path);
ac8e9819
CM		 2046 if (ret || num_found >= sk->nr_items)
2047 break;
2048
2049 }
2050 ret = 0;
2051err:
2052 sk->nr_items = num_found;
2053 btrfs_free_path(path);
2054 return ret;
2055}
2056
2057static noinline int btrfs_ioctl_tree_search(struct file *file,
2058 void __user *argp)
2059{
2060 struct btrfs_ioctl_search_args *args;
2061 struct inode *inode;
2062 int ret;
2063
2064 if (!capable(CAP_SYS_ADMIN))
2065 return -EPERM;
2066
2354d08f
JL		 2067 args = memdup_user(argp, sizeof(*args));
2068 if (IS_ERR(args))
2069 return PTR_ERR(args);
ac8e9819 2070
496ad9aa 2071 inode = file_inode(file);
ac8e9819
CM		 2072 ret = search_ioctl(inode, args);
2073 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2074 ret = -EFAULT;
2075 kfree(args);
2076 return ret;
2077}
2078
98d377a0 2079/*
ac8e9819
CM		 2080 * Search INODE_REFs to identify path name of 'dirid' directory
2081 * in a 'tree_id' tree. and sets path name to 'name'.
2082 */
98d377a0
TH		 2083static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2084 u64 tree_id, u64 dirid, char *name)
2085{
2086 struct btrfs_root *root;
2087 struct btrfs_key key;
ac8e9819 2088 char *ptr;
98d377a0
TH		 2089 int ret = -1;
2090 int slot;
2091 int len;
2092 int total_len = 0;
2093 struct btrfs_inode_ref *iref;
2094 struct extent_buffer *l;
2095 struct btrfs_path *path;
2096
2097 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2098 name[0]='\0';
2099 return 0;
2100 }
2101
2102 path = btrfs_alloc_path();
2103 if (!path)
2104 return -ENOMEM;
2105
ac8e9819 2106 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2107
2108 key.objectid = tree_id;
2109 key.type = BTRFS_ROOT_ITEM_KEY;
2110 key.offset = (u64)-1;
2111 root = btrfs_read_fs_root_no_name(info, &key);
2112 if (IS_ERR(root)) {
efe120a0 2113 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
8ad6fcab
CM		 2114 ret = -ENOENT;
2115 goto out;
98d377a0
TH		 2116 }
2117
2118 key.objectid = dirid;
2119 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2120 key.offset = (u64)-1;
98d377a0 2121
67871254 2122 while (1) {
98d377a0
TH		 2123 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2124 if (ret < 0)
2125 goto out;
18674c6c
FDBM		 2126 else if (ret > 0) {
2127 ret = btrfs_previous_item(root, path, dirid,
2128 BTRFS_INODE_REF_KEY);
2129 if (ret < 0)
2130 goto out;
2131 else if (ret > 0) {
2132 ret = -ENOENT;
2133 goto out;
2134 }
2135 }
98d377a0
TH		 2136
2137 l = path->nodes[0];
2138 slot = path->slots[0];
2139 btrfs_item_key_to_cpu(l, &key, slot);
2140
98d377a0
TH		 2141 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2142 len = btrfs_inode_ref_name_len(l, iref);
2143 ptr -= len + 1;
2144 total_len += len + 1;
a696cf35
FDBM		 2145 if (ptr < name) {
2146 ret = -ENAMETOOLONG;
98d377a0 2147 goto out;
a696cf35 2148 }
98d377a0
TH		 2149
2150 *(ptr + len) = '/';
67871254 2151 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2152
2153 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2154 break;
2155
b3b4aa74 2156 btrfs_release_path(path);
98d377a0 2157 key.objectid = key.offset;
8ad6fcab 2158 key.offset = (u64)-1;
98d377a0 2159 dirid = key.objectid;
98d377a0 2160 }
77906a50 2161 memmove(name, ptr, total_len);
67871254 2162 name[total_len] = '\0';
98d377a0
TH		 2163 ret = 0;
2164out:
2165 btrfs_free_path(path);
ac8e9819
CM		 2166 return ret;
2167}
2168
2169static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2170 void __user *argp)
2171{
2172 struct btrfs_ioctl_ino_lookup_args *args;
2173 struct inode *inode;
2174 int ret;
2175
2176 if (!capable(CAP_SYS_ADMIN))
2177 return -EPERM;
2178
2354d08f
JL		 2179 args = memdup_user(argp, sizeof(*args));
2180 if (IS_ERR(args))
2181 return PTR_ERR(args);
c2b96929 2182
496ad9aa 2183 inode = file_inode(file);
ac8e9819 2184
1b53ac4d
CM		 2185 if (args->treeid == 0)
2186 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2187
ac8e9819
CM		 2188 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2189 args->treeid, args->objectid,
2190 args->name);
2191
2192 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2193 ret = -EFAULT;
2194
2195 kfree(args);
98d377a0
TH		 2196 return ret;
2197}
2198
76dda93c
YZ		 2199static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2200 void __user *arg)
2201{
54563d41 2202 struct dentry *parent = file->f_path.dentry;
76dda93c
YZ		 2203 struct dentry *dentry;
2204 struct inode *dir = parent->d_inode;
2205 struct inode *inode;
2206 struct btrfs_root *root = BTRFS_I(dir)->root;
2207 struct btrfs_root *dest = NULL;
2208 struct btrfs_ioctl_vol_args *vol_args;
2209 struct btrfs_trans_handle *trans;
c58aaad2
MX		 2210 struct btrfs_block_rsv block_rsv;
2211 u64 qgroup_reserved;
76dda93c
YZ		 2212 int namelen;
2213 int ret;
2214 int err = 0;
2215
76dda93c
YZ		 2216 vol_args = memdup_user(arg, sizeof(*vol_args));
2217 if (IS_ERR(vol_args))
2218 return PTR_ERR(vol_args);
2219
2220 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2221 namelen = strlen(vol_args->name);
2222 if (strchr(vol_args->name, '/') ||
2223 strncmp(vol_args->name, "..", namelen) == 0) {
2224 err = -EINVAL;
2225 goto out;
2226 }
2227
a561be71 2228 err = mnt_want_write_file(file);
76dda93c
YZ		 2229 if (err)
2230 goto out;
2231
5c50c9b8
DS		 2232 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2233 if (err == -EINTR)
e43f998e 2234 goto out_drop_write;
76dda93c
YZ		 2235 dentry = lookup_one_len(vol_args->name, parent, namelen);
2236 if (IS_ERR(dentry)) {
2237 err = PTR_ERR(dentry);
2238 goto out_unlock_dir;
2239 }
2240
2241 if (!dentry->d_inode) {
2242 err = -ENOENT;
2243 goto out_dput;
2244 }
2245
2246 inode = dentry->d_inode;
4260f7c7 2247 dest = BTRFS_I(inode)->root;
67871254 2248 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2249 /*
2250 * Regular user. Only allow this with a special mount
2251 * option, when the user has write+exec access to the
2252 * subvol root, and when rmdir(2) would have been
2253 * allowed.
2254 *
2255 * Note that this is _not_ check that the subvol is
2256 * empty or doesn't contain data that we wouldn't
2257 * otherwise be able to delete.
2258 *
2259 * Users who want to delete empty subvols should try
2260 * rmdir(2).
2261 */
2262 err = -EPERM;
2263 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2264 goto out_dput;
2265
2266 /*
2267 * Do not allow deletion if the parent dir is the same
2268 * as the dir to be deleted. That means the ioctl
2269 * must be called on the dentry referencing the root
2270 * of the subvol, not a random directory contained
2271 * within it.
2272 */
2273 err = -EINVAL;
2274 if (root == dest)
2275 goto out_dput;
2276
2277 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2278 if (err)
2279 goto out_dput;
4260f7c7
SW		 2280 }
2281
5c39da5b
MX		 2282 /* check if subvolume may be deleted by a user */
2283 err = btrfs_may_delete(dir, dentry, 1);
2284 if (err)
2285 goto out_dput;
2286
33345d01 2287 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2288 err = -EINVAL;
2289 goto out_dput;
2290 }
2291
76dda93c
YZ		 2292 mutex_lock(&inode->i_mutex);
2293 err = d_invalidate(dentry);
2294 if (err)
2295 goto out_unlock;
2296
2297 down_write(&root->fs_info->subvol_sem);
2298
2299 err = may_destroy_subvol(dest);
2300 if (err)
2301 goto out_up_write;
2302
c58aaad2
MX		 2303 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2304 /*
2305 * One for dir inode, two for dir entries, two for root
2306 * ref/backref.
