projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: add FS_IOC_FSSETXATTR ioctl
[linux-block.git] / fs / btrfs / ioctl.c
CommitLineData
c1d7c514 1// SPDX-License-Identifier: GPL-2.0
f46b5a66
CH		 2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
f46b5a66
CH		 4 */
5
6#include <linux/kernel.h>
7#include <linux/bio.h>
8#include <linux/buffer_head.h>
9#include <linux/file.h>
10#include <linux/fs.h>
cb8e7090 11#include <linux/fsnotify.h>
f46b5a66
CH		 12#include <linux/pagemap.h>
13#include <linux/highmem.h>
14#include <linux/time.h>
15#include <linux/init.h>
16#include <linux/string.h>
f46b5a66 17#include <linux/backing-dev.h>
cb8e7090 18#include <linux/mount.h>
f46b5a66 19#include <linux/mpage.h>
cb8e7090 20#include <linux/namei.h>
f46b5a66
CH		 21#include <linux/swap.h>
22#include <linux/writeback.h>
f46b5a66
CH		 23#include <linux/compat.h>
24#include <linux/bit_spinlock.h>
cb8e7090 25#include <linux/security.h>
f46b5a66 26#include <linux/xattr.h>
f54de068 27#include <linux/mm.h>
5a0e3ad6 28#include <linux/slab.h>
f7039b1d 29#include <linux/blkdev.h>
8ea05e3a 30#include <linux/uuid.h>
55e301fd 31#include <linux/btrfs.h>
416161db 32#include <linux/uaccess.h>
ae5e165d 33#include <linux/iversion.h>
f46b5a66
CH		 34#include "ctree.h"
35#include "disk-io.h"
36#include "transaction.h"
37#include "btrfs_inode.h"
f46b5a66
CH		 38#include "print-tree.h"
39#include "volumes.h"
925baedd 40#include "locking.h"
581bb050 41#include "inode-map.h"
d7728c96 42#include "backref.h"
606686ee 43#include "rcu-string.h"
31db9f7c 44#include "send.h"
3f6bcfbd 45#include "dev-replace.h"
63541927 46#include "props.h"
3b02a68a 47#include "sysfs.h"
fcebe456 48#include "qgroup.h"
1ec9a1ae 49#include "tree-log.h"
ebb8765b 50#include "compression.h"
f46b5a66 51
abccd00f
HM		 52#ifdef CONFIG_64BIT
53/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
54 * structures are incorrect, as the timespec structure from userspace
55 * is 4 bytes too small. We define these alternatives here to teach
56 * the kernel about the 32-bit struct packing.
57 */
58struct btrfs_ioctl_timespec_32 {
59 __u64 sec;
60 __u32 nsec;
61} __attribute__ ((__packed__));
62
63struct btrfs_ioctl_received_subvol_args_32 {
64 char uuid[BTRFS_UUID_SIZE]; /* in */
65 __u64 stransid; /* in */
66 __u64 rtransid; /* out */
67 struct btrfs_ioctl_timespec_32 stime; /* in */
68 struct btrfs_ioctl_timespec_32 rtime; /* out */
69 __u64 flags; /* in */
70 __u64 reserved[16]; /* in */
71} __attribute__ ((__packed__));
72
73#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
74 struct btrfs_ioctl_received_subvol_args_32)
75#endif
76
2351f431
JB		 77#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
78struct btrfs_ioctl_send_args_32 {
79 __s64 send_fd; /* in */
80 __u64 clone_sources_count; /* in */
81 compat_uptr_t clone_sources; /* in */
82 __u64 parent_root; /* in */
83 __u64 flags; /* in */
84 __u64 reserved[4]; /* in */
85} __attribute__ ((__packed__));
86
87#define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
88 struct btrfs_ioctl_send_args_32)
89#endif
abccd00f 90
416161db 91static int btrfs_clone(struct inode *src, struct inode *inode,
1c919a5e
MF		 92 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
93 int no_time_update);
416161db 94
6cbff00f 95/* Mask out flags that are inappropriate for the given type of inode. */
1905a0f7
DS		 96static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
97 unsigned int flags)
6cbff00f 98{
1905a0f7 99 if (S_ISDIR(inode->i_mode))
6cbff00f 100 return flags;
1905a0f7 101 else if (S_ISREG(inode->i_mode))
6cbff00f
CH		 102 return flags & ~FS_DIRSYNC_FL;
103 else
104 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
105}
106
107/*
a157d4fd
DS		 108 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
109 * ioctl.
6cbff00f 110 */
a157d4fd 111static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags)
6cbff00f
CH		 112{
113 unsigned int iflags = 0;
114
115 if (flags & BTRFS_INODE_SYNC)
116 iflags |= FS_SYNC_FL;
117 if (flags & BTRFS_INODE_IMMUTABLE)
118 iflags |= FS_IMMUTABLE_FL;
119 if (flags & BTRFS_INODE_APPEND)
120 iflags |= FS_APPEND_FL;
121 if (flags & BTRFS_INODE_NODUMP)
122 iflags |= FS_NODUMP_FL;
123 if (flags & BTRFS_INODE_NOATIME)
124 iflags |= FS_NOATIME_FL;
125 if (flags & BTRFS_INODE_DIRSYNC)
126 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 127 if (flags & BTRFS_INODE_NODATACOW)
128 iflags |= FS_NOCOW_FL;
129
13f48dc9 130 if (flags & BTRFS_INODE_NOCOMPRESS)
d0092bdd 131 iflags |= FS_NOCOMP_FL;
13f48dc9
ST		 132 else if (flags & BTRFS_INODE_COMPRESS)
133 iflags |= FS_COMPR_FL;
6cbff00f
CH		 134
135 return iflags;
136}
137
138/*
139 * Update inode->i_flags based on the btrfs internal flags.
140 */
7b6a221e 141void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
6cbff00f
CH		 142{
143 struct btrfs_inode *ip = BTRFS_I(inode);
3cc79392 144 unsigned int new_fl = 0;
6cbff00f
CH		 145
146 if (ip->flags & BTRFS_INODE_SYNC)
3cc79392 147 new_fl |= S_SYNC;
6cbff00f 148 if (ip->flags & BTRFS_INODE_IMMUTABLE)
3cc79392 149 new_fl |= S_IMMUTABLE;
6cbff00f 150 if (ip->flags & BTRFS_INODE_APPEND)
3cc79392 151 new_fl |= S_APPEND;
6cbff00f 152 if (ip->flags & BTRFS_INODE_NOATIME)
3cc79392 153 new_fl |= S_NOATIME;
6cbff00f 154 if (ip->flags & BTRFS_INODE_DIRSYNC)
3cc79392
FM		 155 new_fl |= S_DIRSYNC;
156
157 set_mask_bits(&inode->i_flags,
158 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
159 new_fl);
6cbff00f
CH		 160}
161
6cbff00f
CH		 162static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
163{
496ad9aa 164 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
a157d4fd 165 unsigned int flags = btrfs_inode_flags_to_fsflags(ip->flags);
6cbff00f
CH		 166
167 if (copy_to_user(arg, &flags, sizeof(flags)))
168 return -EFAULT;
169 return 0;
170}
171
5ba76abf
DS		 172/* Check if @flags are a supported and valid set of FS_*_FL flags */
173static int check_fsflags(unsigned int flags)
75e7cb7f
LB		 174{
175 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
176 FS_NOATIME_FL | FS_NODUMP_FL | \
177 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 178 FS_NOCOMP_FL | FS_COMPR_FL |
179 FS_NOCOW_FL))
75e7cb7f
LB		 180 return -EOPNOTSUPP;
181
182 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
183 return -EINVAL;
184
75e7cb7f
LB		 185 return 0;
186}
187
6cbff00f
CH		 188static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
189{
496ad9aa 190 struct inode *inode = file_inode(file);
0b246afa 191 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6cbff00f
CH		 192 struct btrfs_inode *ip = BTRFS_I(inode);
193 struct btrfs_root *root = ip->root;
194 struct btrfs_trans_handle *trans;
195 unsigned int flags, oldflags;
196 int ret;
f062abf0
LZ		 197 u64 ip_oldflags;
198 unsigned int i_oldflags;
7e97b8da 199 umode_t mode;
6cbff00f 200
bd60ea0f
DS		 201 if (!inode_owner_or_capable(inode))
202 return -EPERM;
203
b83cc969
LZ		 204 if (btrfs_root_readonly(root))
205 return -EROFS;
206
6cbff00f
CH		 207 if (copy_from_user(&flags, arg, sizeof(flags)))
208 return -EFAULT;
209
5ba76abf 210 ret = check_fsflags(flags);
75e7cb7f
LB		 211 if (ret)
212 return ret;
f46b5a66 213
e7848683
JK		 214 ret = mnt_want_write_file(file);
215 if (ret)
216 return ret;
217
5955102c 218 inode_lock(inode);
6cbff00f 219
f062abf0
LZ		 220 ip_oldflags = ip->flags;
221 i_oldflags = inode->i_flags;
7e97b8da 222 mode = inode->i_mode;
f062abf0 223
1905a0f7 224 flags = btrfs_mask_fsflags_for_type(inode, flags);
a157d4fd 225 oldflags = btrfs_inode_flags_to_fsflags(ip->flags);
6cbff00f
CH		 226 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
227 if (!capable(CAP_LINUX_IMMUTABLE)) {
228 ret = -EPERM;
229 goto out_unlock;
230 }
231 }
232
6cbff00f
CH		 233 if (flags & FS_SYNC_FL)
234 ip->flags |= BTRFS_INODE_SYNC;
235 else
236 ip->flags &= ~BTRFS_INODE_SYNC;
237 if (flags & FS_IMMUTABLE_FL)
238 ip->flags |= BTRFS_INODE_IMMUTABLE;
239 else
240 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
241 if (flags & FS_APPEND_FL)
242 ip->flags |= BTRFS_INODE_APPEND;
243 else
244 ip->flags &= ~BTRFS_INODE_APPEND;
245 if (flags & FS_NODUMP_FL)
246 ip->flags |= BTRFS_INODE_NODUMP;
247 else
248 ip->flags &= ~BTRFS_INODE_NODUMP;
249 if (flags & FS_NOATIME_FL)
250 ip->flags |= BTRFS_INODE_NOATIME;
251 else
252 ip->flags &= ~BTRFS_INODE_NOATIME;
253 if (flags & FS_DIRSYNC_FL)
254 ip->flags |= BTRFS_INODE_DIRSYNC;
255 else
256 ip->flags &= ~BTRFS_INODE_DIRSYNC;
7e97b8da
DS		 257 if (flags & FS_NOCOW_FL) {
258 if (S_ISREG(mode)) {
259 /*
260 * It's safe to turn csums off here, no extents exist.
261 * Otherwise we want the flag to reflect the real COW
262 * status of the file and will not set it.
263 */
264 if (inode->i_size == 0)
265 ip->flags |= BTRFS_INODE_NODATACOW
266 | BTRFS_INODE_NODATASUM;
267 } else {
268 ip->flags |= BTRFS_INODE_NODATACOW;
269 }
270 } else {
271 /*
01327610 272 * Revert back under same assumptions as above
7e97b8da
DS		 273 */
274 if (S_ISREG(mode)) {
275 if (inode->i_size == 0)
276 ip->flags &= ~(BTRFS_INODE_NODATACOW
277 | BTRFS_INODE_NODATASUM);
278 } else {
279 ip->flags &= ~BTRFS_INODE_NODATACOW;
280 }
281 }
6cbff00f 282
75e7cb7f
LB		 283 /*
284 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
285 * flag may be changed automatically if compression code won't make
286 * things smaller.
287 */
288 if (flags & FS_NOCOMP_FL) {
289 ip->flags &= ~BTRFS_INODE_COMPRESS;
290 ip->flags |= BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 291
292 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
293 if (ret && ret != -ENODATA)
294 goto out_drop;
75e7cb7f 295 } else if (flags & FS_COMPR_FL) {
63541927
FDBM		 296 const char *comp;
297
75e7cb7f
LB		 298 ip->flags |= BTRFS_INODE_COMPRESS;
299 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
63541927 300
93370509
DS		 301 comp = btrfs_compress_type2str(fs_info->compress_type);
302 if (!comp || comp[0] == 0)
303 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
304
63541927
FDBM		 305 ret = btrfs_set_prop(inode, "btrfs.compression",
306 comp, strlen(comp), 0);
307 if (ret)
308 goto out_drop;
309
ebcb904d 310 } else {
78a017a2
FM		 311 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
312 if (ret && ret != -ENODATA)
313 goto out_drop;
ebcb904d 314 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 315 }
6cbff00f 316
4da6f1a3 317 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 318 if (IS_ERR(trans)) {
319 ret = PTR_ERR(trans);
320 goto out_drop;
321 }
6cbff00f 322
7b6a221e 323 btrfs_sync_inode_flags_to_i_flags(inode);
0c4d2d95 324 inode_inc_iversion(inode);
c2050a45 325 inode->i_ctime = current_time(inode);
6cbff00f 326 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 327
3a45bb20 328 btrfs_end_transaction(trans);
f062abf0
LZ		 329 out_drop:
330 if (ret) {
331 ip->flags = ip_oldflags;
332 inode->i_flags = i_oldflags;
333 }
6cbff00f 334
6cbff00f 335 out_unlock:
5955102c 336 inode_unlock(inode);
e7848683 337 mnt_drop_write_file(file);
2d4e6f6a 338 return ret;
6cbff00f
CH		 339}
340
19f93b3c
DS		 341/*
342 * Translate btrfs internal inode flags to xflags as expected by the
343 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
344 * silently dropped.
345 */
346static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
347{
348 unsigned int xflags = 0;
349
350 if (flags & BTRFS_INODE_APPEND)
351 xflags |= FS_XFLAG_APPEND;
352 if (flags & BTRFS_INODE_IMMUTABLE)
353 xflags |= FS_XFLAG_IMMUTABLE;
354 if (flags & BTRFS_INODE_NOATIME)
355 xflags |= FS_XFLAG_NOATIME;
356 if (flags & BTRFS_INODE_NODUMP)
357 xflags |= FS_XFLAG_NODUMP;
358 if (flags & BTRFS_INODE_SYNC)
359 xflags |= FS_XFLAG_SYNC;
360
361 return xflags;
362}
363
364/* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
365static int check_xflags(unsigned int flags)
366{
367 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
368 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
369 return -EOPNOTSUPP;
370 return 0;
371}
372
e4202ac9
DS		 373/*
374 * Set the xflags from the internal inode flags. The remaining items of fsxattr
375 * are zeroed.
376 */
377static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
378{
379 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
380 struct fsxattr fa;
381
382 memset(&fa, 0, sizeof(fa));
383 fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags);
384
385 if (copy_to_user(arg, &fa, sizeof(fa)))
386 return -EFAULT;
387
388 return 0;
389}
390
025f2121
DS		 391static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
392{
393 struct inode *inode = file_inode(file);
394 struct btrfs_inode *binode = BTRFS_I(inode);
395 struct btrfs_root *root = binode->root;
396 struct btrfs_trans_handle *trans;
397 struct fsxattr fa;
398 unsigned old_flags;
399 unsigned old_i_flags;
400 int ret = 0;
401
402 if (!inode_owner_or_capable(inode))
403 return -EPERM;
404
405 if (btrfs_root_readonly(root))
406 return -EROFS;
407
408 memset(&fa, 0, sizeof(fa));
409 if (copy_from_user(&fa, arg, sizeof(fa)))
410 return -EFAULT;
411
412 ret = check_xflags(fa.fsx_xflags);
413 if (ret)
414 return ret;
415
416 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
417 return -EOPNOTSUPP;
418
419 ret = mnt_want_write_file(file);
420 if (ret)
421 return ret;
422
423 inode_lock(inode);
424
425 old_flags = binode->flags;
426 old_i_flags = inode->i_flags;
427
428 /* We need the capabilities to change append-only or immutable inode */
429 if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) ||
430 (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) &&
431 !capable(CAP_LINUX_IMMUTABLE)) {
432 ret = -EPERM;
433 goto out_unlock;
434 }
435
436 if (fa.fsx_xflags & FS_XFLAG_SYNC)
437 binode->flags |= BTRFS_INODE_SYNC;
438 else
439 binode->flags &= ~BTRFS_INODE_SYNC;
440 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
441 binode->flags |= BTRFS_INODE_IMMUTABLE;
442 else
443 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
444 if (fa.fsx_xflags & FS_XFLAG_APPEND)
445 binode->flags |= BTRFS_INODE_APPEND;
446 else
447 binode->flags &= ~BTRFS_INODE_APPEND;
448 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
449 binode->flags |= BTRFS_INODE_NODUMP;
450 else
451 binode->flags &= ~BTRFS_INODE_NODUMP;
452 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
453 binode->flags |= BTRFS_INODE_NOATIME;
454 else
455 binode->flags &= ~BTRFS_INODE_NOATIME;
456
457 /* 1 item for the inode */
458 trans = btrfs_start_transaction(root, 1);
459 if (IS_ERR(trans)) {
460 ret = PTR_ERR(trans);
461 goto out_unlock;
462 }
463
464 btrfs_sync_inode_flags_to_i_flags(inode);
465 inode_inc_iversion(inode);
466 inode->i_ctime = current_time(inode);
467 ret = btrfs_update_inode(trans, root, inode);
468
469 btrfs_end_transaction(trans);
470
471out_unlock:
472 if (ret) {
473 binode->flags = old_flags;
474 inode->i_flags = old_i_flags;
475 }
476
477 inode_unlock(inode);
478 mnt_drop_write_file(file);
479
480 return ret;
481}
482
6cbff00f
CH		 483static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
484{
496ad9aa 485 struct inode *inode = file_inode(file);
6cbff00f
CH		 486
487 return put_user(inode->i_generation, arg);
488}
f46b5a66 489
f7039b1d
LD		 490static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
491{
0b246afa
JM		 492 struct inode *inode = file_inode(file);
493 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f7039b1d
LD		 494 struct btrfs_device *device;
495 struct request_queue *q;
496 struct fstrim_range range;
497 u64 minlen = ULLONG_MAX;
498 u64 num_devices = 0;
815745cf 499 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 500 int ret;
501
502 if (!capable(CAP_SYS_ADMIN))
503 return -EPERM;
504
1f78160c
XG		 505 rcu_read_lock();
506 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
507 dev_list) {
f7039b1d
LD		 508 if (!device->bdev)
509 continue;
510 q = bdev_get_queue(device->bdev);
511 if (blk_queue_discard(q)) {
512 num_devices++;
50d0446e 513 minlen = min_t(u64, q->limits.discard_granularity,
f7039b1d
LD		 514 minlen);
515 }
516 }
1f78160c 517 rcu_read_unlock();
f4c697e6 518
f7039b1d
LD		 519 if (!num_devices)
520 return -EOPNOTSUPP;
f7039b1d
LD		 521 if (copy_from_user(&range, arg, sizeof(range)))
522 return -EFAULT;
e515c18b
LC		 523 if (range.start > total_bytes ||
524 range.len < fs_info->sb->s_blocksize)
f4c697e6 525 return -EINVAL;
f7039b1d 526
f4c697e6 527 range.len = min(range.len, total_bytes - range.start);
f7039b1d 528 range.minlen = max(range.minlen, minlen);
2ff7e61e 529 ret = btrfs_trim_fs(fs_info, &range);
f7039b1d
LD		 530 if (ret < 0)
531 return ret;
532
533 if (copy_to_user(arg, &range, sizeof(range)))
534 return -EFAULT;
535
536 return 0;
537}
538
dd5f9615
SB		 539int btrfs_is_empty_uuid(u8 *uuid)
540{
46e0f66a
CM		 541 int i;
542
543 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
544 if (uuid[i])
545 return 0;
546 }
547 return 1;
dd5f9615
SB		 548}
549
d5c12070 550static noinline int create_subvol(struct inode *dir,
cb8e7090 551 struct dentry *dentry,
52f75f4f 552 const char *name, int namelen,
6f72c7e2 553 u64 *async_transid,
8696c533 554 struct btrfs_qgroup_inherit *inherit)
f46b5a66 555{
0b246afa 556 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
f46b5a66
CH		 557 struct btrfs_trans_handle *trans;
558 struct btrfs_key key;
49a3c4d9 559 struct btrfs_root_item *root_item;
f46b5a66
CH		 560 struct btrfs_inode_item *inode_item;
561 struct extent_buffer *leaf;
d5c12070 562 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 563 struct btrfs_root *new_root;
d5c12070 564 struct btrfs_block_rsv block_rsv;
c2050a45 565 struct timespec cur_time = current_time(dir);
5662344b 566 struct inode *inode;
f46b5a66
CH		 567 int ret;
568 int err;
569 u64 objectid;
570 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 571 u64 index = 0;
d5c12070 572 u64 qgroup_reserved;
8ea05e3a 573 uuid_le new_uuid;
f46b5a66 574
49a3c4d9
DS		 575 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
576 if (!root_item)
577 return -ENOMEM;
578
0b246afa 579 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
2fbe8c8a 580 if (ret)
49a3c4d9 581 goto fail_free;
6a912213 582
e09fe2d2
QW		 583 /*
584 * Don't create subvolume whose level is not zero. Or qgroup will be
01327610 585 * screwed up since it assumes subvolume qgroup's level to be 0.
e09fe2d2 586 */
49a3c4d9
DS		 587 if (btrfs_qgroup_level(objectid)) {
588 ret = -ENOSPC;
589 goto fail_free;
590 }
e09fe2d2 591
d5c12070 592 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 593 /*
d5c12070
MX		 594 * The same as the snapshot creation, please see the comment
595 * of create_snapshot().
