projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: fix deadlock with extent-same and readpage
[linux-2.6-block.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
8ea05e3a 44#include <linux/uuid.h>
55e301fd 45#include <linux/btrfs.h>
416161db 46#include <linux/uaccess.h>
f46b5a66
CH		 47#include "ctree.h"
48#include "disk-io.h"
49#include "transaction.h"
50#include "btrfs_inode.h"
f46b5a66
CH		 51#include "print-tree.h"
52#include "volumes.h"
925baedd 53#include "locking.h"
581bb050 54#include "inode-map.h"
d7728c96 55#include "backref.h"
606686ee 56#include "rcu-string.h"
31db9f7c 57#include "send.h"
3f6bcfbd 58#include "dev-replace.h"
63541927 59#include "props.h"
3b02a68a 60#include "sysfs.h"
fcebe456 61#include "qgroup.h"
f46b5a66 62
abccd00f
HM		 63#ifdef CONFIG_64BIT
64/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
68 */
69struct btrfs_ioctl_timespec_32 {
70 __u64 sec;
71 __u32 nsec;
72} __attribute__ ((__packed__));
73
74struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
80 __u64 flags; /* in */
81 __u64 reserved[16]; /* in */
82} __attribute__ ((__packed__));
83
84#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
86#endif
87
88
416161db
MF		 89static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
91
6cbff00f
CH		 92/* Mask out flags that are inappropriate for the given type of inode. */
93static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
94{
95 if (S_ISDIR(mode))
96 return flags;
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
99 else
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
101}
102
103/*
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105 */
106static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
107{
108 unsigned int iflags = 0;
109
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
124
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 129
130 return iflags;
131}
132
133/*
134 * Update inode->i_flags based on the btrfs internal flags.
135 */
136void btrfs_update_iflags(struct inode *inode)
137{
138 struct btrfs_inode *ip = BTRFS_I(inode);
3cc79392 139 unsigned int new_fl = 0;
6cbff00f
CH		 140
141 if (ip->flags & BTRFS_INODE_SYNC)
3cc79392 142 new_fl |= S_SYNC;
6cbff00f 143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
3cc79392 144 new_fl |= S_IMMUTABLE;
6cbff00f 145 if (ip->flags & BTRFS_INODE_APPEND)
3cc79392 146 new_fl |= S_APPEND;
6cbff00f 147 if (ip->flags & BTRFS_INODE_NOATIME)
3cc79392 148 new_fl |= S_NOATIME;
6cbff00f 149 if (ip->flags & BTRFS_INODE_DIRSYNC)
3cc79392
FM		 150 new_fl |= S_DIRSYNC;
151
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
154 new_fl);
6cbff00f
CH		 155}
156
157/*
158 * Inherit flags from the parent inode.
159 *
e27425d6 160 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 161 */
162void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
163{
0b4dcea5
CM		 164 unsigned int flags;
165
166 if (!dir)
167 return;
168
169 flags = BTRFS_I(dir)->flags;
6cbff00f 170
e27425d6
JB		 171 if (flags & BTRFS_INODE_NOCOMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 } else if (flags & BTRFS_INODE_COMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
177 }
178
213490b3 179 if (flags & BTRFS_INODE_NODATACOW) {
e27425d6 180 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
213490b3
LB		 181 if (S_ISREG(inode->i_mode))
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
183 }
6cbff00f 184
6cbff00f
CH		 185 btrfs_update_iflags(inode);
186}
187
188static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
189{
496ad9aa 190 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
6cbff00f
CH		 191 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
192
193 if (copy_to_user(arg, &flags, sizeof(flags)))
194 return -EFAULT;
195 return 0;
196}
197
75e7cb7f
LB		 198static int check_flags(unsigned int flags)
199{
200 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 FS_NOATIME_FL | FS_NODUMP_FL | \
202 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 203 FS_NOCOMP_FL | FS_COMPR_FL |
204 FS_NOCOW_FL))
75e7cb7f
LB		 205 return -EOPNOTSUPP;
206
207 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
208 return -EINVAL;
209
75e7cb7f
LB		 210 return 0;
211}
212
6cbff00f
CH		 213static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
214{
496ad9aa 215 struct inode *inode = file_inode(file);
6cbff00f
CH		 216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
220 int ret;
f062abf0
LZ		 221 u64 ip_oldflags;
222 unsigned int i_oldflags;
7e97b8da 223 umode_t mode;
6cbff00f 224
bd60ea0f
DS		 225 if (!inode_owner_or_capable(inode))
226 return -EPERM;
227
b83cc969
LZ		 228 if (btrfs_root_readonly(root))
229 return -EROFS;
230
6cbff00f
CH		 231 if (copy_from_user(&flags, arg, sizeof(flags)))
232 return -EFAULT;
233
75e7cb7f
LB		 234 ret = check_flags(flags);
235 if (ret)
236 return ret;
f46b5a66 237
e7848683
JK		 238 ret = mnt_want_write_file(file);
239 if (ret)
240 return ret;
241
6cbff00f
CH		 242 mutex_lock(&inode->i_mutex);
243
f062abf0
LZ		 244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
7e97b8da 246 mode = inode->i_mode;
f062abf0 247
6cbff00f
CH		 248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
252 ret = -EPERM;
253 goto out_unlock;
254 }
255 }
256
6cbff00f
CH		 257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
259 else
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
263 else
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
267 else
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
271 else
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
275 else
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
279 else
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
7e97b8da
DS		 281 if (flags & FS_NOCOW_FL) {
282 if (S_ISREG(mode)) {
283 /*
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
287 */
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
291 } else {
292 ip->flags |= BTRFS_INODE_NODATACOW;
293 }
294 } else {
295 /*
296 * Revert back under same assuptions as above
297 */
298 if (S_ISREG(mode)) {
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
302 } else {
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
304 }
305 }
6cbff00f 306
75e7cb7f
LB		 307 /*
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
310 * things smaller.
311 */
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 315
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
318 goto out_drop;
75e7cb7f 319 } else if (flags & FS_COMPR_FL) {
63541927
FDBM		 320 const char *comp;
321
75e7cb7f
LB		 322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 324
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
326 comp = "lzo";
327 else
328 comp = "zlib";
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
331 if (ret)
332 goto out_drop;
333
ebcb904d 334 } else {
78a017a2
FM		 335 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
336 if (ret && ret != -ENODATA)
337 goto out_drop;
ebcb904d 338 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 339 }
6cbff00f 340
4da6f1a3 341 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 342 if (IS_ERR(trans)) {
343 ret = PTR_ERR(trans);
344 goto out_drop;
345 }
6cbff00f 346
306424cc 347 btrfs_update_iflags(inode);
0c4d2d95 348 inode_inc_iversion(inode);
306424cc 349 inode->i_ctime = CURRENT_TIME;
6cbff00f 350 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 351
6cbff00f 352 btrfs_end_transaction(trans, root);
f062abf0
LZ		 353 out_drop:
354 if (ret) {
355 ip->flags = ip_oldflags;
356 inode->i_flags = i_oldflags;
357 }
6cbff00f 358
6cbff00f
CH		 359 out_unlock:
360 mutex_unlock(&inode->i_mutex);
e7848683 361 mnt_drop_write_file(file);
2d4e6f6a 362 return ret;
6cbff00f
CH		 363}
364
365static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
366{
496ad9aa 367 struct inode *inode = file_inode(file);
6cbff00f
CH		 368
369 return put_user(inode->i_generation, arg);
370}
f46b5a66 371
f7039b1d
LD		 372static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
373{
54563d41 374 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
f7039b1d
LD		 375 struct btrfs_device *device;
376 struct request_queue *q;
377 struct fstrim_range range;
378 u64 minlen = ULLONG_MAX;
379 u64 num_devices = 0;
815745cf 380 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 381 int ret;
382
383 if (!capable(CAP_SYS_ADMIN))
384 return -EPERM;
385
1f78160c
XG		 386 rcu_read_lock();
387 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
388 dev_list) {
f7039b1d
LD		 389 if (!device->bdev)
390 continue;
391 q = bdev_get_queue(device->bdev);
392 if (blk_queue_discard(q)) {
393 num_devices++;
394 minlen = min((u64)q->limits.discard_granularity,
395 minlen);
396 }
397 }
1f78160c 398 rcu_read_unlock();
f4c697e6 399
f7039b1d
LD		 400 if (!num_devices)
401 return -EOPNOTSUPP;
f7039b1d
LD		 402 if (copy_from_user(&range, arg, sizeof(range)))
403 return -EFAULT;
e515c18b
LC		 404 if (range.start > total_bytes ||
405 range.len < fs_info->sb->s_blocksize)
f4c697e6 406 return -EINVAL;
f7039b1d 407
f4c697e6 408 range.len = min(range.len, total_bytes - range.start);
f7039b1d 409 range.minlen = max(range.minlen, minlen);
815745cf 410 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 411 if (ret < 0)
412 return ret;
413
414 if (copy_to_user(arg, &range, sizeof(range)))
415 return -EFAULT;
416
417 return 0;
418}
419
dd5f9615
SB		 420int btrfs_is_empty_uuid(u8 *uuid)
421{
46e0f66a
CM		 422 int i;
423
424 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
425 if (uuid[i])
426 return 0;
427 }
428 return 1;
dd5f9615
SB		 429}
430
d5c12070 431static noinline int create_subvol(struct inode *dir,
cb8e7090 432 struct dentry *dentry,
72fd032e 433 char *name, int namelen,
6f72c7e2 434 u64 *async_transid,
8696c533 435 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 436{
437 struct btrfs_trans_handle *trans;
438 struct btrfs_key key;
439 struct btrfs_root_item root_item;
440 struct btrfs_inode_item *inode_item;
441 struct extent_buffer *leaf;
d5c12070 442 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 443 struct btrfs_root *new_root;
d5c12070 444 struct btrfs_block_rsv block_rsv;
8ea05e3a 445 struct timespec cur_time = CURRENT_TIME;
5662344b 446 struct inode *inode;
f46b5a66
CH		 447 int ret;
448 int err;
449 u64 objectid;
450 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 451 u64 index = 0;
d5c12070 452 u64 qgroup_reserved;
8ea05e3a 453 uuid_le new_uuid;
f46b5a66 454
581bb050 455 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 456 if (ret)
a22285a6 457 return ret;
6a912213 458
e09fe2d2
QW		 459 /*
460 * Don't create subvolume whose level is not zero. Or qgroup will be
461 * screwed up since it assume subvolme qgroup's level to be 0.
462 */
463 if (btrfs_qgroup_level(objectid))
464 return -ENOSPC;
465
d5c12070 466 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 467 /*
d5c12070
MX		 468 * The same as the snapshot creation, please see the comment
469 * of create_snapshot().
9ed74f2d 470 */
d5c12070 471 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 472 8, &qgroup_reserved, false);
d5c12070
MX		 473 if (ret)
474 return ret;
475
476 trans = btrfs_start_transaction(root, 0);
477 if (IS_ERR(trans)) {
478 ret = PTR_ERR(trans);
de6e8200
LB		 479 btrfs_subvolume_release_metadata(root, &block_rsv,
480 qgroup_reserved);
481 return ret;
d5c12070
MX		 482 }
483 trans->block_rsv = &block_rsv;
484 trans->bytes_reserved = block_rsv.size;
f46b5a66 485
8696c533 486 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 487 if (ret)
488 goto fail;
489
4d75f8a9 490 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 491 if (IS_ERR(leaf)) {
492 ret = PTR_ERR(leaf);
493 goto fail;
494 }
f46b5a66 495
5d4f98a2 496 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 497 btrfs_set_header_bytenr(leaf, leaf->start);
498 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 499 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 500 btrfs_set_header_owner(leaf, objectid);
501
0a4e5586 502 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
f46b5a66 503 BTRFS_FSID_SIZE);
5d4f98a2 504 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
b308bc2f 505 btrfs_header_chunk_tree_uuid(leaf),
5d4f98a2 506 BTRFS_UUID_SIZE);
f46b5a66
CH		 507 btrfs_mark_buffer_dirty(leaf);
508
8ea05e3a
AB		 509 memset(&root_item, 0, sizeof(root_item));
510
f46b5a66 511 inode_item = &root_item.inode;
3cae210f
QW		 512 btrfs_set_stack_inode_generation(inode_item, 1);
513 btrfs_set_stack_inode_size(inode_item, 3);
514 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 515 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 516 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 517
3cae210f
QW		 518 btrfs_set_root_flags(&root_item, 0);
519 btrfs_set_root_limit(&root_item, 0);
520 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 521
f46b5a66 522 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 523 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 524 btrfs_set_root_level(&root_item, 0);
525 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 526 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 527 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 528
8ea05e3a
AB		 529 btrfs_set_root_generation_v2(&root_item,
530 btrfs_root_generation(&root_item));
531 uuid_le_gen(&new_uuid);
532 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 533 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
534 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 535 root_item.ctime = root_item.otime;
536 btrfs_set_root_ctransid(&root_item, trans->transid);
537 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 538
925baedd 539 btrfs_tree_unlock(leaf);
f46b5a66
CH		 540 free_extent_buffer(leaf);
541 leaf = NULL;
542
543 btrfs_set_root_dirid(&root_item, new_dirid);
544
545 key.objectid = objectid;
5d4f98a2 546 key.offset = 0;
962a298f 547 key.type = BTRFS_ROOT_ITEM_KEY;
f46b5a66
CH		 548 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
549 &root_item);
550 if (ret)
551 goto fail;
552
76dda93c
YZ		 553 key.offset = (u64)-1;
554 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa 555 if (IS_ERR(new_root)) {
79787eaa 556 ret = PTR_ERR(new_root);
6d13f549 557 btrfs_abort_transaction(trans, root, ret);
79787eaa
JM		 558 goto fail;
559 }
76dda93c
YZ		 560
561 btrfs_record_root_in_trans(trans, new_root);
562
63541927 563 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 564 if (ret) {
565 /* We potentially lose an unused inode item here */
79787eaa 566 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 567 goto fail;
568 }
569
f46b5a66
CH		 570 /*
571 * insert the directory item
572 */
3de4586c 573 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 574 if (ret) {
575 btrfs_abort_transaction(trans, root, ret);
576 goto fail;
577 }
3de4586c
CM		 578
579 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 580 name, namelen, dir, &key,
3de4586c 581 BTRFS_FT_DIR, index);
79787eaa
JM		 582 if (ret) {
583 btrfs_abort_transaction(trans, root, ret);
f46b5a66 584 goto fail;
79787eaa 585 }
0660b5af 586
52c26179
YZ		 587 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
588 ret = btrfs_update_inode(trans, root, dir);
589 BUG_ON(ret);
590
0660b5af 591 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 592 objectid, root->root_key.objectid,
33345d01 593 btrfs_ino(dir), index, name, namelen);
76dda93c 594 BUG_ON(ret);
f46b5a66 595
dd5f9615
SB		 596 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
597 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
598 objectid);
599 if (ret)
600 btrfs_abort_transaction(trans, root, ret);
601
f46b5a66 602fail:
d5c12070
MX		 603 trans->block_rsv = NULL;
604 trans->bytes_reserved = 0;
de6e8200
LB		 605 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
606
72fd032e
SW		 607 if (async_transid) {
608 *async_transid = trans->transid;
609 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 610 if (err)
611 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 612 } else {
613 err = btrfs_commit_transaction(trans, root);
614 }
f46b5a66
CH		 615 if (err && !ret)
616 ret = err;
1a65e24b 617
5662344b
TI		 618 if (!ret) {
619 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 620 if (IS_ERR(inode))
621 return PTR_ERR(inode);
5662344b
TI		 622 d_instantiate(dentry, inode);
623 }
f46b5a66
CH		 624 return ret;
625}
626
9ea24bbe 627static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
8257b2dc
MX		 628{
629 s64 writers;
630 DEFINE_WAIT(wait);
631
632 do {
633 prepare_to_wait(&root->subv_writers->wait, &wait,
634 TASK_UNINTERRUPTIBLE);
635
636 writers = percpu_counter_sum(&root->subv_writers->counter);
637 if (writers)
638 schedule();
639
640 finish_wait(&root->subv_writers->wait, &wait);
641 } while (writers);
642}
643
e9662f70
MX		 644static int create_snapshot(struct btrfs_root *root, struct inode *dir,
645 struct dentry *dentry, char *name, int namelen,
646 u64 *async_transid, bool readonly,
647 struct btrfs_qgroup_inherit *inherit)
f46b5a66 648{
2e4bfab9 649 struct inode *inode;
f46b5a66
CH		 650 struct btrfs_pending_snapshot *pending_snapshot;
651 struct btrfs_trans_handle *trans;
2e4bfab9 652 int ret;
f46b5a66 653
27cdeb70 654 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
f46b5a66
CH		 655 return -EINVAL;
656
8257b2dc 657 atomic_inc(&root->will_be_snapshoted);
4e857c58 658 smp_mb__after_atomic();
9ea24bbe 659 btrfs_wait_for_no_snapshoting_writes(root);
8257b2dc 660
6a03843d
MX		 661 ret = btrfs_start_delalloc_inodes(root, 0);
662 if (ret)
8257b2dc 663 goto out;
6a03843d 664
b0244199 665 btrfs_wait_ordered_extents(root, -1);
6a03843d 666
3de4586c 667 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
8257b2dc
MX		 668 if (!pending_snapshot) {
669 ret = -ENOMEM;
670 goto out;
671 }
a22285a6 672
66d8f3dd
MX		 673 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
674 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 675 /*
676 * 1 - parent dir inode
677 * 2 - dir entries
678 * 1 - root item
679 * 2 - root ref/backref
680 * 1 - root of snapshot
dd5f9615 681 * 1 - UUID item
d5c12070
MX		 682 */
683 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 684 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 685 &pending_snapshot->qgroup_reserved,
686 false);
d5c12070 687 if (ret)
8257b2dc 688 goto free;
d5c12070 689
3de4586c 690 pending_snapshot->dentry = dentry;
f46b5a66 691 pending_snapshot->root = root;
b83cc969 692 pending_snapshot->readonly = readonly;
e9662f70 693 pending_snapshot->dir = dir;
8696c533 694 pending_snapshot->inherit = inherit;
a22285a6 695
d5c12070 696 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 697 if (IS_ERR(trans)) {
698 ret = PTR_ERR(trans);
699 goto fail;
700 }
701
8351583e 702 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 703 list_add(&pending_snapshot->list,
704 &trans->transaction->pending_snapshots);
8351583e 705 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 706 if (async_transid) {
707 *async_transid = trans->transid;
708 ret = btrfs_commit_transaction_async(trans,
709 root->fs_info->extent_root, 1);
00d71c9c
MX		 710 if (ret)
711 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 712 } else {
713 ret = btrfs_commit_transaction(trans,
714 root->fs_info->extent_root);
715 }
aec8030a 716 if (ret)
c37b2b62 717 goto fail;
a22285a6
YZ		 718
719 ret = pending_snapshot->error;
720 if (ret)
721 goto fail;
722
d3797308
CM		 723 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
724 if (ret)
725 goto fail;
726
2b0143b5 727 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
2e4bfab9
YZ		 728 if (IS_ERR(inode)) {
729 ret = PTR_ERR(inode);
730 goto fail;
731 }
5662344b 732
2e4bfab9
YZ		 733 d_instantiate(dentry, inode);
734 ret = 0;
735fail:
d5c12070
MX		 736 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
737 &pending_snapshot->block_rsv,
738 pending_snapshot->qgroup_reserved);
8257b2dc 739free:
a22285a6 740 kfree(pending_snapshot);
8257b2dc 741out:
9ea24bbe
FM		 742 if (atomic_dec_and_test(&root->will_be_snapshoted))
743 wake_up_atomic_t(&root->will_be_snapshoted);
f46b5a66
CH		 744 return ret;
745}
746
4260f7c7
SW		 747/* copy of may_delete in fs/namei.c()
748 * Check whether we can remove a link victim from directory dir, check
749 * whether the type of victim is right.
750 * 1. We can't do it if dir is read-only (done in permission())
751 * 2. We should have write and exec permissions on dir
752 * 3. We can't remove anything from append-only dir
753 * 4. We can't do anything with immutable dir (done in permission())
754 * 5. If the sticky bit on dir is set we should either
755 * a. be owner of dir, or
756 * b. be owner of victim, or
757 * c. have CAP_FOWNER capability
758 * 6. If the victim is append-only or immutable we can't do antyhing with
759 * links pointing to it.
760 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
761 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
762 * 9. We can't remove a root or mountpoint.
763 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
764 * nfs_async_unlink().
765 */
766
67871254 767static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 768{
769 int error;
770
2b0143b5 771 if (d_really_is_negative(victim))
4260f7c7
SW		 772 return -ENOENT;
773
2b0143b5 774 BUG_ON(d_inode(victim->d_parent) != dir);
4fa6b5ec 775 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 776
777 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
778 if (error)
779 return error;
780 if (IS_APPEND(dir))
781 return -EPERM;
2b0143b5
DH		 782 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
783 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
4260f7c7
SW		 784 return -EPERM;
785 if (isdir) {
e36cb0b8 786 if (!d_is_dir(victim))
4260f7c7
SW		 787 return -ENOTDIR;
788 if (IS_ROOT(victim))
789 return -EBUSY;
e36cb0b8 790 } else if (d_is_dir(victim))
4260f7c7
SW		 791 return -EISDIR;
792 if (IS_DEADDIR(dir))
793 return -ENOENT;
794 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
795 return -EBUSY;
796 return 0;
797}
798
cb8e7090
CH		 799/* copy of may_create in fs/namei.c() */
800static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
801{
2b0143b5 802 if (d_really_is_positive(child))
cb8e7090
CH		 803 return -EEXIST;
804 if (IS_DEADDIR(dir))
805 return -ENOENT;
806 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
807}
808
809/*
810 * Create a new subvolume below @parent. This is largely modeled after
811 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
812 * inside this filesystem so it's quite a bit simpler.
