projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: Fix a deadlock in btrfs_dev_replace_finishing()
[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
LZ		 334 } else {
335 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 336 }
6cbff00f 337
4da6f1a3 338 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 339 if (IS_ERR(trans)) {
340 ret = PTR_ERR(trans);
341 goto out_drop;
342 }
6cbff00f 343
306424cc 344 btrfs_update_iflags(inode);
0c4d2d95 345 inode_inc_iversion(inode);
306424cc 346 inode->i_ctime = CURRENT_TIME;
6cbff00f 347 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 348
6cbff00f 349 btrfs_end_transaction(trans, root);
f062abf0
LZ		 350 out_drop:
351 if (ret) {
352 ip->flags = ip_oldflags;
353 inode->i_flags = i_oldflags;
354 }
6cbff00f 355
6cbff00f
CH		 356 out_unlock:
357 mutex_unlock(&inode->i_mutex);
e7848683 358 mnt_drop_write_file(file);
2d4e6f6a 359 return ret;
6cbff00f
CH		 360}
361
362static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
363{
496ad9aa 364 struct inode *inode = file_inode(file);
6cbff00f
CH		 365
366 return put_user(inode->i_generation, arg);
367}
f46b5a66 368
f7039b1d
LD		 369static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
370{
54563d41 371 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
f7039b1d
LD		 372 struct btrfs_device *device;
373 struct request_queue *q;
374 struct fstrim_range range;
375 u64 minlen = ULLONG_MAX;
376 u64 num_devices = 0;
815745cf 377 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 378 int ret;
379
380 if (!capable(CAP_SYS_ADMIN))
381 return -EPERM;
382
1f78160c
XG		 383 rcu_read_lock();
384 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
385 dev_list) {
f7039b1d
LD		 386 if (!device->bdev)
387 continue;
388 q = bdev_get_queue(device->bdev);
389 if (blk_queue_discard(q)) {
390 num_devices++;
391 minlen = min((u64)q->limits.discard_granularity,
392 minlen);
393 }
394 }
1f78160c 395 rcu_read_unlock();
f4c697e6 396
f7039b1d
LD		 397 if (!num_devices)
398 return -EOPNOTSUPP;
f7039b1d
LD		 399 if (copy_from_user(&range, arg, sizeof(range)))
400 return -EFAULT;
e515c18b
LC		 401 if (range.start > total_bytes ||
402 range.len < fs_info->sb->s_blocksize)
f4c697e6 403 return -EINVAL;
f7039b1d 404
f4c697e6 405 range.len = min(range.len, total_bytes - range.start);
f7039b1d 406 range.minlen = max(range.minlen, minlen);
815745cf 407 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 408 if (ret < 0)
409 return ret;
410
411 if (copy_to_user(arg, &range, sizeof(range)))
412 return -EFAULT;
413
414 return 0;
415}
416
dd5f9615
SB		 417int btrfs_is_empty_uuid(u8 *uuid)
418{
46e0f66a
CM		 419 int i;
420
421 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
422 if (uuid[i])
423 return 0;
424 }
425 return 1;
dd5f9615
SB		 426}
427
d5c12070 428static noinline int create_subvol(struct inode *dir,
cb8e7090 429 struct dentry *dentry,
72fd032e 430 char *name, int namelen,
6f72c7e2 431 u64 *async_transid,
8696c533 432 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 433{
434 struct btrfs_trans_handle *trans;
435 struct btrfs_key key;
436 struct btrfs_root_item root_item;
437 struct btrfs_inode_item *inode_item;
438 struct extent_buffer *leaf;
d5c12070 439 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 440 struct btrfs_root *new_root;
d5c12070 441 struct btrfs_block_rsv block_rsv;
8ea05e3a 442 struct timespec cur_time = CURRENT_TIME;
5662344b 443 struct inode *inode;
f46b5a66
CH		 444 int ret;
445 int err;
446 u64 objectid;
447 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 448 u64 index = 0;
d5c12070 449 u64 qgroup_reserved;
8ea05e3a 450 uuid_le new_uuid;
f46b5a66 451
581bb050 452 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 453 if (ret)
a22285a6 454 return ret;
6a912213 455
d5c12070 456 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 457 /*
d5c12070
MX		 458 * The same as the snapshot creation, please see the comment
459 * of create_snapshot().
9ed74f2d 460 */
d5c12070 461 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 462 8, &qgroup_reserved, false);
d5c12070
MX		 463 if (ret)
464 return ret;
465
466 trans = btrfs_start_transaction(root, 0);
467 if (IS_ERR(trans)) {
468 ret = PTR_ERR(trans);
de6e8200
LB		 469 btrfs_subvolume_release_metadata(root, &block_rsv,
470 qgroup_reserved);
471 return ret;
d5c12070
MX		 472 }
473 trans->block_rsv = &block_rsv;
474 trans->bytes_reserved = block_rsv.size;
f46b5a66 475
8696c533 476 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 477 if (ret)
478 goto fail;
479
707e8a07 480 leaf = btrfs_alloc_free_block(trans, root, root->nodesize,
5581a51a 481 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 482 if (IS_ERR(leaf)) {
483 ret = PTR_ERR(leaf);
484 goto fail;
485 }
f46b5a66 486
5d4f98a2 487 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 488 btrfs_set_header_bytenr(leaf, leaf->start);
489 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 490 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 491 btrfs_set_header_owner(leaf, objectid);
492
0a4e5586 493 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
f46b5a66 494 BTRFS_FSID_SIZE);
5d4f98a2 495 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
b308bc2f 496 btrfs_header_chunk_tree_uuid(leaf),
5d4f98a2 497 BTRFS_UUID_SIZE);
f46b5a66
CH		 498 btrfs_mark_buffer_dirty(leaf);
499
8ea05e3a
AB		 500 memset(&root_item, 0, sizeof(root_item));
501
f46b5a66 502 inode_item = &root_item.inode;
3cae210f
QW		 503 btrfs_set_stack_inode_generation(inode_item, 1);
504 btrfs_set_stack_inode_size(inode_item, 3);
505 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 506 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 507 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 508
3cae210f
QW		 509 btrfs_set_root_flags(&root_item, 0);
510 btrfs_set_root_limit(&root_item, 0);
511 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 512
f46b5a66 513 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 514 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 515 btrfs_set_root_level(&root_item, 0);
516 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 517 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 518 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 519
8ea05e3a
AB		 520 btrfs_set_root_generation_v2(&root_item,
521 btrfs_root_generation(&root_item));
522 uuid_le_gen(&new_uuid);
523 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 524 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
525 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 526 root_item.ctime = root_item.otime;
527 btrfs_set_root_ctransid(&root_item, trans->transid);
528 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 529
925baedd 530 btrfs_tree_unlock(leaf);
f46b5a66
CH		 531 free_extent_buffer(leaf);
532 leaf = NULL;
533
534 btrfs_set_root_dirid(&root_item, new_dirid);
535
536 key.objectid = objectid;
5d4f98a2 537 key.offset = 0;
962a298f 538 key.type = BTRFS_ROOT_ITEM_KEY;
f46b5a66
CH		 539 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
540 &root_item);
541 if (ret)
542 goto fail;
543
76dda93c
YZ		 544 key.offset = (u64)-1;
545 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 546 if (IS_ERR(new_root)) {
547 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
548 ret = PTR_ERR(new_root);
549 goto fail;
550 }
76dda93c
YZ		 551
552 btrfs_record_root_in_trans(trans, new_root);
553
63541927 554 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 555 if (ret) {
556 /* We potentially lose an unused inode item here */
79787eaa 557 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 558 goto fail;
559 }
560
f46b5a66
CH		 561 /*
562 * insert the directory item
563 */
3de4586c 564 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 565 if (ret) {
566 btrfs_abort_transaction(trans, root, ret);
567 goto fail;
568 }
3de4586c
CM		 569
570 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 571 name, namelen, dir, &key,
3de4586c 572 BTRFS_FT_DIR, index);
79787eaa
JM		 573 if (ret) {
574 btrfs_abort_transaction(trans, root, ret);
f46b5a66 575 goto fail;
79787eaa 576 }
0660b5af 577
52c26179
YZ		 578 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
579 ret = btrfs_update_inode(trans, root, dir);
580 BUG_ON(ret);
581
0660b5af 582 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 583 objectid, root->root_key.objectid,
33345d01 584 btrfs_ino(dir), index, name, namelen);
76dda93c 585 BUG_ON(ret);
f46b5a66 586
dd5f9615
SB		 587 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
588 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
589 objectid);
590 if (ret)
591 btrfs_abort_transaction(trans, root, ret);
592
f46b5a66 593fail:
d5c12070
MX		 594 trans->block_rsv = NULL;
595 trans->bytes_reserved = 0;
de6e8200
LB		 596 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
597
72fd032e
SW		 598 if (async_transid) {
599 *async_transid = trans->transid;
600 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 601 if (err)
602 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 603 } else {
604 err = btrfs_commit_transaction(trans, root);
605 }
f46b5a66
CH		 606 if (err && !ret)
607 ret = err;
1a65e24b 608
5662344b
TI		 609 if (!ret) {
610 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 611 if (IS_ERR(inode))
612 return PTR_ERR(inode);
5662344b
TI		 613 d_instantiate(dentry, inode);
614 }
f46b5a66
CH		 615 return ret;
616}
617
8257b2dc
MX		 618static void btrfs_wait_nocow_write(struct btrfs_root *root)
619{
620 s64 writers;
621 DEFINE_WAIT(wait);
622
623 do {
624 prepare_to_wait(&root->subv_writers->wait, &wait,
625 TASK_UNINTERRUPTIBLE);
626
627 writers = percpu_counter_sum(&root->subv_writers->counter);
628 if (writers)
629 schedule();
630
631 finish_wait(&root->subv_writers->wait, &wait);
632 } while (writers);
633}
634
e9662f70
MX		 635static int create_snapshot(struct btrfs_root *root, struct inode *dir,
636 struct dentry *dentry, char *name, int namelen,
637 u64 *async_transid, bool readonly,
638 struct btrfs_qgroup_inherit *inherit)
f46b5a66 639{
2e4bfab9 640 struct inode *inode;
f46b5a66
CH		 641 struct btrfs_pending_snapshot *pending_snapshot;
642 struct btrfs_trans_handle *trans;
2e4bfab9 643 int ret;
f46b5a66 644
27cdeb70 645 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
f46b5a66
CH		 646 return -EINVAL;
647
8257b2dc 648 atomic_inc(&root->will_be_snapshoted);
4e857c58 649 smp_mb__after_atomic();
8257b2dc
MX		 650 btrfs_wait_nocow_write(root);
651
6a03843d
MX		 652 ret = btrfs_start_delalloc_inodes(root, 0);
653 if (ret)
8257b2dc 654 goto out;
6a03843d 655
b0244199 656 btrfs_wait_ordered_extents(root, -1);
6a03843d 657
3de4586c 658 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
8257b2dc
MX		 659 if (!pending_snapshot) {
660 ret = -ENOMEM;
661 goto out;
662 }
a22285a6 663
66d8f3dd
MX		 664 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
665 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 666 /*
667 * 1 - parent dir inode
668 * 2 - dir entries
669 * 1 - root item
670 * 2 - root ref/backref
671 * 1 - root of snapshot
dd5f9615 672 * 1 - UUID item
d5c12070
MX		 673 */
674 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 675 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 676 &pending_snapshot->qgroup_reserved,
677 false);
d5c12070 678 if (ret)
8257b2dc 679 goto free;
d5c12070 680
3de4586c 681 pending_snapshot->dentry = dentry;
f46b5a66 682 pending_snapshot->root = root;
b83cc969 683 pending_snapshot->readonly = readonly;
e9662f70 684 pending_snapshot->dir = dir;
8696c533 685 pending_snapshot->inherit = inherit;
a22285a6 686
d5c12070 687 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 688 if (IS_ERR(trans)) {
689 ret = PTR_ERR(trans);
690 goto fail;
691 }
692
8351583e 693 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 694 list_add(&pending_snapshot->list,
695 &trans->transaction->pending_snapshots);
8351583e 696 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 697 if (async_transid) {
698 *async_transid = trans->transid;
699 ret = btrfs_commit_transaction_async(trans,
700 root->fs_info->extent_root, 1);
00d71c9c
MX		 701 if (ret)
702 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 703 } else {
704 ret = btrfs_commit_transaction(trans,
705 root->fs_info->extent_root);
706 }
aec8030a 707 if (ret)
c37b2b62 708 goto fail;
a22285a6
YZ		 709
710 ret = pending_snapshot->error;
711 if (ret)
712 goto fail;
713
2fbe8c8a 714 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 715 if (IS_ERR(inode)) {
716 ret = PTR_ERR(inode);
717 goto fail;
718 }
5662344b 719
2e4bfab9
YZ		 720 d_instantiate(dentry, inode);
721 ret = 0;
722fail:
d5c12070
MX		 723 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
724 &pending_snapshot->block_rsv,
725 pending_snapshot->qgroup_reserved);
8257b2dc 726free:
a22285a6 727 kfree(pending_snapshot);
8257b2dc
MX		 728out:
729 atomic_dec(&root->will_be_snapshoted);
f46b5a66
CH		 730 return ret;
731}
732
4260f7c7
SW		 733/* copy of check_sticky in fs/namei.c()
734* It's inline, so penalty for filesystems that don't use sticky bit is
735* minimal.
736*/
737static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
738{
2f2f43d3 739 kuid_t fsuid = current_fsuid();
4260f7c7
SW		 740
741 if (!(dir->i_mode & S_ISVTX))
742 return 0;
2f2f43d3 743 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 744 return 0;
2f2f43d3 745 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW		 746 return 0;
747 return !capable(CAP_FOWNER);
748}
749
750/* copy of may_delete in fs/namei.c()
751 * Check whether we can remove a link victim from directory dir, check
752 * whether the type of victim is right.
753 * 1. We can't do it if dir is read-only (done in permission())
754 * 2. We should have write and exec permissions on dir
755 * 3. We can't remove anything from append-only dir
756 * 4. We can't do anything with immutable dir (done in permission())
757 * 5. If the sticky bit on dir is set we should either
758 * a. be owner of dir, or
759 * b. be owner of victim, or
760 * c. have CAP_FOWNER capability
761 * 6. If the victim is append-only or immutable we can't do antyhing with
762 * links pointing to it.
763 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
764 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
765 * 9. We can't remove a root or mountpoint.
766 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
767 * nfs_async_unlink().
768 */
769
67871254 770static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 771{
772 int error;
773
774 if (!victim->d_inode)
775 return -ENOENT;
776
777 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 778 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 779
780 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
781 if (error)
782 return error;
783 if (IS_APPEND(dir))
784 return -EPERM;
785 if (btrfs_check_sticky(dir, victim->d_inode)||
786 IS_APPEND(victim->d_inode)||
787 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
788 return -EPERM;
789 if (isdir) {
790 if (!S_ISDIR(victim->d_inode->i_mode))
791 return -ENOTDIR;
792 if (IS_ROOT(victim))
793 return -EBUSY;
794 } else if (S_ISDIR(victim->d_inode->i_mode))
795 return -EISDIR;
796 if (IS_DEADDIR(dir))
797 return -ENOENT;
798 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
799 return -EBUSY;
800 return 0;
801}
802
cb8e7090
CH		 803/* copy of may_create in fs/namei.c() */
804static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
805{
806 if (child->d_inode)
807 return -EEXIST;
808 if (IS_DEADDIR(dir))
809 return -ENOENT;
810 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
811}
812
813/*
814 * Create a new subvolume below @parent. This is largely modeled after
815 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
816 * inside this filesystem so it's quite a bit simpler.
817 */
76dda93c
YZ		 818static noinline int btrfs_mksubvol(struct path *parent,
819 char *name, int namelen,
72fd032e 820 struct btrfs_root *snap_src,
6f72c7e2 821 u64 *async_transid, bool readonly,
8696c533 822 struct btrfs_qgroup_inherit *inherit)
cb8e7090 823{
76dda93c 824 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 825 struct dentry *dentry;
826 int error;
827
5c50c9b8
DS		 828 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
829 if (error == -EINTR)
830 return error;
cb8e7090
CH		 831
832 dentry = lookup_one_len(name, parent->dentry, namelen);
833 error = PTR_ERR(dentry);
834 if (IS_ERR(dentry))
835 goto out_unlock;
836
837 error = -EEXIST;
838 if (dentry->d_inode)
839 goto out_dput;
840
76dda93c 841 error = btrfs_may_create(dir, dentry);
cb8e7090 842 if (error)
a874a63e 843 goto out_dput;
cb8e7090 844
9c52057c
CM		 845 /*
846 * even if this name doesn't exist, we may get hash collisions.
847 * check for them now when we can safely fail
848 */
849 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
850 dir->i_ino, name,
851 namelen);
852 if (error)
853 goto out_dput;
854
76dda93c
YZ		 855 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
856
857 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
858 goto out_up_read;
859
3de4586c 860 if (snap_src) {
e9662f70 861 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 862 async_transid, readonly, inherit);
3de4586c 863 } else {
d5c12070
MX		 864 error = create_subvol(dir, dentry, name, namelen,
865 async_transid, inherit);
3de4586c 866 }
76dda93c
YZ		 867 if (!error)
868 fsnotify_mkdir(dir, dentry);
869out_up_read:
870 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 871out_dput:
872 dput(dentry);
873out_unlock:
76dda93c 874 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 875 return error;
876}
877
4cb5300b
CM		 878/*
879 * When we're defragging a range, we don't want to kick it off again
880 * if it is really just waiting for delalloc to send it down.
881 * If we find a nice big extent or delalloc range for the bytes in the
882 * file you want to defrag, we return 0 to let you know to skip this
883 * part of the file
884 */
885static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
886{
887 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
888 struct extent_map *em = NULL;
889 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
890 u64 end;
891
892 read_lock(&em_tree->lock);
893 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
894 read_unlock(&em_tree->lock);
895
896 if (em) {
897 end = extent_map_end(em);
898 free_extent_map(em);
899 if (end - offset > thresh)
900 return 0;
901 }
902 /* if we already have a nice delalloc here, just stop */
903 thresh /= 2;
904 end = count_range_bits(io_tree, &offset, offset + thresh,
905 thresh, EXTENT_DELALLOC, 1);
906 if (end >= thresh)
907 return 0;
908 return 1;
909}
910
911/*
912 * helper function to walk through a file and find extents
913 * newer than a specific transid, and smaller than thresh.