9ed74f2d 596 */
d5c12070 597 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 598 8, &qgroup_reserved, false);
d5c12070 599 if (ret)
49a3c4d9 600 goto fail_free;
d5c12070
MX		 601
602 trans = btrfs_start_transaction(root, 0);
603 if (IS_ERR(trans)) {
604 ret = PTR_ERR(trans);
7775c818 605 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
49a3c4d9 606 goto fail_free;
d5c12070
MX		 607 }
608 trans->block_rsv = &block_rsv;
609 trans->bytes_reserved = block_rsv.size;
f46b5a66 610
0b246afa 611 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 612 if (ret)
613 goto fail;
614
4d75f8a9 615 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 616 if (IS_ERR(leaf)) {
617 ret = PTR_ERR(leaf);
618 goto fail;
619 }
f46b5a66 620
b159fa28 621 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 622 btrfs_set_header_bytenr(leaf, leaf->start);
623 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 624 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 625 btrfs_set_header_owner(leaf, objectid);
626
0b246afa
JM		 627 write_extent_buffer_fsid(leaf, fs_info->fsid);
628 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
f46b5a66
CH		 629 btrfs_mark_buffer_dirty(leaf);
630
49a3c4d9 631 inode_item = &root_item->inode;
3cae210f
QW		 632 btrfs_set_stack_inode_generation(inode_item, 1);
633 btrfs_set_stack_inode_size(inode_item, 3);
634 btrfs_set_stack_inode_nlink(inode_item, 1);
da17066c 635 btrfs_set_stack_inode_nbytes(inode_item,
0b246afa 636 fs_info->nodesize);
3cae210f 637 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 638
49a3c4d9
DS		 639 btrfs_set_root_flags(root_item, 0);
640 btrfs_set_root_limit(root_item, 0);
3cae210f 641 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 642
49a3c4d9
DS		 643 btrfs_set_root_bytenr(root_item, leaf->start);
644 btrfs_set_root_generation(root_item, trans->transid);
645 btrfs_set_root_level(root_item, 0);
646 btrfs_set_root_refs(root_item, 1);
647 btrfs_set_root_used(root_item, leaf->len);
648 btrfs_set_root_last_snapshot(root_item, 0);
f46b5a66 649
49a3c4d9
DS		 650 btrfs_set_root_generation_v2(root_item,
651 btrfs_root_generation(root_item));
8ea05e3a 652 uuid_le_gen(&new_uuid);
49a3c4d9
DS		 653 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
654 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
655 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
656 root_item->ctime = root_item->otime;
657 btrfs_set_root_ctransid(root_item, trans->transid);
658 btrfs_set_root_otransid(root_item, trans->transid);
f46b5a66 659
925baedd 660 btrfs_tree_unlock(leaf);
f46b5a66
CH		 661 free_extent_buffer(leaf);
662 leaf = NULL;
663
49a3c4d9 664 btrfs_set_root_dirid(root_item, new_dirid);
f46b5a66
CH		 665
666 key.objectid = objectid;
5d4f98a2 667 key.offset = 0;
962a298f 668 key.type = BTRFS_ROOT_ITEM_KEY;
0b246afa 669 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
49a3c4d9 670 root_item);
f46b5a66
CH		 671 if (ret)
672 goto fail;
673
76dda93c 674 key.offset = (u64)-1;
0b246afa 675 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
79787eaa 676 if (IS_ERR(new_root)) {
79787eaa 677 ret = PTR_ERR(new_root);
66642832 678 btrfs_abort_transaction(trans, ret);
79787eaa
JM		 679 goto fail;
680 }
76dda93c
YZ		 681
682 btrfs_record_root_in_trans(trans, new_root);
683
63541927 684 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 685 if (ret) {
686 /* We potentially lose an unused inode item here */
66642832 687 btrfs_abort_transaction(trans, ret);
ce598979
MF		 688 goto fail;
689 }
690
f32e48e9
CR		 691 mutex_lock(&new_root->objectid_mutex);
692 new_root->highest_objectid = new_dirid;
693 mutex_unlock(&new_root->objectid_mutex);
694
f46b5a66
CH		 695 /*
696 * insert the directory item
697 */
877574e2 698 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
79787eaa 699 if (ret) {
66642832 700 btrfs_abort_transaction(trans, ret);
79787eaa
JM		 701 goto fail;
702 }
3de4586c
CM		 703
704 ret = btrfs_insert_dir_item(trans, root,
8e7611cf 705 name, namelen, BTRFS_I(dir), &key,
3de4586c 706 BTRFS_FT_DIR, index);
79787eaa 707 if (ret) {
66642832 708 btrfs_abort_transaction(trans, ret);
f46b5a66 709 goto fail;
79787eaa 710 }
0660b5af 711
6ef06d27 712 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
52c26179
YZ		 713 ret = btrfs_update_inode(trans, root, dir);
714 BUG_ON(ret);
715
0b246afa 716 ret = btrfs_add_root_ref(trans, fs_info,
4df27c4d 717 objectid, root->root_key.objectid,
4a0cc7ca 718 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
76dda93c 719 BUG_ON(ret);
f46b5a66 720
0b246afa 721 ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid,
6bccf3ab 722 BTRFS_UUID_KEY_SUBVOL, objectid);
dd5f9615 723 if (ret)
66642832 724 btrfs_abort_transaction(trans, ret);
dd5f9615 725
f46b5a66 726fail:
49a3c4d9 727 kfree(root_item);
d5c12070
MX		 728 trans->block_rsv = NULL;
729 trans->bytes_reserved = 0;
7775c818 730 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
de6e8200 731
72fd032e
SW		 732 if (async_transid) {
733 *async_transid = trans->transid;
3a45bb20 734 err = btrfs_commit_transaction_async(trans, 1);
00d71c9c 735 if (err)
3a45bb20 736 err = btrfs_commit_transaction(trans);
72fd032e 737 } else {
3a45bb20 738 err = btrfs_commit_transaction(trans);
72fd032e 739 }
f46b5a66
CH		 740 if (err && !ret)
741 ret = err;
1a65e24b 742
5662344b
TI		 743 if (!ret) {
744 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 745 if (IS_ERR(inode))
746 return PTR_ERR(inode);
5662344b
TI		 747 d_instantiate(dentry, inode);
748 }
f46b5a66 749 return ret;
49a3c4d9
DS		 750
751fail_free:
752 kfree(root_item);
753 return ret;
f46b5a66
CH		 754}
755
e9662f70 756static int create_snapshot(struct btrfs_root *root, struct inode *dir,
61d7e4cb 757 struct dentry *dentry,
e9662f70
MX		 758 u64 *async_transid, bool readonly,
759 struct btrfs_qgroup_inherit *inherit)
f46b5a66 760{
0b246afa 761 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
2e4bfab9 762 struct inode *inode;
f46b5a66
CH		 763 struct btrfs_pending_snapshot *pending_snapshot;
764 struct btrfs_trans_handle *trans;
2e4bfab9 765 int ret;
f46b5a66 766
27cdeb70 767 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
f46b5a66
CH		 768 return -EINVAL;
769
23269bf5 770 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
a1ee7362
DS		 771 if (!pending_snapshot)
772 return -ENOMEM;
773
b0c0ea63 774 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
23269bf5 775 GFP_KERNEL);
8546b570
DS		 776 pending_snapshot->path = btrfs_alloc_path();
777 if (!pending_snapshot->root_item || !pending_snapshot->path) {
b0c0ea63
DS		 778 ret = -ENOMEM;
779 goto free_pending;
780 }
781
ea14b57f 782 atomic_inc(&root->will_be_snapshotted);
4e857c58 783 smp_mb__after_atomic();
45bac0f3
LB		 784 /* wait for no snapshot writes */
785 wait_event(root->subv_writers->wait,
786 percpu_counter_sum(&root->subv_writers->counter) == 0);
8257b2dc 787
76f32e24 788 ret = btrfs_start_delalloc_inodes(root);
6a03843d 789 if (ret)
a1ee7362 790 goto dec_and_free;
6a03843d 791
6374e57a 792 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
6a03843d 793
66d8f3dd
MX		 794 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
795 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 796 /*
797 * 1 - parent dir inode
798 * 2 - dir entries
799 * 1 - root item
800 * 2 - root ref/backref
801 * 1 - root of snapshot
dd5f9615 802 * 1 - UUID item
d5c12070
MX		 803 */
804 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 805 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 806 &pending_snapshot->qgroup_reserved,
807 false);
d5c12070 808 if (ret)
a1ee7362 809 goto dec_and_free;
d5c12070 810
3de4586c 811 pending_snapshot->dentry = dentry;
f46b5a66 812 pending_snapshot->root = root;
b83cc969 813 pending_snapshot->readonly = readonly;
e9662f70 814 pending_snapshot->dir = dir;
8696c533 815 pending_snapshot->inherit = inherit;
a22285a6 816
d5c12070 817 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 818 if (IS_ERR(trans)) {
819 ret = PTR_ERR(trans);
820 goto fail;
821 }
822
0b246afa 823 spin_lock(&fs_info->trans_lock);
f46b5a66
CH		 824 list_add(&pending_snapshot->list,
825 &trans->transaction->pending_snapshots);
0b246afa 826 spin_unlock(&fs_info->trans_lock);
72fd032e
SW		 827 if (async_transid) {
828 *async_transid = trans->transid;
3a45bb20 829 ret = btrfs_commit_transaction_async(trans, 1);
00d71c9c 830 if (ret)
3a45bb20 831 ret = btrfs_commit_transaction(trans);
72fd032e 832 } else {
3a45bb20 833 ret = btrfs_commit_transaction(trans);
72fd032e 834 }
aec8030a 835 if (ret)
c37b2b62 836 goto fail;
a22285a6
YZ		 837
838 ret = pending_snapshot->error;
839 if (ret)
840 goto fail;
841
d3797308
CM		 842 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
843 if (ret)
844 goto fail;
845
2b0143b5 846 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
2e4bfab9
YZ		 847 if (IS_ERR(inode)) {
848 ret = PTR_ERR(inode);
849 goto fail;
850 }
5662344b 851
2e4bfab9
YZ		 852 d_instantiate(dentry, inode);
853 ret = 0;
854fail:
7775c818 855 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
a1ee7362 856dec_and_free:
ea14b57f 857 if (atomic_dec_and_test(&root->will_be_snapshotted))
4625956a 858 wake_up_var(&root->will_be_snapshotted);
b0c0ea63
DS		 859free_pending:
860 kfree(pending_snapshot->root_item);
8546b570 861 btrfs_free_path(pending_snapshot->path);
a1ee7362
DS		 862 kfree(pending_snapshot);
863
f46b5a66
CH		 864 return ret;
865}
866
4260f7c7
SW		 867/* copy of may_delete in fs/namei.c()
868 * Check whether we can remove a link victim from directory dir, check
869 * whether the type of victim is right.
870 * 1. We can't do it if dir is read-only (done in permission())
871 * 2. We should have write and exec permissions on dir
872 * 3. We can't remove anything from append-only dir
873 * 4. We can't do anything with immutable dir (done in permission())
874 * 5. If the sticky bit on dir is set we should either
875 * a. be owner of dir, or
876 * b. be owner of victim, or
877 * c. have CAP_FOWNER capability
01327610 878 * 6. If the victim is append-only or immutable we can't do anything with
4260f7c7
SW		 879 * links pointing to it.
880 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
881 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
882 * 9. We can't remove a root or mountpoint.
883 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
884 * nfs_async_unlink().
885 */
886
67871254 887static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 888{
889 int error;
890
2b0143b5 891 if (d_really_is_negative(victim))
4260f7c7
SW		 892 return -ENOENT;
893
2b0143b5 894 BUG_ON(d_inode(victim->d_parent) != dir);
4fa6b5ec 895 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 896
897 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
898 if (error)
899 return error;
900 if (IS_APPEND(dir))
901 return -EPERM;
2b0143b5
DH		 902 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
903 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
4260f7c7
SW		 904 return -EPERM;
905 if (isdir) {
e36cb0b8 906 if (!d_is_dir(victim))
4260f7c7
SW		 907 return -ENOTDIR;
908 if (IS_ROOT(victim))
909 return -EBUSY;
e36cb0b8 910 } else if (d_is_dir(victim))
4260f7c7
SW		 911 return -EISDIR;
912 if (IS_DEADDIR(dir))
913 return -ENOENT;
914 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
915 return -EBUSY;
916 return 0;
917}
918
cb8e7090
CH		 919/* copy of may_create in fs/namei.c() */
920static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
921{
2b0143b5 922 if (d_really_is_positive(child))
cb8e7090
CH		 923 return -EEXIST;
924 if (IS_DEADDIR(dir))
925 return -ENOENT;
926 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
927}
928
929/*
930 * Create a new subvolume below @parent. This is largely modeled after
931 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
932 * inside this filesystem so it's quite a bit simpler.
933 */
92872094 934static noinline int btrfs_mksubvol(const struct path *parent,
52f75f4f 935 const char *name, int namelen,
72fd032e 936 struct btrfs_root *snap_src,
6f72c7e2 937 u64 *async_transid, bool readonly,
8696c533 938 struct btrfs_qgroup_inherit *inherit)
cb8e7090 939{
0b246afa
JM		 940 struct inode *dir = d_inode(parent->dentry);
941 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
cb8e7090
CH		 942 struct dentry *dentry;
943 int error;
944
00235411
AV		 945 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
946 if (error == -EINTR)
947 return error;
cb8e7090
CH		 948
949 dentry = lookup_one_len(name, parent->dentry, namelen);
950 error = PTR_ERR(dentry);
951 if (IS_ERR(dentry))
952 goto out_unlock;
953
76dda93c 954 error = btrfs_may_create(dir, dentry);
cb8e7090 955 if (error)
a874a63e 956 goto out_dput;
cb8e7090 957
9c52057c
CM		 958 /*
959 * even if this name doesn't exist, we may get hash collisions.
960 * check for them now when we can safely fail
961 */
962 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
963 dir->i_ino, name,
964 namelen);
965 if (error)
966 goto out_dput;
967
0b246afa 968 down_read(&fs_info->subvol_sem);
76dda93c
YZ		 969
970 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
971 goto out_up_read;
972
3de4586c 973 if (snap_src) {
61d7e4cb 974 error = create_snapshot(snap_src, dir, dentry,
6f72c7e2 975 async_transid, readonly, inherit);
3de4586c 976 } else {
d5c12070
MX		 977 error = create_subvol(dir, dentry, name, namelen,
978 async_transid, inherit);
3de4586c 979 }
76dda93c
YZ		 980 if (!error)
981 fsnotify_mkdir(dir, dentry);
982out_up_read:
0b246afa 983 up_read(&fs_info->subvol_sem);
cb8e7090
CH		 984out_dput:
985 dput(dentry);
986out_unlock:
5955102c 987 inode_unlock(dir);
cb8e7090
CH		 988 return error;
989}
990
4cb5300b
CM		 991/*
992 * When we're defragging a range, we don't want to kick it off again
993 * if it is really just waiting for delalloc to send it down.
994 * If we find a nice big extent or delalloc range for the bytes in the
995 * file you want to defrag, we return 0 to let you know to skip this
996 * part of the file
997 */
aab110ab 998static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
4cb5300b
CM		 999{
1000 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1001 struct extent_map *em = NULL;
1002 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1003 u64 end;
1004
1005 read_lock(&em_tree->lock);
09cbfeaf 1006 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
4cb5300b
CM		 1007 read_unlock(&em_tree->lock);
1008
1009 if (em) {
1010 end = extent_map_end(em);
1011 free_extent_map(em);
1012 if (end - offset > thresh)
1013 return 0;
1014 }
1015 /* if we already have a nice delalloc here, just stop */
1016 thresh /= 2;
1017 end = count_range_bits(io_tree, &offset, offset + thresh,
1018 thresh, EXTENT_DELALLOC, 1);
1019 if (end >= thresh)
1020 return 0;
1021 return 1;
1022}
1023
1024/*
1025 * helper function to walk through a file and find extents
1026 * newer than a specific transid, and smaller than thresh.
1027 *
1028 * This is used by the defragging code to find new and small
1029 * extents
1030 */
1031static int find_new_extents(struct btrfs_root *root,
1032 struct inode *inode, u64 newer_than,
aab110ab 1033 u64 *off, u32 thresh)
4cb5300b
CM		 1034{
1035 struct btrfs_path *path;
1036 struct btrfs_key min_key;
4cb5300b
CM		 1037 struct extent_buffer *leaf;
1038 struct btrfs_file_extent_item *extent;
1039 int type;
1040 int ret;
4a0cc7ca 1041 u64 ino = btrfs_ino(BTRFS_I(inode));
4cb5300b
CM		 1042
1043 path = btrfs_alloc_path();
1044 if (!path)
1045 return -ENOMEM;
1046
a4689d2b 1047 min_key.objectid = ino;
4cb5300b
CM		 1048 min_key.type = BTRFS_EXTENT_DATA_KEY;
1049 min_key.offset = *off;
1050
67871254 1051 while (1) {
6174d3cb 1052 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 1053 if (ret != 0)
1054 goto none;
f094c9bd 1055process_slot:
a4689d2b 1056 if (min_key.objectid != ino)
4cb5300b
CM		 1057 goto none;
1058 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1059 goto none;
1060
1061 leaf = path->nodes[0];
1062 extent = btrfs_item_ptr(leaf, path->slots[0],
1063 struct btrfs_file_extent_item);
1064
1065 type = btrfs_file_extent_type(leaf, extent);
1066 if (type == BTRFS_FILE_EXTENT_REG &&
1067 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1068 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1069 *off = min_key.offset;
1070 btrfs_free_path(path);
1071 return 0;
1072 }
1073
f094c9bd
FM		 1074 path->slots[0]++;
1075 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1076 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1077 goto process_slot;
1078 }
1079
4cb5300b
CM		 1080 if (min_key.offset == (u64)-1)
1081 goto none;
1082
1083 min_key.offset++;
1084 btrfs_release_path(path);
1085 }
1086none:
1087 btrfs_free_path(path);
1088 return -ENOENT;
1089}
1090
6c282eb4 1091static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 1092{
1093 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 1094 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1095 struct extent_map *em;
09cbfeaf 1096 u64 len = PAGE_SIZE;
17ce6ef8 1097
6c282eb4
LZ		 1098 /*
1099 * hopefully we have this extent in the tree already, try without
1100 * the full extent lock
1101 */
17ce6ef8 1102 read_lock(&em_tree->lock);
6c282eb4 1103 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 1104 read_unlock(&em_tree->lock);
1105
6c282eb4 1106 if (!em) {
308d9800
FM		 1107 struct extent_state *cached = NULL;
1108 u64 end = start + len - 1;
1109
6c282eb4 1110 /* get the big lock and read metadata off disk */
ff13db41 1111 lock_extent_bits(io_tree, start, end, &cached);
fc4f21b1 1112 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
e43bbe5e 1113 unlock_extent_cached(io_tree, start, end, &cached);
6c282eb4
LZ		 1114
1115 if (IS_ERR(em))
1116 return NULL;
1117 }
1118
1119 return em;
1120}
17ce6ef8 1121
6c282eb4
LZ		 1122static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1123{
1124 struct extent_map *next;
1125 bool ret = true;
1126
1127 /* this is the last extent */
1128 if (em->start + em->len >= i_size_read(inode))
1129 return false;
1130
1131 next = defrag_lookup_extent(inode, em->start + em->len);
e9512d72
CM		 1132 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1133 ret = false;
1134 else if ((em->block_start + em->block_len == next->block_start) &&
ee22184b 1135 (em->block_len > SZ_128K && next->block_len > SZ_128K))
6c282eb4
LZ		 1136 ret = false;
1137
1138 free_extent_map(next);
17ce6ef8
LB		 1139 return ret;
1140}
1141
aab110ab 1142static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
a43a2111
AM		 1143 u64 *last_len, u64 *skip, u64 *defrag_end,
1144 int compress)
940100a4 1145{
6c282eb4 1146 struct extent_map *em;
940100a4 1147 int ret = 1;
6c282eb4 1148 bool next_mergeable = true;
4a3560c4 1149 bool prev_mergeable = true;
940100a4
CM		 1150
1151 /*
008873ea 1152 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1153 * defragging it
1154 */
1155 if (start < *defrag_end)
1156 return 1;
1157
1158 *skip = 0;
1159
6c282eb4
LZ		 1160 em = defrag_lookup_extent(inode, start);
1161 if (!em)
1162 return 0;
940100a4
CM		 1163
1164 /* this will cover holes, and inline extents */
17ce6ef8 1165 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1166 ret = 0;
17ce6ef8
LB		 1167 goto out;
1168 }
1169
4a3560c4
LB		 1170 if (!*defrag_end)
1171 prev_mergeable = false;
1172
6c282eb4 1173 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1174 /*
6c282eb4
LZ		 1175 * we hit a real extent, if it is big or the next extent is not a
1176 * real extent, don't bother defragging it
940100a4 1177 */
a43a2111 1178 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
4a3560c4 1179 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
940100a4 1180 ret = 0;
17ce6ef8 1181out:
940100a4
CM		 1182 /*
1183 * last_len ends up being a counter of how many bytes we've defragged.
1184 * every time we choose not to defrag an extent, we reset *last_len
1185 * so that the next tiny extent will force a defrag.
1186 *
1187 * The end result of this is that tiny extents before a single big
1188 * extent will force at least part of that big extent to be defragged.
1189 */
1190 if (ret) {
940100a4
CM		 1191 *defrag_end = extent_map_end(em);
1192 } else {
1193 *last_len = 0;
1194 *skip = extent_map_end(em);
1195 *defrag_end = 0;
1196 }
1197
1198 free_extent_map(em);
1199 return ret;
1200}
1201
4cb5300b
CM		 1202/*
1203 * it doesn't do much good to defrag one or two pages
1204 * at a time. This pulls in a nice chunk of pages
1205 * to COW and defrag.
1206 *
1207 * It also makes sure the delalloc code has enough
1208 * dirty data to avoid making new small extents as part
1209 * of the defrag
1210 *
1211 * It's a good idea to start RA on this range
1212 * before calling this.
1213 */
1214static int cluster_pages_for_defrag(struct inode *inode,
1215 struct page **pages,
1216 unsigned long start_index,
c41570c9 1217 unsigned long num_pages)
f46b5a66 1218{
4cb5300b
CM		 1219 unsigned long file_end;
1220 u64 isize = i_size_read(inode);
1221 u64 page_start;
1222 u64 page_end;
1f12bd06 1223 u64 page_cnt;
4cb5300b
CM		 1224 int ret;
1225 int i;
1226 int i_done;
3eaa2885 1227 struct btrfs_ordered_extent *ordered;
4cb5300b 1228 struct extent_state *cached_state = NULL;
600a45e1 1229 struct extent_io_tree *tree;
364ecf36 1230 struct extent_changeset *data_reserved = NULL;
3b16a4e3 1231 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1232
09cbfeaf 1233 file_end = (isize - 1) >> PAGE_SHIFT;
1f12bd06
LB		 1234 if (!isize || start_index > file_end)
1235 return 0;
1236
1237 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b 1238
364ecf36 1239 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
09cbfeaf
KS		 1240 start_index << PAGE_SHIFT,
1241 page_cnt << PAGE_SHIFT);
4cb5300b
CM		 1242 if (ret)
1243 return ret;
4cb5300b 1244 i_done = 0;
600a45e1 1245 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1246
1247 /* step one, lock all the pages */
1f12bd06 1248 for (i = 0; i < page_cnt; i++) {
4cb5300b 1249 struct page *page;
600a45e1 1250again:
a94733d0 1251 page = find_or_create_page(inode->i_mapping,
600a45e1 1252 start_index + i, mask);
4cb5300b
CM		 1253 if (!page)
1254 break;
1255
600a45e1 1256 page_start = page_offset(page);
09cbfeaf 1257 page_end = page_start + PAGE_SIZE - 1;
600a45e1 1258 while (1) {
308d9800 1259 lock_extent_bits(tree, page_start, page_end,
ff13db41 1260 &cached_state);
600a45e1
MX		 1261 ordered = btrfs_lookup_ordered_extent(inode,
1262 page_start);
308d9800 1263 unlock_extent_cached(tree, page_start, page_end,
e43bbe5e 1264 &cached_state);
600a45e1
MX		 1265 if (!ordered)
1266 break;
1267
1268 unlock_page(page);
1269 btrfs_start_ordered_extent(inode, ordered, 1);
1270 btrfs_put_ordered_extent(ordered);
1271 lock_page(page);
1f12bd06
LB		 1272 /*
1273 * we unlocked the page above, so we need check if
1274 * it was released or not.
1275 */
1276 if (page->mapping != inode->i_mapping) {
1277 unlock_page(page);
09cbfeaf 1278 put_page(page);
1f12bd06
LB		 1279 goto again;
1280 }
600a45e1
MX		 1281 }
1282
4cb5300b
CM		 1283 if (!PageUptodate(page)) {
1284 btrfs_readpage(NULL, page);
1285 lock_page(page);
1286 if (!PageUptodate(page)) {
1287 unlock_page(page);
09cbfeaf 1288 put_page(page);
4cb5300b
CM		 1289 ret = -EIO;
1290 break;
1291 }
1292 }
600a45e1 1293
600a45e1
MX		 1294 if (page->mapping != inode->i_mapping) {
1295 unlock_page(page);
09cbfeaf 1296 put_page(page);
600a45e1
MX		 1297 goto again;
1298 }
1299
4cb5300b
CM		 1300 pages[i] = page;
1301 i_done++;
1302 }
1303 if (!i_done || ret)
1304 goto out;
1305
1751e8a6 1306 if (!(inode->i_sb->s_flags & SB_ACTIVE))
4cb5300b
CM		 1307 goto out;
1308
1309 /*
1310 * so now we have a nice long stream of locked
1311 * and up to date pages, lets wait on them
1312 */
1313 for (i = 0; i < i_done; i++)
1314 wait_on_page_writeback(pages[i]);
1315
1316 page_start = page_offset(pages[0]);
09cbfeaf 1317 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
4cb5300b
CM		 1318
1319 lock_extent_bits(&BTRFS_I(inode)->io_tree,
ff13db41 1320 page_start, page_end - 1, &cached_state);
4cb5300b
CM		 1321 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1322 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b 1323 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
ae0f1625 1324 &cached_state);
4cb5300b 1325
1f12bd06 1326 if (i_done != page_cnt) {
9e0baf60
JB		 1327 spin_lock(&BTRFS_I(inode)->lock);
1328 BTRFS_I(inode)->outstanding_extents++;
1329 spin_unlock(&BTRFS_I(inode)->lock);
bc42bda2 1330 btrfs_delalloc_release_space(inode, data_reserved,
09cbfeaf 1331 start_index << PAGE_SHIFT,
43b18595 1332 (page_cnt - i_done) << PAGE_SHIFT, true);
4cb5300b
CM		 1333 }
1334
1335
9e8a4a8b 1336 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
018ed4f7 1337 &cached_state);
4cb5300b
CM		 1338
1339 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
e43bbe5e 1340 page_start, page_end - 1, &cached_state);
4cb5300b
CM		 1341
1342 for (i = 0; i < i_done; i++) {
1343 clear_page_dirty_for_io(pages[i]);
1344 ClearPageChecked(pages[i]);
1345 set_page_extent_mapped(pages[i]);
1346 set_page_dirty(pages[i]);
1347 unlock_page(pages[i]);
09cbfeaf 1348 put_page(pages[i]);
4cb5300b 1349 }
43b18595
QW		 1350 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1351 false);
364ecf36 1352 extent_changeset_free(data_reserved);
4cb5300b
CM		 1353 return i_done;
1354out:
1355 for (i = 0; i < i_done; i++) {
1356 unlock_page(pages[i]);
09cbfeaf 1357 put_page(pages[i]);
4cb5300b 1358 }
bc42bda2 1359 btrfs_delalloc_release_space(inode, data_reserved,
09cbfeaf 1360 start_index << PAGE_SHIFT,
43b18595
QW		 1361 page_cnt << PAGE_SHIFT, true);
1362 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1363 true);
364ecf36 1364 extent_changeset_free(data_reserved);
4cb5300b
CM		 1365 return ret;
1366
1367}
1368
1369int btrfs_defrag_file(struct inode *inode, struct file *file,
1370 struct btrfs_ioctl_defrag_range_args *range,
1371 u64 newer_than, unsigned long max_to_defrag)
1372{
0b246afa 1373 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4cb5300b 1374 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1375 struct file_ra_state *ra = NULL;
f46b5a66 1376 unsigned long last_index;
151a31b2 1377 u64 isize = i_size_read(inode);
940100a4
CM		 1378 u64 last_len = 0;
1379 u64 skip = 0;
1380 u64 defrag_end = 0;
4cb5300b 1381 u64 newer_off = range->start;
f46b5a66 1382 unsigned long i;
008873ea 1383 unsigned long ra_index = 0;
f46b5a66 1384 int ret;
4cb5300b 1385 int defrag_count = 0;
1a419d85 1386 int compress_type = BTRFS_COMPRESS_ZLIB;
aab110ab 1387 u32 extent_thresh = range->extent_thresh;
09cbfeaf 1388 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
c41570c9 1389 unsigned long cluster = max_cluster;
ee22184b 1390 u64 new_align = ~((u64)SZ_128K - 1);
4cb5300b 1391 struct page **pages = NULL;
1e2ef46d 1392 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
4cb5300b 1393
0abd5b17
LB		 1394 if (isize == 0)
1395 return 0;
1396
1397 if (range->start >= isize)
1398 return -EINVAL;
1a419d85 1399
1e2ef46d 1400 if (do_compress) {
1a419d85
LZ		 1401 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1402 return -EINVAL;
1403 if (range->compress_type)
1404 compress_type = range->compress_type;
1405 }
f46b5a66 1406
0abd5b17 1407 if (extent_thresh == 0)
ee22184b 1408 extent_thresh = SZ_256K;
940100a4 1409
4cb5300b 1410 /*
0a52d108
DS		 1411 * If we were not given a file, allocate a readahead context. As
1412 * readahead is just an optimization, defrag will work without it so
1413 * we don't error out.
