]> git.ipfire.org Git - people/ms/linux.git/blame - fs/btrfs/ioctl.c
projects / people / ms / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: Enhance btrfs_trim_fs function to handle error better
[people/ms/linux.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>
f46b5a66
CH		 8#include <linux/file.h>
9#include <linux/fs.h>
cb8e7090 10#include <linux/fsnotify.h>
f46b5a66
CH		 11#include <linux/pagemap.h>
12#include <linux/highmem.h>
13#include <linux/time.h>
f46b5a66 14#include <linux/string.h>
f46b5a66 15#include <linux/backing-dev.h>
cb8e7090 16#include <linux/mount.h>
cb8e7090 17#include <linux/namei.h>
f46b5a66 18#include <linux/writeback.h>
f46b5a66 19#include <linux/compat.h>
cb8e7090 20#include <linux/security.h>
f46b5a66 21#include <linux/xattr.h>
f54de068 22#include <linux/mm.h>
5a0e3ad6 23#include <linux/slab.h>
f7039b1d 24#include <linux/blkdev.h>
8ea05e3a 25#include <linux/uuid.h>
55e301fd 26#include <linux/btrfs.h>
416161db 27#include <linux/uaccess.h>
ae5e165d 28#include <linux/iversion.h>
f46b5a66
CH		 29#include "ctree.h"
30#include "disk-io.h"
31#include "transaction.h"
32#include "btrfs_inode.h"
f46b5a66
CH		 33#include "print-tree.h"
34#include "volumes.h"
925baedd 35#include "locking.h"
581bb050 36#include "inode-map.h"
d7728c96 37#include "backref.h"
606686ee 38#include "rcu-string.h"
31db9f7c 39#include "send.h"
3f6bcfbd 40#include "dev-replace.h"
63541927 41#include "props.h"
3b02a68a 42#include "sysfs.h"
fcebe456 43#include "qgroup.h"
1ec9a1ae 44#include "tree-log.h"
ebb8765b 45#include "compression.h"
f46b5a66 46
abccd00f
HM		 47#ifdef CONFIG_64BIT
48/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
49 * structures are incorrect, as the timespec structure from userspace
50 * is 4 bytes too small. We define these alternatives here to teach
51 * the kernel about the 32-bit struct packing.
52 */
53struct btrfs_ioctl_timespec_32 {
54 __u64 sec;
55 __u32 nsec;
56} __attribute__ ((__packed__));
57
58struct btrfs_ioctl_received_subvol_args_32 {
59 char uuid[BTRFS_UUID_SIZE]; /* in */
60 __u64 stransid; /* in */
61 __u64 rtransid; /* out */
62 struct btrfs_ioctl_timespec_32 stime; /* in */
63 struct btrfs_ioctl_timespec_32 rtime; /* out */
64 __u64 flags; /* in */
65 __u64 reserved[16]; /* in */
66} __attribute__ ((__packed__));
67
68#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
69 struct btrfs_ioctl_received_subvol_args_32)
70#endif
71
2351f431
JB		 72#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
73struct btrfs_ioctl_send_args_32 {
74 __s64 send_fd; /* in */
75 __u64 clone_sources_count; /* in */
76 compat_uptr_t clone_sources; /* in */
77 __u64 parent_root; /* in */
78 __u64 flags; /* in */
79 __u64 reserved[4]; /* in */
80} __attribute__ ((__packed__));
81
82#define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
83 struct btrfs_ioctl_send_args_32)
84#endif
abccd00f 85
416161db 86static int btrfs_clone(struct inode *src, struct inode *inode,
1c919a5e
MF		 87 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
88 int no_time_update);
416161db 89
6cbff00f 90/* Mask out flags that are inappropriate for the given type of inode. */
1905a0f7
DS		 91static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
92 unsigned int flags)
6cbff00f 93{
1905a0f7 94 if (S_ISDIR(inode->i_mode))
6cbff00f 95 return flags;
1905a0f7 96 else if (S_ISREG(inode->i_mode))
6cbff00f
CH		 97 return flags & ~FS_DIRSYNC_FL;
98 else
99 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
100}
101
102/*
a157d4fd
DS		 103 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
104 * ioctl.
6cbff00f 105 */
a157d4fd 106static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags)
6cbff00f
CH		 107{
108 unsigned int iflags = 0;
109
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
124
13f48dc9 125 if (flags & BTRFS_INODE_NOCOMPRESS)
d0092bdd 126 iflags |= FS_NOCOMP_FL;
13f48dc9
ST		 127 else if (flags & BTRFS_INODE_COMPRESS)
128 iflags |= FS_COMPR_FL;
6cbff00f
CH		 129
130 return iflags;
131}
132
133/*
134 * Update inode->i_flags based on the btrfs internal flags.
135 */
7b6a221e 136void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
6cbff00f 137{
5c57b8b6 138 struct btrfs_inode *binode = BTRFS_I(inode);
3cc79392 139 unsigned int new_fl = 0;
6cbff00f 140
5c57b8b6 141 if (binode->flags & BTRFS_INODE_SYNC)
3cc79392 142 new_fl |= S_SYNC;
5c57b8b6 143 if (binode->flags & BTRFS_INODE_IMMUTABLE)
3cc79392 144 new_fl |= S_IMMUTABLE;
5c57b8b6 145 if (binode->flags & BTRFS_INODE_APPEND)
3cc79392 146 new_fl |= S_APPEND;
5c57b8b6 147 if (binode->flags & BTRFS_INODE_NOATIME)
3cc79392 148 new_fl |= S_NOATIME;
5c57b8b6 149 if (binode->flags & BTRFS_INODE_DIRSYNC)
3cc79392
FM		 150 new_fl |= S_DIRSYNC;
151
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
154 new_fl);
6cbff00f
CH		 155}
156
6cbff00f
CH		 157static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
158{
5c57b8b6
DS		 159 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
160 unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags);
6cbff00f
CH		 161
162 if (copy_to_user(arg, &flags, sizeof(flags)))
163 return -EFAULT;
164 return 0;
165}
166
5ba76abf
DS		 167/* Check if @flags are a supported and valid set of FS_*_FL flags */
168static int check_fsflags(unsigned int flags)
75e7cb7f
LB		 169{
170 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
171 FS_NOATIME_FL | FS_NODUMP_FL | \
172 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 173 FS_NOCOMP_FL | FS_COMPR_FL |
174 FS_NOCOW_FL))
75e7cb7f
LB		 175 return -EOPNOTSUPP;
176
177 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
178 return -EINVAL;
179
75e7cb7f
LB		 180 return 0;
181}
182
6cbff00f
CH		 183static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
184{
496ad9aa 185 struct inode *inode = file_inode(file);
0b246afa 186 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5c57b8b6
DS		 187 struct btrfs_inode *binode = BTRFS_I(inode);
188 struct btrfs_root *root = binode->root;
6cbff00f 189 struct btrfs_trans_handle *trans;
5c57b8b6 190 unsigned int fsflags, old_fsflags;
6cbff00f 191 int ret;
5c57b8b6
DS		 192 u64 old_flags;
193 unsigned int old_i_flags;
7e97b8da 194 umode_t mode;
6cbff00f 195
bd60ea0f
DS		 196 if (!inode_owner_or_capable(inode))
197 return -EPERM;
198
b83cc969
LZ		 199 if (btrfs_root_readonly(root))
200 return -EROFS;
201
5c57b8b6 202 if (copy_from_user(&fsflags, arg, sizeof(fsflags)))
6cbff00f
CH		 203 return -EFAULT;
204
5c57b8b6 205 ret = check_fsflags(fsflags);
75e7cb7f
LB		 206 if (ret)
207 return ret;
f46b5a66 208
e7848683
JK		 209 ret = mnt_want_write_file(file);
210 if (ret)
211 return ret;
212
5955102c 213 inode_lock(inode);
6cbff00f 214
5c57b8b6
DS		 215 old_flags = binode->flags;
216 old_i_flags = inode->i_flags;
7e97b8da 217 mode = inode->i_mode;
f062abf0 218
5c57b8b6
DS		 219 fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
220 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
221 if ((fsflags ^ old_fsflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
6cbff00f
CH		 222 if (!capable(CAP_LINUX_IMMUTABLE)) {
223 ret = -EPERM;
224 goto out_unlock;
225 }
226 }
227
5c57b8b6
DS		 228 if (fsflags & FS_SYNC_FL)
229 binode->flags |= BTRFS_INODE_SYNC;
6cbff00f 230 else
5c57b8b6
DS		 231 binode->flags &= ~BTRFS_INODE_SYNC;
232 if (fsflags & FS_IMMUTABLE_FL)
233 binode->flags |= BTRFS_INODE_IMMUTABLE;
6cbff00f 234 else
5c57b8b6
DS		 235 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
236 if (fsflags & FS_APPEND_FL)
237 binode->flags |= BTRFS_INODE_APPEND;
6cbff00f 238 else
5c57b8b6
DS		 239 binode->flags &= ~BTRFS_INODE_APPEND;
240 if (fsflags & FS_NODUMP_FL)
241 binode->flags |= BTRFS_INODE_NODUMP;
6cbff00f 242 else
5c57b8b6
DS		 243 binode->flags &= ~BTRFS_INODE_NODUMP;
244 if (fsflags & FS_NOATIME_FL)
245 binode->flags |= BTRFS_INODE_NOATIME;
6cbff00f 246 else
5c57b8b6
DS		 247 binode->flags &= ~BTRFS_INODE_NOATIME;
248 if (fsflags & FS_DIRSYNC_FL)
249 binode->flags |= BTRFS_INODE_DIRSYNC;
6cbff00f 250 else
5c57b8b6
DS		 251 binode->flags &= ~BTRFS_INODE_DIRSYNC;
252 if (fsflags & FS_NOCOW_FL) {
7e97b8da
DS		 253 if (S_ISREG(mode)) {
254 /*
255 * It's safe to turn csums off here, no extents exist.
256 * Otherwise we want the flag to reflect the real COW
257 * status of the file and will not set it.
258 */
259 if (inode->i_size == 0)
5c57b8b6
DS		 260 binode->flags |= BTRFS_INODE_NODATACOW
261 | BTRFS_INODE_NODATASUM;
7e97b8da 262 } else {
5c57b8b6 263 binode->flags |= BTRFS_INODE_NODATACOW;
7e97b8da
DS		 264 }
265 } else {
266 /*
01327610 267 * Revert back under same assumptions as above
7e97b8da
DS		 268 */
269 if (S_ISREG(mode)) {
270 if (inode->i_size == 0)
5c57b8b6 271 binode->flags &= ~(BTRFS_INODE_NODATACOW
7e97b8da
DS		 272 | BTRFS_INODE_NODATASUM);
273 } else {
5c57b8b6 274 binode->flags &= ~BTRFS_INODE_NODATACOW;
7e97b8da
DS		 275 }
276 }
6cbff00f 277
75e7cb7f
LB		 278 /*
279 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
280 * flag may be changed automatically if compression code won't make
281 * things smaller.
282 */
5c57b8b6
DS		 283 if (fsflags & FS_NOCOMP_FL) {
284 binode->flags &= ~BTRFS_INODE_COMPRESS;
285 binode->flags |= BTRFS_INODE_NOCOMPRESS;
63541927
FDBM		 286
287 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
288 if (ret && ret != -ENODATA)
289 goto out_drop;
5c57b8b6 290 } else if (fsflags & FS_COMPR_FL) {
63541927
FDBM		 291 const char *comp;
292
5c57b8b6
DS		 293 binode->flags |= BTRFS_INODE_COMPRESS;
294 binode->flags &= ~BTRFS_INODE_NOCOMPRESS;
63541927 295
93370509
DS		 296 comp = btrfs_compress_type2str(fs_info->compress_type);
297 if (!comp || comp[0] == 0)
298 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
299
63541927
FDBM		 300 ret = btrfs_set_prop(inode, "btrfs.compression",
301 comp, strlen(comp), 0);
302 if (ret)
303 goto out_drop;
304
ebcb904d 305 } else {
78a017a2
FM		 306 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
307 if (ret && ret != -ENODATA)
308 goto out_drop;
5c57b8b6 309 binode->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 310 }
6cbff00f 311
4da6f1a3 312 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 313 if (IS_ERR(trans)) {
314 ret = PTR_ERR(trans);
315 goto out_drop;
316 }
6cbff00f 317
7b6a221e 318 btrfs_sync_inode_flags_to_i_flags(inode);
0c4d2d95 319 inode_inc_iversion(inode);
c2050a45 320 inode->i_ctime = current_time(inode);
6cbff00f 321 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 322
3a45bb20 323 btrfs_end_transaction(trans);
f062abf0
LZ		 324 out_drop:
325 if (ret) {
5c57b8b6
DS		 326 binode->flags = old_flags;
327 inode->i_flags = old_i_flags;
f062abf0 328 }
6cbff00f 329
6cbff00f 330 out_unlock:
5955102c 331 inode_unlock(inode);
e7848683 332 mnt_drop_write_file(file);
2d4e6f6a 333 return ret;
6cbff00f
CH		 334}
335
19f93b3c
DS		 336/*
337 * Translate btrfs internal inode flags to xflags as expected by the
338 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
339 * silently dropped.