813 */
76dda93c
YZ		 814static noinline int btrfs_mksubvol(struct path *parent,
815 char *name, int namelen,
72fd032e 816 struct btrfs_root *snap_src,
6f72c7e2 817 u64 *async_transid, bool readonly,
8696c533 818 struct btrfs_qgroup_inherit *inherit)
cb8e7090 819{
2b0143b5 820 struct inode *dir = d_inode(parent->dentry);
cb8e7090
CH		 821 struct dentry *dentry;
822 int error;
823
5c50c9b8
DS		 824 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
825 if (error == -EINTR)
826 return error;
cb8e7090
CH		 827
828 dentry = lookup_one_len(name, parent->dentry, namelen);
829 error = PTR_ERR(dentry);
830 if (IS_ERR(dentry))
831 goto out_unlock;
832
833 error = -EEXIST;
2b0143b5 834 if (d_really_is_positive(dentry))
cb8e7090
CH		 835 goto out_dput;
836
76dda93c 837 error = btrfs_may_create(dir, dentry);
cb8e7090 838 if (error)
a874a63e 839 goto out_dput;
cb8e7090 840
9c52057c
CM		 841 /*
842 * even if this name doesn't exist, we may get hash collisions.
843 * check for them now when we can safely fail
844 */
845 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
846 dir->i_ino, name,
847 namelen);
848 if (error)
849 goto out_dput;
850
76dda93c
YZ		 851 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
852
853 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
854 goto out_up_read;
855
3de4586c 856 if (snap_src) {
e9662f70 857 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 858 async_transid, readonly, inherit);
3de4586c 859 } else {
d5c12070
MX		 860 error = create_subvol(dir, dentry, name, namelen,
861 async_transid, inherit);
3de4586c 862 }
76dda93c
YZ		 863 if (!error)
864 fsnotify_mkdir(dir, dentry);
865out_up_read:
866 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 867out_dput:
868 dput(dentry);
869out_unlock:
76dda93c 870 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 871 return error;
872}
873
4cb5300b
CM		 874/*
875 * When we're defragging a range, we don't want to kick it off again
876 * if it is really just waiting for delalloc to send it down.
877 * If we find a nice big extent or delalloc range for the bytes in the
878 * file you want to defrag, we return 0 to let you know to skip this
879 * part of the file
880 */
aab110ab 881static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
4cb5300b
CM		 882{
883 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
884 struct extent_map *em = NULL;
885 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
886 u64 end;
887
888 read_lock(&em_tree->lock);
889 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
890 read_unlock(&em_tree->lock);
891
892 if (em) {
893 end = extent_map_end(em);
894 free_extent_map(em);
895 if (end - offset > thresh)
896 return 0;
897 }
898 /* if we already have a nice delalloc here, just stop */
899 thresh /= 2;
900 end = count_range_bits(io_tree, &offset, offset + thresh,
901 thresh, EXTENT_DELALLOC, 1);
902 if (end >= thresh)
903 return 0;
904 return 1;
905}
906
907/*
908 * helper function to walk through a file and find extents
909 * newer than a specific transid, and smaller than thresh.
910 *
911 * This is used by the defragging code to find new and small
912 * extents
913 */
914static int find_new_extents(struct btrfs_root *root,
915 struct inode *inode, u64 newer_than,
aab110ab 916 u64 *off, u32 thresh)
4cb5300b
CM		 917{
918 struct btrfs_path *path;
919 struct btrfs_key min_key;
4cb5300b
CM		 920 struct extent_buffer *leaf;
921 struct btrfs_file_extent_item *extent;
922 int type;
923 int ret;
a4689d2b 924 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 925
926 path = btrfs_alloc_path();
927 if (!path)
928 return -ENOMEM;
929
a4689d2b 930 min_key.objectid = ino;
4cb5300b
CM		 931 min_key.type = BTRFS_EXTENT_DATA_KEY;
932 min_key.offset = *off;
933
67871254 934 while (1) {
6174d3cb 935 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 936 if (ret != 0)
937 goto none;
f094c9bd 938process_slot:
a4689d2b 939 if (min_key.objectid != ino)
4cb5300b
CM		 940 goto none;
941 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
942 goto none;
943
944 leaf = path->nodes[0];
945 extent = btrfs_item_ptr(leaf, path->slots[0],
946 struct btrfs_file_extent_item);
947
948 type = btrfs_file_extent_type(leaf, extent);
949 if (type == BTRFS_FILE_EXTENT_REG &&
950 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
951 check_defrag_in_cache(inode, min_key.offset, thresh)) {
952 *off = min_key.offset;
953 btrfs_free_path(path);
954 return 0;
955 }
956
f094c9bd
FM		 957 path->slots[0]++;
958 if (path->slots[0] < btrfs_header_nritems(leaf)) {
959 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
960 goto process_slot;
961 }
962
4cb5300b
CM		 963 if (min_key.offset == (u64)-1)
964 goto none;
965
966 min_key.offset++;
967 btrfs_release_path(path);
968 }
969none:
970 btrfs_free_path(path);
971 return -ENOENT;
972}
973
6c282eb4 974static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 975{
976 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 977 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
978 struct extent_map *em;
979 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 980
6c282eb4
LZ		 981 /*
982 * hopefully we have this extent in the tree already, try without
983 * the full extent lock
984 */
17ce6ef8 985 read_lock(&em_tree->lock);
6c282eb4 986 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 987 read_unlock(&em_tree->lock);
988
6c282eb4 989 if (!em) {
308d9800
FM		 990 struct extent_state *cached = NULL;
991 u64 end = start + len - 1;
992
6c282eb4 993 /* get the big lock and read metadata off disk */
308d9800 994 lock_extent_bits(io_tree, start, end, 0, &cached);
6c282eb4 995 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
308d9800 996 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
6c282eb4
LZ		 997
998 if (IS_ERR(em))
999 return NULL;
1000 }
1001
1002 return em;
1003}
17ce6ef8 1004
6c282eb4
LZ		 1005static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1006{
1007 struct extent_map *next;
1008 bool ret = true;
1009
1010 /* this is the last extent */
1011 if (em->start + em->len >= i_size_read(inode))
1012 return false;
1013
1014 next = defrag_lookup_extent(inode, em->start + em->len);
e9512d72
CM		 1015 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1016 ret = false;
1017 else if ((em->block_start + em->block_len == next->block_start) &&
1018 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
6c282eb4
LZ		 1019 ret = false;
1020
1021 free_extent_map(next);
17ce6ef8
LB		 1022 return ret;
1023}
1024
aab110ab 1025static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
a43a2111
AM		 1026 u64 *last_len, u64 *skip, u64 *defrag_end,
1027 int compress)
940100a4 1028{
6c282eb4 1029 struct extent_map *em;
940100a4 1030 int ret = 1;
6c282eb4 1031 bool next_mergeable = true;
940100a4
CM		 1032
1033 /*
008873ea 1034 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1035 * defragging it
1036 */
1037 if (start < *defrag_end)
1038 return 1;
1039
1040 *skip = 0;
1041
6c282eb4
LZ		 1042 em = defrag_lookup_extent(inode, start);
1043 if (!em)
1044 return 0;
940100a4
CM		 1045
1046 /* this will cover holes, and inline extents */
17ce6ef8 1047 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1048 ret = 0;
17ce6ef8
LB		 1049 goto out;
1050 }
1051
6c282eb4 1052 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1053 /*
6c282eb4
LZ		 1054 * we hit a real extent, if it is big or the next extent is not a
1055 * real extent, don't bother defragging it
940100a4 1056 */
a43a2111 1057 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 1058 (em->len >= thresh || !next_mergeable))
940100a4 1059 ret = 0;
17ce6ef8 1060out:
940100a4
CM		 1061 /*
1062 * last_len ends up being a counter of how many bytes we've defragged.
1063 * every time we choose not to defrag an extent, we reset *last_len
1064 * so that the next tiny extent will force a defrag.
1065 *
1066 * The end result of this is that tiny extents before a single big
1067 * extent will force at least part of that big extent to be defragged.
1068 */
1069 if (ret) {
940100a4
CM		 1070 *defrag_end = extent_map_end(em);
1071 } else {
1072 *last_len = 0;
1073 *skip = extent_map_end(em);
1074 *defrag_end = 0;
1075 }
1076
1077 free_extent_map(em);
1078 return ret;
1079}
1080
4cb5300b
CM		 1081/*
1082 * it doesn't do much good to defrag one or two pages
1083 * at a time. This pulls in a nice chunk of pages
1084 * to COW and defrag.
1085 *
1086 * It also makes sure the delalloc code has enough
1087 * dirty data to avoid making new small extents as part
1088 * of the defrag
1089 *
1090 * It's a good idea to start RA on this range
1091 * before calling this.
1092 */
1093static int cluster_pages_for_defrag(struct inode *inode,
1094 struct page **pages,
1095 unsigned long start_index,
c41570c9 1096 unsigned long num_pages)
f46b5a66 1097{
4cb5300b
CM		 1098 unsigned long file_end;
1099 u64 isize = i_size_read(inode);
1100 u64 page_start;
1101 u64 page_end;
1f12bd06 1102 u64 page_cnt;
4cb5300b
CM		 1103 int ret;
1104 int i;
1105 int i_done;
3eaa2885 1106 struct btrfs_ordered_extent *ordered;
4cb5300b 1107 struct extent_state *cached_state = NULL;
600a45e1 1108 struct extent_io_tree *tree;
3b16a4e3 1109 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1110
4cb5300b 1111 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1112 if (!isize || start_index > file_end)
1113 return 0;
1114
1115 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1116
1117 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1118 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1119 if (ret)
1120 return ret;
4cb5300b 1121 i_done = 0;
600a45e1 1122 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1123
1124 /* step one, lock all the pages */
1f12bd06 1125 for (i = 0; i < page_cnt; i++) {
4cb5300b 1126 struct page *page;
600a45e1 1127again:
a94733d0 1128 page = find_or_create_page(inode->i_mapping,
600a45e1 1129 start_index + i, mask);
4cb5300b
CM		 1130 if (!page)
1131 break;
1132
600a45e1
MX		 1133 page_start = page_offset(page);
1134 page_end = page_start + PAGE_CACHE_SIZE - 1;
1135 while (1) {
308d9800
FM		 1136 lock_extent_bits(tree, page_start, page_end,
1137 0, &cached_state);
600a45e1
MX		 1138 ordered = btrfs_lookup_ordered_extent(inode,
1139 page_start);
308d9800
FM		 1140 unlock_extent_cached(tree, page_start, page_end,
1141 &cached_state, GFP_NOFS);
600a45e1
MX		 1142 if (!ordered)
1143 break;
1144
1145 unlock_page(page);
1146 btrfs_start_ordered_extent(inode, ordered, 1);
1147 btrfs_put_ordered_extent(ordered);
1148 lock_page(page);
1f12bd06
LB		 1149 /*
1150 * we unlocked the page above, so we need check if
1151 * it was released or not.
1152 */
1153 if (page->mapping != inode->i_mapping) {
1154 unlock_page(page);
1155 page_cache_release(page);
1156 goto again;
1157 }
600a45e1
MX		 1158 }
1159
4cb5300b
CM		 1160 if (!PageUptodate(page)) {
1161 btrfs_readpage(NULL, page);
1162 lock_page(page);
1163 if (!PageUptodate(page)) {
1164 unlock_page(page);
1165 page_cache_release(page);
1166 ret = -EIO;
1167 break;
1168 }
1169 }
600a45e1 1170
600a45e1
MX		 1171 if (page->mapping != inode->i_mapping) {
1172 unlock_page(page);
1173 page_cache_release(page);
1174 goto again;
1175 }
1176
4cb5300b
CM		 1177 pages[i] = page;
1178 i_done++;
1179 }
1180 if (!i_done || ret)
1181 goto out;
1182
1183 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1184 goto out;
1185
1186 /*
1187 * so now we have a nice long stream of locked
1188 * and up to date pages, lets wait on them
1189 */
1190 for (i = 0; i < i_done; i++)
1191 wait_on_page_writeback(pages[i]);
1192
1193 page_start = page_offset(pages[0]);
1194 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1195
1196 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1197 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1198 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1199 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1200 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1201 &cached_state, GFP_NOFS);
4cb5300b 1202
1f12bd06 1203 if (i_done != page_cnt) {
9e0baf60
JB		 1204 spin_lock(&BTRFS_I(inode)->lock);
1205 BTRFS_I(inode)->outstanding_extents++;
1206 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1207 btrfs_delalloc_release_space(inode,
1f12bd06 1208 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1209 }
1210
1211
9e8a4a8b
LB		 1212 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1213 &cached_state, GFP_NOFS);
4cb5300b
CM		 1214
1215 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1216 page_start, page_end - 1, &cached_state,
1217 GFP_NOFS);
1218
1219 for (i = 0; i < i_done; i++) {
1220 clear_page_dirty_for_io(pages[i]);
1221 ClearPageChecked(pages[i]);
1222 set_page_extent_mapped(pages[i]);
1223 set_page_dirty(pages[i]);
1224 unlock_page(pages[i]);
1225 page_cache_release(pages[i]);
1226 }
1227 return i_done;
1228out:
1229 for (i = 0; i < i_done; i++) {
1230 unlock_page(pages[i]);
1231 page_cache_release(pages[i]);
1232 }
1f12bd06 1233 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1234 return ret;
1235
1236}
1237
1238int btrfs_defrag_file(struct inode *inode, struct file *file,
1239 struct btrfs_ioctl_defrag_range_args *range,
1240 u64 newer_than, unsigned long max_to_defrag)
1241{
1242 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1243 struct file_ra_state *ra = NULL;
f46b5a66 1244 unsigned long last_index;
151a31b2 1245 u64 isize = i_size_read(inode);
940100a4
CM		 1246 u64 last_len = 0;
1247 u64 skip = 0;
1248 u64 defrag_end = 0;
4cb5300b 1249 u64 newer_off = range->start;
f46b5a66 1250 unsigned long i;
008873ea 1251 unsigned long ra_index = 0;
f46b5a66 1252 int ret;
4cb5300b 1253 int defrag_count = 0;
1a419d85 1254 int compress_type = BTRFS_COMPRESS_ZLIB;
aab110ab 1255 u32 extent_thresh = range->extent_thresh;
c41570c9
JM		 1256 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1257 unsigned long cluster = max_cluster;
4cb5300b
CM		 1258 u64 new_align = ~((u64)128 * 1024 - 1);
1259 struct page **pages = NULL;
1260
0abd5b17
LB		 1261 if (isize == 0)
1262 return 0;
1263
1264 if (range->start >= isize)
1265 return -EINVAL;
1a419d85
LZ		 1266
1267 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1268 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1269 return -EINVAL;
1270 if (range->compress_type)
1271 compress_type = range->compress_type;
1272 }
f46b5a66 1273
0abd5b17
LB		 1274 if (extent_thresh == 0)
1275 extent_thresh = 256 * 1024;
940100a4 1276
4cb5300b
CM		 1277 /*
1278 * if we were not given a file, allocate a readahead
1279 * context
1280 */
1281 if (!file) {
1282 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1283 if (!ra)
1284 return -ENOMEM;
1285 file_ra_state_init(ra, inode->i_mapping);
1286 } else {
1287 ra = &file->f_ra;
1288 }
1289
d9b0d9ba 1290 pages = kmalloc_array(max_cluster, sizeof(struct page *),
4cb5300b
CM		 1291 GFP_NOFS);
1292 if (!pages) {
1293 ret = -ENOMEM;
1294 goto out_ra;
1295 }
1296
1297 /* find the last page to defrag */
1e701a32 1298 if (range->start + range->len > range->start) {
151a31b2 1299 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1300 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1301 } else {
151a31b2 1302 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1303 }
1304
4cb5300b
CM		 1305 if (newer_than) {
1306 ret = find_new_extents(root, inode, newer_than,
1307 &newer_off, 64 * 1024);
1308 if (!ret) {
1309 range->start = newer_off;
1310 /*
1311 * we always align our defrag to help keep
1312 * the extents in the file evenly spaced
1313 */
1314 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1315 } else
1316 goto out_ra;
1317 } else {
1318 i = range->start >> PAGE_CACHE_SHIFT;
1319 }
1320 if (!max_to_defrag)
070034bd 1321 max_to_defrag = last_index - i + 1;
4cb5300b 1322
2a0f7f57
LZ		 1323 /*
1324 * make writeback starts from i, so the defrag range can be
1325 * written sequentially.