914 *
915 * This is used by the defragging code to find new and small
916 * extents
917 */
918static int find_new_extents(struct btrfs_root *root,
919 struct inode *inode, u64 newer_than,
920 u64 *off, int thresh)
921{
922 struct btrfs_path *path;
923 struct btrfs_key min_key;
4cb5300b
CM		 924 struct extent_buffer *leaf;
925 struct btrfs_file_extent_item *extent;
926 int type;
927 int ret;
a4689d2b 928 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 929
930 path = btrfs_alloc_path();
931 if (!path)
932 return -ENOMEM;
933
a4689d2b 934 min_key.objectid = ino;
4cb5300b
CM		 935 min_key.type = BTRFS_EXTENT_DATA_KEY;
936 min_key.offset = *off;
937
67871254 938 while (1) {
6174d3cb 939 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 940 if (ret != 0)
941 goto none;
f094c9bd 942process_slot:
a4689d2b 943 if (min_key.objectid != ino)
4cb5300b
CM		 944 goto none;
945 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
946 goto none;
947
948 leaf = path->nodes[0];
949 extent = btrfs_item_ptr(leaf, path->slots[0],
950 struct btrfs_file_extent_item);
951
952 type = btrfs_file_extent_type(leaf, extent);
953 if (type == BTRFS_FILE_EXTENT_REG &&
954 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
955 check_defrag_in_cache(inode, min_key.offset, thresh)) {
956 *off = min_key.offset;
957 btrfs_free_path(path);
958 return 0;
959 }
960
f094c9bd
FM		 961 path->slots[0]++;
962 if (path->slots[0] < btrfs_header_nritems(leaf)) {
963 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
964 goto process_slot;
965 }
966
4cb5300b
CM		 967 if (min_key.offset == (u64)-1)
968 goto none;
969
970 min_key.offset++;
971 btrfs_release_path(path);
972 }
973none:
974 btrfs_free_path(path);
975 return -ENOENT;
976}
977
6c282eb4 978static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 979{
980 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 981 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
982 struct extent_map *em;
983 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 984
6c282eb4
LZ		 985 /*
986 * hopefully we have this extent in the tree already, try without
987 * the full extent lock
988 */
17ce6ef8 989 read_lock(&em_tree->lock);
6c282eb4 990 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 991 read_unlock(&em_tree->lock);
992
6c282eb4 993 if (!em) {
308d9800
FM		 994 struct extent_state *cached = NULL;
995 u64 end = start + len - 1;
996
6c282eb4 997 /* get the big lock and read metadata off disk */
308d9800 998 lock_extent_bits(io_tree, start, end, 0, &cached);
6c282eb4 999 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
308d9800 1000 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
6c282eb4
LZ		 1001
1002 if (IS_ERR(em))
1003 return NULL;
1004 }
1005
1006 return em;
1007}
17ce6ef8 1008
6c282eb4
LZ		 1009static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1010{
1011 struct extent_map *next;
1012 bool ret = true;
1013
1014 /* this is the last extent */
1015 if (em->start + em->len >= i_size_read(inode))
1016 return false;
1017
1018 next = defrag_lookup_extent(inode, em->start + em->len);
e9512d72
CM		 1019 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1020 ret = false;
1021 else if ((em->block_start + em->block_len == next->block_start) &&
1022 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
6c282eb4
LZ		 1023 ret = false;
1024
1025 free_extent_map(next);
17ce6ef8
LB		 1026 return ret;
1027}
1028
6c282eb4 1029static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 1030 u64 *last_len, u64 *skip, u64 *defrag_end,
1031 int compress)
940100a4 1032{
6c282eb4 1033 struct extent_map *em;
940100a4 1034 int ret = 1;
6c282eb4 1035 bool next_mergeable = true;
940100a4
CM		 1036
1037 /*
008873ea 1038 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1039 * defragging it
1040 */
1041 if (start < *defrag_end)
1042 return 1;
1043
1044 *skip = 0;
1045
6c282eb4
LZ		 1046 em = defrag_lookup_extent(inode, start);
1047 if (!em)
1048 return 0;
940100a4
CM		 1049
1050 /* this will cover holes, and inline extents */
17ce6ef8 1051 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1052 ret = 0;
17ce6ef8
LB		 1053 goto out;
1054 }
1055
6c282eb4 1056 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1057 /*
6c282eb4
LZ		 1058 * we hit a real extent, if it is big or the next extent is not a
1059 * real extent, don't bother defragging it
940100a4 1060 */
a43a2111 1061 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 1062 (em->len >= thresh || !next_mergeable))
940100a4 1063 ret = 0;
17ce6ef8 1064out:
940100a4
CM		 1065 /*
1066 * last_len ends up being a counter of how many bytes we've defragged.
1067 * every time we choose not to defrag an extent, we reset *last_len
1068 * so that the next tiny extent will force a defrag.
1069 *
1070 * The end result of this is that tiny extents before a single big
1071 * extent will force at least part of that big extent to be defragged.
1072 */
1073 if (ret) {
940100a4
CM		 1074 *defrag_end = extent_map_end(em);
1075 } else {
1076 *last_len = 0;
1077 *skip = extent_map_end(em);
1078 *defrag_end = 0;
1079 }
1080
1081 free_extent_map(em);
1082 return ret;
1083}
1084
4cb5300b
CM		 1085/*
1086 * it doesn't do much good to defrag one or two pages
1087 * at a time. This pulls in a nice chunk of pages
1088 * to COW and defrag.
1089 *
1090 * It also makes sure the delalloc code has enough
1091 * dirty data to avoid making new small extents as part
1092 * of the defrag
1093 *
1094 * It's a good idea to start RA on this range
1095 * before calling this.
1096 */
1097static int cluster_pages_for_defrag(struct inode *inode,
1098 struct page **pages,
1099 unsigned long start_index,
c41570c9 1100 unsigned long num_pages)
f46b5a66 1101{
4cb5300b
CM		 1102 unsigned long file_end;
1103 u64 isize = i_size_read(inode);
1104 u64 page_start;
1105 u64 page_end;
1f12bd06 1106 u64 page_cnt;
4cb5300b
CM		 1107 int ret;
1108 int i;
1109 int i_done;
3eaa2885 1110 struct btrfs_ordered_extent *ordered;
4cb5300b 1111 struct extent_state *cached_state = NULL;
600a45e1 1112 struct extent_io_tree *tree;
3b16a4e3 1113 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1114
4cb5300b 1115 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1116 if (!isize || start_index > file_end)
1117 return 0;
1118
1119 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1120
1121 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1122 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1123 if (ret)
1124 return ret;
4cb5300b 1125 i_done = 0;
600a45e1 1126 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1127
1128 /* step one, lock all the pages */
1f12bd06 1129 for (i = 0; i < page_cnt; i++) {
4cb5300b 1130 struct page *page;
600a45e1 1131again:
a94733d0 1132 page = find_or_create_page(inode->i_mapping,
600a45e1 1133 start_index + i, mask);
4cb5300b
CM		 1134 if (!page)
1135 break;
1136
600a45e1
MX		 1137 page_start = page_offset(page);
1138 page_end = page_start + PAGE_CACHE_SIZE - 1;
1139 while (1) {
308d9800
FM		 1140 lock_extent_bits(tree, page_start, page_end,
1141 0, &cached_state);
600a45e1
MX		 1142 ordered = btrfs_lookup_ordered_extent(inode,
1143 page_start);
308d9800
FM		 1144 unlock_extent_cached(tree, page_start, page_end,
1145 &cached_state, GFP_NOFS);
600a45e1
MX		 1146 if (!ordered)
1147 break;
1148
1149 unlock_page(page);
1150 btrfs_start_ordered_extent(inode, ordered, 1);
1151 btrfs_put_ordered_extent(ordered);
1152 lock_page(page);
1f12bd06
LB		 1153 /*
1154 * we unlocked the page above, so we need check if
1155 * it was released or not.
1156 */
1157 if (page->mapping != inode->i_mapping) {
1158 unlock_page(page);
1159 page_cache_release(page);
1160 goto again;
1161 }
600a45e1
MX		 1162 }
1163
4cb5300b
CM		 1164 if (!PageUptodate(page)) {
1165 btrfs_readpage(NULL, page);
1166 lock_page(page);
1167 if (!PageUptodate(page)) {
1168 unlock_page(page);
1169 page_cache_release(page);
1170 ret = -EIO;
1171 break;
1172 }
1173 }
600a45e1 1174
600a45e1
MX		 1175 if (page->mapping != inode->i_mapping) {
1176 unlock_page(page);
1177 page_cache_release(page);
1178 goto again;
1179 }
1180
4cb5300b
CM		 1181 pages[i] = page;
1182 i_done++;
1183 }
1184 if (!i_done || ret)
1185 goto out;
1186
1187 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1188 goto out;
1189
1190 /*
1191 * so now we have a nice long stream of locked
1192 * and up to date pages, lets wait on them
1193 */
1194 for (i = 0; i < i_done; i++)
1195 wait_on_page_writeback(pages[i]);
1196
1197 page_start = page_offset(pages[0]);
1198 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1199
1200 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1201 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1202 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1203 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1204 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1205 &cached_state, GFP_NOFS);
4cb5300b 1206
1f12bd06 1207 if (i_done != page_cnt) {
9e0baf60
JB		 1208 spin_lock(&BTRFS_I(inode)->lock);
1209 BTRFS_I(inode)->outstanding_extents++;
1210 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1211 btrfs_delalloc_release_space(inode,
1f12bd06 1212 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1213 }
1214
1215
9e8a4a8b
LB		 1216 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1217 &cached_state, GFP_NOFS);
4cb5300b
CM		 1218
1219 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1220 page_start, page_end - 1, &cached_state,
1221 GFP_NOFS);
1222
1223 for (i = 0; i < i_done; i++) {
1224 clear_page_dirty_for_io(pages[i]);
1225 ClearPageChecked(pages[i]);
1226 set_page_extent_mapped(pages[i]);
1227 set_page_dirty(pages[i]);
1228 unlock_page(pages[i]);
1229 page_cache_release(pages[i]);
1230 }
1231 return i_done;
1232out:
1233 for (i = 0; i < i_done; i++) {
1234 unlock_page(pages[i]);
1235 page_cache_release(pages[i]);
1236 }
1f12bd06 1237 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1238 return ret;
1239
1240}
1241
1242int btrfs_defrag_file(struct inode *inode, struct file *file,
1243 struct btrfs_ioctl_defrag_range_args *range,
1244 u64 newer_than, unsigned long max_to_defrag)
1245{
1246 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1247 struct file_ra_state *ra = NULL;
f46b5a66 1248 unsigned long last_index;
151a31b2 1249 u64 isize = i_size_read(inode);
940100a4
CM		 1250 u64 last_len = 0;
1251 u64 skip = 0;
1252 u64 defrag_end = 0;
4cb5300b 1253 u64 newer_off = range->start;
f46b5a66 1254 unsigned long i;
008873ea 1255 unsigned long ra_index = 0;
f46b5a66 1256 int ret;
4cb5300b 1257 int defrag_count = 0;
1a419d85 1258 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1259 int extent_thresh = range->extent_thresh;
c41570c9
JM		 1260 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1261 unsigned long cluster = max_cluster;
4cb5300b
CM		 1262 u64 new_align = ~((u64)128 * 1024 - 1);
1263 struct page **pages = NULL;
1264
0abd5b17
LB		 1265 if (isize == 0)
1266 return 0;
1267
1268 if (range->start >= isize)
1269 return -EINVAL;
1a419d85
LZ		 1270
1271 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1272 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1273 return -EINVAL;
1274 if (range->compress_type)
1275 compress_type = range->compress_type;
1276 }
f46b5a66 1277
0abd5b17
LB		 1278 if (extent_thresh == 0)
1279 extent_thresh = 256 * 1024;
940100a4 1280
4cb5300b
CM		 1281 /*
1282 * if we were not given a file, allocate a readahead
1283 * context
1284 */
1285 if (!file) {
1286 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1287 if (!ra)
1288 return -ENOMEM;
1289 file_ra_state_init(ra, inode->i_mapping);
1290 } else {
1291 ra = &file->f_ra;
1292 }
1293
d9b0d9ba 1294 pages = kmalloc_array(max_cluster, sizeof(struct page *),
4cb5300b
CM		 1295 GFP_NOFS);
1296 if (!pages) {
1297 ret = -ENOMEM;
1298 goto out_ra;
1299 }
1300
1301 /* find the last page to defrag */
1e701a32 1302 if (range->start + range->len > range->start) {
151a31b2 1303 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1304 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1305 } else {
151a31b2 1306 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1307 }
1308
4cb5300b
CM		 1309 if (newer_than) {
1310 ret = find_new_extents(root, inode, newer_than,
1311 &newer_off, 64 * 1024);
1312 if (!ret) {
1313 range->start = newer_off;
1314 /*
1315 * we always align our defrag to help keep
1316 * the extents in the file evenly spaced
1317 */
1318 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1319 } else
1320 goto out_ra;
1321 } else {
1322 i = range->start >> PAGE_CACHE_SHIFT;
1323 }
1324 if (!max_to_defrag)
7ec31b54 1325 max_to_defrag = last_index + 1;
4cb5300b 1326
2a0f7f57
LZ		 1327 /*
1328 * make writeback starts from i, so the defrag range can be
1329 * written sequentially.
1330 */
1331 if (i < inode->i_mapping->writeback_index)
1332 inode->i_mapping->writeback_index = i;
1333
f7f43cc8 1334 while (i <= last_index && defrag_count < max_to_defrag &&
ed6078f7 1335 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
4cb5300b
CM		 1336 /*
1337 * make sure we stop running if someone unmounts
1338 * the FS
1339 */
1340 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1341 break;
1342
210549eb 1343 if (btrfs_defrag_cancelled(root->fs_info)) {
efe120a0 1344 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
210549eb
DS		 1345 ret = -EAGAIN;
1346 break;
1347 }
1348
66c26892 1349 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1350 extent_thresh, &last_len, &skip,
a43a2111
AM		 1351 &defrag_end, range->flags &
1352 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1353 unsigned long next;
1354 /*
1355 * the should_defrag function tells us how much to skip
1356 * bump our counter by the suggested amount
1357 */
ed6078f7 1358 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
940100a4
CM		 1359 i = max(i + 1, next);
1360 continue;
1361 }
008873ea
LZ		 1362
1363 if (!newer_than) {
1364 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1365 PAGE_CACHE_SHIFT) - i;
1366 cluster = min(cluster, max_cluster);
1367 } else {
1368 cluster = max_cluster;
1369 }
1370
008873ea
LZ		 1371 if (i + cluster > ra_index) {
1372 ra_index = max(i, ra_index);
1373 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1374 cluster);
1375 ra_index += max_cluster;
1376 }
940100a4 1377
ecb8bea8 1378 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1379 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1380 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1381 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1382 if (ret < 0) {
1383 mutex_unlock(&inode->i_mutex);
4cb5300b 1384 goto out_ra;
ecb8bea8 1385 }
4cb5300b
CM		 1386
1387 defrag_count += ret;
d0e1d66b 1388 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1389 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1390
1391 if (newer_than) {
1392 if (newer_off == (u64)-1)
1393 break;
1394
e1f041e1
LB		 1395 if (ret > 0)
1396 i += ret;
1397
4cb5300b
CM		 1398 newer_off = max(newer_off + 1,
1399 (u64)i << PAGE_CACHE_SHIFT);
1400
1401 ret = find_new_extents(root, inode,
1402 newer_than, &newer_off,
1403 64 * 1024);
1404 if (!ret) {
1405 range->start = newer_off;
1406 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1407 } else {
1408 break;
f46b5a66 1409 }
4cb5300b 1410 } else {
008873ea 1411 if (ret > 0) {
cbcc8326 1412 i += ret;
008873ea
LZ		 1413 last_len += ret << PAGE_CACHE_SHIFT;
1414 } else {
cbcc8326 1415 i++;
008873ea
LZ		 1416 last_len = 0;
1417 }
f46b5a66 1418 }
f46b5a66
CH		 1419 }
1420
dec8ef90 1421 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1422 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1423 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1424 &BTRFS_I(inode)->runtime_flags))
1425 filemap_flush(inode->i_mapping);
1426 }
1e701a32
CM		 1427
1428 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1429 /* the filemap_flush will queue IO into the worker threads, but
1430 * we have to make sure the IO is actually started and that
1431 * ordered extents get created before we return
1432 */
1433 atomic_inc(&root->fs_info->async_submit_draining);
1434 while (atomic_read(&root->fs_info->nr_async_submits) ||
1435 atomic_read(&root->fs_info->async_delalloc_pages)) {
1436 wait_event(root->fs_info->async_submit_wait,
1437 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1438 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1439 }
1440 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1441 }
1442
1a419d85 1443 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1444 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1445 }
1446
60ccf82f 1447 ret = defrag_count;
940100a4 1448
4cb5300b 1449out_ra:
633085c7
FDBM		 1450 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1451 mutex_lock(&inode->i_mutex);
1452 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1453 mutex_unlock(&inode->i_mutex);
1454 }
4cb5300b
CM		 1455 if (!file)
1456 kfree(ra);
1457 kfree(pages);
940100a4 1458 return ret;
f46b5a66
CH		 1459}
1460
198605a8 1461static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1462 void __user *arg)
f46b5a66
CH		 1463{
1464 u64 new_size;
1465 u64 old_size;
1466 u64 devid = 1;
496ad9aa 1467 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1468 struct btrfs_ioctl_vol_args *vol_args;
1469 struct btrfs_trans_handle *trans;
1470 struct btrfs_device *device = NULL;
1471 char *sizestr;
9a40f122 1472 char *retptr;
f46b5a66
CH		 1473 char *devstr = NULL;
1474 int ret = 0;
f46b5a66
CH		 1475 int mod = 0;
1476
e441d54d
CM		 1477 if (!capable(CAP_SYS_ADMIN))
1478 return -EPERM;
1479
198605a8
MX		 1480 ret = mnt_want_write_file(file);
1481 if (ret)
1482 return ret;
1483
5ac00add
SB		 1484 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1485 1)) {
97547676 1486 mnt_drop_write_file(file);
e57138b3 1487 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1488 }
1489
5ac00add 1490 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1491 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1492 if (IS_ERR(vol_args)) {
1493 ret = PTR_ERR(vol_args);
1494 goto out;
1495 }
5516e595
MF		 1496
1497 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1498
f46b5a66
CH		 1499 sizestr = vol_args->name;
1500 devstr = strchr(sizestr, ':');
1501 if (devstr) {
f46b5a66
CH		 1502 sizestr = devstr + 1;
1503 *devstr = '\0';
1504 devstr = vol_args->name;
58dfae63
Z		 1505 ret = kstrtoull(devstr, 10, &devid);
1506 if (ret)
1507 goto out_free;
dfd79829
MX		 1508 if (!devid) {
1509 ret = -EINVAL;
1510 goto out_free;
1511 }
efe120a0 1512 btrfs_info(root->fs_info, "resizing devid %llu", devid);
f46b5a66 1513 }
dba60f3f 1514
aa1b8cd4 1515 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1516 if (!device) {
efe120a0 1517 btrfs_info(root->fs_info, "resizer unable to find device %llu",
c1c9ff7c 1518 devid);
dfd79829 1519 ret = -ENODEV;
c9e9f97b 1520 goto out_free;
f46b5a66 1521 }
dba60f3f
MX		 1522
1523 if (!device->writeable) {
efe120a0
FH		 1524 btrfs_info(root->fs_info,
1525 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1526 devid);
dfd79829 1527 ret = -EPERM;
4e42ae1b
LB		 1528 goto out_free;
1529 }
1530
f46b5a66
CH		 1531 if (!strcmp(sizestr, "max"))
1532 new_size = device->bdev->bd_inode->i_size;
1533 else {
1534 if (sizestr[0] == '-') {
1535 mod = -1;