340 */
341static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
342{
343 unsigned int xflags = 0;
344
345 if (flags & BTRFS_INODE_APPEND)
346 xflags |= FS_XFLAG_APPEND;
347 if (flags & BTRFS_INODE_IMMUTABLE)
348 xflags |= FS_XFLAG_IMMUTABLE;
349 if (flags & BTRFS_INODE_NOATIME)
350 xflags |= FS_XFLAG_NOATIME;
351 if (flags & BTRFS_INODE_NODUMP)
352 xflags |= FS_XFLAG_NODUMP;
353 if (flags & BTRFS_INODE_SYNC)
354 xflags |= FS_XFLAG_SYNC;
355
356 return xflags;
357}
358
359/* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
360static int check_xflags(unsigned int flags)
361{
362 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
363 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
364 return -EOPNOTSUPP;
365 return 0;
366}
367
e4202ac9
DS		 368/*
369 * Set the xflags from the internal inode flags. The remaining items of fsxattr
370 * are zeroed.
371 */
372static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
373{
374 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
375 struct fsxattr fa;
376
377 memset(&fa, 0, sizeof(fa));
378 fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags);
379
380 if (copy_to_user(arg, &fa, sizeof(fa)))
381 return -EFAULT;
382
383 return 0;
384}
385
025f2121
DS		 386static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
387{
388 struct inode *inode = file_inode(file);
389 struct btrfs_inode *binode = BTRFS_I(inode);
390 struct btrfs_root *root = binode->root;
391 struct btrfs_trans_handle *trans;
392 struct fsxattr fa;
393 unsigned old_flags;
394 unsigned old_i_flags;
395 int ret = 0;
396
397 if (!inode_owner_or_capable(inode))
398 return -EPERM;
399
400 if (btrfs_root_readonly(root))
401 return -EROFS;
402
403 memset(&fa, 0, sizeof(fa));
404 if (copy_from_user(&fa, arg, sizeof(fa)))
405 return -EFAULT;
406
407 ret = check_xflags(fa.fsx_xflags);
408 if (ret)
409 return ret;
410
411 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
412 return -EOPNOTSUPP;
413
414 ret = mnt_want_write_file(file);
415 if (ret)
416 return ret;
417
418 inode_lock(inode);
419
420 old_flags = binode->flags;
421 old_i_flags = inode->i_flags;
422
423 /* We need the capabilities to change append-only or immutable inode */
424 if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) ||
425 (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) &&
426 !capable(CAP_LINUX_IMMUTABLE)) {
427 ret = -EPERM;
428 goto out_unlock;
429 }
430
431 if (fa.fsx_xflags & FS_XFLAG_SYNC)
432 binode->flags |= BTRFS_INODE_SYNC;
433 else
434 binode->flags &= ~BTRFS_INODE_SYNC;
435 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
436 binode->flags |= BTRFS_INODE_IMMUTABLE;
437 else
438 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
439 if (fa.fsx_xflags & FS_XFLAG_APPEND)
440 binode->flags |= BTRFS_INODE_APPEND;
441 else
442 binode->flags &= ~BTRFS_INODE_APPEND;
443 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
444 binode->flags |= BTRFS_INODE_NODUMP;
445 else
446 binode->flags &= ~BTRFS_INODE_NODUMP;
447 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
448 binode->flags |= BTRFS_INODE_NOATIME;
449 else
450 binode->flags &= ~BTRFS_INODE_NOATIME;
451
452 /* 1 item for the inode */
453 trans = btrfs_start_transaction(root, 1);
454 if (IS_ERR(trans)) {
455 ret = PTR_ERR(trans);
456 goto out_unlock;
457 }
458
459 btrfs_sync_inode_flags_to_i_flags(inode);
460 inode_inc_iversion(inode);
461 inode->i_ctime = current_time(inode);
462 ret = btrfs_update_inode(trans, root, inode);
463
464 btrfs_end_transaction(trans);
465
466out_unlock:
467 if (ret) {
468 binode->flags = old_flags;
469 inode->i_flags = old_i_flags;
470 }
471
472 inode_unlock(inode);
473 mnt_drop_write_file(file);
474
475 return ret;
476}
477
6cbff00f
CH		 478static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
479{
496ad9aa 480 struct inode *inode = file_inode(file);
6cbff00f
CH		 481
482 return put_user(inode->i_generation, arg);
483}
f46b5a66 484
f7039b1d
LD		 485static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
486{
0b246afa
JM		 487 struct inode *inode = file_inode(file);
488 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f7039b1d
LD		 489 struct btrfs_device *device;
490 struct request_queue *q;
491 struct fstrim_range range;
492 u64 minlen = ULLONG_MAX;
493 u64 num_devices = 0;
815745cf 494 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 495 int ret;
496
497 if (!capable(CAP_SYS_ADMIN))
498 return -EPERM;
499
1f78160c
XG		 500 rcu_read_lock();
501 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
502 dev_list) {
f7039b1d
LD		 503 if (!device->bdev)
504 continue;
505 q = bdev_get_queue(device->bdev);
506 if (blk_queue_discard(q)) {
507 num_devices++;
50d0446e 508 minlen = min_t(u64, q->limits.discard_granularity,
f7039b1d
LD		 509 minlen);
510 }
511 }
1f78160c 512 rcu_read_unlock();
f4c697e6 513
f7039b1d
LD		 514 if (!num_devices)
515 return -EOPNOTSUPP;
f7039b1d
LD		 516 if (copy_from_user(&range, arg, sizeof(range)))
517 return -EFAULT;
e515c18b
LC		 518 if (range.start > total_bytes ||
519 range.len < fs_info->sb->s_blocksize)
f4c697e6 520 return -EINVAL;
f7039b1d 521
f4c697e6 522 range.len = min(range.len, total_bytes - range.start);
f7039b1d 523 range.minlen = max(range.minlen, minlen);
2ff7e61e 524 ret = btrfs_trim_fs(fs_info, &range);
f7039b1d
LD		 525 if (ret < 0)
526 return ret;
527
528 if (copy_to_user(arg, &range, sizeof(range)))
529 return -EFAULT;
530
531 return 0;
532}
533
dd5f9615
SB		 534int btrfs_is_empty_uuid(u8 *uuid)
535{
46e0f66a
CM		 536 int i;
537
538 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
539 if (uuid[i])
540 return 0;
541 }
542 return 1;
dd5f9615
SB		 543}
544
d5c12070 545static noinline int create_subvol(struct inode *dir,
cb8e7090 546 struct dentry *dentry,
52f75f4f 547 const char *name, int namelen,
6f72c7e2 548 u64 *async_transid,
8696c533 549 struct btrfs_qgroup_inherit *inherit)
f46b5a66 550{
0b246afa 551 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
f46b5a66
CH		 552 struct btrfs_trans_handle *trans;
553 struct btrfs_key key;
49a3c4d9 554 struct btrfs_root_item *root_item;
f46b5a66
CH		 555 struct btrfs_inode_item *inode_item;
556 struct extent_buffer *leaf;
d5c12070 557 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 558 struct btrfs_root *new_root;
d5c12070 559 struct btrfs_block_rsv block_rsv;
95582b00 560 struct timespec64 cur_time = current_time(dir);
5662344b 561 struct inode *inode;
f46b5a66
CH		 562 int ret;
563 int err;
564 u64 objectid;
565 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 566 u64 index = 0;
8ea05e3a 567 uuid_le new_uuid;
f46b5a66 568
49a3c4d9
DS		 569 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
570 if (!root_item)
571 return -ENOMEM;
572
0b246afa 573 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
2fbe8c8a 574 if (ret)
49a3c4d9 575 goto fail_free;
6a912213 576
e09fe2d2
QW		 577 /*
578 * Don't create subvolume whose level is not zero. Or qgroup will be
01327610 579 * screwed up since it assumes subvolume qgroup's level to be 0.
e09fe2d2 580 */
49a3c4d9
DS		 581 if (btrfs_qgroup_level(objectid)) {
582 ret = -ENOSPC;
583 goto fail_free;
584 }
e09fe2d2 585
d5c12070 586 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 587 /*
d5c12070
MX		 588 * The same as the snapshot creation, please see the comment
589 * of create_snapshot().
9ed74f2d 590 */
c4c129db 591 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
d5c12070 592 if (ret)
49a3c4d9 593 goto fail_free;
d5c12070
MX		 594
595 trans = btrfs_start_transaction(root, 0);
596 if (IS_ERR(trans)) {
597 ret = PTR_ERR(trans);
7775c818 598 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
49a3c4d9 599 goto fail_free;
d5c12070
MX		 600 }
601 trans->block_rsv = &block_rsv;
602 trans->bytes_reserved = block_rsv.size;
f46b5a66 603
a9377422 604 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
6f72c7e2
AJ		 605 if (ret)
606 goto fail;
607
4d75f8a9 608 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 609 if (IS_ERR(leaf)) {
610 ret = PTR_ERR(leaf);
611 goto fail;
612 }
f46b5a66 613
f46b5a66
CH		 614 btrfs_mark_buffer_dirty(leaf);
615
49a3c4d9 616 inode_item = &root_item->inode;
3cae210f
QW		 617 btrfs_set_stack_inode_generation(inode_item, 1);
618 btrfs_set_stack_inode_size(inode_item, 3);
619 btrfs_set_stack_inode_nlink(inode_item, 1);
da17066c 620 btrfs_set_stack_inode_nbytes(inode_item,
0b246afa 621 fs_info->nodesize);
3cae210f 622 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 623
49a3c4d9
DS		 624 btrfs_set_root_flags(root_item, 0);
625 btrfs_set_root_limit(root_item, 0);
3cae210f 626 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 627
49a3c4d9
DS		 628 btrfs_set_root_bytenr(root_item, leaf->start);
629 btrfs_set_root_generation(root_item, trans->transid);
630 btrfs_set_root_level(root_item, 0);
631 btrfs_set_root_refs(root_item, 1);
632 btrfs_set_root_used(root_item, leaf->len);
633 btrfs_set_root_last_snapshot(root_item, 0);
f46b5a66 634
49a3c4d9
DS		 635 btrfs_set_root_generation_v2(root_item,
636 btrfs_root_generation(root_item));
8ea05e3a 637 uuid_le_gen(&new_uuid);
49a3c4d9
DS		 638 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
639 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
640 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
641 root_item->ctime = root_item->otime;
642 btrfs_set_root_ctransid(root_item, trans->transid);
643 btrfs_set_root_otransid(root_item, trans->transid);
f46b5a66 644
925baedd 645 btrfs_tree_unlock(leaf);
f46b5a66
CH		 646 free_extent_buffer(leaf);
647 leaf = NULL;
648
49a3c4d9 649 btrfs_set_root_dirid(root_item, new_dirid);
f46b5a66
CH		 650
651 key.objectid = objectid;
5d4f98a2 652 key.offset = 0;
962a298f 653 key.type = BTRFS_ROOT_ITEM_KEY;
0b246afa 654 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
49a3c4d9 655 root_item);
f46b5a66
CH		 656 if (ret)
657 goto fail;
658
76dda93c 659 key.offset = (u64)-1;
0b246afa 660 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
79787eaa 661 if (IS_ERR(new_root)) {
79787eaa 662 ret = PTR_ERR(new_root);
66642832 663 btrfs_abort_transaction(trans, ret);
79787eaa
JM		 664 goto fail;
665 }
76dda93c
YZ		 666
667 btrfs_record_root_in_trans(trans, new_root);
668
63541927 669 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 670 if (ret) {
671 /* We potentially lose an unused inode item here */
66642832 672 btrfs_abort_transaction(trans, ret);
ce598979
MF		 673 goto fail;
674 }
675
f32e48e9
CR		 676 mutex_lock(&new_root->objectid_mutex);
677 new_root->highest_objectid = new_dirid;
678 mutex_unlock(&new_root->objectid_mutex);
679
f46b5a66
CH		 680 /*
681 * insert the directory item
682 */
877574e2 683 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
79787eaa 684 if (ret) {
66642832 685 btrfs_abort_transaction(trans, ret);
79787eaa
JM		 686 goto fail;
687 }
3de4586c 688
684572df 689 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
3de4586c 690 BTRFS_FT_DIR, index);
79787eaa 691 if (ret) {
66642832 692 btrfs_abort_transaction(trans, ret);
f46b5a66 693 goto fail;
79787eaa 694 }
0660b5af 695
6ef06d27 696 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
52c26179
YZ		 697 ret = btrfs_update_inode(trans, root, dir);
698 BUG_ON(ret);
699
6025c19f 700 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
4a0cc7ca 701 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
76dda93c 702 BUG_ON(ret);
f46b5a66 703
cdb345a8 704 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
6bccf3ab 705 BTRFS_UUID_KEY_SUBVOL, objectid);
dd5f9615 706 if (ret)
66642832 707 btrfs_abort_transaction(trans, ret);
dd5f9615 708
f46b5a66 709fail:
49a3c4d9 710 kfree(root_item);
d5c12070
MX		 711 trans->block_rsv = NULL;
712 trans->bytes_reserved = 0;
7775c818 713 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
de6e8200 714
72fd032e
SW		 715 if (async_transid) {
716 *async_transid = trans->transid;
3a45bb20 717 err = btrfs_commit_transaction_async(trans, 1);
00d71c9c 718 if (err)
3a45bb20 719 err = btrfs_commit_transaction(trans);
72fd032e 720 } else {
3a45bb20 721 err = btrfs_commit_transaction(trans);
72fd032e 722 }
f46b5a66
CH		 723 if (err && !ret)
724 ret = err;
1a65e24b 725
5662344b
TI		 726 if (!ret) {
727 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 728 if (IS_ERR(inode))
729 return PTR_ERR(inode);
5662344b
TI		 730 d_instantiate(dentry, inode);
731 }
f46b5a66 732 return ret;
49a3c4d9
DS		 733
734fail_free:
735 kfree(root_item);
736 return ret;
f46b5a66
CH		 737}
738
e9662f70 739static int create_snapshot(struct btrfs_root *root, struct inode *dir,
61d7e4cb 740 struct dentry *dentry,
e9662f70
MX		 741 u64 *async_transid, bool readonly,
742 struct btrfs_qgroup_inherit *inherit)
f46b5a66 743{
0b246afa 744 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
2e4bfab9 745 struct inode *inode;
f46b5a66
CH		 746 struct btrfs_pending_snapshot *pending_snapshot;
747 struct btrfs_trans_handle *trans;
2e4bfab9 748 int ret;
8ecebf4d 749 bool snapshot_force_cow = false;
f46b5a66 750
27cdeb70 751 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
f46b5a66
CH		 752 return -EINVAL;
753
23269bf5 754 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
a1ee7362
DS		 755 if (!pending_snapshot)
756 return -ENOMEM;
757
b0c0ea63 758 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
23269bf5 759 GFP_KERNEL);
8546b570
DS		 760 pending_snapshot->path = btrfs_alloc_path();
761 if (!pending_snapshot->root_item || !pending_snapshot->path) {
b0c0ea63
DS		 762 ret = -ENOMEM;
763 goto free_pending;
764 }
765
8ecebf4d
RK		 766 /*
767 * Force new buffered writes to reserve space even when NOCOW is
768 * possible. This is to avoid later writeback (running dealloc) to
769 * fallback to COW mode and unexpectedly fail with ENOSPC.