2307 */
2308 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2309 5, &qgroup_reserved, true);
c58aaad2
MX		 2310 if (err)
2311 goto out_up_write;
2312
a22285a6
YZ		 2313 trans = btrfs_start_transaction(root, 0);
2314 if (IS_ERR(trans)) {
2315 err = PTR_ERR(trans);
c58aaad2 2316 goto out_release;
a22285a6 2317 }
c58aaad2
MX		 2318 trans->block_rsv = &block_rsv;
2319 trans->bytes_reserved = block_rsv.size;
a22285a6 2320
76dda93c
YZ		 2321 ret = btrfs_unlink_subvol(trans, root, dir,
2322 dest->root_key.objectid,
2323 dentry->d_name.name,
2324 dentry->d_name.len);
79787eaa
JM		 2325 if (ret) {
2326 err = ret;
2327 btrfs_abort_transaction(trans, root, ret);
2328 goto out_end_trans;
2329 }
76dda93c
YZ		 2330
2331 btrfs_record_root_in_trans(trans, dest);
2332
2333 memset(&dest->root_item.drop_progress, 0,
2334 sizeof(dest->root_item.drop_progress));
2335 dest->root_item.drop_level = 0;
2336 btrfs_set_root_refs(&dest->root_item, 0);
2337
d68fc57b
YZ		 2338 if (!xchg(&dest->orphan_item_inserted, 1)) {
2339 ret = btrfs_insert_orphan_item(trans,
2340 root->fs_info->tree_root,
2341 dest->root_key.objectid);
79787eaa
JM		 2342 if (ret) {
2343 btrfs_abort_transaction(trans, root, ret);
2344 err = ret;
2345 goto out_end_trans;
2346 }
d68fc57b 2347 }
dd5f9615
SB		 2348
2349 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2350 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2351 dest->root_key.objectid);
2352 if (ret && ret != -ENOENT) {
2353 btrfs_abort_transaction(trans, root, ret);
2354 err = ret;
2355 goto out_end_trans;
2356 }
2357 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2358 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2359 dest->root_item.received_uuid,
2360 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2361 dest->root_key.objectid);
2362 if (ret && ret != -ENOENT) {
2363 btrfs_abort_transaction(trans, root, ret);
2364 err = ret;
2365 goto out_end_trans;
2366 }
2367 }
2368
79787eaa 2369out_end_trans:
c58aaad2
MX		 2370 trans->block_rsv = NULL;
2371 trans->bytes_reserved = 0;
531cb13f 2372 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2373 if (ret && !err)
2374 err = ret;
76dda93c 2375 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2376out_release:
2377 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2378out_up_write:
2379 up_write(&root->fs_info->subvol_sem);
2380out_unlock:
2381 mutex_unlock(&inode->i_mutex);
2382 if (!err) {
efefb143 2383 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2384 btrfs_invalidate_inodes(dest);
2385 d_delete(dentry);
fa6ac876
LB		 2386
2387 /* the last ref */
2388 if (dest->cache_inode) {
2389 iput(dest->cache_inode);
2390 dest->cache_inode = NULL;
2391 }
76dda93c
YZ		 2392 }
2393out_dput:
2394 dput(dentry);
2395out_unlock_dir:
2396 mutex_unlock(&dir->i_mutex);
e43f998e 2397out_drop_write:
2a79f17e 2398 mnt_drop_write_file(file);
76dda93c
YZ		 2399out:
2400 kfree(vol_args);
2401 return err;
2402}
2403
1e701a32 2404static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2405{
496ad9aa 2406 struct inode *inode = file_inode(file);
f46b5a66 2407 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2408 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2409 int ret;
2410
25122d15
ID		 2411 ret = mnt_want_write_file(file);
2412 if (ret)
2413 return ret;
b83cc969 2414
25122d15
ID		 2415 if (btrfs_root_readonly(root)) {
2416 ret = -EROFS;
2417 goto out;
5ac00add 2418 }
f46b5a66
CH		 2419
2420 switch (inode->i_mode & S_IFMT) {
2421 case S_IFDIR:
e441d54d
CM		 2422 if (!capable(CAP_SYS_ADMIN)) {
2423 ret = -EPERM;
2424 goto out;
2425 }
de78b51a 2426 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2427 if (ret)
2428 goto out;
de78b51a 2429 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2430 break;
2431 case S_IFREG:
e441d54d
CM		 2432 if (!(file->f_mode & FMODE_WRITE)) {
2433 ret = -EINVAL;
2434 goto out;
2435 }
1e701a32
CM		 2436
2437 range = kzalloc(sizeof(*range), GFP_KERNEL);
2438 if (!range) {
2439 ret = -ENOMEM;
2440 goto out;
2441 }
2442
2443 if (argp) {
2444 if (copy_from_user(range, argp,
2445 sizeof(*range))) {
2446 ret = -EFAULT;
2447 kfree(range);
683be16e 2448 goto out;
1e701a32
CM		 2449 }
2450 /* compression requires us to start the IO */
2451 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2452 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2453 range->extent_thresh = (u32)-1;
2454 }
2455 } else {
2456 /* the rest are all set to zero by kzalloc */
2457 range->len = (u64)-1;
2458 }
496ad9aa 2459 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2460 range, 0, 0);
2461 if (ret > 0)
2462 ret = 0;
1e701a32 2463 kfree(range);
f46b5a66 2464 break;
8929ecfa
YZ		 2465 default:
2466 ret = -EINVAL;
f46b5a66 2467 }
e441d54d 2468out:
25122d15 2469 mnt_drop_write_file(file);
e441d54d 2470 return ret;
f46b5a66
CH		 2471}
2472
b2950863 2473static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2474{
2475 struct btrfs_ioctl_vol_args *vol_args;
2476 int ret;
2477
e441d54d
CM		 2478 if (!capable(CAP_SYS_ADMIN))
2479 return -EPERM;
2480
5ac00add
SB		 2481 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2482 1)) {
e57138b3 2483 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 2484 }
2485
5ac00add 2486 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2487 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2488 if (IS_ERR(vol_args)) {
2489 ret = PTR_ERR(vol_args);
2490 goto out;
2491 }
f46b5a66 2492
5516e595 2493 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2494 ret = btrfs_init_new_device(root, vol_args->name);
2495
f46b5a66 2496 kfree(vol_args);
c9e9f97b
ID		 2497out:
2498 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2499 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2500 return ret;
2501}
2502
da24927b 2503static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2504{
496ad9aa 2505 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2506 struct btrfs_ioctl_vol_args *vol_args;
2507 int ret;
2508
e441d54d
CM		 2509 if (!capable(CAP_SYS_ADMIN))
2510 return -EPERM;
2511
da24927b
MX		 2512 ret = mnt_want_write_file(file);
2513 if (ret)
2514 return ret;
c146afad 2515
dae7b665 2516 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2517 if (IS_ERR(vol_args)) {
2518 ret = PTR_ERR(vol_args);
2519 goto out;
2520 }
f46b5a66 2521
5516e595 2522 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2523
183860f6
AJ		 2524 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2525 1)) {
2526 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2527 goto out;
2528 }
2529
2530 mutex_lock(&root->fs_info->volume_mutex);
2531 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2532 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2533 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6
AJ		 2534
2535out:
2536 kfree(vol_args);
4ac20c70 2537 mnt_drop_write_file(file);
f46b5a66
CH		 2538 return ret;
2539}
2540
475f6387
JS		 2541static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2542{
027ed2f0 2543 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2544 struct btrfs_device *device;
2545 struct btrfs_device *next;
2546 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2547 int ret = 0;
475f6387
JS		 2548
2549 if (!capable(CAP_SYS_ADMIN))
2550 return -EPERM;
2551
027ed2f0
LZ		 2552 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2553 if (!fi_args)
2554 return -ENOMEM;
2555
f7171750 2556 mutex_lock(&fs_devices->device_list_mutex);
027ed2f0
LZ		 2557 fi_args->num_devices = fs_devices->num_devices;
2558 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387 2559
475f6387 2560 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2561 if (device->devid > fi_args->max_id)
2562 fi_args->max_id = device->devid;
475f6387
JS		 2563 }
2564 mutex_unlock(&fs_devices->device_list_mutex);
2565
027ed2f0
LZ		 2566 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2567 ret = -EFAULT;
475f6387 2568
027ed2f0
LZ		 2569 kfree(fi_args);
2570 return ret;
475f6387
JS		 2571}
2572
2573static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2574{
2575 struct btrfs_ioctl_dev_info_args *di_args;
2576 struct btrfs_device *dev;
2577 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2578 int ret = 0;
2579 char *s_uuid = NULL;
475f6387
JS		 2580
2581 if (!capable(CAP_SYS_ADMIN))
2582 return -EPERM;
2583
2584 di_args = memdup_user(arg, sizeof(*di_args));
2585 if (IS_ERR(di_args))
2586 return PTR_ERR(di_args);
2587
dd5f9615 2588 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2589 s_uuid = di_args->uuid;
2590
2591 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2592 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2593
2594 if (!dev) {
2595 ret = -ENODEV;
2596 goto out;
2597 }
2598
2599 di_args->devid = dev->devid;
2600 di_args->bytes_used = dev->bytes_used;
2601 di_args->total_bytes = dev->total_bytes;
2602 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2603 if (dev->name) {
606686ee
JB		 2604 struct rcu_string *name;
2605
2606 rcu_read_lock();
2607 name = rcu_dereference(dev->name);
2608 strncpy(di_args->path, name->str, sizeof(di_args->path));
2609 rcu_read_unlock();
a27202fb
JM		 2610 di_args->path[sizeof(di_args->path) - 1] = 0;
2611 } else {
99ba55ad 2612 di_args->path[0] = '\0';
a27202fb 2613 }
475f6387
JS		 2614
2615out:
55793c0d 2616 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2617 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2618 ret = -EFAULT;
2619
2620 kfree(di_args);
2621 return ret;
2622}
2623
416161db
MF		 2624static struct page *extent_same_get_page(struct inode *inode, u64 off)
2625{
2626 struct page *page;
2627 pgoff_t index;
2628 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2629
2630 index = off >> PAGE_CACHE_SHIFT;
2631
2632 page = grab_cache_page(inode->i_mapping, index);
2633 if (!page)
2634 return NULL;
2635
2636 if (!PageUptodate(page)) {
2637 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2638 0))
2639 return NULL;
2640 lock_page(page);
2641 if (!PageUptodate(page)) {
2642 unlock_page(page);
2643 page_cache_release(page);
2644 return NULL;
2645 }
2646 }
2647 unlock_page(page);
2648
2649 return page;
2650}
2651
77fe20dc
MF		 2652static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2653{
2654 /* do any pending delalloc/csum calc on src, one way or
2655 another, and lock file content */
2656 while (1) {
2657 struct btrfs_ordered_extent *ordered;
2658 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2659 ordered = btrfs_lookup_first_ordered_extent(inode,
2660 off + len - 1);
2661 if (!ordered &&
2662 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2663 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2664 break;
2665 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2666 if (ordered)
2667 btrfs_put_ordered_extent(ordered);
2668 btrfs_wait_ordered_range(inode, off, len);
2669 }
2670}
2671
416161db
MF		 2672static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2673 struct inode *inode2, u64 loff2, u64 len)
2674{
2675 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2676 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2677
2678 mutex_unlock(&inode1->i_mutex);
2679 mutex_unlock(&inode2->i_mutex);
2680}
2681
2682static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2683 struct inode *inode2, u64 loff2, u64 len)
2684{
2685 if (inode1 < inode2) {
2686 swap(inode1, inode2);
2687 swap(loff1, loff2);
2688 }
2689
2690 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2691 lock_extent_range(inode1, loff1, len);
2692 if (inode1 != inode2) {
2693 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2694 lock_extent_range(inode2, loff2, len);
2695 }
2696}
2697
2698static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2699 u64 dst_loff, u64 len)
2700{
2701 int ret = 0;
2702 struct page *src_page, *dst_page;
2703 unsigned int cmp_len = PAGE_CACHE_SIZE;
2704 void *addr, *dst_addr;
2705
2706 while (len) {
2707 if (len < PAGE_CACHE_SIZE)
2708 cmp_len = len;
2709
2710 src_page = extent_same_get_page(src, loff);
2711 if (!src_page)
2712 return -EINVAL;
2713 dst_page = extent_same_get_page(dst, dst_loff);
2714 if (!dst_page) {
2715 page_cache_release(src_page);
2716 return -EINVAL;
2717 }
2718 addr = kmap_atomic(src_page);
2719 dst_addr = kmap_atomic(dst_page);
2720
2721 flush_dcache_page(src_page);
2722 flush_dcache_page(dst_page);
2723
2724 if (memcmp(addr, dst_addr, cmp_len))
2725 ret = BTRFS_SAME_DATA_DIFFERS;
2726
2727 kunmap_atomic(addr);
2728 kunmap_atomic(dst_addr);
2729 page_cache_release(src_page);
2730 page_cache_release(dst_page);
2731
2732 if (ret)
2733 break;
2734
2735 loff += cmp_len;
2736 dst_loff += cmp_len;
2737 len -= cmp_len;
2738 }
2739
2740 return ret;
2741}
2742
2743static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2744{
2745 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2746
2747 if (off + len > inode->i_size || off + len < off)
2748 return -EINVAL;
2749 /* Check that we are block aligned - btrfs_clone() requires this */
2750 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2751 return -EINVAL;
2752
2753 return 0;
2754}
2755
2756static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2757 struct inode *dst, u64 dst_loff)
2758{
2759 int ret;
2760
2761 /*
2762 * btrfs_clone() can't handle extents in the same file
2763 * yet. Once that works, we can drop this check and replace it
2764 * with a check for the same inode, but overlapping extents.