1536 sizestr++;
1537 } else if (sizestr[0] == '+') {
1538 mod = 1;
1539 sizestr++;
1540 }
9a40f122
GH		 1541 new_size = memparse(sizestr, &retptr);
1542 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1543 ret = -EINVAL;
c9e9f97b 1544 goto out_free;
f46b5a66
CH		 1545 }
1546 }
1547
63a212ab 1548 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1549 ret = -EPERM;
63a212ab
SB		 1550 goto out_free;
1551 }
1552
f46b5a66
CH		 1553 old_size = device->total_bytes;
1554
1555 if (mod < 0) {
1556 if (new_size > old_size) {
1557 ret = -EINVAL;
c9e9f97b 1558 goto out_free;
f46b5a66
CH		 1559 }
1560 new_size = old_size - new_size;
1561 } else if (mod > 0) {
eb8052e0 1562 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1563 ret = -ERANGE;
eb8052e0
WF		 1564 goto out_free;
1565 }
f46b5a66
CH		 1566 new_size = old_size + new_size;
1567 }
1568
1569 if (new_size < 256 * 1024 * 1024) {
1570 ret = -EINVAL;
c9e9f97b 1571 goto out_free;
f46b5a66
CH		 1572 }
1573 if (new_size > device->bdev->bd_inode->i_size) {
1574 ret = -EFBIG;
c9e9f97b 1575 goto out_free;
f46b5a66
CH		 1576 }
1577
1578 do_div(new_size, root->sectorsize);
1579 new_size *= root->sectorsize;
1580
efe120a0 1581 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
c1c9ff7c 1582 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1583
1584 if (new_size > old_size) {
a22285a6 1585 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1586 if (IS_ERR(trans)) {
1587 ret = PTR_ERR(trans);
c9e9f97b 1588 goto out_free;
98d5dc13 1589 }
f46b5a66
CH		 1590 ret = btrfs_grow_device(trans, device, new_size);
1591 btrfs_commit_transaction(trans, root);
ece7d20e 1592 } else if (new_size < old_size) {
f46b5a66 1593 ret = btrfs_shrink_device(device, new_size);
0253f40e 1594 } /* equal, nothing need to do */
f46b5a66 1595
c9e9f97b 1596out_free:
f46b5a66 1597 kfree(vol_args);
c9e9f97b
ID		 1598out:
1599 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1600 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1601 mnt_drop_write_file(file);
f46b5a66
CH		 1602 return ret;
1603}
1604
72fd032e 1605static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1606 char *name, unsigned long fd, int subvol,
1607 u64 *transid, bool readonly,
8696c533 1608 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1609{
f46b5a66 1610 int namelen;
3de4586c 1611 int ret = 0;
f46b5a66 1612
a874a63e
LB		 1613 ret = mnt_want_write_file(file);
1614 if (ret)
1615 goto out;
1616
72fd032e
SW		 1617 namelen = strlen(name);
1618 if (strchr(name, '/')) {
f46b5a66 1619 ret = -EINVAL;
a874a63e 1620 goto out_drop_write;
f46b5a66
CH		 1621 }
1622
16780cab
CM		 1623 if (name[0] == '.' &&
1624 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1625 ret = -EEXIST;
a874a63e 1626 goto out_drop_write;
16780cab
CM		 1627 }
1628
3de4586c 1629 if (subvol) {
72fd032e 1630 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1631 NULL, transid, readonly, inherit);
cb8e7090 1632 } else {
2903ff01 1633 struct fd src = fdget(fd);
3de4586c 1634 struct inode *src_inode;
2903ff01 1635 if (!src.file) {
3de4586c 1636 ret = -EINVAL;
a874a63e 1637 goto out_drop_write;
3de4586c
CM		 1638 }
1639
496ad9aa
AV		 1640 src_inode = file_inode(src.file);
1641 if (src_inode->i_sb != file_inode(file)->i_sb) {
efe120a0
FH		 1642 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1643 "Snapshot src from another FS");
23ad5b17 1644 ret = -EXDEV;
d0242061
DS		 1645 } else if (!inode_owner_or_capable(src_inode)) {
1646 /*
1647 * Subvolume creation is not restricted, but snapshots
1648 * are limited to own subvolumes only
1649 */
1650 ret = -EPERM;
ecd18815
AV		 1651 } else {
1652 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1653 BTRFS_I(src_inode)->root,
1654 transid, readonly, inherit);
3de4586c 1655 }
2903ff01 1656 fdput(src);
cb8e7090 1657 }
a874a63e
LB		 1658out_drop_write:
1659 mnt_drop_write_file(file);
f46b5a66 1660out:
72fd032e
SW		 1661 return ret;
1662}
1663
1664static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1665 void __user *arg, int subvol)
72fd032e 1666{
fa0d2b9b 1667 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1668 int ret;
1669
fa0d2b9b
LZ		 1670 vol_args = memdup_user(arg, sizeof(*vol_args));
1671 if (IS_ERR(vol_args))
1672 return PTR_ERR(vol_args);
1673 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1674
fa0d2b9b 1675 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1676 vol_args->fd, subvol,
6f72c7e2 1677 NULL, false, NULL);
fdfb1e4f 1678
fa0d2b9b
LZ		 1679 kfree(vol_args);
1680 return ret;
1681}
fdfb1e4f 1682
fa0d2b9b
LZ		 1683static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1684 void __user *arg, int subvol)
1685{
1686 struct btrfs_ioctl_vol_args_v2 *vol_args;
1687 int ret;
1688 u64 transid = 0;
1689 u64 *ptr = NULL;
b83cc969 1690 bool readonly = false;
6f72c7e2 1691 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1692
fa0d2b9b
LZ		 1693 vol_args = memdup_user(arg, sizeof(*vol_args));
1694 if (IS_ERR(vol_args))
1695 return PTR_ERR(vol_args);
1696 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1697
b83cc969 1698 if (vol_args->flags &
6f72c7e2
AJ		 1699 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1700 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1701 ret = -EOPNOTSUPP;
c47ca32d 1702 goto free_args;
72fd032e 1703 }
fa0d2b9b
LZ		 1704
1705 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1706 ptr = &transid;
b83cc969
LZ		 1707 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1708 readonly = true;
6f72c7e2
AJ		 1709 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1710 if (vol_args->size > PAGE_CACHE_SIZE) {
1711 ret = -EINVAL;
c47ca32d 1712 goto free_args;
6f72c7e2
AJ		 1713 }
1714 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1715 if (IS_ERR(inherit)) {
1716 ret = PTR_ERR(inherit);
c47ca32d 1717 goto free_args;
6f72c7e2
AJ		 1718 }
1719 }
fa0d2b9b
LZ		 1720
1721 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1722 vol_args->fd, subvol, ptr,
8696c533 1723 readonly, inherit);
c47ca32d
DC		 1724 if (ret)
1725 goto free_inherit;
fa0d2b9b 1726
c47ca32d
DC		 1727 if (ptr && copy_to_user(arg +
1728 offsetof(struct btrfs_ioctl_vol_args_v2,
1729 transid),
1730 ptr, sizeof(*ptr)))
fa0d2b9b 1731 ret = -EFAULT;
c47ca32d
DC		 1732
1733free_inherit:
6f72c7e2 1734 kfree(inherit);
c47ca32d
DC		 1735free_args:
1736 kfree(vol_args);
f46b5a66
CH		 1737 return ret;
1738}
1739
0caa102d
LZ		 1740static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1741 void __user *arg)
1742{
496ad9aa 1743 struct inode *inode = file_inode(file);
0caa102d
LZ		 1744 struct btrfs_root *root = BTRFS_I(inode)->root;
1745 int ret = 0;
1746 u64 flags = 0;
1747
33345d01 1748 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1749 return -EINVAL;
1750
1751 down_read(&root->fs_info->subvol_sem);
1752 if (btrfs_root_readonly(root))
1753 flags |= BTRFS_SUBVOL_RDONLY;
1754 up_read(&root->fs_info->subvol_sem);
1755
1756 if (copy_to_user(arg, &flags, sizeof(flags)))
1757 ret = -EFAULT;
1758
1759 return ret;
1760}
1761
1762static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1763 void __user *arg)
1764{
496ad9aa 1765 struct inode *inode = file_inode(file);
0caa102d
LZ		 1766 struct btrfs_root *root = BTRFS_I(inode)->root;
1767 struct btrfs_trans_handle *trans;
1768 u64 root_flags;
1769 u64 flags;
1770 int ret = 0;
1771
bd60ea0f
DS		 1772 if (!inode_owner_or_capable(inode))
1773 return -EPERM;
1774
b9ca0664
LB		 1775 ret = mnt_want_write_file(file);
1776 if (ret)
1777 goto out;
0caa102d 1778
b9ca0664
LB		 1779 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1780 ret = -EINVAL;
1781 goto out_drop_write;
1782 }
0caa102d 1783
b9ca0664
LB		 1784 if (copy_from_user(&flags, arg, sizeof(flags))) {
1785 ret = -EFAULT;
1786 goto out_drop_write;
1787 }
0caa102d 1788
b9ca0664
LB		 1789 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1790 ret = -EINVAL;
1791 goto out_drop_write;
1792 }
0caa102d 1793
b9ca0664
LB		 1794 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1795 ret = -EOPNOTSUPP;
1796 goto out_drop_write;
1797 }
0caa102d
LZ		 1798
1799 down_write(&root->fs_info->subvol_sem);
1800
1801 /* nothing to do */
1802 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1803 goto out_drop_sem;
0caa102d
LZ		 1804
1805 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1806 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1807 btrfs_set_root_flags(&root->root_item,
1808 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1809 } else {
1810 /*
1811 * Block RO -> RW transition if this subvolume is involved in
1812 * send
1813 */
1814 spin_lock(&root->root_item_lock);
1815 if (root->send_in_progress == 0) {
1816 btrfs_set_root_flags(&root->root_item,
0caa102d 1817 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1818 spin_unlock(&root->root_item_lock);
1819 } else {
1820 spin_unlock(&root->root_item_lock);
1821 btrfs_warn(root->fs_info,
1822 "Attempt to set subvolume %llu read-write during send",
1823 root->root_key.objectid);
1824 ret = -EPERM;
1825 goto out_drop_sem;
1826 }
1827 }
0caa102d
LZ		 1828
1829 trans = btrfs_start_transaction(root, 1);
1830 if (IS_ERR(trans)) {
1831 ret = PTR_ERR(trans);
1832 goto out_reset;
1833 }
1834
b4dc2b8c 1835 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1836 &root->root_key, &root->root_item);
1837
1838 btrfs_commit_transaction(trans, root);
1839out_reset:
1840 if (ret)
1841 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1842out_drop_sem:
0caa102d 1843 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1844out_drop_write:
1845 mnt_drop_write_file(file);
1846out:
0caa102d
LZ		 1847 return ret;
1848}
1849
76dda93c
YZ		 1850/*
1851 * helper to check if the subvolume references other subvolumes
1852 */
1853static noinline int may_destroy_subvol(struct btrfs_root *root)
1854{
1855 struct btrfs_path *path;
175a2b87 1856 struct btrfs_dir_item *di;
76dda93c 1857 struct btrfs_key key;
175a2b87 1858 u64 dir_id;
76dda93c
YZ		 1859 int ret;
1860
1861 path = btrfs_alloc_path();
1862 if (!path)
1863 return -ENOMEM;
1864
175a2b87
JB		 1865 /* Make sure this root isn't set as the default subvol */
1866 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1867 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1868 dir_id, "default", 7, 0);
1869 if (di && !IS_ERR(di)) {
1870 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1871 if (key.objectid == root->root_key.objectid) {
72de6b53
GS		 1872 ret = -EPERM;
1873 btrfs_err(root->fs_info, "deleting default subvolume "
1874 "%llu is not allowed", key.objectid);
175a2b87
JB		 1875 goto out;
1876 }
1877 btrfs_release_path(path);
1878 }
1879
76dda93c
YZ		 1880 key.objectid = root->root_key.objectid;
1881 key.type = BTRFS_ROOT_REF_KEY;
1882 key.offset = (u64)-1;
1883
1884 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1885 &key, path, 0, 0);
1886 if (ret < 0)
1887 goto out;
1888 BUG_ON(ret == 0);
1889
1890 ret = 0;
1891 if (path->slots[0] > 0) {
1892 path->slots[0]--;
1893 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1894 if (key.objectid == root->root_key.objectid &&
1895 key.type == BTRFS_ROOT_REF_KEY)
1896 ret = -ENOTEMPTY;
1897 }
1898out:
1899 btrfs_free_path(path);
1900 return ret;
1901}
1902
ac8e9819
CM		 1903static noinline int key_in_sk(struct btrfs_key *key,
1904 struct btrfs_ioctl_search_key *sk)
1905{
abc6e134
CM		 1906 struct btrfs_key test;
1907 int ret;
1908
1909 test.objectid = sk->min_objectid;
1910 test.type = sk->min_type;
1911 test.offset = sk->min_offset;
1912
1913 ret = btrfs_comp_cpu_keys(key, &test);
1914 if (ret < 0)
ac8e9819 1915 return 0;
abc6e134
CM		 1916
1917 test.objectid = sk->max_objectid;
1918 test.type = sk->max_type;
1919 test.offset = sk->max_offset;
1920
1921 ret = btrfs_comp_cpu_keys(key, &test);
1922 if (ret > 0)
ac8e9819
CM		 1923 return 0;
1924 return 1;
1925}
1926
1927static noinline int copy_to_sk(struct btrfs_root *root,
1928 struct btrfs_path *path,
1929 struct btrfs_key *key,
1930 struct btrfs_ioctl_search_key *sk,
9b6e817d 1931 size_t *buf_size,
ba346b35 1932 char __user *ubuf,
ac8e9819
CM		 1933 unsigned long *sk_offset,
1934 int *num_found)
1935{
1936 u64 found_transid;
1937 struct extent_buffer *leaf;
1938 struct btrfs_ioctl_search_header sh;
1939 unsigned long item_off;
1940 unsigned long item_len;
1941 int nritems;
1942 int i;
1943 int slot;
ac8e9819
CM		 1944 int ret = 0;
1945
1946 leaf = path->nodes[0];
1947 slot = path->slots[0];
1948 nritems = btrfs_header_nritems(leaf);
1949
1950 if (btrfs_header_generation(leaf) > sk->max_transid) {
1951 i = nritems;
1952 goto advance_key;
1953 }
1954 found_transid = btrfs_header_generation(leaf);
1955
1956 for (i = slot; i < nritems; i++) {
1957 item_off = btrfs_item_ptr_offset(leaf, i);
1958 item_len = btrfs_item_size_nr(leaf, i);
1959
03b71c6c
GP		 1960 btrfs_item_key_to_cpu(leaf, key, i);
1961 if (!key_in_sk(key, sk))
1962 continue;
1963
9b6e817d 1964 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 1965 if (*num_found) {
1966 ret = 1;
1967 goto out;
1968 }
1969
1970 /*
1971 * return one empty item back for v1, which does not
1972 * handle -EOVERFLOW
1973 */
1974
9b6e817d 1975 *buf_size = sizeof(sh) + item_len;
ac8e9819 1976 item_len = 0;
8f5f6178
GH		 1977 ret = -EOVERFLOW;
1978 }
ac8e9819 1979
9b6e817d 1980 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 1981 ret = 1;
25c9bc2e 1982 goto out;
ac8e9819
CM		 1983 }
1984
ac8e9819
CM		 1985 sh.objectid = key->objectid;
1986 sh.offset = key->offset;
1987 sh.type = key->type;
1988 sh.len = item_len;
1989 sh.transid = found_transid;
1990
1991 /* copy search result header */
ba346b35
GH		 1992 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1993 ret = -EFAULT;
1994 goto out;
1995 }
1996
ac8e9819
CM		 1997 *sk_offset += sizeof(sh);
1998
1999 if (item_len) {
ba346b35 2000 char __user *up = ubuf + *sk_offset;
ac8e9819 2001 /* copy the item */
ba346b35
GH		 2002 if (read_extent_buffer_to_user(leaf, up,
2003 item_off, item_len)) {
2004 ret = -EFAULT;
2005 goto out;
2006 }
2007
ac8e9819 2008 *sk_offset += item_len;
ac8e9819 2009 }
e2156867 2010 (*num_found)++;
ac8e9819 2011
8f5f6178
GH		 2012 if (ret) /* -EOVERFLOW from above */
2013 goto out;
2014
25c9bc2e
GH		 2015 if (*num_found >= sk->nr_items) {
2016 ret = 1;
2017 goto out;
2018 }
ac8e9819
CM		 2019 }
2020advance_key:
abc6e134
CM		 2021 ret = 0;
2022 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 2023 key->offset++;
abc6e134
CM		 2024 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2025 key->offset = 0;
ac8e9819 2026 key->type++;
abc6e134
CM		 2027 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2028 key->offset = 0;
2029 key->type = 0;
ac8e9819 2030 key->objectid++;
abc6e134
CM		 2031 } else
2032 ret = 1;
25c9bc2e 2033out:
ba346b35
GH		 2034 /*
2035 * 0: all items from this leaf copied, continue with next
2036 * 1: * more items can be copied, but unused buffer is too small
2037 * * all items were found
2038 * Either way, it will stops the loop which iterates to the next
2039 * leaf
2040 * -EOVERFLOW: item was to large for buffer
2041 * -EFAULT: could not copy extent buffer back to userspace
2042 */
ac8e9819
CM		 2043 return ret;
2044}
2045
2046static noinline int search_ioctl(struct inode *inode,
12544442 2047 struct btrfs_ioctl_search_key *sk,
9b6e817d 2048 size_t *buf_size,
ba346b35 2049 char __user *ubuf)
ac8e9819
CM		 2050{
2051 struct btrfs_root *root;
2052 struct btrfs_key key;
ac8e9819 2053 struct btrfs_path *path;
ac8e9819
CM		 2054 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2055 int ret;
2056 int num_found = 0;
2057 unsigned long sk_offset = 0;
2058
9b6e817d
GH		 2059 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2060 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2061 return -EOVERFLOW;
9b6e817d 2062 }
12544442 2063
ac8e9819
CM		 2064 path = btrfs_alloc_path();
2065 if (!path)
2066 return -ENOMEM;
2067
2068 if (sk->tree_id == 0) {
2069 /* search the root of the inode that was passed */
2070 root = BTRFS_I(inode)->root;
2071 } else {
2072 key.objectid = sk->tree_id;
2073 key.type = BTRFS_ROOT_ITEM_KEY;
2074 key.offset = (u64)-1;
2075 root = btrfs_read_fs_root_no_name(info, &key);
2076 if (IS_ERR(root)) {
efe120a0 2077 printk(KERN_ERR "BTRFS: could not find root %llu\n",
ac8e9819
CM		 2078 sk->tree_id);
2079 btrfs_free_path(path);
2080 return -ENOENT;
2081 }
2082 }
2083
2084 key.objectid = sk->min_objectid;
2085 key.type = sk->min_type;
2086 key.offset = sk->min_offset;
2087
67871254 2088 while (1) {
6174d3cb 2089 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2090 if (ret != 0) {
2091 if (ret > 0)
2092 ret = 0;
2093 goto err;
2094 }
ba346b35 2095 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
ac8e9819 2096 &sk_offset, &num_found);
b3b4aa74 2097 btrfs_release_path(path);
25c9bc2e 2098 if (ret)
ac8e9819
CM		 2099 break;
2100
2101 }
8f5f6178
GH		 2102 if (ret > 0)
2103 ret = 0;
ac8e9819
CM		 2104err:
2105 sk->nr_items = num_found;
2106 btrfs_free_path(path);
2107 return ret;
2108}
2109
2110static noinline int btrfs_ioctl_tree_search(struct file *file,
2111 void __user *argp)
2112{
ba346b35
GH		 2113 struct btrfs_ioctl_search_args __user *uargs;
2114 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2115 struct inode *inode;
2116 int ret;
2117 size_t buf_size;
ac8e9819
CM		 2118
2119 if (!capable(CAP_SYS_ADMIN))
2120 return -EPERM;
2121
ba346b35
GH		 2122 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2123
2124 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2125 return -EFAULT;
ac8e9819 2126
ba346b35 2127 buf_size = sizeof(uargs->buf);
ac8e9819 2128
496ad9aa 2129 inode = file_inode(file);
ba346b35 2130 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2131
2132 /*
2133 * In the origin implementation an overflow is handled by returning a
2134 * search header with a len of zero, so reset ret.
2135 */
2136 if (ret == -EOVERFLOW)
2137 ret = 0;
2138
ba346b35 2139 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2140 ret = -EFAULT;
ac8e9819
CM		 2141 return ret;
2142}
2143
cc68a8a5
GH		 2144static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2145 void __user *argp)
2146{
2147 struct btrfs_ioctl_search_args_v2 __user *uarg;
2148 struct btrfs_ioctl_search_args_v2 args;
2149 struct inode *inode;
2150 int ret;
2151 size_t buf_size;
2152 const size_t buf_limit = 16 * 1024 * 1024;
2153
2154 if (!capable(CAP_SYS_ADMIN))
2155 return -EPERM;
2156
2157 /* copy search header and buffer size */
2158 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2159 if (copy_from_user(&args, uarg, sizeof(args)))
2160 return -EFAULT;
2161
2162 buf_size = args.buf_size;
2163
2164 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2165 return -EOVERFLOW;
2166
2167 /* limit result size to 16MB */
2168 if (buf_size > buf_limit)
2169 buf_size = buf_limit;
2170
2171 inode = file_inode(file);
2172 ret = search_ioctl(inode, &args.key, &buf_size,
2173 (char *)(&uarg->buf[0]));
2174 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2175 ret = -EFAULT;
2176 else if (ret == -EOVERFLOW &&
2177 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2178 ret = -EFAULT;
2179
ac8e9819
CM		 2180 return ret;
2181}
2182
98d377a0 2183/*
ac8e9819
CM		 2184 * Search INODE_REFs to identify path name of 'dirid' directory
2185 * in a 'tree_id' tree. and sets path name to 'name'.