770 */
ea14b57f 771 atomic_inc(&root->will_be_snapshotted);
4e857c58 772 smp_mb__after_atomic();
45bac0f3
LB		 773 /* wait for no snapshot writes */
774 wait_event(root->subv_writers->wait,
775 percpu_counter_sum(&root->subv_writers->counter) == 0);
8257b2dc 776
76f32e24 777 ret = btrfs_start_delalloc_inodes(root);
6a03843d 778 if (ret)
a1ee7362 779 goto dec_and_free;
6a03843d 780
8ecebf4d
RK		 781 /*
782 * All previous writes have started writeback in NOCOW mode, so now
783 * we force future writes to fallback to COW mode during snapshot
784 * creation.
785 */
786 atomic_inc(&root->snapshot_force_cow);
787 snapshot_force_cow = true;
788
6374e57a 789 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
6a03843d 790
66d8f3dd
MX		 791 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
792 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 793 /*
794 * 1 - parent dir inode
795 * 2 - dir entries
796 * 1 - root item
797 * 2 - root ref/backref
798 * 1 - root of snapshot
dd5f9615 799 * 1 - UUID item
d5c12070
MX		 800 */
801 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 802 &pending_snapshot->block_rsv, 8,
ee3441b4 803 false);
d5c12070 804 if (ret)
a1ee7362 805 goto dec_and_free;
d5c12070 806
3de4586c 807 pending_snapshot->dentry = dentry;
f46b5a66 808 pending_snapshot->root = root;
b83cc969 809 pending_snapshot->readonly = readonly;
e9662f70 810 pending_snapshot->dir = dir;
8696c533 811 pending_snapshot->inherit = inherit;
a22285a6 812
d5c12070 813 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 814 if (IS_ERR(trans)) {
815 ret = PTR_ERR(trans);
816 goto fail;
817 }
818
0b246afa 819 spin_lock(&fs_info->trans_lock);
f46b5a66
CH		 820 list_add(&pending_snapshot->list,
821 &trans->transaction->pending_snapshots);
0b246afa 822 spin_unlock(&fs_info->trans_lock);
72fd032e
SW		 823 if (async_transid) {
824 *async_transid = trans->transid;
3a45bb20 825 ret = btrfs_commit_transaction_async(trans, 1);
00d71c9c 826 if (ret)
3a45bb20 827 ret = btrfs_commit_transaction(trans);
72fd032e 828 } else {
3a45bb20 829 ret = btrfs_commit_transaction(trans);
72fd032e 830 }
aec8030a 831 if (ret)
c37b2b62 832 goto fail;
a22285a6
YZ		 833
834 ret = pending_snapshot->error;
835 if (ret)
836 goto fail;
837
d3797308
CM		 838 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
839 if (ret)
840 goto fail;
841
2b0143b5 842 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
2e4bfab9
YZ		 843 if (IS_ERR(inode)) {
844 ret = PTR_ERR(inode);
845 goto fail;
846 }
5662344b 847
2e4bfab9
YZ		 848 d_instantiate(dentry, inode);
849 ret = 0;
850fail:
7775c818 851 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
a1ee7362 852dec_and_free:
8ecebf4d
RK		 853 if (snapshot_force_cow)
854 atomic_dec(&root->snapshot_force_cow);
ea14b57f 855 if (atomic_dec_and_test(&root->will_be_snapshotted))
4625956a 856 wake_up_var(&root->will_be_snapshotted);
b0c0ea63
DS		 857free_pending:
858 kfree(pending_snapshot->root_item);
8546b570 859 btrfs_free_path(pending_snapshot->path);
a1ee7362
DS		 860 kfree(pending_snapshot);
861
f46b5a66
CH		 862 return ret;
863}
864
4260f7c7
SW		 865/* copy of may_delete in fs/namei.c()
866 * Check whether we can remove a link victim from directory dir, check
867 * whether the type of victim is right.
868 * 1. We can't do it if dir is read-only (done in permission())
869 * 2. We should have write and exec permissions on dir
870 * 3. We can't remove anything from append-only dir
871 * 4. We can't do anything with immutable dir (done in permission())
872 * 5. If the sticky bit on dir is set we should either
873 * a. be owner of dir, or
874 * b. be owner of victim, or
875 * c. have CAP_FOWNER capability
01327610 876 * 6. If the victim is append-only or immutable we can't do anything with
4260f7c7
SW		 877 * links pointing to it.
878 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
879 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
880 * 9. We can't remove a root or mountpoint.
881 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
882 * nfs_async_unlink().
883 */
884
67871254 885static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 886{
887 int error;
888
2b0143b5 889 if (d_really_is_negative(victim))
4260f7c7
SW		 890 return -ENOENT;
891
2b0143b5 892 BUG_ON(d_inode(victim->d_parent) != dir);
4fa6b5ec 893 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 894
895 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
896 if (error)
897 return error;
898 if (IS_APPEND(dir))
899 return -EPERM;
2b0143b5
DH		 900 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
901 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
4260f7c7
SW		 902 return -EPERM;
903 if (isdir) {
e36cb0b8 904 if (!d_is_dir(victim))
4260f7c7
SW		 905 return -ENOTDIR;
906 if (IS_ROOT(victim))
907 return -EBUSY;
e36cb0b8 908 } else if (d_is_dir(victim))
4260f7c7
SW		 909 return -EISDIR;
910 if (IS_DEADDIR(dir))
911 return -ENOENT;
912 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
913 return -EBUSY;
914 return 0;
915}
916
cb8e7090
CH		 917/* copy of may_create in fs/namei.c() */
918static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
919{
2b0143b5 920 if (d_really_is_positive(child))
cb8e7090
CH		 921 return -EEXIST;
922 if (IS_DEADDIR(dir))
923 return -ENOENT;
924 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
925}
926
927/*
928 * Create a new subvolume below @parent. This is largely modeled after
929 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
930 * inside this filesystem so it's quite a bit simpler.
931 */
92872094 932static noinline int btrfs_mksubvol(const struct path *parent,
52f75f4f 933 const char *name, int namelen,
72fd032e 934 struct btrfs_root *snap_src,
6f72c7e2 935 u64 *async_transid, bool readonly,
8696c533 936 struct btrfs_qgroup_inherit *inherit)
cb8e7090 937{
0b246afa
JM		 938 struct inode *dir = d_inode(parent->dentry);
939 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
cb8e7090
CH		 940 struct dentry *dentry;
941 int error;
942
00235411
AV		 943 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
944 if (error == -EINTR)
945 return error;
cb8e7090
CH		 946
947 dentry = lookup_one_len(name, parent->dentry, namelen);
948 error = PTR_ERR(dentry);
949 if (IS_ERR(dentry))
950 goto out_unlock;
951
76dda93c 952 error = btrfs_may_create(dir, dentry);
cb8e7090 953 if (error)
a874a63e 954 goto out_dput;
cb8e7090 955
9c52057c
CM		 956 /*
957 * even if this name doesn't exist, we may get hash collisions.
958 * check for them now when we can safely fail
959 */
960 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
961 dir->i_ino, name,
962 namelen);
963 if (error)
964 goto out_dput;
965
0b246afa 966 down_read(&fs_info->subvol_sem);
76dda93c
YZ		 967
968 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
969 goto out_up_read;
970
3de4586c 971 if (snap_src) {
61d7e4cb 972 error = create_snapshot(snap_src, dir, dentry,
6f72c7e2 973 async_transid, readonly, inherit);
3de4586c 974 } else {
d5c12070
MX		 975 error = create_subvol(dir, dentry, name, namelen,
976 async_transid, inherit);
3de4586c 977 }
76dda93c
YZ		 978 if (!error)
979 fsnotify_mkdir(dir, dentry);
980out_up_read:
0b246afa 981 up_read(&fs_info->subvol_sem);
cb8e7090
CH		 982out_dput:
983 dput(dentry);
984out_unlock:
5955102c 985 inode_unlock(dir);
cb8e7090
CH		 986 return error;
987}
988
4cb5300b
CM		 989/*
990 * When we're defragging a range, we don't want to kick it off again
991 * if it is really just waiting for delalloc to send it down.
992 * If we find a nice big extent or delalloc range for the bytes in the
993 * file you want to defrag, we return 0 to let you know to skip this
994 * part of the file
995 */
aab110ab 996static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
4cb5300b
CM		 997{
998 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
999 struct extent_map *em = NULL;
1000 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1001 u64 end;
1002
1003 read_lock(&em_tree->lock);
09cbfeaf 1004 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
4cb5300b
CM		 1005 read_unlock(&em_tree->lock);
1006
1007 if (em) {
1008 end = extent_map_end(em);
1009 free_extent_map(em);
1010 if (end - offset > thresh)
1011 return 0;
1012 }
1013 /* if we already have a nice delalloc here, just stop */
1014 thresh /= 2;
1015 end = count_range_bits(io_tree, &offset, offset + thresh,
1016 thresh, EXTENT_DELALLOC, 1);
1017 if (end >= thresh)
1018 return 0;
1019 return 1;
1020}
1021
1022/*
1023 * helper function to walk through a file and find extents
1024 * newer than a specific transid, and smaller than thresh.