1326 */
1327 if (i < inode->i_mapping->writeback_index)
1328 inode->i_mapping->writeback_index = i;
1329
f7f43cc8 1330 while (i <= last_index && defrag_count < max_to_defrag &&
ed6078f7 1331 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
4cb5300b
CM		 1332 /*
1333 * make sure we stop running if someone unmounts
1334 * the FS
1335 */
1336 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1337 break;
1338
210549eb 1339 if (btrfs_defrag_cancelled(root->fs_info)) {
efe120a0 1340 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
210549eb
DS		 1341 ret = -EAGAIN;
1342 break;
1343 }
1344
66c26892 1345 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1346 extent_thresh, &last_len, &skip,
a43a2111
AM		 1347 &defrag_end, range->flags &
1348 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1349 unsigned long next;
1350 /*
1351 * the should_defrag function tells us how much to skip
1352 * bump our counter by the suggested amount
1353 */
ed6078f7 1354 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
940100a4
CM		 1355 i = max(i + 1, next);
1356 continue;
1357 }
008873ea
LZ		 1358
1359 if (!newer_than) {
1360 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1361 PAGE_CACHE_SHIFT) - i;
1362 cluster = min(cluster, max_cluster);
1363 } else {
1364 cluster = max_cluster;
1365 }
1366
008873ea
LZ		 1367 if (i + cluster > ra_index) {
1368 ra_index = max(i, ra_index);
1369 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1370 cluster);
e4826a5b 1371 ra_index += cluster;
008873ea 1372 }
940100a4 1373
ecb8bea8 1374 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1375 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1376 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1377 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1378 if (ret < 0) {
1379 mutex_unlock(&inode->i_mutex);
4cb5300b 1380 goto out_ra;
ecb8bea8 1381 }
4cb5300b
CM		 1382
1383 defrag_count += ret;
d0e1d66b 1384 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1385 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1386
1387 if (newer_than) {
1388 if (newer_off == (u64)-1)
1389 break;
1390
e1f041e1
LB		 1391 if (ret > 0)
1392 i += ret;
1393
4cb5300b
CM		 1394 newer_off = max(newer_off + 1,
1395 (u64)i << PAGE_CACHE_SHIFT);
1396
1397 ret = find_new_extents(root, inode,
1398 newer_than, &newer_off,
1399 64 * 1024);
1400 if (!ret) {
1401 range->start = newer_off;
1402 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1403 } else {
1404 break;
f46b5a66 1405 }
4cb5300b 1406 } else {
008873ea 1407 if (ret > 0) {
cbcc8326 1408 i += ret;
008873ea
LZ		 1409 last_len += ret << PAGE_CACHE_SHIFT;
1410 } else {
cbcc8326 1411 i++;
008873ea
LZ		 1412 last_len = 0;
1413 }
f46b5a66 1414 }
f46b5a66
CH		 1415 }
1416
dec8ef90 1417 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1418 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1419 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1420 &BTRFS_I(inode)->runtime_flags))
1421 filemap_flush(inode->i_mapping);
1422 }
1e701a32
CM		 1423
1424 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1425 /* the filemap_flush will queue IO into the worker threads, but
1426 * we have to make sure the IO is actually started and that
1427 * ordered extents get created before we return
1428 */
1429 atomic_inc(&root->fs_info->async_submit_draining);
1430 while (atomic_read(&root->fs_info->nr_async_submits) ||
1431 atomic_read(&root->fs_info->async_delalloc_pages)) {
1432 wait_event(root->fs_info->async_submit_wait,
1433 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1434 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1435 }
1436 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1437 }
1438
1a419d85 1439 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1440 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1441 }
1442
60ccf82f 1443 ret = defrag_count;
940100a4 1444
4cb5300b 1445out_ra:
633085c7
FDBM		 1446 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1447 mutex_lock(&inode->i_mutex);
1448 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1449 mutex_unlock(&inode->i_mutex);
1450 }
4cb5300b
CM		 1451 if (!file)
1452 kfree(ra);
1453 kfree(pages);
940100a4 1454 return ret;
f46b5a66
CH		 1455}
1456
198605a8 1457static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1458 void __user *arg)
f46b5a66
CH		 1459{
1460 u64 new_size;
1461 u64 old_size;
1462 u64 devid = 1;
496ad9aa 1463 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1464 struct btrfs_ioctl_vol_args *vol_args;
1465 struct btrfs_trans_handle *trans;
1466 struct btrfs_device *device = NULL;
1467 char *sizestr;
9a40f122 1468 char *retptr;
f46b5a66
CH		 1469 char *devstr = NULL;
1470 int ret = 0;
f46b5a66
CH		 1471 int mod = 0;
1472
e441d54d
CM		 1473 if (!capable(CAP_SYS_ADMIN))
1474 return -EPERM;
1475
198605a8
MX		 1476 ret = mnt_want_write_file(file);
1477 if (ret)
1478 return ret;
1479
5ac00add
SB		 1480 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1481 1)) {
97547676 1482 mnt_drop_write_file(file);
e57138b3 1483 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1484 }
1485
5ac00add 1486 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1487 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1488 if (IS_ERR(vol_args)) {
1489 ret = PTR_ERR(vol_args);
1490 goto out;
1491 }
5516e595
MF		 1492
1493 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1494
f46b5a66
CH		 1495 sizestr = vol_args->name;
1496 devstr = strchr(sizestr, ':');
1497 if (devstr) {
f46b5a66
CH		 1498 sizestr = devstr + 1;
1499 *devstr = '\0';
1500 devstr = vol_args->name;
58dfae63
Z		 1501 ret = kstrtoull(devstr, 10, &devid);
1502 if (ret)
1503 goto out_free;
dfd79829
MX		 1504 if (!devid) {
1505 ret = -EINVAL;
1506 goto out_free;
1507 }
efe120a0 1508 btrfs_info(root->fs_info, "resizing devid %llu", devid);
f46b5a66 1509 }
dba60f3f 1510
aa1b8cd4 1511 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1512 if (!device) {
efe120a0 1513 btrfs_info(root->fs_info, "resizer unable to find device %llu",
c1c9ff7c 1514 devid);
dfd79829 1515 ret = -ENODEV;
c9e9f97b 1516 goto out_free;
f46b5a66 1517 }
dba60f3f
MX		 1518
1519 if (!device->writeable) {
efe120a0
FH		 1520 btrfs_info(root->fs_info,
1521 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1522 devid);
dfd79829 1523 ret = -EPERM;
4e42ae1b
LB		 1524 goto out_free;
1525 }
1526
f46b5a66
CH		 1527 if (!strcmp(sizestr, "max"))
1528 new_size = device->bdev->bd_inode->i_size;
1529 else {
1530 if (sizestr[0] == '-') {
1531 mod = -1;
1532 sizestr++;
1533 } else if (sizestr[0] == '+') {
1534 mod = 1;
1535 sizestr++;
1536 }
9a40f122
GH		 1537 new_size = memparse(sizestr, &retptr);
1538 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1539 ret = -EINVAL;
c9e9f97b 1540 goto out_free;
f46b5a66
CH		 1541 }
1542 }
1543
63a212ab 1544 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1545 ret = -EPERM;
63a212ab
SB		 1546 goto out_free;
1547 }
1548
7cc8e58d 1549 old_size = btrfs_device_get_total_bytes(device);
f46b5a66
CH		 1550
1551 if (mod < 0) {
1552 if (new_size > old_size) {
1553 ret = -EINVAL;
c9e9f97b 1554 goto out_free;
f46b5a66
CH		 1555 }
1556 new_size = old_size - new_size;
1557 } else if (mod > 0) {
eb8052e0 1558 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1559 ret = -ERANGE;
eb8052e0
WF		 1560 goto out_free;
1561 }
f46b5a66
CH		 1562 new_size = old_size + new_size;
1563 }
1564
1565 if (new_size < 256 * 1024 * 1024) {
1566 ret = -EINVAL;
c9e9f97b 1567 goto out_free;
f46b5a66
CH		 1568 }
1569 if (new_size > device->bdev->bd_inode->i_size) {
1570 ret = -EFBIG;
c9e9f97b 1571 goto out_free;
f46b5a66
CH		 1572 }
1573
b8b93add 1574 new_size = div_u64(new_size, root->sectorsize);
f46b5a66
CH		 1575 new_size *= root->sectorsize;
1576
efe120a0 1577 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
c1c9ff7c 1578 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1579
1580 if (new_size > old_size) {
a22285a6 1581 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1582 if (IS_ERR(trans)) {
1583 ret = PTR_ERR(trans);
c9e9f97b 1584 goto out_free;
98d5dc13 1585 }
f46b5a66
CH		 1586 ret = btrfs_grow_device(trans, device, new_size);
1587 btrfs_commit_transaction(trans, root);
ece7d20e 1588 } else if (new_size < old_size) {
f46b5a66 1589 ret = btrfs_shrink_device(device, new_size);
0253f40e 1590 } /* equal, nothing need to do */
f46b5a66 1591
c9e9f97b 1592out_free:
f46b5a66 1593 kfree(vol_args);
c9e9f97b
ID		 1594out:
1595 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1596 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1597 mnt_drop_write_file(file);
f46b5a66
CH		 1598 return ret;
1599}
1600
72fd032e 1601static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1602 char *name, unsigned long fd, int subvol,
1603 u64 *transid, bool readonly,
8696c533 1604 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1605{
f46b5a66 1606 int namelen;
3de4586c 1607 int ret = 0;
f46b5a66 1608
a874a63e
LB		 1609 ret = mnt_want_write_file(file);
1610 if (ret)
1611 goto out;
1612
72fd032e
SW		 1613 namelen = strlen(name);
1614 if (strchr(name, '/')) {
f46b5a66 1615 ret = -EINVAL;
a874a63e 1616 goto out_drop_write;
f46b5a66
CH		 1617 }
1618
16780cab
CM		 1619 if (name[0] == '.' &&
1620 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1621 ret = -EEXIST;
a874a63e 1622 goto out_drop_write;
16780cab
CM		 1623 }
1624
3de4586c 1625 if (subvol) {
72fd032e 1626 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1627 NULL, transid, readonly, inherit);
cb8e7090 1628 } else {
2903ff01 1629 struct fd src = fdget(fd);
3de4586c 1630 struct inode *src_inode;
2903ff01 1631 if (!src.file) {
3de4586c 1632 ret = -EINVAL;
a874a63e 1633 goto out_drop_write;
3de4586c
CM		 1634 }
1635
496ad9aa
AV		 1636 src_inode = file_inode(src.file);
1637 if (src_inode->i_sb != file_inode(file)->i_sb) {
efe120a0
FH		 1638 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1639 "Snapshot src from another FS");
23ad5b17 1640 ret = -EXDEV;
d0242061
DS		 1641 } else if (!inode_owner_or_capable(src_inode)) {
1642 /*
1643 * Subvolume creation is not restricted, but snapshots
1644 * are limited to own subvolumes only
1645 */
1646 ret = -EPERM;
ecd18815
AV		 1647 } else {
1648 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1649 BTRFS_I(src_inode)->root,
1650 transid, readonly, inherit);
3de4586c 1651 }
2903ff01 1652 fdput(src);
cb8e7090 1653 }
a874a63e
LB		 1654out_drop_write:
1655 mnt_drop_write_file(file);
f46b5a66 1656out:
72fd032e
SW		 1657 return ret;
1658}
1659
1660static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1661 void __user *arg, int subvol)
72fd032e 1662{
fa0d2b9b 1663 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1664 int ret;
1665
fa0d2b9b
LZ		 1666 vol_args = memdup_user(arg, sizeof(*vol_args));
1667 if (IS_ERR(vol_args))
1668 return PTR_ERR(vol_args);
1669 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1670
fa0d2b9b 1671 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1672 vol_args->fd, subvol,
6f72c7e2 1673 NULL, false, NULL);
fdfb1e4f 1674
fa0d2b9b
LZ		 1675 kfree(vol_args);
1676 return ret;
1677}
fdfb1e4f 1678
fa0d2b9b
LZ		 1679static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1680 void __user *arg, int subvol)
1681{
1682 struct btrfs_ioctl_vol_args_v2 *vol_args;
1683 int ret;
1684 u64 transid = 0;
1685 u64 *ptr = NULL;
b83cc969 1686 bool readonly = false;
6f72c7e2 1687 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1688
fa0d2b9b
LZ		 1689 vol_args = memdup_user(arg, sizeof(*vol_args));
1690 if (IS_ERR(vol_args))
1691 return PTR_ERR(vol_args);
1692 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1693
b83cc969 1694 if (vol_args->flags &
6f72c7e2
AJ		 1695 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1696 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1697 ret = -EOPNOTSUPP;
c47ca32d 1698 goto free_args;
72fd032e 1699 }
fa0d2b9b
LZ		 1700
1701 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1702 ptr = &transid;
b83cc969
LZ		 1703 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1704 readonly = true;
6f72c7e2
AJ		 1705 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1706 if (vol_args->size > PAGE_CACHE_SIZE) {
1707 ret = -EINVAL;
c47ca32d 1708 goto free_args;
6f72c7e2
AJ		 1709 }
1710 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1711 if (IS_ERR(inherit)) {
1712 ret = PTR_ERR(inherit);
c47ca32d 1713 goto free_args;
6f72c7e2
AJ		 1714 }
1715 }
fa0d2b9b
LZ		 1716
1717 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1718 vol_args->fd, subvol, ptr,
8696c533 1719 readonly, inherit);
c47ca32d
DC		 1720 if (ret)
1721 goto free_inherit;
fa0d2b9b 1722
c47ca32d
DC		 1723 if (ptr && copy_to_user(arg +
1724 offsetof(struct btrfs_ioctl_vol_args_v2,
1725 transid),
1726 ptr, sizeof(*ptr)))
fa0d2b9b 1727 ret = -EFAULT;
c47ca32d
DC		 1728
1729free_inherit:
6f72c7e2 1730 kfree(inherit);
c47ca32d
DC		 1731free_args:
1732 kfree(vol_args);
f46b5a66
CH		 1733 return ret;
1734}
1735
0caa102d
LZ		 1736static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1737 void __user *arg)
1738{
496ad9aa 1739 struct inode *inode = file_inode(file);
0caa102d
LZ		 1740 struct btrfs_root *root = BTRFS_I(inode)->root;
1741 int ret = 0;
1742 u64 flags = 0;
1743
33345d01 1744 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1745 return -EINVAL;
1746
1747 down_read(&root->fs_info->subvol_sem);
1748 if (btrfs_root_readonly(root))
1749 flags |= BTRFS_SUBVOL_RDONLY;
1750 up_read(&root->fs_info->subvol_sem);
1751
1752 if (copy_to_user(arg, &flags, sizeof(flags)))
1753 ret = -EFAULT;
1754
1755 return ret;
1756}
1757
1758static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1759 void __user *arg)
1760{
496ad9aa 1761 struct inode *inode = file_inode(file);
0caa102d
LZ		 1762 struct btrfs_root *root = BTRFS_I(inode)->root;
1763 struct btrfs_trans_handle *trans;
1764 u64 root_flags;
1765 u64 flags;
1766 int ret = 0;
1767
bd60ea0f
DS		 1768 if (!inode_owner_or_capable(inode))
1769 return -EPERM;
1770
b9ca0664
LB		 1771 ret = mnt_want_write_file(file);
1772 if (ret)
1773 goto out;
0caa102d 1774
b9ca0664
LB		 1775 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1776 ret = -EINVAL;
1777 goto out_drop_write;
1778 }
0caa102d 1779
b9ca0664
LB		 1780 if (copy_from_user(&flags, arg, sizeof(flags))) {
1781 ret = -EFAULT;
1782 goto out_drop_write;
1783 }
0caa102d 1784
b9ca0664
LB		 1785 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1786 ret = -EINVAL;
1787 goto out_drop_write;
1788 }
0caa102d 1789
b9ca0664
LB		 1790 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1791 ret = -EOPNOTSUPP;
1792 goto out_drop_write;
1793 }
0caa102d
LZ		 1794
1795 down_write(&root->fs_info->subvol_sem);
1796
1797 /* nothing to do */
1798 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1799 goto out_drop_sem;
0caa102d
LZ		 1800
1801 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1802 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1803 btrfs_set_root_flags(&root->root_item,
1804 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1805 } else {
1806 /*
1807 * Block RO -> RW transition if this subvolume is involved in
1808 * send
1809 */
1810 spin_lock(&root->root_item_lock);
1811 if (root->send_in_progress == 0) {
1812 btrfs_set_root_flags(&root->root_item,
0caa102d 1813 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1814 spin_unlock(&root->root_item_lock);
1815 } else {
1816 spin_unlock(&root->root_item_lock);
1817 btrfs_warn(root->fs_info,
1818 "Attempt to set subvolume %llu read-write during send",
1819 root->root_key.objectid);
1820 ret = -EPERM;
1821 goto out_drop_sem;
1822 }
1823 }
0caa102d
LZ		 1824
1825 trans = btrfs_start_transaction(root, 1);
1826 if (IS_ERR(trans)) {
1827 ret = PTR_ERR(trans);
1828 goto out_reset;
1829 }
1830
b4dc2b8c 1831 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1832 &root->root_key, &root->root_item);
1833
1834 btrfs_commit_transaction(trans, root);
1835out_reset:
1836 if (ret)
1837 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1838out_drop_sem:
0caa102d 1839 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1840out_drop_write:
1841 mnt_drop_write_file(file);
1842out:
0caa102d
LZ		 1843 return ret;
1844}
1845
76dda93c
YZ		 1846/*
1847 * helper to check if the subvolume references other subvolumes
1848 */
1849static noinline int may_destroy_subvol(struct btrfs_root *root)
1850{
1851 struct btrfs_path *path;
175a2b87 1852 struct btrfs_dir_item *di;
76dda93c 1853 struct btrfs_key key;
175a2b87 1854 u64 dir_id;
76dda93c
YZ		 1855 int ret;
1856
1857 path = btrfs_alloc_path();
1858 if (!path)
1859 return -ENOMEM;
1860
175a2b87
JB		 1861 /* Make sure this root isn't set as the default subvol */
1862 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1863 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1864 dir_id, "default", 7, 0);
1865 if (di && !IS_ERR(di)) {
1866 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1867 if (key.objectid == root->root_key.objectid) {
72de6b53
GS		 1868 ret = -EPERM;
1869 btrfs_err(root->fs_info, "deleting default subvolume "
1870 "%llu is not allowed", key.objectid);
175a2b87
JB		 1871 goto out;
1872 }
1873 btrfs_release_path(path);
1874 }
1875
76dda93c
YZ		 1876 key.objectid = root->root_key.objectid;
1877 key.type = BTRFS_ROOT_REF_KEY;
1878 key.offset = (u64)-1;
1879
1880 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1881 &key, path, 0, 0);
1882 if (ret < 0)
1883 goto out;
1884 BUG_ON(ret == 0);
1885
1886 ret = 0;
1887 if (path->slots[0] > 0) {
1888 path->slots[0]--;
1889 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1890 if (key.objectid == root->root_key.objectid &&
1891 key.type == BTRFS_ROOT_REF_KEY)
1892 ret = -ENOTEMPTY;
1893 }
1894out:
1895 btrfs_free_path(path);
1896 return ret;
1897}
1898
ac8e9819
CM		 1899static noinline int key_in_sk(struct btrfs_key *key,
1900 struct btrfs_ioctl_search_key *sk)
1901{
abc6e134
CM		 1902 struct btrfs_key test;
1903 int ret;
1904
1905 test.objectid = sk->min_objectid;
1906 test.type = sk->min_type;
1907 test.offset = sk->min_offset;
1908
1909 ret = btrfs_comp_cpu_keys(key, &test);
1910 if (ret < 0)
ac8e9819 1911 return 0;
abc6e134
CM		 1912
1913 test.objectid = sk->max_objectid;
1914 test.type = sk->max_type;
1915 test.offset = sk->max_offset;
1916
1917 ret = btrfs_comp_cpu_keys(key, &test);
1918 if (ret > 0)
ac8e9819
CM		 1919 return 0;
1920 return 1;
1921}
1922
1923static noinline int copy_to_sk(struct btrfs_root *root,
1924 struct btrfs_path *path,
1925 struct btrfs_key *key,
1926 struct btrfs_ioctl_search_key *sk,
9b6e817d 1927 size_t *buf_size,
ba346b35 1928 char __user *ubuf,
ac8e9819
CM		 1929 unsigned long *sk_offset,
1930 int *num_found)
1931{
1932 u64 found_transid;
1933 struct extent_buffer *leaf;
1934 struct btrfs_ioctl_search_header sh;
1935 unsigned long item_off;
1936 unsigned long item_len;
1937 int nritems;
1938 int i;
1939 int slot;
ac8e9819
CM		 1940 int ret = 0;
1941
1942 leaf = path->nodes[0];
1943 slot = path->slots[0];
1944 nritems = btrfs_header_nritems(leaf);
1945
1946 if (btrfs_header_generation(leaf) > sk->max_transid) {
1947 i = nritems;
1948 goto advance_key;
1949 }
1950 found_transid = btrfs_header_generation(leaf);
1951
1952 for (i = slot; i < nritems; i++) {
1953 item_off = btrfs_item_ptr_offset(leaf, i);
1954 item_len = btrfs_item_size_nr(leaf, i);
1955
03b71c6c
GP		 1956 btrfs_item_key_to_cpu(leaf, key, i);
1957 if (!key_in_sk(key, sk))
1958 continue;
1959
9b6e817d 1960 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 1961 if (*num_found) {
1962 ret = 1;
1963 goto out;
1964 }
1965
1966 /*
1967 * return one empty item back for v1, which does not
1968 * handle -EOVERFLOW
1969 */
1970
9b6e817d 1971 *buf_size = sizeof(sh) + item_len;
ac8e9819 1972 item_len = 0;
8f5f6178
GH		 1973 ret = -EOVERFLOW;
1974 }
ac8e9819 1975
9b6e817d 1976 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 1977 ret = 1;
25c9bc2e 1978 goto out;
ac8e9819
CM		 1979 }
1980
ac8e9819
CM		 1981 sh.objectid = key->objectid;
1982 sh.offset = key->offset;
1983 sh.type = key->type;
1984 sh.len = item_len;
1985 sh.transid = found_transid;
1986
1987 /* copy search result header */
ba346b35
GH		 1988 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1989 ret = -EFAULT;
1990 goto out;
1991 }
1992
ac8e9819
CM		 1993 *sk_offset += sizeof(sh);
1994
1995 if (item_len) {
ba346b35 1996 char __user *up = ubuf + *sk_offset;
ac8e9819 1997 /* copy the item */
ba346b35
GH		 1998 if (read_extent_buffer_to_user(leaf, up,
1999 item_off, item_len)) {
2000 ret = -EFAULT;
2001 goto out;
2002 }
2003
ac8e9819 2004 *sk_offset += item_len;
ac8e9819 2005 }
e2156867 2006 (*num_found)++;
ac8e9819 2007
8f5f6178
GH		 2008 if (ret) /* -EOVERFLOW from above */
2009 goto out;
2010
25c9bc2e
GH		 2011 if (*num_found >= sk->nr_items) {
2012 ret = 1;
2013 goto out;
2014 }
ac8e9819
CM		 2015 }
2016advance_key:
abc6e134
CM		 2017 ret = 0;
2018 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 2019 key->offset++;
abc6e134
CM		 2020 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2021 key->offset = 0;
ac8e9819 2022 key->type++;
abc6e134
CM		 2023 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2024 key->offset = 0;
2025 key->type = 0;
ac8e9819 2026 key->objectid++;
abc6e134
CM		 2027 } else
2028 ret = 1;
25c9bc2e 2029out:
ba346b35
GH		 2030 /*
2031 * 0: all items from this leaf copied, continue with next
2032 * 1: * more items can be copied, but unused buffer is too small
2033 * * all items were found
2034 * Either way, it will stops the loop which iterates to the next
2035 * leaf
2036 * -EOVERFLOW: item was to large for buffer
2037 * -EFAULT: could not copy extent buffer back to userspace
2038 */
ac8e9819
CM		 2039 return ret;
2040}
2041
2042static noinline int search_ioctl(struct inode *inode,
12544442 2043 struct btrfs_ioctl_search_key *sk,
9b6e817d 2044 size_t *buf_size,
ba346b35 2045 char __user *ubuf)
ac8e9819
CM		 2046{
2047 struct btrfs_root *root;
2048 struct btrfs_key key;
ac8e9819 2049 struct btrfs_path *path;
ac8e9819
CM		 2050 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2051 int ret;
2052 int num_found = 0;
2053 unsigned long sk_offset = 0;
2054
9b6e817d
GH		 2055 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2056 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2057 return -EOVERFLOW;
9b6e817d 2058 }
12544442 2059
ac8e9819
CM		 2060 path = btrfs_alloc_path();
2061 if (!path)
2062 return -ENOMEM;
2063
2064 if (sk->tree_id == 0) {
2065 /* search the root of the inode that was passed */
2066 root = BTRFS_I(inode)->root;
2067 } else {
2068 key.objectid = sk->tree_id;
2069 key.type = BTRFS_ROOT_ITEM_KEY;
2070 key.offset = (u64)-1;
2071 root = btrfs_read_fs_root_no_name(info, &key);
2072 if (IS_ERR(root)) {
efe120a0 2073 printk(KERN_ERR "BTRFS: could not find root %llu\n",
ac8e9819
CM		 2074 sk->tree_id);
2075 btrfs_free_path(path);
2076 return -ENOENT;
2077 }
2078 }
2079
2080 key.objectid = sk->min_objectid;
2081 key.type = sk->min_type;
2082 key.offset = sk->min_offset;
2083
67871254 2084 while (1) {
6174d3cb 2085 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2086 if (ret != 0) {
2087 if (ret > 0)
2088 ret = 0;
2089 goto err;
2090 }
ba346b35 2091 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
ac8e9819 2092 &sk_offset, &num_found);
b3b4aa74 2093 btrfs_release_path(path);
25c9bc2e 2094 if (ret)
ac8e9819
CM		 2095 break;
2096
2097 }
8f5f6178
GH		 2098 if (ret > 0)
2099 ret = 0;
ac8e9819
CM		 2100err:
2101 sk->nr_items = num_found;
2102 btrfs_free_path(path);
2103 return ret;
2104}
2105
2106static noinline int btrfs_ioctl_tree_search(struct file *file,
2107 void __user *argp)
2108{
ba346b35
GH		 2109 struct btrfs_ioctl_search_args __user *uargs;
2110 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2111 struct inode *inode;
2112 int ret;
2113 size_t buf_size;
ac8e9819
CM		 2114
2115 if (!capable(CAP_SYS_ADMIN))
2116 return -EPERM;
2117
ba346b35
GH		 2118 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2119
2120 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2121 return -EFAULT;
ac8e9819 2122
ba346b35 2123 buf_size = sizeof(uargs->buf);
ac8e9819 2124
496ad9aa 2125 inode = file_inode(file);
ba346b35 2126 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2127
2128 /*
2129 * In the origin implementation an overflow is handled by returning a
2130 * search header with a len of zero, so reset ret.