2765 */
2766 if (src == dst)
2767 return -EINVAL;
2768
2769 btrfs_double_lock(src, loff, dst, dst_loff, len);
2770
2771 ret = extent_same_check_offsets(src, loff, len);
2772 if (ret)
2773 goto out_unlock;
2774
2775 ret = extent_same_check_offsets(dst, dst_loff, len);
2776 if (ret)
2777 goto out_unlock;
2778
2779 /* don't make the dst file partly checksummed */
2780 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2781 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2782 ret = -EINVAL;
2783 goto out_unlock;
2784 }
2785
2786 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2787 if (ret == 0)
2788 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2789
2790out_unlock:
2791 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2792
2793 return ret;
2794}
2795
2796#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2797
2798static long btrfs_ioctl_file_extent_same(struct file *file,
2799 void __user *argp)
2800{
cbf8b8ca
MF		 2801 struct btrfs_ioctl_same_args tmp;
2802 struct btrfs_ioctl_same_args *same;
2803 struct btrfs_ioctl_same_extent_info *info;
416161db
MF		 2804 struct inode *src = file->f_dentry->d_inode;
2805 struct file *dst_file = NULL;
2806 struct inode *dst;
2807 u64 off;
2808 u64 len;
2809 int i;
2810 int ret;
cbf8b8ca 2811 unsigned long size;
416161db
MF		 2812 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2813 bool is_admin = capable(CAP_SYS_ADMIN);
2814
2815 if (!(file->f_mode & FMODE_READ))
2816 return -EINVAL;
2817
2818 ret = mnt_want_write_file(file);
2819 if (ret)
2820 return ret;
2821
cbf8b8ca 2822 if (copy_from_user(&tmp,
416161db 2823 (struct btrfs_ioctl_same_args __user *)argp,
cbf8b8ca 2824 sizeof(tmp))) {
416161db
MF		 2825 ret = -EFAULT;
2826 goto out;
2827 }
2828
cbf8b8ca
MF		 2829 size = sizeof(tmp) +
2830 tmp.dest_count * sizeof(struct btrfs_ioctl_same_extent_info);
2831
229eed43 2832 same = memdup_user((struct btrfs_ioctl_same_args __user *)argp, size);
cbf8b8ca 2833
229eed43
GB		 2834 if (IS_ERR(same)) {
2835 ret = PTR_ERR(same);
cbf8b8ca
MF		 2836 goto out;
2837 }
2838
2839 off = same->logical_offset;
2840 len = same->length;
416161db
MF		 2841
2842 /*
2843 * Limit the total length we will dedupe for each operation.
2844 * This is intended to bound the total time spent in this
2845 * ioctl to something sane.
2846 */
2847 if (len > BTRFS_MAX_DEDUPE_LEN)
2848 len = BTRFS_MAX_DEDUPE_LEN;
2849
2850 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2851 /*
2852 * Btrfs does not support blocksize < page_size. As a
2853 * result, btrfs_cmp_data() won't correctly handle
2854 * this situation without an update.
2855 */
2856 ret = -EINVAL;
2857 goto out;
2858 }
2859
2860 ret = -EISDIR;
2861 if (S_ISDIR(src->i_mode))
2862 goto out;
2863
2864 ret = -EACCES;
2865 if (!S_ISREG(src->i_mode))
2866 goto out;
2867
cbf8b8ca
MF		 2868 /* pre-format output fields to sane values */
2869 for (i = 0; i < same->dest_count; i++) {
2870 same->info[i].bytes_deduped = 0ULL;
2871 same->info[i].status = 0;
2872 }
416161db 2873
cbf8b8ca
MF		 2874 ret = 0;
2875 for (i = 0; i < same->dest_count; i++) {
2876 info = &same->info[i];
416161db 2877
cbf8b8ca 2878 dst_file = fget(info->fd);
416161db 2879 if (!dst_file) {
cbf8b8ca 2880 info->status = -EBADF;
416161db
MF		 2881 goto next;
2882 }
2883
2884 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
cbf8b8ca 2885 info->status = -EINVAL;
416161db
MF		 2886 goto next;
2887 }
2888
cbf8b8ca 2889 info->status = -EXDEV;
416161db
MF		 2890 if (file->f_path.mnt != dst_file->f_path.mnt)
2891 goto next;
2892
2893 dst = dst_file->f_dentry->d_inode;
2894 if (src->i_sb != dst->i_sb)
2895 goto next;
2896
2897 if (S_ISDIR(dst->i_mode)) {
cbf8b8ca 2898 info->status = -EISDIR;
416161db
MF		 2899 goto next;
2900 }
2901
2902 if (!S_ISREG(dst->i_mode)) {
cbf8b8ca 2903 info->status = -EACCES;
416161db
MF		 2904 goto next;
2905 }
2906
cbf8b8ca
MF		 2907 info->status = btrfs_extent_same(src, off, len, dst,
2908 info->logical_offset);
2909 if (info->status == 0)
2910 info->bytes_deduped += len;
416161db
MF		 2911
2912next:
2913 if (dst_file)
2914 fput(dst_file);
416161db
MF		 2915 }
2916
cbf8b8ca
MF		 2917 ret = copy_to_user(argp, same, size);
2918 if (ret)
2919 ret = -EFAULT;
2920
416161db
MF		 2921out:
2922 mnt_drop_write_file(file);
2923 return ret;
2924}
2925
32b7c687
MF		 2926/**
2927 * btrfs_clone() - clone a range from inode file to another
2928 *
2929 * @src: Inode to clone from
2930 * @inode: Inode to clone to
2931 * @off: Offset within source to start clone from
2932 * @olen: Original length, passed by user, of range to clone
2933 * @olen_aligned: Block-aligned value of olen, extent_same uses
2934 * identical values here
2935 * @destoff: Offset within @inode to start clone
2936 */
2937static int btrfs_clone(struct inode *src, struct inode *inode,
2938 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
f46b5a66 2939{
f46b5a66 2940 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 2941 struct btrfs_path *path = NULL;
f46b5a66 2942 struct extent_buffer *leaf;
32b7c687
MF		 2943 struct btrfs_trans_handle *trans;
2944 char *buf = NULL;
ae01a0ab 2945 struct btrfs_key key;
f46b5a66
CH		 2946 u32 nritems;
2947 int slot;
ae01a0ab 2948 int ret;
32b7c687 2949 u64 len = olen_aligned;
ae01a0ab
YZ		 2950
2951 ret = -ENOMEM;
2952 buf = vmalloc(btrfs_level_size(root, 0));
2953 if (!buf)
32b7c687 2954 return ret;
ae01a0ab
YZ		 2955
2956 path = btrfs_alloc_path();
2957 if (!path) {
2958 vfree(buf);
32b7c687 2959 return ret;
d525e8ab
LZ		 2960 }
2961
32b7c687 2962 path->reada = 2;
c5c9cd4d 2963 /* clone data */
33345d01 2964 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2965 key.type = BTRFS_EXTENT_DATA_KEY;
2966 key.offset = 0;
f46b5a66
CH		 2967
2968 while (1) {
2969 /*
2970 * note the key will change type as we walk through the
2971 * tree.
2972 */
e4355f34 2973 path->leave_spinning = 1;
362a20c5
DS		 2974 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2975 0, 0);
f46b5a66
CH		 2976 if (ret < 0)
2977 goto out;
2978
ae01a0ab 2979 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 2980process_slot:
ae01a0ab 2981 if (path->slots[0] >= nritems) {
362a20c5 2982 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 2983 if (ret < 0)
2984 goto out;
2985 if (ret > 0)
2986 break;
ae01a0ab 2987 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2988 }
2989 leaf = path->nodes[0];
2990 slot = path->slots[0];
f46b5a66 2991
ae01a0ab 2992 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2993 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2994 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2995 break;
2996
c5c9cd4d
SW		 2997 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2998 struct btrfs_file_extent_item *extent;
2999 int type;
31840ae1
ZY		 3000 u32 size;
3001 struct btrfs_key new_key;
c5c9cd4d
SW		 3002 u64 disko = 0, diskl = 0;
3003 u64 datao = 0, datal = 0;
3004 u8 comp;
b5384d48 3005 u64 endoff;
31840ae1 3006
c5c9cd4d
SW		 3007 extent = btrfs_item_ptr(leaf, slot,
3008 struct btrfs_file_extent_item);
3009 comp = btrfs_file_extent_compression(leaf, extent);
3010 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3011 if (type == BTRFS_FILE_EXTENT_REG ||
3012 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3013 disko = btrfs_file_extent_disk_bytenr(leaf,
3014 extent);
3015 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3016 extent);
c5c9cd4d 3017 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3018 datal = btrfs_file_extent_num_bytes(leaf,
3019 extent);
c5c9cd4d
SW		 3020 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3021 /* take upper bound, may be compressed */
3022 datal = btrfs_file_extent_ram_bytes(leaf,
3023 extent);
3024 }
31840ae1 3025
050006a7 3026 if (key.offset + datal <= off ||
e4355f34
FDBM		 3027 key.offset >= off + len - 1) {
3028 path->slots[0]++;
3029 goto process_slot;
3030 }
3031
3032 size = btrfs_item_size_nr(leaf, slot);
3033 read_extent_buffer(leaf, buf,
3034 btrfs_item_ptr_offset(leaf, slot),
3035 size);
3036
3037 btrfs_release_path(path);
3038 path->leave_spinning = 0;
c5c9cd4d 3039
31840ae1 3040 memcpy(&new_key, &key, sizeof(new_key));
33345d01 3041 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 3042 if (off <= key.offset)
3043 new_key.offset = key.offset + destoff - off;
3044 else
3045 new_key.offset = destoff;
31840ae1 3046
b6f3409b
SW		 3047 /*
3048 * 1 - adjusting old extent (we may have to split it)
3049 * 1 - add new extent
3050 * 1 - inode update
3051 */
3052 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3053 if (IS_ERR(trans)) {
3054 ret = PTR_ERR(trans);
3055 goto out;
3056 }
3057
c8a894d7
CM		 3058 if (type == BTRFS_FILE_EXTENT_REG ||
3059 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3060 /*
3061 * a | --- range to clone ---| b
3062 * | ------------- extent ------------- |
3063 */
3064
3065 /* substract range b */
3066 if (key.offset + datal > off + len)
3067 datal = off + len - key.offset;
3068
3069 /* substract range a */
a22285a6
YZ		 3070 if (off > key.offset) {
3071 datao += off - key.offset;
3072 datal -= off - key.offset;
3073 }
3074
5dc562c5 3075 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 3076 new_key.offset,
3077 new_key.offset + datal,
2671485d 3078 1);
79787eaa
JM		 3079 if (ret) {
3080 btrfs_abort_transaction(trans, root,
3081 ret);
3082 btrfs_end_transaction(trans, root);
3083 goto out;
3084 }
a22285a6 3085
c5c9cd4d
SW		 3086 ret = btrfs_insert_empty_item(trans, root, path,
3087 &new_key, size);
79787eaa
JM		 3088 if (ret) {
3089 btrfs_abort_transaction(trans, root,
3090 ret);
3091 btrfs_end_transaction(trans, root);
3092 goto out;
3093 }
c5c9cd4d
SW		 3094
3095 leaf = path->nodes[0];
3096 slot = path->slots[0];
3097 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3098 btrfs_item_ptr_offset(leaf, slot),
3099 size);
ae01a0ab 3100
c5c9cd4d 3101 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3102 struct btrfs_file_extent_item);
c5c9cd4d 3103
c5c9cd4d
SW		 3104 /* disko == 0 means it's a hole */
3105 if (!disko)
3106 datao = 0;
c5c9cd4d
SW		 3107
3108 btrfs_set_file_extent_offset(leaf, extent,
3109 datao);
3110 btrfs_set_file_extent_num_bytes(leaf, extent,
3111 datal);
3112 if (disko) {
3113 inode_add_bytes(inode, datal);
ae01a0ab 3114 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3115 disko, diskl, 0,
3116 root->root_key.objectid,
33345d01 3117 btrfs_ino(inode),
66d7e7f0
AJ		 3118 new_key.offset - datao,
3119 0);
79787eaa
JM		 3120 if (ret) {
3121 btrfs_abort_transaction(trans,
3122 root,
3123 ret);
3124 btrfs_end_transaction(trans,
3125 root);
3126 goto out;
3127
3128 }
f46b5a66 3129 }
c5c9cd4d
SW		 3130 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3131 u64 skip = 0;
3132 u64 trim = 0;
3133 if (off > key.offset) {
3134 skip = off - key.offset;
3135 new_key.offset += skip;
3136 }
d397712b 3137
aa42ffd9
LB		 3138 if (key.offset + datal > off + len)
3139 trim = key.offset + datal - (off + len);