1025 *
1026 * This is used by the defragging code to find new and small
1027 * extents
1028 */
1029static int find_new_extents(struct btrfs_root *root,
1030 struct inode *inode, u64 newer_than,
aab110ab 1031 u64 *off, u32 thresh)
4cb5300b
CM		 1032{
1033 struct btrfs_path *path;
1034 struct btrfs_key min_key;
4cb5300b
CM		 1035 struct extent_buffer *leaf;
1036 struct btrfs_file_extent_item *extent;
1037 int type;
1038 int ret;
4a0cc7ca 1039 u64 ino = btrfs_ino(BTRFS_I(inode));
4cb5300b
CM		 1040
1041 path = btrfs_alloc_path();
1042 if (!path)
1043 return -ENOMEM;
1044
a4689d2b 1045 min_key.objectid = ino;
4cb5300b
CM		 1046 min_key.type = BTRFS_EXTENT_DATA_KEY;
1047 min_key.offset = *off;
1048
67871254 1049 while (1) {
6174d3cb 1050 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 1051 if (ret != 0)
1052 goto none;
f094c9bd 1053process_slot:
a4689d2b 1054 if (min_key.objectid != ino)
4cb5300b
CM		 1055 goto none;
1056 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1057 goto none;
1058
1059 leaf = path->nodes[0];
1060 extent = btrfs_item_ptr(leaf, path->slots[0],
1061 struct btrfs_file_extent_item);
1062
1063 type = btrfs_file_extent_type(leaf, extent);
1064 if (type == BTRFS_FILE_EXTENT_REG &&
1065 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1066 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1067 *off = min_key.offset;
1068 btrfs_free_path(path);
1069 return 0;
1070 }
1071
f094c9bd
FM		 1072 path->slots[0]++;
1073 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1074 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1075 goto process_slot;
1076 }
1077
4cb5300b
CM		 1078 if (min_key.offset == (u64)-1)
1079 goto none;
1080
1081 min_key.offset++;
1082 btrfs_release_path(path);
1083 }
1084none:
1085 btrfs_free_path(path);
1086 return -ENOENT;
1087}
1088
6c282eb4 1089static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 1090{
1091 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 1092 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1093 struct extent_map *em;
09cbfeaf 1094 u64 len = PAGE_SIZE;
17ce6ef8 1095
6c282eb4
LZ		 1096 /*
1097 * hopefully we have this extent in the tree already, try without
1098 * the full extent lock
1099 */
17ce6ef8 1100 read_lock(&em_tree->lock);
6c282eb4 1101 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 1102 read_unlock(&em_tree->lock);
1103
6c282eb4 1104 if (!em) {
308d9800
FM		 1105 struct extent_state *cached = NULL;
1106 u64 end = start + len - 1;
1107
6c282eb4 1108 /* get the big lock and read metadata off disk */
ff13db41 1109 lock_extent_bits(io_tree, start, end, &cached);
fc4f21b1 1110 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
e43bbe5e 1111 unlock_extent_cached(io_tree, start, end, &cached);
6c282eb4
LZ		 1112
1113 if (IS_ERR(em))
1114 return NULL;
1115 }
1116
1117 return em;
1118}
17ce6ef8 1119
6c282eb4
LZ		 1120static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1121{
1122 struct extent_map *next;
1123 bool ret = true;
1124
1125 /* this is the last extent */
1126 if (em->start + em->len >= i_size_read(inode))
1127 return false;
1128
1129 next = defrag_lookup_extent(inode, em->start + em->len);
e9512d72
CM		 1130 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1131 ret = false;
1132 else if ((em->block_start + em->block_len == next->block_start) &&
ee22184b 1133 (em->block_len > SZ_128K && next->block_len > SZ_128K))
6c282eb4
LZ		 1134 ret = false;
1135
1136 free_extent_map(next);
17ce6ef8
LB		 1137 return ret;
1138}
1139
aab110ab 1140static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
a43a2111
AM		 1141 u64 *last_len, u64 *skip, u64 *defrag_end,
1142 int compress)
940100a4 1143{
6c282eb4 1144 struct extent_map *em;
940100a4 1145 int ret = 1;
6c282eb4 1146 bool next_mergeable = true;
4a3560c4 1147 bool prev_mergeable = true;
940100a4
CM		 1148
1149 /*
008873ea 1150 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1151 * defragging it
1152 */
1153 if (start < *defrag_end)
1154 return 1;
1155
1156 *skip = 0;
1157
6c282eb4
LZ		 1158 em = defrag_lookup_extent(inode, start);
1159 if (!em)
1160 return 0;
940100a4
CM		 1161
1162 /* this will cover holes, and inline extents */
17ce6ef8 1163 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1164 ret = 0;
17ce6ef8
LB		 1165 goto out;
1166 }
1167
4a3560c4
LB		 1168 if (!*defrag_end)
1169 prev_mergeable = false;
1170
6c282eb4 1171 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1172 /*
6c282eb4
LZ		 1173 * we hit a real extent, if it is big or the next extent is not a
1174 * real extent, don't bother defragging it
940100a4 1175 */
a43a2111 1176 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
4a3560c4 1177 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
940100a4 1178 ret = 0;
17ce6ef8 1179out:
940100a4
CM		 1180 /*
1181 * last_len ends up being a counter of how many bytes we've defragged.
1182 * every time we choose not to defrag an extent, we reset *last_len
1183 * so that the next tiny extent will force a defrag.
1184 *
1185 * The end result of this is that tiny extents before a single big
1186 * extent will force at least part of that big extent to be defragged.
1187 */
1188 if (ret) {
940100a4
CM		 1189 *defrag_end = extent_map_end(em);
1190 } else {
1191 *last_len = 0;
1192 *skip = extent_map_end(em);
1193 *defrag_end = 0;
1194 }
1195
1196 free_extent_map(em);
1197 return ret;
1198}
1199
4cb5300b
CM		 1200/*
1201 * it doesn't do much good to defrag one or two pages
1202 * at a time. This pulls in a nice chunk of pages
1203 * to COW and defrag.
1204 *
1205 * It also makes sure the delalloc code has enough
1206 * dirty data to avoid making new small extents as part
1207 * of the defrag
1208 *
1209 * It's a good idea to start RA on this range
1210 * before calling this.
1211 */
1212static int cluster_pages_for_defrag(struct inode *inode,
1213 struct page **pages,
1214 unsigned long start_index,
c41570c9 1215 unsigned long num_pages)
f46b5a66 1216{
4cb5300b
CM		 1217 unsigned long file_end;
1218 u64 isize = i_size_read(inode);
1219 u64 page_start;
1220 u64 page_end;
1f12bd06 1221 u64 page_cnt;
4cb5300b
CM		 1222 int ret;
1223 int i;
1224 int i_done;
3eaa2885 1225 struct btrfs_ordered_extent *ordered;
4cb5300b 1226 struct extent_state *cached_state = NULL;
600a45e1 1227 struct extent_io_tree *tree;
364ecf36 1228 struct extent_changeset *data_reserved = NULL;
3b16a4e3 1229 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1230
09cbfeaf 1231 file_end = (isize - 1) >> PAGE_SHIFT;
1f12bd06
LB		 1232 if (!isize || start_index > file_end)
1233 return 0;
1234
1235 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b 1236
364ecf36 1237 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
09cbfeaf
KS		 1238 start_index << PAGE_SHIFT,
1239 page_cnt << PAGE_SHIFT);
4cb5300b
CM		 1240 if (ret)
1241 return ret;
4cb5300b 1242 i_done = 0;
600a45e1 1243 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1244
1245 /* step one, lock all the pages */
1f12bd06 1246 for (i = 0; i < page_cnt; i++) {
4cb5300b 1247 struct page *page;
600a45e1 1248again:
a94733d0 1249 page = find_or_create_page(inode->i_mapping,
600a45e1 1250 start_index + i, mask);
4cb5300b
CM		 1251 if (!page)
1252 break;
1253
600a45e1 1254 page_start = page_offset(page);
09cbfeaf 1255 page_end = page_start + PAGE_SIZE - 1;
600a45e1 1256 while (1) {
308d9800 1257 lock_extent_bits(tree, page_start, page_end,
ff13db41 1258 &cached_state);
600a45e1
MX		 1259 ordered = btrfs_lookup_ordered_extent(inode,
1260 page_start);
308d9800 1261 unlock_extent_cached(tree, page_start, page_end,
e43bbe5e 1262 &cached_state);
600a45e1
MX		 1263 if (!ordered)
1264 break;
1265
1266 unlock_page(page);
1267 btrfs_start_ordered_extent(inode, ordered, 1);
1268 btrfs_put_ordered_extent(ordered);
1269 lock_page(page);
1f12bd06
LB		 1270 /*
1271 * we unlocked the page above, so we need check if
1272 * it was released or not.
1273 */
1274 if (page->mapping != inode->i_mapping) {
1275 unlock_page(page);
09cbfeaf 1276 put_page(page);
1f12bd06
LB		 1277 goto again;
1278 }
600a45e1
MX		 1279 }
1280
4cb5300b
CM		 1281 if (!PageUptodate(page)) {
1282 btrfs_readpage(NULL, page);
1283 lock_page(page);
1284 if (!PageUptodate(page)) {
1285 unlock_page(page);
09cbfeaf 1286 put_page(page);
4cb5300b
CM		 1287 ret = -EIO;
1288 break;
1289 }
1290 }
600a45e1 1291
600a45e1
MX		 1292 if (page->mapping != inode->i_mapping) {
1293 unlock_page(page);
09cbfeaf 1294 put_page(page);
600a45e1
MX		 1295 goto again;
1296 }
1297
4cb5300b
CM		 1298 pages[i] = page;
1299 i_done++;
1300 }
1301 if (!i_done || ret)
1302 goto out;
1303
1751e8a6 1304 if (!(inode->i_sb->s_flags & SB_ACTIVE))
4cb5300b
CM		 1305 goto out;
1306
1307 /*
1308 * so now we have a nice long stream of locked
1309 * and up to date pages, lets wait on them
1310 */
1311 for (i = 0; i < i_done; i++)
1312 wait_on_page_writeback(pages[i]);
1313
1314 page_start = page_offset(pages[0]);
09cbfeaf 1315 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
4cb5300b
CM		 1316
1317 lock_extent_bits(&BTRFS_I(inode)->io_tree,
ff13db41 1318 page_start, page_end - 1, &cached_state);
4cb5300b
CM		 1319 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1320 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b 1321 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
ae0f1625 1322 &cached_state);
4cb5300b 1323
1f12bd06 1324 if (i_done != page_cnt) {
9e0baf60 1325 spin_lock(&BTRFS_I(inode)->lock);
28c4a3e2 1326 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
9e0baf60 1327 spin_unlock(&BTRFS_I(inode)->lock);
bc42bda2 1328 btrfs_delalloc_release_space(inode, data_reserved,
09cbfeaf 1329 start_index << PAGE_SHIFT,
43b18595 1330 (page_cnt - i_done) << PAGE_SHIFT, true);
4cb5300b
CM		 1331 }
1332
1333
9e8a4a8b 1334 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
018ed4f7 1335 &cached_state);
4cb5300b
CM		 1336
1337 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
e43bbe5e 1338 page_start, page_end - 1, &cached_state);
4cb5300b
CM		 1339
1340 for (i = 0; i < i_done; i++) {
1341 clear_page_dirty_for_io(pages[i]);
1342 ClearPageChecked(pages[i]);
1343 set_page_extent_mapped(pages[i]);
1344 set_page_dirty(pages[i]);
1345 unlock_page(pages[i]);
09cbfeaf 1346 put_page(pages[i]);
4cb5300b 1347 }
43b18595
QW		 1348 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1349 false);
364ecf36 1350 extent_changeset_free(data_reserved);
4cb5300b
CM		 1351 return i_done;
1352out:
1353 for (i = 0; i < i_done; i++) {
1354 unlock_page(pages[i]);
09cbfeaf 1355 put_page(pages[i]);
4cb5300b 1356 }
bc42bda2 1357 btrfs_delalloc_release_space(inode, data_reserved,
09cbfeaf 1358 start_index << PAGE_SHIFT,
43b18595
QW		 1359 page_cnt << PAGE_SHIFT, true);
1360 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1361 true);
364ecf36 1362 extent_changeset_free(data_reserved);
4cb5300b
CM		 1363 return ret;
1364
1365}
1366
1367int btrfs_defrag_file(struct inode *inode, struct file *file,
1368 struct btrfs_ioctl_defrag_range_args *range,
1369 u64 newer_than, unsigned long max_to_defrag)
1370{
0b246afa 1371 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4cb5300b 1372 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1373 struct file_ra_state *ra = NULL;
f46b5a66 1374 unsigned long last_index;
151a31b2 1375 u64 isize = i_size_read(inode);
940100a4
CM		 1376 u64 last_len = 0;
1377 u64 skip = 0;
1378 u64 defrag_end = 0;
4cb5300b 1379 u64 newer_off = range->start;
f46b5a66 1380 unsigned long i;
008873ea 1381 unsigned long ra_index = 0;
f46b5a66 1382 int ret;
4cb5300b 1383 int defrag_count = 0;
1a419d85 1384 int compress_type = BTRFS_COMPRESS_ZLIB;
aab110ab 1385 u32 extent_thresh = range->extent_thresh;
09cbfeaf 1386 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
c41570c9 1387 unsigned long cluster = max_cluster;
ee22184b 1388 u64 new_align = ~((u64)SZ_128K - 1);
4cb5300b 1389 struct page **pages = NULL;
1e2ef46d 1390 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
4cb5300b 1391
0abd5b17
LB		 1392 if (isize == 0)
1393 return 0;
1394
1395 if (range->start >= isize)
1396 return -EINVAL;
1a419d85 1397
1e2ef46d 1398 if (do_compress) {
1a419d85
LZ		 1399 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1400 return -EINVAL;
1401 if (range->compress_type)
1402 compress_type = range->compress_type;
1403 }
f46b5a66 1404
0abd5b17 1405 if (extent_thresh == 0)
ee22184b 1406 extent_thresh = SZ_256K;
940100a4 1407
4cb5300b 1408 /*
0a52d108
DS		 1409 * If we were not given a file, allocate a readahead context. As
1410 * readahead is just an optimization, defrag will work without it so
1411 * we don't error out.