2186 */
98d377a0
TH		 2187static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2188 u64 tree_id, u64 dirid, char *name)
2189{
2190 struct btrfs_root *root;
2191 struct btrfs_key key;
ac8e9819 2192 char *ptr;
98d377a0
TH		 2193 int ret = -1;
2194 int slot;
2195 int len;
2196 int total_len = 0;
2197 struct btrfs_inode_ref *iref;
2198 struct extent_buffer *l;
2199 struct btrfs_path *path;
2200
2201 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2202 name[0]='\0';
2203 return 0;
2204 }
2205
2206 path = btrfs_alloc_path();
2207 if (!path)
2208 return -ENOMEM;
2209
ac8e9819 2210 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2211
2212 key.objectid = tree_id;
2213 key.type = BTRFS_ROOT_ITEM_KEY;
2214 key.offset = (u64)-1;
2215 root = btrfs_read_fs_root_no_name(info, &key);
2216 if (IS_ERR(root)) {
efe120a0 2217 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
8ad6fcab
CM		 2218 ret = -ENOENT;
2219 goto out;
98d377a0
TH		 2220 }
2221
2222 key.objectid = dirid;
2223 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2224 key.offset = (u64)-1;
98d377a0 2225
67871254 2226 while (1) {
98d377a0
TH		 2227 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2228 if (ret < 0)
2229 goto out;
18674c6c
FDBM		 2230 else if (ret > 0) {
2231 ret = btrfs_previous_item(root, path, dirid,
2232 BTRFS_INODE_REF_KEY);
2233 if (ret < 0)
2234 goto out;
2235 else if (ret > 0) {
2236 ret = -ENOENT;
2237 goto out;
2238 }
2239 }
98d377a0
TH		 2240
2241 l = path->nodes[0];
2242 slot = path->slots[0];
2243 btrfs_item_key_to_cpu(l, &key, slot);
2244
98d377a0
TH		 2245 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2246 len = btrfs_inode_ref_name_len(l, iref);
2247 ptr -= len + 1;
2248 total_len += len + 1;
a696cf35
FDBM		 2249 if (ptr < name) {
2250 ret = -ENAMETOOLONG;
98d377a0 2251 goto out;
a696cf35 2252 }
98d377a0
TH		 2253
2254 *(ptr + len) = '/';
67871254 2255 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2256
2257 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2258 break;
2259
b3b4aa74 2260 btrfs_release_path(path);
98d377a0 2261 key.objectid = key.offset;
8ad6fcab 2262 key.offset = (u64)-1;
98d377a0 2263 dirid = key.objectid;
98d377a0 2264 }
77906a50 2265 memmove(name, ptr, total_len);
67871254 2266 name[total_len] = '\0';
98d377a0
TH		 2267 ret = 0;
2268out:
2269 btrfs_free_path(path);
ac8e9819
CM		 2270 return ret;
2271}
2272
2273static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2274 void __user *argp)
2275{
2276 struct btrfs_ioctl_ino_lookup_args *args;
2277 struct inode *inode;
2278 int ret;
2279
2280 if (!capable(CAP_SYS_ADMIN))
2281 return -EPERM;
2282
2354d08f
JL		 2283 args = memdup_user(argp, sizeof(*args));
2284 if (IS_ERR(args))
2285 return PTR_ERR(args);
c2b96929 2286
496ad9aa 2287 inode = file_inode(file);
ac8e9819 2288
1b53ac4d
CM		 2289 if (args->treeid == 0)
2290 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2291
ac8e9819
CM		 2292 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2293 args->treeid, args->objectid,
2294 args->name);
2295
2296 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2297 ret = -EFAULT;
2298
2299 kfree(args);
98d377a0
TH		 2300 return ret;
2301}
2302
76dda93c
YZ		 2303static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2304 void __user *arg)
2305{
54563d41 2306 struct dentry *parent = file->f_path.dentry;
76dda93c
YZ		 2307 struct dentry *dentry;
2308 struct inode *dir = parent->d_inode;
2309 struct inode *inode;
2310 struct btrfs_root *root = BTRFS_I(dir)->root;
2311 struct btrfs_root *dest = NULL;
2312 struct btrfs_ioctl_vol_args *vol_args;
2313 struct btrfs_trans_handle *trans;
c58aaad2 2314 struct btrfs_block_rsv block_rsv;
521e0546 2315 u64 root_flags;
c58aaad2 2316 u64 qgroup_reserved;
76dda93c
YZ		 2317 int namelen;
2318 int ret;
2319 int err = 0;
2320
76dda93c
YZ		 2321 vol_args = memdup_user(arg, sizeof(*vol_args));
2322 if (IS_ERR(vol_args))
2323 return PTR_ERR(vol_args);
2324
2325 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2326 namelen = strlen(vol_args->name);
2327 if (strchr(vol_args->name, '/') ||
2328 strncmp(vol_args->name, "..", namelen) == 0) {
2329 err = -EINVAL;
2330 goto out;
2331 }
2332
a561be71 2333 err = mnt_want_write_file(file);
76dda93c
YZ		 2334 if (err)
2335 goto out;
2336
521e0546 2337
5c50c9b8
DS		 2338 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2339 if (err == -EINTR)
e43f998e 2340 goto out_drop_write;
76dda93c
YZ		 2341 dentry = lookup_one_len(vol_args->name, parent, namelen);
2342 if (IS_ERR(dentry)) {
2343 err = PTR_ERR(dentry);
2344 goto out_unlock_dir;
2345 }
2346
2347 if (!dentry->d_inode) {
2348 err = -ENOENT;
2349 goto out_dput;
2350 }
2351
2352 inode = dentry->d_inode;
4260f7c7 2353 dest = BTRFS_I(inode)->root;
67871254 2354 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2355 /*
2356 * Regular user. Only allow this with a special mount
2357 * option, when the user has write+exec access to the
2358 * subvol root, and when rmdir(2) would have been
2359 * allowed.
2360 *
2361 * Note that this is _not_ check that the subvol is
2362 * empty or doesn't contain data that we wouldn't
2363 * otherwise be able to delete.
2364 *
2365 * Users who want to delete empty subvols should try
2366 * rmdir(2).
2367 */
2368 err = -EPERM;
2369 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2370 goto out_dput;
2371
2372 /*
2373 * Do not allow deletion if the parent dir is the same
2374 * as the dir to be deleted. That means the ioctl
2375 * must be called on the dentry referencing the root
2376 * of the subvol, not a random directory contained
2377 * within it.
2378 */
2379 err = -EINVAL;
2380 if (root == dest)
2381 goto out_dput;
2382
2383 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2384 if (err)
2385 goto out_dput;
4260f7c7
SW		 2386 }
2387
5c39da5b
MX		 2388 /* check if subvolume may be deleted by a user */
2389 err = btrfs_may_delete(dir, dentry, 1);
2390 if (err)
2391 goto out_dput;
2392
33345d01 2393 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2394 err = -EINVAL;
2395 goto out_dput;
2396 }
2397
76dda93c 2398 mutex_lock(&inode->i_mutex);
521e0546
DS		 2399
2400 /*
2401 * Don't allow to delete a subvolume with send in progress. This is
2402 * inside the i_mutex so the error handling that has to drop the bit
2403 * again is not run concurrently.
2404 */
2405 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2406 root_flags = btrfs_root_flags(&dest->root_item);
2407 if (dest->send_in_progress == 0) {
2408 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2409 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2410 spin_unlock(&dest->root_item_lock);
2411 } else {
2412 spin_unlock(&dest->root_item_lock);
2413 btrfs_warn(root->fs_info,
2414 "Attempt to delete subvolume %llu during send",
c55bfa67 2415 dest->root_key.objectid);
521e0546
DS		 2416 err = -EPERM;
2417 goto out_dput;
2418 }
2419
76dda93c
YZ		 2420 err = d_invalidate(dentry);
2421 if (err)
2422 goto out_unlock;
2423
2424 down_write(&root->fs_info->subvol_sem);
2425
2426 err = may_destroy_subvol(dest);
2427 if (err)
2428 goto out_up_write;
2429
c58aaad2
MX		 2430 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2431 /*
2432 * One for dir inode, two for dir entries, two for root
2433 * ref/backref.
2434 */
2435 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2436 5, &qgroup_reserved, true);
c58aaad2
MX		 2437 if (err)
2438 goto out_up_write;
2439
a22285a6
YZ		 2440 trans = btrfs_start_transaction(root, 0);
2441 if (IS_ERR(trans)) {
2442 err = PTR_ERR(trans);
c58aaad2 2443 goto out_release;
a22285a6 2444 }
c58aaad2
MX		 2445 trans->block_rsv = &block_rsv;
2446 trans->bytes_reserved = block_rsv.size;
a22285a6 2447
76dda93c
YZ		 2448 ret = btrfs_unlink_subvol(trans, root, dir,
2449 dest->root_key.objectid,
2450 dentry->d_name.name,
2451 dentry->d_name.len);
79787eaa
JM		 2452 if (ret) {
2453 err = ret;
2454 btrfs_abort_transaction(trans, root, ret);
2455 goto out_end_trans;
2456 }
76dda93c
YZ		 2457
2458 btrfs_record_root_in_trans(trans, dest);
2459
2460 memset(&dest->root_item.drop_progress, 0,
2461 sizeof(dest->root_item.drop_progress));
2462 dest->root_item.drop_level = 0;
2463 btrfs_set_root_refs(&dest->root_item, 0);
2464
27cdeb70 2465 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
d68fc57b
YZ		 2466 ret = btrfs_insert_orphan_item(trans,
2467 root->fs_info->tree_root,
2468 dest->root_key.objectid);
79787eaa
JM		 2469 if (ret) {
2470 btrfs_abort_transaction(trans, root, ret);
2471 err = ret;
2472 goto out_end_trans;
2473 }
d68fc57b 2474 }
dd5f9615
SB		 2475
2476 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2477 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2478 dest->root_key.objectid);
2479 if (ret && ret != -ENOENT) {
2480 btrfs_abort_transaction(trans, root, ret);
2481 err = ret;
2482 goto out_end_trans;
2483 }
2484 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2485 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2486 dest->root_item.received_uuid,
2487 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2488 dest->root_key.objectid);
2489 if (ret && ret != -ENOENT) {
2490 btrfs_abort_transaction(trans, root, ret);
2491 err = ret;
2492 goto out_end_trans;
2493 }
2494 }
2495
79787eaa 2496out_end_trans:
c58aaad2
MX		 2497 trans->block_rsv = NULL;
2498 trans->bytes_reserved = 0;
531cb13f 2499 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2500 if (ret && !err)
2501 err = ret;
76dda93c 2502 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2503out_release:
2504 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2505out_up_write:
2506 up_write(&root->fs_info->subvol_sem);
2507out_unlock:
521e0546
DS		 2508 if (err) {
2509 spin_lock(&dest->root_item_lock);
c55bfa67
FM		 2510 root_flags = btrfs_root_flags(&dest->root_item);
2511 btrfs_set_root_flags(&dest->root_item,
521e0546
DS		 2512 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2513 spin_unlock(&dest->root_item_lock);
2514 }
76dda93c
YZ		 2515 mutex_unlock(&inode->i_mutex);
2516 if (!err) {
efefb143 2517 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2518 btrfs_invalidate_inodes(dest);
2519 d_delete(dentry);
61155aa0 2520 ASSERT(dest->send_in_progress == 0);
fa6ac876
LB		 2521
2522 /* the last ref */
57cdc8db
DS		 2523 if (dest->ino_cache_inode) {
2524 iput(dest->ino_cache_inode);
2525 dest->ino_cache_inode = NULL;
fa6ac876 2526 }
76dda93c
YZ		 2527 }
2528out_dput:
2529 dput(dentry);
2530out_unlock_dir:
2531 mutex_unlock(&dir->i_mutex);
e43f998e 2532out_drop_write:
2a79f17e 2533 mnt_drop_write_file(file);
76dda93c
YZ		 2534out:
2535 kfree(vol_args);
2536 return err;
2537}
2538
1e701a32 2539static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2540{
496ad9aa 2541 struct inode *inode = file_inode(file);
f46b5a66 2542 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2543 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2544 int ret;
2545
25122d15
ID		 2546 ret = mnt_want_write_file(file);
2547 if (ret)
2548 return ret;
b83cc969 2549
25122d15
ID		 2550 if (btrfs_root_readonly(root)) {
2551 ret = -EROFS;
2552 goto out;
5ac00add 2553 }
f46b5a66
CH		 2554
2555 switch (inode->i_mode & S_IFMT) {
2556 case S_IFDIR:
e441d54d
CM		 2557 if (!capable(CAP_SYS_ADMIN)) {
2558 ret = -EPERM;
2559 goto out;
2560 }
de78b51a 2561 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2562 if (ret)
2563 goto out;
de78b51a 2564 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2565 break;
2566 case S_IFREG:
e441d54d
CM		 2567 if (!(file->f_mode & FMODE_WRITE)) {
2568 ret = -EINVAL;
2569 goto out;
2570 }
1e701a32
CM		 2571
2572 range = kzalloc(sizeof(*range), GFP_KERNEL);
2573 if (!range) {
2574 ret = -ENOMEM;
2575 goto out;
2576 }
2577
2578 if (argp) {
2579 if (copy_from_user(range, argp,
2580 sizeof(*range))) {
2581 ret = -EFAULT;
2582 kfree(range);
683be16e 2583 goto out;
1e701a32
CM		 2584 }
2585 /* compression requires us to start the IO */
2586 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2587 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2588 range->extent_thresh = (u32)-1;
2589 }
2590 } else {
2591 /* the rest are all set to zero by kzalloc */
2592 range->len = (u64)-1;
2593 }
496ad9aa 2594 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2595 range, 0, 0);
2596 if (ret > 0)
2597 ret = 0;
1e701a32 2598 kfree(range);
f46b5a66 2599 break;
8929ecfa
YZ		 2600 default:
2601 ret = -EINVAL;
f46b5a66 2602 }
e441d54d 2603out:
25122d15 2604 mnt_drop_write_file(file);
e441d54d 2605 return ret;
f46b5a66
CH		 2606}
2607
b2950863 2608static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2609{
2610 struct btrfs_ioctl_vol_args *vol_args;
2611 int ret;
2612
e441d54d
CM		 2613 if (!capable(CAP_SYS_ADMIN))
2614 return -EPERM;
2615
5ac00add
SB		 2616 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2617 1)) {
e57138b3 2618 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 2619 }
2620
5ac00add 2621 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2622 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2623 if (IS_ERR(vol_args)) {
2624 ret = PTR_ERR(vol_args);
2625 goto out;
2626 }
f46b5a66 2627
5516e595 2628 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2629 ret = btrfs_init_new_device(root, vol_args->name);
2630
43d20761
AJ		 2631 if (!ret)
2632 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2633
f46b5a66 2634 kfree(vol_args);
c9e9f97b
ID		 2635out:
2636 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2637 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2638 return ret;
2639}
2640
da24927b 2641static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2642{
496ad9aa 2643 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2644 struct btrfs_ioctl_vol_args *vol_args;
2645 int ret;
2646
e441d54d
CM		 2647 if (!capable(CAP_SYS_ADMIN))
2648 return -EPERM;
2649
da24927b
MX		 2650 ret = mnt_want_write_file(file);
2651 if (ret)
2652 return ret;
c146afad 2653
dae7b665 2654 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2655 if (IS_ERR(vol_args)) {
2656 ret = PTR_ERR(vol_args);
c47ca32d 2657 goto err_drop;
c9e9f97b 2658 }
f46b5a66 2659
5516e595 2660 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2661
183860f6
AJ		 2662 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2663 1)) {
2664 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2665 goto out;
2666 }
2667
2668 mutex_lock(&root->fs_info->volume_mutex);
2669 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2670 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2671 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6 2672
ec95d491
AJ		 2673 if (!ret)
2674 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2675
183860f6
AJ		 2676out:
2677 kfree(vol_args);
c47ca32d 2678err_drop:
4ac20c70 2679 mnt_drop_write_file(file);
f46b5a66
CH		 2680 return ret;
2681}
2682
475f6387
JS		 2683static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2684{
027ed2f0 2685 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2686 struct btrfs_device *device;
2687 struct btrfs_device *next;
2688 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2689 int ret = 0;
475f6387 2690
027ed2f0
LZ		 2691 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2692 if (!fi_args)
2693 return -ENOMEM;
2694
f7171750 2695 mutex_lock(&fs_devices->device_list_mutex);
027ed2f0
LZ		 2696 fi_args->num_devices = fs_devices->num_devices;
2697 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387 2698
475f6387 2699 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2700 if (device->devid > fi_args->max_id)
2701 fi_args->max_id = device->devid;
475f6387
JS		 2702 }
2703 mutex_unlock(&fs_devices->device_list_mutex);
2704
80a773fb
DS		 2705 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2706 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2707 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2708
027ed2f0
LZ		 2709 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2710 ret = -EFAULT;
475f6387 2711
027ed2f0
LZ		 2712 kfree(fi_args);
2713 return ret;
475f6387
JS		 2714}
2715
2716static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2717{
2718 struct btrfs_ioctl_dev_info_args *di_args;
2719 struct btrfs_device *dev;
2720 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2721 int ret = 0;
2722 char *s_uuid = NULL;
475f6387 2723
475f6387
JS		 2724 di_args = memdup_user(arg, sizeof(*di_args));
2725 if (IS_ERR(di_args))
2726 return PTR_ERR(di_args);
2727
dd5f9615 2728 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2729 s_uuid = di_args->uuid;
2730
2731 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2732 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2733
2734 if (!dev) {
2735 ret = -ENODEV;
2736 goto out;
2737 }
2738
2739 di_args->devid = dev->devid;
2740 di_args->bytes_used = dev->bytes_used;
2741 di_args->total_bytes = dev->total_bytes;
2742 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2743 if (dev->name) {
606686ee
JB		 2744 struct rcu_string *name;
2745
2746 rcu_read_lock();
2747 name = rcu_dereference(dev->name);
2748 strncpy(di_args->path, name->str, sizeof(di_args->path));
2749 rcu_read_unlock();
a27202fb
JM		 2750 di_args->path[sizeof(di_args->path) - 1] = 0;
2751 } else {
99ba55ad 2752 di_args->path[0] = '\0';
a27202fb 2753 }
475f6387
JS		 2754
2755out:
55793c0d 2756 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2757 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2758 ret = -EFAULT;
2759
2760 kfree(di_args);
2761 return ret;
2762}
2763
416161db
MF		 2764static struct page *extent_same_get_page(struct inode *inode, u64 off)
2765{
2766 struct page *page;
2767 pgoff_t index;
2768 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2769
2770 index = off >> PAGE_CACHE_SHIFT;
2771
2772 page = grab_cache_page(inode->i_mapping, index);
2773 if (!page)
2774 return NULL;
2775
2776 if (!PageUptodate(page)) {
2777 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2778 0))
2779 return NULL;
2780 lock_page(page);
2781 if (!PageUptodate(page)) {
2782 unlock_page(page);
2783 page_cache_release(page);
2784 return NULL;
2785 }
2786 }
2787 unlock_page(page);
2788
2789 return page;
2790}
2791
77fe20dc
MF		 2792static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2793{
2794 /* do any pending delalloc/csum calc on src, one way or
2795 another, and lock file content */
2796 while (1) {
2797 struct btrfs_ordered_extent *ordered;
2798 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2799 ordered = btrfs_lookup_first_ordered_extent(inode,
2800 off + len - 1);
ff5df9b8
FM		 2801 if ((!ordered ||
2802 ordered->file_offset + ordered->len <= off ||
2803 ordered->file_offset >= off + len) &&
77fe20dc 2804 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
ff5df9b8
FM		 2805 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2806 if (ordered)
2807 btrfs_put_ordered_extent(ordered);
77fe20dc 2808 break;
ff5df9b8 2809 }
77fe20dc
MF		 2810 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2811 if (ordered)
2812 btrfs_put_ordered_extent(ordered);
2813 btrfs_wait_ordered_range(inode, off, len);
2814 }
2815}
2816
416161db
MF		 2817static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2818 struct inode *inode2, u64 loff2, u64 len)
2819{
2820 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2821 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2822
2823 mutex_unlock(&inode1->i_mutex);
2824 mutex_unlock(&inode2->i_mutex);
2825}
2826
2827static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2828 struct inode *inode2, u64 loff2, u64 len)
2829{
2830 if (inode1 < inode2) {
2831 swap(inode1, inode2);
2832 swap(loff1, loff2);
2833 }
2834
2835 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2836 lock_extent_range(inode1, loff1, len);
2837 if (inode1 != inode2) {
2838 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2839 lock_extent_range(inode2, loff2, len);
2840 }
2841}
2842
2843static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2844 u64 dst_loff, u64 len)
2845{
2846 int ret = 0;
2847 struct page *src_page, *dst_page;
2848 unsigned int cmp_len = PAGE_CACHE_SIZE;
2849 void *addr, *dst_addr;
2850
2851 while (len) {
2852 if (len < PAGE_CACHE_SIZE)
2853 cmp_len = len;
2854
2855 src_page = extent_same_get_page(src, loff);
2856 if (!src_page)
2857 return -EINVAL;
2858 dst_page = extent_same_get_page(dst, dst_loff);
2859 if (!dst_page) {
2860 page_cache_release(src_page);
2861 return -EINVAL;
2862 }
2863 addr = kmap_atomic(src_page);
2864 dst_addr = kmap_atomic(dst_page);
2865
2866 flush_dcache_page(src_page);
2867 flush_dcache_page(dst_page);
2868
2869 if (memcmp(addr, dst_addr, cmp_len))
2870 ret = BTRFS_SAME_DATA_DIFFERS;
2871
2872 kunmap_atomic(addr);
2873 kunmap_atomic(dst_addr);
2874 page_cache_release(src_page);
2875 page_cache_release(dst_page);
2876
2877 if (ret)
2878 break;
2879
2880 loff += cmp_len;
2881 dst_loff += cmp_len;
2882 len -= cmp_len;
2883 }
2884
2885 return ret;
2886}
2887
2888static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2889{
2890 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2891
2892 if (off + len > inode->i_size || off + len < off)
2893 return -EINVAL;
2894 /* Check that we are block aligned - btrfs_clone() requires this */
2895 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2896 return -EINVAL;
2897
2898 return 0;
2899}
2900
2901static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2902 struct inode *dst, u64 dst_loff)
2903{
2904 int ret;
2905
2906 /*
2907 * btrfs_clone() can't handle extents in the same file
2908 * yet. Once that works, we can drop this check and replace it
2909 * with a check for the same inode, but overlapping extents.