d397712b 3140
c5c9cd4d 3141 if (comp && (skip || trim)) {
c5c9cd4d 3142 ret = -EINVAL;
a22285a6 3143 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3144 goto out;
3145 }
3146 size -= skip + trim;
3147 datal -= skip + trim;
a22285a6 3148
5dc562c5 3149 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 3150 new_key.offset,
3151 new_key.offset + datal,
2671485d 3152 1);
79787eaa
JM		 3153 if (ret) {
3154 btrfs_abort_transaction(trans, root,
3155 ret);
3156 btrfs_end_transaction(trans, root);
3157 goto out;
3158 }
a22285a6 3159
c5c9cd4d
SW		 3160 ret = btrfs_insert_empty_item(trans, root, path,
3161 &new_key, size);
79787eaa
JM		 3162 if (ret) {
3163 btrfs_abort_transaction(trans, root,
3164 ret);
3165 btrfs_end_transaction(trans, root);
3166 goto out;
3167 }
c5c9cd4d
SW		 3168
3169 if (skip) {
d397712b
CM		 3170 u32 start =
3171 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3172 memmove(buf+start, buf+start+skip,
3173 datal);
3174 }
3175
3176 leaf = path->nodes[0];
3177 slot = path->slots[0];
3178 write_extent_buffer(leaf, buf,
3179 btrfs_item_ptr_offset(leaf, slot),
3180 size);
3181 inode_add_bytes(inode, datal);
f46b5a66 3182 }
c5c9cd4d
SW		 3183
3184 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3185 btrfs_release_path(path);
c5c9cd4d 3186
0c4d2d95 3187 inode_inc_iversion(inode);
a22285a6 3188 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 3189
3190 /*
3191 * we round up to the block size at eof when
3192 * determining which extents to clone above,
3193 * but shouldn't round up the file size
3194 */
3195 endoff = new_key.offset + datal;
5f3888ff
LZ		 3196 if (endoff > destoff+olen)
3197 endoff = destoff+olen;
b5384d48
SW		 3198 if (endoff > inode->i_size)
3199 btrfs_i_size_write(inode, endoff);
3200
a22285a6 3201 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 3202 if (ret) {
3203 btrfs_abort_transaction(trans, root, ret);
3204 btrfs_end_transaction(trans, root);
3205 goto out;
3206 }
3207 ret = btrfs_end_transaction(trans, root);
a22285a6 3208 }
b3b4aa74 3209 btrfs_release_path(path);
f46b5a66 3210 key.offset++;
f46b5a66 3211 }
f46b5a66 3212 ret = 0;
32b7c687 3213
f46b5a66 3214out:
b3b4aa74 3215 btrfs_release_path(path);
32b7c687
MF		 3216 btrfs_free_path(path);
3217 vfree(buf);
3218 return ret;
3219}
3220
3221static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3222 u64 off, u64 olen, u64 destoff)
3223{
54563d41 3224 struct inode *inode = file_inode(file);
32b7c687
MF		 3225 struct btrfs_root *root = BTRFS_I(inode)->root;
3226 struct fd src_file;
3227 struct inode *src;
3228 int ret;
3229 u64 len = olen;
3230 u64 bs = root->fs_info->sb->s_blocksize;
3231 int same_inode = 0;
3232
3233 /*
3234 * TODO:
3235 * - split compressed inline extents. annoying: we need to
3236 * decompress into destination's address_space (the file offset
3237 * may change, so source mapping won't do), then recompress (or
3238 * otherwise reinsert) a subrange.
3239 * - allow ranges within the same file to be cloned (provided
3240 * they don't overlap)?
3241 */
3242
3243 /* the destination must be opened for writing */
3244 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3245 return -EINVAL;
3246
3247 if (btrfs_root_readonly(root))
3248 return -EROFS;
3249
3250 ret = mnt_want_write_file(file);
3251 if (ret)
3252 return ret;
3253
3254 src_file = fdget(srcfd);
3255 if (!src_file.file) {
3256 ret = -EBADF;
3257 goto out_drop_write;
3258 }
3259
3260 ret = -EXDEV;
3261 if (src_file.file->f_path.mnt != file->f_path.mnt)
3262 goto out_fput;
3263
3264 src = file_inode(src_file.file);
3265
3266 ret = -EINVAL;
3267 if (src == inode)
3268 same_inode = 1;
3269
3270 /* the src must be open for reading */
3271 if (!(src_file.file->f_mode & FMODE_READ))
3272 goto out_fput;
3273
3274 /* don't make the dst file partly checksummed */
3275 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3276 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3277 goto out_fput;
3278
3279 ret = -EISDIR;
3280 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3281 goto out_fput;
3282
3283 ret = -EXDEV;
3284 if (src->i_sb != inode->i_sb)
3285 goto out_fput;
3286
3287 if (!same_inode) {
3288 if (inode < src) {
3289 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3290 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3291 } else {
3292 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3293 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3294 }
3295 } else {
3296 mutex_lock(&src->i_mutex);
3297 }
3298
3299 /* determine range to clone */
3300 ret = -EINVAL;
3301 if (off + len > src->i_size || off + len < off)
3302 goto out_unlock;
3303 if (len == 0)
3304 olen = len = src->i_size - off;
3305 /* if we extend to eof, continue to block boundary */
3306 if (off + len == src->i_size)
3307 len = ALIGN(src->i_size, bs) - off;
3308
3309 /* verify the end result is block aligned */
3310 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3311 !IS_ALIGNED(destoff, bs))
3312 goto out_unlock;
3313
3314 /* verify if ranges are overlapped within the same file */
3315 if (same_inode) {
3316 if (destoff + len > off && destoff < off + len)
3317 goto out_unlock;
3318 }
3319
3320 if (destoff > inode->i_size) {
3321 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3322 if (ret)
3323 goto out_unlock;
3324 }
3325
3326 /* truncate page cache pages from target inode range */
3327 truncate_inode_pages_range(&inode->i_data, destoff,
3328 PAGE_CACHE_ALIGN(destoff + len) - 1);
3329
3330 lock_extent_range(src, off, len);
3331
3332 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3333
aa42ffd9 3334 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
f46b5a66 3335out_unlock:
c57c2b3e
FDBM		 3336 if (!same_inode) {
3337 if (inode < src) {
3338 mutex_unlock(&src->i_mutex);
3339 mutex_unlock(&inode->i_mutex);
3340 } else {
3341 mutex_unlock(&inode->i_mutex);
3342 mutex_unlock(&src->i_mutex);
3343 }
3344 } else {
3345 mutex_unlock(&src->i_mutex);
3346 }
f46b5a66 3347out_fput:
2903ff01 3348 fdput(src_file);
ab67b7c1 3349out_drop_write:
2a79f17e 3350 mnt_drop_write_file(file);
f46b5a66
CH		 3351 return ret;
3352}
3353
7a865e8a 3354static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3355{
3356 struct btrfs_ioctl_clone_range_args args;
3357
7a865e8a 3358 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3359 return -EFAULT;
3360 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3361 args.src_length, args.dest_offset);
3362}
3363
f46b5a66
CH		 3364/*
3365 * there are many ways the trans_start and trans_end ioctls can lead
3366 * to deadlocks. They should only be used by applications that
3367 * basically own the machine, and have a very in depth understanding
3368 * of all the possible deadlocks and enospc problems.
3369 */
b2950863 3370static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3371{
496ad9aa 3372 struct inode *inode = file_inode(file);
f46b5a66
CH		 3373 struct btrfs_root *root = BTRFS_I(inode)->root;
3374 struct btrfs_trans_handle *trans;
1ab86aed 3375 int ret;
f46b5a66 3376
1ab86aed 3377 ret = -EPERM;
df5b5520 3378 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3379 goto out;
df5b5520 3380
1ab86aed
SW		 3381 ret = -EINPROGRESS;
3382 if (file->private_data)
f46b5a66 3383 goto out;
9ca9ee09 3384
b83cc969
LZ		 3385 ret = -EROFS;
3386 if (btrfs_root_readonly(root))
3387 goto out;
3388
a561be71 3389 ret = mnt_want_write_file(file);
c146afad
YZ		 3390 if (ret)
3391 goto out;
3392
a4abeea4 3393 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3394
1ab86aed 3395 ret = -ENOMEM;
7a7eaa40 3396 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3397 if (IS_ERR(trans))
1ab86aed
SW		 3398 goto out_drop;
3399
3400 file->private_data = trans;
3401 return 0;
3402
3403out_drop:
a4abeea4 3404 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3405 mnt_drop_write_file(file);
f46b5a66 3406out:
f46b5a66
CH		 3407 return ret;
3408}
3409
6ef5ed0d
JB		 3410static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3411{
496ad9aa 3412 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3413 struct btrfs_root *root = BTRFS_I(inode)->root;
3414 struct btrfs_root *new_root;
3415 struct btrfs_dir_item *di;
3416 struct btrfs_trans_handle *trans;
3417 struct btrfs_path *path;
3418 struct btrfs_key location;
3419 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3420 u64 objectid = 0;
3421 u64 dir_id;
3c04ce01 3422 int ret;
6ef5ed0d
JB		 3423
3424 if (!capable(CAP_SYS_ADMIN))
3425 return -EPERM;
3426
3c04ce01
MX		 3427 ret = mnt_want_write_file(file);
3428 if (ret)
3429 return ret;
3430
3431 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3432 ret = -EFAULT;
3433 goto out;
3434 }
6ef5ed0d
JB		 3435
3436 if (!objectid)
1cecf579 3437 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3438
3439 location.objectid = objectid;
3440 location.type = BTRFS_ROOT_ITEM_KEY;
3441 location.offset = (u64)-1;
3442
3443 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3444 if (IS_ERR(new_root)) {
3445 ret = PTR_ERR(new_root);
3446 goto out;
3447 }
6ef5ed0d 3448
6ef5ed0d 3449 path = btrfs_alloc_path();
3c04ce01
MX		 3450 if (!path) {
3451 ret = -ENOMEM;
3452 goto out;
3453 }
6ef5ed0d
JB		 3454 path->leave_spinning = 1;
3455
3456 trans = btrfs_start_transaction(root, 1);
98d5dc13 3457 if (IS_ERR(trans)) {
6ef5ed0d 3458 btrfs_free_path(path);
3c04ce01
MX		 3459 ret = PTR_ERR(trans);
3460 goto out;
6ef5ed0d
JB		 3461 }
3462
6c41761f 3463 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3464 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3465 dir_id, "default", 7, 1);
cf1e99a4 3466 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3467 btrfs_free_path(path);
3468 btrfs_end_transaction(trans, root);
efe120a0
FH		 3469 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3470 "item, this isn't going to work");
3c04ce01
MX		 3471 ret = -ENOENT;
3472 goto out;
6ef5ed0d
JB		 3473 }
3474
3475 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3476 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3477 btrfs_mark_buffer_dirty(path->nodes[0]);
3478 btrfs_free_path(path);
3479
2b0ce2c2 3480 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3481 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3482out:
3483 mnt_drop_write_file(file);
3484 return ret;
6ef5ed0d
JB		 3485}
3486
5af3e8cc
SB		 3487void btrfs_get_block_group_info(struct list_head *groups_list,
3488 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3489{
3490 struct btrfs_block_group_cache *block_group;
3491
3492 space->total_bytes = 0;
3493 space->used_bytes = 0;
3494 space->flags = 0;
3495 list_for_each_entry(block_group, groups_list, list) {
3496 space->flags = block_group->flags;
3497 space->total_bytes += block_group->key.offset;
3498 space->used_bytes +=
3499 btrfs_block_group_used(&block_group->item);
3500 }
3501}
3502
48a3b636 3503static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3504{