4cb5300b
CM		 1412 */
1413 if (!file) {
63e727ec 1414 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
0a52d108
DS		 1415 if (ra)
1416 file_ra_state_init(ra, inode->i_mapping);
4cb5300b
CM		 1417 } else {
1418 ra = &file->f_ra;
1419 }
1420
63e727ec 1421 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
4cb5300b
CM		 1422 if (!pages) {
1423 ret = -ENOMEM;
1424 goto out_ra;
1425 }
1426
1427 /* find the last page to defrag */
1e701a32 1428 if (range->start + range->len > range->start) {
151a31b2 1429 last_index = min_t(u64, isize - 1,
09cbfeaf 1430 range->start + range->len - 1) >> PAGE_SHIFT;
1e701a32 1431 } else {
09cbfeaf 1432 last_index = (isize - 1) >> PAGE_SHIFT;
1e701a32
CM		 1433 }
1434
4cb5300b
CM		 1435 if (newer_than) {
1436 ret = find_new_extents(root, inode, newer_than,
ee22184b 1437 &newer_off, SZ_64K);
4cb5300b
CM		 1438 if (!ret) {
1439 range->start = newer_off;
1440 /*
1441 * we always align our defrag to help keep
1442 * the extents in the file evenly spaced
1443 */
09cbfeaf 1444 i = (newer_off & new_align) >> PAGE_SHIFT;
4cb5300b
CM		 1445 } else
1446 goto out_ra;
1447 } else {
09cbfeaf 1448 i = range->start >> PAGE_SHIFT;
4cb5300b
CM		 1449 }
1450 if (!max_to_defrag)
070034bd 1451 max_to_defrag = last_index - i + 1;
4cb5300b 1452
2a0f7f57
LZ		 1453 /*
1454 * make writeback starts from i, so the defrag range can be
1455 * written sequentially.
1456 */
1457 if (i < inode->i_mapping->writeback_index)
1458 inode->i_mapping->writeback_index = i;
1459
f7f43cc8 1460 while (i <= last_index && defrag_count < max_to_defrag &&
09cbfeaf 1461 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
4cb5300b
CM		 1462 /*
1463 * make sure we stop running if someone unmounts
1464 * the FS
1465 */
1751e8a6 1466 if (!(inode->i_sb->s_flags & SB_ACTIVE))
4cb5300b
CM		 1467 break;
1468
0b246afa
JM		 1469 if (btrfs_defrag_cancelled(fs_info)) {
1470 btrfs_debug(fs_info, "defrag_file cancelled");
210549eb
DS		 1471 ret = -EAGAIN;
1472 break;
1473 }
1474
09cbfeaf 1475 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
6c282eb4 1476 extent_thresh, &last_len, &skip,
1e2ef46d 1477 &defrag_end, do_compress)){
940100a4
CM		 1478 unsigned long next;
1479 /*
1480 * the should_defrag function tells us how much to skip
1481 * bump our counter by the suggested amount
1482 */
09cbfeaf 1483 next = DIV_ROUND_UP(skip, PAGE_SIZE);
940100a4
CM		 1484 i = max(i + 1, next);
1485 continue;
1486 }
008873ea
LZ		 1487
1488 if (!newer_than) {
09cbfeaf
KS		 1489 cluster = (PAGE_ALIGN(defrag_end) >>
1490 PAGE_SHIFT) - i;
008873ea
LZ		 1491 cluster = min(cluster, max_cluster);
1492 } else {
1493 cluster = max_cluster;
1494 }
1495
008873ea
LZ		 1496 if (i + cluster > ra_index) {
1497 ra_index = max(i, ra_index);
0a52d108 1498 if (ra)
d3c0bab5
DS		 1499 page_cache_sync_readahead(inode->i_mapping, ra,
1500 file, ra_index, cluster);
e4826a5b 1501 ra_index += cluster;
008873ea 1502 }
940100a4 1503
5955102c 1504 inode_lock(inode);
1e2ef46d 1505 if (do_compress)
eec63c65 1506 BTRFS_I(inode)->defrag_compress = compress_type;
008873ea 1507 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8 1508 if (ret < 0) {
5955102c 1509 inode_unlock(inode);
4cb5300b 1510 goto out_ra;
ecb8bea8 1511 }
4cb5300b
CM		 1512
1513 defrag_count += ret;
d0e1d66b 1514 balance_dirty_pages_ratelimited(inode->i_mapping);
5955102c 1515 inode_unlock(inode);
4cb5300b
CM		 1516
1517 if (newer_than) {
1518 if (newer_off == (u64)-1)
1519 break;
1520
e1f041e1
LB		 1521 if (ret > 0)
1522 i += ret;
1523
4cb5300b 1524 newer_off = max(newer_off + 1,
09cbfeaf 1525 (u64)i << PAGE_SHIFT);
4cb5300b 1526
ee22184b
BL		 1527 ret = find_new_extents(root, inode, newer_than,
1528 &newer_off, SZ_64K);
4cb5300b
CM		 1529 if (!ret) {
1530 range->start = newer_off;
09cbfeaf 1531 i = (newer_off & new_align) >> PAGE_SHIFT;
4cb5300b
CM		 1532 } else {
1533 break;
f46b5a66 1534 }
4cb5300b 1535 } else {
008873ea 1536 if (ret > 0) {
cbcc8326 1537 i += ret;
09cbfeaf 1538 last_len += ret << PAGE_SHIFT;
008873ea 1539 } else {
cbcc8326 1540 i++;
008873ea
LZ		 1541 last_len = 0;
1542 }
f46b5a66 1543 }
f46b5a66
CH		 1544 }
1545
dec8ef90 1546 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1547 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1548 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1549 &BTRFS_I(inode)->runtime_flags))
1550 filemap_flush(inode->i_mapping);
1551 }
1e701a32 1552
1a419d85 1553 if (range->compress_type == BTRFS_COMPRESS_LZO) {
0b246afa 1554 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
5c1aab1d
NT		 1555 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1556 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1a419d85
LZ		 1557 }
1558
60ccf82f 1559 ret = defrag_count;
940100a4 1560
4cb5300b 1561out_ra:
1e2ef46d 1562 if (do_compress) {
5955102c 1563 inode_lock(inode);
eec63c65 1564 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
5955102c 1565 inode_unlock(inode);
633085c7 1566 }
4cb5300b
CM		 1567 if (!file)
1568 kfree(ra);
1569 kfree(pages);
940100a4 1570 return ret;
f46b5a66
CH		 1571}
1572
198605a8 1573static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1574 void __user *arg)
f46b5a66 1575{
0b246afa
JM		 1576 struct inode *inode = file_inode(file);
1577 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66
CH		 1578 u64 new_size;
1579 u64 old_size;
1580 u64 devid = 1;
0b246afa 1581 struct btrfs_root *root = BTRFS_I(inode)->root;
f46b5a66
CH		 1582 struct btrfs_ioctl_vol_args *vol_args;
1583 struct btrfs_trans_handle *trans;
1584 struct btrfs_device *device = NULL;
1585 char *sizestr;
9a40f122 1586 char *retptr;
f46b5a66
CH		 1587 char *devstr = NULL;
1588 int ret = 0;
f46b5a66
CH		 1589 int mod = 0;
1590
e441d54d
CM		 1591 if (!capable(CAP_SYS_ADMIN))
1592 return -EPERM;
1593
198605a8
MX		 1594 ret = mnt_want_write_file(file);
1595 if (ret)
1596 return ret;
1597
171938e5 1598 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
97547676 1599 mnt_drop_write_file(file);
e57138b3 1600 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1601 }
1602
dae7b665 1603 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1604 if (IS_ERR(vol_args)) {
1605 ret = PTR_ERR(vol_args);
1606 goto out;
1607 }
5516e595
MF		 1608
1609 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1610
f46b5a66
CH		 1611 sizestr = vol_args->name;
1612 devstr = strchr(sizestr, ':');
1613 if (devstr) {
f46b5a66
CH		 1614 sizestr = devstr + 1;
1615 *devstr = '\0';
1616 devstr = vol_args->name;
58dfae63
Z		 1617 ret = kstrtoull(devstr, 10, &devid);
1618 if (ret)
1619 goto out_free;
dfd79829
MX		 1620 if (!devid) {
1621 ret = -EINVAL;
1622 goto out_free;
1623 }
0b246afa 1624 btrfs_info(fs_info, "resizing devid %llu", devid);
f46b5a66 1625 }
dba60f3f 1626
0b246afa 1627 device = btrfs_find_device(fs_info, devid, NULL, NULL);
f46b5a66 1628 if (!device) {
0b246afa
JM		 1629 btrfs_info(fs_info, "resizer unable to find device %llu",
1630 devid);
dfd79829 1631 ret = -ENODEV;
c9e9f97b 1632 goto out_free;
f46b5a66 1633 }
dba60f3f 1634
ebbede42 1635 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
0b246afa 1636 btrfs_info(fs_info,
efe120a0 1637 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1638 devid);
dfd79829 1639 ret = -EPERM;
4e42ae1b
LB		 1640 goto out_free;
1641 }
1642
f46b5a66
CH		 1643 if (!strcmp(sizestr, "max"))
1644 new_size = device->bdev->bd_inode->i_size;
1645 else {
1646 if (sizestr[0] == '-') {
1647 mod = -1;
1648 sizestr++;
1649 } else if (sizestr[0] == '+') {
1650 mod = 1;
1651 sizestr++;
1652 }
9a40f122
GH		 1653 new_size = memparse(sizestr, &retptr);
1654 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1655 ret = -EINVAL;
c9e9f97b 1656 goto out_free;
f46b5a66
CH		 1657 }
1658 }
1659
401e29c1 1660 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
dfd79829 1661 ret = -EPERM;
63a212ab
SB		 1662 goto out_free;
1663 }
1664
7cc8e58d 1665 old_size = btrfs_device_get_total_bytes(device);
f46b5a66
CH		 1666
1667 if (mod < 0) {
1668 if (new_size > old_size) {
1669 ret = -EINVAL;
c9e9f97b 1670 goto out_free;
f46b5a66
CH		 1671 }
1672 new_size = old_size - new_size;
1673 } else if (mod > 0) {
eb8052e0 1674 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1675 ret = -ERANGE;
eb8052e0
WF		 1676 goto out_free;
1677 }
f46b5a66
CH		 1678 new_size = old_size + new_size;
1679 }
1680
ee22184b 1681 if (new_size < SZ_256M) {
f46b5a66 1682 ret = -EINVAL;
c9e9f97b 1683 goto out_free;
f46b5a66
CH		 1684 }
1685 if (new_size > device->bdev->bd_inode->i_size) {
1686 ret = -EFBIG;
c9e9f97b 1687 goto out_free;
f46b5a66
CH		 1688 }
1689
47f08b96 1690 new_size = round_down(new_size, fs_info->sectorsize);
f46b5a66 1691
0b246afa
JM		 1692 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1693 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1694
1695 if (new_size > old_size) {
a22285a6 1696 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1697 if (IS_ERR(trans)) {
1698 ret = PTR_ERR(trans);
c9e9f97b 1699 goto out_free;
98d5dc13 1700 }
f46b5a66 1701 ret = btrfs_grow_device(trans, device, new_size);
3a45bb20 1702 btrfs_commit_transaction(trans);
ece7d20e 1703 } else if (new_size < old_size) {
f46b5a66 1704 ret = btrfs_shrink_device(device, new_size);
0253f40e 1705 } /* equal, nothing need to do */
f46b5a66 1706
c9e9f97b 1707out_free:
f46b5a66 1708 kfree(vol_args);
c9e9f97b 1709out:
171938e5 1710 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
18f39c41 1711 mnt_drop_write_file(file);
f46b5a66
CH		 1712 return ret;
1713}
1714
72fd032e 1715static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
52f75f4f 1716 const char *name, unsigned long fd, int subvol,
6f72c7e2 1717 u64 *transid, bool readonly,
8696c533 1718 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1719{
f46b5a66 1720 int namelen;
3de4586c 1721 int ret = 0;
f46b5a66 1722
325c50e3
JM		 1723 if (!S_ISDIR(file_inode(file)->i_mode))
1724 return -ENOTDIR;
1725
a874a63e
LB		 1726 ret = mnt_want_write_file(file);
1727 if (ret)
1728 goto out;
1729
72fd032e
SW		 1730 namelen = strlen(name);
1731 if (strchr(name, '/')) {
f46b5a66 1732 ret = -EINVAL;
a874a63e 1733 goto out_drop_write;
f46b5a66
CH		 1734 }
1735
16780cab
CM		 1736 if (name[0] == '.' &&
1737 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1738 ret = -EEXIST;
a874a63e 1739 goto out_drop_write;
16780cab
CM		 1740 }
1741
3de4586c 1742 if (subvol) {
72fd032e 1743 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1744 NULL, transid, readonly, inherit);
cb8e7090 1745 } else {
2903ff01 1746 struct fd src = fdget(fd);
3de4586c 1747 struct inode *src_inode;
2903ff01 1748 if (!src.file) {
3de4586c 1749 ret = -EINVAL;
a874a63e 1750 goto out_drop_write;
3de4586c
CM		 1751 }
1752
496ad9aa
AV		 1753 src_inode = file_inode(src.file);
1754 if (src_inode->i_sb != file_inode(file)->i_sb) {
c79b4713 1755 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
efe120a0 1756 "Snapshot src from another FS");
23ad5b17 1757 ret = -EXDEV;
d0242061
DS		 1758 } else if (!inode_owner_or_capable(src_inode)) {
1759 /*
1760 * Subvolume creation is not restricted, but snapshots