2131 */
2132 if (ret == -EOVERFLOW)
2133 ret = 0;
2134
ba346b35 2135 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2136 ret = -EFAULT;
ac8e9819
CM		 2137 return ret;
2138}
2139
cc68a8a5
GH		 2140static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2141 void __user *argp)
2142{
2143 struct btrfs_ioctl_search_args_v2 __user *uarg;
2144 struct btrfs_ioctl_search_args_v2 args;
2145 struct inode *inode;
2146 int ret;
2147 size_t buf_size;
2148 const size_t buf_limit = 16 * 1024 * 1024;
2149
2150 if (!capable(CAP_SYS_ADMIN))
2151 return -EPERM;
2152
2153 /* copy search header and buffer size */
2154 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2155 if (copy_from_user(&args, uarg, sizeof(args)))
2156 return -EFAULT;
2157
2158 buf_size = args.buf_size;
2159
2160 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2161 return -EOVERFLOW;
2162
2163 /* limit result size to 16MB */
2164 if (buf_size > buf_limit)
2165 buf_size = buf_limit;
2166
2167 inode = file_inode(file);
2168 ret = search_ioctl(inode, &args.key, &buf_size,
2169 (char *)(&uarg->buf[0]));
2170 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2171 ret = -EFAULT;
2172 else if (ret == -EOVERFLOW &&
2173 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2174 ret = -EFAULT;
2175
ac8e9819
CM		 2176 return ret;
2177}
2178
98d377a0 2179/*
ac8e9819
CM		 2180 * Search INODE_REFs to identify path name of 'dirid' directory
2181 * in a 'tree_id' tree. and sets path name to 'name'.
2182 */
98d377a0
TH		 2183static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2184 u64 tree_id, u64 dirid, char *name)
2185{
2186 struct btrfs_root *root;
2187 struct btrfs_key key;
ac8e9819 2188 char *ptr;
98d377a0
TH		 2189 int ret = -1;
2190 int slot;
2191 int len;
2192 int total_len = 0;
2193 struct btrfs_inode_ref *iref;
2194 struct extent_buffer *l;
2195 struct btrfs_path *path;
2196
2197 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2198 name[0]='\0';
2199 return 0;
2200 }
2201
2202 path = btrfs_alloc_path();
2203 if (!path)
2204 return -ENOMEM;
2205
ac8e9819 2206 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2207
2208 key.objectid = tree_id;
2209 key.type = BTRFS_ROOT_ITEM_KEY;
2210 key.offset = (u64)-1;
2211 root = btrfs_read_fs_root_no_name(info, &key);
2212 if (IS_ERR(root)) {
efe120a0 2213 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
8ad6fcab
CM		 2214 ret = -ENOENT;
2215 goto out;
98d377a0
TH		 2216 }
2217
2218 key.objectid = dirid;
2219 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2220 key.offset = (u64)-1;
98d377a0 2221
67871254 2222 while (1) {
98d377a0
TH		 2223 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2224 if (ret < 0)
2225 goto out;
18674c6c
FDBM		 2226 else if (ret > 0) {
2227 ret = btrfs_previous_item(root, path, dirid,
2228 BTRFS_INODE_REF_KEY);
2229 if (ret < 0)
2230 goto out;
2231 else if (ret > 0) {
2232 ret = -ENOENT;
2233 goto out;
2234 }
2235 }
98d377a0
TH		 2236
2237 l = path->nodes[0];
2238 slot = path->slots[0];
2239 btrfs_item_key_to_cpu(l, &key, slot);
2240
98d377a0
TH		 2241 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2242 len = btrfs_inode_ref_name_len(l, iref);
2243 ptr -= len + 1;
2244 total_len += len + 1;
a696cf35
FDBM		 2245 if (ptr < name) {
2246 ret = -ENAMETOOLONG;
98d377a0 2247 goto out;
a696cf35 2248 }
98d377a0
TH		 2249
2250 *(ptr + len) = '/';
67871254 2251 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2252
2253 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2254 break;
2255
b3b4aa74 2256 btrfs_release_path(path);
98d377a0 2257 key.objectid = key.offset;
8ad6fcab 2258 key.offset = (u64)-1;
98d377a0 2259 dirid = key.objectid;
98d377a0 2260 }
77906a50 2261 memmove(name, ptr, total_len);
67871254 2262 name[total_len] = '\0';
98d377a0
TH		 2263 ret = 0;
2264out:
2265 btrfs_free_path(path);
ac8e9819
CM		 2266 return ret;
2267}
2268
2269static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2270 void __user *argp)
2271{
2272 struct btrfs_ioctl_ino_lookup_args *args;
2273 struct inode *inode;
01b810b8 2274 int ret = 0;
ac8e9819 2275
2354d08f
JL		 2276 args = memdup_user(argp, sizeof(*args));
2277 if (IS_ERR(args))
2278 return PTR_ERR(args);
c2b96929 2279
496ad9aa 2280 inode = file_inode(file);
ac8e9819 2281
01b810b8
DS		 2282 /*
2283 * Unprivileged query to obtain the containing subvolume root id. The
2284 * path is reset so it's consistent with btrfs_search_path_in_tree.
2285 */
1b53ac4d
CM		 2286 if (args->treeid == 0)
2287 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2288
01b810b8
DS		 2289 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2290 args->name[0] = 0;
2291 goto out;
2292 }
2293
2294 if (!capable(CAP_SYS_ADMIN)) {
2295 ret = -EPERM;
2296 goto out;
2297 }
2298
ac8e9819
CM		 2299 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2300 args->treeid, args->objectid,
2301 args->name);
2302
01b810b8 2303out:
ac8e9819
CM		 2304 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2305 ret = -EFAULT;
2306
2307 kfree(args);
98d377a0
TH		 2308 return ret;
2309}
2310
76dda93c
YZ		 2311static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2312 void __user *arg)
2313{
54563d41 2314 struct dentry *parent = file->f_path.dentry;
76dda93c 2315 struct dentry *dentry;
2b0143b5 2316 struct inode *dir = d_inode(parent);
76dda93c
YZ		 2317 struct inode *inode;
2318 struct btrfs_root *root = BTRFS_I(dir)->root;
2319 struct btrfs_root *dest = NULL;
2320 struct btrfs_ioctl_vol_args *vol_args;
2321 struct btrfs_trans_handle *trans;
c58aaad2 2322 struct btrfs_block_rsv block_rsv;
521e0546 2323 u64 root_flags;
c58aaad2 2324 u64 qgroup_reserved;
76dda93c
YZ		 2325 int namelen;
2326 int ret;
2327 int err = 0;
2328
76dda93c
YZ		 2329 vol_args = memdup_user(arg, sizeof(*vol_args));
2330 if (IS_ERR(vol_args))
2331 return PTR_ERR(vol_args);
2332
2333 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2334 namelen = strlen(vol_args->name);
2335 if (strchr(vol_args->name, '/') ||
2336 strncmp(vol_args->name, "..", namelen) == 0) {
2337 err = -EINVAL;
2338 goto out;
2339 }
2340
a561be71 2341 err = mnt_want_write_file(file);
76dda93c
YZ		 2342 if (err)
2343 goto out;
2344
521e0546 2345
5c50c9b8
DS		 2346 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2347 if (err == -EINTR)
e43f998e 2348 goto out_drop_write;
76dda93c
YZ		 2349 dentry = lookup_one_len(vol_args->name, parent, namelen);
2350 if (IS_ERR(dentry)) {
2351 err = PTR_ERR(dentry);
2352 goto out_unlock_dir;
2353 }
2354
2b0143b5 2355 if (d_really_is_negative(dentry)) {
76dda93c
YZ		 2356 err = -ENOENT;
2357 goto out_dput;
2358 }
2359
2b0143b5 2360 inode = d_inode(dentry);
4260f7c7 2361 dest = BTRFS_I(inode)->root;
67871254 2362 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2363 /*
2364 * Regular user. Only allow this with a special mount
2365 * option, when the user has write+exec access to the
2366 * subvol root, and when rmdir(2) would have been
2367 * allowed.
2368 *
2369 * Note that this is _not_ check that the subvol is
2370 * empty or doesn't contain data that we wouldn't
2371 * otherwise be able to delete.
2372 *
2373 * Users who want to delete empty subvols should try
2374 * rmdir(2).
2375 */
2376 err = -EPERM;
2377 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2378 goto out_dput;
2379
2380 /*
2381 * Do not allow deletion if the parent dir is the same
2382 * as the dir to be deleted. That means the ioctl
2383 * must be called on the dentry referencing the root
2384 * of the subvol, not a random directory contained
2385 * within it.
2386 */
2387 err = -EINVAL;
2388 if (root == dest)
2389 goto out_dput;
2390
2391 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2392 if (err)
2393 goto out_dput;
4260f7c7
SW		 2394 }
2395
5c39da5b
MX		 2396 /* check if subvolume may be deleted by a user */
2397 err = btrfs_may_delete(dir, dentry, 1);
2398 if (err)
2399 goto out_dput;
2400
33345d01 2401 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2402 err = -EINVAL;
2403 goto out_dput;
2404 }
2405
76dda93c 2406 mutex_lock(&inode->i_mutex);
521e0546
DS		 2407
2408 /*
2409 * Don't allow to delete a subvolume with send in progress. This is
2410 * inside the i_mutex so the error handling that has to drop the bit
2411 * again is not run concurrently.
2412 */
2413 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2414 root_flags = btrfs_root_flags(&dest->root_item);
2415 if (dest->send_in_progress == 0) {
2416 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2417 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2418 spin_unlock(&dest->root_item_lock);
2419 } else {
2420 spin_unlock(&dest->root_item_lock);
2421 btrfs_warn(root->fs_info,
2422 "Attempt to delete subvolume %llu during send",
c55bfa67 2423 dest->root_key.objectid);
521e0546 2424 err = -EPERM;
909e26dc 2425 goto out_unlock_inode;
521e0546
DS		 2426 }
2427
76dda93c
YZ		 2428 down_write(&root->fs_info->subvol_sem);
2429
2430 err = may_destroy_subvol(dest);
2431 if (err)
2432 goto out_up_write;
2433
c58aaad2
MX		 2434 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2435 /*
2436 * One for dir inode, two for dir entries, two for root
2437 * ref/backref.
2438 */
2439 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2440 5, &qgroup_reserved, true);
c58aaad2
MX		 2441 if (err)
2442 goto out_up_write;
2443
a22285a6
YZ		 2444 trans = btrfs_start_transaction(root, 0);
2445 if (IS_ERR(trans)) {
2446 err = PTR_ERR(trans);
c58aaad2 2447 goto out_release;
a22285a6 2448 }
c58aaad2
MX		 2449 trans->block_rsv = &block_rsv;
2450 trans->bytes_reserved = block_rsv.size;
a22285a6 2451
76dda93c
YZ		 2452 ret = btrfs_unlink_subvol(trans, root, dir,
2453 dest->root_key.objectid,
2454 dentry->d_name.name,
2455 dentry->d_name.len);
79787eaa
JM		 2456 if (ret) {
2457 err = ret;
2458 btrfs_abort_transaction(trans, root, ret);
2459 goto out_end_trans;
2460 }
76dda93c
YZ		 2461
2462 btrfs_record_root_in_trans(trans, dest);
2463
2464 memset(&dest->root_item.drop_progress, 0,
2465 sizeof(dest->root_item.drop_progress));
2466 dest->root_item.drop_level = 0;
2467 btrfs_set_root_refs(&dest->root_item, 0);
2468
27cdeb70 2469 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
d68fc57b
YZ		 2470 ret = btrfs_insert_orphan_item(trans,
2471 root->fs_info->tree_root,
2472 dest->root_key.objectid);
79787eaa
JM		 2473 if (ret) {
2474 btrfs_abort_transaction(trans, root, ret);
2475 err = ret;
2476 goto out_end_trans;
2477 }
d68fc57b 2478 }
dd5f9615
SB		 2479
2480 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2481 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2482 dest->root_key.objectid);
2483 if (ret && ret != -ENOENT) {
2484 btrfs_abort_transaction(trans, root, ret);
2485 err = ret;
2486 goto out_end_trans;
2487 }
2488 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2489 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2490 dest->root_item.received_uuid,
2491 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2492 dest->root_key.objectid);
2493 if (ret && ret != -ENOENT) {
2494 btrfs_abort_transaction(trans, root, ret);
2495 err = ret;
2496 goto out_end_trans;
2497 }
2498 }
2499
79787eaa 2500out_end_trans:
c58aaad2
MX		 2501 trans->block_rsv = NULL;
2502 trans->bytes_reserved = 0;
531cb13f 2503 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2504 if (ret && !err)
2505 err = ret;
76dda93c 2506 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2507out_release:
2508 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2509out_up_write:
2510 up_write(&root->fs_info->subvol_sem);
521e0546
DS		 2511 if (err) {
2512 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2513 root_flags = btrfs_root_flags(&dest->root_item);
2514 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2515 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2516 spin_unlock(&dest->root_item_lock);
2517 }
909e26dc 2518out_unlock_inode:
76dda93c
YZ		 2519 mutex_unlock(&inode->i_mutex);
2520 if (!err) {
64ad6c48 2521 d_invalidate(dentry);
76dda93c
YZ		 2522 btrfs_invalidate_inodes(dest);
2523 d_delete(dentry);
61155aa0 2524 ASSERT(dest->send_in_progress == 0);
fa6ac876
LB		 2525
2526 /* the last ref */
57cdc8db
DS		 2527 if (dest->ino_cache_inode) {
2528 iput(dest->ino_cache_inode);
2529 dest->ino_cache_inode = NULL;
fa6ac876 2530 }
76dda93c
YZ		 2531 }
2532out_dput:
2533 dput(dentry);
2534out_unlock_dir:
2535 mutex_unlock(&dir->i_mutex);
e43f998e 2536out_drop_write:
2a79f17e 2537 mnt_drop_write_file(file);
76dda93c
YZ		 2538out:
2539 kfree(vol_args);
2540 return err;
2541}
2542
1e701a32 2543static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2544{
496ad9aa 2545 struct inode *inode = file_inode(file);
f46b5a66 2546 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2547 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2548 int ret;
2549
25122d15
ID		 2550 ret = mnt_want_write_file(file);
2551 if (ret)
2552 return ret;
b83cc969 2553
25122d15
ID		 2554 if (btrfs_root_readonly(root)) {
2555 ret = -EROFS;
2556 goto out;
5ac00add 2557 }
f46b5a66
CH		 2558
2559 switch (inode->i_mode & S_IFMT) {
2560 case S_IFDIR:
e441d54d
CM		 2561 if (!capable(CAP_SYS_ADMIN)) {
2562 ret = -EPERM;
2563 goto out;
2564 }
de78b51a 2565 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2566 if (ret)
2567 goto out;
de78b51a 2568 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2569 break;
2570 case S_IFREG:
e441d54d
CM		 2571 if (!(file->f_mode & FMODE_WRITE)) {
2572 ret = -EINVAL;
2573 goto out;
2574 }
1e701a32
CM		 2575
2576 range = kzalloc(sizeof(*range), GFP_KERNEL);
2577 if (!range) {
2578 ret = -ENOMEM;
2579 goto out;
2580 }
2581
2582 if (argp) {
2583 if (copy_from_user(range, argp,
2584 sizeof(*range))) {
2585 ret = -EFAULT;
2586 kfree(range);
683be16e 2587 goto out;
1e701a32
CM		 2588 }
2589 /* compression requires us to start the IO */
2590 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2591 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2592 range->extent_thresh = (u32)-1;
2593 }
2594 } else {
2595 /* the rest are all set to zero by kzalloc */
2596 range->len = (u64)-1;
2597 }
496ad9aa 2598 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2599 range, 0, 0);
2600 if (ret > 0)
2601 ret = 0;
1e701a32 2602 kfree(range);
f46b5a66 2603 break;
8929ecfa
YZ		 2604 default:
2605 ret = -EINVAL;
f46b5a66 2606 }
e441d54d 2607out:
25122d15 2608 mnt_drop_write_file(file);
e441d54d 2609 return ret;
f46b5a66
CH		 2610}
2611
b2950863 2612static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2613{
2614 struct btrfs_ioctl_vol_args *vol_args;
2615 int ret;
2616
e441d54d
CM		 2617 if (!capable(CAP_SYS_ADMIN))
2618 return -EPERM;
2619
5ac00add
SB		 2620 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2621 1)) {
e57138b3 2622 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 2623 }
2624
5ac00add 2625 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2626 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2627 if (IS_ERR(vol_args)) {
2628 ret = PTR_ERR(vol_args);
2629 goto out;
2630 }
f46b5a66 2631
5516e595 2632 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2633 ret = btrfs_init_new_device(root, vol_args->name);
2634
43d20761
AJ		 2635 if (!ret)
2636 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2637
f46b5a66 2638 kfree(vol_args);
c9e9f97b
ID		 2639out:
2640 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2641 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2642 return ret;
2643}
2644
da24927b 2645static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2646{
496ad9aa 2647 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2648 struct btrfs_ioctl_vol_args *vol_args;
2649 int ret;
2650
e441d54d
CM		 2651 if (!capable(CAP_SYS_ADMIN))
2652 return -EPERM;
2653
da24927b
MX		 2654 ret = mnt_want_write_file(file);
2655 if (ret)
2656 return ret;
c146afad 2657
dae7b665 2658 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2659 if (IS_ERR(vol_args)) {
2660 ret = PTR_ERR(vol_args);
c47ca32d 2661 goto err_drop;
c9e9f97b 2662 }
f46b5a66 2663
5516e595 2664 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2665
183860f6
AJ		 2666 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2667 1)) {
2668 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2669 goto out;
2670 }
2671
2672 mutex_lock(&root->fs_info->volume_mutex);
2673 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2674 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2675 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6 2676
ec95d491
AJ		 2677 if (!ret)
2678 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2679
183860f6
AJ		 2680out:
2681 kfree(vol_args);
c47ca32d 2682err_drop:
4ac20c70 2683 mnt_drop_write_file(file);
f46b5a66
CH		 2684 return ret;
2685}
2686
475f6387
JS		 2687static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2688{
027ed2f0 2689 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2690 struct btrfs_device *device;
2691 struct btrfs_device *next;
2692 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2693 int ret = 0;
475f6387 2694
027ed2f0
LZ		 2695 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2696 if (!fi_args)
2697 return -ENOMEM;
2698
f7171750 2699 mutex_lock(&fs_devices->device_list_mutex);
027ed2f0
LZ		 2700 fi_args->num_devices = fs_devices->num_devices;
2701 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387 2702
475f6387 2703 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2704 if (device->devid > fi_args->max_id)
2705 fi_args->max_id = device->devid;
475f6387
JS		 2706 }
2707 mutex_unlock(&fs_devices->device_list_mutex);
2708
80a773fb
DS		 2709 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2710 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2711 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2712
027ed2f0
LZ		 2713 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2714 ret = -EFAULT;
475f6387 2715
027ed2f0
LZ		 2716 kfree(fi_args);
2717 return ret;
475f6387
JS		 2718}
2719
2720static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2721{
2722 struct btrfs_ioctl_dev_info_args *di_args;
2723 struct btrfs_device *dev;
2724 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2725 int ret = 0;
2726 char *s_uuid = NULL;
475f6387 2727
475f6387
JS		 2728 di_args = memdup_user(arg, sizeof(*di_args));
2729 if (IS_ERR(di_args))
2730 return PTR_ERR(di_args);
2731
dd5f9615 2732 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2733 s_uuid = di_args->uuid;
2734
2735 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2736 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2737
2738 if (!dev) {
2739 ret = -ENODEV;
2740 goto out;
2741 }
2742
2743 di_args->devid = dev->devid;
7cc8e58d
MX		 2744 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2745 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
475f6387 2746 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2747 if (dev->name) {
606686ee
JB		 2748 struct rcu_string *name;
2749
2750 rcu_read_lock();
2751 name = rcu_dereference(dev->name);
2752 strncpy(di_args->path, name->str, sizeof(di_args->path));
2753 rcu_read_unlock();
a27202fb
JM		 2754 di_args->path[sizeof(di_args->path) - 1] = 0;
2755 } else {
99ba55ad 2756 di_args->path[0] = '\0';
a27202fb 2757 }
475f6387
JS		 2758
2759out:
55793c0d 2760 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2761 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2762 ret = -EFAULT;
2763
2764 kfree(di_args);
2765 return ret;
2766}
2767
f4414602 2768static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
416161db
MF		 2769{
2770 struct page *page;
416161db
MF		 2771 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2772
416161db
MF		 2773 page = grab_cache_page(inode->i_mapping, index);
2774 if (!page)
2775 return NULL;
2776
2777 if (!PageUptodate(page)) {
2778 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2779 0))
2780 return NULL;
2781 lock_page(page);
2782 if (!PageUptodate(page)) {
2783 unlock_page(page);
2784 page_cache_release(page);
2785 return NULL;
2786 }
2787 }
2788 unlock_page(page);
2789
2790 return page;
2791}
2792
f4414602
MF		 2793static int gather_extent_pages(struct inode *inode, struct page **pages,
2794 int num_pages, u64 off)
2795{
2796 int i;
2797 pgoff_t index = off >> PAGE_CACHE_SHIFT;
2798
2799 for (i = 0; i < num_pages; i++) {
2800 pages[i] = extent_same_get_page(inode, index + i);
2801 if (!pages[i])
2802 return -ENOMEM;
2803 }
2804 return 0;
2805}
2806
77fe20dc
MF		 2807static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2808{
2809 /* do any pending delalloc/csum calc on src, one way or
2810 another, and lock file content */
2811 while (1) {
2812 struct btrfs_ordered_extent *ordered;
2813 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2814 ordered = btrfs_lookup_first_ordered_extent(inode,
2815 off + len - 1);
ff5df9b8
FM		 2816 if ((!ordered ||
2817 ordered->file_offset + ordered->len <= off ||
2818 ordered->file_offset >= off + len) &&
77fe20dc 2819 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
ff5df9b8
FM		 2820 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2821 if (ordered)
2822 btrfs_put_ordered_extent(ordered);
77fe20dc 2823 break;
ff5df9b8 2824 }
77fe20dc
MF		 2825 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2826 if (ordered)
2827 btrfs_put_ordered_extent(ordered);
2828 btrfs_wait_ordered_range(inode, off, len);
2829 }
2830}
2831
f4414602 2832static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
416161db 2833{
416161db
MF		 2834 mutex_unlock(&inode1->i_mutex);
2835 mutex_unlock(&inode2->i_mutex);
2836}
2837
f4414602
MF		 2838static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2839{
2840 if (inode1 < inode2)
2841 swap(inode1, inode2);
2842
2843 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2844 if (inode1 != inode2)
2845 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2846}
2847
2848static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2849 struct inode *inode2, u64 loff2, u64 len)
2850{
2851 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2852 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2853}
2854
2855static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2856 struct inode *inode2, u64 loff2, u64 len)
416161db
MF		 2857{
2858 if (inode1 < inode2) {
2859 swap(inode1, inode2);
2860 swap(loff1, loff2);
2861 }
416161db 2862 lock_extent_range(inode1, loff1, len);
f4414602 2863 if (inode1 != inode2)
416161db 2864 lock_extent_range(inode2, loff2, len);
f4414602
MF		 2865}
2866
2867struct cmp_pages {
2868 int num_pages;
2869 struct page **src_pages;
2870 struct page **dst_pages;
2871};
2872
2873static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2874{
2875 int i;
2876 struct page *pg;
2877
2878 for (i = 0; i < cmp->num_pages; i++) {
2879 pg = cmp->src_pages[i];
2880 if (pg)
2881 page_cache_release(pg);
2882 pg = cmp->dst_pages[i];
2883 if (pg)
2884 page_cache_release(pg);
2885 }
2886 kfree(cmp->src_pages);
2887 kfree(cmp->dst_pages);
2888}
2889
2890static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2891 struct inode *dst, u64 dst_loff,
2892 u64 len, struct cmp_pages *cmp)
2893{
2894 int ret;
2895 int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
2896 struct page **src_pgarr, **dst_pgarr;
2897
2898 /*
2899 * We must gather up all the pages before we initiate our
2900 * extent locking. We use an array for the page pointers. Size
2901 * of the array is bounded by len, which is in turn bounded by
2902 * BTRFS_MAX_DEDUPE_LEN.