4cb5300b
CM		 1414 */
1415 if (!file) {
63e727ec 1416 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
0a52d108
DS		 1417 if (ra)
1418 file_ra_state_init(ra, inode->i_mapping);
4cb5300b
CM		 1419 } else {
1420 ra = &file->f_ra;
1421 }
1422
63e727ec 1423 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
4cb5300b
CM		 1424 if (!pages) {
1425 ret = -ENOMEM;
1426 goto out_ra;
1427 }
1428
1429 /* find the last page to defrag */
1e701a32 1430 if (range->start + range->len > range->start) {
151a31b2 1431 last_index = min_t(u64, isize - 1,
09cbfeaf 1432 range->start + range->len - 1) >> PAGE_SHIFT;
1e701a32 1433 } else {
09cbfeaf 1434 last_index = (isize - 1) >> PAGE_SHIFT;
1e701a32
CM		 1435 }
1436
4cb5300b
CM		 1437 if (newer_than) {
1438 ret = find_new_extents(root, inode, newer_than,
ee22184b 1439 &newer_off, SZ_64K);
4cb5300b
CM		 1440 if (!ret) {
1441 range->start = newer_off;
1442 /*
1443 * we always align our defrag to help keep
1444 * the extents in the file evenly spaced
1445 */
09cbfeaf 1446 i = (newer_off & new_align) >> PAGE_SHIFT;
4cb5300b
CM		 1447 } else
1448 goto out_ra;
1449 } else {
09cbfeaf 1450 i = range->start >> PAGE_SHIFT;
4cb5300b
CM		 1451 }
1452 if (!max_to_defrag)
070034bd 1453 max_to_defrag = last_index - i + 1;
4cb5300b 1454
2a0f7f57
LZ		 1455 /*
1456 * make writeback starts from i, so the defrag range can be
1457 * written sequentially.
1458 */
1459 if (i < inode->i_mapping->writeback_index)
1460 inode->i_mapping->writeback_index = i;
1461
f7f43cc8 1462 while (i <= last_index && defrag_count < max_to_defrag &&
09cbfeaf 1463 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
4cb5300b
CM		 1464 /*
1465 * make sure we stop running if someone unmounts
1466 * the FS
1467 */
1751e8a6 1468 if (!(inode->i_sb->s_flags & SB_ACTIVE))
4cb5300b
CM		 1469 break;
1470
0b246afa
JM		 1471 if (btrfs_defrag_cancelled(fs_info)) {
1472 btrfs_debug(fs_info, "defrag_file cancelled");
210549eb
DS		 1473 ret = -EAGAIN;
1474 break;
1475 }
1476
09cbfeaf 1477 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
6c282eb4 1478 extent_thresh, &last_len, &skip,
1e2ef46d 1479 &defrag_end, do_compress)){
940100a4
CM		 1480 unsigned long next;
1481 /*
1482 * the should_defrag function tells us how much to skip
1483 * bump our counter by the suggested amount
1484 */
09cbfeaf 1485 next = DIV_ROUND_UP(skip, PAGE_SIZE);
940100a4
CM		 1486 i = max(i + 1, next);
1487 continue;
1488 }
008873ea
LZ		 1489
1490 if (!newer_than) {
09cbfeaf
KS		 1491 cluster = (PAGE_ALIGN(defrag_end) >>
1492 PAGE_SHIFT) - i;
008873ea
LZ		 1493 cluster = min(cluster, max_cluster);
1494 } else {
1495 cluster = max_cluster;
1496 }
1497
008873ea
LZ		 1498 if (i + cluster > ra_index) {
1499 ra_index = max(i, ra_index);
0a52d108 1500 if (ra)
d3c0bab5
DS		 1501 page_cache_sync_readahead(inode->i_mapping, ra,
1502 file, ra_index, cluster);
e4826a5b 1503 ra_index += cluster;
008873ea 1504 }
940100a4 1505
5955102c 1506 inode_lock(inode);
1e2ef46d 1507 if (do_compress)
eec63c65 1508 BTRFS_I(inode)->defrag_compress = compress_type;
008873ea 1509 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8 1510 if (ret < 0) {
5955102c 1511 inode_unlock(inode);
4cb5300b 1512 goto out_ra;
ecb8bea8 1513 }
4cb5300b
CM		 1514
1515 defrag_count += ret;
d0e1d66b 1516 balance_dirty_pages_ratelimited(inode->i_mapping);
5955102c 1517 inode_unlock(inode);
4cb5300b
CM		 1518
1519 if (newer_than) {
1520 if (newer_off == (u64)-1)
1521 break;
1522
e1f041e1
LB		 1523 if (ret > 0)
1524 i += ret;
1525
4cb5300b 1526 newer_off = max(newer_off + 1,
09cbfeaf 1527 (u64)i << PAGE_SHIFT);
4cb5300b 1528
ee22184b
BL		 1529 ret = find_new_extents(root, inode, newer_than,
1530 &newer_off, SZ_64K);
4cb5300b
CM		 1531 if (!ret) {
1532 range->start = newer_off;
09cbfeaf 1533 i = (newer_off & new_align) >> PAGE_SHIFT;
4cb5300b
CM		 1534 } else {
1535 break;
f46b5a66 1536 }
4cb5300b 1537 } else {
008873ea 1538 if (ret > 0) {
cbcc8326 1539 i += ret;
09cbfeaf 1540 last_len += ret << PAGE_SHIFT;
008873ea 1541 } else {
cbcc8326 1542 i++;
008873ea
LZ		 1543 last_len = 0;
1544 }
f46b5a66 1545 }
f46b5a66
CH		 1546 }
1547
dec8ef90 1548 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1549 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1550 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1551 &BTRFS_I(inode)->runtime_flags))
1552 filemap_flush(inode->i_mapping);
1553 }
1e701a32 1554
1a419d85 1555 if (range->compress_type == BTRFS_COMPRESS_LZO) {
0b246afa 1556 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
5c1aab1d
NT		 1557 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1558 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1a419d85
LZ		 1559 }
1560
60ccf82f 1561 ret = defrag_count;
940100a4 1562
4cb5300b 1563out_ra:
1e2ef46d 1564 if (do_compress) {
5955102c 1565 inode_lock(inode);
eec63c65 1566 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
5955102c 1567 inode_unlock(inode);
633085c7 1568 }
4cb5300b
CM		 1569 if (!file)
1570 kfree(ra);
1571 kfree(pages);
940100a4 1572 return ret;
f46b5a66
CH		 1573}
1574
198605a8 1575static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1576 void __user *arg)
f46b5a66 1577{
0b246afa
JM		 1578 struct inode *inode = file_inode(file);
1579 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66
CH		 1580 u64 new_size;
1581 u64 old_size;
1582 u64 devid = 1;
0b246afa 1583 struct btrfs_root *root = BTRFS_I(inode)->root;
f46b5a66
CH		 1584 struct btrfs_ioctl_vol_args *vol_args;
1585 struct btrfs_trans_handle *trans;
1586 struct btrfs_device *device = NULL;
1587 char *sizestr;
9a40f122 1588 char *retptr;
f46b5a66
CH		 1589 char *devstr = NULL;
1590 int ret = 0;
f46b5a66
CH		 1591 int mod = 0;
1592
e441d54d
CM		 1593 if (!capable(CAP_SYS_ADMIN))
1594 return -EPERM;
1595
198605a8
MX		 1596 ret = mnt_want_write_file(file);
1597 if (ret)
1598 return ret;
1599
171938e5 1600 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
97547676 1601 mnt_drop_write_file(file);
e57138b3 1602 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1603 }
1604
dae7b665 1605 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1606 if (IS_ERR(vol_args)) {
1607 ret = PTR_ERR(vol_args);
1608 goto out;
1609 }
5516e595
MF		 1610
1611 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1612
f46b5a66
CH		 1613 sizestr = vol_args->name;
1614 devstr = strchr(sizestr, ':');
1615 if (devstr) {
f46b5a66
CH		 1616 sizestr = devstr + 1;
1617 *devstr = '\0';
1618 devstr = vol_args->name;
58dfae63
Z		 1619 ret = kstrtoull(devstr, 10, &devid);
1620 if (ret)
1621 goto out_free;
dfd79829
MX		 1622 if (!devid) {
1623 ret = -EINVAL;
1624 goto out_free;
1625 }
0b246afa 1626 btrfs_info(fs_info, "resizing devid %llu", devid);
f46b5a66 1627 }
dba60f3f 1628
0b246afa 1629 device = btrfs_find_device(fs_info, devid, NULL, NULL);
f46b5a66 1630 if (!device) {
0b246afa
JM		 1631 btrfs_info(fs_info, "resizer unable to find device %llu",
1632 devid);
dfd79829 1633 ret = -ENODEV;
c9e9f97b 1634 goto out_free;
f46b5a66 1635 }
dba60f3f 1636
ebbede42 1637 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
0b246afa 1638 btrfs_info(fs_info,
efe120a0 1639 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1640 devid);
dfd79829 1641 ret = -EPERM;
4e42ae1b
LB		 1642 goto out_free;
1643 }
1644
f46b5a66
CH		 1645 if (!strcmp(sizestr, "max"))
1646 new_size = device->bdev->bd_inode->i_size;
1647 else {
1648 if (sizestr[0] == '-') {
1649 mod = -1;
1650 sizestr++;
1651 } else if (sizestr[0] == '+') {
1652 mod = 1;
1653 sizestr++;
1654 }
9a40f122
GH		 1655 new_size = memparse(sizestr, &retptr);
1656 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1657 ret = -EINVAL;
c9e9f97b 1658 goto out_free;
f46b5a66
CH		 1659 }
1660 }
1661
401e29c1 1662 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
dfd79829 1663 ret = -EPERM;
63a212ab
SB		 1664 goto out_free;
1665 }
1666
7cc8e58d 1667 old_size = btrfs_device_get_total_bytes(device);
f46b5a66
CH		 1668
1669 if (mod < 0) {
1670 if (new_size > old_size) {
1671 ret = -EINVAL;
c9e9f97b 1672 goto out_free;
f46b5a66
CH		 1673 }
1674 new_size = old_size - new_size;
1675 } else if (mod > 0) {
eb8052e0 1676 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1677 ret = -ERANGE;
eb8052e0
WF		 1678 goto out_free;
1679 }
f46b5a66
CH		 1680 new_size = old_size + new_size;
1681 }
1682
ee22184b 1683 if (new_size < SZ_256M) {
f46b5a66 1684 ret = -EINVAL;
c9e9f97b 1685 goto out_free;
f46b5a66
CH		 1686 }
1687 if (new_size > device->bdev->bd_inode->i_size) {
1688 ret = -EFBIG;
c9e9f97b 1689 goto out_free;
f46b5a66
CH		 1690 }
1691
47f08b96 1692 new_size = round_down(new_size, fs_info->sectorsize);
f46b5a66 1693
0b246afa
JM		 1694 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1695 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1696
1697 if (new_size > old_size) {
a22285a6 1698 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1699 if (IS_ERR(trans)) {
1700 ret = PTR_ERR(trans);
c9e9f97b 1701 goto out_free;
98d5dc13 1702 }
f46b5a66 1703 ret = btrfs_grow_device(trans, device, new_size);
3a45bb20 1704 btrfs_commit_transaction(trans);
ece7d20e 1705 } else if (new_size < old_size) {
f46b5a66 1706 ret = btrfs_shrink_device(device, new_size);
0253f40e 1707 } /* equal, nothing need to do */
f46b5a66 1708
c9e9f97b 1709out_free:
f46b5a66 1710 kfree(vol_args);
c9e9f97b 1711out:
171938e5 1712 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
18f39c41 1713 mnt_drop_write_file(file);
f46b5a66
CH		 1714 return ret;
1715}
1716
72fd032e 1717static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
52f75f4f 1718 const char *name, unsigned long fd, int subvol,
6f72c7e2 1719 u64 *transid, bool readonly,
8696c533 1720 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1721{
f46b5a66 1722 int namelen;
3de4586c 1723 int ret = 0;
f46b5a66 1724
325c50e3
JM		 1725 if (!S_ISDIR(file_inode(file)->i_mode))
1726 return -ENOTDIR;
1727
a874a63e
LB		 1728 ret = mnt_want_write_file(file);
1729 if (ret)
1730 goto out;
1731
72fd032e
SW		 1732 namelen = strlen(name);
1733 if (strchr(name, '/')) {
f46b5a66 1734 ret = -EINVAL;
a874a63e 1735 goto out_drop_write;
f46b5a66
CH		 1736 }
1737
16780cab
CM		 1738 if (name[0] == '.' &&
1739 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1740 ret = -EEXIST;
a874a63e 1741 goto out_drop_write;
16780cab
CM		 1742 }
1743
3de4586c 1744 if (subvol) {
72fd032e 1745 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1746 NULL, transid, readonly, inherit);
cb8e7090 1747 } else {
2903ff01 1748 struct fd src = fdget(fd);
3de4586c 1749 struct inode *src_inode;
2903ff01 1750 if (!src.file) {
3de4586c 1751 ret = -EINVAL;
a874a63e 1752 goto out_drop_write;
3de4586c
CM		 1753 }
1754
496ad9aa
AV		 1755 src_inode = file_inode(src.file);
1756 if (src_inode->i_sb != file_inode(file)->i_sb) {
c79b4713 1757 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
efe120a0 1758 "Snapshot src from another FS");
23ad5b17 1759 ret = -EXDEV;
d0242061
DS		 1760 } else if (!inode_owner_or_capable(src_inode)) {
1761 /*
1762 * Subvolume creation is not restricted, but snapshots
1763 * are limited to own subvolumes only
1764 */
1765 ret = -EPERM;
ecd18815
AV		 1766 } else {
1767 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1768 BTRFS_I(src_inode)->root,
1769 transid, readonly, inherit);
3de4586c 1770 }
2903ff01 1771 fdput(src);
cb8e7090 1772 }
a874a63e
LB		 1773out_drop_write:
1774 mnt_drop_write_file(file);
f46b5a66 1775out:
72fd032e
SW		 1776 return ret;
1777}
1778
1779static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1780 void __user *arg, int subvol)
72fd032e 1781{
fa0d2b9b 1782 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1783 int ret;
1784
325c50e3
JM		 1785 if (!S_ISDIR(file_inode(file)->i_mode))
1786 return -ENOTDIR;
1787
fa0d2b9b
LZ		 1788 vol_args = memdup_user(arg, sizeof(*vol_args));
1789 if (IS_ERR(vol_args))
1790 return PTR_ERR(vol_args);
1791 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1792
fa0d2b9b 1793 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1794 vol_args->fd, subvol,
6f72c7e2 1795 NULL, false, NULL);
fdfb1e4f 1796
fa0d2b9b
LZ		 1797 kfree(vol_args);
1798 return ret;
1799}
fdfb1e4f 1800
fa0d2b9b
LZ		 1801static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1802 void __user *arg, int subvol)
1803{
1804 struct btrfs_ioctl_vol_args_v2 *vol_args;
1805 int ret;
1806 u64 transid = 0;
1807 u64 *ptr = NULL;
b83cc969 1808 bool readonly = false;
6f72c7e2 1809 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1810
325c50e3
JM		 1811 if (!S_ISDIR(file_inode(file)->i_mode))
1812 return -ENOTDIR;
1813
fa0d2b9b
LZ		 1814 vol_args = memdup_user(arg, sizeof(*vol_args));
1815 if (IS_ERR(vol_args))
1816 return PTR_ERR(vol_args);
1817 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1818
b83cc969 1819 if (vol_args->flags &
6f72c7e2
AJ		 1820 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1821 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1822 ret = -EOPNOTSUPP;
c47ca32d 1823 goto free_args;
72fd032e 1824 }
fa0d2b9b
LZ		 1825
1826 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1827 ptr = &transid;
b83cc969
LZ		 1828 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1829 readonly = true;
6f72c7e2 1830 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
09cbfeaf 1831 if (vol_args->size > PAGE_SIZE) {
6f72c7e2 1832 ret = -EINVAL;
c47ca32d 1833 goto free_args;
6f72c7e2
AJ		 1834 }
1835 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1836 if (IS_ERR(inherit)) {
1837 ret = PTR_ERR(inherit);
c47ca32d 1838 goto free_args;
6f72c7e2
AJ		 1839 }
1840 }
fa0d2b9b
LZ		 1841
1842 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1843 vol_args->fd, subvol, ptr,
8696c533 1844 readonly, inherit);
c47ca32d
DC		 1845 if (ret)
1846 goto free_inherit;
fa0d2b9b 1847
c47ca32d
DC		 1848 if (ptr && copy_to_user(arg +
1849 offsetof(struct btrfs_ioctl_vol_args_v2,
1850 transid),
1851 ptr, sizeof(*ptr)))
fa0d2b9b 1852 ret = -EFAULT;
c47ca32d
DC		 1853
1854free_inherit:
6f72c7e2 1855 kfree(inherit);
c47ca32d
DC		 1856free_args:
1857 kfree(vol_args);
f46b5a66
CH		 1858 return ret;
1859}
1860
0caa102d
LZ		 1861static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1862 void __user *arg)
1863{
496ad9aa 1864 struct inode *inode = file_inode(file);
0b246afa 1865 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
0caa102d
LZ		 1866 struct btrfs_root *root = BTRFS_I(inode)->root;
1867 int ret = 0;
1868 u64 flags = 0;
1869
4a0cc7ca 1870 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1871 return -EINVAL;
1872
0b246afa 1873 down_read(&fs_info->subvol_sem);
0caa102d
LZ		 1874 if (btrfs_root_readonly(root))
1875 flags |= BTRFS_SUBVOL_RDONLY;
0b246afa 1876 up_read(&fs_info->subvol_sem);
0caa102d
LZ		 1877
1878 if (copy_to_user(arg, &flags, sizeof(flags)))
1879 ret = -EFAULT;
1880
1881 return ret;
1882}
1883
1884static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1885 void __user *arg)
1886{
496ad9aa 1887 struct inode *inode = file_inode(file);
0b246afa 1888 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
0caa102d
LZ		 1889 struct btrfs_root *root = BTRFS_I(inode)->root;
1890 struct btrfs_trans_handle *trans;
1891 u64 root_flags;
1892 u64 flags;
1893 int ret = 0;
1894
bd60ea0f
DS		 1895 if (!inode_owner_or_capable(inode))
1896 return -EPERM;
1897
b9ca0664
LB		 1898 ret = mnt_want_write_file(file);
1899 if (ret)
1900 goto out;
0caa102d 1901
4a0cc7ca 1902 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
b9ca0664
LB		 1903 ret = -EINVAL;
1904 goto out_drop_write;
1905 }
0caa102d 1906
b9ca0664
LB		 1907 if (copy_from_user(&flags, arg, sizeof(flags))) {
1908 ret = -EFAULT;
1909 goto out_drop_write;
1910 }
0caa102d 1911
b9ca0664
LB		 1912 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1913 ret = -EINVAL;
1914 goto out_drop_write;
1915 }
0caa102d 1916
b9ca0664
LB		 1917 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1918 ret = -EOPNOTSUPP;
1919 goto out_drop_write;
1920 }
0caa102d 1921
0b246afa 1922 down_write(&fs_info->subvol_sem);
0caa102d
LZ		 1923
1924 /* nothing to do */
1925 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1926 goto out_drop_sem;
0caa102d
LZ		 1927
1928 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1929 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1930 btrfs_set_root_flags(&root->root_item,
1931 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1932 } else {
1933 /*
1934 * Block RO -> RW transition if this subvolume is involved in
1935 * send
1936 */
1937 spin_lock(&root->root_item_lock);
1938 if (root->send_in_progress == 0) {
1939 btrfs_set_root_flags(&root->root_item,
0caa102d 1940 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1941 spin_unlock(&root->root_item_lock);
1942 } else {
1943 spin_unlock(&root->root_item_lock);
0b246afa
JM		 1944 btrfs_warn(fs_info,
1945 "Attempt to set subvolume %llu read-write during send",
1946 root->root_key.objectid);
2c686537
DS		 1947 ret = -EPERM;
1948 goto out_drop_sem;
1949 }
1950 }
0caa102d
LZ		 1951
1952 trans = btrfs_start_transaction(root, 1);
1953 if (IS_ERR(trans)) {
1954 ret = PTR_ERR(trans);
1955 goto out_reset;
1956 }
1957
0b246afa 1958 ret = btrfs_update_root(trans, fs_info->tree_root,
0caa102d 1959 &root->root_key, &root->root_item);
9417ebc8
NB		 1960 if (ret < 0) {
1961 btrfs_end_transaction(trans);
1962 goto out_reset;
1963 }
1964
1965 ret = btrfs_commit_transaction(trans);
0caa102d 1966
0caa102d
LZ		 1967out_reset:
1968 if (ret)
1969 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1970out_drop_sem:
0b246afa 1971 up_write(&fs_info->subvol_sem);
b9ca0664
LB		 1972out_drop_write:
1973 mnt_drop_write_file(file);
1974out:
0caa102d
LZ		 1975 return ret;
1976}
1977
ac8e9819
CM		 1978static noinline int key_in_sk(struct btrfs_key *key,
1979 struct btrfs_ioctl_search_key *sk)
1980{
abc6e134
CM		 1981 struct btrfs_key test;
1982 int ret;
1983
1984 test.objectid = sk->min_objectid;
1985 test.type = sk->min_type;
1986 test.offset = sk->min_offset;
1987
1988 ret = btrfs_comp_cpu_keys(key, &test);
1989 if (ret < 0)
ac8e9819 1990 return 0;
abc6e134
CM		 1991
1992 test.objectid = sk->max_objectid;
1993 test.type = sk->max_type;
1994 test.offset = sk->max_offset;
1995
1996 ret = btrfs_comp_cpu_keys(key, &test);
1997 if (ret > 0)
ac8e9819
CM		 1998 return 0;
1999 return 1;
2000}
2001
df397565 2002static noinline int copy_to_sk(struct btrfs_path *path,
ac8e9819
CM		 2003 struct btrfs_key *key,
2004 struct btrfs_ioctl_search_key *sk,
9b6e817d 2005 size_t *buf_size,
ba346b35 2006 char __user *ubuf,
ac8e9819
CM		 2007 unsigned long *sk_offset,
2008 int *num_found)
2009{
2010 u64 found_transid;
2011 struct extent_buffer *leaf;
2012 struct btrfs_ioctl_search_header sh;
dd81d459 2013 struct btrfs_key test;
ac8e9819
CM		 2014 unsigned long item_off;
2015 unsigned long item_len;
2016 int nritems;
2017 int i;
2018 int slot;
ac8e9819
CM		 2019 int ret = 0;
2020
2021 leaf = path->nodes[0];
2022 slot = path->slots[0];
2023 nritems = btrfs_header_nritems(leaf);
2024
2025 if (btrfs_header_generation(leaf) > sk->max_transid) {
2026 i = nritems;
2027 goto advance_key;
2028 }
2029 found_transid = btrfs_header_generation(leaf);
2030
2031 for (i = slot; i < nritems; i++) {
2032 item_off = btrfs_item_ptr_offset(leaf, i);
2033 item_len = btrfs_item_size_nr(leaf, i);
2034
03b71c6c
GP		 2035 btrfs_item_key_to_cpu(leaf, key, i);
2036 if (!key_in_sk(key, sk))
2037 continue;
2038
9b6e817d 2039 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 2040 if (*num_found) {
2041 ret = 1;
2042 goto out;
2043 }
2044
2045 /*
2046 * return one empty item back for v1, which does not
2047 * handle -EOVERFLOW
2048 */
2049
9b6e817d 2050 *buf_size = sizeof(sh) + item_len;
ac8e9819 2051 item_len = 0;
8f5f6178
GH		 2052 ret = -EOVERFLOW;
2053 }
ac8e9819 2054
9b6e817d 2055 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 2056 ret = 1;
25c9bc2e 2057 goto out;
ac8e9819
CM		 2058 }
2059
ac8e9819
CM		 2060 sh.objectid = key->objectid;
2061 sh.offset = key->offset;
2062 sh.type = key->type;
2063 sh.len = item_len;
2064 sh.transid = found_transid;
2065
2066 /* copy search result header */
ba346b35
GH		 2067 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2068 ret = -EFAULT;
2069 goto out;
2070 }
2071
ac8e9819
CM		 2072 *sk_offset += sizeof(sh);
2073
2074 if (item_len) {
ba346b35 2075 char __user *up = ubuf + *sk_offset;
ac8e9819 2076 /* copy the item */
ba346b35
GH		 2077 if (read_extent_buffer_to_user(leaf, up,
2078 item_off, item_len)) {
2079 ret = -EFAULT;
2080 goto out;
2081 }
2082
ac8e9819 2083 *sk_offset += item_len;
ac8e9819 2084 }
e2156867 2085 (*num_found)++;
ac8e9819 2086
8f5f6178
GH		 2087 if (ret) /* -EOVERFLOW from above */
2088 goto out;
2089
25c9bc2e
GH		 2090 if (*num_found >= sk->nr_items) {
2091 ret = 1;
2092 goto out;
2093 }
ac8e9819
CM		 2094 }
2095advance_key:
abc6e134 2096 ret = 0;
dd81d459
NA		 2097 test.objectid = sk->max_objectid;
2098 test.type = sk->max_type;
2099 test.offset = sk->max_offset;
2100 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2101 ret = 1;
2102 else if (key->offset < (u64)-1)
ac8e9819 2103 key->offset++;
dd81d459 2104 else if (key->type < (u8)-1) {
abc6e134 2105 key->offset = 0;
ac8e9819 2106 key->type++;
dd81d459 2107 } else if (key->objectid < (u64)-1) {
abc6e134
CM		 2108 key->offset = 0;
2109 key->type = 0;
ac8e9819 2110 key->objectid++;
abc6e134
CM		 2111 } else
2112 ret = 1;
25c9bc2e 2113out:
ba346b35
GH		 2114 /*
2115 * 0: all items from this leaf copied, continue with next
2116 * 1: * more items can be copied, but unused buffer is too small
2117 * * all items were found
2118 * Either way, it will stops the loop which iterates to the next
2119 * leaf
2120 * -EOVERFLOW: item was to large for buffer
2121 * -EFAULT: could not copy extent buffer back to userspace
2122 */
ac8e9819
CM		 2123 return ret;
2124}
2125
2126static noinline int search_ioctl(struct inode *inode,
12544442 2127 struct btrfs_ioctl_search_key *sk,
9b6e817d 2128 size_t *buf_size,
ba346b35 2129 char __user *ubuf)
ac8e9819 2130{
0b246afa 2131 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
ac8e9819
CM		 2132 struct btrfs_root *root;
2133 struct btrfs_key key;
ac8e9819 2134 struct btrfs_path *path;
ac8e9819
CM		 2135 int ret;
2136 int num_found = 0;
2137 unsigned long sk_offset = 0;
2138
9b6e817d
GH		 2139 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2140 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2141 return -EOVERFLOW;
9b6e817d 2142 }
12544442 2143
ac8e9819
CM		 2144 path = btrfs_alloc_path();
2145 if (!path)
2146 return -ENOMEM;
2147
2148 if (sk->tree_id == 0) {
2149 /* search the root of the inode that was passed */
2150 root = BTRFS_I(inode)->root;
2151 } else {
2152 key.objectid = sk->tree_id;
2153 key.type = BTRFS_ROOT_ITEM_KEY;
2154 key.offset = (u64)-1;
2155 root = btrfs_read_fs_root_no_name(info, &key);
2156 if (IS_ERR(root)) {
ac8e9819
CM		 2157 btrfs_free_path(path);
2158 return -ENOENT;
2159 }
2160 }
2161
2162 key.objectid = sk->min_objectid;
2163 key.type = sk->min_type;
2164 key.offset = sk->min_offset;
2165
67871254 2166 while (1) {
6174d3cb 2167 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2168 if (ret != 0) {
2169 if (ret > 0)
2170 ret = 0;
2171 goto err;
2172 }
df397565 2173 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
ac8e9819 2174 &sk_offset, &num_found);
b3b4aa74 2175 btrfs_release_path(path);
25c9bc2e 2176 if (ret)
ac8e9819
CM		 2177 break;
2178
2179 }
8f5f6178
GH		 2180 if (ret > 0)
2181 ret = 0;
ac8e9819
CM		 2182err:
2183 sk->nr_items = num_found;
2184 btrfs_free_path(path);
2185 return ret;
2186}
2187
2188static noinline int btrfs_ioctl_tree_search(struct file *file,
2189 void __user *argp)
2190{
ba346b35
GH		 2191 struct btrfs_ioctl_search_args __user *uargs;
2192 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2193 struct inode *inode;
2194 int ret;
2195 size_t buf_size;
ac8e9819
CM		 2196
2197 if (!capable(CAP_SYS_ADMIN))
2198 return -EPERM;
2199
ba346b35
GH		 2200 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2201
2202 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2203 return -EFAULT;
ac8e9819 2204
ba346b35 2205 buf_size = sizeof(uargs->buf);
ac8e9819 2206
496ad9aa 2207 inode = file_inode(file);
ba346b35 2208 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2209
2210 /*
2211 * In the origin implementation an overflow is handled by returning a
2212 * search header with a len of zero, so reset ret.