1761 * are limited to own subvolumes only
1762 */
1763 ret = -EPERM;
ecd18815
AV		 1764 } else {
1765 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1766 BTRFS_I(src_inode)->root,
1767 transid, readonly, inherit);
3de4586c 1768 }
2903ff01 1769 fdput(src);
cb8e7090 1770 }
a874a63e
LB		 1771out_drop_write:
1772 mnt_drop_write_file(file);
f46b5a66 1773out:
72fd032e
SW		 1774 return ret;
1775}
1776
1777static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1778 void __user *arg, int subvol)
72fd032e 1779{
fa0d2b9b 1780 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1781 int ret;
1782
325c50e3
JM		 1783 if (!S_ISDIR(file_inode(file)->i_mode))
1784 return -ENOTDIR;
1785
fa0d2b9b
LZ		 1786 vol_args = memdup_user(arg, sizeof(*vol_args));
1787 if (IS_ERR(vol_args))
1788 return PTR_ERR(vol_args);
1789 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1790
fa0d2b9b 1791 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1792 vol_args->fd, subvol,
6f72c7e2 1793 NULL, false, NULL);
fdfb1e4f 1794
fa0d2b9b
LZ		 1795 kfree(vol_args);
1796 return ret;
1797}
fdfb1e4f 1798
fa0d2b9b
LZ		 1799static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1800 void __user *arg, int subvol)
1801{
1802 struct btrfs_ioctl_vol_args_v2 *vol_args;
1803 int ret;
1804 u64 transid = 0;
1805 u64 *ptr = NULL;
b83cc969 1806 bool readonly = false;
6f72c7e2 1807 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1808
325c50e3
JM		 1809 if (!S_ISDIR(file_inode(file)->i_mode))
1810 return -ENOTDIR;
1811
fa0d2b9b
LZ		 1812 vol_args = memdup_user(arg, sizeof(*vol_args));
1813 if (IS_ERR(vol_args))
1814 return PTR_ERR(vol_args);
1815 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1816
b83cc969 1817 if (vol_args->flags &
6f72c7e2
AJ		 1818 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1819 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1820 ret = -EOPNOTSUPP;
c47ca32d 1821 goto free_args;
72fd032e 1822 }
fa0d2b9b
LZ		 1823
1824 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1825 ptr = &transid;
b83cc969
LZ		 1826 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1827 readonly = true;
6f72c7e2 1828 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
09cbfeaf 1829 if (vol_args->size > PAGE_SIZE) {
6f72c7e2 1830 ret = -EINVAL;
c47ca32d 1831 goto free_args;
6f72c7e2
AJ		 1832 }
1833 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1834 if (IS_ERR(inherit)) {
1835 ret = PTR_ERR(inherit);
c47ca32d 1836 goto free_args;
6f72c7e2
AJ		 1837 }
1838 }
fa0d2b9b
LZ		 1839
1840 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1841 vol_args->fd, subvol, ptr,
8696c533 1842 readonly, inherit);
c47ca32d
DC		 1843 if (ret)
1844 goto free_inherit;
fa0d2b9b 1845
c47ca32d
DC		 1846 if (ptr && copy_to_user(arg +
1847 offsetof(struct btrfs_ioctl_vol_args_v2,
1848 transid),
1849 ptr, sizeof(*ptr)))
fa0d2b9b 1850 ret = -EFAULT;
c47ca32d
DC		 1851
1852free_inherit:
6f72c7e2 1853 kfree(inherit);
c47ca32d
DC		 1854free_args:
1855 kfree(vol_args);
f46b5a66
CH		 1856 return ret;
1857}
1858
0caa102d
LZ		 1859static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1860 void __user *arg)
1861{
496ad9aa 1862 struct inode *inode = file_inode(file);
0b246afa 1863 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
0caa102d
LZ		 1864 struct btrfs_root *root = BTRFS_I(inode)->root;
1865 int ret = 0;
1866 u64 flags = 0;
1867
4a0cc7ca 1868 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1869 return -EINVAL;
1870
0b246afa 1871 down_read(&fs_info->subvol_sem);
0caa102d
LZ		 1872 if (btrfs_root_readonly(root))
1873 flags |= BTRFS_SUBVOL_RDONLY;
0b246afa 1874 up_read(&fs_info->subvol_sem);
0caa102d
LZ		 1875
1876 if (copy_to_user(arg, &flags, sizeof(flags)))
1877 ret = -EFAULT;
1878
1879 return ret;
1880}
1881
1882static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1883 void __user *arg)
1884{
496ad9aa 1885 struct inode *inode = file_inode(file);
0b246afa 1886 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
0caa102d
LZ		 1887 struct btrfs_root *root = BTRFS_I(inode)->root;
1888 struct btrfs_trans_handle *trans;
1889 u64 root_flags;
1890 u64 flags;
1891 int ret = 0;
1892
bd60ea0f
DS		 1893 if (!inode_owner_or_capable(inode))
1894 return -EPERM;
1895
b9ca0664
LB		 1896 ret = mnt_want_write_file(file);
1897 if (ret)
1898 goto out;
0caa102d 1899
4a0cc7ca 1900 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
b9ca0664
LB		 1901 ret = -EINVAL;
1902 goto out_drop_write;
1903 }
0caa102d 1904
b9ca0664
LB		 1905 if (copy_from_user(&flags, arg, sizeof(flags))) {
1906 ret = -EFAULT;
1907 goto out_drop_write;
1908 }
0caa102d 1909
b9ca0664
LB		 1910 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1911 ret = -EINVAL;
1912 goto out_drop_write;
1913 }
0caa102d 1914
b9ca0664
LB		 1915 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1916 ret = -EOPNOTSUPP;
1917 goto out_drop_write;
1918 }
0caa102d 1919
0b246afa 1920 down_write(&fs_info->subvol_sem);
0caa102d
LZ		 1921
1922 /* nothing to do */
1923 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1924 goto out_drop_sem;
0caa102d
LZ		 1925
1926 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1927 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1928 btrfs_set_root_flags(&root->root_item,
1929 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1930 } else {
1931 /*
1932 * Block RO -> RW transition if this subvolume is involved in
1933 * send
1934 */
1935 spin_lock(&root->root_item_lock);
1936 if (root->send_in_progress == 0) {
1937 btrfs_set_root_flags(&root->root_item,
0caa102d 1938 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1939 spin_unlock(&root->root_item_lock);
1940 } else {
1941 spin_unlock(&root->root_item_lock);
0b246afa
JM		 1942 btrfs_warn(fs_info,
1943 "Attempt to set subvolume %llu read-write during send",
1944 root->root_key.objectid);
2c686537
DS		 1945 ret = -EPERM;
1946 goto out_drop_sem;
1947 }
1948 }
0caa102d
LZ		 1949
1950 trans = btrfs_start_transaction(root, 1);
1951 if (IS_ERR(trans)) {
1952 ret = PTR_ERR(trans);
1953 goto out_reset;
1954 }
1955
0b246afa 1956 ret = btrfs_update_root(trans, fs_info->tree_root,
0caa102d 1957 &root->root_key, &root->root_item);
9417ebc8
NB		 1958 if (ret < 0) {
1959 btrfs_end_transaction(trans);
1960 goto out_reset;
1961 }
1962
1963 ret = btrfs_commit_transaction(trans);
0caa102d 1964
0caa102d
LZ		 1965out_reset:
1966 if (ret)
1967 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1968out_drop_sem:
0b246afa 1969 up_write(&fs_info->subvol_sem);
b9ca0664
LB		 1970out_drop_write:
1971 mnt_drop_write_file(file);
1972out:
0caa102d
LZ		 1973 return ret;
1974}
1975
ac8e9819
CM		 1976static noinline int key_in_sk(struct btrfs_key *key,
1977 struct btrfs_ioctl_search_key *sk)
1978{
abc6e134
CM		 1979 struct btrfs_key test;
1980 int ret;
1981
1982 test.objectid = sk->min_objectid;
1983 test.type = sk->min_type;
1984 test.offset = sk->min_offset;
1985
1986 ret = btrfs_comp_cpu_keys(key, &test);
1987 if (ret < 0)
ac8e9819 1988 return 0;
abc6e134
CM		 1989
1990 test.objectid = sk->max_objectid;
1991 test.type = sk->max_type;
1992 test.offset = sk->max_offset;
1993
1994 ret = btrfs_comp_cpu_keys(key, &test);
1995 if (ret > 0)
ac8e9819
CM		 1996 return 0;
1997 return 1;
1998}
1999
df397565 2000static noinline int copy_to_sk(struct btrfs_path *path,
ac8e9819
CM		 2001 struct btrfs_key *key,
2002 struct btrfs_ioctl_search_key *sk,
9b6e817d 2003 size_t *buf_size,
ba346b35 2004 char __user *ubuf,
ac8e9819
CM		 2005 unsigned long *sk_offset,
2006 int *num_found)
2007{
2008 u64 found_transid;
2009 struct extent_buffer *leaf;
2010 struct btrfs_ioctl_search_header sh;
dd81d459 2011 struct btrfs_key test;
ac8e9819
CM		 2012 unsigned long item_off;
2013 unsigned long item_len;
2014 int nritems;
2015 int i;
2016 int slot;
ac8e9819
CM		 2017 int ret = 0;
2018
2019 leaf = path->nodes[0];
2020 slot = path->slots[0];
2021 nritems = btrfs_header_nritems(leaf);
2022
2023 if (btrfs_header_generation(leaf) > sk->max_transid) {
2024 i = nritems;
2025 goto advance_key;
2026 }
2027 found_transid = btrfs_header_generation(leaf);
2028
2029 for (i = slot; i < nritems; i++) {
2030 item_off = btrfs_item_ptr_offset(leaf, i);
2031 item_len = btrfs_item_size_nr(leaf, i);
2032
03b71c6c
GP		 2033 btrfs_item_key_to_cpu(leaf, key, i);
2034 if (!key_in_sk(key, sk))
2035 continue;
2036
9b6e817d 2037 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 2038 if (*num_found) {
2039 ret = 1;
2040 goto out;
2041 }
2042
2043 /*
2044 * return one empty item back for v1, which does not
2045 * handle -EOVERFLOW
2046 */
2047
9b6e817d 2048 *buf_size = sizeof(sh) + item_len;
ac8e9819 2049 item_len = 0;
8f5f6178
GH		 2050 ret = -EOVERFLOW;
2051 }
ac8e9819 2052
9b6e817d 2053 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 2054 ret = 1;
25c9bc2e 2055 goto out;
ac8e9819
CM		 2056 }
2057
ac8e9819
CM		 2058 sh.objectid = key->objectid;
2059 sh.offset = key->offset;
2060 sh.type = key->type;
2061 sh.len = item_len;
2062 sh.transid = found_transid;
2063
2064 /* copy search result header */
ba346b35
GH		 2065 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2066 ret = -EFAULT;
2067 goto out;
2068 }
2069
ac8e9819
CM		 2070 *sk_offset += sizeof(sh);
2071
2072 if (item_len) {
ba346b35 2073 char __user *up = ubuf + *sk_offset;
ac8e9819 2074 /* copy the item */
ba346b35
GH		 2075 if (read_extent_buffer_to_user(leaf, up,
2076 item_off, item_len)) {
2077 ret = -EFAULT;
2078 goto out;
2079 }
2080
ac8e9819 2081 *sk_offset += item_len;
ac8e9819 2082 }
e2156867 2083 (*num_found)++;
ac8e9819 2084
8f5f6178
GH		 2085 if (ret) /* -EOVERFLOW from above */
2086 goto out;
2087
25c9bc2e
GH		 2088 if (*num_found >= sk->nr_items) {
2089 ret = 1;
2090 goto out;
2091 }
ac8e9819
CM		 2092 }
2093advance_key:
abc6e134 2094 ret = 0;
dd81d459
NA		 2095 test.objectid = sk->max_objectid;
2096 test.type = sk->max_type;
2097 test.offset = sk->max_offset;
2098 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2099 ret = 1;
2100 else if (key->offset < (u64)-1)
ac8e9819 2101 key->offset++;
dd81d459 2102 else if (key->type < (u8)-1) {
abc6e134 2103 key->offset = 0;
ac8e9819 2104 key->type++;
dd81d459 2105 } else if (key->objectid < (u64)-1) {
abc6e134
CM		 2106 key->offset = 0;
2107 key->type = 0;
ac8e9819 2108 key->objectid++;
abc6e134
CM		 2109 } else
2110 ret = 1;
25c9bc2e 2111out:
ba346b35
GH		 2112 /*
2113 * 0: all items from this leaf copied, continue with next
2114 * 1: * more items can be copied, but unused buffer is too small
2115 * * all items were found
2116 * Either way, it will stops the loop which iterates to the next
2117 * leaf
2118 * -EOVERFLOW: item was to large for buffer
2119 * -EFAULT: could not copy extent buffer back to userspace
2120 */
ac8e9819
CM		 2121 return ret;
2122}
2123
2124static noinline int search_ioctl(struct inode *inode,
12544442 2125 struct btrfs_ioctl_search_key *sk,
9b6e817d 2126 size_t *buf_size,
ba346b35 2127 char __user *ubuf)
ac8e9819 2128{
0b246afa 2129 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
ac8e9819
CM		 2130 struct btrfs_root *root;
2131 struct btrfs_key key;