2903 */
2904 src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2905 dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2906 if (!src_pgarr || !dst_pgarr) {
2907 kfree(src_pgarr);
2908 kfree(dst_pgarr);
2909 return -ENOMEM;
416161db 2910 }
f4414602
MF		 2911 cmp->num_pages = num_pages;
2912 cmp->src_pages = src_pgarr;
2913 cmp->dst_pages = dst_pgarr;
2914
2915 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2916 if (ret)
2917 goto out;
2918
2919 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2920
2921out:
2922 if (ret)
2923 btrfs_cmp_data_free(cmp);
2924 return 0;
416161db
MF		 2925}
2926
2927static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
f4414602 2928 u64 dst_loff, u64 len, struct cmp_pages *cmp)
416161db
MF		 2929{
2930 int ret = 0;
f4414602 2931 int i;
416161db
MF		 2932 struct page *src_page, *dst_page;
2933 unsigned int cmp_len = PAGE_CACHE_SIZE;
2934 void *addr, *dst_addr;
2935
f4414602 2936 i = 0;
416161db
MF		 2937 while (len) {
2938 if (len < PAGE_CACHE_SIZE)
2939 cmp_len = len;
2940
f4414602
MF		 2941 BUG_ON(i >= cmp->num_pages);
2942
2943 src_page = cmp->src_pages[i];
2944 dst_page = cmp->dst_pages[i];
2945
416161db
MF		 2946 addr = kmap_atomic(src_page);
2947 dst_addr = kmap_atomic(dst_page);
2948
2949 flush_dcache_page(src_page);
2950 flush_dcache_page(dst_page);
2951
2952 if (memcmp(addr, dst_addr, cmp_len))
2953 ret = BTRFS_SAME_DATA_DIFFERS;
2954
2955 kunmap_atomic(addr);
2956 kunmap_atomic(dst_addr);
416161db
MF		 2957
2958 if (ret)
2959 break;
2960
416161db 2961 len -= cmp_len;
f4414602 2962 i++;
416161db
MF		 2963 }
2964
2965 return ret;
2966}
2967
e1d227a4
MF		 2968static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
2969 u64 olen)
416161db 2970{
e1d227a4 2971 u64 len = *plen;
416161db
MF		 2972 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2973
e1d227a4 2974 if (off + olen > inode->i_size || off + olen < off)
416161db 2975 return -EINVAL;
e1d227a4
MF		 2976
2977 /* if we extend to eof, continue to block boundary */
2978 if (off + len == inode->i_size)
2979 *plen = len = ALIGN(inode->i_size, bs) - off;
2980
416161db
MF		 2981 /* Check that we are block aligned - btrfs_clone() requires this */
2982 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2983 return -EINVAL;
2984
2985 return 0;
2986}
2987
e1d227a4 2988static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
416161db
MF		 2989 struct inode *dst, u64 dst_loff)
2990{
2991 int ret;
e1d227a4 2992 u64 len = olen;
f4414602 2993 struct cmp_pages cmp;
416161db
MF		 2994
2995 /*
2996 * btrfs_clone() can't handle extents in the same file
2997 * yet. Once that works, we can drop this check and replace it
2998 * with a check for the same inode, but overlapping extents.
2999 */
3000 if (src == dst)
3001 return -EINVAL;
3002
113e8283
FM		 3003 if (len == 0)
3004 return 0;
3005
f4414602 3006 btrfs_double_inode_lock(src, dst);
416161db 3007
e1d227a4 3008 ret = extent_same_check_offsets(src, loff, &len, olen);
416161db
MF		 3009 if (ret)
3010 goto out_unlock;
3011
e1d227a4 3012 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
416161db
MF		 3013 if (ret)
3014 goto out_unlock;
3015
3016 /* don't make the dst file partly checksummed */
3017 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3018 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3019 ret = -EINVAL;
3020 goto out_unlock;
3021 }
3022
f4414602
MF		 3023 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3024 if (ret)
3025 goto out_unlock;
3026
3027 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
3028
207910dd 3029 /* pass original length for comparison so we stay within i_size */
f4414602 3030 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
416161db 3031 if (ret == 0)
e1d227a4 3032 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff);
416161db 3033
f4414602
MF		 3034 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3035
3036 btrfs_cmp_data_free(&cmp);
416161db 3037out_unlock:
f4414602 3038 btrfs_double_inode_unlock(src, dst);
416161db
MF		 3039
3040 return ret;
3041}
3042
3043#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
3044
3045static long btrfs_ioctl_file_extent_same(struct file *file,
1c1c8747 3046 struct btrfs_ioctl_same_args __user *argp)
416161db 3047{
cbf8b8ca
MF		 3048 struct btrfs_ioctl_same_args *same;
3049 struct btrfs_ioctl_same_extent_info *info;
1c1c8747 3050 struct inode *src = file_inode(file);
416161db
MF		 3051 u64 off;
3052 u64 len;
3053 int i;
3054 int ret;
cbf8b8ca 3055 unsigned long size;
416161db
MF		 3056 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3057 bool is_admin = capable(CAP_SYS_ADMIN);
1c1c8747 3058 u16 count;
416161db
MF		 3059
3060 if (!(file->f_mode & FMODE_READ))
3061 return -EINVAL;
3062
3063 ret = mnt_want_write_file(file);
3064 if (ret)
3065 return ret;
3066
1c1c8747 3067 if (get_user(count, &argp->dest_count)) {
416161db
MF		 3068 ret = -EFAULT;
3069 goto out;
3070 }
3071
1c1c8747 3072 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
cbf8b8ca 3073
1c1c8747 3074 same = memdup_user(argp, size);
cbf8b8ca 3075
229eed43
GB		 3076 if (IS_ERR(same)) {
3077 ret = PTR_ERR(same);
cbf8b8ca
MF		 3078 goto out;
3079 }
3080
3081 off = same->logical_offset;
3082 len = same->length;
416161db
MF		 3083
3084 /*
3085 * Limit the total length we will dedupe for each operation.
3086 * This is intended to bound the total time spent in this
3087 * ioctl to something sane.
3088 */
3089 if (len > BTRFS_MAX_DEDUPE_LEN)
3090 len = BTRFS_MAX_DEDUPE_LEN;
3091
3092 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3093 /*
3094 * Btrfs does not support blocksize < page_size. As a
3095 * result, btrfs_cmp_data() won't correctly handle
3096 * this situation without an update.
3097 */
3098 ret = -EINVAL;
3099 goto out;
3100 }
3101
3102 ret = -EISDIR;
3103 if (S_ISDIR(src->i_mode))
3104 goto out;
3105
3106 ret = -EACCES;
3107 if (!S_ISREG(src->i_mode))
3108 goto out;
3109
cbf8b8ca 3110 /* pre-format output fields to sane values */
1c1c8747 3111 for (i = 0; i < count; i++) {
cbf8b8ca
MF		 3112 same->info[i].bytes_deduped = 0ULL;
3113 same->info[i].status = 0;
3114 }
416161db 3115
1c1c8747
AV		 3116 for (i = 0, info = same->info; i < count; i++, info++) {
3117 struct inode *dst;
3118 struct fd dst_file = fdget(info->fd);
3119 if (!dst_file.file) {
cbf8b8ca 3120 info->status = -EBADF;
1c1c8747 3121 continue;
416161db 3122 }
1c1c8747 3123 dst = file_inode(dst_file.file);
416161db 3124
1c1c8747 3125 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
cbf8b8ca 3126 info->status = -EINVAL;
1c1c8747
AV		 3127 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3128 info->status = -EXDEV;
3129 } else if (S_ISDIR(dst->i_mode)) {
cbf8b8ca 3130 info->status = -EISDIR;
1c1c8747 3131 } else if (!S_ISREG(dst->i_mode)) {
cbf8b8ca 3132 info->status = -EACCES;
1c1c8747
AV		 3133 } else {
3134 info->status = btrfs_extent_same(src, off, len, dst,
3135 info->logical_offset);
3136 if (info->status == 0)
3137 info->bytes_deduped += len;
416161db 3138 }
1c1c8747 3139 fdput(dst_file);
416161db
MF		 3140 }
3141
cbf8b8ca
MF		 3142 ret = copy_to_user(argp, same, size);
3143 if (ret)
3144 ret = -EFAULT;
3145
416161db
MF		 3146out:
3147 mnt_drop_write_file(file);
3148 return ret;
3149}
3150
fcebe456
JB		 3151/* Helper to check and see if this root currently has a ref on the given disk
3152 * bytenr. If it does then we need to update the quota for this root. This
3153 * doesn't do anything if quotas aren't enabled.
3154 */
3155static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3156 u64 disko)
3157{
3284da7b 3158 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
fcebe456
JB		 3159 struct ulist *roots;
3160 struct ulist_iterator uiter;
3161 struct ulist_node *root_node = NULL;
3162 int ret;
3163
3164 if (!root->fs_info->quota_enabled)
3165 return 1;
3166
3167 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3168 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3169 tree_mod_seq_elem.seq, &roots);
3170 if (ret < 0)
3171 goto out;
3172 ret = 0;
3173 ULIST_ITER_INIT(&uiter);
3174 while ((root_node = ulist_next(roots, &uiter))) {
3175 if (root_node->val == root->objectid) {
3176 ret = 1;
3177 break;
3178 }
3179 }
3180 ulist_free(roots);
3181out:
3182 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3183 return ret;
3184}
3185
f82a9901
FM		 3186static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3187 struct inode *inode,
3188 u64 endoff,
3189 const u64 destoff,
3190 const u64 olen)
3191{
3192 struct btrfs_root *root = BTRFS_I(inode)->root;
3193 int ret;
3194
3195 inode_inc_iversion(inode);
3196 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3197 /*
3198 * We round up to the block size at eof when determining which
3199 * extents to clone above, but shouldn't round up the file size.
3200 */
3201 if (endoff > destoff + olen)
3202 endoff = destoff + olen;
3203 if (endoff > inode->i_size)
3204 btrfs_i_size_write(inode, endoff);
3205
3206 ret = btrfs_update_inode(trans, root, inode);
3207 if (ret) {
3208 btrfs_abort_transaction(trans, root, ret);
3209 btrfs_end_transaction(trans, root);
3210 goto out;
3211 }
3212 ret = btrfs_end_transaction(trans, root);
3213out:
3214 return ret;
3215}
3216
7ffbb598
FM		 3217static void clone_update_extent_map(struct inode *inode,
3218 const struct btrfs_trans_handle *trans,
3219 const struct btrfs_path *path,
7ffbb598
FM		 3220 const u64 hole_offset,
3221 const u64 hole_len)
3222{
3223 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3224 struct extent_map *em;
3225 int ret;
3226
3227 em = alloc_extent_map();
3228 if (!em) {
3229 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3230 &BTRFS_I(inode)->runtime_flags);
3231 return;
3232 }
3233
14f59796
FM		 3234 if (path) {
3235 struct btrfs_file_extent_item *fi;
3236
3237 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3238 struct btrfs_file_extent_item);
7ffbb598
FM		 3239 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3240 em->generation = -1;
3241 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3242 BTRFS_FILE_EXTENT_INLINE)
3243 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3244 &BTRFS_I(inode)->runtime_flags);
3245 } else {
3246 em->start = hole_offset;
3247 em->len = hole_len;
3248 em->ram_bytes = em->len;
3249 em->orig_start = hole_offset;
3250 em->block_start = EXTENT_MAP_HOLE;
3251 em->block_len = 0;
3252 em->orig_block_len = 0;
3253 em->compress_type = BTRFS_COMPRESS_NONE;
3254 em->generation = trans->transid;
3255 }
3256
3257 while (1) {
3258 write_lock(&em_tree->lock);
3259 ret = add_extent_mapping(em_tree, em, 1);
3260 write_unlock(&em_tree->lock);
3261 if (ret != -EEXIST) {
3262 free_extent_map(em);
3263 break;
3264 }
3265 btrfs_drop_extent_cache(inode, em->start,
3266 em->start + em->len - 1, 0);
3267 }
3268
ee39b432 3269 if (ret)
7ffbb598
FM		 3270 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3271 &BTRFS_I(inode)->runtime_flags);
3272}
3273
32b7c687
MF		 3274/**
3275 * btrfs_clone() - clone a range from inode file to another
3276 *
3277 * @src: Inode to clone from
3278 * @inode: Inode to clone to
3279 * @off: Offset within source to start clone from
3280 * @olen: Original length, passed by user, of range to clone
3281 * @olen_aligned: Block-aligned value of olen, extent_same uses
3282 * identical values here
3283 * @destoff: Offset within @inode to start clone
3284 */
3285static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901
FM		 3286 const u64 off, const u64 olen, const u64 olen_aligned,
3287 const u64 destoff)
f46b5a66 3288{
f46b5a66 3289 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3290 struct btrfs_path *path = NULL;
f46b5a66 3291 struct extent_buffer *leaf;
32b7c687
MF		 3292 struct btrfs_trans_handle *trans;
3293 char *buf = NULL;
ae01a0ab 3294 struct btrfs_key key;
f46b5a66
CH		 3295 u32 nritems;
3296 int slot;
ae01a0ab 3297 int ret;
fcebe456 3298 int no_quota;
f82a9901 3299 const u64 len = olen_aligned;
fcebe456 3300 u64 last_disko = 0;
f82a9901 3301 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3302
3303 ret = -ENOMEM;
707e8a07 3304 buf = vmalloc(root->nodesize);
ae01a0ab 3305 if (!buf)
32b7c687 3306 return ret;
ae01a0ab
YZ		 3307
3308 path = btrfs_alloc_path();
3309 if (!path) {
3310 vfree(buf);
32b7c687 3311 return ret;
d525e8ab
LZ		 3312 }
3313
32b7c687 3314 path->reada = 2;
c5c9cd4d 3315 /* clone data */
33345d01 3316 key.objectid = btrfs_ino(src);
ae01a0ab 3317 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3318 key.offset = off;
f46b5a66
CH		 3319
3320 while (1) {
de249e66 3321 u64 next_key_min_offset = key.offset + 1;
df858e76 3322
f46b5a66
CH		 3323 /*
3324 * note the key will change type as we walk through the
3325 * tree.
3326 */
e4355f34 3327 path->leave_spinning = 1;
362a20c5
DS		 3328 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3329 0, 0);
f46b5a66
CH		 3330 if (ret < 0)
3331 goto out;
2c463823
FM		 3332 /*
3333 * First search, if no extent item that starts at offset off was
3334 * found but the previous item is an extent item, it's possible
3335 * it might overlap our target range, therefore process it.
3336 */
3337 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3338 btrfs_item_key_to_cpu(path->nodes[0], &key,
3339 path->slots[0] - 1);
3340 if (key.type == BTRFS_EXTENT_DATA_KEY)
3341 path->slots[0]--;
3342 }
f46b5a66 3343
ae01a0ab 3344 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3345process_slot:
fcebe456 3346 no_quota = 1;
ae01a0ab 3347 if (path->slots[0] >= nritems) {
362a20c5 3348 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3349 if (ret < 0)
3350 goto out;
3351 if (ret > 0)
3352 break;
ae01a0ab 3353 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3354 }
3355 leaf = path->nodes[0];
3356 slot = path->slots[0];
f46b5a66 3357
ae01a0ab 3358 btrfs_item_key_to_cpu(leaf, &key, slot);
962a298f 3359 if (key.type > BTRFS_EXTENT_DATA_KEY ||
33345d01 3360 key.objectid != btrfs_ino(src))
f46b5a66
CH		 3361 break;
3362
962a298f 3363 if (key.type == BTRFS_EXTENT_DATA_KEY) {
c5c9cd4d
SW		 3364 struct btrfs_file_extent_item *extent;
3365 int type;
31840ae1
ZY		 3366 u32 size;
3367 struct btrfs_key new_key;
c5c9cd4d
SW		 3368 u64 disko = 0, diskl = 0;
3369 u64 datao = 0, datal = 0;
3370 u8 comp;
f82a9901 3371 u64 drop_start;
31840ae1 3372
c5c9cd4d
SW		 3373 extent = btrfs_item_ptr(leaf, slot,
3374 struct btrfs_file_extent_item);
3375 comp = btrfs_file_extent_compression(leaf, extent);
3376 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3377 if (type == BTRFS_FILE_EXTENT_REG ||
3378 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3379 disko = btrfs_file_extent_disk_bytenr(leaf,
3380 extent);
3381 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3382 extent);
c5c9cd4d 3383 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3384 datal = btrfs_file_extent_num_bytes(leaf,
3385 extent);
c5c9cd4d
SW		 3386 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3387 /* take upper bound, may be compressed */
3388 datal = btrfs_file_extent_ram_bytes(leaf,
3389 extent);
3390 }
31840ae1 3391
2c463823
FM		 3392 /*
3393 * The first search might have left us at an extent
3394 * item that ends before our target range's start, can
3395 * happen if we have holes and NO_HOLES feature enabled.
3396 */
3397 if (key.offset + datal <= off) {
e4355f34
FDBM		 3398 path->slots[0]++;
3399 goto process_slot;
2c463823
FM		 3400 } else if (key.offset >= off + len) {
3401 break;
e4355f34 3402 }
df858e76 3403 next_key_min_offset = key.offset + datal;
e4355f34
FDBM		 3404 size = btrfs_item_size_nr(leaf, slot);
3405 read_extent_buffer(leaf, buf,
3406 btrfs_item_ptr_offset(leaf, slot),
3407 size);
3408
3409 btrfs_release_path(path);
3410 path->leave_spinning = 0;
c5c9cd4d 3411
31840ae1 3412 memcpy(&new_key, &key, sizeof(new_key));
33345d01 3413 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 3414 if (off <= key.offset)
3415 new_key.offset = key.offset + destoff - off;
3416 else
3417 new_key.offset = destoff;
31840ae1 3418
f82a9901
FM		 3419 /*
3420 * Deal with a hole that doesn't have an extent item
3421 * that represents it (NO_HOLES feature enabled).