3505 struct btrfs_ioctl_space_args space_args;
3506 struct btrfs_ioctl_space_info space;
3507 struct btrfs_ioctl_space_info *dest;
7fde62bf 3508 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3509 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3510 struct btrfs_space_info *info;
bf5fc093
JB		 3511 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3512 BTRFS_BLOCK_GROUP_SYSTEM,
3513 BTRFS_BLOCK_GROUP_METADATA,
3514 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3515 int num_types = 4;
7fde62bf 3516 int alloc_size;
1406e432 3517 int ret = 0;
51788b1b 3518 u64 slot_count = 0;
bf5fc093 3519 int i, c;
1406e432
JB		 3520
3521 if (copy_from_user(&space_args,
3522 (struct btrfs_ioctl_space_args __user *)arg,
3523 sizeof(space_args)))
3524 return -EFAULT;
3525
bf5fc093
JB		 3526 for (i = 0; i < num_types; i++) {
3527 struct btrfs_space_info *tmp;
3528
3529 info = NULL;
3530 rcu_read_lock();
3531 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3532 list) {
3533 if (tmp->flags == types[i]) {
3534 info = tmp;
3535 break;
3536 }
3537 }
3538 rcu_read_unlock();
3539
3540 if (!info)
3541 continue;
3542
3543 down_read(&info->groups_sem);
3544 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3545 if (!list_empty(&info->block_groups[c]))
3546 slot_count++;
3547 }
3548 up_read(&info->groups_sem);
3549 }
7fde62bf
CM		 3550
3551 /* space_slots == 0 means they are asking for a count */
3552 if (space_args.space_slots == 0) {
3553 space_args.total_spaces = slot_count;
3554 goto out;
3555 }
bf5fc093 3556
51788b1b 3557 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3558
7fde62bf 3559 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3560
7fde62bf
CM		 3561 /* we generally have at most 6 or so space infos, one for each raid
3562 * level. So, a whole page should be more than enough for everyone
3563 */
3564 if (alloc_size > PAGE_CACHE_SIZE)
3565 return -ENOMEM;
3566
1406e432 3567 space_args.total_spaces = 0;
7fde62bf
CM		 3568 dest = kmalloc(alloc_size, GFP_NOFS);
3569 if (!dest)
3570 return -ENOMEM;
3571 dest_orig = dest;
1406e432 3572
7fde62bf 3573 /* now we have a buffer to copy into */
bf5fc093
JB		 3574 for (i = 0; i < num_types; i++) {
3575 struct btrfs_space_info *tmp;
3576
51788b1b
DR		 3577 if (!slot_count)
3578 break;
3579
bf5fc093
JB		 3580 info = NULL;
3581 rcu_read_lock();
3582 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3583 list) {
3584 if (tmp->flags == types[i]) {
3585 info = tmp;
3586 break;
3587 }
3588 }
3589 rcu_read_unlock();
7fde62bf 3590
bf5fc093
JB		 3591 if (!info)
3592 continue;
3593 down_read(&info->groups_sem);
3594 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3595 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3596 btrfs_get_block_group_info(
3597 &info->block_groups[c], &space);
bf5fc093
JB		 3598 memcpy(dest, &space, sizeof(space));
3599 dest++;
3600 space_args.total_spaces++;
51788b1b 3601 slot_count--;
bf5fc093 3602 }
51788b1b
DR		 3603 if (!slot_count)
3604 break;
bf5fc093
JB		 3605 }
3606 up_read(&info->groups_sem);
1406e432 3607 }
1406e432 3608
2eec6c81 3609 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3610 (arg + sizeof(struct btrfs_ioctl_space_args));
3611
3612 if (copy_to_user(user_dest, dest_orig, alloc_size))
3613 ret = -EFAULT;
3614
3615 kfree(dest_orig);
3616out:
3617 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 3618 ret = -EFAULT;
3619
3620 return ret;
3621}
3622
f46b5a66
CH		 3623/*
3624 * there are many ways the trans_start and trans_end ioctls can lead
3625 * to deadlocks. They should only be used by applications that
3626 * basically own the machine, and have a very in depth understanding
3627 * of all the possible deadlocks and enospc problems.
3628 */
3629long btrfs_ioctl_trans_end(struct file *file)
3630{
496ad9aa 3631 struct inode *inode = file_inode(file);
f46b5a66
CH		 3632 struct btrfs_root *root = BTRFS_I(inode)->root;
3633 struct btrfs_trans_handle *trans;
f46b5a66 3634
f46b5a66 3635 trans = file->private_data;
1ab86aed
SW		 3636 if (!trans)
3637 return -EINVAL;
b214107e 3638 file->private_data = NULL;
9ca9ee09 3639
1ab86aed
SW		 3640 btrfs_end_transaction(trans, root);
3641
a4abeea4 3642 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 3643
2a79f17e 3644 mnt_drop_write_file(file);
1ab86aed 3645 return 0;
f46b5a66
CH		 3646}
3647
9a8c28be
MX		 3648static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3649 void __user *argp)
46204592 3650{
46204592
SW		 3651 struct btrfs_trans_handle *trans;
3652 u64 transid;
db5b493a 3653 int ret;
46204592 3654
d4edf39b 3655 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 3656 if (IS_ERR(trans)) {
3657 if (PTR_ERR(trans) != -ENOENT)
3658 return PTR_ERR(trans);
3659
3660 /* No running transaction, don't bother */
3661 transid = root->fs_info->last_trans_committed;
3662 goto out;
3663 }
46204592 3664 transid = trans->transid;
db5b493a 3665 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 3666 if (ret) {
3667 btrfs_end_transaction(trans, root);
db5b493a 3668 return ret;
8b2b2d3c 3669 }
ff7c1d33 3670out:
46204592
SW		 3671 if (argp)
3672 if (copy_to_user(argp, &transid, sizeof(transid)))
3673 return -EFAULT;
3674 return 0;
3675}
3676
9a8c28be
MX		 3677static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3678 void __user *argp)
46204592 3679{
46204592
SW		 3680 u64 transid;
3681
3682 if (argp) {
3683 if (copy_from_user(&transid, argp, sizeof(transid)))
3684 return -EFAULT;
3685 } else {
3686 transid = 0; /* current trans */
3687 }
3688 return btrfs_wait_for_commit(root, transid);
3689}
3690
b8e95489 3691static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 3692{
496ad9aa 3693 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 3694 struct btrfs_ioctl_scrub_args *sa;
b8e95489 3695 int ret;
475f6387
JS		 3696
3697 if (!capable(CAP_SYS_ADMIN))
3698 return -EPERM;
3699
3700 sa = memdup_user(arg, sizeof(*sa));
3701 if (IS_ERR(sa))
3702 return PTR_ERR(sa);
3703
b8e95489
MX		 3704 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3705 ret = mnt_want_write_file(file);
3706 if (ret)
3707 goto out;
3708 }
3709
aa1b8cd4 3710 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 3711 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3712 0);
475f6387
JS		 3713
3714 if (copy_to_user(arg, sa, sizeof(*sa)))
3715 ret = -EFAULT;
3716
b8e95489
MX		 3717 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3718 mnt_drop_write_file(file);
3719out:
475f6387
JS		 3720 kfree(sa);
3721 return ret;
3722}
3723
3724static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3725{
3726 if (!capable(CAP_SYS_ADMIN))
3727 return -EPERM;
3728
aa1b8cd4 3729 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 3730}
3731
3732static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3733 void __user *arg)
3734{
3735 struct btrfs_ioctl_scrub_args *sa;
3736 int ret;
3737
3738 if (!capable(CAP_SYS_ADMIN))
3739 return -EPERM;
3740
3741 sa = memdup_user(arg, sizeof(*sa));
3742 if (IS_ERR(sa))
3743 return PTR_ERR(sa);
3744
3745 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3746
3747 if (copy_to_user(arg, sa, sizeof(*sa)))
3748 ret = -EFAULT;
3749
3750 kfree(sa);
3751 return ret;
3752}
3753
c11d2c23 3754static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 3755 void __user *arg)
c11d2c23
SB		 3756{
3757 struct btrfs_ioctl_get_dev_stats *sa;
3758 int ret;
3759
c11d2c23
SB		 3760 sa = memdup_user(arg, sizeof(*sa));
3761 if (IS_ERR(sa))
3762 return PTR_ERR(sa);
3763
b27f7c0c
DS		 3764 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3765 kfree(sa);
3766 return -EPERM;
3767 }
3768
3769 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 3770
3771 if (copy_to_user(arg, sa, sizeof(*sa)))
3772 ret = -EFAULT;
3773
3774 kfree(sa);
3775 return ret;
3776}
3777
3f6bcfbd
SB		 3778static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3779{
3780 struct btrfs_ioctl_dev_replace_args *p;
3781 int ret;
3782
3783 if (!capable(CAP_SYS_ADMIN))
3784 return -EPERM;
3785
3786 p = memdup_user(arg, sizeof(*p));
3787 if (IS_ERR(p))
3788 return PTR_ERR(p);
3789
3790 switch (p->cmd) {
3791 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
adfa97cb
ID		 3792 if (root->fs_info->sb->s_flags & MS_RDONLY) {
3793 ret = -EROFS;
3794 goto out;
3795 }
3f6bcfbd
SB		 3796 if (atomic_xchg(
3797 &root->fs_info->mutually_exclusive_operation_running,
3798 1)) {
e57138b3 3799 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd
SB		 3800 } else {
3801 ret = btrfs_dev_replace_start(root, p);
3802 atomic_set(
3803 &root->fs_info->mutually_exclusive_operation_running,
3804 0);
3805 }
3806 break;
3807 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3808 btrfs_dev_replace_status(root->fs_info, p);
3809 ret = 0;
3810 break;
3811 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3812 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3813 break;
3814 default:
3815 ret = -EINVAL;
3816 break;
3817 }
3818
3819 if (copy_to_user(arg, p, sizeof(*p)))
3820 ret = -EFAULT;
adfa97cb 3821out:
3f6bcfbd
SB		 3822 kfree(p);
3823 return ret;
3824}
3825
d7728c96
JS		 3826static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3827{
3828 int ret = 0;
3829 int i;
740c3d22 3830 u64 rel_ptr;
d7728c96 3831 int size;
806468f8 3832 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3833 struct inode_fs_paths *ipath = NULL;
3834 struct btrfs_path *path;
3835
82b22ac8 3836 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 3837 return -EPERM;
3838
3839 path = btrfs_alloc_path();
3840 if (!path) {
3841 ret = -ENOMEM;
3842 goto out;
3843 }
3844
3845 ipa = memdup_user(arg, sizeof(*ipa));
3846 if (IS_ERR(ipa)) {
3847 ret = PTR_ERR(ipa);
3848 ipa = NULL;
3849 goto out;
3850 }
3851
3852 size = min_t(u32, ipa->size, 4096);
3853 ipath = init_ipath(size, root, path);
3854 if (IS_ERR(ipath)) {
3855 ret = PTR_ERR(ipath);
3856 ipath = NULL;
3857 goto out;
3858 }
3859
3860 ret = paths_from_inode(ipa->inum, ipath);
3861 if (ret < 0)
3862 goto out;
3863
3864 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3865 rel_ptr = ipath->fspath->val[i] -
3866 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3867 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3868 }
3869
745c4d8e
JM		 3870 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3871 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3872 if (ret) {
3873 ret = -EFAULT;
3874 goto out;
3875 }
3876
3877out:
3878 btrfs_free_path(path);
3879 free_ipath(ipath);
3880 kfree(ipa);
3881
3882 return ret;
3883}
3884
3885static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3886{
3887 struct btrfs_data_container *inodes = ctx;
3888 const size_t c = 3 * sizeof(u64);
3889
3890 if (inodes->bytes_left >= c) {
3891 inodes->bytes_left -= c;
3892 inodes->val[inodes->elem_cnt] = inum;
3893 inodes->val[inodes->elem_cnt + 1] = offset;
3894 inodes->val[inodes->elem_cnt + 2] = root;
3895 inodes->elem_cnt += 3;
3896 } else {
3897 inodes->bytes_missing += c - inodes->bytes_left;