2213 */
2214 if (ret == -EOVERFLOW)
2215 ret = 0;
2216
ba346b35 2217 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2218 ret = -EFAULT;
ac8e9819
CM		 2219 return ret;
2220}
2221
cc68a8a5
GH		 2222static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2223 void __user *argp)
2224{
2225 struct btrfs_ioctl_search_args_v2 __user *uarg;
2226 struct btrfs_ioctl_search_args_v2 args;
2227 struct inode *inode;
2228 int ret;
2229 size_t buf_size;
ee22184b 2230 const size_t buf_limit = SZ_16M;
cc68a8a5
GH		 2231
2232 if (!capable(CAP_SYS_ADMIN))
2233 return -EPERM;
2234
2235 /* copy search header and buffer size */
2236 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2237 if (copy_from_user(&args, uarg, sizeof(args)))
2238 return -EFAULT;
2239
2240 buf_size = args.buf_size;
2241
cc68a8a5
GH		 2242 /* limit result size to 16MB */
2243 if (buf_size > buf_limit)
2244 buf_size = buf_limit;
2245
2246 inode = file_inode(file);
2247 ret = search_ioctl(inode, &args.key, &buf_size,
718dc5fa 2248 (char __user *)(&uarg->buf[0]));
cc68a8a5
GH		 2249 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2250 ret = -EFAULT;
2251 else if (ret == -EOVERFLOW &&
2252 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2253 ret = -EFAULT;
2254
ac8e9819
CM		 2255 return ret;
2256}
2257
98d377a0 2258/*
ac8e9819
CM		 2259 * Search INODE_REFs to identify path name of 'dirid' directory
2260 * in a 'tree_id' tree. and sets path name to 'name'.
2261 */
98d377a0
TH		 2262static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2263 u64 tree_id, u64 dirid, char *name)
2264{
2265 struct btrfs_root *root;
2266 struct btrfs_key key;
ac8e9819 2267 char *ptr;
98d377a0
TH		 2268 int ret = -1;
2269 int slot;
2270 int len;
2271 int total_len = 0;
2272 struct btrfs_inode_ref *iref;
2273 struct extent_buffer *l;
2274 struct btrfs_path *path;
2275
2276 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2277 name[0]='\0';
2278 return 0;
2279 }
2280
2281 path = btrfs_alloc_path();
2282 if (!path)
2283 return -ENOMEM;
2284
c8bcbfbd 2285 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
98d377a0
TH		 2286
2287 key.objectid = tree_id;
2288 key.type = BTRFS_ROOT_ITEM_KEY;
2289 key.offset = (u64)-1;
2290 root = btrfs_read_fs_root_no_name(info, &key);
2291 if (IS_ERR(root)) {
f14d104d 2292 btrfs_err(info, "could not find root %llu", tree_id);
8ad6fcab
CM		 2293 ret = -ENOENT;
2294 goto out;
98d377a0
TH		 2295 }
2296
2297 key.objectid = dirid;
2298 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2299 key.offset = (u64)-1;
98d377a0 2300
67871254 2301 while (1) {
98d377a0
TH		 2302 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2303 if (ret < 0)
2304 goto out;
18674c6c
FDBM		 2305 else if (ret > 0) {
2306 ret = btrfs_previous_item(root, path, dirid,
2307 BTRFS_INODE_REF_KEY);
2308 if (ret < 0)
2309 goto out;
2310 else if (ret > 0) {
2311 ret = -ENOENT;
2312 goto out;
2313 }
2314 }
98d377a0
TH		 2315
2316 l = path->nodes[0];
2317 slot = path->slots[0];
2318 btrfs_item_key_to_cpu(l, &key, slot);
2319
98d377a0
TH		 2320 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2321 len = btrfs_inode_ref_name_len(l, iref);
2322 ptr -= len + 1;
2323 total_len += len + 1;
a696cf35
FDBM		 2324 if (ptr < name) {
2325 ret = -ENAMETOOLONG;
98d377a0 2326 goto out;
a696cf35 2327 }
98d377a0
TH		 2328
2329 *(ptr + len) = '/';
67871254 2330 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2331
2332 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2333 break;
2334
b3b4aa74 2335 btrfs_release_path(path);
98d377a0 2336 key.objectid = key.offset;
8ad6fcab 2337 key.offset = (u64)-1;
98d377a0 2338 dirid = key.objectid;
98d377a0 2339 }
77906a50 2340 memmove(name, ptr, total_len);
67871254 2341 name[total_len] = '\0';
98d377a0
TH		 2342 ret = 0;
2343out:
2344 btrfs_free_path(path);
ac8e9819
CM		 2345 return ret;
2346}
2347
2348static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2349 void __user *argp)
2350{
2351 struct btrfs_ioctl_ino_lookup_args *args;
2352 struct inode *inode;
01b810b8 2353 int ret = 0;
ac8e9819 2354
2354d08f
JL		 2355 args = memdup_user(argp, sizeof(*args));
2356 if (IS_ERR(args))
2357 return PTR_ERR(args);
c2b96929 2358
496ad9aa 2359 inode = file_inode(file);
ac8e9819 2360
01b810b8
DS		 2361 /*
2362 * Unprivileged query to obtain the containing subvolume root id. The
2363 * path is reset so it's consistent with btrfs_search_path_in_tree.
2364 */
1b53ac4d
CM		 2365 if (args->treeid == 0)
2366 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2367
01b810b8
DS		 2368 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2369 args->name[0] = 0;
2370 goto out;
2371 }
2372
2373 if (!capable(CAP_SYS_ADMIN)) {
2374 ret = -EPERM;
2375 goto out;
2376 }
2377
ac8e9819
CM		 2378 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2379 args->treeid, args->objectid,
2380 args->name);
2381
01b810b8 2382out:
ac8e9819
CM		 2383 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2384 ret = -EFAULT;
2385
2386 kfree(args);
98d377a0
TH		 2387 return ret;
2388}
2389
76dda93c
YZ		 2390static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2391 void __user *arg)
2392{
54563d41 2393 struct dentry *parent = file->f_path.dentry;
0b246afa 2394 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
76dda93c 2395 struct dentry *dentry;
2b0143b5 2396 struct inode *dir = d_inode(parent);
76dda93c
YZ		 2397 struct inode *inode;
2398 struct btrfs_root *root = BTRFS_I(dir)->root;
2399 struct btrfs_root *dest = NULL;
2400 struct btrfs_ioctl_vol_args *vol_args;
76dda93c 2401 int namelen;
76dda93c
YZ		 2402 int err = 0;
2403
325c50e3
JM		 2404 if (!S_ISDIR(dir->i_mode))
2405 return -ENOTDIR;
2406
76dda93c
YZ		 2407 vol_args = memdup_user(arg, sizeof(*vol_args));
2408 if (IS_ERR(vol_args))
2409 return PTR_ERR(vol_args);
2410
2411 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2412 namelen = strlen(vol_args->name);
2413 if (strchr(vol_args->name, '/') ||
2414 strncmp(vol_args->name, "..", namelen) == 0) {
2415 err = -EINVAL;
2416 goto out;
2417 }
2418
a561be71 2419 err = mnt_want_write_file(file);
76dda93c
YZ		 2420 if (err)
2421 goto out;
2422
521e0546 2423
00235411
AV		 2424 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2425 if (err == -EINTR)
2426 goto out_drop_write;
76dda93c
YZ		 2427 dentry = lookup_one_len(vol_args->name, parent, namelen);
2428 if (IS_ERR(dentry)) {
2429 err = PTR_ERR(dentry);
2430 goto out_unlock_dir;
2431 }
2432
2b0143b5 2433 if (d_really_is_negative(dentry)) {
76dda93c
YZ		 2434 err = -ENOENT;
2435 goto out_dput;
2436 }
2437
2b0143b5 2438 inode = d_inode(dentry);
4260f7c7 2439 dest = BTRFS_I(inode)->root;
67871254 2440 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2441 /*
2442 * Regular user. Only allow this with a special mount
2443 * option, when the user has write+exec access to the
2444 * subvol root, and when rmdir(2) would have been
2445 * allowed.
2446 *
2447 * Note that this is _not_ check that the subvol is
2448 * empty or doesn't contain data that we wouldn't
2449 * otherwise be able to delete.
2450 *
2451 * Users who want to delete empty subvols should try
2452 * rmdir(2).
2453 */
2454 err = -EPERM;
0b246afa 2455 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
4260f7c7
SW		 2456 goto out_dput;
2457
2458 /*
2459 * Do not allow deletion if the parent dir is the same
2460 * as the dir to be deleted. That means the ioctl
2461 * must be called on the dentry referencing the root
2462 * of the subvol, not a random directory contained
2463 * within it.
2464 */
2465 err = -EINVAL;
2466 if (root == dest)
2467 goto out_dput;
2468
2469 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2470 if (err)
2471 goto out_dput;
4260f7c7
SW		 2472 }
2473
5c39da5b
MX		 2474 /* check if subvolume may be deleted by a user */
2475 err = btrfs_may_delete(dir, dentry, 1);
2476 if (err)
2477 goto out_dput;
2478
4a0cc7ca 2479 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2480 err = -EINVAL;
2481 goto out_dput;
2482 }
2483
5955102c 2484 inode_lock(inode);
f60a2364 2485 err = btrfs_delete_subvolume(dir, dentry);
5955102c 2486 inode_unlock(inode);
f60a2364 2487 if (!err)
76dda93c 2488 d_delete(dentry);
fa6ac876 2489
76dda93c
YZ		 2490out_dput:
2491 dput(dentry);
2492out_unlock_dir:
5955102c 2493 inode_unlock(dir);
00235411 2494out_drop_write:
2a79f17e 2495 mnt_drop_write_file(file);
76dda93c
YZ		 2496out:
2497 kfree(vol_args);
2498 return err;
2499}
2500
1e701a32 2501static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2502{
496ad9aa 2503 struct inode *inode = file_inode(file);
f46b5a66 2504 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2505 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2506 int ret;
2507
25122d15
ID		 2508 ret = mnt_want_write_file(file);
2509 if (ret)
2510 return ret;
b83cc969 2511
25122d15
ID		 2512 if (btrfs_root_readonly(root)) {
2513 ret = -EROFS;
2514 goto out;
5ac00add 2515 }
f46b5a66
CH		 2516
2517 switch (inode->i_mode & S_IFMT) {
2518 case S_IFDIR:
e441d54d
CM		 2519 if (!capable(CAP_SYS_ADMIN)) {
2520 ret = -EPERM;
2521 goto out;
2522 }
de78b51a 2523 ret = btrfs_defrag_root(root);
f46b5a66
CH		 2524 break;
2525 case S_IFREG:
e441d54d
CM		 2526 if (!(file->f_mode & FMODE_WRITE)) {
2527 ret = -EINVAL;
2528 goto out;
2529 }
1e701a32
CM		 2530
2531 range = kzalloc(sizeof(*range), GFP_KERNEL);
2532 if (!range) {
2533 ret = -ENOMEM;
2534 goto out;
2535 }
2536
2537 if (argp) {
2538 if (copy_from_user(range, argp,
2539 sizeof(*range))) {
2540 ret = -EFAULT;
2541 kfree(range);
683be16e 2542 goto out;
1e701a32
CM		 2543 }
2544 /* compression requires us to start the IO */
2545 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2546 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2547 range->extent_thresh = (u32)-1;
2548 }
2549 } else {
2550 /* the rest are all set to zero by kzalloc */
2551 range->len = (u64)-1;
2552 }
496ad9aa 2553 ret = btrfs_defrag_file(file_inode(file), file,
7c829b72 2554 range, BTRFS_OLDEST_GENERATION, 0);
4cb5300b
CM		 2555 if (ret > 0)
2556 ret = 0;
1e701a32 2557 kfree(range);
f46b5a66 2558 break;
8929ecfa
YZ		 2559 default:
2560 ret = -EINVAL;
f46b5a66 2561 }
e441d54d 2562out:
25122d15 2563 mnt_drop_write_file(file);
e441d54d 2564 return ret;
f46b5a66
CH		 2565}
2566
2ff7e61e 2567static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
f46b5a66
CH		 2568{
2569 struct btrfs_ioctl_vol_args *vol_args;
2570 int ret;
2571
e441d54d
CM		 2572 if (!capable(CAP_SYS_ADMIN))
2573 return -EPERM;
2574
171938e5 2575 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
e57138b3 2576 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b 2577
dae7b665 2578 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2579 if (IS_ERR(vol_args)) {
2580 ret = PTR_ERR(vol_args);
2581 goto out;
2582 }
f46b5a66 2583
5516e595 2584 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2ff7e61e 2585 ret = btrfs_init_new_device(fs_info, vol_args->name);
f46b5a66 2586
43d20761 2587 if (!ret)
0b246afa 2588 btrfs_info(fs_info, "disk added %s", vol_args->name);
43d20761 2589
f46b5a66 2590 kfree(vol_args);
c9e9f97b 2591out:
171938e5 2592 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
f46b5a66
CH		 2593 return ret;
2594}
2595
6b526ed7 2596static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
f46b5a66 2597{
0b246afa
JM		 2598 struct inode *inode = file_inode(file);
2599 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6b526ed7 2600 struct btrfs_ioctl_vol_args_v2 *vol_args;
f46b5a66
CH		 2601 int ret;
2602
e441d54d
CM		 2603 if (!capable(CAP_SYS_ADMIN))
2604 return -EPERM;
2605
da24927b
MX		 2606 ret = mnt_want_write_file(file);
2607 if (ret)
2608 return ret;
c146afad 2609
dae7b665 2610 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2611 if (IS_ERR(vol_args)) {
2612 ret = PTR_ERR(vol_args);
c47ca32d 2613 goto err_drop;
c9e9f97b 2614 }
f46b5a66 2615
6b526ed7 2616 /* Check for compatibility reject unknown flags */
735654ea
DS		 2617 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2618 return -EOPNOTSUPP;
f46b5a66 2619
171938e5 2620 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
183860f6
AJ		 2621 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2622 goto out;
2623 }
2624
735654ea 2625 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2ff7e61e 2626 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
6b526ed7
AJ		 2627 } else {
2628 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2ff7e61e 2629 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
6b526ed7 2630 }
171938e5 2631 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
183860f6 2632
6b526ed7 2633 if (!ret) {
735654ea 2634 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
0b246afa 2635 btrfs_info(fs_info, "device deleted: id %llu",
6b526ed7
AJ		 2636 vol_args->devid);
2637 else
0b246afa 2638 btrfs_info(fs_info, "device deleted: %s",
6b526ed7
AJ		 2639 vol_args->name);
2640 }
183860f6
AJ		 2641out:
2642 kfree(vol_args);
c47ca32d 2643err_drop:
4ac20c70 2644 mnt_drop_write_file(file);
f46b5a66
CH		 2645 return ret;
2646}
2647
da24927b 2648static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2649{
0b246afa
JM		 2650 struct inode *inode = file_inode(file);
2651 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66
CH		 2652 struct btrfs_ioctl_vol_args *vol_args;
2653 int ret;
2654
e441d54d
CM		 2655 if (!capable(CAP_SYS_ADMIN))
2656 return -EPERM;
2657
da24927b
MX		 2658 ret = mnt_want_write_file(file);
2659 if (ret)
2660 return ret;
c146afad 2661
171938e5 2662 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
183860f6 2663 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
58d7bbf8
DS		 2664 goto out_drop_write;
2665 }
2666
2667 vol_args = memdup_user(arg, sizeof(*vol_args));
2668 if (IS_ERR(vol_args)) {
2669 ret = PTR_ERR(vol_args);
183860f6
AJ		 2670 goto out;
2671 }
2672
58d7bbf8 2673 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2ff7e61e 2674 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
183860f6 2675
ec95d491 2676 if (!ret)
0b246afa 2677 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
183860f6 2678 kfree(vol_args);
58d7bbf8 2679out:
171938e5 2680 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
58d7bbf8 2681out_drop_write:
4ac20c70 2682 mnt_drop_write_file(file);
58d7bbf8 2683
f46b5a66
CH		 2684 return ret;
2685}
2686
2ff7e61e
JM		 2687static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2688 void __user *arg)
475f6387 2689{
027ed2f0 2690 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387 2691 struct btrfs_device *device;
0b246afa 2692 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
027ed2f0 2693 int ret = 0;
475f6387 2694
027ed2f0
LZ		 2695 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2696 if (!fi_args)
2697 return -ENOMEM;
2698
d03262c7 2699 rcu_read_lock();
027ed2f0 2700 fi_args->num_devices = fs_devices->num_devices;
475f6387 2701
d03262c7 2702 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2703 if (device->devid > fi_args->max_id)
2704 fi_args->max_id = device->devid;
475f6387 2705 }
d03262c7 2706 rcu_read_unlock();
475f6387 2707
d03262c7 2708 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
bea7eafd
OS		 2709 fi_args->nodesize = fs_info->nodesize;
2710 fi_args->sectorsize = fs_info->sectorsize;
2711 fi_args->clone_alignment = fs_info->sectorsize;
80a773fb 2712
027ed2f0
LZ		 2713 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2714 ret = -EFAULT;
475f6387 2715
027ed2f0
LZ		 2716 kfree(fi_args);
2717 return ret;
475f6387
JS		 2718}
2719
2ff7e61e
JM		 2720static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2721 void __user *arg)
475f6387
JS		 2722{
2723 struct btrfs_ioctl_dev_info_args *di_args;
2724 struct btrfs_device *dev;
475f6387
JS		 2725 int ret = 0;
2726 char *s_uuid = NULL;
475f6387 2727
475f6387
JS		 2728 di_args = memdup_user(arg, sizeof(*di_args));
2729 if (IS_ERR(di_args))
2730 return PTR_ERR(di_args);
2731
dd5f9615 2732 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2733 s_uuid = di_args->uuid;
2734
c5593ca3 2735 rcu_read_lock();
0b246afa 2736 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2737
2738 if (!dev) {
2739 ret = -ENODEV;
2740 goto out;
2741 }
2742
2743 di_args->devid = dev->devid;
7cc8e58d
MX		 2744 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2745 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
475f6387 2746 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2747 if (dev->name) {
606686ee
JB		 2748 struct rcu_string *name;
2749
606686ee 2750 name = rcu_dereference(dev->name);
6670d4c2 2751 strncpy(di_args->path, name->str, sizeof(di_args->path) - 1);
a27202fb
JM		 2752 di_args->path[sizeof(di_args->path) - 1] = 0;
2753 } else {
99ba55ad 2754 di_args->path[0] = '\0';
a27202fb 2755 }
475f6387
JS		 2756
2757out:
c5593ca3 2758 rcu_read_unlock();
475f6387
JS		 2759 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2760 ret = -EFAULT;
2761
2762 kfree(di_args);
2763 return ret;
2764}
2765
f4414602 2766static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
416161db
MF		 2767{
2768 struct page *page;
416161db 2769
416161db
MF		 2770 page = grab_cache_page(inode->i_mapping, index);
2771 if (!page)
31314002 2772 return ERR_PTR(-ENOMEM);
416161db
MF		 2773
2774 if (!PageUptodate(page)) {
31314002
FM		 2775 int ret;
2776
2777 ret = btrfs_readpage(NULL, page);
2778 if (ret)
2779 return ERR_PTR(ret);
416161db
MF		 2780 lock_page(page);
2781 if (!PageUptodate(page)) {
2782 unlock_page(page);
09cbfeaf 2783 put_page(page);
31314002
FM		 2784 return ERR_PTR(-EIO);
2785 }
2786 if (page->mapping != inode->i_mapping) {
2787 unlock_page(page);
09cbfeaf 2788 put_page(page);
31314002 2789 return ERR_PTR(-EAGAIN);
416161db
MF		 2790 }
2791 }
416161db
MF		 2792
2793 return page;
2794}
2795
f4414602
MF		 2796static int gather_extent_pages(struct inode *inode, struct page **pages,
2797 int num_pages, u64 off)
2798{
2799 int i;
09cbfeaf 2800 pgoff_t index = off >> PAGE_SHIFT;
f4414602
MF		 2801
2802 for (i = 0; i < num_pages; i++) {
31314002 2803again:
f4414602 2804 pages[i] = extent_same_get_page(inode, index + i);
31314002
FM		 2805 if (IS_ERR(pages[i])) {
2806 int err = PTR_ERR(pages[i]);
2807
2808 if (err == -EAGAIN)
2809 goto again;
2810 pages[i] = NULL;
2811 return err;
2812 }
f4414602
MF		 2813 }
2814 return 0;
2815}
2816
e0bd70c6
FM		 2817static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2818 bool retry_range_locking)
77fe20dc 2819{
e0bd70c6
FM		 2820 /*
2821 * Do any pending delalloc/csum calculations on inode, one way or
2822 * another, and lock file content.