3422 * This hole is either in the middle of the cloning
3423 * range or at the beginning (fully overlaps it or
3424 * partially overlaps it).
3425 */
3426 if (new_key.offset != last_dest_end)
3427 drop_start = last_dest_end;
3428 else
3429 drop_start = new_key.offset;
3430
b6f3409b
SW		 3431 /*
3432 * 1 - adjusting old extent (we may have to split it)
3433 * 1 - add new extent
3434 * 1 - inode update
3435 */
3436 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3437 if (IS_ERR(trans)) {
3438 ret = PTR_ERR(trans);
3439 goto out;
3440 }
3441
c8a894d7
CM		 3442 if (type == BTRFS_FILE_EXTENT_REG ||
3443 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3444 /*
3445 * a | --- range to clone ---| b
3446 * | ------------- extent ------------- |
3447 */
3448
93915584 3449 /* subtract range b */
d72c0842
LZ		 3450 if (key.offset + datal > off + len)
3451 datal = off + len - key.offset;
3452
93915584 3453 /* subtract range a */
a22285a6
YZ		 3454 if (off > key.offset) {
3455 datao += off - key.offset;
3456 datal -= off - key.offset;
3457 }
3458
5dc562c5 3459 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3460 drop_start,
a22285a6 3461 new_key.offset + datal,
2671485d 3462 1);
79787eaa 3463 if (ret) {
3f9e3df8 3464 if (ret != -EOPNOTSUPP)
00fdf13a
LB		 3465 btrfs_abort_transaction(trans,
3466 root, ret);
79787eaa
JM		 3467 btrfs_end_transaction(trans, root);
3468 goto out;
3469 }
a22285a6 3470
c5c9cd4d
SW		 3471 ret = btrfs_insert_empty_item(trans, root, path,
3472 &new_key, size);
79787eaa
JM		 3473 if (ret) {
3474 btrfs_abort_transaction(trans, root,
3475 ret);
3476 btrfs_end_transaction(trans, root);
3477 goto out;
3478 }
c5c9cd4d
SW		 3479
3480 leaf = path->nodes[0];
3481 slot = path->slots[0];
3482 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3483 btrfs_item_ptr_offset(leaf, slot),
3484 size);
ae01a0ab 3485
c5c9cd4d 3486 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3487 struct btrfs_file_extent_item);
c5c9cd4d 3488
c5c9cd4d
SW		 3489 /* disko == 0 means it's a hole */
3490 if (!disko)
3491 datao = 0;
c5c9cd4d
SW		 3492
3493 btrfs_set_file_extent_offset(leaf, extent,
3494 datao);
3495 btrfs_set_file_extent_num_bytes(leaf, extent,
3496 datal);
fcebe456
JB		 3497
3498 /*
3499 * We need to look up the roots that point at
3500 * this bytenr and see if the new root does. If
3501 * it does not we need to make sure we update
3502 * quotas appropriately.
3503 */
3504 if (disko && root != BTRFS_I(src)->root &&
3505 disko != last_disko) {
3506 no_quota = check_ref(trans, root,
3507 disko);
3508 if (no_quota < 0) {
3509 btrfs_abort_transaction(trans,
3510 root,
3511 ret);
3512 btrfs_end_transaction(trans,
3513 root);
3514 ret = no_quota;
3515 goto out;
3516 }
3517 }
3518
c5c9cd4d
SW		 3519 if (disko) {
3520 inode_add_bytes(inode, datal);
ae01a0ab 3521 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3522 disko, diskl, 0,
3523 root->root_key.objectid,
33345d01 3524 btrfs_ino(inode),
66d7e7f0 3525 new_key.offset - datao,
fcebe456 3526 no_quota);
79787eaa
JM		 3527 if (ret) {
3528 btrfs_abort_transaction(trans,
3529 root,
3530 ret);
3531 btrfs_end_transaction(trans,
3532 root);
3533 goto out;
3534
3535 }
f46b5a66 3536 }
c5c9cd4d
SW		 3537 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3538 u64 skip = 0;
3539 u64 trim = 0;
d3ecfcdf
LB		 3540 u64 aligned_end = 0;
3541
c5c9cd4d
SW		 3542 if (off > key.offset) {
3543 skip = off - key.offset;
3544 new_key.offset += skip;
3545 }
d397712b 3546
aa42ffd9
LB		 3547 if (key.offset + datal > off + len)
3548 trim = key.offset + datal - (off + len);
d397712b 3549
c5c9cd4d 3550 if (comp && (skip || trim)) {
c5c9cd4d 3551 ret = -EINVAL;
a22285a6 3552 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3553 goto out;
3554 }
3555 size -= skip + trim;
3556 datal -= skip + trim;
a22285a6 3557
d3ecfcdf
LB		 3558 aligned_end = ALIGN(new_key.offset + datal,
3559 root->sectorsize);
5dc562c5 3560 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3561 drop_start,
d3ecfcdf 3562 aligned_end,
2671485d 3563 1);
79787eaa 3564 if (ret) {
3f9e3df8 3565 if (ret != -EOPNOTSUPP)
3a29bc09
CM		 3566 btrfs_abort_transaction(trans,
3567 root, ret);
79787eaa
JM		 3568 btrfs_end_transaction(trans, root);
3569 goto out;
3570 }
a22285a6 3571
c5c9cd4d
SW		 3572 ret = btrfs_insert_empty_item(trans, root, path,
3573 &new_key, size);
79787eaa
JM		 3574 if (ret) {
3575 btrfs_abort_transaction(trans, root,
3576 ret);
3577 btrfs_end_transaction(trans, root);
3578 goto out;
3579 }
c5c9cd4d
SW		 3580
3581 if (skip) {
d397712b
CM		 3582 u32 start =
3583 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3584 memmove(buf+start, buf+start+skip,
3585 datal);
3586 }
3587
3588 leaf = path->nodes[0];
3589 slot = path->slots[0];
3590 write_extent_buffer(leaf, buf,
3591 btrfs_item_ptr_offset(leaf, slot),
3592 size);
3593 inode_add_bytes(inode, datal);
f46b5a66 3594 }
c5c9cd4d 3595
7ffbb598
FM		 3596 /* If we have an implicit hole (NO_HOLES feature). */
3597 if (drop_start < new_key.offset)
3598 clone_update_extent_map(inode, trans,
14f59796 3599 NULL, drop_start,
7ffbb598
FM		 3600 new_key.offset - drop_start);
3601
14f59796 3602 clone_update_extent_map(inode, trans, path, 0, 0);
7ffbb598 3603
c5c9cd4d 3604 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3605 btrfs_release_path(path);
c5c9cd4d 3606
62e2390e
FM		 3607 last_dest_end = ALIGN(new_key.offset + datal,
3608 root->sectorsize);
f82a9901
FM		 3609 ret = clone_finish_inode_update(trans, inode,
3610 last_dest_end,
3611 destoff, olen);
3612 if (ret)
79787eaa 3613 goto out;
2c463823
FM		 3614 if (new_key.offset + datal >= destoff + len)
3615 break;
a22285a6 3616 }
b3b4aa74 3617 btrfs_release_path(path);
df858e76 3618 key.offset = next_key_min_offset;
f46b5a66 3619 }
f46b5a66 3620 ret = 0;
32b7c687 3621
f82a9901
FM		 3622 if (last_dest_end < destoff + len) {
3623 /*
3624 * We have an implicit hole (NO_HOLES feature is enabled) that
3625 * fully or partially overlaps our cloning range at its end.
3626 */
3627 btrfs_release_path(path);
3628
3629 /*
3630 * 1 - remove extent(s)
3631 * 1 - inode update
3632 */
3633 trans = btrfs_start_transaction(root, 2);
3634 if (IS_ERR(trans)) {
3635 ret = PTR_ERR(trans);
3636 goto out;
3637 }
3638 ret = btrfs_drop_extents(trans, root, inode,
3639 last_dest_end, destoff + len, 1);
3640 if (ret) {
3641 if (ret != -EOPNOTSUPP)
3642 btrfs_abort_transaction(trans, root, ret);
3643 btrfs_end_transaction(trans, root);
3644 goto out;
3645 }
14f59796
FM		 3646 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3647 destoff + len - last_dest_end);
f82a9901
FM		 3648 ret = clone_finish_inode_update(trans, inode, destoff + len,
3649 destoff, olen);
3650 }
3651
f46b5a66 3652out:
32b7c687
MF		 3653 btrfs_free_path(path);
3654 vfree(buf);
3655 return ret;
3656}
3657
3658static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3659 u64 off, u64 olen, u64 destoff)
3660{
54563d41 3661 struct inode *inode = file_inode(file);
32b7c687
MF		 3662 struct btrfs_root *root = BTRFS_I(inode)->root;
3663 struct fd src_file;
3664 struct inode *src;
3665 int ret;
3666 u64 len = olen;
3667 u64 bs = root->fs_info->sb->s_blocksize;
3668 int same_inode = 0;
3669
3670 /*
3671 * TODO:
3672 * - split compressed inline extents. annoying: we need to
3673 * decompress into destination's address_space (the file offset
3674 * may change, so source mapping won't do), then recompress (or
3675 * otherwise reinsert) a subrange.
00fdf13a
LB		 3676 *
3677 * - split destination inode's inline extents. The inline extents can
3678 * be either compressed or non-compressed.
32b7c687
MF		 3679 */
3680
3681 /* the destination must be opened for writing */
3682 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3683 return -EINVAL;
3684
3685 if (btrfs_root_readonly(root))
3686 return -EROFS;
3687
3688 ret = mnt_want_write_file(file);
3689 if (ret)
3690 return ret;
3691
3692 src_file = fdget(srcfd);
3693 if (!src_file.file) {
3694 ret = -EBADF;
3695 goto out_drop_write;
3696 }
3697
3698 ret = -EXDEV;
3699 if (src_file.file->f_path.mnt != file->f_path.mnt)
3700 goto out_fput;
3701
3702 src = file_inode(src_file.file);
3703
3704 ret = -EINVAL;
3705 if (src == inode)
3706 same_inode = 1;
3707
3708 /* the src must be open for reading */
3709 if (!(src_file.file->f_mode & FMODE_READ))
3710 goto out_fput;
3711
3712 /* don't make the dst file partly checksummed */
3713 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3714 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3715 goto out_fput;
3716
3717 ret = -EISDIR;
3718 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3719 goto out_fput;
3720
3721 ret = -EXDEV;
3722 if (src->i_sb != inode->i_sb)
3723 goto out_fput;
3724
3725 if (!same_inode) {
3726 if (inode < src) {
3727 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3728 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3729 } else {
3730 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3731 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3732 }
3733 } else {
3734 mutex_lock(&src->i_mutex);
3735 }
3736
3737 /* determine range to clone */
3738 ret = -EINVAL;
3739 if (off + len > src->i_size || off + len < off)
3740 goto out_unlock;
3741 if (len == 0)
3742 olen = len = src->i_size - off;
3743 /* if we extend to eof, continue to block boundary */
3744 if (off + len == src->i_size)
3745 len = ALIGN(src->i_size, bs) - off;
3746
ccccf3d6
FM		 3747 if (len == 0) {
3748 ret = 0;
3749 goto out_unlock;
3750 }
3751
32b7c687
MF		 3752 /* verify the end result is block aligned */
3753 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3754 !IS_ALIGNED(destoff, bs))
3755 goto out_unlock;
3756
3757 /* verify if ranges are overlapped within the same file */
3758 if (same_inode) {
3759 if (destoff + len > off && destoff < off + len)
3760 goto out_unlock;
3761 }
3762
3763 if (destoff > inode->i_size) {
3764 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3765 if (ret)
3766 goto out_unlock;
3767 }
3768
c125b8bf
FM		 3769 /*
3770 * Lock the target range too. Right after we replace the file extent
3771 * items in the fs tree (which now point to the cloned data), we might
3772 * have a worker replace them with extent items relative to a write
3773 * operation that was issued before this clone operation (i.e. confront
3774 * with inode.c:btrfs_finish_ordered_io).
3775 */
3776 if (same_inode) {
3777 u64 lock_start = min_t(u64, off, destoff);
3778 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
32b7c687 3779
c125b8bf
FM		 3780 lock_extent_range(src, lock_start, lock_len);
3781 } else {
3782 lock_extent_range(src, off, len);
3783 lock_extent_range(inode, destoff, len);
3784 }
32b7c687
MF		 3785
3786 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3787
c125b8bf
FM		 3788 if (same_inode) {
3789 u64 lock_start = min_t(u64, off, destoff);
3790 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3791
3792 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3793 } else {
3794 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3795 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3796 destoff + len - 1);
3797 }
3798 /*
3799 * Truncate page cache pages so that future reads will see the cloned
3800 * data immediately and not the previous data.
3801 */
3802 truncate_inode_pages_range(&inode->i_data, destoff,
3803 PAGE_CACHE_ALIGN(destoff + len) - 1);
f46b5a66 3804out_unlock:
c57c2b3e
FDBM		 3805 if (!same_inode) {
3806 if (inode < src) {
3807 mutex_unlock(&src->i_mutex);
3808 mutex_unlock(&inode->i_mutex);
3809 } else {
3810 mutex_unlock(&inode->i_mutex);
3811 mutex_unlock(&src->i_mutex);
3812 }
3813 } else {
3814 mutex_unlock(&src->i_mutex);
3815 }
f46b5a66 3816out_fput:
2903ff01 3817 fdput(src_file);
ab67b7c1 3818out_drop_write:
2a79f17e 3819 mnt_drop_write_file(file);
f46b5a66
CH		 3820 return ret;
3821}
3822
7a865e8a 3823static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3824{
3825 struct btrfs_ioctl_clone_range_args args;
3826
7a865e8a 3827 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3828 return -EFAULT;
3829 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3830 args.src_length, args.dest_offset);
3831}
3832
f46b5a66
CH		 3833/*
3834 * there are many ways the trans_start and trans_end ioctls can lead
3835 * to deadlocks. They should only be used by applications that
3836 * basically own the machine, and have a very in depth understanding
3837 * of all the possible deadlocks and enospc problems.
3838 */
b2950863 3839static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3840{
496ad9aa 3841 struct inode *inode = file_inode(file);
f46b5a66
CH		 3842 struct btrfs_root *root = BTRFS_I(inode)->root;
3843 struct btrfs_trans_handle *trans;
1ab86aed 3844 int ret;
f46b5a66 3845
1ab86aed 3846 ret = -EPERM;
df5b5520 3847 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3848 goto out;
df5b5520 3849
1ab86aed
SW		 3850 ret = -EINPROGRESS;
3851 if (file->private_data)
f46b5a66 3852 goto out;
9ca9ee09 3853
b83cc969
LZ		 3854 ret = -EROFS;
3855 if (btrfs_root_readonly(root))
3856 goto out;
3857
a561be71 3858 ret = mnt_want_write_file(file);
c146afad
YZ		 3859 if (ret)
3860 goto out;
3861
a4abeea4 3862 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3863
1ab86aed 3864 ret = -ENOMEM;
7a7eaa40 3865 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3866 if (IS_ERR(trans))
1ab86aed
SW		 3867 goto out_drop;
3868
3869 file->private_data = trans;
3870 return 0;
3871
3872out_drop:
a4abeea4 3873 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3874 mnt_drop_write_file(file);
f46b5a66 3875out:
f46b5a66
CH		 3876 return ret;
3877}
3878
6ef5ed0d
JB		 3879static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3880{
496ad9aa 3881 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3882 struct btrfs_root *root = BTRFS_I(inode)->root;
3883 struct btrfs_root *new_root;
3884 struct btrfs_dir_item *di;
3885 struct btrfs_trans_handle *trans;
3886 struct btrfs_path *path;
3887 struct btrfs_key location;
3888 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3889 u64 objectid = 0;
3890 u64 dir_id;
3c04ce01 3891 int ret;
6ef5ed0d
JB		 3892
3893 if (!capable(CAP_SYS_ADMIN))
3894 return -EPERM;
3895
3c04ce01
MX		 3896 ret = mnt_want_write_file(file);
3897 if (ret)
3898 return ret;
3899
3900 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3901 ret = -EFAULT;
3902 goto out;
3903 }
6ef5ed0d
JB		 3904
3905 if (!objectid)
1cecf579 3906 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3907
3908 location.objectid = objectid;
3909 location.type = BTRFS_ROOT_ITEM_KEY;
3910 location.offset = (u64)-1;
3911
3912 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3913 if (IS_ERR(new_root)) {
3914 ret = PTR_ERR(new_root);
3915 goto out;
3916 }
6ef5ed0d 3917
6ef5ed0d 3918 path = btrfs_alloc_path();
3c04ce01
MX		 3919 if (!path) {
3920 ret = -ENOMEM;
3921 goto out;
3922 }
6ef5ed0d
JB		 3923 path->leave_spinning = 1;
3924
3925 trans = btrfs_start_transaction(root, 1);
98d5dc13 3926 if (IS_ERR(trans)) {
6ef5ed0d 3927 btrfs_free_path(path);
3c04ce01
MX		 3928 ret = PTR_ERR(trans);
3929 goto out;
6ef5ed0d
JB		 3930 }
3931
6c41761f 3932 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3933 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3934 dir_id, "default", 7, 1);
cf1e99a4 3935 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3936 btrfs_free_path(path);
3937 btrfs_end_transaction(trans, root);
efe120a0
FH		 3938 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3939 "item, this isn't going to work");
3c04ce01
MX		 3940 ret = -ENOENT;
3941 goto out;
6ef5ed0d
JB		 3942 }
3943
3944 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3945 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3946 btrfs_mark_buffer_dirty(path->nodes[0]);
3947 btrfs_free_path(path);
3948
2b0ce2c2 3949 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3950 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3951out:
3952 mnt_drop_write_file(file);
3953 return ret;
6ef5ed0d
JB		 3954}
3955
5af3e8cc
SB		 3956void btrfs_get_block_group_info(struct list_head *groups_list,
3957 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3958{
3959 struct btrfs_block_group_cache *block_group;
3960
3961 space->total_bytes = 0;
3962 space->used_bytes = 0;
3963 space->flags = 0;
3964 list_for_each_entry(block_group, groups_list, list) {
3965 space->flags = block_group->flags;
3966 space->total_bytes += block_group->key.offset;
3967 space->used_bytes +=
3968 btrfs_block_group_used(&block_group->item);
3969 }
3970}
3971
48a3b636 3972static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3973{
3974 struct btrfs_ioctl_space_args space_args;
3975 struct btrfs_ioctl_space_info space;
3976 struct btrfs_ioctl_space_info *dest;
7fde62bf 3977 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3978 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3979 struct btrfs_space_info *info;
bf5fc093
JB		 3980 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3981 BTRFS_BLOCK_GROUP_SYSTEM,
3982 BTRFS_BLOCK_GROUP_METADATA,
3983 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3984 int num_types = 4;
7fde62bf 3985 int alloc_size;
1406e432 3986 int ret = 0;
51788b1b 3987 u64 slot_count = 0;
bf5fc093 3988 int i, c;
1406e432
JB		 3989
3990 if (copy_from_user(&space_args,
3991 (struct btrfs_ioctl_space_args __user *)arg,
3992 sizeof(space_args)))
3993 return -EFAULT;
3994
bf5fc093
JB		 3995 for (i = 0; i < num_types; i++) {
3996 struct btrfs_space_info *tmp;
3997
3998 info = NULL;
3999 rcu_read_lock();
4000 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4001 list) {
4002 if (tmp->flags == types[i]) {
4003 info = tmp;
4004 break;
4005 }
4006 }
4007 rcu_read_unlock();
4008
4009 if (!info)
4010 continue;
4011
4012 down_read(&info->groups_sem);
4013 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4014 if (!list_empty(&info->block_groups[c]))
4015 slot_count++;
4016 }
4017 up_read(&info->groups_sem);
4018 }
7fde62bf 4019
36523e95
DS		 4020 /*
4021 * Global block reserve, exported as a space_info
4022 */
4023 slot_count++;
4024
7fde62bf
CM		 4025 /* space_slots == 0 means they are asking for a count */
4026 if (space_args.space_slots == 0) {
4027 space_args.total_spaces = slot_count;
4028 goto out;
4029 }
bf5fc093 4030
51788b1b 4031 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 4032
7fde62bf 4033 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 4034
7fde62bf
CM		 4035 /* we generally have at most 6 or so space infos, one for each raid
4036 * level. So, a whole page should be more than enough for everyone
4037 */
4038 if (alloc_size > PAGE_CACHE_SIZE)
4039 return -ENOMEM;
4040
1406e432 4041 space_args.total_spaces = 0;
7fde62bf
CM		 4042 dest = kmalloc(alloc_size, GFP_NOFS);
4043 if (!dest)
4044 return -ENOMEM;
4045 dest_orig = dest;
1406e432 4046
7fde62bf 4047 /* now we have a buffer to copy into */
bf5fc093
JB		 4048 for (i = 0; i < num_types; i++) {
4049 struct btrfs_space_info *tmp;
4050
51788b1b
DR		 4051 if (!slot_count)
4052 break;
4053
bf5fc093
JB		 4054 info = NULL;
4055 rcu_read_lock();
4056 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4057 list) {
4058 if (tmp->flags == types[i]) {
4059 info = tmp;
4060 break;
4061 }
4062 }
4063 rcu_read_unlock();
7fde62bf 4064
bf5fc093
JB		 4065 if (!info)
4066 continue;
4067 down_read(&info->groups_sem);
4068 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4069 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 4070 btrfs_get_block_group_info(
4071 &info->block_groups[c], &space);
bf5fc093
JB		 4072 memcpy(dest, &space, sizeof(space));
4073 dest++;
4074 space_args.total_spaces++;
51788b1b 4075 slot_count--;
bf5fc093 4076 }
51788b1b
DR		 4077 if (!slot_count)
4078 break;
bf5fc093
JB		 4079 }
4080 up_read(&info->groups_sem);
1406e432 4081 }
1406e432 4082
36523e95
DS		 4083 /*
4084 * Add global block reserve
4085 */
4086 if (slot_count) {
4087 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4088
4089 spin_lock(&block_rsv->lock);
4090 space.total_bytes = block_rsv->size;
4091 space.used_bytes = block_rsv->size - block_rsv->reserved;
4092 spin_unlock(&block_rsv->lock);
4093 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4094 memcpy(dest, &space, sizeof(space));
4095 space_args.total_spaces++;
4096 }
4097
2eec6c81 4098 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 4099 (arg + sizeof(struct btrfs_ioctl_space_args));
4100
4101 if (copy_to_user(user_dest, dest_orig, alloc_size))
4102 ret = -EFAULT;
4103
4104 kfree(dest_orig);
4105out:
4106 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 4107 ret = -EFAULT;
4108
4109 return ret;
4110}
4111
f46b5a66
CH		 4112/*
4113 * there are many ways the trans_start and trans_end ioctls can lead
4114 * to deadlocks. They should only be used by applications that
4115 * basically own the machine, and have a very in depth understanding
4116 * of all the possible deadlocks and enospc problems.