2910 */
2911 if (src == dst)
2912 return -EINVAL;
2913
2914 btrfs_double_lock(src, loff, dst, dst_loff, len);
2915
2916 ret = extent_same_check_offsets(src, loff, len);
2917 if (ret)
2918 goto out_unlock;
2919
2920 ret = extent_same_check_offsets(dst, dst_loff, len);
2921 if (ret)
2922 goto out_unlock;
2923
2924 /* don't make the dst file partly checksummed */
2925 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2926 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2927 ret = -EINVAL;
2928 goto out_unlock;
2929 }
2930
2931 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2932 if (ret == 0)
2933 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2934
2935out_unlock:
2936 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2937
2938 return ret;
2939}
2940
2941#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2942
2943static long btrfs_ioctl_file_extent_same(struct file *file,
1c1c8747 2944 struct btrfs_ioctl_same_args __user *argp)
416161db 2945{
cbf8b8ca
MF		 2946 struct btrfs_ioctl_same_args *same;
2947 struct btrfs_ioctl_same_extent_info *info;
1c1c8747 2948 struct inode *src = file_inode(file);
416161db
MF		 2949 u64 off;
2950 u64 len;
2951 int i;
2952 int ret;
cbf8b8ca 2953 unsigned long size;
416161db
MF		 2954 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2955 bool is_admin = capable(CAP_SYS_ADMIN);
1c1c8747 2956 u16 count;
416161db
MF		 2957
2958 if (!(file->f_mode & FMODE_READ))
2959 return -EINVAL;
2960
2961 ret = mnt_want_write_file(file);
2962 if (ret)
2963 return ret;
2964
1c1c8747 2965 if (get_user(count, &argp->dest_count)) {
416161db
MF		 2966 ret = -EFAULT;
2967 goto out;
2968 }
2969
1c1c8747 2970 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
cbf8b8ca 2971
1c1c8747 2972 same = memdup_user(argp, size);
cbf8b8ca 2973
229eed43
GB		 2974 if (IS_ERR(same)) {
2975 ret = PTR_ERR(same);
cbf8b8ca
MF		 2976 goto out;
2977 }
2978
2979 off = same->logical_offset;
2980 len = same->length;
416161db
MF		 2981
2982 /*
2983 * Limit the total length we will dedupe for each operation.
2984 * This is intended to bound the total time spent in this
2985 * ioctl to something sane.
2986 */
2987 if (len > BTRFS_MAX_DEDUPE_LEN)
2988 len = BTRFS_MAX_DEDUPE_LEN;
2989
2990 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2991 /*
2992 * Btrfs does not support blocksize < page_size. As a
2993 * result, btrfs_cmp_data() won't correctly handle
2994 * this situation without an update.
2995 */
2996 ret = -EINVAL;
2997 goto out;
2998 }
2999
3000 ret = -EISDIR;
3001 if (S_ISDIR(src->i_mode))
3002 goto out;
3003
3004 ret = -EACCES;
3005 if (!S_ISREG(src->i_mode))
3006 goto out;
3007
cbf8b8ca 3008 /* pre-format output fields to sane values */
1c1c8747 3009 for (i = 0; i < count; i++) {
cbf8b8ca
MF		 3010 same->info[i].bytes_deduped = 0ULL;
3011 same->info[i].status = 0;
3012 }
416161db 3013
1c1c8747
AV		 3014 for (i = 0, info = same->info; i < count; i++, info++) {
3015 struct inode *dst;
3016 struct fd dst_file = fdget(info->fd);
3017 if (!dst_file.file) {
cbf8b8ca 3018 info->status = -EBADF;
1c1c8747 3019 continue;
416161db 3020 }
1c1c8747 3021 dst = file_inode(dst_file.file);
416161db 3022
1c1c8747 3023 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
cbf8b8ca 3024 info->status = -EINVAL;
1c1c8747
AV		 3025 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3026 info->status = -EXDEV;
3027 } else if (S_ISDIR(dst->i_mode)) {
cbf8b8ca 3028 info->status = -EISDIR;
1c1c8747 3029 } else if (!S_ISREG(dst->i_mode)) {
cbf8b8ca 3030 info->status = -EACCES;
1c1c8747
AV		 3031 } else {
3032 info->status = btrfs_extent_same(src, off, len, dst,
3033 info->logical_offset);
3034 if (info->status == 0)
3035 info->bytes_deduped += len;
416161db 3036 }
1c1c8747 3037 fdput(dst_file);
416161db
MF		 3038 }
3039
cbf8b8ca
MF		 3040 ret = copy_to_user(argp, same, size);
3041 if (ret)
3042 ret = -EFAULT;
3043
416161db
MF		 3044out:
3045 mnt_drop_write_file(file);
3046 return ret;
3047}
3048
fcebe456
JB		 3049/* Helper to check and see if this root currently has a ref on the given disk
3050 * bytenr. If it does then we need to update the quota for this root. This
3051 * doesn't do anything if quotas aren't enabled.
3052 */
3053static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3054 u64 disko)
3055{
3056 struct seq_list tree_mod_seq_elem = {};
3057 struct ulist *roots;
3058 struct ulist_iterator uiter;
3059 struct ulist_node *root_node = NULL;
3060 int ret;
3061
3062 if (!root->fs_info->quota_enabled)
3063 return 1;
3064
3065 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3066 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3067 tree_mod_seq_elem.seq, &roots);
3068 if (ret < 0)
3069 goto out;
3070 ret = 0;
3071 ULIST_ITER_INIT(&uiter);
3072 while ((root_node = ulist_next(roots, &uiter))) {
3073 if (root_node->val == root->objectid) {
3074 ret = 1;
3075 break;
3076 }
3077 }
3078 ulist_free(roots);
3079out:
3080 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3081 return ret;
3082}
3083
f82a9901
FM		 3084static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3085 struct inode *inode,
3086 u64 endoff,
3087 const u64 destoff,
3088 const u64 olen)
3089{
3090 struct btrfs_root *root = BTRFS_I(inode)->root;
3091 int ret;
3092
3093 inode_inc_iversion(inode);
3094 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3095 /*
3096 * We round up to the block size at eof when determining which
3097 * extents to clone above, but shouldn't round up the file size.
3098 */
3099 if (endoff > destoff + olen)
3100 endoff = destoff + olen;
3101 if (endoff > inode->i_size)
3102 btrfs_i_size_write(inode, endoff);
3103
3104 ret = btrfs_update_inode(trans, root, inode);
3105 if (ret) {
3106 btrfs_abort_transaction(trans, root, ret);
3107 btrfs_end_transaction(trans, root);
3108 goto out;
3109 }
3110 ret = btrfs_end_transaction(trans, root);
3111out:
3112 return ret;
3113}
3114
7ffbb598
FM		 3115static void clone_update_extent_map(struct inode *inode,
3116 const struct btrfs_trans_handle *trans,
3117 const struct btrfs_path *path,
7ffbb598
FM		 3118 const u64 hole_offset,
3119 const u64 hole_len)
3120{
3121 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3122 struct extent_map *em;
3123 int ret;
3124
3125 em = alloc_extent_map();
3126 if (!em) {
3127 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3128 &BTRFS_I(inode)->runtime_flags);
3129 return;
3130 }
3131
14f59796
FM		 3132 if (path) {
3133 struct btrfs_file_extent_item *fi;
3134
3135 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3136 struct btrfs_file_extent_item);
7ffbb598
FM		 3137 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3138 em->generation = -1;
3139 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3140 BTRFS_FILE_EXTENT_INLINE)
3141 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3142 &BTRFS_I(inode)->runtime_flags);
3143 } else {
3144 em->start = hole_offset;
3145 em->len = hole_len;
3146 em->ram_bytes = em->len;
3147 em->orig_start = hole_offset;
3148 em->block_start = EXTENT_MAP_HOLE;
3149 em->block_len = 0;
3150 em->orig_block_len = 0;
3151 em->compress_type = BTRFS_COMPRESS_NONE;
3152 em->generation = trans->transid;
3153 }
3154
3155 while (1) {
3156 write_lock(&em_tree->lock);
3157 ret = add_extent_mapping(em_tree, em, 1);
3158 write_unlock(&em_tree->lock);
3159 if (ret != -EEXIST) {
3160 free_extent_map(em);
3161 break;
3162 }
3163 btrfs_drop_extent_cache(inode, em->start,
3164 em->start + em->len - 1, 0);
3165 }
3166
3167 if (unlikely(ret))
3168 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3169 &BTRFS_I(inode)->runtime_flags);
3170}
3171
32b7c687
MF		 3172/**
3173 * btrfs_clone() - clone a range from inode file to another
3174 *
3175 * @src: Inode to clone from
3176 * @inode: Inode to clone to
3177 * @off: Offset within source to start clone from
3178 * @olen: Original length, passed by user, of range to clone
3179 * @olen_aligned: Block-aligned value of olen, extent_same uses
3180 * identical values here
3181 * @destoff: Offset within @inode to start clone
3182 */
3183static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901
FM		 3184 const u64 off, const u64 olen, const u64 olen_aligned,
3185 const u64 destoff)
f46b5a66 3186{
f46b5a66 3187 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3188 struct btrfs_path *path = NULL;
f46b5a66 3189 struct extent_buffer *leaf;
32b7c687
MF		 3190 struct btrfs_trans_handle *trans;
3191 char *buf = NULL;
ae01a0ab 3192 struct btrfs_key key;
f46b5a66
CH		 3193 u32 nritems;
3194 int slot;
ae01a0ab 3195 int ret;
fcebe456 3196 int no_quota;
f82a9901 3197 const u64 len = olen_aligned;
fcebe456 3198 u64 last_disko = 0;
f82a9901 3199 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3200
3201 ret = -ENOMEM;
707e8a07 3202 buf = vmalloc(root->nodesize);
ae01a0ab 3203 if (!buf)
32b7c687 3204 return ret;
ae01a0ab
YZ		 3205
3206 path = btrfs_alloc_path();
3207 if (!path) {
3208 vfree(buf);
32b7c687 3209 return ret;
d525e8ab
LZ		 3210 }
3211
32b7c687 3212 path->reada = 2;
c5c9cd4d 3213 /* clone data */
33345d01 3214 key.objectid = btrfs_ino(src);
ae01a0ab 3215 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3216 key.offset = off;
f46b5a66
CH		 3217
3218 while (1) {
3219 /*
3220 * note the key will change type as we walk through the
3221 * tree.
3222 */
e4355f34 3223 path->leave_spinning = 1;
362a20c5
DS		 3224 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3225 0, 0);
f46b5a66
CH		 3226 if (ret < 0)
3227 goto out;
2c463823
FM		 3228 /*
3229 * First search, if no extent item that starts at offset off was
3230 * found but the previous item is an extent item, it's possible
3231 * it might overlap our target range, therefore process it.
3232 */
3233 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3234 btrfs_item_key_to_cpu(path->nodes[0], &key,
3235 path->slots[0] - 1);
3236 if (key.type == BTRFS_EXTENT_DATA_KEY)
3237 path->slots[0]--;
3238 }
f46b5a66 3239
ae01a0ab 3240 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3241process_slot:
fcebe456 3242 no_quota = 1;
ae01a0ab 3243 if (path->slots[0] >= nritems) {
362a20c5 3244 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3245 if (ret < 0)
3246 goto out;
3247 if (ret > 0)
3248 break;
ae01a0ab 3249 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3250 }
3251 leaf = path->nodes[0];
3252 slot = path->slots[0];
f46b5a66 3253
ae01a0ab 3254 btrfs_item_key_to_cpu(leaf, &key, slot);
962a298f 3255 if (key.type > BTRFS_EXTENT_DATA_KEY ||
33345d01 3256 key.objectid != btrfs_ino(src))
f46b5a66
CH		 3257 break;
3258
962a298f 3259 if (key.type == BTRFS_EXTENT_DATA_KEY) {
c5c9cd4d
SW		 3260 struct btrfs_file_extent_item *extent;
3261 int type;
31840ae1
ZY		 3262 u32 size;
3263 struct btrfs_key new_key;
c5c9cd4d
SW		 3264 u64 disko = 0, diskl = 0;
3265 u64 datao = 0, datal = 0;
3266 u8 comp;
f82a9901 3267 u64 drop_start;
31840ae1 3268
c5c9cd4d
SW		 3269 extent = btrfs_item_ptr(leaf, slot,
3270 struct btrfs_file_extent_item);
3271 comp = btrfs_file_extent_compression(leaf, extent);
3272 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3273 if (type == BTRFS_FILE_EXTENT_REG ||
3274 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3275 disko = btrfs_file_extent_disk_bytenr(leaf,
3276 extent);
3277 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3278 extent);
c5c9cd4d 3279 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3280 datal = btrfs_file_extent_num_bytes(leaf,
3281 extent);
c5c9cd4d
SW		 3282 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3283 /* take upper bound, may be compressed */
3284 datal = btrfs_file_extent_ram_bytes(leaf,
3285 extent);
3286 }
31840ae1 3287
2c463823
FM		 3288 /*
3289 * The first search might have left us at an extent
3290 * item that ends before our target range's start, can
3291 * happen if we have holes and NO_HOLES feature enabled.
3292 */
3293 if (key.offset + datal <= off) {
e4355f34
FDBM		 3294 path->slots[0]++;
3295 goto process_slot;
2c463823
FM		 3296 } else if (key.offset >= off + len) {
3297 break;
e4355f34
FDBM		 3298 }
3299
3300 size = btrfs_item_size_nr(leaf, slot);
3301 read_extent_buffer(leaf, buf,
3302 btrfs_item_ptr_offset(leaf, slot),
3303 size);
3304
3305 btrfs_release_path(path);
3306 path->leave_spinning = 0;
c5c9cd4d 3307
31840ae1 3308 memcpy(&new_key, &key, sizeof(new_key));
33345d01 3309 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 3310 if (off <= key.offset)
3311 new_key.offset = key.offset + destoff - off;
3312 else
3313 new_key.offset = destoff;
31840ae1 3314
f82a9901
FM		 3315 /*
3316 * Deal with a hole that doesn't have an extent item
3317 * that represents it (NO_HOLES feature enabled).
3318 * This hole is either in the middle of the cloning
3319 * range or at the beginning (fully overlaps it or
3320 * partially overlaps it).
3321 */
3322 if (new_key.offset != last_dest_end)
3323 drop_start = last_dest_end;
3324 else
3325 drop_start = new_key.offset;
3326
b6f3409b
SW		 3327 /*
3328 * 1 - adjusting old extent (we may have to split it)
3329 * 1 - add new extent
3330 * 1 - inode update
3331 */
3332 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3333 if (IS_ERR(trans)) {
3334 ret = PTR_ERR(trans);
3335 goto out;
3336 }
3337
c8a894d7
CM		 3338 if (type == BTRFS_FILE_EXTENT_REG ||
3339 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3340 /*
3341 * a | --- range to clone ---| b
3342 * | ------------- extent ------------- |
3343 */
3344
93915584 3345 /* subtract range b */
d72c0842
LZ		 3346 if (key.offset + datal > off + len)
3347 datal = off + len - key.offset;
3348
93915584 3349 /* subtract range a */
a22285a6
YZ		 3350 if (off > key.offset) {
3351 datao += off - key.offset;
3352 datal -= off - key.offset;
3353 }
3354
5dc562c5 3355 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3356 drop_start,
a22285a6 3357 new_key.offset + datal,
2671485d 3358 1);
79787eaa 3359 if (ret) {
3f9e3df8 3360 if (ret != -EOPNOTSUPP)
00fdf13a
LB		 3361 btrfs_abort_transaction(trans,
3362 root, ret);
79787eaa
JM		 3363 btrfs_end_transaction(trans, root);
3364 goto out;
3365 }
a22285a6 3366
c5c9cd4d
SW		 3367 ret = btrfs_insert_empty_item(trans, root, path,
3368 &new_key, size);
79787eaa
JM		 3369 if (ret) {
3370 btrfs_abort_transaction(trans, root,
3371 ret);
3372 btrfs_end_transaction(trans, root);
3373 goto out;
3374 }
c5c9cd4d
SW		 3375
3376 leaf = path->nodes[0];
3377 slot = path->slots[0];
3378 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3379 btrfs_item_ptr_offset(leaf, slot),
3380 size);
ae01a0ab 3381
c5c9cd4d 3382 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3383 struct btrfs_file_extent_item);
c5c9cd4d 3384
c5c9cd4d
SW		 3385 /* disko == 0 means it's a hole */
3386 if (!disko)
3387 datao = 0;
c5c9cd4d
SW		 3388
3389 btrfs_set_file_extent_offset(leaf, extent,
3390 datao);
3391 btrfs_set_file_extent_num_bytes(leaf, extent,
3392 datal);
fcebe456
JB		 3393
3394 /*
3395 * We need to look up the roots that point at
3396 * this bytenr and see if the new root does. If
3397 * it does not we need to make sure we update
3398 * quotas appropriately.