3898 inodes->bytes_left = 0;
3899 inodes->elem_missed += 3;
3900 }
3901
3902 return 0;
3903}
3904
3905static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3906 void __user *arg)
3907{
3908 int ret = 0;
3909 int size;
d7728c96
JS		 3910 struct btrfs_ioctl_logical_ino_args *loi;
3911 struct btrfs_data_container *inodes = NULL;
3912 struct btrfs_path *path = NULL;
d7728c96
JS		 3913
3914 if (!capable(CAP_SYS_ADMIN))
3915 return -EPERM;
3916
3917 loi = memdup_user(arg, sizeof(*loi));
3918 if (IS_ERR(loi)) {
3919 ret = PTR_ERR(loi);
3920 loi = NULL;
3921 goto out;
3922 }
3923
3924 path = btrfs_alloc_path();
3925 if (!path) {
3926 ret = -ENOMEM;
3927 goto out;
3928 }
3929
425d17a2 3930 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 3931 inodes = init_data_container(size);
3932 if (IS_ERR(inodes)) {
3933 ret = PTR_ERR(inodes);
3934 inodes = NULL;
3935 goto out;
3936 }
3937
df031f07
LB		 3938 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3939 build_ino_list, inodes);
3940 if (ret == -EINVAL)
d7728c96
JS		 3941 ret = -ENOENT;
3942 if (ret < 0)
3943 goto out;
3944
745c4d8e
JM		 3945 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3946 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3947 if (ret)
3948 ret = -EFAULT;
3949
3950out:
3951 btrfs_free_path(path);
425d17a2 3952 vfree(inodes);
d7728c96
JS		 3953 kfree(loi);
3954
3955 return ret;
3956}
3957
19a39dce 3958void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3959 struct btrfs_ioctl_balance_args *bargs)
3960{
3961 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3962
3963 bargs->flags = bctl->flags;
3964
837d5b6e
ID		 3965 if (atomic_read(&fs_info->balance_running))
3966 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3967 if (atomic_read(&fs_info->balance_pause_req))
3968 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3969 if (atomic_read(&fs_info->balance_cancel_req))
3970 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3971
c9e9f97b
ID		 3972 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3973 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3974 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3975
3976 if (lock) {
3977 spin_lock(&fs_info->balance_lock);
3978 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3979 spin_unlock(&fs_info->balance_lock);
3980 } else {
3981 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3982 }
c9e9f97b
ID		 3983}
3984
9ba1f6e4 3985static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3986{
496ad9aa 3987 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 3988 struct btrfs_fs_info *fs_info = root->fs_info;
3989 struct btrfs_ioctl_balance_args *bargs;
3990 struct btrfs_balance_control *bctl;
ed0fb78f 3991 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 3992 int ret;
3993
3994 if (!capable(CAP_SYS_ADMIN))
3995 return -EPERM;
3996
e54bfa31 3997 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 3998 if (ret)
3999 return ret;
4000
ed0fb78f
ID		 4001again:
4002 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4003 mutex_lock(&fs_info->volume_mutex);
4004 mutex_lock(&fs_info->balance_mutex);
4005 need_unlock = true;
4006 goto locked;
4007 }
4008
4009 /*
4010 * mut. excl. ops lock is locked. Three possibilites:
4011 * (1) some other op is running
4012 * (2) balance is running
4013 * (3) balance is paused -- special case (think resume)
4014 */
c9e9f97b 4015 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4016 if (fs_info->balance_ctl) {
4017 /* this is either (2) or (3) */
4018 if (!atomic_read(&fs_info->balance_running)) {
4019 mutex_unlock(&fs_info->balance_mutex);
4020 if (!mutex_trylock(&fs_info->volume_mutex))
4021 goto again;
4022 mutex_lock(&fs_info->balance_mutex);
4023
4024 if (fs_info->balance_ctl &&
4025 !atomic_read(&fs_info->balance_running)) {
4026 /* this is (3) */
4027 need_unlock = false;
4028 goto locked;
4029 }
4030
4031 mutex_unlock(&fs_info->balance_mutex);
4032 mutex_unlock(&fs_info->volume_mutex);
4033 goto again;
4034 } else {
4035 /* this is (2) */
4036 mutex_unlock(&fs_info->balance_mutex);
4037 ret = -EINPROGRESS;
4038 goto out;
4039 }
4040 } else {
4041 /* this is (1) */
4042 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4043 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4044 goto out;
4045 }
4046
4047locked:
4048 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 4049
4050 if (arg) {
4051 bargs = memdup_user(arg, sizeof(*bargs));
4052 if (IS_ERR(bargs)) {
4053 ret = PTR_ERR(bargs);
ed0fb78f 4054 goto out_unlock;
c9e9f97b 4055 }
de322263
ID		 4056
4057 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4058 if (!fs_info->balance_ctl) {
4059 ret = -ENOTCONN;
4060 goto out_bargs;
4061 }
4062
4063 bctl = fs_info->balance_ctl;
4064 spin_lock(&fs_info->balance_lock);
4065 bctl->flags |= BTRFS_BALANCE_RESUME;
4066 spin_unlock(&fs_info->balance_lock);
4067
4068 goto do_balance;
4069 }
c9e9f97b
ID		 4070 } else {
4071 bargs = NULL;
4072 }
4073
ed0fb78f 4074 if (fs_info->balance_ctl) {
837d5b6e
ID		 4075 ret = -EINPROGRESS;
4076 goto out_bargs;
4077 }
4078
c9e9f97b
ID		 4079 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4080 if (!bctl) {
4081 ret = -ENOMEM;
4082 goto out_bargs;
4083 }
4084
4085 bctl->fs_info = fs_info;
4086 if (arg) {
4087 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4088 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4089 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4090
4091 bctl->flags = bargs->flags;
f43ffb60
ID		 4092 } else {
4093 /* balance everything - no filters */
4094 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4095 }
4096
de322263 4097do_balance:
c9e9f97b 4098 /*
ed0fb78f
ID		 4099 * Ownership of bctl and mutually_exclusive_operation_running
4100 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4101 * or, if restriper was paused all the way until unmount, in
4102 * free_fs_info. mutually_exclusive_operation_running is
4103 * cleared in __cancel_balance.
c9e9f97b 4104 */
ed0fb78f
ID		 4105 need_unlock = false;
4106
4107 ret = btrfs_balance(bctl, bargs);
4108
c9e9f97b
ID		 4109 if (arg) {
4110 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4111 ret = -EFAULT;
4112 }
4113
4114out_bargs:
4115 kfree(bargs);
ed0fb78f 4116out_unlock:
c9e9f97b
ID		 4117 mutex_unlock(&fs_info->balance_mutex);
4118 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 4119 if (need_unlock)
4120 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4121out:
e54bfa31 4122 mnt_drop_write_file(file);
c9e9f97b
ID		 4123 return ret;
4124}
4125
837d5b6e
ID		 4126static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4127{
4128 if (!capable(CAP_SYS_ADMIN))
4129 return -EPERM;
4130
4131 switch (cmd) {
4132 case BTRFS_BALANCE_CTL_PAUSE:
4133 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4134 case BTRFS_BALANCE_CTL_CANCEL:
4135 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4136 }
4137
4138 return -EINVAL;
4139}
4140
19a39dce
ID		 4141static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4142 void __user *arg)
4143{
4144 struct btrfs_fs_info *fs_info = root->fs_info;
4145 struct btrfs_ioctl_balance_args *bargs;
4146 int ret = 0;
4147
4148 if (!capable(CAP_SYS_ADMIN))
4149 return -EPERM;
4150
4151 mutex_lock(&fs_info->balance_mutex);
4152 if (!fs_info->balance_ctl) {
4153 ret = -ENOTCONN;
4154 goto out;
4155 }
4156
4157 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4158 if (!bargs) {
4159 ret = -ENOMEM;
4160 goto out;
4161 }
4162
4163 update_ioctl_balance_args(fs_info, 1, bargs);
4164
4165 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4166 ret = -EFAULT;
4167
4168 kfree(bargs);
4169out:
4170 mutex_unlock(&fs_info->balance_mutex);
4171 return ret;
4172}
4173
905b0dda 4174static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4175{
496ad9aa 4176 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4177 struct btrfs_ioctl_quota_ctl_args *sa;
4178 struct btrfs_trans_handle *trans = NULL;
4179 int ret;
4180 int err;
4181
4182 if (!capable(CAP_SYS_ADMIN))
4183 return -EPERM;
4184
905b0dda
MX		 4185 ret = mnt_want_write_file(file);
4186 if (ret)
4187 return ret;
5d13a37b
AJ		 4188
4189 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4190 if (IS_ERR(sa)) {
4191 ret = PTR_ERR(sa);
4192 goto drop_write;
4193 }
5d13a37b 4194
7708f029 4195 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4196 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4197 if (IS_ERR(trans)) {
4198 ret = PTR_ERR(trans);
4199 goto out;
5d13a37b
AJ		 4200 }
4201
4202 switch (sa->cmd) {
4203 case BTRFS_QUOTA_CTL_ENABLE:
4204 ret = btrfs_quota_enable(trans, root->fs_info);
4205 break;
4206 case BTRFS_QUOTA_CTL_DISABLE:
4207 ret = btrfs_quota_disable(trans, root->fs_info);
4208 break;
5d13a37b
AJ		 4209 default:
4210 ret = -EINVAL;
4211 break;
4212 }
4213
2f232036
JS		 4214 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4215 if (err && !ret)
4216 ret = err;
5d13a37b
AJ		 4217out:
4218 kfree(sa);
7708f029 4219 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4220drop_write:
4221 mnt_drop_write_file(file);
5d13a37b
AJ		 4222 return ret;
4223}
4224
905b0dda 4225static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4226{
496ad9aa 4227 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4228 struct btrfs_ioctl_qgroup_assign_args *sa;
4229 struct btrfs_trans_handle *trans;
4230 int ret;
4231 int err;
4232
4233 if (!capable(CAP_SYS_ADMIN))
4234 return -EPERM;
4235
905b0dda
MX		 4236 ret = mnt_want_write_file(file);
4237 if (ret)
4238 return ret;
5d13a37b
AJ		 4239
4240 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4241 if (IS_ERR(sa)) {
4242 ret = PTR_ERR(sa);
4243 goto drop_write;
4244 }
5d13a37b
AJ		 4245
4246 trans = btrfs_join_transaction(root);
4247 if (IS_ERR(trans)) {
4248 ret = PTR_ERR(trans);
4249 goto out;
4250 }
4251
4252 /* FIXME: check if the IDs really exist */
4253 if (sa->assign) {
4254 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4255 sa->src, sa->dst);
4256 } else {
4257 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4258 sa->src, sa->dst);
4259 }
4260
4261 err = btrfs_end_transaction(trans, root);
4262 if (err && !ret)
4263 ret = err;
4264
4265out:
4266 kfree(sa);
905b0dda
MX		 4267drop_write:
4268 mnt_drop_write_file(file);
5d13a37b
AJ		 4269 return ret;
4270}
4271
905b0dda 4272static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4273{
496ad9aa 4274 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4275 struct btrfs_ioctl_qgroup_create_args *sa;
4276 struct btrfs_trans_handle *trans;
4277 int ret;
4278 int err;
4279
4280 if (!capable(CAP_SYS_ADMIN))
4281 return -EPERM;
4282
905b0dda
MX		 4283 ret = mnt_want_write_file(file);
4284 if (ret)
4285 return ret;
5d13a37b
AJ		 4286
4287 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4288 if (IS_ERR(sa)) {
4289 ret = PTR_ERR(sa);
4290 goto drop_write;
4291 }
5d13a37b 4292
d86e56cf
MX		 4293 if (!sa->qgroupid) {