2823 * The locking order is:
2824 *
2825 * 1) pages
2826 * 2) range in the inode's io tree
2827 */
77fe20dc
MF		 2828 while (1) {
2829 struct btrfs_ordered_extent *ordered;
2830 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2831 ordered = btrfs_lookup_first_ordered_extent(inode,
2832 off + len - 1);
ff5df9b8
FM		 2833 if ((!ordered ||
2834 ordered->file_offset + ordered->len <= off ||
2835 ordered->file_offset >= off + len) &&
77fe20dc 2836 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
ff5df9b8
FM		 2837 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2838 if (ordered)
2839 btrfs_put_ordered_extent(ordered);
77fe20dc 2840 break;
ff5df9b8 2841 }
77fe20dc
MF		 2842 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2843 if (ordered)
2844 btrfs_put_ordered_extent(ordered);
e0bd70c6
FM		 2845 if (!retry_range_locking)
2846 return -EAGAIN;
77fe20dc
MF		 2847 btrfs_wait_ordered_range(inode, off, len);
2848 }
e0bd70c6 2849 return 0;
77fe20dc
MF		 2850}
2851
f4414602 2852static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
416161db 2853{
5955102c
AV		 2854 inode_unlock(inode1);
2855 inode_unlock(inode2);
416161db
MF		 2856}
2857
f4414602
MF		 2858static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2859{
2860 if (inode1 < inode2)
2861 swap(inode1, inode2);
2862
5955102c
AV		 2863 inode_lock_nested(inode1, I_MUTEX_PARENT);
2864 inode_lock_nested(inode2, I_MUTEX_CHILD);
f4414602
MF		 2865}
2866
2867static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2868 struct inode *inode2, u64 loff2, u64 len)
2869{
2870 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2871 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2872}
2873
e0bd70c6
FM		 2874static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2875 struct inode *inode2, u64 loff2, u64 len,
2876 bool retry_range_locking)
416161db 2877{
e0bd70c6
FM		 2878 int ret;
2879
416161db
MF		 2880 if (inode1 < inode2) {
2881 swap(inode1, inode2);
2882 swap(loff1, loff2);
2883 }
e0bd70c6
FM		 2884 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2885 if (ret)
2886 return ret;
2887 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2888 if (ret)
2889 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2890 loff1 + len - 1);
2891 return ret;
f4414602
MF		 2892}
2893
2894struct cmp_pages {
2895 int num_pages;
2896 struct page **src_pages;
2897 struct page **dst_pages;
2898};
2899
2900static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2901{
2902 int i;
2903 struct page *pg;
2904
2905 for (i = 0; i < cmp->num_pages; i++) {
2906 pg = cmp->src_pages[i];
e0bd70c6
FM		 2907 if (pg) {
2908 unlock_page(pg);
09cbfeaf 2909 put_page(pg);
e0bd70c6 2910 }
f4414602 2911 pg = cmp->dst_pages[i];
e0bd70c6
FM		 2912 if (pg) {
2913 unlock_page(pg);
09cbfeaf 2914 put_page(pg);
e0bd70c6 2915 }
f4414602
MF		 2916 }
2917 kfree(cmp->src_pages);
2918 kfree(cmp->dst_pages);
2919}
2920
2921static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2922 struct inode *dst, u64 dst_loff,
2923 u64 len, struct cmp_pages *cmp)
2924{
2925 int ret;
09cbfeaf 2926 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
f4414602
MF		 2927 struct page **src_pgarr, **dst_pgarr;
2928
2929 /*
2930 * We must gather up all the pages before we initiate our
2931 * extent locking. We use an array for the page pointers. Size
2932 * of the array is bounded by len, which is in turn bounded by
2933 * BTRFS_MAX_DEDUPE_LEN.
2934 */
66722f7c
DS		 2935 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2936 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
f4414602
MF		 2937 if (!src_pgarr || !dst_pgarr) {
2938 kfree(src_pgarr);
2939 kfree(dst_pgarr);
2940 return -ENOMEM;
416161db 2941 }
f4414602
MF		 2942 cmp->num_pages = num_pages;
2943 cmp->src_pages = src_pgarr;
2944 cmp->dst_pages = dst_pgarr;
2945
b1517622
FM		 2946 /*
2947 * If deduping ranges in the same inode, locking rules make it mandatory
2948 * to always lock pages in ascending order to avoid deadlocks with
2949 * concurrent tasks (such as starting writeback/delalloc).
2950 */
2951 if (src == dst && dst_loff < loff) {
2952 swap(src_pgarr, dst_pgarr);
2953 swap(loff, dst_loff);
2954 }
2955
2956 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
f4414602
MF		 2957 if (ret)
2958 goto out;
2959
b1517622 2960 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
f4414602
MF		 2961
2962out:
2963 if (ret)
2964 btrfs_cmp_data_free(cmp);
78ad4ce0 2965 return ret;
416161db
MF		 2966}
2967
1a287cfe 2968static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
416161db
MF		 2969{
2970 int ret = 0;
f4414602 2971 int i;
416161db 2972 struct page *src_page, *dst_page;
09cbfeaf 2973 unsigned int cmp_len = PAGE_SIZE;
416161db
MF		 2974 void *addr, *dst_addr;
2975
f4414602 2976 i = 0;
416161db 2977 while (len) {
09cbfeaf 2978 if (len < PAGE_SIZE)
416161db
MF		 2979 cmp_len = len;
2980
f4414602
MF		 2981 BUG_ON(i >= cmp->num_pages);
2982
2983 src_page = cmp->src_pages[i];
2984 dst_page = cmp->dst_pages[i];
e0bd70c6
FM		 2985 ASSERT(PageLocked(src_page));
2986 ASSERT(PageLocked(dst_page));
f4414602 2987
416161db
MF		 2988 addr = kmap_atomic(src_page);
2989 dst_addr = kmap_atomic(dst_page);
2990
2991 flush_dcache_page(src_page);
2992 flush_dcache_page(dst_page);
2993
2994 if (memcmp(addr, dst_addr, cmp_len))
2b3909f8 2995 ret = -EBADE;
416161db
MF		 2996
2997 kunmap_atomic(addr);
2998 kunmap_atomic(dst_addr);
416161db
MF		 2999
3000 if (ret)
3001 break;
3002
416161db 3003 len -= cmp_len;
f4414602 3004 i++;
416161db
MF		 3005 }
3006
3007 return ret;
3008}
3009
e1d227a4
MF		 3010static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3011 u64 olen)
416161db 3012{
e1d227a4 3013 u64 len = *plen;
416161db
MF		 3014 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3015
e1d227a4 3016 if (off + olen > inode->i_size || off + olen < off)
416161db 3017 return -EINVAL;
e1d227a4
MF		 3018
3019 /* if we extend to eof, continue to block boundary */
3020 if (off + len == inode->i_size)
3021 *plen = len = ALIGN(inode->i_size, bs) - off;
3022
416161db
MF		 3023 /* Check that we are block aligned - btrfs_clone() requires this */
3024 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3025 return -EINVAL;
3026
3027 return 0;
3028}
3029
e1d227a4 3030static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
416161db
MF		 3031 struct inode *dst, u64 dst_loff)
3032{
3033 int ret;
e1d227a4 3034 u64 len = olen;
f4414602 3035 struct cmp_pages cmp;
fc4badd9 3036 bool same_inode = (src == dst);
0efa9f48
MF		 3037 u64 same_lock_start = 0;
3038 u64 same_lock_len = 0;
416161db 3039
113e8283
FM		 3040 if (len == 0)
3041 return 0;
3042
fc4badd9 3043 if (same_inode)
5955102c 3044 inode_lock(src);
fc4badd9
OS		 3045 else
3046 btrfs_double_inode_lock(src, dst);
416161db 3047
fc4badd9
OS		 3048 ret = extent_same_check_offsets(src, loff, &len, olen);
3049 if (ret)
3050 goto out_unlock;
416161db 3051
fc4badd9
OS		 3052 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3053 if (ret)
3054 goto out_unlock;
3055
3056 if (same_inode) {
0efa9f48
MF		 3057 /*
3058 * Single inode case wants the same checks, except we
3059 * don't want our length pushed out past i_size as
3060 * comparing that data range makes no sense.
3061 *
3062 * extent_same_check_offsets() will do this for an
3063 * unaligned length at i_size, so catch it here and
3064 * reject the request.
3065 *
3066 * This effectively means we require aligned extents
3067 * for the single-inode case, whereas the other cases
3068 * allow an unaligned length so long as it ends at
3069 * i_size.
3070 */
3071 if (len != olen) {
3072 ret = -EINVAL;
3073 goto out_unlock;
3074 }
3075
3076 /* Check for overlapping ranges */
3077 if (dst_loff + len > loff && dst_loff < loff + len) {
3078 ret = -EINVAL;
3079 goto out_unlock;
3080 }
3081
3082 same_lock_start = min_t(u64, loff, dst_loff);
3083 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
0efa9f48 3084 }
416161db
MF		 3085
3086 /* don't make the dst file partly checksummed */
3087 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3088 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3089 ret = -EINVAL;
3090 goto out_unlock;
3091 }
3092
e0bd70c6 3093again:
f4414602
MF		 3094 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3095 if (ret)
3096 goto out_unlock;
3097
0efa9f48 3098 if (same_inode)
e0bd70c6
FM		 3099 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3100 false);
0efa9f48 3101 else
e0bd70c6
FM		 3102 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3103 false);
3104 /*
3105 * If one of the inodes has dirty pages in the respective range or
3106 * ordered extents, we need to flush dellaloc and wait for all ordered
3107 * extents in the range. We must unlock the pages and the ranges in the
3108 * io trees to avoid deadlocks when flushing delalloc (requires locking
3109 * pages) and when waiting for ordered extents to complete (they require
3110 * range locking).
3111 */
3112 if (ret == -EAGAIN) {
3113 /*
3114 * Ranges in the io trees already unlocked. Now unlock all
3115 * pages before waiting for all IO to complete.
3116 */
3117 btrfs_cmp_data_free(&cmp);
3118 if (same_inode) {
3119 btrfs_wait_ordered_range(src, same_lock_start,
3120 same_lock_len);
3121 } else {
3122 btrfs_wait_ordered_range(src, loff, len);
3123 btrfs_wait_ordered_range(dst, dst_loff, len);
3124 }
3125 goto again;
3126 }
3127 ASSERT(ret == 0);
3128 if (WARN_ON(ret)) {
3129 /* ranges in the io trees already unlocked */
3130 btrfs_cmp_data_free(&cmp);
3131 return ret;
3132 }
f4414602 3133
207910dd 3134 /* pass original length for comparison so we stay within i_size */
1a287cfe 3135 ret = btrfs_cmp_data(olen, &cmp);
416161db 3136 if (ret == 0)
1c919a5e 3137 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
416161db 3138
0efa9f48
MF		 3139 if (same_inode)
3140 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3141 same_lock_start + same_lock_len - 1);
3142 else
3143 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
f4414602
MF		 3144
3145 btrfs_cmp_data_free(&cmp);
416161db 3146out_unlock:
0efa9f48 3147 if (same_inode)
5955102c 3148 inode_unlock(src);
0efa9f48
MF		 3149 else
3150 btrfs_double_inode_unlock(src, dst);
416161db
MF		 3151
3152 return ret;
3153}
3154
ee22184b 3155#define BTRFS_MAX_DEDUPE_LEN SZ_16M
416161db 3156
2b3909f8
DW		 3157ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3158 struct file *dst_file, u64 dst_loff)
416161db 3159{
2b3909f8
DW		 3160 struct inode *src = file_inode(src_file);
3161 struct inode *dst = file_inode(dst_file);
416161db 3162 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2b3909f8 3163 ssize_t res;
416161db 3164
2b3909f8
DW		 3165 if (olen > BTRFS_MAX_DEDUPE_LEN)
3166 olen = BTRFS_MAX_DEDUPE_LEN;
416161db 3167
09cbfeaf 3168 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
416161db
MF		 3169 /*
3170 * Btrfs does not support blocksize < page_size. As a
3171 * result, btrfs_cmp_data() won't correctly handle
3172 * this situation without an update.
3173 */
2b3909f8 3174 return -EINVAL;
416161db
MF		 3175 }
3176
2b3909f8
DW		 3177 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3178 if (res)
3179 return res;
3180 return olen;
416161db
MF		 3181}
3182
f82a9901
FM		 3183static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3184 struct inode *inode,
3185 u64 endoff,
3186 const u64 destoff,
1c919a5e
MF		 3187 const u64 olen,
3188 int no_time_update)
f82a9901
FM		 3189{
3190 struct btrfs_root *root = BTRFS_I(inode)->root;
3191 int ret;
3192
3193 inode_inc_iversion(inode);
1c919a5e 3194 if (!no_time_update)
c2050a45 3195 inode->i_mtime = inode->i_ctime = current_time(inode);
f82a9901
FM		 3196 /*
3197 * We round up to the block size at eof when determining which
3198 * extents to clone above, but shouldn't round up the file size.
3199 */
3200 if (endoff > destoff + olen)
3201 endoff = destoff + olen;
3202 if (endoff > inode->i_size)
6ef06d27 3203 btrfs_i_size_write(BTRFS_I(inode), endoff);
f82a9901
FM		 3204
3205 ret = btrfs_update_inode(trans, root, inode);
3206 if (ret) {
66642832 3207 btrfs_abort_transaction(trans, ret);
3a45bb20 3208 btrfs_end_transaction(trans);
f82a9901
FM		 3209 goto out;
3210 }
3a45bb20 3211 ret = btrfs_end_transaction(trans);
f82a9901
FM		 3212out:
3213 return ret;
3214}
3215
a2f392e4 3216static void clone_update_extent_map(struct btrfs_inode *inode,
7ffbb598
FM		 3217 const struct btrfs_trans_handle *trans,
3218 const struct btrfs_path *path,
7ffbb598
FM		 3219 const u64 hole_offset,
3220 const u64 hole_len)
3221{
a2f392e4 3222 struct extent_map_tree *em_tree = &inode->extent_tree;
7ffbb598
FM		 3223 struct extent_map *em;
3224 int ret;
3225
3226 em = alloc_extent_map();
3227 if (!em) {
a2f392e4 3228 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
7ffbb598
FM		 3229 return;
3230 }
3231
14f59796
FM		 3232 if (path) {
3233 struct btrfs_file_extent_item *fi;
3234
3235 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3236 struct btrfs_file_extent_item);
a2f392e4 3237 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
7ffbb598
FM		 3238 em->generation = -1;
3239 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3240 BTRFS_FILE_EXTENT_INLINE)
3241 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
a2f392e4 3242 &inode->runtime_flags);
7ffbb598
FM		 3243 } else {
3244 em->start = hole_offset;
3245 em->len = hole_len;
3246 em->ram_bytes = em->len;
3247 em->orig_start = hole_offset;
3248 em->block_start = EXTENT_MAP_HOLE;
3249 em->block_len = 0;
3250 em->orig_block_len = 0;
3251 em->compress_type = BTRFS_COMPRESS_NONE;
3252 em->generation = trans->transid;
3253 }
3254
3255 while (1) {
3256 write_lock(&em_tree->lock);
3257 ret = add_extent_mapping(em_tree, em, 1);
3258 write_unlock(&em_tree->lock);
3259 if (ret != -EEXIST) {
3260 free_extent_map(em);
3261 break;
3262 }
a2f392e4 3263 btrfs_drop_extent_cache(inode, em->start,
7ffbb598
FM		 3264 em->start + em->len - 1, 0);
3265 }
3266
ee39b432 3267 if (ret)
a2f392e4 3268 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
7ffbb598
FM		 3269}
3270
8039d87d
FM		 3271/*
3272 * Make sure we do not end up inserting an inline extent into a file that has
3273 * already other (non-inline) extents. If a file has an inline extent it can
3274 * not have any other extents and the (single) inline extent must start at the
3275 * file offset 0. Failing to respect these rules will lead to file corruption,
3276 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3277 *
3278 * We can have extents that have been already written to disk or we can have
3279 * dirty ranges still in delalloc, in which case the extent maps and items are
3280 * created only when we run delalloc, and the delalloc ranges might fall outside
3281 * the range we are currently locking in the inode's io tree. So we check the
3282 * inode's i_size because of that (i_size updates are done while holding the
3283 * i_mutex, which we are holding here).
3284 * We also check to see if the inode has a size not greater than "datal" but has
3285 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3286 * protected against such concurrent fallocate calls by the i_mutex).
3287 *
3288 * If the file has no extents but a size greater than datal, do not allow the
3289 * copy because we would need turn the inline extent into a non-inline one (even
3290 * with NO_HOLES enabled). If we find our destination inode only has one inline
3291 * extent, just overwrite it with the source inline extent if its size is less
3292 * than the source extent's size, or we could copy the source inline extent's
3293 * data into the destination inode's inline extent if the later is greater then
3294 * the former.
3295 */
4a0ab9d7 3296static int clone_copy_inline_extent(struct inode *dst,
8039d87d
FM		 3297 struct btrfs_trans_handle *trans,
3298 struct btrfs_path *path,
3299 struct btrfs_key *new_key,
3300 const u64 drop_start,
3301 const u64 datal,
3302 const u64 skip,
3303 const u64 size,
3304 char *inline_data)
3305{
0b246afa 3306 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
8039d87d
FM		 3307 struct btrfs_root *root = BTRFS_I(dst)->root;
3308 const u64 aligned_end = ALIGN(new_key->offset + datal,
0b246afa 3309 fs_info->sectorsize);
8039d87d
FM		 3310 int ret;
3311 struct btrfs_key key;
3312
3313 if (new_key->offset > 0)
3314 return -EOPNOTSUPP;
3315
4a0cc7ca 3316 key.objectid = btrfs_ino(BTRFS_I(dst));
8039d87d
FM		 3317 key.type = BTRFS_EXTENT_DATA_KEY;
3318 key.offset = 0;
3319 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3320 if (ret < 0) {
3321 return ret;
3322 } else if (ret > 0) {
3323 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3324 ret = btrfs_next_leaf(root, path);
3325 if (ret < 0)
3326 return ret;
3327 else if (ret > 0)
3328 goto copy_inline_extent;
3329 }
3330 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
4a0cc7ca 3331 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
8039d87d
FM		 3332 key.type == BTRFS_EXTENT_DATA_KEY) {
3333 ASSERT(key.offset > 0);
3334 return -EOPNOTSUPP;
3335 }
3336 } else if (i_size_read(dst) <= datal) {
3337 struct btrfs_file_extent_item *ei;
3338 u64 ext_len;
3339
3340 /*
3341 * If the file size is <= datal, make sure there are no other
3342 * extents following (can happen do to an fallocate call with
3343 * the flag FALLOC_FL_KEEP_SIZE).
3344 */
3345 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3346 struct btrfs_file_extent_item);
3347 /*
3348 * If it's an inline extent, it can not have other extents
3349 * following it.
3350 */
3351 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3352 BTRFS_FILE_EXTENT_INLINE)
3353 goto copy_inline_extent;
3354
3355 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3356 if (ext_len > aligned_end)
3357 return -EOPNOTSUPP;
3358
3359 ret = btrfs_next_item(root, path);
3360 if (ret < 0) {
3361 return ret;
3362 } else if (ret == 0) {
3363 btrfs_item_key_to_cpu(path->nodes[0], &key,
3364 path->slots[0]);
4a0cc7ca 3365 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
8039d87d
FM		 3366 key.type == BTRFS_EXTENT_DATA_KEY)
3367 return -EOPNOTSUPP;
3368 }
3369 }
3370
3371copy_inline_extent:
3372 /*
3373 * We have no extent items, or we have an extent at offset 0 which may
3374 * or may not be inlined. All these cases are dealt the same way.
3375 */
3376 if (i_size_read(dst) > datal) {
3377 /*
3378 * If the destination inode has an inline extent...
3379 * This would require copying the data from the source inline
3380 * extent into the beginning of the destination's inline extent.
3381 * But this is really complex, both extents can be compressed
3382 * or just one of them, which would require decompressing and
3383 * re-compressing data (which could increase the new compressed
3384 * size, not allowing the compressed data to fit anymore in an
3385 * inline extent).
3386 * So just don't support this case for now (it should be rare,
3387 * we are not really saving space when cloning inline extents).
3388 */
3389 return -EOPNOTSUPP;
3390 }
3391
3392 btrfs_release_path(path);
3393 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3394 if (ret)
3395 return ret;
3396 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3397 if (ret)
3398 return ret;
3399
3400 if (skip) {
3401 const u32 start = btrfs_file_extent_calc_inline_size(0);
3402
3403 memmove(inline_data + start, inline_data + start + skip, datal);
3404 }
3405
3406 write_extent_buffer(path->nodes[0], inline_data,
3407 btrfs_item_ptr_offset(path->nodes[0],
3408 path->slots[0]),
3409 size);
3410 inode_add_bytes(dst, datal);
3411
3412 return 0;
3413}
3414
32b7c687
MF		 3415/**
3416 * btrfs_clone() - clone a range from inode file to another
3417 *
3418 * @src: Inode to clone from
3419 * @inode: Inode to clone to
3420 * @off: Offset within source to start clone from
3421 * @olen: Original length, passed by user, of range to clone
1c919a5e 3422 * @olen_aligned: Block-aligned value of olen
32b7c687 3423 * @destoff: Offset within @inode to start clone
1c919a5e 3424 * @no_time_update: Whether to update mtime/ctime on the target inode
32b7c687
MF		 3425 */
3426static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901 3427 const u64 off, const u64 olen, const u64 olen_aligned,
1c919a5e 3428 const u64 destoff, int no_time_update)
f46b5a66 3429{
0b246afa 3430 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66 3431 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3432 struct btrfs_path *path = NULL;
f46b5a66 3433 struct extent_buffer *leaf;
32b7c687
MF		 3434 struct btrfs_trans_handle *trans;
3435 char *buf = NULL;
ae01a0ab 3436 struct btrfs_key key;
f46b5a66
CH		 3437 u32 nritems;
3438 int slot;
ae01a0ab 3439 int ret;
f82a9901 3440 const u64 len = olen_aligned;
f82a9901 3441 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3442
3443 ret = -ENOMEM;
752ade68
MH		 3444 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3445 if (!buf)
3446 return ret;
ae01a0ab
YZ		 3447
3448 path = btrfs_alloc_path();
3449 if (!path) {
15351955 3450 kvfree(buf);
32b7c687 3451 return ret;
d525e8ab
LZ		 3452 }
3453
e4058b54 3454 path->reada = READA_FORWARD;
c5c9cd4d 3455 /* clone data */
4a0cc7ca 3456 key.objectid = btrfs_ino(BTRFS_I(src));
ae01a0ab 3457 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3458 key.offset = off;
f46b5a66
CH		 3459
3460 while (1) {
de249e66 3461 u64 next_key_min_offset = key.offset + 1;
df858e76 3462
f46b5a66
CH		 3463 /*
3464 * note the key will change type as we walk through the
3465 * tree.