3399 */
3400 if (disko && root != BTRFS_I(src)->root &&
3401 disko != last_disko) {
3402 no_quota = check_ref(trans, root,
3403 disko);
3404 if (no_quota < 0) {
3405 btrfs_abort_transaction(trans,
3406 root,
3407 ret);
3408 btrfs_end_transaction(trans,
3409 root);
3410 ret = no_quota;
3411 goto out;
3412 }
3413 }
3414
c5c9cd4d
SW		 3415 if (disko) {
3416 inode_add_bytes(inode, datal);
ae01a0ab 3417 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3418 disko, diskl, 0,
3419 root->root_key.objectid,
33345d01 3420 btrfs_ino(inode),
66d7e7f0 3421 new_key.offset - datao,
fcebe456 3422 no_quota);
79787eaa
JM		 3423 if (ret) {
3424 btrfs_abort_transaction(trans,
3425 root,
3426 ret);
3427 btrfs_end_transaction(trans,
3428 root);
3429 goto out;
3430
3431 }
f46b5a66 3432 }
c5c9cd4d
SW		 3433 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3434 u64 skip = 0;
3435 u64 trim = 0;
d3ecfcdf
LB		 3436 u64 aligned_end = 0;
3437
c5c9cd4d
SW		 3438 if (off > key.offset) {
3439 skip = off - key.offset;
3440 new_key.offset += skip;
3441 }
d397712b 3442
aa42ffd9
LB		 3443 if (key.offset + datal > off + len)
3444 trim = key.offset + datal - (off + len);
d397712b 3445
c5c9cd4d 3446 if (comp && (skip || trim)) {
c5c9cd4d 3447 ret = -EINVAL;
a22285a6 3448 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3449 goto out;
3450 }
3451 size -= skip + trim;
3452 datal -= skip + trim;
a22285a6 3453
d3ecfcdf
LB		 3454 aligned_end = ALIGN(new_key.offset + datal,
3455 root->sectorsize);
5dc562c5 3456 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3457 drop_start,
d3ecfcdf 3458 aligned_end,
2671485d 3459 1);
79787eaa 3460 if (ret) {
3f9e3df8 3461 if (ret != -EOPNOTSUPP)
3a29bc09
CM		 3462 btrfs_abort_transaction(trans,
3463 root, ret);
79787eaa
JM		 3464 btrfs_end_transaction(trans, root);
3465 goto out;
3466 }
a22285a6 3467
c5c9cd4d
SW		 3468 ret = btrfs_insert_empty_item(trans, root, path,
3469 &new_key, size);
79787eaa
JM		 3470 if (ret) {
3471 btrfs_abort_transaction(trans, root,
3472 ret);
3473 btrfs_end_transaction(trans, root);
3474 goto out;
3475 }
c5c9cd4d
SW		 3476
3477 if (skip) {
d397712b
CM		 3478 u32 start =
3479 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3480 memmove(buf+start, buf+start+skip,
3481 datal);
3482 }
3483
3484 leaf = path->nodes[0];
3485 slot = path->slots[0];
3486 write_extent_buffer(leaf, buf,
3487 btrfs_item_ptr_offset(leaf, slot),
3488 size);
3489 inode_add_bytes(inode, datal);
f46b5a66 3490 }
c5c9cd4d 3491
7ffbb598
FM		 3492 /* If we have an implicit hole (NO_HOLES feature). */
3493 if (drop_start < new_key.offset)
3494 clone_update_extent_map(inode, trans,
14f59796 3495 NULL, drop_start,
7ffbb598
FM		 3496 new_key.offset - drop_start);
3497
14f59796 3498 clone_update_extent_map(inode, trans, path, 0, 0);
7ffbb598 3499
c5c9cd4d 3500 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3501 btrfs_release_path(path);
c5c9cd4d 3502
62e2390e
FM		 3503 last_dest_end = ALIGN(new_key.offset + datal,
3504 root->sectorsize);
f82a9901
FM		 3505 ret = clone_finish_inode_update(trans, inode,
3506 last_dest_end,
3507 destoff, olen);
3508 if (ret)
79787eaa 3509 goto out;
2c463823
FM		 3510 if (new_key.offset + datal >= destoff + len)
3511 break;
a22285a6 3512 }
b3b4aa74 3513 btrfs_release_path(path);
f46b5a66 3514 key.offset++;
f46b5a66 3515 }
f46b5a66 3516 ret = 0;
32b7c687 3517
f82a9901
FM		 3518 if (last_dest_end < destoff + len) {
3519 /*
3520 * We have an implicit hole (NO_HOLES feature is enabled) that
3521 * fully or partially overlaps our cloning range at its end.
3522 */
3523 btrfs_release_path(path);
3524
3525 /*
3526 * 1 - remove extent(s)
3527 * 1 - inode update
3528 */
3529 trans = btrfs_start_transaction(root, 2);
3530 if (IS_ERR(trans)) {
3531 ret = PTR_ERR(trans);
3532 goto out;
3533 }
3534 ret = btrfs_drop_extents(trans, root, inode,
3535 last_dest_end, destoff + len, 1);
3536 if (ret) {
3537 if (ret != -EOPNOTSUPP)
3538 btrfs_abort_transaction(trans, root, ret);
3539 btrfs_end_transaction(trans, root);
3540 goto out;
3541 }
14f59796
FM		 3542 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3543 destoff + len - last_dest_end);
f82a9901
FM		 3544 ret = clone_finish_inode_update(trans, inode, destoff + len,
3545 destoff, olen);
3546 }
3547
f46b5a66 3548out:
32b7c687
MF		 3549 btrfs_free_path(path);
3550 vfree(buf);
3551 return ret;
3552}
3553
3554static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3555 u64 off, u64 olen, u64 destoff)
3556{
54563d41 3557 struct inode *inode = file_inode(file);
32b7c687
MF		 3558 struct btrfs_root *root = BTRFS_I(inode)->root;
3559 struct fd src_file;
3560 struct inode *src;
3561 int ret;
3562 u64 len = olen;
3563 u64 bs = root->fs_info->sb->s_blocksize;
3564 int same_inode = 0;
3565
3566 /*
3567 * TODO:
3568 * - split compressed inline extents. annoying: we need to
3569 * decompress into destination's address_space (the file offset
3570 * may change, so source mapping won't do), then recompress (or
3571 * otherwise reinsert) a subrange.
00fdf13a
LB		 3572 *
3573 * - split destination inode's inline extents. The inline extents can
3574 * be either compressed or non-compressed.
32b7c687
MF		 3575 */
3576
3577 /* the destination must be opened for writing */
3578 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3579 return -EINVAL;
3580
3581 if (btrfs_root_readonly(root))
3582 return -EROFS;
3583
3584 ret = mnt_want_write_file(file);
3585 if (ret)
3586 return ret;
3587
3588 src_file = fdget(srcfd);
3589 if (!src_file.file) {
3590 ret = -EBADF;
3591 goto out_drop_write;
3592 }
3593
3594 ret = -EXDEV;
3595 if (src_file.file->f_path.mnt != file->f_path.mnt)
3596 goto out_fput;
3597
3598 src = file_inode(src_file.file);
3599
3600 ret = -EINVAL;
3601 if (src == inode)
3602 same_inode = 1;
3603
3604 /* the src must be open for reading */
3605 if (!(src_file.file->f_mode & FMODE_READ))
3606 goto out_fput;
3607
3608 /* don't make the dst file partly checksummed */
3609 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3610 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3611 goto out_fput;
3612
3613 ret = -EISDIR;
3614 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3615 goto out_fput;
3616
3617 ret = -EXDEV;
3618 if (src->i_sb != inode->i_sb)
3619 goto out_fput;
3620
3621 if (!same_inode) {
3622 if (inode < src) {
3623 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3624 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3625 } else {
3626 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3627 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3628 }
3629 } else {
3630 mutex_lock(&src->i_mutex);
3631 }
3632
3633 /* determine range to clone */
3634 ret = -EINVAL;
3635 if (off + len > src->i_size || off + len < off)
3636 goto out_unlock;
3637 if (len == 0)
3638 olen = len = src->i_size - off;
3639 /* if we extend to eof, continue to block boundary */
3640 if (off + len == src->i_size)
3641 len = ALIGN(src->i_size, bs) - off;
3642
3643 /* verify the end result is block aligned */
3644 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3645 !IS_ALIGNED(destoff, bs))
3646 goto out_unlock;
3647
3648 /* verify if ranges are overlapped within the same file */
3649 if (same_inode) {
3650 if (destoff + len > off && destoff < off + len)
3651 goto out_unlock;
3652 }
3653
3654 if (destoff > inode->i_size) {
3655 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3656 if (ret)
3657 goto out_unlock;
3658 }
3659
c125b8bf
FM		 3660 /*
3661 * Lock the target range too. Right after we replace the file extent
3662 * items in the fs tree (which now point to the cloned data), we might
3663 * have a worker replace them with extent items relative to a write
3664 * operation that was issued before this clone operation (i.e. confront
3665 * with inode.c:btrfs_finish_ordered_io).
3666 */
3667 if (same_inode) {
3668 u64 lock_start = min_t(u64, off, destoff);
3669 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
32b7c687 3670
c125b8bf
FM		 3671 lock_extent_range(src, lock_start, lock_len);
3672 } else {
3673 lock_extent_range(src, off, len);
3674 lock_extent_range(inode, destoff, len);
3675 }
32b7c687
MF		 3676
3677 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3678
c125b8bf
FM		 3679 if (same_inode) {
3680 u64 lock_start = min_t(u64, off, destoff);
3681 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3682
3683 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3684 } else {
3685 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3686 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3687 destoff + len - 1);
3688 }
3689 /*
3690 * Truncate page cache pages so that future reads will see the cloned
3691 * data immediately and not the previous data.
3692 */
3693 truncate_inode_pages_range(&inode->i_data, destoff,
3694 PAGE_CACHE_ALIGN(destoff + len) - 1);
f46b5a66 3695out_unlock:
c57c2b3e
FDBM		 3696 if (!same_inode) {
3697 if (inode < src) {
3698 mutex_unlock(&src->i_mutex);
3699 mutex_unlock(&inode->i_mutex);
3700 } else {
3701 mutex_unlock(&inode->i_mutex);
3702 mutex_unlock(&src->i_mutex);
3703 }
3704 } else {
3705 mutex_unlock(&src->i_mutex);
3706 }
f46b5a66 3707out_fput:
2903ff01 3708 fdput(src_file);
ab67b7c1 3709out_drop_write:
2a79f17e 3710 mnt_drop_write_file(file);
f46b5a66
CH		 3711 return ret;
3712}
3713
7a865e8a 3714static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3715{
3716 struct btrfs_ioctl_clone_range_args args;
3717
7a865e8a 3718 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3719 return -EFAULT;
3720 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3721 args.src_length, args.dest_offset);
3722}
3723
f46b5a66
CH		 3724/*
3725 * there are many ways the trans_start and trans_end ioctls can lead
3726 * to deadlocks. They should only be used by applications that
3727 * basically own the machine, and have a very in depth understanding
3728 * of all the possible deadlocks and enospc problems.
3729 */
b2950863 3730static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3731{
496ad9aa 3732 struct inode *inode = file_inode(file);
f46b5a66
CH		 3733 struct btrfs_root *root = BTRFS_I(inode)->root;
3734 struct btrfs_trans_handle *trans;
1ab86aed 3735 int ret;
f46b5a66 3736
1ab86aed 3737 ret = -EPERM;
df5b5520 3738 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3739 goto out;
df5b5520 3740
1ab86aed
SW		 3741 ret = -EINPROGRESS;
3742 if (file->private_data)
f46b5a66 3743 goto out;
9ca9ee09 3744
b83cc969
LZ		 3745 ret = -EROFS;
3746 if (btrfs_root_readonly(root))
3747 goto out;
3748
a561be71 3749 ret = mnt_want_write_file(file);
c146afad
YZ		 3750 if (ret)
3751 goto out;
3752
a4abeea4 3753 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3754
1ab86aed 3755 ret = -ENOMEM;
7a7eaa40 3756 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3757 if (IS_ERR(trans))
1ab86aed
SW		 3758 goto out_drop;
3759
3760 file->private_data = trans;
3761 return 0;
3762
3763out_drop:
a4abeea4 3764 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3765 mnt_drop_write_file(file);
f46b5a66 3766out:
f46b5a66
CH		 3767 return ret;
3768}
3769
6ef5ed0d
JB		 3770static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3771{
496ad9aa 3772 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3773 struct btrfs_root *root = BTRFS_I(inode)->root;
3774 struct btrfs_root *new_root;
3775 struct btrfs_dir_item *di;
3776 struct btrfs_trans_handle *trans;
3777 struct btrfs_path *path;
3778 struct btrfs_key location;
3779 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3780 u64 objectid = 0;
3781 u64 dir_id;
3c04ce01 3782 int ret;
6ef5ed0d
JB		 3783
3784 if (!capable(CAP_SYS_ADMIN))
3785 return -EPERM;
3786
3c04ce01
MX		 3787 ret = mnt_want_write_file(file);
3788 if (ret)
3789 return ret;
3790
3791 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3792 ret = -EFAULT;
3793 goto out;
3794 }
6ef5ed0d
JB		 3795
3796 if (!objectid)
1cecf579 3797 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3798
3799 location.objectid = objectid;
3800 location.type = BTRFS_ROOT_ITEM_KEY;
3801 location.offset = (u64)-1;
3802
3803 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3804 if (IS_ERR(new_root)) {
3805 ret = PTR_ERR(new_root);
3806 goto out;
3807 }
6ef5ed0d 3808
6ef5ed0d 3809 path = btrfs_alloc_path();
3c04ce01
MX		 3810 if (!path) {
3811 ret = -ENOMEM;
3812 goto out;
3813 }
6ef5ed0d
JB		 3814 path->leave_spinning = 1;
3815
3816 trans = btrfs_start_transaction(root, 1);
98d5dc13 3817 if (IS_ERR(trans)) {
6ef5ed0d 3818 btrfs_free_path(path);
3c04ce01
MX		 3819 ret = PTR_ERR(trans);
3820 goto out;
6ef5ed0d
JB		 3821 }
3822
6c41761f 3823 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3824 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3825 dir_id, "default", 7, 1);
cf1e99a4 3826 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3827 btrfs_free_path(path);
3828 btrfs_end_transaction(trans, root);
efe120a0
FH		 3829 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3830 "item, this isn't going to work");
3c04ce01
MX		 3831 ret = -ENOENT;
3832 goto out;
6ef5ed0d
JB		 3833 }
3834
3835 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3836 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3837 btrfs_mark_buffer_dirty(path->nodes[0]);
3838 btrfs_free_path(path);
3839
2b0ce2c2 3840 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3841 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3842out:
3843 mnt_drop_write_file(file);
3844 return ret;
6ef5ed0d
JB		 3845}
3846
5af3e8cc
SB		 3847void btrfs_get_block_group_info(struct list_head *groups_list,
3848 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3849{
3850 struct btrfs_block_group_cache *block_group;
3851
3852 space->total_bytes = 0;
3853 space->used_bytes = 0;
3854 space->flags = 0;
3855 list_for_each_entry(block_group, groups_list, list) {
3856 space->flags = block_group->flags;
3857 space->total_bytes += block_group->key.offset;
3858 space->used_bytes +=
3859 btrfs_block_group_used(&block_group->item);
3860 }
3861}
3862
48a3b636 3863static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3864{
3865 struct btrfs_ioctl_space_args space_args;
3866 struct btrfs_ioctl_space_info space;
3867 struct btrfs_ioctl_space_info *dest;
7fde62bf 3868 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3869 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3870 struct btrfs_space_info *info;
bf5fc093
JB		 3871 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3872 BTRFS_BLOCK_GROUP_SYSTEM,
3873 BTRFS_BLOCK_GROUP_METADATA,
3874 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3875 int num_types = 4;
7fde62bf 3876 int alloc_size;
1406e432 3877 int ret = 0;
51788b1b 3878 u64 slot_count = 0;
bf5fc093 3879 int i, c;
1406e432
JB		 3880
3881 if (copy_from_user(&space_args,
3882 (struct btrfs_ioctl_space_args __user *)arg,
3883 sizeof(space_args)))
3884 return -EFAULT;
3885
bf5fc093
JB		 3886 for (i = 0; i < num_types; i++) {
3887 struct btrfs_space_info *tmp;
3888
3889 info = NULL;
3890 rcu_read_lock();
3891 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3892 list) {
3893 if (tmp->flags == types[i]) {
3894 info = tmp;
3895 break;
3896 }
3897 }
3898 rcu_read_unlock();
3899
3900 if (!info)
3901 continue;
3902
3903 down_read(&info->groups_sem);
3904 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3905 if (!list_empty(&info->block_groups[c]))
3906 slot_count++;
3907 }
3908 up_read(&info->groups_sem);
3909 }
7fde62bf 3910
36523e95
DS		 3911 /*
3912 * Global block reserve, exported as a space_info
3913 */
3914 slot_count++;
3915
7fde62bf
CM		 3916 /* space_slots == 0 means they are asking for a count */
3917 if (space_args.space_slots == 0) {
3918 space_args.total_spaces = slot_count;
3919 goto out;
3920 }
bf5fc093 3921
51788b1b 3922 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3923
7fde62bf 3924 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3925
7fde62bf
CM		 3926 /* we generally have at most 6 or so space infos, one for each raid
3927 * level. So, a whole page should be more than enough for everyone
3928 */
3929 if (alloc_size > PAGE_CACHE_SIZE)
3930 return -ENOMEM;
3931
1406e432 3932 space_args.total_spaces = 0;
7fde62bf
CM		 3933 dest = kmalloc(alloc_size, GFP_NOFS);
3934 if (!dest)
3935 return -ENOMEM;
3936 dest_orig = dest;
1406e432 3937
7fde62bf 3938 /* now we have a buffer to copy into */
bf5fc093
JB		 3939 for (i = 0; i < num_types; i++) {
3940 struct btrfs_space_info *tmp;
3941
51788b1b
DR		 3942 if (!slot_count)
3943 break;
3944
bf5fc093
JB		 3945 info = NULL;
3946 rcu_read_lock();
3947 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3948 list) {
3949 if (tmp->flags == types[i]) {
3950 info = tmp;
3951 break;
3952 }
3953 }
3954 rcu_read_unlock();
7fde62bf 3955
bf5fc093
JB		 3956 if (!info)
3957 continue;
3958 down_read(&info->groups_sem);
3959 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3960 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3961 btrfs_get_block_group_info(
3962 &info->block_groups[c], &space);
bf5fc093
JB		 3963 memcpy(dest, &space, sizeof(space));
3964 dest++;
3965 space_args.total_spaces++;
51788b1b 3966 slot_count--;
bf5fc093 3967 }
51788b1b
DR		 3968 if (!slot_count)
3969 break;
bf5fc093
JB		 3970 }
3971 up_read(&info->groups_sem);
1406e432 3972 }
1406e432 3973
36523e95
DS		 3974 /*
3975 * Add global block reserve
3976 */
3977 if (slot_count) {
3978 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3979
3980 spin_lock(&block_rsv->lock);
3981 space.total_bytes = block_rsv->size;
3982 space.used_bytes = block_rsv->size - block_rsv->reserved;
3983 spin_unlock(&block_rsv->lock);
3984 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3985 memcpy(dest, &space, sizeof(space));
3986 space_args.total_spaces++;
3987 }
3988
2eec6c81 3989 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3990 (arg + sizeof(struct btrfs_ioctl_space_args));
3991
3992 if (copy_to_user(user_dest, dest_orig, alloc_size))
3993 ret = -EFAULT;
3994
3995 kfree(dest_orig);
3996out:
3997 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 3998 ret = -EFAULT;
3999
4000 return ret;
4001}
4002
f46b5a66
CH		 4003/*
4004 * there are many ways the trans_start and trans_end ioctls can lead
4005 * to deadlocks. They should only be used by applications that
4006 * basically own the machine, and have a very in depth understanding
4007 * of all the possible deadlocks and enospc problems.