4294 ret = -EINVAL;
4295 goto out;
4296 }
4297
5d13a37b
AJ		 4298 trans = btrfs_join_transaction(root);
4299 if (IS_ERR(trans)) {
4300 ret = PTR_ERR(trans);
4301 goto out;
4302 }
4303
4304 /* FIXME: check if the IDs really exist */
4305 if (sa->create) {
4306 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4307 NULL);
4308 } else {
4309 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4310 }
4311
4312 err = btrfs_end_transaction(trans, root);
4313 if (err && !ret)
4314 ret = err;
4315
4316out:
4317 kfree(sa);
905b0dda
MX		 4318drop_write:
4319 mnt_drop_write_file(file);
5d13a37b
AJ		 4320 return ret;
4321}
4322
905b0dda 4323static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4324{
496ad9aa 4325 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4326 struct btrfs_ioctl_qgroup_limit_args *sa;
4327 struct btrfs_trans_handle *trans;
4328 int ret;
4329 int err;
4330 u64 qgroupid;
4331
4332 if (!capable(CAP_SYS_ADMIN))
4333 return -EPERM;
4334
905b0dda
MX		 4335 ret = mnt_want_write_file(file);
4336 if (ret)
4337 return ret;
5d13a37b
AJ		 4338
4339 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4340 if (IS_ERR(sa)) {
4341 ret = PTR_ERR(sa);
4342 goto drop_write;
4343 }
5d13a37b
AJ		 4344
4345 trans = btrfs_join_transaction(root);
4346 if (IS_ERR(trans)) {
4347 ret = PTR_ERR(trans);
4348 goto out;
4349 }
4350
4351 qgroupid = sa->qgroupid;
4352 if (!qgroupid) {
4353 /* take the current subvol as qgroup */
4354 qgroupid = root->root_key.objectid;
4355 }
4356
4357 /* FIXME: check if the IDs really exist */
4358 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4359
4360 err = btrfs_end_transaction(trans, root);
4361 if (err && !ret)
4362 ret = err;
4363
4364out:
4365 kfree(sa);
905b0dda
MX		 4366drop_write:
4367 mnt_drop_write_file(file);
5d13a37b
AJ		 4368 return ret;
4369}
4370
2f232036
JS		 4371static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4372{
6d0379ec 4373 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4374 struct btrfs_ioctl_quota_rescan_args *qsa;
4375 int ret;
4376
4377 if (!capable(CAP_SYS_ADMIN))
4378 return -EPERM;
4379
4380 ret = mnt_want_write_file(file);
4381 if (ret)
4382 return ret;
4383
4384 qsa = memdup_user(arg, sizeof(*qsa));
4385 if (IS_ERR(qsa)) {
4386 ret = PTR_ERR(qsa);
4387 goto drop_write;
4388 }
4389
4390 if (qsa->flags) {
4391 ret = -EINVAL;
4392 goto out;
4393 }
4394
4395 ret = btrfs_qgroup_rescan(root->fs_info);
4396
4397out:
4398 kfree(qsa);
4399drop_write:
4400 mnt_drop_write_file(file);
4401 return ret;
4402}
4403
4404static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4405{
6d0379ec 4406 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4407 struct btrfs_ioctl_quota_rescan_args *qsa;
4408 int ret = 0;
4409
4410 if (!capable(CAP_SYS_ADMIN))
4411 return -EPERM;
4412
4413 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4414 if (!qsa)
4415 return -ENOMEM;
4416
4417 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4418 qsa->flags = 1;
4419 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4420 }
4421
4422 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4423 ret = -EFAULT;
4424
4425 kfree(qsa);
4426 return ret;
4427}
4428
57254b6e
JS		 4429static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4430{
54563d41 4431 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
57254b6e
JS		 4432
4433 if (!capable(CAP_SYS_ADMIN))
4434 return -EPERM;
4435
4436 return btrfs_qgroup_wait_for_completion(root->fs_info);
4437}
4438
abccd00f
HM		 4439static long _btrfs_ioctl_set_received_subvol(struct file *file,
4440 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4441{
496ad9aa 4442 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4443 struct btrfs_root *root = BTRFS_I(inode)->root;
4444 struct btrfs_root_item *root_item = &root->root_item;
4445 struct btrfs_trans_handle *trans;
4446 struct timespec ct = CURRENT_TIME;
4447 int ret = 0;
dd5f9615 4448 int received_uuid_changed;
8ea05e3a 4449
bd60ea0f
DS		 4450 if (!inode_owner_or_capable(inode))
4451 return -EPERM;
4452
8ea05e3a
AB		 4453 ret = mnt_want_write_file(file);
4454 if (ret < 0)
4455 return ret;
4456
4457 down_write(&root->fs_info->subvol_sem);
4458
4459 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4460 ret = -EINVAL;
4461 goto out;
4462 }
4463
4464 if (btrfs_root_readonly(root)) {
4465 ret = -EROFS;
4466 goto out;
4467 }
4468
dd5f9615
SB		 4469 /*
4470 * 1 - root item
4471 * 2 - uuid items (received uuid + subvol uuid)
4472 */
4473 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4474 if (IS_ERR(trans)) {
4475 ret = PTR_ERR(trans);
4476 trans = NULL;
4477 goto out;
4478 }
4479
4480 sa->rtransid = trans->transid;
4481 sa->rtime.sec = ct.tv_sec;
4482 sa->rtime.nsec = ct.tv_nsec;
4483
dd5f9615
SB		 4484 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4485 BTRFS_UUID_SIZE);
4486 if (received_uuid_changed &&
4487 !btrfs_is_empty_uuid(root_item->received_uuid))
4488 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4489 root_item->received_uuid,
4490 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4491 root->root_key.objectid);
8ea05e3a
AB		 4492 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4493 btrfs_set_root_stransid(root_item, sa->stransid);
4494 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4495 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4496 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4497 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4498 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4499
4500 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4501 &root->root_key, &root->root_item);
4502 if (ret < 0) {
4503 btrfs_end_transaction(trans, root);
8ea05e3a 4504 goto out;
dd5f9615
SB		 4505 }
4506 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4507 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4508 sa->uuid,
4509 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4510 root->root_key.objectid);
4511 if (ret < 0 && ret != -EEXIST) {
4512 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 4513 goto out;
dd5f9615
SB		 4514 }
4515 }
4516 ret = btrfs_commit_transaction(trans, root);
4517 if (ret < 0) {
4518 btrfs_abort_transaction(trans, root, ret);
4519 goto out;
8ea05e3a
AB		 4520 }
4521
abccd00f
HM		 4522out:
4523 up_write(&root->fs_info->subvol_sem);
4524 mnt_drop_write_file(file);
4525 return ret;
4526}
4527
4528#ifdef CONFIG_64BIT
4529static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4530 void __user *arg)
4531{
4532 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4533 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4534 int ret = 0;
4535
4536 args32 = memdup_user(arg, sizeof(*args32));
4537 if (IS_ERR(args32)) {
4538 ret = PTR_ERR(args32);
4539 args32 = NULL;
4540 goto out;
4541 }
4542
4543 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4544 if (IS_ERR(args64)) {
4545 ret = PTR_ERR(args64);
4546 args64 = NULL;
4547 goto out;
4548 }
4549
4550 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4551 args64->stransid = args32->stransid;
4552 args64->rtransid = args32->rtransid;
4553 args64->stime.sec = args32->stime.sec;
4554 args64->stime.nsec = args32->stime.nsec;
4555 args64->rtime.sec = args32->rtime.sec;
4556 args64->rtime.nsec = args32->rtime.nsec;
4557 args64->flags = args32->flags;
4558
4559 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4560 if (ret)
4561 goto out;
4562
4563 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4564 args32->stransid = args64->stransid;
4565 args32->rtransid = args64->rtransid;
4566 args32->stime.sec = args64->stime.sec;
4567 args32->stime.nsec = args64->stime.nsec;
4568 args32->rtime.sec = args64->rtime.sec;
4569 args32->rtime.nsec = args64->rtime.nsec;
4570 args32->flags = args64->flags;
4571
4572 ret = copy_to_user(arg, args32, sizeof(*args32));
4573 if (ret)
4574 ret = -EFAULT;
4575
4576out:
4577 kfree(args32);
4578 kfree(args64);
4579 return ret;
4580}
4581#endif
4582
4583static long btrfs_ioctl_set_received_subvol(struct file *file,
4584 void __user *arg)
4585{
4586 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4587 int ret = 0;
4588
4589 sa = memdup_user(arg, sizeof(*sa));
4590 if (IS_ERR(sa)) {
4591 ret = PTR_ERR(sa);
4592 sa = NULL;
4593 goto out;
4594 }
4595
4596 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4597
4598 if (ret)
4599 goto out;
4600
8ea05e3a
AB		 4601 ret = copy_to_user(arg, sa, sizeof(*sa));
4602 if (ret)
4603 ret = -EFAULT;
4604
4605out:
4606 kfree(sa);
8ea05e3a
AB		 4607 return ret;
4608}
4609
867ab667 4610static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4611{
6d0379ec 4612 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 4613 size_t len;
867ab667 4614 int ret;
a1b83ac5
AJ		 4615 char label[BTRFS_LABEL_SIZE];
4616
4617 spin_lock(&root->fs_info->super_lock);
4618 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4619 spin_unlock(&root->fs_info->super_lock);
4620
4621 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 4622
4623 if (len == BTRFS_LABEL_SIZE) {
efe120a0
FH		 4624 btrfs_warn(root->fs_info,
4625 "label is too long, return the first %zu bytes", --len);
867ab667 4626 }
4627
867ab667 4628 ret = copy_to_user(arg, label, len);
867ab667 4629
4630 return ret ? -EFAULT : 0;
4631}
4632
a8bfd4ab 4633static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4634{
6d0379ec 4635 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 4636 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4637 struct btrfs_trans_handle *trans;
4638 char label[BTRFS_LABEL_SIZE];
4639 int ret;
4640
4641 if (!capable(CAP_SYS_ADMIN))
4642 return -EPERM;
4643
4644 if (copy_from_user(label, arg, sizeof(label)))
4645 return -EFAULT;
4646
4647 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
efe120a0 4648 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
a8bfd4ab 4649 BTRFS_LABEL_SIZE - 1);
4650 return -EINVAL;
4651 }
4652
4653 ret = mnt_want_write_file(file);
4654 if (ret)
4655 return ret;
4656
a8bfd4ab 4657 trans = btrfs_start_transaction(root, 0);
4658 if (IS_ERR(trans)) {
4659 ret = PTR_ERR(trans);
4660 goto out_unlock;
4661 }
4662
a1b83ac5 4663 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 4664 strcpy(super_block->label, label);
a1b83ac5 4665 spin_unlock(&root->fs_info->super_lock);
d0270aca 4666 ret = btrfs_commit_transaction(trans, root);
a8bfd4ab 4667
4668out_unlock:
a8bfd4ab 4669 mnt_drop_write_file(file);
4670 return ret;
4671}
4672
2eaa055f
JM		 4673#define INIT_FEATURE_FLAGS(suffix) \
4674 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4675 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4676 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4677
4678static int btrfs_ioctl_get_supported_features(struct file *file,
4679 void __user *arg)
4680{
4681 static struct btrfs_ioctl_feature_flags features[3] = {
4682 INIT_FEATURE_FLAGS(SUPP),
4683 INIT_FEATURE_FLAGS(SAFE_SET),
4684 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4685 };