3466 */
e4355f34 3467 path->leave_spinning = 1;
362a20c5
DS		 3468 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3469 0, 0);
f46b5a66
CH		 3470 if (ret < 0)
3471 goto out;
2c463823
FM		 3472 /*
3473 * First search, if no extent item that starts at offset off was
3474 * found but the previous item is an extent item, it's possible
3475 * it might overlap our target range, therefore process it.
3476 */
3477 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3478 btrfs_item_key_to_cpu(path->nodes[0], &key,
3479 path->slots[0] - 1);
3480 if (key.type == BTRFS_EXTENT_DATA_KEY)
3481 path->slots[0]--;
3482 }
f46b5a66 3483
ae01a0ab 3484 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3485process_slot:
ae01a0ab 3486 if (path->slots[0] >= nritems) {
362a20c5 3487 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3488 if (ret < 0)
3489 goto out;
3490 if (ret > 0)
3491 break;
ae01a0ab 3492 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3493 }
3494 leaf = path->nodes[0];
3495 slot = path->slots[0];
f46b5a66 3496
ae01a0ab 3497 btrfs_item_key_to_cpu(leaf, &key, slot);
962a298f 3498 if (key.type > BTRFS_EXTENT_DATA_KEY ||
4a0cc7ca 3499 key.objectid != btrfs_ino(BTRFS_I(src)))
f46b5a66
CH		 3500 break;
3501
962a298f 3502 if (key.type == BTRFS_EXTENT_DATA_KEY) {
c5c9cd4d
SW		 3503 struct btrfs_file_extent_item *extent;
3504 int type;
31840ae1
ZY		 3505 u32 size;
3506 struct btrfs_key new_key;
c5c9cd4d
SW		 3507 u64 disko = 0, diskl = 0;
3508 u64 datao = 0, datal = 0;
3509 u8 comp;
f82a9901 3510 u64 drop_start;
31840ae1 3511
c5c9cd4d
SW		 3512 extent = btrfs_item_ptr(leaf, slot,
3513 struct btrfs_file_extent_item);
3514 comp = btrfs_file_extent_compression(leaf, extent);
3515 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3516 if (type == BTRFS_FILE_EXTENT_REG ||
3517 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3518 disko = btrfs_file_extent_disk_bytenr(leaf,
3519 extent);
3520 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3521 extent);
c5c9cd4d 3522 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3523 datal = btrfs_file_extent_num_bytes(leaf,
3524 extent);
c5c9cd4d
SW		 3525 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3526 /* take upper bound, may be compressed */
3527 datal = btrfs_file_extent_ram_bytes(leaf,
3528 extent);
3529 }
31840ae1 3530
2c463823
FM		 3531 /*
3532 * The first search might have left us at an extent
3533 * item that ends before our target range's start, can
3534 * happen if we have holes and NO_HOLES feature enabled.
3535 */
3536 if (key.offset + datal <= off) {
e4355f34
FDBM		 3537 path->slots[0]++;
3538 goto process_slot;
2c463823
FM		 3539 } else if (key.offset >= off + len) {
3540 break;
e4355f34 3541 }
df858e76 3542 next_key_min_offset = key.offset + datal;
e4355f34
FDBM		 3543 size = btrfs_item_size_nr(leaf, slot);
3544 read_extent_buffer(leaf, buf,
3545 btrfs_item_ptr_offset(leaf, slot),
3546 size);
3547
3548 btrfs_release_path(path);
3549 path->leave_spinning = 0;
c5c9cd4d 3550
31840ae1 3551 memcpy(&new_key, &key, sizeof(new_key));
4a0cc7ca 3552 new_key.objectid = btrfs_ino(BTRFS_I(inode));
4d728ec7
LZ		 3553 if (off <= key.offset)
3554 new_key.offset = key.offset + destoff - off;
3555 else
3556 new_key.offset = destoff;
31840ae1 3557
f82a9901
FM		 3558 /*
3559 * Deal with a hole that doesn't have an extent item
3560 * that represents it (NO_HOLES feature enabled).
3561 * This hole is either in the middle of the cloning
3562 * range or at the beginning (fully overlaps it or
3563 * partially overlaps it).
3564 */
3565 if (new_key.offset != last_dest_end)
3566 drop_start = last_dest_end;
3567 else
3568 drop_start = new_key.offset;
3569
b6f3409b
SW		 3570 /*
3571 * 1 - adjusting old extent (we may have to split it)
3572 * 1 - add new extent
3573 * 1 - inode update
3574 */
3575 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3576 if (IS_ERR(trans)) {
3577 ret = PTR_ERR(trans);
3578 goto out;
3579 }
3580
c8a894d7
CM		 3581 if (type == BTRFS_FILE_EXTENT_REG ||
3582 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3583 /*
3584 * a | --- range to clone ---| b
3585 * | ------------- extent ------------- |
3586 */
3587
93915584 3588 /* subtract range b */
d72c0842
LZ		 3589 if (key.offset + datal > off + len)
3590 datal = off + len - key.offset;
3591
93915584 3592 /* subtract range a */
a22285a6
YZ		 3593 if (off > key.offset) {
3594 datao += off - key.offset;
3595 datal -= off - key.offset;
3596 }
3597
5dc562c5 3598 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3599 drop_start,
a22285a6 3600 new_key.offset + datal,
2671485d 3601 1);
79787eaa 3602 if (ret) {
3f9e3df8 3603 if (ret != -EOPNOTSUPP)
00fdf13a 3604 btrfs_abort_transaction(trans,
66642832 3605 ret);
3a45bb20 3606 btrfs_end_transaction(trans);
79787eaa
JM		 3607 goto out;
3608 }
a22285a6 3609
c5c9cd4d
SW		 3610 ret = btrfs_insert_empty_item(trans, root, path,
3611 &new_key, size);
79787eaa 3612 if (ret) {
66642832 3613 btrfs_abort_transaction(trans, ret);
3a45bb20 3614 btrfs_end_transaction(trans);
79787eaa
JM		 3615 goto out;
3616 }
c5c9cd4d
SW		 3617
3618 leaf = path->nodes[0];
3619 slot = path->slots[0];
3620 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3621 btrfs_item_ptr_offset(leaf, slot),
3622 size);
ae01a0ab 3623
c5c9cd4d 3624 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3625 struct btrfs_file_extent_item);
c5c9cd4d 3626
c5c9cd4d
SW		 3627 /* disko == 0 means it's a hole */
3628 if (!disko)
3629 datao = 0;
c5c9cd4d
SW		 3630
3631 btrfs_set_file_extent_offset(leaf, extent,
3632 datao);
3633 btrfs_set_file_extent_num_bytes(leaf, extent,
3634 datal);
fcebe456 3635
c5c9cd4d
SW		 3636 if (disko) {
3637 inode_add_bytes(inode, datal);
2ff7e61e 3638 ret = btrfs_inc_extent_ref(trans,
84f7d8e6 3639 root,
5d4f98a2
YZ		 3640 disko, diskl, 0,
3641 root->root_key.objectid,
4a0cc7ca 3642 btrfs_ino(BTRFS_I(inode)),
b06c4bf5 3643 new_key.offset - datao);
79787eaa
JM		 3644 if (ret) {
3645 btrfs_abort_transaction(trans,
79787eaa 3646 ret);
3a45bb20 3647 btrfs_end_transaction(trans);
79787eaa
JM		 3648 goto out;
3649
3650 }
f46b5a66 3651 }
c5c9cd4d
SW		 3652 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3653 u64 skip = 0;
3654 u64 trim = 0;
ed958762 3655
c5c9cd4d
SW		 3656 if (off > key.offset) {
3657 skip = off - key.offset;
3658 new_key.offset += skip;
3659 }
d397712b 3660
aa42ffd9
LB		 3661 if (key.offset + datal > off + len)
3662 trim = key.offset + datal - (off + len);
d397712b 3663
c5c9cd4d 3664 if (comp && (skip || trim)) {
c5c9cd4d 3665 ret = -EINVAL;
3a45bb20 3666 btrfs_end_transaction(trans);
c5c9cd4d
SW		 3667 goto out;
3668 }
3669 size -= skip + trim;
3670 datal -= skip + trim;
a22285a6 3671
4a0ab9d7 3672 ret = clone_copy_inline_extent(inode,
8039d87d
FM		 3673 trans, path,
3674 &new_key,
3675 drop_start,
3676 datal,
3677 skip, size, buf);
79787eaa 3678 if (ret) {
3f9e3df8 3679 if (ret != -EOPNOTSUPP)
3a29bc09 3680 btrfs_abort_transaction(trans,
8039d87d 3681 ret);
3a45bb20 3682 btrfs_end_transaction(trans);
79787eaa
JM		 3683 goto out;
3684 }
c5c9cd4d
SW		 3685 leaf = path->nodes[0];
3686 slot = path->slots[0];
f46b5a66 3687 }
c5c9cd4d 3688
7ffbb598
FM		 3689 /* If we have an implicit hole (NO_HOLES feature). */
3690 if (drop_start < new_key.offset)
a2f392e4 3691 clone_update_extent_map(BTRFS_I(inode), trans,
14f59796 3692 NULL, drop_start,
7ffbb598
FM		 3693 new_key.offset - drop_start);
3694
a2f392e4
NB		 3695 clone_update_extent_map(BTRFS_I(inode), trans,
3696 path, 0, 0);
7ffbb598 3697
c5c9cd4d 3698 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3699 btrfs_release_path(path);
c5c9cd4d 3700
62e2390e 3701 last_dest_end = ALIGN(new_key.offset + datal,
0b246afa 3702 fs_info->sectorsize);
f82a9901
FM		 3703 ret = clone_finish_inode_update(trans, inode,
3704 last_dest_end,
1c919a5e
MF		 3705 destoff, olen,
3706 no_time_update);
f82a9901 3707 if (ret)
79787eaa 3708 goto out;
2c463823
FM		 3709 if (new_key.offset + datal >= destoff + len)
3710 break;
a22285a6 3711 }
b3b4aa74 3712 btrfs_release_path(path);
df858e76 3713 key.offset = next_key_min_offset;
69ae5e44
WX		 3714
3715 if (fatal_signal_pending(current)) {
3716 ret = -EINTR;
3717 goto out;
3718 }
f46b5a66 3719 }
f46b5a66 3720 ret = 0;
32b7c687 3721
f82a9901
FM		 3722 if (last_dest_end < destoff + len) {
3723 /*
3724 * We have an implicit hole (NO_HOLES feature is enabled) that
3725 * fully or partially overlaps our cloning range at its end.
3726 */
3727 btrfs_release_path(path);
3728
3729 /*
3730 * 1 - remove extent(s)
3731 * 1 - inode update
3732 */
3733 trans = btrfs_start_transaction(root, 2);
3734 if (IS_ERR(trans)) {
3735 ret = PTR_ERR(trans);
3736 goto out;
3737 }
3738 ret = btrfs_drop_extents(trans, root, inode,
3739 last_dest_end, destoff + len, 1);
3740 if (ret) {
3741 if (ret != -EOPNOTSUPP)
66642832 3742 btrfs_abort_transaction(trans, ret);
3a45bb20 3743 btrfs_end_transaction(trans);
f82a9901
FM		 3744 goto out;
3745 }
a2f392e4
NB		 3746 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3747 last_dest_end,
3748 destoff + len - last_dest_end);
f82a9901 3749 ret = clone_finish_inode_update(trans, inode, destoff + len,
1c919a5e 3750 destoff, olen, no_time_update);
f82a9901
FM		 3751 }
3752
f46b5a66 3753out:
32b7c687 3754 btrfs_free_path(path);
15351955 3755 kvfree(buf);
32b7c687
MF		 3756 return ret;
3757}
3758
3db11b2e
ZB		 3759static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3760 u64 off, u64 olen, u64 destoff)
32b7c687 3761{
54563d41 3762 struct inode *inode = file_inode(file);
3db11b2e 3763 struct inode *src = file_inode(file_src);
0b246afa 3764 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
32b7c687 3765 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687
MF		 3766 int ret;
3767 u64 len = olen;
0b246afa 3768 u64 bs = fs_info->sb->s_blocksize;
3db11b2e 3769 int same_inode = src == inode;
32b7c687
MF		 3770
3771 /*
3772 * TODO:
3773 * - split compressed inline extents. annoying: we need to
3774 * decompress into destination's address_space (the file offset
3775 * may change, so source mapping won't do), then recompress (or
3776 * otherwise reinsert) a subrange.
00fdf13a
LB		 3777 *
3778 * - split destination inode's inline extents. The inline extents can
3779 * be either compressed or non-compressed.
32b7c687
MF		 3780 */
3781
32b7c687
MF		 3782 if (btrfs_root_readonly(root))
3783 return -EROFS;
3784
3db11b2e
ZB		 3785 if (file_src->f_path.mnt != file->f_path.mnt ||
3786 src->i_sb != inode->i_sb)
3787 return -EXDEV;
32b7c687
MF		 3788
3789 /* don't make the dst file partly checksummed */
3790 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3791 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3db11b2e 3792 return -EINVAL;
32b7c687 3793
32b7c687 3794 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3db11b2e 3795 return -EISDIR;
32b7c687
MF		 3796
3797 if (!same_inode) {
293a8489 3798 btrfs_double_inode_lock(src, inode);
32b7c687 3799 } else {
5955102c 3800 inode_lock(src);
32b7c687
MF		 3801 }
3802
3803 /* determine range to clone */
3804 ret = -EINVAL;
3805 if (off + len > src->i_size || off + len < off)
3806 goto out_unlock;
3807 if (len == 0)
3808 olen = len = src->i_size - off;
3809 /* if we extend to eof, continue to block boundary */
3810 if (off + len == src->i_size)
3811 len = ALIGN(src->i_size, bs) - off;
3812
ccccf3d6
FM		 3813 if (len == 0) {
3814 ret = 0;
3815 goto out_unlock;
3816 }
3817
32b7c687
MF		 3818 /* verify the end result is block aligned */
3819 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3820 !IS_ALIGNED(destoff, bs))
3821 goto out_unlock;
3822
3823 /* verify if ranges are overlapped within the same file */
3824 if (same_inode) {
3825 if (destoff + len > off && destoff < off + len)
3826 goto out_unlock;
3827 }
3828
3829 if (destoff > inode->i_size) {
3830 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3831 if (ret)
3832 goto out_unlock;
3833 }
3834
c125b8bf
FM		 3835 /*
3836 * Lock the target range too. Right after we replace the file extent
3837 * items in the fs tree (which now point to the cloned data), we might
3838 * have a worker replace them with extent items relative to a write
3839 * operation that was issued before this clone operation (i.e. confront
3840 * with inode.c:btrfs_finish_ordered_io).
3841 */
3842 if (same_inode) {
3843 u64 lock_start = min_t(u64, off, destoff);
3844 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
32b7c687 3845
e0bd70c6 3846 ret = lock_extent_range(src, lock_start, lock_len, true);
c125b8bf 3847 } else {
e0bd70c6
FM		 3848 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3849 true);
3850 }
3851 ASSERT(ret == 0);
3852 if (WARN_ON(ret)) {
3853 /* ranges in the io trees already unlocked */
3854 goto out_unlock;
c125b8bf 3855 }
32b7c687 3856
1c919a5e 3857 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
32b7c687 3858
c125b8bf
FM		 3859 if (same_inode) {
3860 u64 lock_start = min_t(u64, off, destoff);
3861 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3862
3863 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3864 } else {
293a8489 3865 btrfs_double_extent_unlock(src, off, inode, destoff, len);
c125b8bf
FM		 3866 }
3867 /*
3868 * Truncate page cache pages so that future reads will see the cloned
3869 * data immediately and not the previous data.
3870 */
65bfa658 3871 truncate_inode_pages_range(&inode->i_data,
09cbfeaf
KS		 3872 round_down(destoff, PAGE_SIZE),
3873 round_up(destoff + len, PAGE_SIZE) - 1);
f46b5a66 3874out_unlock:
293a8489
MF		 3875 if (!same_inode)
3876 btrfs_double_inode_unlock(src, inode);
3877 else
5955102c 3878 inode_unlock(src);
3db11b2e
ZB		 3879 return ret;
3880}
3881
04b38d60
CH		 3882int btrfs_clone_file_range(struct file *src_file, loff_t off,
3883 struct file *dst_file, loff_t destoff, u64 len)
3db11b2e 3884{
04b38d60 3885 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
c5c9cd4d
SW		 3886}
3887
6ef5ed0d
JB		 3888static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3889{
496ad9aa 3890 struct inode *inode = file_inode(file);
0b246afa 3891 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6ef5ed0d
JB		 3892 struct btrfs_root *root = BTRFS_I(inode)->root;
3893 struct btrfs_root *new_root;
3894 struct btrfs_dir_item *di;
3895 struct btrfs_trans_handle *trans;
3896 struct btrfs_path *path;
3897 struct btrfs_key location;
3898 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3899 u64 objectid = 0;
3900 u64 dir_id;
3c04ce01 3901 int ret;
6ef5ed0d
JB		 3902
3903 if (!capable(CAP_SYS_ADMIN))
3904 return -EPERM;
3905
3c04ce01
MX		 3906 ret = mnt_want_write_file(file);
3907 if (ret)
3908 return ret;
3909
3910 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3911 ret = -EFAULT;
3912 goto out;
3913 }
6ef5ed0d
JB		 3914
3915 if (!objectid)
1cecf579 3916 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 3917
3918 location.objectid = objectid;
3919 location.type = BTRFS_ROOT_ITEM_KEY;
3920 location.offset = (u64)-1;
3921
0b246afa 3922 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
3c04ce01
MX		 3923 if (IS_ERR(new_root)) {
3924 ret = PTR_ERR(new_root);
3925 goto out;
3926 }
6d6d2829 3927 if (!is_fstree(new_root->objectid)) {
3928 ret = -ENOENT;
3929 goto out;
3930 }
6ef5ed0d 3931
6ef5ed0d 3932 path = btrfs_alloc_path();
3c04ce01
MX		 3933 if (!path) {
3934 ret = -ENOMEM;
3935 goto out;
3936 }
6ef5ed0d
JB		 3937 path->leave_spinning = 1;
3938
3939 trans = btrfs_start_transaction(root, 1);
98d5dc13 3940 if (IS_ERR(trans)) {
6ef5ed0d 3941 btrfs_free_path(path);
3c04ce01
MX		 3942 ret = PTR_ERR(trans);
3943 goto out;
6ef5ed0d
JB		 3944 }
3945
0b246afa
JM		 3946 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3947 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
6ef5ed0d 3948 dir_id, "default", 7, 1);
cf1e99a4 3949 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d 3950 btrfs_free_path(path);
3a45bb20 3951 btrfs_end_transaction(trans);
0b246afa 3952 btrfs_err(fs_info,
5d163e0e 3953 "Umm, you don't have the default diritem, this isn't going to work");
3c04ce01
MX		 3954 ret = -ENOENT;
3955 goto out;
6ef5ed0d
JB		 3956 }
3957
3958 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3959 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3960 btrfs_mark_buffer_dirty(path->nodes[0]);
3961 btrfs_free_path(path);
3962
0b246afa 3963 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3a45bb20 3964 btrfs_end_transaction(trans);
3c04ce01
MX		 3965out:
3966 mnt_drop_write_file(file);
3967 return ret;
6ef5ed0d
JB		 3968}
3969
c065f5b1
SY		 3970static void get_block_group_info(struct list_head *groups_list,
3971 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3972{
3973 struct btrfs_block_group_cache *block_group;
3974
3975 space->total_bytes = 0;
3976 space->used_bytes = 0;
3977 space->flags = 0;
3978 list_for_each_entry(block_group, groups_list, list) {
3979 space->flags = block_group->flags;
3980 space->total_bytes += block_group->key.offset;
3981 space->used_bytes +=
3982 btrfs_block_group_used(&block_group->item);
3983 }
3984}
3985
2ff7e61e
JM		 3986static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3987 void __user *arg)
1406e432
JB		 3988{
3989 struct btrfs_ioctl_space_args space_args;
3990 struct btrfs_ioctl_space_info space;
3991 struct btrfs_ioctl_space_info *dest;
7fde62bf 3992 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3993 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3994 struct btrfs_space_info *info;
315d8e98
CIK		 3995 static const u64 types[] = {
3996 BTRFS_BLOCK_GROUP_DATA,
3997 BTRFS_BLOCK_GROUP_SYSTEM,
3998 BTRFS_BLOCK_GROUP_METADATA,
3999 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4000 };
bf5fc093 4001 int num_types = 4;
7fde62bf 4002 int alloc_size;
1406e432 4003 int ret = 0;
51788b1b 4004 u64 slot_count = 0;
bf5fc093 4005 int i, c;
1406e432
JB		 4006
4007 if (copy_from_user(&space_args,
4008 (struct btrfs_ioctl_space_args __user *)arg,
4009 sizeof(space_args)))
4010 return -EFAULT;
4011
bf5fc093
JB		 4012 for (i = 0; i < num_types; i++) {
4013 struct btrfs_space_info *tmp;
4014
4015 info = NULL;
4016 rcu_read_lock();
0b246afa 4017 list_for_each_entry_rcu(tmp, &fs_info->space_info,
bf5fc093
JB		 4018 list) {
4019 if (tmp->flags == types[i]) {
4020 info = tmp;
4021 break;
4022 }
4023 }
4024 rcu_read_unlock();
4025
4026 if (!info)
4027 continue;
4028
4029 down_read(&info->groups_sem);
4030 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4031 if (!list_empty(&info->block_groups[c]))
4032 slot_count++;
4033 }
4034 up_read(&info->groups_sem);
4035 }
7fde62bf 4036
36523e95
DS		 4037 /*
4038 * Global block reserve, exported as a space_info
4039 */
4040 slot_count++;
4041
7fde62bf
CM		 4042 /* space_slots == 0 means they are asking for a count */
4043 if (space_args.space_slots == 0) {
4044 space_args.total_spaces = slot_count;
4045 goto out;
4046 }
bf5fc093 4047
51788b1b 4048 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 4049
7fde62bf 4050 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 4051
7fde62bf
CM		 4052 /* we generally have at most 6 or so space infos, one for each raid
4053 * level. So, a whole page should be more than enough for everyone
4054 */
09cbfeaf 4055 if (alloc_size > PAGE_SIZE)
7fde62bf
CM		 4056 return -ENOMEM;
4057
1406e432 4058 space_args.total_spaces = 0;
8d2db785 4059 dest = kmalloc(alloc_size, GFP_KERNEL);
7fde62bf
CM		 4060 if (!dest)
4061 return -ENOMEM;
4062 dest_orig = dest;
1406e432 4063
7fde62bf 4064 /* now we have a buffer to copy into */
bf5fc093
JB		 4065 for (i = 0; i < num_types; i++) {
4066 struct btrfs_space_info *tmp;
4067
51788b1b
DR		 4068 if (!slot_count)
4069 break;
4070
bf5fc093
JB		 4071 info = NULL;
4072 rcu_read_lock();
0b246afa 4073 list_for_each_entry_rcu(tmp, &fs_info->space_info,
bf5fc093
JB		 4074 list) {
4075 if (tmp->flags == types[i]) {
4076 info = tmp;
4077 break;
4078 }
4079 }
4080 rcu_read_unlock();
7fde62bf 4081
bf5fc093
JB		 4082 if (!info)
4083 continue;
4084 down_read(&info->groups_sem);
4085 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4086 if (!list_empty(&info->block_groups[c])) {
c065f5b1
SY		 4087 get_block_group_info(&info->block_groups[c],
4088 &space);
bf5fc093
JB		 4089 memcpy(dest, &space, sizeof(space));
4090 dest++;
4091 space_args.total_spaces++;
51788b1b 4092 slot_count--;
bf5fc093 4093 }
51788b1b
DR		 4094 if (!slot_count)
4095 break;
bf5fc093
JB		 4096 }
4097 up_read(&info->groups_sem);
1406e432 4098 }
1406e432 4099
36523e95
DS		 4100 /*
4101 * Add global block reserve
4102 */
4103 if (slot_count) {
0b246afa 4104 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
36523e95
DS		 4105
4106 spin_lock(&block_rsv->lock);
4107 space.total_bytes = block_rsv->size;
4108 space.used_bytes = block_rsv->size - block_rsv->reserved;
4109 spin_unlock(&block_rsv->lock);
4110 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4111 memcpy(dest, &space, sizeof(space));
4112 space_args.total_spaces++;
4113 }
4114
2eec6c81 4115 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 4116 (arg + sizeof(struct btrfs_ioctl_space_args));
4117
4118 if (copy_to_user(user_dest, dest_orig, alloc_size))
4119 ret = -EFAULT;
4120
4121 kfree(dest_orig);
4122out:
4123 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 4124 ret = -EFAULT;
4125
4126 return ret;
4127}
4128
9a8c28be
MX		 4129static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4130 void __user *argp)
46204592 4131{
46204592
SW		 4132 struct btrfs_trans_handle *trans;
4133 u64 transid;
db5b493a 4134 int ret;
46204592 4135
d4edf39b 4136 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4137 if (IS_ERR(trans)) {
4138 if (PTR_ERR(trans) != -ENOENT)
4139 return PTR_ERR(trans);
4140
4141 /* No running transaction, don't bother */
4142 transid = root->fs_info->last_trans_committed;
4143 goto out;
4144 }
46204592 4145 transid = trans->transid;
3a45bb20 4146 ret = btrfs_commit_transaction_async(trans, 0);
8b2b2d3c 4147 if (ret) {
3a45bb20 4148 btrfs_end_transaction(trans);
db5b493a 4149 return ret;
8b2b2d3c 4150 }
ff7c1d33 4151out:
46204592
SW		 4152 if (argp)
4153 if (copy_to_user(argp, &transid, sizeof(transid)))
4154 return -EFAULT;
4155 return 0;
4156}
4157
2ff7e61e 4158static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
9a8c28be 4159 void __user *argp)
46204592 4160{
46204592
SW		 4161 u64 transid;
4162
4163 if (argp) {
4164 if (copy_from_user(&transid, argp, sizeof(transid)))
4165 return -EFAULT;
4166 } else {
4167 transid = 0; /* current trans */
4168 }
2ff7e61e 4169 return btrfs_wait_for_commit(fs_info, transid);
46204592
SW		 4170}
4171
b8e95489 4172static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4173{
0b246afa 4174 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
475f6387 4175 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4176 int ret;
475f6387
JS		 4177
4178 if (!capable(CAP_SYS_ADMIN))
4179 return -EPERM;
4180
4181 sa = memdup_user(arg, sizeof(*sa));
4182 if (IS_ERR(sa))
4183 return PTR_ERR(sa);
4184
b8e95489
MX		 4185 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4186 ret = mnt_want_write_file(file);
4187 if (ret)
4188 goto out;
4189 }
4190
0b246afa 4191 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4192 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4193 0);
475f6387
JS		 4194
4195 if (copy_to_user(arg, sa, sizeof(*sa)))
4196 ret = -EFAULT;
4197
b8e95489
MX		 4198 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4199 mnt_drop_write_file(file);
4200out:
475f6387
JS		 4201 kfree(sa);
4202 return ret;
4203}
4204
2ff7e61e 4205static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
475f6387
JS		 4206{
4207 if (!capable(CAP_SYS_ADMIN))
4208 return -EPERM;
4209
2ff7e61e 4210 return btrfs_scrub_cancel(fs_info);
475f6387
JS		 4211}
4212
2ff7e61e 4213static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
475f6387
JS		 4214 void __user *arg)
4215{
4216 struct btrfs_ioctl_scrub_args *sa;
4217 int ret;
4218
4219 if (!capable(CAP_SYS_ADMIN))
4220 return -EPERM;
4221
4222 sa = memdup_user(arg, sizeof(*sa));
4223 if (IS_ERR(sa))
4224 return PTR_ERR(sa);
4225
2ff7e61e 4226 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
475f6387
JS		 4227
4228 if (copy_to_user(arg, sa, sizeof(*sa)))
4229 ret = -EFAULT;
4230
4231 kfree(sa);
4232 return ret;
4233}
4234
2ff7e61e 4235static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
b27f7c0c 4236 void __user *arg)
c11d2c23
SB		 4237{
4238 struct btrfs_ioctl_get_dev_stats *sa;
4239 int ret;
4240
c11d2c23
SB		 4241 sa = memdup_user(arg, sizeof(*sa));
4242 if (IS_ERR(sa))
4243 return PTR_ERR(sa);
4244
b27f7c0c
DS		 4245 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4246 kfree(sa);
4247 return -EPERM;
4248 }
4249
2ff7e61e 4250 ret = btrfs_get_dev_stats(fs_info, sa);
c11d2c23
SB		 4251
4252 if (copy_to_user(arg, sa, sizeof(*sa)))
4253 ret = -EFAULT;
4254
4255 kfree(sa);
4256 return ret;
4257}
4258
2ff7e61e
JM		 4259static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4260 void __user *arg)
3f6bcfbd
SB		 4261{
4262 struct btrfs_ioctl_dev_replace_args *p;
4263 int ret;
4264
4265 if (!capable(CAP_SYS_ADMIN))
4266 return -EPERM;
4267
4268 p = memdup_user(arg, sizeof(*p));
4269 if (IS_ERR(p))
4270 return PTR_ERR(p);
4271
4272 switch (p->cmd) {
4273 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
bc98a42c 4274 if (sb_rdonly(fs_info->sb)) {
adfa97cb
ID		 4275 ret = -EROFS;
4276 goto out;
4277 }
171938e5 4278 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
e57138b3 4279 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd 4280 } else {
2ff7e61e 4281 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
171938e5 4282 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3f6bcfbd
SB		 4283 }
4284 break;
4285 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
0b246afa 4286 btrfs_dev_replace_status(fs_info, p);
3f6bcfbd
SB		 4287 ret = 0;
4288 break;
4289 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
17d202b9 4290 p->result = btrfs_dev_replace_cancel(fs_info);
97282031 4291 ret = 0;
3f6bcfbd
SB		 4292 break;
4293 default:
4294 ret = -EINVAL;
4295 break;
4296 }
4297
4298 if (copy_to_user(arg, p, sizeof(*p)))
4299 ret = -EFAULT;
adfa97cb 4300out:
3f6bcfbd
SB		 4301 kfree(p);
4302 return ret;
4303}
4304
d7728c96
JS		 4305static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4306{
4307 int ret = 0;
4308 int i;
740c3d22 4309 u64 rel_ptr;
d7728c96 4310 int size;
806468f8 4311 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4312 struct inode_fs_paths *ipath = NULL;
4313 struct btrfs_path *path;
4314
82b22ac8 4315 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4316 return -EPERM;
4317
4318 path = btrfs_alloc_path();
4319 if (!path) {
4320 ret = -ENOMEM;
4321 goto out;
4322 }
4323
4324 ipa = memdup_user(arg, sizeof(*ipa));
4325 if (IS_ERR(ipa)) {
4326 ret = PTR_ERR(ipa);
4327 ipa = NULL;
4328 goto out;
4329 }
4330
4331 size = min_t(u32, ipa->size, 4096);
4332 ipath = init_ipath(size, root, path);
4333 if (IS_ERR(ipath)) {
4334 ret = PTR_ERR(ipath);
4335 ipath = NULL;
4336 goto out;
4337 }
4338
4339 ret = paths_from_inode(ipa->inum, ipath);
4340 if (ret < 0)
4341 goto out;
4342
4343 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4344 rel_ptr = ipath->fspath->val[i] -
4345 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4346 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4347 }
4348
718dc5fa
OS		 4349 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4350 ipath->fspath, size);
d7728c96
JS		 4351 if (ret) {
4352 ret = -EFAULT;
4353 goto out;
4354 }
4355
4356out:
4357 btrfs_free_path(path);
4358 free_ipath(ipath);
4359 kfree(ipa);
4360
4361 return ret;
4362}
4363
4364static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4365{
4366 struct btrfs_data_container *inodes = ctx;
4367 const size_t c = 3 * sizeof(u64);
4368
4369 if (inodes->bytes_left >= c) {
4370 inodes->bytes_left -= c;
4371 inodes->val[inodes->elem_cnt] = inum;
4372 inodes->val[inodes->elem_cnt + 1] = offset;
4373 inodes->val[inodes->elem_cnt + 2] = root;
4374 inodes->elem_cnt += 3;
4375 } else {
4376 inodes->bytes_missing += c - inodes->bytes_left;
4377 inodes->bytes_left = 0;
4378 inodes->elem_missed += 3;
4379 }
4380
4381 return 0;
4382}
4383
2ff7e61e 4384static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
d24a67b2 4385 void __user *arg, int version)
d7728c96
JS		 4386{
4387 int ret = 0;
4388 int size;
d7728c96
JS		 4389 struct btrfs_ioctl_logical_ino_args *loi;
4390 struct btrfs_data_container *inodes = NULL;
4391 struct btrfs_path *path = NULL;
d24a67b2 4392 bool ignore_offset;
d7728c96
JS		 4393
4394 if (!capable(CAP_SYS_ADMIN))
4395 return -EPERM;
4396
4397 loi = memdup_user(arg, sizeof(*loi));
7b9ea627
SV		 4398 if (IS_ERR(loi))
4399 return PTR_ERR(loi);
d7728c96 4400
d24a67b2
ZB		 4401 if (version == 1) {
4402 ignore_offset = false;
b115e3bc 4403 size = min_t(u32, loi->size, SZ_64K);
d24a67b2
ZB		 4404 } else {
4405 /* All reserved bits must be 0 for now */
4406 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4407 ret = -EINVAL;
4408 goto out_loi;
4409 }
4410 /* Only accept flags we have defined so far */
4411 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4412 ret = -EINVAL;
4413 goto out_loi;
4414 }
4415 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
b115e3bc 4416 size = min_t(u32, loi->size, SZ_16M);
d24a67b2
ZB		 4417 }
4418
d7728c96
JS		 4419 path = btrfs_alloc_path();
4420 if (!path) {
4421 ret = -ENOMEM;
4422 goto out;
4423 }
4424
d7728c96
JS		 4425 inodes = init_data_container(size);
4426 if (IS_ERR(inodes)) {
4427 ret = PTR_ERR(inodes);
4428 inodes = NULL;
4429 goto out;
4430 }
4431
2ff7e61e 4432 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
d24a67b2 4433 build_ino_list, inodes, ignore_offset);
df031f07 4434 if (ret == -EINVAL)
d7728c96
JS		 4435 ret = -ENOENT;
4436 if (ret < 0)
4437 goto out;
4438
718dc5fa
OS		 4439 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4440 size);
d7728c96
JS		 4441 if (ret)
4442 ret = -EFAULT;
4443
4444out:
4445 btrfs_free_path(path);
f54de068 4446 kvfree(inodes);
d24a67b2 4447out_loi:
d7728c96
JS		 4448 kfree(loi);
4449
4450 return ret;
4451}
4452
008ef096 4453void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
c9e9f97b
ID		 4454 struct btrfs_ioctl_balance_args *bargs)
4455{
4456 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4457
4458 bargs->flags = bctl->flags;
4459
3009a62f 4460 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
837d5b6e
ID		 4461 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4462 if (atomic_read(&fs_info->balance_pause_req))
4463 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4464 if (atomic_read(&fs_info->balance_cancel_req))
4465 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4466
c9e9f97b
ID		 4467 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4468 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4469 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce 4470
008ef096
DS		 4471 spin_lock(&fs_info->balance_lock);
4472 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4473 spin_unlock(&fs_info->balance_lock);
c9e9f97b
ID		 4474}
4475
9ba1f6e4 4476static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4477{
496ad9aa 4478 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4479 struct btrfs_fs_info *fs_info = root->fs_info;
4480 struct btrfs_ioctl_balance_args *bargs;
4481 struct btrfs_balance_control *bctl;
ed0fb78f 4482 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4483 int ret;
4484
4485 if (!capable(CAP_SYS_ADMIN))
4486 return -EPERM;
4487
e54bfa31 4488 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4489 if (ret)
4490 return ret;
4491
ed0fb78f 4492again:
171938e5 4493 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
ed0fb78f
ID		 4494 mutex_lock(&fs_info->balance_mutex);
4495 need_unlock = true;
4496 goto locked;
4497 }
4498
4499 /*
01327610 4500 * mut. excl. ops lock is locked. Three possibilities:
ed0fb78f
ID		 4501 * (1) some other op is running
4502 * (2) balance is running
4503 * (3) balance is paused -- special case (think resume)
4504 */
c9e9f97b 4505 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4506 if (fs_info->balance_ctl) {
4507 /* this is either (2) or (3) */
3009a62f 4508 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
ed0fb78f 4509 mutex_unlock(&fs_info->balance_mutex);
dccdb07b
DS		 4510 /*
4511 * Lock released to allow other waiters to continue,
4512 * we'll reexamine the status again.
4513 */
ed0fb78f
ID		 4514 mutex_lock(&fs_info->balance_mutex);
4515
4516 if (fs_info->balance_ctl &&
3009a62f 4517 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
ed0fb78f
ID		 4518 /* this is (3) */
4519 need_unlock = false;
4520 goto locked;
4521 }
4522
4523 mutex_unlock(&fs_info->balance_mutex);
ed0fb78f
ID		 4524 goto again;
4525 } else {
4526 /* this is (2) */
4527 mutex_unlock(&fs_info->balance_mutex);
4528 ret = -EINPROGRESS;
4529 goto out;
4530 }
4531 } else {
4532 /* this is (1) */
4533 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4534 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4535 goto out;
4536 }
4537
4538locked:
171938e5 4539 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
c9e9f97b
ID		 4540
4541 if (arg) {
4542 bargs = memdup_user(arg, sizeof(*bargs));
4543 if (IS_ERR(bargs)) {
4544 ret = PTR_ERR(bargs);
ed0fb78f 4545 goto out_unlock;
c9e9f97b 4546 }
de322263
ID		 4547
4548 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4549 if (!fs_info->balance_ctl) {
4550 ret = -ENOTCONN;
4551 goto out_bargs;
4552 }
4553
4554 bctl = fs_info->balance_ctl;
4555 spin_lock(&fs_info->balance_lock);
4556 bctl->flags |= BTRFS_BALANCE_RESUME;
4557 spin_unlock(&fs_info->balance_lock);
4558
4559 goto do_balance;
4560 }
c9e9f97b
ID		 4561 } else {
4562 bargs = NULL;
4563 }
4564
ed0fb78f 4565 if (fs_info->balance_ctl) {
837d5b6e
ID		 4566 ret = -EINPROGRESS;
4567 goto out_bargs;
4568 }
4569
8d2db785 4570 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
c9e9f97b
ID		 4571 if (!bctl) {
4572 ret = -ENOMEM;
4573 goto out_bargs;
4574 }
4575
c9e9f97b
ID		 4576 if (arg) {
4577 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4578 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4579 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4580
4581 bctl->flags = bargs->flags;
f43ffb60
ID		 4582 } else {
4583 /* balance everything - no filters */
4584 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4585 }
4586
8eb93459
DS		 4587 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4588 ret = -EINVAL;
0f89abf5 4589 goto out_bctl;
8eb93459
DS		 4590 }
4591
de322263 4592do_balance:
c9e9f97b 4593 /*
149196a2
DS		 4594 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
4595 * btrfs_balance. bctl is freed in reset_balance_state, or, if
4596 * restriper was paused all the way until unmount, in free_fs_info.
4597 * The flag should be cleared after reset_balance_state.
c9e9f97b 4598 */
ed0fb78f
ID		 4599 need_unlock = false;
4600
6fcf6e2b 4601 ret = btrfs_balance(fs_info, bctl, bargs);
0f89abf5 4602 bctl = NULL;
ed0fb78f 4603
c9e9f97b
ID		 4604 if (arg) {
4605 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4606 ret = -EFAULT;
4607 }
4608
0f89abf5
CE		 4609out_bctl:
4610 kfree(bctl);
c9e9f97b
ID		 4611out_bargs:
4612 kfree(bargs);
ed0fb78f 4613out_unlock:
c9e9f97b 4614 mutex_unlock(&fs_info->balance_mutex);
ed0fb78f 4615 if (need_unlock)
171938e5 4616 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
ed0fb78f 4617out:
e54bfa31 4618 mnt_drop_write_file(file);
c9e9f97b
ID		 4619 return ret;
4620}
4621
2ff7e61e 4622static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
837d5b6e
ID		 4623{
4624 if (!capable(CAP_SYS_ADMIN))
4625 return -EPERM;
4626
4627 switch (cmd) {
4628 case BTRFS_BALANCE_CTL_PAUSE:
0b246afa 4629 return btrfs_pause_balance(fs_info);
a7e99c69 4630 case BTRFS_BALANCE_CTL_CANCEL:
0b246afa 4631 return btrfs_cancel_balance(fs_info);
837d5b6e
ID		 4632 }
4633
4634 return -EINVAL;
4635}
4636
2ff7e61e 4637static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
19a39dce
ID		 4638 void __user *arg)
4639{
19a39dce
ID		 4640 struct btrfs_ioctl_balance_args *bargs;
4641 int ret = 0;
4642
4643 if (!capable(CAP_SYS_ADMIN))
4644 return -EPERM;
4645
4646 mutex_lock(&fs_info->balance_mutex);
4647 if (!fs_info->balance_ctl) {
4648 ret = -ENOTCONN;
4649 goto out;
4650 }
4651
8d2db785 4652 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
19a39dce
ID		 4653 if (!bargs) {
4654 ret = -ENOMEM;
4655 goto out;
4656 }
4657
008ef096 4658 btrfs_update_ioctl_balance_args(fs_info, bargs);
19a39dce
ID		 4659
4660 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4661 ret = -EFAULT;
4662
4663 kfree(bargs);
4664out:
4665 mutex_unlock(&fs_info->balance_mutex);
4666 return ret;
4667}
4668
905b0dda 4669static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4670{
0b246afa
JM		 4671 struct inode *inode = file_inode(file);
4672 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5d13a37b
AJ		 4673 struct btrfs_ioctl_quota_ctl_args *sa;
4674 struct btrfs_trans_handle *trans = NULL;
4675 int ret;
4676 int err;
4677
4678 if (!capable(CAP_SYS_ADMIN))
4679 return -EPERM;
4680
905b0dda
MX		 4681 ret = mnt_want_write_file(file);
4682 if (ret)
4683 return ret;
5d13a37b
AJ		 4684
4685 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4686 if (IS_ERR(sa)) {
4687 ret = PTR_ERR(sa);
4688 goto drop_write;
4689 }
5d13a37b 4690
0b246afa
JM		 4691 down_write(&fs_info->subvol_sem);
4692 trans = btrfs_start_transaction(fs_info->tree_root, 2);
2f232036
JS		 4693 if (IS_ERR(trans)) {
4694 ret = PTR_ERR(trans);
4695 goto out;
5d13a37b
AJ		 4696 }
4697
4698 switch (sa->cmd) {
4699 case BTRFS_QUOTA_CTL_ENABLE:
0b246afa 4700 ret = btrfs_quota_enable(trans, fs_info);
5d13a37b
AJ		 4701 break;
4702 case BTRFS_QUOTA_CTL_DISABLE:
0b246afa 4703 ret = btrfs_quota_disable(trans, fs_info);
5d13a37b 4704 break;
5d13a37b
AJ		 4705 default:
4706 ret = -EINVAL;
4707 break;
4708 }
4709
3a45bb20 4710 err = btrfs_commit_transaction(trans);
2f232036
JS		 4711 if (err && !ret)
4712 ret = err;
5d13a37b
AJ		 4713out:
4714 kfree(sa);
0b246afa 4715 up_write(&fs_info->subvol_sem);
905b0dda
MX		 4716drop_write:
4717 mnt_drop_write_file(file);
5d13a37b
AJ		 4718 return ret;
4719}
4720
905b0dda 4721static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4722{
0b246afa
JM		 4723 struct inode *inode = file_inode(file);
4724 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4725 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4726 struct btrfs_ioctl_qgroup_assign_args *sa;
4727 struct btrfs_trans_handle *trans;
4728 int ret;
4729 int err;
4730
4731 if (!capable(CAP_SYS_ADMIN))
4732 return -EPERM;
4733
905b0dda
MX		 4734 ret = mnt_want_write_file(file);
4735 if (ret)
4736 return ret;
5d13a37b
AJ		 4737
4738 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4739 if (IS_ERR(sa)) {
4740 ret = PTR_ERR(sa);
4741 goto drop_write;
4742 }
5d13a37b
AJ		 4743
4744 trans = btrfs_join_transaction(root);
4745 if (IS_ERR(trans)) {
4746 ret = PTR_ERR(trans);
4747 goto out;
4748 }
4749
5d13a37b 4750 if (sa->assign) {
0b246afa 4751 ret = btrfs_add_qgroup_relation(trans, fs_info,
5d13a37b
AJ		 4752 sa->src, sa->dst);
4753 } else {
0b246afa 4754 ret = btrfs_del_qgroup_relation(trans, fs_info,
5d13a37b
AJ		 4755 sa->src, sa->dst);
4756 }
4757
e082f563 4758 /* update qgroup status and info */
0b246afa 4759 err = btrfs_run_qgroups(trans, fs_info);
e082f563 4760 if (err < 0)
0b246afa
JM		 4761 btrfs_handle_fs_error(fs_info, err,
4762 "failed to update qgroup status and info");
3a45bb20 4763 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 4764 if (err && !ret)
4765 ret = err;
4766
4767out:
4768 kfree(sa);
905b0dda
MX		 4769drop_write:
4770 mnt_drop_write_file(file);
5d13a37b
AJ		 4771 return ret;
4772}
4773
905b0dda 4774static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4775{
0b246afa
JM		 4776 struct inode *inode = file_inode(file);
4777 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4778 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4779 struct btrfs_ioctl_qgroup_create_args *sa;
4780 struct btrfs_trans_handle *trans;
4781 int ret;
4782 int err;
4783
4784 if (!capable(CAP_SYS_ADMIN))
4785 return -EPERM;
4786
905b0dda
MX		 4787 ret = mnt_want_write_file(file);
4788 if (ret)
4789 return ret;
5d13a37b
AJ		 4790
4791 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4792 if (IS_ERR(sa)) {
4793 ret = PTR_ERR(sa);
4794 goto drop_write;
4795 }
5d13a37b 4796
d86e56cf
MX		 4797 if (!sa->qgroupid) {
4798 ret = -EINVAL;
4799 goto out;
4800 }
4801
5d13a37b
AJ		 4802 trans = btrfs_join_transaction(root);
4803 if (IS_ERR(trans)) {
4804 ret = PTR_ERR(trans);
4805 goto out;
4806 }
4807
5d13a37b 4808 if (sa->create) {
0b246afa 4809 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
5d13a37b 4810 } else {
0b246afa 4811 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
5d13a37b
AJ		 4812 }
4813
3a45bb20 4814 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 4815 if (err && !ret)
4816 ret = err;
4817
4818out:
4819 kfree(sa);
905b0dda
MX		 4820drop_write:
4821 mnt_drop_write_file(file);
5d13a37b
AJ		 4822 return ret;
4823}
4824
905b0dda 4825static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4826{
0b246afa
JM		 4827 struct inode *inode = file_inode(file);
4828 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4829 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4830 struct btrfs_ioctl_qgroup_limit_args *sa;
4831 struct btrfs_trans_handle *trans;
4832 int ret;
4833 int err;
4834 u64 qgroupid;
4835
4836 if (!capable(CAP_SYS_ADMIN))
4837 return -EPERM;
4838
905b0dda
MX		 4839 ret = mnt_want_write_file(file);
4840 if (ret)
4841 return ret;
5d13a37b
AJ		 4842
4843 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4844 if (IS_ERR(sa)) {
4845 ret = PTR_ERR(sa);
4846 goto drop_write;
4847 }
5d13a37b
AJ		 4848
4849 trans = btrfs_join_transaction(root);
4850 if (IS_ERR(trans)) {
4851 ret = PTR_ERR(trans);
4852 goto out;
4853 }
4854
4855 qgroupid = sa->qgroupid;
4856 if (!qgroupid) {
4857 /* take the current subvol as qgroup */
4858 qgroupid = root->root_key.objectid;
4859 }
4860
0b246afa 4861 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
5d13a37b 4862
3a45bb20 4863 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 4864 if (err && !ret)
4865 ret = err;
4866
4867out:
4868 kfree(sa);
905b0dda
MX		 4869drop_write:
4870 mnt_drop_write_file(file);
5d13a37b
AJ		 4871 return ret;
4872}
4873
2f232036
JS		 4874static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4875{
0b246afa
JM		 4876 struct inode *inode = file_inode(file);
4877 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2f232036
JS		 4878 struct btrfs_ioctl_quota_rescan_args *qsa;
4879 int ret;
4880
4881 if (!capable(CAP_SYS_ADMIN))
4882 return -EPERM;
4883
4884 ret = mnt_want_write_file(file);
4885 if (ret)
4886 return ret;
4887
4888 qsa = memdup_user(arg, sizeof(*qsa));
4889 if (IS_ERR(qsa)) {
4890 ret = PTR_ERR(qsa);
4891 goto drop_write;
4892 }
4893
4894 if (qsa->flags) {
4895 ret = -EINVAL;
4896 goto out;
4897 }
4898
0b246afa 4899 ret = btrfs_qgroup_rescan(fs_info);
2f232036
JS		 4900
4901out:
4902 kfree(qsa);
4903drop_write:
4904 mnt_drop_write_file(file);
4905 return ret;
4906}
4907
4908static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4909{
0b246afa
JM		 4910 struct inode *inode = file_inode(file);
4911 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2f232036
JS		 4912 struct btrfs_ioctl_quota_rescan_args *qsa;
4913 int ret = 0;
4914
4915 if (!capable(CAP_SYS_ADMIN))
4916 return -EPERM;
4917
8d2db785 4918 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
2f232036
JS		 4919 if (!qsa)
4920 return -ENOMEM;
4921
0b246afa 4922 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
2f232036 4923 qsa->flags = 1;
0b246afa 4924 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
2f232036
JS		 4925 }
4926
4927 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4928 ret = -EFAULT;
4929
4930 kfree(qsa);
4931 return ret;
4932}
4933
57254b6e