4008 */
4009long btrfs_ioctl_trans_end(struct file *file)
4010{
496ad9aa 4011 struct inode *inode = file_inode(file);
f46b5a66
CH		 4012 struct btrfs_root *root = BTRFS_I(inode)->root;
4013 struct btrfs_trans_handle *trans;
f46b5a66 4014
f46b5a66 4015 trans = file->private_data;
1ab86aed
SW		 4016 if (!trans)
4017 return -EINVAL;
b214107e 4018 file->private_data = NULL;
9ca9ee09 4019
1ab86aed
SW		 4020 btrfs_end_transaction(trans, root);
4021
a4abeea4 4022 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 4023
2a79f17e 4024 mnt_drop_write_file(file);
1ab86aed 4025 return 0;
f46b5a66
CH		 4026}
4027
9a8c28be
MX		 4028static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4029 void __user *argp)
46204592 4030{
46204592
SW		 4031 struct btrfs_trans_handle *trans;
4032 u64 transid;
db5b493a 4033 int ret;
46204592 4034
d4edf39b 4035 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4036 if (IS_ERR(trans)) {
4037 if (PTR_ERR(trans) != -ENOENT)
4038 return PTR_ERR(trans);
4039
4040 /* No running transaction, don't bother */
4041 transid = root->fs_info->last_trans_committed;
4042 goto out;
4043 }
46204592 4044 transid = trans->transid;
db5b493a 4045 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 4046 if (ret) {
4047 btrfs_end_transaction(trans, root);
db5b493a 4048 return ret;
8b2b2d3c 4049 }
ff7c1d33 4050out:
46204592
SW		 4051 if (argp)
4052 if (copy_to_user(argp, &transid, sizeof(transid)))
4053 return -EFAULT;
4054 return 0;
4055}
4056
9a8c28be
MX		 4057static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4058 void __user *argp)
46204592 4059{
46204592
SW		 4060 u64 transid;
4061
4062 if (argp) {
4063 if (copy_from_user(&transid, argp, sizeof(transid)))
4064 return -EFAULT;
4065 } else {
4066 transid = 0; /* current trans */
4067 }
4068 return btrfs_wait_for_commit(root, transid);
4069}
4070
b8e95489 4071static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4072{
496ad9aa 4073 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 4074 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4075 int ret;
475f6387
JS		 4076
4077 if (!capable(CAP_SYS_ADMIN))
4078 return -EPERM;
4079
4080 sa = memdup_user(arg, sizeof(*sa));
4081 if (IS_ERR(sa))
4082 return PTR_ERR(sa);
4083
b8e95489
MX		 4084 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4085 ret = mnt_want_write_file(file);
4086 if (ret)
4087 goto out;
4088 }
4089
aa1b8cd4 4090 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4091 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4092 0);
475f6387
JS		 4093
4094 if (copy_to_user(arg, sa, sizeof(*sa)))
4095 ret = -EFAULT;
4096
b8e95489
MX		 4097 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4098 mnt_drop_write_file(file);
4099out:
475f6387
JS		 4100 kfree(sa);
4101 return ret;
4102}
4103
4104static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4105{
4106 if (!capable(CAP_SYS_ADMIN))
4107 return -EPERM;
4108
aa1b8cd4 4109 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 4110}
4111
4112static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4113 void __user *arg)
4114{
4115 struct btrfs_ioctl_scrub_args *sa;
4116 int ret;
4117
4118 if (!capable(CAP_SYS_ADMIN))
4119 return -EPERM;
4120
4121 sa = memdup_user(arg, sizeof(*sa));
4122 if (IS_ERR(sa))
4123 return PTR_ERR(sa);
4124
4125 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4126
4127 if (copy_to_user(arg, sa, sizeof(*sa)))
4128 ret = -EFAULT;
4129
4130 kfree(sa);
4131 return ret;
4132}
4133
c11d2c23 4134static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 4135 void __user *arg)
c11d2c23
SB		 4136{
4137 struct btrfs_ioctl_get_dev_stats *sa;
4138 int ret;
4139
c11d2c23
SB		 4140 sa = memdup_user(arg, sizeof(*sa));
4141 if (IS_ERR(sa))
4142 return PTR_ERR(sa);
4143
b27f7c0c
DS		 4144 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4145 kfree(sa);
4146 return -EPERM;
4147 }
4148
4149 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 4150
4151 if (copy_to_user(arg, sa, sizeof(*sa)))
4152 ret = -EFAULT;
4153
4154 kfree(sa);
4155 return ret;
4156}
4157
3f6bcfbd
SB		 4158static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4159{
4160 struct btrfs_ioctl_dev_replace_args *p;
4161 int ret;
4162
4163 if (!capable(CAP_SYS_ADMIN))
4164 return -EPERM;
4165
4166 p = memdup_user(arg, sizeof(*p));
4167 if (IS_ERR(p))
4168 return PTR_ERR(p);
4169
4170 switch (p->cmd) {
4171 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
adfa97cb
ID		 4172 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4173 ret = -EROFS;
4174 goto out;
4175 }
3f6bcfbd
SB		 4176 if (atomic_xchg(
4177 &root->fs_info->mutually_exclusive_operation_running,
4178 1)) {
e57138b3 4179 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd
SB		 4180 } else {
4181 ret = btrfs_dev_replace_start(root, p);
4182 atomic_set(
4183 &root->fs_info->mutually_exclusive_operation_running,
4184 0);
4185 }
4186 break;
4187 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4188 btrfs_dev_replace_status(root->fs_info, p);
4189 ret = 0;
4190 break;
4191 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4192 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4193 break;
4194 default:
4195 ret = -EINVAL;
4196 break;
4197 }
4198
4199 if (copy_to_user(arg, p, sizeof(*p)))
4200 ret = -EFAULT;
adfa97cb 4201out:
3f6bcfbd
SB		 4202 kfree(p);
4203 return ret;
4204}
4205
d7728c96
JS		 4206static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4207{
4208 int ret = 0;
4209 int i;
740c3d22 4210 u64 rel_ptr;
d7728c96 4211 int size;
806468f8 4212 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4213 struct inode_fs_paths *ipath = NULL;
4214 struct btrfs_path *path;
4215
82b22ac8 4216 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4217 return -EPERM;
4218
4219 path = btrfs_alloc_path();
4220 if (!path) {
4221 ret = -ENOMEM;
4222 goto out;
4223 }
4224
4225 ipa = memdup_user(arg, sizeof(*ipa));
4226 if (IS_ERR(ipa)) {
4227 ret = PTR_ERR(ipa);
4228 ipa = NULL;
4229 goto out;
4230 }
4231
4232 size = min_t(u32, ipa->size, 4096);
4233 ipath = init_ipath(size, root, path);
4234 if (IS_ERR(ipath)) {
4235 ret = PTR_ERR(ipath);
4236 ipath = NULL;
4237 goto out;
4238 }
4239
4240 ret = paths_from_inode(ipa->inum, ipath);
4241 if (ret < 0)
4242 goto out;
4243
4244 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4245 rel_ptr = ipath->fspath->val[i] -
4246 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4247 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4248 }
4249
745c4d8e
JM		 4250 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4251 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 4252 if (ret) {
4253 ret = -EFAULT;
4254 goto out;
4255 }
4256
4257out:
4258 btrfs_free_path(path);
4259 free_ipath(ipath);
4260 kfree(ipa);
4261
4262 return ret;
4263}
4264
4265static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4266{
4267 struct btrfs_data_container *inodes = ctx;
4268 const size_t c = 3 * sizeof(u64);
4269
4270 if (inodes->bytes_left >= c) {
4271 inodes->bytes_left -= c;
4272 inodes->val[inodes->elem_cnt] = inum;
4273 inodes->val[inodes->elem_cnt + 1] = offset;
4274 inodes->val[inodes->elem_cnt + 2] = root;
4275 inodes->elem_cnt += 3;
4276 } else {
4277 inodes->bytes_missing += c - inodes->bytes_left;
4278 inodes->bytes_left = 0;
4279 inodes->elem_missed += 3;
4280 }
4281
4282 return 0;
4283}
4284
4285static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4286 void __user *arg)
4287{
4288 int ret = 0;
4289 int size;
d7728c96
JS		 4290 struct btrfs_ioctl_logical_ino_args *loi;
4291 struct btrfs_data_container *inodes = NULL;
4292 struct btrfs_path *path = NULL;
d7728c96
JS		 4293
4294 if (!capable(CAP_SYS_ADMIN))
4295 return -EPERM;
4296
4297 loi = memdup_user(arg, sizeof(*loi));
4298 if (IS_ERR(loi)) {
4299 ret = PTR_ERR(loi);
4300 loi = NULL;
4301 goto out;
4302 }
4303
4304 path = btrfs_alloc_path();
4305 if (!path) {
4306 ret = -ENOMEM;
4307 goto out;
4308 }
4309
425d17a2 4310 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 4311 inodes = init_data_container(size);
4312 if (IS_ERR(inodes)) {
4313 ret = PTR_ERR(inodes);
4314 inodes = NULL;
4315 goto out;
4316 }
4317
df031f07
LB		 4318 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4319 build_ino_list, inodes);
4320 if (ret == -EINVAL)
d7728c96
JS		 4321 ret = -ENOENT;
4322 if (ret < 0)
4323 goto out;
4324
745c4d8e
JM		 4325 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4326 (void *)(unsigned long)inodes, size);
d7728c96
JS		 4327 if (ret)
4328 ret = -EFAULT;
4329
4330out:
4331 btrfs_free_path(path);
425d17a2 4332 vfree(inodes);
d7728c96
JS		 4333 kfree(loi);
4334
4335 return ret;
4336}
4337
19a39dce 4338void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 4339 struct btrfs_ioctl_balance_args *bargs)
4340{
4341 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4342
4343 bargs->flags = bctl->flags;
4344
837d5b6e
ID		 4345 if (atomic_read(&fs_info->balance_running))
4346 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4347 if (atomic_read(&fs_info->balance_pause_req))
4348 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4349 if (atomic_read(&fs_info->balance_cancel_req))
4350 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4351
c9e9f97b
ID		 4352 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4353 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4354 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 4355
4356 if (lock) {
4357 spin_lock(&fs_info->balance_lock);
4358 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4359 spin_unlock(&fs_info->balance_lock);
4360 } else {
4361 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4362 }
c9e9f97b
ID		 4363}
4364
9ba1f6e4 4365static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4366{
496ad9aa 4367 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4368 struct btrfs_fs_info *fs_info = root->fs_info;
4369 struct btrfs_ioctl_balance_args *bargs;
4370 struct btrfs_balance_control *bctl;
ed0fb78f 4371 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4372 int ret;
4373
4374 if (!capable(CAP_SYS_ADMIN))
4375 return -EPERM;
4376
e54bfa31 4377 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4378 if (ret)
4379 return ret;
4380
ed0fb78f
ID		 4381again:
4382 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4383 mutex_lock(&fs_info->volume_mutex);
4384 mutex_lock(&fs_info->balance_mutex);
4385 need_unlock = true;
4386 goto locked;
4387 }
4388
4389 /*
4390 * mut. excl. ops lock is locked. Three possibilites:
4391 * (1) some other op is running
4392 * (2) balance is running
4393 * (3) balance is paused -- special case (think resume)
4394 */
c9e9f97b 4395 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4396 if (fs_info->balance_ctl) {
4397 /* this is either (2) or (3) */
4398 if (!atomic_read(&fs_info->balance_running)) {
4399 mutex_unlock(&fs_info->balance_mutex);
4400 if (!mutex_trylock(&fs_info->volume_mutex))
4401 goto again;
4402 mutex_lock(&fs_info->balance_mutex);
4403
4404 if (fs_info->balance_ctl &&
4405 !atomic_read(&fs_info->balance_running)) {
4406 /* this is (3) */
4407 need_unlock = false;
4408 goto locked;
4409 }
4410
4411 mutex_unlock(&fs_info->balance_mutex);
4412 mutex_unlock(&fs_info->volume_mutex);
4413 goto again;
4414 } else {
4415 /* this is (2) */
4416 mutex_unlock(&fs_info->balance_mutex);
4417 ret = -EINPROGRESS;
4418 goto out;
4419 }
4420 } else {
4421 /* this is (1) */
4422 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4423 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4424 goto out;
4425 }
4426
4427locked:
4428 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 4429
4430 if (arg) {
4431 bargs = memdup_user(arg, sizeof(*bargs));
4432 if (IS_ERR(bargs)) {
4433 ret = PTR_ERR(bargs);
ed0fb78f 4434 goto out_unlock;
c9e9f97b 4435 }
de322263
ID		 4436
4437 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4438 if (!fs_info->balance_ctl) {
4439 ret = -ENOTCONN;
4440 goto out_bargs;
4441 }
4442
4443 bctl = fs_info->balance_ctl;
4444 spin_lock(&fs_info->balance_lock);
4445 bctl->flags |= BTRFS_BALANCE_RESUME;
4446 spin_unlock(&fs_info->balance_lock);
4447
4448 goto do_balance;
4449 }
c9e9f97b
ID		 4450 } else {
4451 bargs = NULL;
4452 }
4453
ed0fb78f 4454 if (fs_info->balance_ctl) {
837d5b6e
ID		 4455 ret = -EINPROGRESS;
4456 goto out_bargs;
4457 }
4458
c9e9f97b
ID		 4459 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4460 if (!bctl) {
4461 ret = -ENOMEM;
4462 goto out_bargs;
4463 }
4464
4465 bctl->fs_info = fs_info;
4466 if (arg) {
4467 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4468 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4469 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4470
4471 bctl->flags = bargs->flags;
f43ffb60
ID		 4472 } else {
4473 /* balance everything - no filters */
4474 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4475 }
4476
de322263 4477do_balance:
c9e9f97b 4478 /*
ed0fb78f
ID		 4479 * Ownership of bctl and mutually_exclusive_operation_running
4480 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4481 * or, if restriper was paused all the way until unmount, in
4482 * free_fs_info. mutually_exclusive_operation_running is
4483 * cleared in __cancel_balance.
c9e9f97b 4484 */
ed0fb78f
ID		 4485 need_unlock = false;
4486
4487 ret = btrfs_balance(bctl, bargs);
4488
c9e9f97b
ID		 4489 if (arg) {
4490 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4491 ret = -EFAULT;
4492 }
4493
4494out_bargs:
4495 kfree(bargs);
ed0fb78f 4496out_unlock:
c9e9f97b
ID		 4497 mutex_unlock(&fs_info->balance_mutex);
4498 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 4499 if (need_unlock)
4500 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4501out:
e54bfa31 4502 mnt_drop_write_file(file);
c9e9f97b
ID		 4503 return ret;
4504}
4505
837d5b6e
ID		 4506static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4507{
4508 if (!capable(CAP_SYS_ADMIN))
4509 return -EPERM;
4510
4511 switch (cmd) {
4512 case BTRFS_BALANCE_CTL_PAUSE:
4513 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4514 case BTRFS_BALANCE_CTL_CANCEL:
4515 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4516 }
4517
4518 return -EINVAL;
4519}
4520
19a39dce
ID		 4521static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4522 void __user *arg)
4523{
4524 struct btrfs_fs_info *fs_info = root->fs_info;
4525 struct btrfs_ioctl_balance_args *bargs;
4526 int ret = 0;
4527
4528 if (!capable(CAP_SYS_ADMIN))
4529 return -EPERM;
4530
4531 mutex_lock(&fs_info->balance_mutex);
4532 if (!fs_info->balance_ctl) {
4533 ret = -ENOTCONN;
4534 goto out;
4535 }
4536
4537 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4538 if (!bargs) {
4539 ret = -ENOMEM;
4540 goto out;
4541 }
4542
4543 update_ioctl_balance_args(fs_info, 1, bargs);
4544
4545 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4546 ret = -EFAULT;
4547
4548 kfree(bargs);
4549out:
4550 mutex_unlock(&fs_info->balance_mutex);
4551 return ret;
4552}
4553
905b0dda 4554static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4555{
496ad9aa 4556 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4557 struct btrfs_ioctl_quota_ctl_args *sa;
4558 struct btrfs_trans_handle *trans = NULL;
4559 int ret;
4560 int err;
4561
4562 if (!capable(CAP_SYS_ADMIN))
4563 return -EPERM;
4564
905b0dda
MX		 4565 ret = mnt_want_write_file(file);
4566 if (ret)
4567 return ret;
5d13a37b
AJ		 4568
4569 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4570 if (IS_ERR(sa)) {
4571 ret = PTR_ERR(sa);
4572 goto drop_write;
4573 }
5d13a37b 4574
7708f029 4575 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4576 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4577 if (IS_ERR(trans)) {
4578 ret = PTR_ERR(trans);
4579 goto out;
5d13a37b
AJ		 4580 }
4581
4582 switch (sa->cmd) {
4583 case BTRFS_QUOTA_CTL_ENABLE:
4584 ret = btrfs_quota_enable(trans, root->fs_info);
4585 break;
4586 case BTRFS_QUOTA_CTL_DISABLE:
4587 ret = btrfs_quota_disable(trans, root->fs_info);
4588 break;
5d13a37b
AJ		 4589 default:
4590 ret = -EINVAL;
4591 break;
4592 }
4593
2f232036
JS		 4594 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4595 if (err && !ret)
4596 ret = err;
5d13a37b
AJ		 4597out:
4598 kfree(sa);
7708f029 4599 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4600drop_write:
4601 mnt_drop_write_file(file);
5d13a37b
AJ		 4602 return ret;
4603}
4604
905b0dda 4605static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4606{
496ad9aa 4607 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4608 struct btrfs_ioctl_qgroup_assign_args *sa;
4609 struct btrfs_trans_handle *trans;
4610 int ret;
4611 int err;
4612
4613 if (!capable(CAP_SYS_ADMIN))
4614 return -EPERM;
4615
905b0dda
MX		 4616 ret = mnt_want_write_file(file);
4617 if (ret)
4618 return ret;
5d13a37b
AJ		 4619
4620 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4621 if (IS_ERR(sa)) {
4622 ret = PTR_ERR(sa);
4623 goto drop_write;
4624 }
5d13a37b
AJ		 4625
4626 trans = btrfs_join_transaction(root);
4627 if (IS_ERR(trans)) {
4628 ret = PTR_ERR(trans);
4629 goto out;
4630 }
4631
4632 /* FIXME: check if the IDs really exist */
4633 if (sa->assign) {
4634 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4635 sa->src, sa->dst);
4636 } else {
4637 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4638 sa->src, sa->dst);
4639 }
4640
4641 err = btrfs_end_transaction(trans, root);
4642 if (err && !ret)
4643 ret = err;
4644
4645out:
4646 kfree(sa);
905b0dda
MX		 4647drop_write:
4648 mnt_drop_write_file(file);
5d13a37b
AJ		 4649 return ret;
4650}
4651
905b0dda 4652static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4653{
496ad9aa 4654 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4655 struct btrfs_ioctl_qgroup_create_args *sa;
4656 struct btrfs_trans_handle *trans;
4657 int ret;
4658 int err;
4659
4660 if (!capable(CAP_SYS_ADMIN))
4661 return -EPERM;
4662
905b0dda
MX		 4663 ret = mnt_want_write_file(file);
4664 if (ret)
4665 return ret;
5d13a37b
AJ		 4666
4667 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4668 if (IS_ERR(sa)) {
4669 ret = PTR_ERR(sa);
4670 goto drop_write;
4671 }
5d13a37b 4672
d86e56cf
MX		 4673 if (!sa->qgroupid) {
4674 ret = -EINVAL;
4675 goto out;
4676 }
4677
5d13a37b
AJ		 4678 trans = btrfs_join_transaction(root);
4679 if (IS_ERR(trans)) {
4680 ret = PTR_ERR(trans);
4681 goto out;
4682 }
4683
4684 /* FIXME: check if the IDs really exist */
4685 if (sa->create) {
4686 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4687 NULL);
4688 } else {
4689 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4690 }
4691
4692 err = btrfs_end_transaction(trans, root);
4693 if (err && !ret)
4694 ret = err;
4695
4696out:
4697 kfree(sa);
905b0dda
MX		 4698drop_write:
4699 mnt_drop_write_file(file);
5d13a37b
AJ		 4700 return ret;
4701}
4702
905b0dda 4703static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4704{
496ad9aa 4705 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4706 struct btrfs_ioctl_qgroup_limit_args *sa;
4707 struct btrfs_trans_handle *trans;
4708 int ret;
4709 int err;
4710 u64 qgroupid;
4711
4712 if (!capable(CAP_SYS_ADMIN))
4713 return -EPERM;
4714
905b0dda
MX		 4715 ret = mnt_want_write_file(file);
4716 if (ret)
4717 return ret;
5d13a37b
AJ		 4718
4719 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4720 if (IS_ERR(sa)) {
4721 ret = PTR_ERR(sa);
4722 goto drop_write;
4723 }
5d13a37b
AJ		 4724
4725 trans = btrfs_join_transaction(root);
4726 if (IS_ERR(trans)) {
4727 ret = PTR_ERR(trans);
4728 goto out;
4729 }
4730
4731 qgroupid = sa->qgroupid;
4732 if (!qgroupid) {
4733 /* take the current subvol as qgroup */
4734 qgroupid = root->root_key.objectid;
4735 }
4736
4737 /* FIXME: check if the IDs really exist */
4738 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4739
4740 err = btrfs_end_transaction(trans, root);
4741 if (err && !ret)
4742 ret = err;
4743
4744out:
4745 kfree(sa);
905b0dda
MX		 4746drop_write:
4747 mnt_drop_write_file(file);
5d13a37b
AJ		 4748 return ret;
4749}
4750
2f232036
JS		 4751static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4752{
6d0379ec 4753 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4754 struct btrfs_ioctl_quota_rescan_args *qsa;