ac8e9819 2132 struct btrfs_path *path;
ac8e9819
CM		 2133 int ret;
2134 int num_found = 0;
2135 unsigned long sk_offset = 0;
2136
9b6e817d
GH		 2137 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2138 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2139 return -EOVERFLOW;
9b6e817d 2140 }
12544442 2141
ac8e9819
CM		 2142 path = btrfs_alloc_path();
2143 if (!path)
2144 return -ENOMEM;
2145
2146 if (sk->tree_id == 0) {
2147 /* search the root of the inode that was passed */
2148 root = BTRFS_I(inode)->root;
2149 } else {
2150 key.objectid = sk->tree_id;
2151 key.type = BTRFS_ROOT_ITEM_KEY;
2152 key.offset = (u64)-1;
2153 root = btrfs_read_fs_root_no_name(info, &key);
2154 if (IS_ERR(root)) {
ac8e9819 2155 btrfs_free_path(path);
ad1e3d56 2156 return PTR_ERR(root);
ac8e9819
CM		 2157 }
2158 }
2159
2160 key.objectid = sk->min_objectid;
2161 key.type = sk->min_type;
2162 key.offset = sk->min_offset;
2163
67871254 2164 while (1) {
6174d3cb 2165 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2166 if (ret != 0) {
2167 if (ret > 0)
2168 ret = 0;
2169 goto err;
2170 }
df397565 2171 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
ac8e9819 2172 &sk_offset, &num_found);
b3b4aa74 2173 btrfs_release_path(path);
25c9bc2e 2174 if (ret)
ac8e9819
CM		 2175 break;
2176
2177 }
8f5f6178
GH		 2178 if (ret > 0)
2179 ret = 0;
ac8e9819
CM		 2180err:
2181 sk->nr_items = num_found;
2182 btrfs_free_path(path);
2183 return ret;
2184}
2185
2186static noinline int btrfs_ioctl_tree_search(struct file *file,
2187 void __user *argp)
2188{
ba346b35
GH		 2189 struct btrfs_ioctl_search_args __user *uargs;
2190 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2191 struct inode *inode;
2192 int ret;
2193 size_t buf_size;
ac8e9819
CM		 2194
2195 if (!capable(CAP_SYS_ADMIN))
2196 return -EPERM;
2197
ba346b35
GH		 2198 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2199
2200 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2201 return -EFAULT;
ac8e9819 2202
ba346b35 2203 buf_size = sizeof(uargs->buf);
ac8e9819 2204
496ad9aa 2205 inode = file_inode(file);
ba346b35 2206 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2207
2208 /*
2209 * In the origin implementation an overflow is handled by returning a
2210 * search header with a len of zero, so reset ret.
2211 */
2212 if (ret == -EOVERFLOW)
2213 ret = 0;
2214
ba346b35 2215 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2216 ret = -EFAULT;
ac8e9819
CM		 2217 return ret;
2218}
2219
cc68a8a5
GH		 2220static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2221 void __user *argp)
2222{
2223 struct btrfs_ioctl_search_args_v2 __user *uarg;
2224 struct btrfs_ioctl_search_args_v2 args;
2225 struct inode *inode;
2226 int ret;
2227 size_t buf_size;
ee22184b 2228 const size_t buf_limit = SZ_16M;
cc68a8a5
GH		 2229
2230 if (!capable(CAP_SYS_ADMIN))
2231 return -EPERM;
2232
2233 /* copy search header and buffer size */
2234 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2235 if (copy_from_user(&args, uarg, sizeof(args)))
2236 return -EFAULT;
2237
2238 buf_size = args.buf_size;
2239
cc68a8a5
GH		 2240 /* limit result size to 16MB */
2241 if (buf_size > buf_limit)
2242 buf_size = buf_limit;
2243
2244 inode = file_inode(file);
2245 ret = search_ioctl(inode, &args.key, &buf_size,
718dc5fa 2246 (char __user *)(&uarg->buf[0]));
cc68a8a5
GH		 2247 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2248 ret = -EFAULT;
2249 else if (ret == -EOVERFLOW &&
2250 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2251 ret = -EFAULT;
2252
ac8e9819
CM		 2253 return ret;
2254}
2255
98d377a0 2256/*
ac8e9819
CM		 2257 * Search INODE_REFs to identify path name of 'dirid' directory
2258 * in a 'tree_id' tree. and sets path name to 'name'.
2259 */
98d377a0
TH		 2260static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2261 u64 tree_id, u64 dirid, char *name)
2262{
2263 struct btrfs_root *root;
2264 struct btrfs_key key;
ac8e9819 2265 char *ptr;
98d377a0
TH		 2266 int ret = -1;
2267 int slot;
2268 int len;
2269 int total_len = 0;
2270 struct btrfs_inode_ref *iref;
2271 struct extent_buffer *l;
2272 struct btrfs_path *path;
2273
2274 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2275 name[0]='\0';
2276 return 0;
2277 }
2278
2279 path = btrfs_alloc_path();
2280 if (!path)
2281 return -ENOMEM;
2282
c8bcbfbd 2283 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
98d377a0
TH		 2284
2285 key.objectid = tree_id;
2286 key.type = BTRFS_ROOT_ITEM_KEY;
2287 key.offset = (u64)-1;
2288 root = btrfs_read_fs_root_no_name(info, &key);
2289 if (IS_ERR(root)) {
ad1e3d56 2290 ret = PTR_ERR(root);
8ad6fcab 2291 goto out;
98d377a0
TH		 2292 }
2293
2294 key.objectid = dirid;
2295 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2296 key.offset = (u64)-1;
98d377a0 2297
67871254 2298 while (1) {
98d377a0
TH		 2299 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2300 if (ret < 0)
2301 goto out;
18674c6c
FDBM		 2302 else if (ret > 0) {
2303 ret = btrfs_previous_item(root, path, dirid,
2304 BTRFS_INODE_REF_KEY);
2305 if (ret < 0)
2306 goto out;
2307 else if (ret > 0) {
2308 ret = -ENOENT;
2309 goto out;
2310 }
2311 }
98d377a0
TH		 2312
2313 l = path->nodes[0];
2314 slot = path->slots[0];
2315 btrfs_item_key_to_cpu(l, &key, slot);
2316
98d377a0
TH		 2317 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2318 len = btrfs_inode_ref_name_len(l, iref);
2319 ptr -= len + 1;
2320 total_len += len + 1;
a696cf35
FDBM		 2321 if (ptr < name) {
2322 ret = -ENAMETOOLONG;
98d377a0 2323 goto out;
a696cf35 2324 }
98d377a0
TH		 2325
2326 *(ptr + len) = '/';
67871254 2327 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2328
2329 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2330 break;
2331
b3b4aa74 2332 btrfs_release_path(path);
98d377a0 2333 key.objectid = key.offset;
8ad6fcab 2334 key.offset = (u64)-1;
98d377a0 2335 dirid = key.objectid;
98d377a0 2336 }
77906a50 2337 memmove(name, ptr, total_len);
67871254 2338 name[total_len] = '\0';
98d377a0
TH		 2339 ret = 0;
2340out:
2341 btrfs_free_path(path);
ac8e9819
CM		 2342 return ret;
2343}
2344
23d0b79d
TM		 2345static int btrfs_search_path_in_tree_user(struct inode *inode,
2346 struct btrfs_ioctl_ino_lookup_user_args *args)
2347{
2348 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2349 struct super_block *sb = inode->i_sb;
2350 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2351 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2352 u64 dirid = args->dirid;
2353 unsigned long item_off;
2354 unsigned long item_len;
2355 struct btrfs_inode_ref *iref;
2356 struct btrfs_root_ref *rref;
2357 struct btrfs_root *root;
2358 struct btrfs_path *path;
2359 struct btrfs_key key, key2;
2360 struct extent_buffer *leaf;
2361 struct inode *temp_inode;
2362 char *ptr;
2363 int slot;
2364 int len;
2365 int total_len = 0;
2366 int ret;
2367
2368 path = btrfs_alloc_path();
2369 if (!path)
2370 return -ENOMEM;
2371
2372 /*
2373 * If the bottom subvolume does not exist directly under upper_limit,
2374 * construct the path in from the bottom up.
2375 */
2376 if (dirid != upper_limit.objectid) {
2377 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2378
2379 key.objectid = treeid;
2380 key.type = BTRFS_ROOT_ITEM_KEY;
2381 key.offset = (u64)-1;
2382 root = btrfs_read_fs_root_no_name(fs_info, &key);
2383 if (IS_ERR(root)) {
2384 ret = PTR_ERR(root);
2385 goto out;
2386 }
2387
2388 key.objectid = dirid;
2389 key.type = BTRFS_INODE_REF_KEY;
2390 key.offset = (u64)-1;
2391 while (1) {
2392 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2393 if (ret < 0) {
2394 goto out;
2395 } else if (ret > 0) {
2396 ret = btrfs_previous_item(root, path, dirid,
2397 BTRFS_INODE_REF_KEY);
2398 if (ret < 0) {
2399 goto out;
2400 } else if (ret > 0) {
2401 ret = -ENOENT;
2402 goto out;
2403 }
2404 }
2405
2406 leaf = path->nodes[0];
2407 slot = path->slots[0];
2408 btrfs_item_key_to_cpu(leaf, &key, slot);
2409
2410 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2411 len = btrfs_inode_ref_name_len(leaf, iref);
2412 ptr -= len + 1;
2413 total_len += len + 1;
2414 if (ptr < args->path) {
2415 ret = -ENAMETOOLONG;
2416 goto out;
2417 }
2418
2419 *(ptr + len) = '/';
2420 read_extent_buffer(leaf, ptr,
2421 (unsigned long)(iref + 1), len);
2422
2423 /* Check the read+exec permission of this directory */
2424 ret = btrfs_previous_item(root, path, dirid,
2425 BTRFS_INODE_ITEM_KEY);
2426 if (ret < 0) {
2427 goto out;
2428 } else if (ret > 0) {
2429 ret = -ENOENT;
2430 goto out;
2431 }
2432
2433 leaf = path->nodes[0];
2434 slot = path->slots[0];
2435 btrfs_item_key_to_cpu(leaf, &key2, slot);
2436 if (key2.objectid != dirid) {
2437 ret = -ENOENT;
2438 goto out;
2439 }
2440
2441 temp_inode = btrfs_iget(sb, &key2, root, NULL);
3ca57bd6
MT		 2442 if (IS_ERR(temp_inode)) {
2443 ret = PTR_ERR(temp_inode);
2444 goto out;
2445 }
23d0b79d
TM		 2446 ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2447 iput(temp_inode);
2448 if (ret) {
2449 ret = -EACCES;
2450 goto out;
2451 }
2452
2453 if (key.offset == upper_limit.objectid)
2454 break;
2455 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2456 ret = -EACCES;
2457 goto out;
2458 }
2459
2460 btrfs_release_path(path);
2461 key.objectid = key.offset;
2462 key.offset = (u64)-1;
2463 dirid = key.objectid;
2464 }
2465
2466 memmove(args->path, ptr, total_len);
2467 args->path[total_len] = '\0';
2468 btrfs_release_path(path);
2469 }
2470
2471 /* Get the bottom subvolume's name from ROOT_REF */
2472 root = fs_info->tree_root;
2473 key.objectid = treeid;
2474 key.type = BTRFS_ROOT_REF_KEY;
2475 key.offset = args->treeid;
2476 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2477 if (ret < 0) {
2478 goto out;
2479 } else if (ret > 0) {
2480 ret = -ENOENT;
2481 goto out;
2482 }
2483
2484 leaf = path->nodes[0];
2485 slot = path->slots[0];
2486 btrfs_item_key_to_cpu(leaf, &key, slot);
2487
2488 item_off = btrfs_item_ptr_offset(leaf, slot);
2489 item_len = btrfs_item_size_nr(leaf, slot);
2490 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2491 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2492 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2493 ret = -EINVAL;
2494 goto out;
2495 }
2496
2497 /* Copy subvolume's name */
2498 item_off += sizeof(struct btrfs_root_ref);
2499 item_len -= sizeof(struct btrfs_root_ref);
2500 read_extent_buffer(leaf, args->name, item_off, item_len);
2501 args->name[item_len] = 0;
2502
2503out:
2504 btrfs_free_path(path);
2505 return ret;
2506}
2507
ac8e9819
CM		 2508static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2509 void __user *argp)
2510{
bece2e82
BVA		 2511 struct btrfs_ioctl_ino_lookup_args *args;
2512 struct inode *inode;
01b810b8 2513 int ret = 0;
ac8e9819 2514
2354d08f
JL		 2515 args = memdup_user(argp, sizeof(*args));
2516 if (IS_ERR(args))
2517 return PTR_ERR(args);
c2b96929 2518
496ad9aa 2519 inode = file_inode(file);
ac8e9819 2520
01b810b8
DS		 2521 /*
2522 * Unprivileged query to obtain the containing subvolume root id. The
2523 * path is reset so it's consistent with btrfs_search_path_in_tree.