JS		 4934static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4935{
0b246afa
JM		 4936 struct inode *inode = file_inode(file);
4937 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
57254b6e
JS		 4938
4939 if (!capable(CAP_SYS_ADMIN))
4940 return -EPERM;
4941
0b246afa 4942 return btrfs_qgroup_wait_for_completion(fs_info, true);
57254b6e
JS		 4943}
4944
abccd00f
HM		 4945static long _btrfs_ioctl_set_received_subvol(struct file *file,
4946 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4947{
496ad9aa 4948 struct inode *inode = file_inode(file);
0b246afa 4949 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
8ea05e3a
AB		 4950 struct btrfs_root *root = BTRFS_I(inode)->root;
4951 struct btrfs_root_item *root_item = &root->root_item;
4952 struct btrfs_trans_handle *trans;
c2050a45 4953 struct timespec ct = current_time(inode);
8ea05e3a 4954 int ret = 0;
dd5f9615 4955 int received_uuid_changed;
8ea05e3a 4956
bd60ea0f
DS		 4957 if (!inode_owner_or_capable(inode))
4958 return -EPERM;
4959
8ea05e3a
AB		 4960 ret = mnt_want_write_file(file);
4961 if (ret < 0)
4962 return ret;
4963
0b246afa 4964 down_write(&fs_info->subvol_sem);
8ea05e3a 4965
4a0cc7ca 4966 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
8ea05e3a
AB		 4967 ret = -EINVAL;
4968 goto out;
4969 }
4970
4971 if (btrfs_root_readonly(root)) {
4972 ret = -EROFS;
4973 goto out;
4974 }
4975
dd5f9615
SB		 4976 /*
4977 * 1 - root item
4978 * 2 - uuid items (received uuid + subvol uuid)
4979 */
4980 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4981 if (IS_ERR(trans)) {
4982 ret = PTR_ERR(trans);
4983 trans = NULL;
4984 goto out;
4985 }
4986
4987 sa->rtransid = trans->transid;
4988 sa->rtime.sec = ct.tv_sec;
4989 sa->rtime.nsec = ct.tv_nsec;
4990
dd5f9615
SB		 4991 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4992 BTRFS_UUID_SIZE);
4993 if (received_uuid_changed &&
d87ff758
NB		 4994 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4995 ret = btrfs_uuid_tree_rem(trans, fs_info,
4996 root_item->received_uuid,
4997 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4998 root->root_key.objectid);
4999 if (ret && ret != -ENOENT) {
5000 btrfs_abort_transaction(trans, ret);
5001 btrfs_end_transaction(trans);
5002 goto out;
5003 }
5004 }
8ea05e3a
AB		 5005 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5006 btrfs_set_root_stransid(root_item, sa->stransid);
5007 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 5008 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5009 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5010 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5011 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a 5012
0b246afa 5013 ret = btrfs_update_root(trans, fs_info->tree_root,
8ea05e3a
AB		 5014 &root->root_key, &root->root_item);
5015 if (ret < 0) {
3a45bb20 5016 btrfs_end_transaction(trans);
8ea05e3a 5017 goto out;
dd5f9615
SB		 5018 }
5019 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
0b246afa 5020 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
dd5f9615
SB		 5021 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5022 root->root_key.objectid);
5023 if (ret < 0 && ret != -EEXIST) {
66642832 5024 btrfs_abort_transaction(trans, ret);
efd38150 5025 btrfs_end_transaction(trans);
8ea05e3a 5026 goto out;
dd5f9615
SB		 5027 }
5028 }
3a45bb20 5029 ret = btrfs_commit_transaction(trans);
abccd00f 5030out:
0b246afa 5031 up_write(&fs_info->subvol_sem);
abccd00f
HM		 5032 mnt_drop_write_file(file);
5033 return ret;
5034}
5035
5036#ifdef CONFIG_64BIT
5037static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5038 void __user *arg)
5039{
5040 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5041 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5042 int ret = 0;
5043
5044 args32 = memdup_user(arg, sizeof(*args32));
7b9ea627
SV		 5045 if (IS_ERR(args32))
5046 return PTR_ERR(args32);
abccd00f 5047
8d2db785 5048 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
84dbeb87
DC		 5049 if (!args64) {
5050 ret = -ENOMEM;
abccd00f
HM		 5051 goto out;
5052 }
5053
5054 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5055 args64->stransid = args32->stransid;
5056 args64->rtransid = args32->rtransid;
5057 args64->stime.sec = args32->stime.sec;
5058 args64->stime.nsec = args32->stime.nsec;
5059 args64->rtime.sec = args32->rtime.sec;
5060 args64->rtime.nsec = args32->rtime.nsec;
5061 args64->flags = args32->flags;
5062
5063 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5064 if (ret)
5065 goto out;
5066
5067 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5068 args32->stransid = args64->stransid;
5069 args32->rtransid = args64->rtransid;
5070 args32->stime.sec = args64->stime.sec;
5071 args32->stime.nsec = args64->stime.nsec;
5072 args32->rtime.sec = args64->rtime.sec;
5073 args32->rtime.nsec = args64->rtime.nsec;
5074 args32->flags = args64->flags;
5075
5076 ret = copy_to_user(arg, args32, sizeof(*args32));
5077 if (ret)
5078 ret = -EFAULT;
5079
5080out:
5081 kfree(args32);
5082 kfree(args64);
5083 return ret;
5084}
5085#endif
5086
5087static long btrfs_ioctl_set_received_subvol(struct file *file,
5088 void __user *arg)
5089{
5090 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5091 int ret = 0;
5092
5093 sa = memdup_user(arg, sizeof(*sa));
7b9ea627
SV		 5094 if (IS_ERR(sa))
5095 return PTR_ERR(sa);
abccd00f
HM		 5096
5097 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5098
5099 if (ret)
5100 goto out;
5101
8ea05e3a
AB		 5102 ret = copy_to_user(arg, sa, sizeof(*sa));
5103 if (ret)
5104 ret = -EFAULT;
5105
5106out:
5107 kfree(sa);
8ea05e3a
AB		 5108 return ret;
5109}
5110
867ab667 5111static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5112{
0b246afa
JM		 5113 struct inode *inode = file_inode(file);
5114 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
a1b83ac5 5115 size_t len;
867ab667 5116 int ret;
a1b83ac5
AJ		 5117 char label[BTRFS_LABEL_SIZE];
5118
0b246afa
JM		 5119 spin_lock(&fs_info->super_lock);
5120 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5121 spin_unlock(&fs_info->super_lock);
a1b83ac5
AJ		 5122
5123 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5124
5125 if (len == BTRFS_LABEL_SIZE) {
0b246afa
JM		 5126 btrfs_warn(fs_info,
5127 "label is too long, return the first %zu bytes",
5128 --len);
867ab667 5129 }
5130
867ab667 5131 ret = copy_to_user(arg, label, len);
867ab667 5132
5133 return ret ? -EFAULT : 0;
5134}
5135
a8bfd4ab 5136static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5137{
0b246afa
JM		 5138 struct inode *inode = file_inode(file);
5139 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5140 struct btrfs_root *root = BTRFS_I(inode)->root;
5141 struct btrfs_super_block *super_block = fs_info->super_copy;
a8bfd4ab 5142 struct btrfs_trans_handle *trans;
5143 char label[BTRFS_LABEL_SIZE];
5144 int ret;
5145
5146 if (!capable(CAP_SYS_ADMIN))
5147 return -EPERM;
5148
5149 if (copy_from_user(label, arg, sizeof(label)))
5150 return -EFAULT;
5151
5152 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
0b246afa 5153 btrfs_err(fs_info,
5d163e0e
JM		 5154 "unable to set label with more than %d bytes",
5155 BTRFS_LABEL_SIZE - 1);
a8bfd4ab 5156 return -EINVAL;
5157 }
5158
5159 ret = mnt_want_write_file(file);
5160 if (ret)
5161 return ret;
5162
a8bfd4ab 5163 trans = btrfs_start_transaction(root, 0);
5164 if (IS_ERR(trans)) {
5165 ret = PTR_ERR(trans);
5166 goto out_unlock;
5167 }
5168
0b246afa 5169 spin_lock(&fs_info->super_lock);
a8bfd4ab 5170 strcpy(super_block->label, label);
0b246afa 5171 spin_unlock(&fs_info->super_lock);
3a45bb20 5172 ret = btrfs_commit_transaction(trans);
a8bfd4ab 5173
5174out_unlock:
a8bfd4ab 5175 mnt_drop_write_file(file);
5176 return ret;
5177}
5178
2eaa055f
JM		 5179#define INIT_FEATURE_FLAGS(suffix) \
5180 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5181 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5182 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5183
d5131b65 5184int btrfs_ioctl_get_supported_features(void __user *arg)
2eaa055f 5185{
4d4ab6d6 5186 static const struct btrfs_ioctl_feature_flags features[3] = {
2eaa055f
JM		 5187 INIT_FEATURE_FLAGS(SUPP),
5188 INIT_FEATURE_FLAGS(SAFE_SET),
5189 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5190 };
5191
5192 if (copy_to_user(arg, &features, sizeof(features)))
5193 return -EFAULT;
5194
5195 return 0;
5196}
5197
5198static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5199{
0b246afa
JM		 5200 struct inode *inode = file_inode(file);
5201 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5202 struct btrfs_super_block *super_block = fs_info->super_copy;
2eaa055f
JM		 5203 struct btrfs_ioctl_feature_flags features;
5204
5205 features.compat_flags = btrfs_super_compat_flags(super_block);
5206 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5207 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5208
5209 if (copy_to_user(arg, &features, sizeof(features)))
5210 return -EFAULT;
5211
5212 return 0;
5213}
5214
2ff7e61e 5215static int check_feature_bits(struct btrfs_fs_info *fs_info,
3b02a68a 5216 enum btrfs_feature_set set,
2eaa055f
JM		 5217 u64 change_mask, u64 flags, u64 supported_flags,
5218 u64 safe_set, u64 safe_clear)
5219{
3b02a68a
JM		 5220 const char *type = btrfs_feature_set_names[set];
5221 char *names;
2eaa055f
JM		 5222 u64 disallowed, unsupported;
5223 u64 set_mask = flags & change_mask;
5224 u64 clear_mask = ~flags & change_mask;
5225
5226 unsupported = set_mask & ~supported_flags;
5227 if (unsupported) {
3b02a68a
JM		 5228 names = btrfs_printable_features(set, unsupported);
5229 if (names) {
0b246afa
JM		 5230 btrfs_warn(fs_info,
5231 "this kernel does not support the %s feature bit%s",
5232 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5233 kfree(names);
5234 } else
0b246afa
JM		 5235 btrfs_warn(fs_info,
5236 "this kernel does not support %s bits 0x%llx",
5237 type, unsupported);
2eaa055f
JM		 5238 return -EOPNOTSUPP;
5239 }
5240
5241 disallowed = set_mask & ~safe_set;
5242 if (disallowed) {
3b02a68a
JM		 5243 names = btrfs_printable_features(set, disallowed);
5244 if (names) {
0b246afa
JM		 5245 btrfs_warn(fs_info,
5246 "can't set the %s feature bit%s while mounted",
5247 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5248 kfree(names);
5249 } else
0b246afa
JM		 5250 btrfs_warn(fs_info,
5251 "can't set %s bits 0x%llx while mounted",
5252 type, disallowed);
2eaa055f
JM		 5253 return -EPERM;
5254 }
5255
5256 disallowed = clear_mask & ~safe_clear;
5257 if (disallowed) {
3b02a68a
JM		 5258 names = btrfs_printable_features(set, disallowed);
5259 if (names) {
0b246afa
JM		 5260 btrfs_warn(fs_info,
5261 "can't clear the %s feature bit%s while mounted",
5262 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5263 kfree(names);
5264 } else
0b246afa
JM		 5265 btrfs_warn(fs_info,
5266 "can't clear %s bits 0x%llx while mounted",
5267 type, disallowed);
2eaa055f
JM		 5268 return -EPERM;
5269 }
5270
5271 return 0;
5272}
5273
2ff7e61e
JM		 5274#define check_feature(fs_info, change_mask, flags, mask_base) \
5275check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5276 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5277 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5278 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5279
5280static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5281{
0b246afa
JM		 5282 struct inode *inode = file_inode(file);
5283 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5284 struct btrfs_root *root = BTRFS_I(inode)->root;
5285 struct btrfs_super_block *super_block = fs_info->super_copy;
2eaa055f
JM		 5286 struct btrfs_ioctl_feature_flags flags[2];
5287 struct btrfs_trans_handle *trans;
5288 u64 newflags;
5289 int ret;
5290
5291 if (!capable(CAP_SYS_ADMIN))
5292 return -EPERM;
5293
5294 if (copy_from_user(flags, arg, sizeof(flags)))
5295 return -EFAULT;
5296
5297 /* Nothing to do */
5298 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5299 !flags[0].incompat_flags)
5300 return 0;
5301
2ff7e61e 5302 ret = check_feature(fs_info, flags[0].compat_flags,
2eaa055f
JM		 5303 flags[1].compat_flags, COMPAT);
5304 if (ret)
5305 return ret;
5306
2ff7e61e 5307 ret = check_feature(fs_info, flags[0].compat_ro_flags,
2eaa055f
JM		 5308 flags[1].compat_ro_flags, COMPAT_RO);
5309 if (ret)
5310 return ret;
5311
2ff7e61e 5312 ret = check_feature(fs_info, flags[0].incompat_flags,
2eaa055f
JM		 5313 flags[1].incompat_flags, INCOMPAT);
5314 if (ret)
5315 return ret;
5316
7ab19625
DS		 5317 ret = mnt_want_write_file(file);
5318 if (ret)
5319 return ret;
5320
8051aa1a 5321 trans = btrfs_start_transaction(root, 0);
7ab19625
DS		 5322 if (IS_ERR(trans)) {
5323 ret = PTR_ERR(trans);
5324 goto out_drop_write;
5325 }
2eaa055f 5326
0b246afa 5327 spin_lock(&fs_info->super_lock);
2eaa055f
JM		 5328 newflags = btrfs_super_compat_flags(super_block);
5329 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5330 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5331 btrfs_set_super_compat_flags(super_block, newflags);
5332
5333 newflags = btrfs_super_compat_ro_flags(super_block);
5334 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5335 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5336 btrfs_set_super_compat_ro_flags(super_block, newflags);
5337
5338 newflags = btrfs_super_incompat_flags(super_block);
5339 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5340 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5341 btrfs_set_super_incompat_flags(super_block, newflags);
0b246afa 5342 spin_unlock(&fs_info->super_lock);
2eaa055f 5343
3a45bb20 5344 ret = btrfs_commit_transaction(trans);
7ab19625
DS		 5345out_drop_write:
5346 mnt_drop_write_file(file);
5347
5348 return ret;
2eaa055f
JM		 5349}
5350
2351f431
JB		 5351static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5352{
5353 struct btrfs_ioctl_send_args *arg;
5354 int ret;
5355
5356 if (compat) {
5357#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5358 struct btrfs_ioctl_send_args_32 args32;
5359
5360 ret = copy_from_user(&args32, argp, sizeof(args32));
5361 if (ret)
5362 return -EFAULT;
5363 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5364 if (!arg)
5365 return -ENOMEM;
5366 arg->send_fd = args32.send_fd;
5367 arg->clone_sources_count = args32.clone_sources_count;
5368 arg->clone_sources = compat_ptr(args32.clone_sources);
5369 arg->parent_root = args32.parent_root;
5370 arg->flags = args32.flags;
5371 memcpy(arg->reserved, args32.reserved,
5372 sizeof(args32.reserved));
5373#else
5374 return -ENOTTY;
5375#endif
5376 } else {
5377 arg = memdup_user(argp, sizeof(*arg));
5378 if (IS_ERR(arg))
5379 return PTR_ERR(arg);
5380 }
5381 ret = btrfs_ioctl_send(file, arg);
5382 kfree(arg);
5383 return ret;
5384}
5385
f46b5a66
CH		 5386long btrfs_ioctl(struct file *file, unsigned int
5387 cmd, unsigned long arg)
5388{
0b246afa
JM		 5389 struct inode *inode = file_inode(file);
5390 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5391 struct btrfs_root *root = BTRFS_I(inode)->root;
4bcabaa3 5392 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5393
5394 switch (cmd) {
6cbff00f
CH		 5395 case FS_IOC_GETFLAGS:
5396 return btrfs_ioctl_getflags(file, argp);
5397 case FS_IOC_SETFLAGS:
5398 return btrfs_ioctl_setflags(file, argp);
5399 case FS_IOC_GETVERSION:
5400 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5401 case FITRIM:
5402 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5403 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5404 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5405 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5406 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5407 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5408 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5409 case BTRFS_IOC_SUBVOL_CREATE_V2:
5410 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5411 case BTRFS_IOC_SNAP_DESTROY:
5412 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5413 case BTRFS_IOC_SUBVOL_GETFLAGS:
5414 return btrfs_ioctl_subvol_getflags(file, argp);
5415 case BTRFS_IOC_SUBVOL_SETFLAGS:
5416 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5417 case BTRFS_IOC_DEFAULT_SUBVOL:
5418 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5419 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5420 return btrfs_ioctl_defrag(file, NULL);
5421 case BTRFS_IOC_DEFRAG_RANGE:
5422 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5423 case BTRFS_IOC_RESIZE:
198605a8 5424 return btrfs_ioctl_resize(file, argp);
f46b5a66 5425 case BTRFS_IOC_ADD_DEV:
2ff7e61e 5426 return btrfs_ioctl_add_dev(fs_info, argp);
f46b5a66 5427 case BTRFS_IOC_RM_DEV:
da24927b 5428 return btrfs_ioctl_rm_dev(file, argp);
6b526ed7
AJ		 5429 case BTRFS_IOC_RM_DEV_V2:
5430 return btrfs_ioctl_rm_dev_v2(file, argp);
475f6387 5431 case BTRFS_IOC_FS_INFO:
2ff7e61e 5432 return btrfs_ioctl_fs_info(fs_info, argp);
475f6387 5433 case BTRFS_IOC_DEV_INFO:
2ff7e61e 5434 return btrfs_ioctl_dev_info(fs_info, argp);
f46b5a66 5435 case BTRFS_IOC_BALANCE:
9ba1f6e4 5436 return btrfs_ioctl_balance(file, NULL);
ac8e9819
CM		 5437 case BTRFS_IOC_TREE_SEARCH:
5438 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5439 case BTRFS_IOC_TREE_SEARCH_V2:
5440 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5441 case BTRFS_IOC_INO_LOOKUP:
5442 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5443 case BTRFS_IOC_INO_PATHS:
5444 return btrfs_ioctl_ino_to_path(root, argp);
5445 case BTRFS_IOC_LOGICAL_INO:
d24a67b2
ZB		 5446 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5447 case BTRFS_IOC_LOGICAL_INO_V2:
5448 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
1406e432 5449 case BTRFS_IOC_SPACE_INFO:
2ff7e61e 5450 return btrfs_ioctl_space_info(fs_info, argp);
9b199859
FDBM		 5451 case BTRFS_IOC_SYNC: {
5452 int ret;
5453
82b3e53b 5454 ret = btrfs_start_delalloc_roots(fs_info, -1);
9b199859
FDBM		 5455 if (ret)
5456 return ret;
0b246afa 5457 ret = btrfs_sync_fs(inode->i_sb, 1);
2fad4e83
DS		 5458 /*
5459 * The transaction thread may want to do more work,
01327610 5460 * namely it pokes the cleaner kthread that will start
2fad4e83
DS		 5461 * processing uncleaned subvols.
5462 */
0b246afa 5463 wake_up_process(fs_info->transaction_kthread);
9b199859
FDBM		 5464 return ret;
5465 }
46204592 5466 case BTRFS_IOC_START_SYNC:
9a8c28be 5467 return btrfs_ioctl_start_sync(root, argp);
46204592 5468 case BTRFS_IOC_WAIT_SYNC:
2ff7e61e 5469 return btrfs_ioctl_wait_sync(fs_info, argp);
475f6387 5470 case BTRFS_IOC_SCRUB:
b8e95489 5471 return btrfs_ioctl_scrub(file, argp);
475f6387 5472 case BTRFS_IOC_SCRUB_CANCEL:
2ff7e61e 5473 return btrfs_ioctl_scrub_cancel(fs_info);
475f6387 5474 case BTRFS_IOC_SCRUB_PROGRESS:
2ff7e61e 5475 return btrfs_ioctl_scrub_progress(fs_info, argp);
c9e9f97b 5476 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5477 return btrfs_ioctl_balance(file, argp);
837d5b6e 5478 case BTRFS_IOC_BALANCE_CTL:
2ff7e61e 5479 return btrfs_ioctl_balance_ctl(fs_info, arg);
19a39dce 5480 case BTRFS_IOC_BALANCE_PROGRESS:
2ff7e61e 5481 return btrfs_ioctl_balance_progress(fs_info, argp);
8ea05e3a
AB		 5482 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5483 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5484#ifdef CONFIG_64BIT
5485 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5486 return btrfs_ioctl_set_received_subvol_32(file, argp);
5487#endif
31db9f7c 5488 case BTRFS_IOC_SEND:
2351f431
JB		 5489 return _btrfs_ioctl_send(file, argp, false);
5490#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5491 case BTRFS_IOC_SEND_32:
5492 return _btrfs_ioctl_send(file, argp, true);
5493#endif
c11d2c23 5494 case BTRFS_IOC_GET_DEV_STATS:
2ff7e61e 5495 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5d13a37b 5496 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5497 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5498 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5499 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5500 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5501 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5502 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5503 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5504 case BTRFS_IOC_QUOTA_RESCAN:
5505 return btrfs_ioctl_quota_rescan(file, argp);
5506 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5507 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5508 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5509 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd 5510 case BTRFS_IOC_DEV_REPLACE:
2ff7e61e 5511 return btrfs_ioctl_dev_replace(fs_info, argp);
867ab667 5512 case BTRFS_IOC_GET_FSLABEL:
5513 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5514 case BTRFS_IOC_SET_FSLABEL:
5515 return btrfs_ioctl_set_fslabel(file, argp);
2eaa055f 5516 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
d5131b65 5517 return btrfs_ioctl_get_supported_features(argp);
2eaa055f
JM		 5518 case BTRFS_IOC_GET_FEATURES:
5519 return btrfs_ioctl_get_features(file, argp);
5520 case BTRFS_IOC_SET_FEATURES:
5521 return btrfs_ioctl_set_features(file, argp);
e4202ac9
DS		 5522 case FS_IOC_FSGETXATTR:
5523 return btrfs_ioctl_fsgetxattr(file, argp);
025f2121
DS		 5524 case FS_IOC_FSSETXATTR:
5525 return btrfs_ioctl_fssetxattr(file, argp);
f46b5a66
CH		 5526 }
5527
5528 return -ENOTTY;
5529}
4c63c245
LD		 5530
5531#ifdef CONFIG_COMPAT
5532long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5533{
2a362249
JM		 5534 /*
5535 * These all access 32-bit values anyway so no further
5536 * handling is necessary.
5537 */
4c63c245
LD		 5538 switch (cmd) {
5539 case FS_IOC32_GETFLAGS:
5540 cmd = FS_IOC_GETFLAGS;
5541 break;
5542 case FS_IOC32_SETFLAGS:
5543 cmd = FS_IOC_SETFLAGS;
5544 break;
5545 case FS_IOC32_GETVERSION:
5546 cmd = FS_IOC_GETVERSION;
5547 break;
4c63c245
LD		 5548 }
5549
5550 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5551}
5552#endif
Linux 6.x block layer and io_uring tree(s)