4755 int ret;
4756
4757 if (!capable(CAP_SYS_ADMIN))
4758 return -EPERM;
4759
4760 ret = mnt_want_write_file(file);
4761 if (ret)
4762 return ret;
4763
4764 qsa = memdup_user(arg, sizeof(*qsa));
4765 if (IS_ERR(qsa)) {
4766 ret = PTR_ERR(qsa);
4767 goto drop_write;
4768 }
4769
4770 if (qsa->flags) {
4771 ret = -EINVAL;
4772 goto out;
4773 }
4774
4775 ret = btrfs_qgroup_rescan(root->fs_info);
4776
4777out:
4778 kfree(qsa);
4779drop_write:
4780 mnt_drop_write_file(file);
4781 return ret;
4782}
4783
4784static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4785{
6d0379ec 4786 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4787 struct btrfs_ioctl_quota_rescan_args *qsa;
4788 int ret = 0;
4789
4790 if (!capable(CAP_SYS_ADMIN))
4791 return -EPERM;
4792
4793 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4794 if (!qsa)
4795 return -ENOMEM;
4796
4797 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4798 qsa->flags = 1;
4799 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4800 }
4801
4802 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4803 ret = -EFAULT;
4804
4805 kfree(qsa);
4806 return ret;
4807}
4808
57254b6e
JS		 4809static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4810{
54563d41 4811 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
57254b6e
JS		 4812
4813 if (!capable(CAP_SYS_ADMIN))
4814 return -EPERM;
4815
4816 return btrfs_qgroup_wait_for_completion(root->fs_info);
4817}
4818
abccd00f
HM		 4819static long _btrfs_ioctl_set_received_subvol(struct file *file,
4820 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4821{
496ad9aa 4822 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4823 struct btrfs_root *root = BTRFS_I(inode)->root;
4824 struct btrfs_root_item *root_item = &root->root_item;
4825 struct btrfs_trans_handle *trans;
4826 struct timespec ct = CURRENT_TIME;
4827 int ret = 0;
dd5f9615 4828 int received_uuid_changed;
8ea05e3a 4829
bd60ea0f
DS		 4830 if (!inode_owner_or_capable(inode))
4831 return -EPERM;
4832
8ea05e3a
AB		 4833 ret = mnt_want_write_file(file);
4834 if (ret < 0)
4835 return ret;
4836
4837 down_write(&root->fs_info->subvol_sem);
4838
4839 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4840 ret = -EINVAL;
4841 goto out;
4842 }
4843
4844 if (btrfs_root_readonly(root)) {
4845 ret = -EROFS;
4846 goto out;
4847 }
4848
dd5f9615
SB		 4849 /*
4850 * 1 - root item
4851 * 2 - uuid items (received uuid + subvol uuid)
4852 */
4853 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4854 if (IS_ERR(trans)) {
4855 ret = PTR_ERR(trans);
4856 trans = NULL;
4857 goto out;
4858 }
4859
4860 sa->rtransid = trans->transid;
4861 sa->rtime.sec = ct.tv_sec;
4862 sa->rtime.nsec = ct.tv_nsec;
4863
dd5f9615
SB		 4864 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4865 BTRFS_UUID_SIZE);
4866 if (received_uuid_changed &&
4867 !btrfs_is_empty_uuid(root_item->received_uuid))
4868 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4869 root_item->received_uuid,
4870 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4871 root->root_key.objectid);
8ea05e3a
AB		 4872 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4873 btrfs_set_root_stransid(root_item, sa->stransid);
4874 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4875 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4876 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4877 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4878 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4879
4880 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4881 &root->root_key, &root->root_item);
4882 if (ret < 0) {
4883 btrfs_end_transaction(trans, root);
8ea05e3a 4884 goto out;
dd5f9615
SB		 4885 }
4886 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4887 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4888 sa->uuid,
4889 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4890 root->root_key.objectid);
4891 if (ret < 0 && ret != -EEXIST) {
4892 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 4893 goto out;
dd5f9615
SB		 4894 }
4895 }
4896 ret = btrfs_commit_transaction(trans, root);
4897 if (ret < 0) {
4898 btrfs_abort_transaction(trans, root, ret);
4899 goto out;
8ea05e3a
AB		 4900 }
4901
abccd00f
HM		 4902out:
4903 up_write(&root->fs_info->subvol_sem);
4904 mnt_drop_write_file(file);
4905 return ret;
4906}
4907
4908#ifdef CONFIG_64BIT
4909static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4910 void __user *arg)
4911{
4912 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4913 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4914 int ret = 0;
4915
4916 args32 = memdup_user(arg, sizeof(*args32));
4917 if (IS_ERR(args32)) {
4918 ret = PTR_ERR(args32);
4919 args32 = NULL;
4920 goto out;
4921 }
4922
4923 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
84dbeb87
DC		 4924 if (!args64) {
4925 ret = -ENOMEM;
abccd00f
HM		 4926 goto out;
4927 }
4928
4929 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4930 args64->stransid = args32->stransid;
4931 args64->rtransid = args32->rtransid;
4932 args64->stime.sec = args32->stime.sec;
4933 args64->stime.nsec = args32->stime.nsec;
4934 args64->rtime.sec = args32->rtime.sec;
4935 args64->rtime.nsec = args32->rtime.nsec;
4936 args64->flags = args32->flags;
4937
4938 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4939 if (ret)
4940 goto out;
4941
4942 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4943 args32->stransid = args64->stransid;
4944 args32->rtransid = args64->rtransid;
4945 args32->stime.sec = args64->stime.sec;
4946 args32->stime.nsec = args64->stime.nsec;
4947 args32->rtime.sec = args64->rtime.sec;
4948 args32->rtime.nsec = args64->rtime.nsec;
4949 args32->flags = args64->flags;
4950
4951 ret = copy_to_user(arg, args32, sizeof(*args32));
4952 if (ret)
4953 ret = -EFAULT;
4954
4955out:
4956 kfree(args32);
4957 kfree(args64);
4958 return ret;
4959}
4960#endif
4961
4962static long btrfs_ioctl_set_received_subvol(struct file *file,
4963 void __user *arg)
4964{
4965 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4966 int ret = 0;
4967
4968 sa = memdup_user(arg, sizeof(*sa));
4969 if (IS_ERR(sa)) {
4970 ret = PTR_ERR(sa);
4971 sa = NULL;
4972 goto out;
4973 }
4974
4975 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4976
4977 if (ret)
4978 goto out;
4979
8ea05e3a
AB		 4980 ret = copy_to_user(arg, sa, sizeof(*sa));
4981 if (ret)
4982 ret = -EFAULT;
4983
4984out:
4985 kfree(sa);
8ea05e3a
AB		 4986 return ret;
4987}
4988
867ab667 4989static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4990{
6d0379ec 4991 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 4992 size_t len;
867ab667 4993 int ret;
a1b83ac5
AJ		 4994 char label[BTRFS_LABEL_SIZE];
4995
4996 spin_lock(&root->fs_info->super_lock);
4997 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4998 spin_unlock(&root->fs_info->super_lock);
4999
5000 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5001
5002 if (len == BTRFS_LABEL_SIZE) {
efe120a0
FH		 5003 btrfs_warn(root->fs_info,
5004 "label is too long, return the first %zu bytes", --len);
867ab667 5005 }
5006
867ab667 5007 ret = copy_to_user(arg, label, len);
867ab667 5008
5009 return ret ? -EFAULT : 0;
5010}
5011
a8bfd4ab 5012static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5013{
6d0379ec 5014 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 5015 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5016 struct btrfs_trans_handle *trans;
5017 char label[BTRFS_LABEL_SIZE];
5018 int ret;
5019
5020 if (!capable(CAP_SYS_ADMIN))
5021 return -EPERM;
5022
5023 if (copy_from_user(label, arg, sizeof(label)))
5024 return -EFAULT;
5025
5026 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
efe120a0 5027 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
a8bfd4ab 5028 BTRFS_LABEL_SIZE - 1);
5029 return -EINVAL;
5030 }
5031
5032 ret = mnt_want_write_file(file);
5033 if (ret)
5034 return ret;
5035
a8bfd4ab 5036 trans = btrfs_start_transaction(root, 0);
5037 if (IS_ERR(trans)) {
5038 ret = PTR_ERR(trans);
5039 goto out_unlock;
5040 }
5041
a1b83ac5 5042 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 5043 strcpy(super_block->label, label);
a1b83ac5 5044 spin_unlock(&root->fs_info->super_lock);
d0270aca 5045 ret = btrfs_commit_transaction(trans, root);
a8bfd4ab 5046
5047out_unlock:
a8bfd4ab 5048 mnt_drop_write_file(file);
5049 return ret;
5050}
5051
2eaa055f
JM		 5052#define INIT_FEATURE_FLAGS(suffix) \
5053 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5054 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5055 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5056
5057static int btrfs_ioctl_get_supported_features(struct file *file,
5058 void __user *arg)
5059{
5060 static struct btrfs_ioctl_feature_flags features[3] = {
5061 INIT_FEATURE_FLAGS(SUPP),
5062 INIT_FEATURE_FLAGS(SAFE_SET),
5063 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5064 };
5065
5066 if (copy_to_user(arg, &features, sizeof(features)))
5067 return -EFAULT;
5068
5069 return 0;
5070}
5071
5072static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5073{
5074 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5075 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5076 struct btrfs_ioctl_feature_flags features;
5077
5078 features.compat_flags = btrfs_super_compat_flags(super_block);
5079 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5080 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5081
5082 if (copy_to_user(arg, &features, sizeof(features)))
5083 return -EFAULT;
5084
5085 return 0;
5086}
5087
3b02a68a
JM		 5088static int check_feature_bits(struct btrfs_root *root,
5089 enum btrfs_feature_set set,
2eaa055f
JM		 5090 u64 change_mask, u64 flags, u64 supported_flags,
5091 u64 safe_set, u64 safe_clear)
5092{
3b02a68a
JM		 5093 const char *type = btrfs_feature_set_names[set];
5094 char *names;
2eaa055f
JM		 5095 u64 disallowed, unsupported;
5096 u64 set_mask = flags & change_mask;
5097 u64 clear_mask = ~flags & change_mask;
5098
5099 unsupported = set_mask & ~supported_flags;
5100 if (unsupported) {
3b02a68a
JM		 5101 names = btrfs_printable_features(set, unsupported);
5102 if (names) {
5103 btrfs_warn(root->fs_info,
5104 "this kernel does not support the %s feature bit%s",
5105 names, strchr(names, ',') ? "s" : "");
5106 kfree(names);
5107 } else
5108 btrfs_warn(root->fs_info,
2eaa055f
JM		 5109 "this kernel does not support %s bits 0x%llx",
5110 type, unsupported);
5111 return -EOPNOTSUPP;
5112 }
5113
5114 disallowed = set_mask & ~safe_set;
5115 if (disallowed) {
3b02a68a
JM		 5116 names = btrfs_printable_features(set, disallowed);
5117 if (names) {
5118 btrfs_warn(root->fs_info,
5119 "can't set the %s feature bit%s while mounted",
5120 names, strchr(names, ',') ? "s" : "");
5121 kfree(names);
5122 } else
5123 btrfs_warn(root->fs_info,
2eaa055f
JM		 5124 "can't set %s bits 0x%llx while mounted",
5125 type, disallowed);
5126 return -EPERM;
5127 }
5128
5129 disallowed = clear_mask & ~safe_clear;
5130 if (disallowed) {
3b02a68a
JM		 5131 names = btrfs_printable_features(set, disallowed);
5132 if (names) {
5133 btrfs_warn(root->fs_info,
5134 "can't clear the %s feature bit%s while mounted",
5135 names, strchr(names, ',') ? "s" : "");
5136 kfree(names);
5137 } else
5138 btrfs_warn(root->fs_info,
2eaa055f
JM		 5139 "can't clear %s bits 0x%llx while mounted",
5140 type, disallowed);
5141 return -EPERM;
5142 }
5143
5144 return 0;
5145}
5146
5147#define check_feature(root, change_mask, flags, mask_base) \
3b02a68a 5148check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5149 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5150 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5151 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5152
5153static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5154{
5155 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5156 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5157 struct btrfs_ioctl_feature_flags flags[2];
5158 struct btrfs_trans_handle *trans;
5159 u64 newflags;
5160 int ret;
5161
5162 if (!capable(CAP_SYS_ADMIN))
5163 return -EPERM;
5164
5165 if (copy_from_user(flags, arg, sizeof(flags)))
5166 return -EFAULT;
5167
5168 /* Nothing to do */
5169 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5170 !flags[0].incompat_flags)
5171 return 0;
5172
5173 ret = check_feature(root, flags[0].compat_flags,
5174 flags[1].compat_flags, COMPAT);
5175 if (ret)
5176 return ret;
5177
5178 ret = check_feature(root, flags[0].compat_ro_flags,
5179 flags[1].compat_ro_flags, COMPAT_RO);
5180 if (ret)
5181 return ret;
5182
5183 ret = check_feature(root, flags[0].incompat_flags,
5184 flags[1].incompat_flags, INCOMPAT);
5185 if (ret)
5186 return ret;
5187
8051aa1a 5188 trans = btrfs_start_transaction(root, 0);
2eaa055f
JM		 5189 if (IS_ERR(trans))
5190 return PTR_ERR(trans);
5191
5192 spin_lock(&root->fs_info->super_lock);
5193 newflags = btrfs_super_compat_flags(super_block);
5194 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5195 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5196 btrfs_set_super_compat_flags(super_block, newflags);
5197
5198 newflags = btrfs_super_compat_ro_flags(super_block);
5199 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5200 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5201 btrfs_set_super_compat_ro_flags(super_block, newflags);
5202
5203 newflags = btrfs_super_incompat_flags(super_block);
5204 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5205 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5206 btrfs_set_super_incompat_flags(super_block, newflags);
5207 spin_unlock(&root->fs_info->super_lock);
5208
d0270aca 5209 return btrfs_commit_transaction(trans, root);
2eaa055f
JM		 5210}
5211
f46b5a66
CH		 5212long btrfs_ioctl(struct file *file, unsigned int
5213 cmd, unsigned long arg)
5214{
496ad9aa 5215 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 5216 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5217
5218 switch (cmd) {
6cbff00f
CH		 5219 case FS_IOC_GETFLAGS:
5220 return btrfs_ioctl_getflags(file, argp);
5221 case FS_IOC_SETFLAGS:
5222 return btrfs_ioctl_setflags(file, argp);
5223 case FS_IOC_GETVERSION:
5224 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5225 case FITRIM:
5226 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5227 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5228 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5229 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5230 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5231 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5232 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5233 case BTRFS_IOC_SUBVOL_CREATE_V2:
5234 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5235 case BTRFS_IOC_SNAP_DESTROY:
5236 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5237 case BTRFS_IOC_SUBVOL_GETFLAGS:
5238 return btrfs_ioctl_subvol_getflags(file, argp);
5239 case BTRFS_IOC_SUBVOL_SETFLAGS:
5240 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5241 case BTRFS_IOC_DEFAULT_SUBVOL:
5242 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5243 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5244 return btrfs_ioctl_defrag(file, NULL);
5245 case BTRFS_IOC_DEFRAG_RANGE:
5246 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5247 case BTRFS_IOC_RESIZE:
198605a8 5248 return btrfs_ioctl_resize(file, argp);
f46b5a66 5249 case BTRFS_IOC_ADD_DEV:
4bcabaa3 5250 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 5251 case BTRFS_IOC_RM_DEV:
da24927b 5252 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 5253 case BTRFS_IOC_FS_INFO:
5254 return btrfs_ioctl_fs_info(root, argp);
5255 case BTRFS_IOC_DEV_INFO:
5256 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 5257 case BTRFS_IOC_BALANCE:
9ba1f6e4 5258 return btrfs_ioctl_balance(file, NULL);
f46b5a66 5259 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 5260 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5261 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 5262 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 5263 case BTRFS_IOC_TRANS_START:
5264 return btrfs_ioctl_trans_start(file);
5265 case BTRFS_IOC_TRANS_END:
5266 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 5267 case BTRFS_IOC_TREE_SEARCH:
5268 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5269 case BTRFS_IOC_TREE_SEARCH_V2:
5270 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5271 case BTRFS_IOC_INO_LOOKUP:
5272 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5273 case BTRFS_IOC_INO_PATHS:
5274 return btrfs_ioctl_ino_to_path(root, argp);
5275 case BTRFS_IOC_LOGICAL_INO:
5276 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 5277 case BTRFS_IOC_SPACE_INFO:
5278 return btrfs_ioctl_space_info(root, argp);
9b199859
FDBM		 5279 case BTRFS_IOC_SYNC: {
5280 int ret;
5281
6c255e67 5282 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
9b199859
FDBM		 5283 if (ret)
5284 return ret;
5285 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
2fad4e83
DS		 5286 /*
5287 * The transaction thread may want to do more work,
5288 * namely it pokes the cleaner ktread that will start
5289 * processing uncleaned subvols.
5290 */
5291 wake_up_process(root->fs_info->transaction_kthread);
9b199859
FDBM		 5292 return ret;
5293 }
46204592 5294 case BTRFS_IOC_START_SYNC:
9a8c28be 5295 return btrfs_ioctl_start_sync(root, argp);
46204592 5296 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 5297 return btrfs_ioctl_wait_sync(root, argp);
475f6387 5298 case BTRFS_IOC_SCRUB:
b8e95489 5299 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 5300 case BTRFS_IOC_SCRUB_CANCEL:
5301 return btrfs_ioctl_scrub_cancel(root, argp);
5302 case BTRFS_IOC_SCRUB_PROGRESS:
5303 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 5304 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5305 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 5306 case BTRFS_IOC_BALANCE_CTL:
5307 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 5308 case BTRFS_IOC_BALANCE_PROGRESS:
5309 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 5310 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5311 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5312#ifdef CONFIG_64BIT
5313 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5314 return btrfs_ioctl_set_received_subvol_32(file, argp);
5315#endif
31db9f7c
AB		 5316 case BTRFS_IOC_SEND:
5317 return btrfs_ioctl_send(file, argp);
c11d2c23 5318 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 5319 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 5320 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5321 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5322 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5323 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5324 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5325 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5326 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5327 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5328 case BTRFS_IOC_QUOTA_RESCAN:
5329 return btrfs_ioctl_quota_rescan(file, argp);
5330 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5331 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5332 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5333 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 5334 case BTRFS_IOC_DEV_REPLACE:
5335 return btrfs_ioctl_dev_replace(root, argp);
867ab667 5336 case BTRFS_IOC_GET_FSLABEL:
5337 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5338 case BTRFS_IOC_SET_FSLABEL:
5339 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 5340 case BTRFS_IOC_FILE_EXTENT_SAME:
5341 return btrfs_ioctl_file_extent_same(file, argp);
2eaa055f
JM		 5342 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5343 return btrfs_ioctl_get_supported_features(file, argp);
5344 case BTRFS_IOC_GET_FEATURES:
5345 return btrfs_ioctl_get_features(file, argp);
5346 case BTRFS_IOC_SET_FEATURES:
5347 return btrfs_ioctl_set_features(file, argp);
f46b5a66
CH		 5348 }
5349
5350 return -ENOTTY;
5351}
Linux 6.x block layer and io_uring tree(s)