4117 */
4118long btrfs_ioctl_trans_end(struct file *file)
4119{
496ad9aa 4120 struct inode *inode = file_inode(file);
f46b5a66
CH		 4121 struct btrfs_root *root = BTRFS_I(inode)->root;
4122 struct btrfs_trans_handle *trans;
f46b5a66 4123
f46b5a66 4124 trans = file->private_data;
1ab86aed
SW		 4125 if (!trans)
4126 return -EINVAL;
b214107e 4127 file->private_data = NULL;
9ca9ee09 4128
1ab86aed
SW		 4129 btrfs_end_transaction(trans, root);
4130
a4abeea4 4131 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 4132
2a79f17e 4133 mnt_drop_write_file(file);
1ab86aed 4134 return 0;
f46b5a66
CH		 4135}
4136
9a8c28be
MX		 4137static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4138 void __user *argp)
46204592 4139{
46204592
SW		 4140 struct btrfs_trans_handle *trans;
4141 u64 transid;
db5b493a 4142 int ret;
46204592 4143
d4edf39b 4144 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4145 if (IS_ERR(trans)) {
4146 if (PTR_ERR(trans) != -ENOENT)
4147 return PTR_ERR(trans);
4148
4149 /* No running transaction, don't bother */
4150 transid = root->fs_info->last_trans_committed;
4151 goto out;
4152 }
46204592 4153 transid = trans->transid;
db5b493a 4154 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 4155 if (ret) {
4156 btrfs_end_transaction(trans, root);
db5b493a 4157 return ret;
8b2b2d3c 4158 }
ff7c1d33 4159out:
46204592
SW		 4160 if (argp)
4161 if (copy_to_user(argp, &transid, sizeof(transid)))
4162 return -EFAULT;
4163 return 0;
4164}
4165
9a8c28be
MX		 4166static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4167 void __user *argp)
46204592 4168{
46204592
SW		 4169 u64 transid;
4170
4171 if (argp) {
4172 if (copy_from_user(&transid, argp, sizeof(transid)))
4173 return -EFAULT;
4174 } else {
4175 transid = 0; /* current trans */
4176 }
4177 return btrfs_wait_for_commit(root, transid);
4178}
4179
b8e95489 4180static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4181{
496ad9aa 4182 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 4183 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4184 int ret;
475f6387
JS		 4185
4186 if (!capable(CAP_SYS_ADMIN))
4187 return -EPERM;
4188
4189 sa = memdup_user(arg, sizeof(*sa));
4190 if (IS_ERR(sa))
4191 return PTR_ERR(sa);
4192
b8e95489
MX		 4193 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4194 ret = mnt_want_write_file(file);
4195 if (ret)
4196 goto out;
4197 }
4198
aa1b8cd4 4199 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4200 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4201 0);
475f6387
JS		 4202
4203 if (copy_to_user(arg, sa, sizeof(*sa)))
4204 ret = -EFAULT;
4205
b8e95489
MX		 4206 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4207 mnt_drop_write_file(file);
4208out:
475f6387
JS		 4209 kfree(sa);
4210 return ret;
4211}
4212
4213static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4214{
4215 if (!capable(CAP_SYS_ADMIN))
4216 return -EPERM;
4217
aa1b8cd4 4218 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 4219}
4220
4221static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4222 void __user *arg)
4223{
4224 struct btrfs_ioctl_scrub_args *sa;
4225 int ret;
4226
4227 if (!capable(CAP_SYS_ADMIN))
4228 return -EPERM;
4229
4230 sa = memdup_user(arg, sizeof(*sa));
4231 if (IS_ERR(sa))
4232 return PTR_ERR(sa);
4233
4234 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4235
4236 if (copy_to_user(arg, sa, sizeof(*sa)))
4237 ret = -EFAULT;
4238
4239 kfree(sa);
4240 return ret;
4241}
4242
c11d2c23 4243static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 4244 void __user *arg)
c11d2c23
SB		 4245{
4246 struct btrfs_ioctl_get_dev_stats *sa;
4247 int ret;
4248
c11d2c23
SB		 4249 sa = memdup_user(arg, sizeof(*sa));
4250 if (IS_ERR(sa))
4251 return PTR_ERR(sa);
4252
b27f7c0c
DS		 4253 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4254 kfree(sa);
4255 return -EPERM;
4256 }
4257
4258 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 4259
4260 if (copy_to_user(arg, sa, sizeof(*sa)))
4261 ret = -EFAULT;
4262
4263 kfree(sa);
4264 return ret;
4265}
4266
3f6bcfbd
SB		 4267static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4268{
4269 struct btrfs_ioctl_dev_replace_args *p;
4270 int ret;
4271
4272 if (!capable(CAP_SYS_ADMIN))
4273 return -EPERM;
4274
4275 p = memdup_user(arg, sizeof(*p));
4276 if (IS_ERR(p))
4277 return PTR_ERR(p);
4278
4279 switch (p->cmd) {
4280 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
adfa97cb
ID		 4281 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4282 ret = -EROFS;
4283 goto out;
4284 }
3f6bcfbd
SB		 4285 if (atomic_xchg(
4286 &root->fs_info->mutually_exclusive_operation_running,
4287 1)) {
e57138b3 4288 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd
SB		 4289 } else {
4290 ret = btrfs_dev_replace_start(root, p);
4291 atomic_set(
4292 &root->fs_info->mutually_exclusive_operation_running,
4293 0);
4294 }
4295 break;
4296 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4297 btrfs_dev_replace_status(root->fs_info, p);
4298 ret = 0;
4299 break;
4300 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4301 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4302 break;
4303 default:
4304 ret = -EINVAL;
4305 break;
4306 }
4307
4308 if (copy_to_user(arg, p, sizeof(*p)))
4309 ret = -EFAULT;
adfa97cb 4310out:
3f6bcfbd
SB		 4311 kfree(p);
4312 return ret;
4313}
4314
d7728c96
JS		 4315static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4316{
4317 int ret = 0;
4318 int i;
740c3d22 4319 u64 rel_ptr;
d7728c96 4320 int size;
806468f8 4321 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4322 struct inode_fs_paths *ipath = NULL;
4323 struct btrfs_path *path;
4324
82b22ac8 4325 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4326 return -EPERM;
4327
4328 path = btrfs_alloc_path();
4329 if (!path) {
4330 ret = -ENOMEM;
4331 goto out;
4332 }
4333
4334 ipa = memdup_user(arg, sizeof(*ipa));
4335 if (IS_ERR(ipa)) {
4336 ret = PTR_ERR(ipa);
4337 ipa = NULL;
4338 goto out;
4339 }
4340
4341 size = min_t(u32, ipa->size, 4096);
4342 ipath = init_ipath(size, root, path);
4343 if (IS_ERR(ipath)) {
4344 ret = PTR_ERR(ipath);
4345 ipath = NULL;
4346 goto out;
4347 }
4348
4349 ret = paths_from_inode(ipa->inum, ipath);
4350 if (ret < 0)
4351 goto out;
4352
4353 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4354 rel_ptr = ipath->fspath->val[i] -
4355 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4356 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4357 }
4358
745c4d8e
JM		 4359 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4360 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 4361 if (ret) {
4362 ret = -EFAULT;
4363 goto out;
4364 }
4365
4366out:
4367 btrfs_free_path(path);
4368 free_ipath(ipath);
4369 kfree(ipa);
4370
4371 return ret;
4372}
4373
4374static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4375{
4376 struct btrfs_data_container *inodes = ctx;
4377 const size_t c = 3 * sizeof(u64);
4378
4379 if (inodes->bytes_left >= c) {
4380 inodes->bytes_left -= c;
4381 inodes->val[inodes->elem_cnt] = inum;
4382 inodes->val[inodes->elem_cnt + 1] = offset;
4383 inodes->val[inodes->elem_cnt + 2] = root;
4384 inodes->elem_cnt += 3;
4385 } else {
4386 inodes->bytes_missing += c - inodes->bytes_left;
4387 inodes->bytes_left = 0;
4388 inodes->elem_missed += 3;
4389 }
4390
4391 return 0;
4392}
4393
4394static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4395 void __user *arg)
4396{
4397 int ret = 0;
4398 int size;
d7728c96
JS		 4399 struct btrfs_ioctl_logical_ino_args *loi;
4400 struct btrfs_data_container *inodes = NULL;
4401 struct btrfs_path *path = NULL;
d7728c96
JS		 4402
4403 if (!capable(CAP_SYS_ADMIN))
4404 return -EPERM;
4405
4406 loi = memdup_user(arg, sizeof(*loi));
4407 if (IS_ERR(loi)) {
4408 ret = PTR_ERR(loi);
4409 loi = NULL;
4410 goto out;
4411 }
4412
4413 path = btrfs_alloc_path();
4414 if (!path) {
4415 ret = -ENOMEM;
4416 goto out;
4417 }
4418
425d17a2 4419 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 4420 inodes = init_data_container(size);
4421 if (IS_ERR(inodes)) {
4422 ret = PTR_ERR(inodes);
4423 inodes = NULL;
4424 goto out;
4425 }
4426
df031f07
LB		 4427 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4428 build_ino_list, inodes);
4429 if (ret == -EINVAL)
d7728c96
JS		 4430 ret = -ENOENT;
4431 if (ret < 0)
4432 goto out;
4433
745c4d8e
JM		 4434 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4435 (void *)(unsigned long)inodes, size);
d7728c96
JS		 4436 if (ret)
4437 ret = -EFAULT;
4438
4439out:
4440 btrfs_free_path(path);
425d17a2 4441 vfree(inodes);
d7728c96
JS		 4442 kfree(loi);
4443
4444 return ret;
4445}
4446
19a39dce 4447void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 4448 struct btrfs_ioctl_balance_args *bargs)
4449{
4450 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4451
4452 bargs->flags = bctl->flags;
4453
837d5b6e
ID		 4454 if (atomic_read(&fs_info->balance_running))
4455 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4456 if (atomic_read(&fs_info->balance_pause_req))
4457 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4458 if (atomic_read(&fs_info->balance_cancel_req))
4459 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4460
c9e9f97b
ID		 4461 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4462 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4463 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 4464
4465 if (lock) {
4466 spin_lock(&fs_info->balance_lock);
4467 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4468 spin_unlock(&fs_info->balance_lock);
4469 } else {
4470 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4471 }
c9e9f97b
ID		 4472}
4473
9ba1f6e4 4474static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4475{
496ad9aa 4476 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4477 struct btrfs_fs_info *fs_info = root->fs_info;
4478 struct btrfs_ioctl_balance_args *bargs;
4479 struct btrfs_balance_control *bctl;
ed0fb78f 4480 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4481 int ret;
4482
4483 if (!capable(CAP_SYS_ADMIN))
4484 return -EPERM;
4485
e54bfa31 4486 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4487 if (ret)
4488 return ret;
4489
ed0fb78f
ID		 4490again:
4491 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4492 mutex_lock(&fs_info->volume_mutex);
4493 mutex_lock(&fs_info->balance_mutex);
4494 need_unlock = true;
4495 goto locked;
4496 }
4497
4498 /*
4499 * mut. excl. ops lock is locked. Three possibilites:
4500 * (1) some other op is running
4501 * (2) balance is running
4502 * (3) balance is paused -- special case (think resume)
4503 */
c9e9f97b 4504 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4505 if (fs_info->balance_ctl) {
4506 /* this is either (2) or (3) */
4507 if (!atomic_read(&fs_info->balance_running)) {
4508 mutex_unlock(&fs_info->balance_mutex);
4509 if (!mutex_trylock(&fs_info->volume_mutex))
4510 goto again;
4511 mutex_lock(&fs_info->balance_mutex);
4512
4513 if (fs_info->balance_ctl &&
4514 !atomic_read(&fs_info->balance_running)) {
4515 /* this is (3) */
4516 need_unlock = false;
4517 goto locked;
4518 }
4519
4520 mutex_unlock(&fs_info->balance_mutex);
4521 mutex_unlock(&fs_info->volume_mutex);
4522 goto again;
4523 } else {
4524 /* this is (2) */
4525 mutex_unlock(&fs_info->balance_mutex);
4526 ret = -EINPROGRESS;
4527 goto out;
4528 }
4529 } else {
4530 /* this is (1) */
4531 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4532 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4533 goto out;
4534 }
4535
4536locked:
4537 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 4538
4539 if (arg) {
4540 bargs = memdup_user(arg, sizeof(*bargs));
4541 if (IS_ERR(bargs)) {
4542 ret = PTR_ERR(bargs);
ed0fb78f 4543 goto out_unlock;
c9e9f97b 4544 }
de322263
ID		 4545
4546 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4547 if (!fs_info->balance_ctl) {
4548 ret = -ENOTCONN;
4549 goto out_bargs;
4550 }
4551
4552 bctl = fs_info->balance_ctl;
4553 spin_lock(&fs_info->balance_lock);
4554 bctl->flags |= BTRFS_BALANCE_RESUME;
4555 spin_unlock(&fs_info->balance_lock);
4556
4557 goto do_balance;
4558 }
c9e9f97b
ID		 4559 } else {
4560 bargs = NULL;
4561 }
4562
ed0fb78f 4563 if (fs_info->balance_ctl) {
837d5b6e
ID		 4564 ret = -EINPROGRESS;
4565 goto out_bargs;
4566 }
4567
c9e9f97b
ID		 4568 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4569 if (!bctl) {
4570 ret = -ENOMEM;
4571 goto out_bargs;
4572 }
4573
4574 bctl->fs_info = fs_info;
4575 if (arg) {
4576 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4577 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4578 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4579
4580 bctl->flags = bargs->flags;
f43ffb60
ID		 4581 } else {
4582 /* balance everything - no filters */
4583 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4584 }
4585
de322263 4586do_balance:
c9e9f97b 4587 /*
ed0fb78f
ID		 4588 * Ownership of bctl and mutually_exclusive_operation_running
4589 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4590 * or, if restriper was paused all the way until unmount, in
4591 * free_fs_info. mutually_exclusive_operation_running is
4592 * cleared in __cancel_balance.