4686
4687 if (copy_to_user(arg, &features, sizeof(features)))
4688 return -EFAULT;
4689
4690 return 0;
4691}
4692
4693static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
4694{
4695 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4696 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4697 struct btrfs_ioctl_feature_flags features;
4698
4699 features.compat_flags = btrfs_super_compat_flags(super_block);
4700 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4701 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4702
4703 if (copy_to_user(arg, &features, sizeof(features)))
4704 return -EFAULT;
4705
4706 return 0;
4707}
4708
3b02a68a
JM		 4709static int check_feature_bits(struct btrfs_root *root,
4710 enum btrfs_feature_set set,
2eaa055f
JM		 4711 u64 change_mask, u64 flags, u64 supported_flags,
4712 u64 safe_set, u64 safe_clear)
4713{
3b02a68a
JM		 4714 const char *type = btrfs_feature_set_names[set];
4715 char *names;
2eaa055f
JM		 4716 u64 disallowed, unsupported;
4717 u64 set_mask = flags & change_mask;
4718 u64 clear_mask = ~flags & change_mask;
4719
4720 unsupported = set_mask & ~supported_flags;
4721 if (unsupported) {
3b02a68a
JM		 4722 names = btrfs_printable_features(set, unsupported);
4723 if (names) {
4724 btrfs_warn(root->fs_info,
4725 "this kernel does not support the %s feature bit%s",
4726 names, strchr(names, ',') ? "s" : "");
4727 kfree(names);
4728 } else
4729 btrfs_warn(root->fs_info,
2eaa055f
JM		 4730 "this kernel does not support %s bits 0x%llx",
4731 type, unsupported);
4732 return -EOPNOTSUPP;
4733 }
4734
4735 disallowed = set_mask & ~safe_set;
4736 if (disallowed) {
3b02a68a
JM		 4737 names = btrfs_printable_features(set, disallowed);
4738 if (names) {
4739 btrfs_warn(root->fs_info,
4740 "can't set the %s feature bit%s while mounted",
4741 names, strchr(names, ',') ? "s" : "");
4742 kfree(names);
4743 } else
4744 btrfs_warn(root->fs_info,
2eaa055f
JM		 4745 "can't set %s bits 0x%llx while mounted",
4746 type, disallowed);
4747 return -EPERM;
4748 }
4749
4750 disallowed = clear_mask & ~safe_clear;
4751 if (disallowed) {
3b02a68a
JM		 4752 names = btrfs_printable_features(set, disallowed);
4753 if (names) {
4754 btrfs_warn(root->fs_info,
4755 "can't clear the %s feature bit%s while mounted",
4756 names, strchr(names, ',') ? "s" : "");
4757 kfree(names);
4758 } else
4759 btrfs_warn(root->fs_info,
2eaa055f
JM		 4760 "can't clear %s bits 0x%llx while mounted",
4761 type, disallowed);
4762 return -EPERM;
4763 }
4764
4765 return 0;
4766}
4767
4768#define check_feature(root, change_mask, flags, mask_base) \
3b02a68a 4769check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 4770 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4771 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4772 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4773
4774static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4775{
4776 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4777 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4778 struct btrfs_ioctl_feature_flags flags[2];
4779 struct btrfs_trans_handle *trans;
4780 u64 newflags;
4781 int ret;
4782
4783 if (!capable(CAP_SYS_ADMIN))
4784 return -EPERM;
4785
4786 if (copy_from_user(flags, arg, sizeof(flags)))
4787 return -EFAULT;
4788
4789 /* Nothing to do */
4790 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4791 !flags[0].incompat_flags)
4792 return 0;
4793
4794 ret = check_feature(root, flags[0].compat_flags,
4795 flags[1].compat_flags, COMPAT);
4796 if (ret)
4797 return ret;
4798
4799 ret = check_feature(root, flags[0].compat_ro_flags,
4800 flags[1].compat_ro_flags, COMPAT_RO);
4801 if (ret)
4802 return ret;
4803
4804 ret = check_feature(root, flags[0].incompat_flags,
4805 flags[1].incompat_flags, INCOMPAT);
4806 if (ret)
4807 return ret;
4808
8051aa1a 4809 trans = btrfs_start_transaction(root, 0);
2eaa055f
JM		 4810 if (IS_ERR(trans))
4811 return PTR_ERR(trans);
4812
4813 spin_lock(&root->fs_info->super_lock);
4814 newflags = btrfs_super_compat_flags(super_block);
4815 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4816 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4817 btrfs_set_super_compat_flags(super_block, newflags);
4818
4819 newflags = btrfs_super_compat_ro_flags(super_block);
4820 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4821 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4822 btrfs_set_super_compat_ro_flags(super_block, newflags);
4823
4824 newflags = btrfs_super_incompat_flags(super_block);
4825 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4826 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4827 btrfs_set_super_incompat_flags(super_block, newflags);
4828 spin_unlock(&root->fs_info->super_lock);
4829
d0270aca 4830 return btrfs_commit_transaction(trans, root);
2eaa055f
JM		 4831}
4832
f46b5a66
CH		 4833long btrfs_ioctl(struct file *file, unsigned int
4834 cmd, unsigned long arg)
4835{
496ad9aa 4836 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 4837 void __user *argp = (void __user *)arg;
f46b5a66
CH		 4838
4839 switch (cmd) {
6cbff00f
CH		 4840 case FS_IOC_GETFLAGS:
4841 return btrfs_ioctl_getflags(file, argp);
4842 case FS_IOC_SETFLAGS:
4843 return btrfs_ioctl_setflags(file, argp);
4844 case FS_IOC_GETVERSION:
4845 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 4846 case FITRIM:
4847 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 4848 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 4849 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 4850 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 4851 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 4852 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 4853 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 4854 case BTRFS_IOC_SUBVOL_CREATE_V2:
4855 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 4856 case BTRFS_IOC_SNAP_DESTROY:
4857 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 4858 case BTRFS_IOC_SUBVOL_GETFLAGS:
4859 return btrfs_ioctl_subvol_getflags(file, argp);
4860 case BTRFS_IOC_SUBVOL_SETFLAGS:
4861 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 4862 case BTRFS_IOC_DEFAULT_SUBVOL:
4863 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 4864 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 4865 return btrfs_ioctl_defrag(file, NULL);
4866 case BTRFS_IOC_DEFRAG_RANGE:
4867 return btrfs_ioctl_defrag(file, argp);
f46b5a66 4868 case BTRFS_IOC_RESIZE:
198605a8 4869 return btrfs_ioctl_resize(file, argp);
f46b5a66 4870 case BTRFS_IOC_ADD_DEV:
4bcabaa3 4871 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 4872 case BTRFS_IOC_RM_DEV:
da24927b 4873 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 4874 case BTRFS_IOC_FS_INFO:
4875 return btrfs_ioctl_fs_info(root, argp);
4876 case BTRFS_IOC_DEV_INFO:
4877 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 4878 case BTRFS_IOC_BALANCE:
9ba1f6e4 4879 return btrfs_ioctl_balance(file, NULL);
f46b5a66 4880 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 4881 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4882 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 4883 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 4884 case BTRFS_IOC_TRANS_START:
4885 return btrfs_ioctl_trans_start(file);
4886 case BTRFS_IOC_TRANS_END:
4887 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 4888 case BTRFS_IOC_TREE_SEARCH:
4889 return btrfs_ioctl_tree_search(file, argp);
4890 case BTRFS_IOC_INO_LOOKUP:
4891 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 4892 case BTRFS_IOC_INO_PATHS:
4893 return btrfs_ioctl_ino_to_path(root, argp);
4894 case BTRFS_IOC_LOGICAL_INO:
4895 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 4896 case BTRFS_IOC_SPACE_INFO:
4897 return btrfs_ioctl_space_info(root, argp);
9b199859
FDBM		 4898 case BTRFS_IOC_SYNC: {
4899 int ret;
4900
6c255e67 4901 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
9b199859
FDBM		 4902 if (ret)
4903 return ret;
4904 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
4905 return ret;
4906 }
46204592 4907 case BTRFS_IOC_START_SYNC:
9a8c28be 4908 return btrfs_ioctl_start_sync(root, argp);
46204592 4909 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 4910 return btrfs_ioctl_wait_sync(root, argp);
475f6387 4911 case BTRFS_IOC_SCRUB:
b8e95489 4912 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 4913 case BTRFS_IOC_SCRUB_CANCEL:
4914 return btrfs_ioctl_scrub_cancel(root, argp);
4915 case BTRFS_IOC_SCRUB_PROGRESS:
4916 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 4917 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 4918 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 4919 case BTRFS_IOC_BALANCE_CTL:
4920 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 4921 case BTRFS_IOC_BALANCE_PROGRESS:
4922 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 4923 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4924 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 4925#ifdef CONFIG_64BIT
4926 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4927 return btrfs_ioctl_set_received_subvol_32(file, argp);
4928#endif
31db9f7c
AB		 4929 case BTRFS_IOC_SEND:
4930 return btrfs_ioctl_send(file, argp);
c11d2c23 4931 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 4932 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 4933 case BTRFS_IOC_QUOTA_CTL:
905b0dda 4934 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 4935 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 4936 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 4937 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 4938 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 4939 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 4940 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 4941 case BTRFS_IOC_QUOTA_RESCAN:
4942 return btrfs_ioctl_quota_rescan(file, argp);
4943 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4944 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 4945 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4946 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 4947 case BTRFS_IOC_DEV_REPLACE:
4948 return btrfs_ioctl_dev_replace(root, argp);
867ab667 4949 case BTRFS_IOC_GET_FSLABEL:
4950 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 4951 case BTRFS_IOC_SET_FSLABEL:
4952 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 4953 case BTRFS_IOC_FILE_EXTENT_SAME:
4954 return btrfs_ioctl_file_extent_same(file, argp);
2eaa055f
JM		 4955 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4956 return btrfs_ioctl_get_supported_features(file, argp);
4957 case BTRFS_IOC_GET_FEATURES:
4958 return btrfs_ioctl_get_features(file, argp);
4959 case BTRFS_IOC_SET_FEATURES:
4960 return btrfs_ioctl_set_features(file, argp);
f46b5a66
CH		 4961 }
4962
4963 return -ENOTTY;
4964}
Linux 6.x block layer and io_uring tree(s)