c9e9f97b 4593 */
ed0fb78f
ID		 4594 need_unlock = false;
4595
4596 ret = btrfs_balance(bctl, bargs);
4597
c9e9f97b
ID		 4598 if (arg) {
4599 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4600 ret = -EFAULT;
4601 }
4602
4603out_bargs:
4604 kfree(bargs);
ed0fb78f 4605out_unlock:
c9e9f97b
ID		 4606 mutex_unlock(&fs_info->balance_mutex);
4607 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 4608 if (need_unlock)
4609 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4610out:
e54bfa31 4611 mnt_drop_write_file(file);
c9e9f97b
ID		 4612 return ret;
4613}
4614
837d5b6e
ID		 4615static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4616{
4617 if (!capable(CAP_SYS_ADMIN))
4618 return -EPERM;
4619
4620 switch (cmd) {
4621 case BTRFS_BALANCE_CTL_PAUSE:
4622 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4623 case BTRFS_BALANCE_CTL_CANCEL:
4624 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4625 }
4626
4627 return -EINVAL;
4628}
4629
19a39dce
ID		 4630static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4631 void __user *arg)
4632{
4633 struct btrfs_fs_info *fs_info = root->fs_info;
4634 struct btrfs_ioctl_balance_args *bargs;
4635 int ret = 0;
4636
4637 if (!capable(CAP_SYS_ADMIN))
4638 return -EPERM;
4639
4640 mutex_lock(&fs_info->balance_mutex);
4641 if (!fs_info->balance_ctl) {
4642 ret = -ENOTCONN;
4643 goto out;
4644 }
4645
4646 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4647 if (!bargs) {
4648 ret = -ENOMEM;
4649 goto out;
4650 }
4651
4652 update_ioctl_balance_args(fs_info, 1, bargs);
4653
4654 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4655 ret = -EFAULT;
4656
4657 kfree(bargs);
4658out:
4659 mutex_unlock(&fs_info->balance_mutex);
4660 return ret;
4661}
4662
905b0dda 4663static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4664{
496ad9aa 4665 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4666 struct btrfs_ioctl_quota_ctl_args *sa;
4667 struct btrfs_trans_handle *trans = NULL;
4668 int ret;
4669 int err;
4670
4671 if (!capable(CAP_SYS_ADMIN))
4672 return -EPERM;
4673
905b0dda
MX		 4674 ret = mnt_want_write_file(file);
4675 if (ret)
4676 return ret;
5d13a37b
AJ		 4677
4678 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4679 if (IS_ERR(sa)) {
4680 ret = PTR_ERR(sa);
4681 goto drop_write;
4682 }
5d13a37b 4683
7708f029 4684 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4685 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4686 if (IS_ERR(trans)) {
4687 ret = PTR_ERR(trans);
4688 goto out;
5d13a37b
AJ		 4689 }
4690
4691 switch (sa->cmd) {
4692 case BTRFS_QUOTA_CTL_ENABLE:
4693 ret = btrfs_quota_enable(trans, root->fs_info);
4694 break;
4695 case BTRFS_QUOTA_CTL_DISABLE:
4696 ret = btrfs_quota_disable(trans, root->fs_info);
4697 break;
5d13a37b
AJ		 4698 default:
4699 ret = -EINVAL;
4700 break;
4701 }
4702
2f232036
JS		 4703 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4704 if (err && !ret)
4705 ret = err;
5d13a37b
AJ		 4706out:
4707 kfree(sa);
7708f029 4708 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4709drop_write:
4710 mnt_drop_write_file(file);
5d13a37b
AJ		 4711 return ret;
4712}
4713
905b0dda 4714static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4715{
496ad9aa 4716 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4717 struct btrfs_ioctl_qgroup_assign_args *sa;
4718 struct btrfs_trans_handle *trans;
4719 int ret;
4720 int err;
4721
4722 if (!capable(CAP_SYS_ADMIN))
4723 return -EPERM;
4724
905b0dda
MX		 4725 ret = mnt_want_write_file(file);
4726 if (ret)
4727 return ret;
5d13a37b
AJ		 4728
4729 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4730 if (IS_ERR(sa)) {
4731 ret = PTR_ERR(sa);
4732 goto drop_write;
4733 }
5d13a37b
AJ		 4734
4735 trans = btrfs_join_transaction(root);
4736 if (IS_ERR(trans)) {
4737 ret = PTR_ERR(trans);
4738 goto out;
4739 }
4740
4741 /* FIXME: check if the IDs really exist */
4742 if (sa->assign) {
4743 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4744 sa->src, sa->dst);
4745 } else {
4746 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4747 sa->src, sa->dst);
4748 }
4749
e082f563
QW		 4750 /* update qgroup status and info */
4751 err = btrfs_run_qgroups(trans, root->fs_info);
4752 if (err < 0)
4753 btrfs_error(root->fs_info, ret,
4754 "failed to update qgroup status and info\n");
5d13a37b
AJ		 4755 err = btrfs_end_transaction(trans, root);
4756 if (err && !ret)
4757 ret = err;
4758
4759out:
4760 kfree(sa);
905b0dda
MX		 4761drop_write:
4762 mnt_drop_write_file(file);
5d13a37b
AJ		 4763 return ret;
4764}
4765
905b0dda 4766static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4767{
496ad9aa 4768 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4769 struct btrfs_ioctl_qgroup_create_args *sa;
4770 struct btrfs_trans_handle *trans;
4771 int ret;
4772 int err;
4773
4774 if (!capable(CAP_SYS_ADMIN))
4775 return -EPERM;
4776
905b0dda
MX		 4777 ret = mnt_want_write_file(file);
4778 if (ret)
4779 return ret;
5d13a37b
AJ		 4780
4781 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4782 if (IS_ERR(sa)) {
4783 ret = PTR_ERR(sa);
4784 goto drop_write;
4785 }
5d13a37b 4786
d86e56cf
MX		 4787 if (!sa->qgroupid) {
4788 ret = -EINVAL;
4789 goto out;
4790 }
4791
5d13a37b
AJ		 4792 trans = btrfs_join_transaction(root);
4793 if (IS_ERR(trans)) {
4794 ret = PTR_ERR(trans);
4795 goto out;
4796 }
4797
4798 /* FIXME: check if the IDs really exist */
4799 if (sa->create) {
4087cf24 4800 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
5d13a37b
AJ		 4801 } else {
4802 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4803 }
4804
4805 err = btrfs_end_transaction(trans, root);
4806 if (err && !ret)
4807 ret = err;
4808
4809out:
4810 kfree(sa);
905b0dda
MX		 4811drop_write:
4812 mnt_drop_write_file(file);
5d13a37b
AJ		 4813 return ret;
4814}
4815
905b0dda 4816static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4817{
496ad9aa 4818 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4819 struct btrfs_ioctl_qgroup_limit_args *sa;
4820 struct btrfs_trans_handle *trans;
4821 int ret;
4822 int err;
4823 u64 qgroupid;
4824
4825 if (!capable(CAP_SYS_ADMIN))
4826 return -EPERM;
4827
905b0dda
MX		 4828 ret = mnt_want_write_file(file);
4829 if (ret)
4830 return ret;
5d13a37b
AJ		 4831
4832 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4833 if (IS_ERR(sa)) {
4834 ret = PTR_ERR(sa);
4835 goto drop_write;
4836 }
5d13a37b
AJ		 4837
4838 trans = btrfs_join_transaction(root);
4839 if (IS_ERR(trans)) {
4840 ret = PTR_ERR(trans);
4841 goto out;
4842 }
4843
4844 qgroupid = sa->qgroupid;
4845 if (!qgroupid) {
4846 /* take the current subvol as qgroup */
4847 qgroupid = root->root_key.objectid;
4848 }
4849
4850 /* FIXME: check if the IDs really exist */
4851 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4852
4853 err = btrfs_end_transaction(trans, root);
4854 if (err && !ret)
4855 ret = err;
4856
4857out:
4858 kfree(sa);
905b0dda
MX		 4859drop_write:
4860 mnt_drop_write_file(file);
5d13a37b
AJ		 4861 return ret;
4862}
4863
2f232036
JS		 4864static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4865{
6d0379ec 4866 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4867 struct btrfs_ioctl_quota_rescan_args *qsa;
4868 int ret;
4869
4870 if (!capable(CAP_SYS_ADMIN))
4871 return -EPERM;
4872
4873 ret = mnt_want_write_file(file);
4874 if (ret)
4875 return ret;
4876
4877 qsa = memdup_user(arg, sizeof(*qsa));
4878 if (IS_ERR(qsa)) {
4879 ret = PTR_ERR(qsa);
4880 goto drop_write;
4881 }
4882
4883 if (qsa->flags) {
4884 ret = -EINVAL;
4885 goto out;
4886 }
4887
4888 ret = btrfs_qgroup_rescan(root->fs_info);
4889
4890out:
4891 kfree(qsa);
4892drop_write:
4893 mnt_drop_write_file(file);
4894 return ret;
4895}
4896
4897static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4898{
6d0379ec 4899 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4900 struct btrfs_ioctl_quota_rescan_args *qsa;
4901 int ret = 0;
4902
4903 if (!capable(CAP_SYS_ADMIN))
4904 return -EPERM;
4905
4906 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4907 if (!qsa)
4908 return -ENOMEM;
4909
4910 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4911 qsa->flags = 1;
4912 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4913 }
4914
4915 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4916 ret = -EFAULT;
4917
4918 kfree(qsa);
4919 return ret;
4920}
4921
57254b6e
JS		 4922static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4923{
54563d41 4924 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
57254b6e
JS		 4925
4926 if (!capable(CAP_SYS_ADMIN))
4927 return -EPERM;
4928
4929 return btrfs_qgroup_wait_for_completion(root->fs_info);
4930}
4931
abccd00f
HM		 4932static long _btrfs_ioctl_set_received_subvol(struct file *file,
4933 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4934{
496ad9aa 4935 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4936 struct btrfs_root *root = BTRFS_I(inode)->root;
4937 struct btrfs_root_item *root_item = &root->root_item;
4938 struct btrfs_trans_handle *trans;
4939 struct timespec ct = CURRENT_TIME;
4940 int ret = 0;
dd5f9615 4941 int received_uuid_changed;
8ea05e3a 4942
bd60ea0f
DS		 4943 if (!inode_owner_or_capable(inode))
4944 return -EPERM;
4945
8ea05e3a
AB		 4946 ret = mnt_want_write_file(file);
4947 if (ret < 0)
4948 return ret;
4949
4950 down_write(&root->fs_info->subvol_sem);
4951
4952 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4953 ret = -EINVAL;
4954 goto out;
4955 }
4956
4957 if (btrfs_root_readonly(root)) {
4958 ret = -EROFS;
4959 goto out;
4960 }
4961
dd5f9615
SB		 4962 /*
4963 * 1 - root item
4964 * 2 - uuid items (received uuid + subvol uuid)
4965 */
4966 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4967 if (IS_ERR(trans)) {
4968 ret = PTR_ERR(trans);
4969 trans = NULL;
4970 goto out;
4971 }
4972
4973 sa->rtransid = trans->transid;
4974 sa->rtime.sec = ct.tv_sec;
4975 sa->rtime.nsec = ct.tv_nsec;
4976
dd5f9615
SB		 4977 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4978 BTRFS_UUID_SIZE);
4979 if (received_uuid_changed &&
4980 !btrfs_is_empty_uuid(root_item->received_uuid))
4981 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4982 root_item->received_uuid,
4983 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4984 root->root_key.objectid);
8ea05e3a
AB		 4985 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4986 btrfs_set_root_stransid(root_item, sa->stransid);
4987 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4988 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4989 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4990 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4991 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4992
4993 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4994 &root->root_key, &root->root_item);
4995 if (ret < 0) {
4996 btrfs_end_transaction(trans, root);
8ea05e3a 4997 goto out;
dd5f9615
SB		 4998 }
4999 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5000 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5001 sa->uuid,
5002 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5003 root->root_key.objectid);
5004 if (ret < 0 && ret != -EEXIST) {
5005 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 5006 goto out;
dd5f9615
SB		 5007 }
5008 }
5009 ret = btrfs_commit_transaction(trans, root);
5010 if (ret < 0) {
5011 btrfs_abort_transaction(trans, root, ret);
5012 goto out;
8ea05e3a
AB		 5013 }
5014
abccd00f
HM		 5015out:
5016 up_write(&root->fs_info->subvol_sem);
5017 mnt_drop_write_file(file);
5018 return ret;
5019}
5020
5021#ifdef CONFIG_64BIT
5022static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5023 void __user *arg)
5024{
5025 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5026 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5027 int ret = 0;
5028
5029 args32 = memdup_user(arg, sizeof(*args32));
5030 if (IS_ERR(args32)) {
5031 ret = PTR_ERR(args32);
5032 args32 = NULL;
5033 goto out;
5034 }
5035
5036 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
84dbeb87
DC		 5037 if (!args64) {
5038 ret = -ENOMEM;
abccd00f
HM		 5039 goto out;
5040 }
5041
5042 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5043 args64->stransid = args32->stransid;
5044 args64->rtransid = args32->rtransid;
5045 args64->stime.sec = args32->stime.sec;
5046 args64->stime.nsec = args32->stime.nsec;
5047 args64->rtime.sec = args32->rtime.sec;
5048 args64->rtime.nsec = args32->rtime.nsec;
5049 args64->flags = args32->flags;
5050
5051 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5052 if (ret)
5053 goto out;
5054
5055 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5056 args32->stransid = args64->stransid;
5057 args32->rtransid = args64->rtransid;
5058 args32->stime.sec = args64->stime.sec;
5059 args32->stime.nsec = args64->stime.nsec;
5060 args32->rtime.sec = args64->rtime.sec;
5061 args32->rtime.nsec = args64->rtime.nsec;
5062 args32->flags = args64->flags;
5063
5064 ret = copy_to_user(arg, args32, sizeof(*args32));
5065 if (ret)
5066 ret = -EFAULT;
5067
5068out:
5069 kfree(args32);
5070 kfree(args64);
5071 return ret;
5072}
5073#endif
5074
5075static long btrfs_ioctl_set_received_subvol(struct file *file,
5076 void __user *arg)
5077{
5078 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5079 int ret = 0;
5080
5081 sa = memdup_user(arg, sizeof(*sa));
5082 if (IS_ERR(sa)) {
5083 ret = PTR_ERR(sa);
5084 sa = NULL;
5085 goto out;
5086 }
5087
5088 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5089
5090 if (ret)
5091 goto out;
5092
8ea05e3a
AB		 5093 ret = copy_to_user(arg, sa, sizeof(*sa));
5094 if (ret)
5095 ret = -EFAULT;
5096
5097out:
5098 kfree(sa);
8ea05e3a
AB		 5099 return ret;
5100}
5101
867ab667 5102static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5103{
6d0379ec 5104 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 5105 size_t len;
867ab667 5106 int ret;
a1b83ac5
AJ		 5107 char label[BTRFS_LABEL_SIZE];
5108
5109 spin_lock(&root->fs_info->super_lock);
5110 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5111 spin_unlock(&root->fs_info->super_lock);
5112
5113 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5114
5115 if (len == BTRFS_LABEL_SIZE) {
efe120a0
FH		 5116 btrfs_warn(root->fs_info,
5117 "label is too long, return the first %zu bytes", --len);
867ab667 5118 }
5119
867ab667 5120 ret = copy_to_user(arg, label, len);
867ab667 5121
5122 return ret ? -EFAULT : 0;
5123}
5124
a8bfd4ab 5125static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5126{
6d0379ec 5127 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 5128 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5129 struct btrfs_trans_handle *trans;
5130 char label[BTRFS_LABEL_SIZE];
5131 int ret;
5132
5133 if (!capable(CAP_SYS_ADMIN))
5134 return -EPERM;
5135
5136 if (copy_from_user(label, arg, sizeof(label)))
5137 return -EFAULT;
5138
5139 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
efe120a0 5140 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
a8bfd4ab 5141 BTRFS_LABEL_SIZE - 1);
5142 return -EINVAL;
5143 }
5144
5145 ret = mnt_want_write_file(file);
5146 if (ret)
5147 return ret;
5148
a8bfd4ab 5149 trans = btrfs_start_transaction(root, 0);
5150 if (IS_ERR(trans)) {
5151 ret = PTR_ERR(trans);
5152 goto out_unlock;
5153 }
5154
a1b83ac5 5155 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 5156 strcpy(super_block->label, label);
a1b83ac5 5157 spin_unlock(&root->fs_info->super_lock);
d0270aca 5158 ret = btrfs_commit_transaction(trans, root);
a8bfd4ab 5159
5160out_unlock:
a8bfd4ab 5161 mnt_drop_write_file(file);
5162 return ret;
5163}
5164
2eaa055f
JM		 5165#define INIT_FEATURE_FLAGS(suffix) \
5166 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5167 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5168 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5169
5170static int btrfs_ioctl_get_supported_features(struct file *file,
5171 void __user *arg)
5172{
5173 static struct btrfs_ioctl_feature_flags features[3] = {
5174 INIT_FEATURE_FLAGS(SUPP),
5175 INIT_FEATURE_FLAGS(SAFE_SET),
5176 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5177 };
5178
5179 if (copy_to_user(arg, &features, sizeof(features)))
5180 return -EFAULT;
5181
5182 return 0;
5183}
5184
5185static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5186{
5187 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5188 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5189 struct btrfs_ioctl_feature_flags features;
5190
5191 features.compat_flags = btrfs_super_compat_flags(super_block);
5192 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5193 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5194
5195 if (copy_to_user(arg, &features, sizeof(features)))
5196 return -EFAULT;
5197
5198 return 0;
5199}
5200
3b02a68a
JM		 5201static int check_feature_bits(struct btrfs_root *root,
5202 enum btrfs_feature_set set,
2eaa055f
JM		 5203 u64 change_mask, u64 flags, u64 supported_flags,
5204 u64 safe_set, u64 safe_clear)
5205{
3b02a68a
JM		 5206 const char *type = btrfs_feature_set_names[set];
5207 char *names;
2eaa055f
JM		 5208 u64 disallowed, unsupported;
5209 u64 set_mask = flags & change_mask;
5210 u64 clear_mask = ~flags & change_mask;
5211
5212 unsupported = set_mask & ~supported_flags;
5213 if (unsupported) {
3b02a68a
JM		 5214 names = btrfs_printable_features(set, unsupported);
5215 if (names) {
5216 btrfs_warn(root->fs_info,
5217 "this kernel does not support the %s feature bit%s",
5218 names, strchr(names, ',') ? "s" : "");
5219 kfree(names);
5220 } else
5221 btrfs_warn(root->fs_info,
2eaa055f
JM		 5222 "this kernel does not support %s bits 0x%llx",
5223 type, unsupported);
5224 return -EOPNOTSUPP;
5225 }
5226
5227 disallowed = set_mask & ~safe_set;
5228 if (disallowed) {
3b02a68a
JM		 5229 names = btrfs_printable_features(set, disallowed);
5230 if (names) {
5231 btrfs_warn(root->fs_info,
5232 "can't set the %s feature bit%s while mounted",
5233 names, strchr(names, ',') ? "s" : "");
5234 kfree(names);
5235 } else
5236 btrfs_warn(root->fs_info,
2eaa055f
JM		 5237 "can't set %s bits 0x%llx while mounted",
5238 type, disallowed);
5239 return -EPERM;
5240 }
5241
5242 disallowed = clear_mask & ~safe_clear;
5243 if (disallowed) {
3b02a68a
JM		 5244 names = btrfs_printable_features(set, disallowed);
5245 if (names) {
5246 btrfs_warn(root->fs_info,
5247 "can't clear the %s feature bit%s while mounted",
5248 names, strchr(names, ',') ? "s" : "");
5249 kfree(names);
5250 } else
5251 btrfs_warn(root->fs_info,
2eaa055f
JM		 5252 "can't clear %s bits 0x%llx while mounted",
5253 type, disallowed);
5254 return -EPERM;
5255 }
5256
5257 return 0;
5258}
5259
5260#define check_feature(root, change_mask, flags, mask_base) \
3b02a68a 5261check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5262 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5263 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5264 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5265
5266static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5267{
5268 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5269 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5270 struct btrfs_ioctl_feature_flags flags[2];
5271 struct btrfs_trans_handle *trans;
5272 u64 newflags;
5273 int ret;
5274
5275 if (!capable(CAP_SYS_ADMIN))
5276 return -EPERM;
5277
5278 if (copy_from_user(flags, arg, sizeof(flags)))
5279 return -EFAULT;
5280
5281 /* Nothing to do */
5282 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5283 !flags[0].incompat_flags)
5284 return 0;
5285
5286 ret = check_feature(root, flags[0].compat_flags,
5287 flags[1].compat_flags, COMPAT);
5288 if (ret)
5289 return ret;
5290
5291 ret = check_feature(root, flags[0].compat_ro_flags,
5292 flags[1].compat_ro_flags, COMPAT_RO);
5293 if (ret)
5294 return ret;
5295
5296 ret = check_feature(root, flags[0].incompat_flags,
5297 flags[1].incompat_flags, INCOMPAT);
5298 if (ret)
5299 return ret;
5300
8051aa1a 5301 trans = btrfs_start_transaction(root, 0);
2eaa055f
JM		 5302 if (IS_ERR(trans))
5303 return PTR_ERR(trans);
5304
5305 spin_lock(&root->fs_info->super_lock);
5306 newflags = btrfs_super_compat_flags(super_block);
5307 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5308 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5309 btrfs_set_super_compat_flags(super_block, newflags);
5310
5311 newflags = btrfs_super_compat_ro_flags(super_block);
5312 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5313 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5314 btrfs_set_super_compat_ro_flags(super_block, newflags);
5315
5316 newflags = btrfs_super_incompat_flags(super_block);
5317 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5318 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5319 btrfs_set_super_incompat_flags(super_block, newflags);
5320 spin_unlock(&root->fs_info->super_lock);
5321
d0270aca 5322 return btrfs_commit_transaction(trans, root);
2eaa055f
JM		 5323}
5324
f46b5a66
CH		 5325long btrfs_ioctl(struct file *file, unsigned int
5326 cmd, unsigned long arg)
5327{
496ad9aa 5328 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 5329 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5330
5331 switch (cmd) {
6cbff00f
CH		 5332 case FS_IOC_GETFLAGS:
5333 return btrfs_ioctl_getflags(file, argp);
5334 case FS_IOC_SETFLAGS:
5335 return btrfs_ioctl_setflags(file, argp);
5336 case FS_IOC_GETVERSION:
5337 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5338 case FITRIM:
5339 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5340 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5341 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5342 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5343 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5344 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5345 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5346 case BTRFS_IOC_SUBVOL_CREATE_V2:
5347 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5348 case BTRFS_IOC_SNAP_DESTROY:
5349 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5350 case BTRFS_IOC_SUBVOL_GETFLAGS:
5351 return btrfs_ioctl_subvol_getflags(file, argp);
5352 case BTRFS_IOC_SUBVOL_SETFLAGS:
5353 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5354 case BTRFS_IOC_DEFAULT_SUBVOL:
5355 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5356 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5357 return btrfs_ioctl_defrag(file, NULL);
5358 case BTRFS_IOC_DEFRAG_RANGE:
5359 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5360 case BTRFS_IOC_RESIZE:
198605a8 5361 return btrfs_ioctl_resize(file, argp);
f46b5a66 5362 case BTRFS_IOC_ADD_DEV:
4bcabaa3 5363 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 5364 case BTRFS_IOC_RM_DEV:
da24927b 5365 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 5366 case BTRFS_IOC_FS_INFO:
5367 return btrfs_ioctl_fs_info(root, argp);
5368 case BTRFS_IOC_DEV_INFO:
5369 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 5370 case BTRFS_IOC_BALANCE:
9ba1f6e4 5371 return btrfs_ioctl_balance(file, NULL);
f46b5a66 5372 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 5373 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5374 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 5375 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 5376 case BTRFS_IOC_TRANS_START:
5377 return btrfs_ioctl_trans_start(file);
5378 case BTRFS_IOC_TRANS_END:
5379 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 5380 case BTRFS_IOC_TREE_SEARCH:
5381 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5382 case BTRFS_IOC_TREE_SEARCH_V2:
5383 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5384 case BTRFS_IOC_INO_LOOKUP:
5385 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5386 case BTRFS_IOC_INO_PATHS:
5387 return btrfs_ioctl_ino_to_path(root, argp);
5388 case BTRFS_IOC_LOGICAL_INO:
5389 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 5390 case BTRFS_IOC_SPACE_INFO:
5391 return btrfs_ioctl_space_info(root, argp);
9b199859
FDBM		 5392 case BTRFS_IOC_SYNC: {
5393 int ret;
5394
6c255e67 5395 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
9b199859
FDBM		 5396 if (ret)
5397 return ret;
ddb52f4f 5398 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
2fad4e83
DS		 5399 /*
5400 * The transaction thread may want to do more work,
5401 * namely it pokes the cleaner ktread that will start
5402 * processing uncleaned subvols.
5403 */
5404 wake_up_process(root->fs_info->transaction_kthread);
9b199859
FDBM		 5405 return ret;
5406 }
46204592 5407 case BTRFS_IOC_START_SYNC:
9a8c28be 5408 return btrfs_ioctl_start_sync(root, argp);
46204592 5409 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 5410 return btrfs_ioctl_wait_sync(root, argp);
475f6387 5411 case BTRFS_IOC_SCRUB:
b8e95489 5412 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 5413 case BTRFS_IOC_SCRUB_CANCEL:
5414 return btrfs_ioctl_scrub_cancel(root, argp);
5415 case BTRFS_IOC_SCRUB_PROGRESS:
5416 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 5417 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5418 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 5419 case BTRFS_IOC_BALANCE_CTL:
5420 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 5421 case BTRFS_IOC_BALANCE_PROGRESS:
5422 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 5423 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5424 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5425#ifdef CONFIG_64BIT
5426 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5427 return btrfs_ioctl_set_received_subvol_32(file, argp);
5428#endif
31db9f7c
AB		 5429 case BTRFS_IOC_SEND:
5430 return btrfs_ioctl_send(file, argp);
c11d2c23 5431 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 5432 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 5433 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5434 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5435 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5436 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5437 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5438 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5439 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5440 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5441 case BTRFS_IOC_QUOTA_RESCAN:
5442 return btrfs_ioctl_quota_rescan(file, argp);
5443 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5444 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5445 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5446 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 5447 case BTRFS_IOC_DEV_REPLACE:
5448 return btrfs_ioctl_dev_replace(root, argp);
867ab667 5449 case BTRFS_IOC_GET_FSLABEL:
5450 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5451 case BTRFS_IOC_SET_FSLABEL:
5452 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 5453 case BTRFS_IOC_FILE_EXTENT_SAME:
5454 return btrfs_ioctl_file_extent_same(file, argp);
2eaa055f
JM		 5455 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5456 return btrfs_ioctl_get_supported_features(file, argp);
5457 case BTRFS_IOC_GET_FEATURES:
5458 return btrfs_ioctl_get_features(file, argp);
5459 case BTRFS_IOC_SET_FEATURES:
5460 return btrfs_ioctl_set_features(file, argp);
f46b5a66
CH		 5461 }
5462
5463 return -ENOTTY;
5464}
Linux 6.x block layer and io_uring tree(s)