2524 */
1b53ac4d
CM		 2525 if (args->treeid == 0)
2526 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2527
01b810b8
DS		 2528 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2529 args->name[0] = 0;
2530 goto out;
2531 }
2532
2533 if (!capable(CAP_SYS_ADMIN)) {
2534 ret = -EPERM;
2535 goto out;
2536 }
2537
ac8e9819
CM		 2538 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2539 args->treeid, args->objectid,
2540 args->name);
2541
01b810b8 2542out:
ac8e9819
CM		 2543 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2544 ret = -EFAULT;
2545
2546 kfree(args);
98d377a0
TH		 2547 return ret;
2548}
2549
23d0b79d
TM		 2550/*
2551 * Version of ino_lookup ioctl (unprivileged)
2552 *
2553 * The main differences from ino_lookup ioctl are:
2554 *
2555 * 1. Read + Exec permission will be checked using inode_permission() during
2556 * path construction. -EACCES will be returned in case of failure.
2557 * 2. Path construction will be stopped at the inode number which corresponds
2558 * to the fd with which this ioctl is called. If constructed path does not
2559 * exist under fd's inode, -EACCES will be returned.
2560 * 3. The name of bottom subvolume is also searched and filled.
2561 */
2562static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2563{
2564 struct btrfs_ioctl_ino_lookup_user_args *args;
2565 struct inode *inode;
2566 int ret;
2567
2568 args = memdup_user(argp, sizeof(*args));
2569 if (IS_ERR(args))
2570 return PTR_ERR(args);
2571
2572 inode = file_inode(file);
2573
2574 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2575 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2576 /*
2577 * The subvolume does not exist under fd with which this is
2578 * called
2579 */
2580 kfree(args);
2581 return -EACCES;
2582 }
2583
2584 ret = btrfs_search_path_in_tree_user(inode, args);
2585
2586 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2587 ret = -EFAULT;
2588
2589 kfree(args);
2590 return ret;
2591}
2592
b64ec075
TM		 2593/* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2594static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2595{
2596 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2597 struct btrfs_fs_info *fs_info;
2598 struct btrfs_root *root;
2599 struct btrfs_path *path;
2600 struct btrfs_key key;
2601 struct btrfs_root_item *root_item;
2602 struct btrfs_root_ref *rref;
2603 struct extent_buffer *leaf;
2604 unsigned long item_off;
2605 unsigned long item_len;
2606 struct inode *inode;
2607 int slot;
2608 int ret = 0;
2609
2610 path = btrfs_alloc_path();
2611 if (!path)
2612 return -ENOMEM;
2613
2614 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2615 if (!subvol_info) {
2616 btrfs_free_path(path);
2617 return -ENOMEM;
2618 }
2619
2620 inode = file_inode(file);
2621 fs_info = BTRFS_I(inode)->root->fs_info;
2622
2623 /* Get root_item of inode's subvolume */
2624 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2625 key.type = BTRFS_ROOT_ITEM_KEY;
2626 key.offset = (u64)-1;
2627 root = btrfs_read_fs_root_no_name(fs_info, &key);
2628 if (IS_ERR(root)) {
2629 ret = PTR_ERR(root);
2630 goto out;
2631 }
2632 root_item = &root->root_item;
2633
2634 subvol_info->treeid = key.objectid;
2635
2636 subvol_info->generation = btrfs_root_generation(root_item);
2637 subvol_info->flags = btrfs_root_flags(root_item);
2638
2639 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2640 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2641 BTRFS_UUID_SIZE);
2642 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2643 BTRFS_UUID_SIZE);
2644
2645 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2646 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2647 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2648
2649 subvol_info->otransid = btrfs_root_otransid(root_item);
2650 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2651 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2652
2653 subvol_info->stransid = btrfs_root_stransid(root_item);
2654 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2655 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2656
2657 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2658 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2659 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2660
2661 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2662 /* Search root tree for ROOT_BACKREF of this subvolume */
2663 root = fs_info->tree_root;
2664
2665 key.type = BTRFS_ROOT_BACKREF_KEY;
2666 key.offset = 0;
2667 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2668 if (ret < 0) {
2669 goto out;
2670 } else if (path->slots[0] >=
2671 btrfs_header_nritems(path->nodes[0])) {
2672 ret = btrfs_next_leaf(root, path);
2673 if (ret < 0) {
2674 goto out;
2675 } else if (ret > 0) {
2676 ret = -EUCLEAN;
2677 goto out;
2678 }
2679 }
2680
2681 leaf = path->nodes[0];
2682 slot = path->slots[0];
2683 btrfs_item_key_to_cpu(leaf, &key, slot);
2684 if (key.objectid == subvol_info->treeid &&
2685 key.type == BTRFS_ROOT_BACKREF_KEY) {
2686 subvol_info->parent_id = key.offset;
2687
2688 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2689 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2690
2691 item_off = btrfs_item_ptr_offset(leaf, slot)
2692 + sizeof(struct btrfs_root_ref);
2693 item_len = btrfs_item_size_nr(leaf, slot)
2694 - sizeof(struct btrfs_root_ref);
2695 read_extent_buffer(leaf, subvol_info->name,
2696 item_off, item_len);
2697 } else {
2698 ret = -ENOENT;
2699 goto out;
2700 }
2701 }
2702
2703 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2704 ret = -EFAULT;
2705
2706out:
2707 btrfs_free_path(path);
2708 kzfree(subvol_info);
2709 return ret;
2710}
2711
42e4b520
TM		 2712/*
2713 * Return ROOT_REF information of the subvolume containing this inode
2714 * except the subvolume name.
2715 */
2716static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2717{
2718 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2719 struct btrfs_root_ref *rref;
2720 struct btrfs_root *root;
2721 struct btrfs_path *path;
2722 struct btrfs_key key;
2723 struct extent_buffer *leaf;
2724 struct inode *inode;
2725 u64 objectid;
2726 int slot;
2727 int ret;
2728 u8 found;
2729
2730 path = btrfs_alloc_path();
2731 if (!path)
2732 return -ENOMEM;
2733
2734 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2735 if (IS_ERR(rootrefs)) {
2736 btrfs_free_path(path);
2737 return PTR_ERR(rootrefs);
2738 }
2739
2740 inode = file_inode(file);
2741 root = BTRFS_I(inode)->root->fs_info->tree_root;
2742 objectid = BTRFS_I(inode)->root->root_key.objectid;
2743
2744 key.objectid = objectid;
2745 key.type = BTRFS_ROOT_REF_KEY;
2746 key.offset = rootrefs->min_treeid;
2747 found = 0;
2748
2749 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2750 if (ret < 0) {
2751 goto out;
2752 } else if (path->slots[0] >=
2753 btrfs_header_nritems(path->nodes[0])) {
2754 ret = btrfs_next_leaf(root, path);
2755 if (ret < 0) {
2756 goto out;
2757 } else if (ret > 0) {
2758 ret = -EUCLEAN;
2759 goto out;
2760 }
2761 }
2762 while (1) {
2763 leaf = path->nodes[0];
2764 slot = path->slots[0];
2765
2766 btrfs_item_key_to_cpu(leaf, &key, slot);
2767 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2768 ret = 0;
2769 goto out;
2770 }
2771
2772 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2773 ret = -EOVERFLOW;
2774 goto out;
2775 }
2776
2777 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2778 rootrefs->rootref[found].treeid = key.offset;
2779 rootrefs->rootref[found].dirid =
2780 btrfs_root_ref_dirid(leaf, rref);
2781 found++;
2782
2783 ret = btrfs_next_item(root, path);
2784 if (ret < 0) {
2785 goto out;
2786 } else if (ret > 0) {
2787 ret = -EUCLEAN;
2788 goto out;
2789 }
2790 }
2791
2792out:
2793 if (!ret || ret == -EOVERFLOW) {
2794 rootrefs->num_items = found;
2795 /* update min_treeid for next search */
2796 if (found)
2797 rootrefs->min_treeid =
2798 rootrefs->rootref[found - 1].treeid + 1;
2799 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2800 ret = -EFAULT;
2801 }
2802
2803 kfree(rootrefs);
2804 btrfs_free_path(path);
2805
2806 return ret;
2807}
2808
76dda93c
YZ		 2809static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2810 void __user *arg)
2811{
54563d41 2812 struct dentry *parent = file->f_path.dentry;
0b246afa 2813 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
76dda93c 2814 struct dentry *dentry;
2b0143b5 2815 struct inode *dir = d_inode(parent);
76dda93c
YZ		 2816 struct inode *inode;
2817 struct btrfs_root *root = BTRFS_I(dir)->root;
2818 struct btrfs_root *dest = NULL;
2819 struct btrfs_ioctl_vol_args *vol_args;
76dda93c 2820 int namelen;
76dda93c
YZ		 2821 int err = 0;
2822
325c50e3
JM		 2823 if (!S_ISDIR(dir->i_mode))
2824 return -ENOTDIR;
2825
76dda93c
YZ		 2826 vol_args = memdup_user(arg, sizeof(*vol_args));
2827 if (IS_ERR(vol_args))
2828 return PTR_ERR(vol_args);
2829
2830 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2831 namelen = strlen(vol_args->name);
2832 if (strchr(vol_args->name, '/') ||
2833 strncmp(vol_args->name, "..", namelen) == 0) {
2834 err = -EINVAL;
2835 goto out;
2836 }
2837
a561be71 2838 err = mnt_want_write_file(file);
76dda93c
YZ		 2839 if (err)
2840 goto out;
2841
521e0546 2842
00235411
AV		 2843 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2844 if (err == -EINTR)
2845 goto out_drop_write;
76dda93c
YZ		 2846 dentry = lookup_one_len(vol_args->name, parent, namelen);
2847 if (IS_ERR(dentry)) {
2848 err = PTR_ERR(dentry);
2849 goto out_unlock_dir;
2850 }
2851
2b0143b5 2852 if (d_really_is_negative(dentry)) {
76dda93c
YZ		 2853 err = -ENOENT;
2854 goto out_dput;
2855 }
2856
2b0143b5 2857 inode = d_inode(dentry);
4260f7c7 2858 dest = BTRFS_I(inode)->root;
67871254 2859 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2860 /*
2861 * Regular user. Only allow this with a special mount
2862 * option, when the user has write+exec access to the
2863 * subvol root, and when rmdir(2) would have been
2864 * allowed.
2865 *
2866 * Note that this is _not_ check that the subvol is
2867 * empty or doesn't contain data that we wouldn't
2868 * otherwise be able to delete.
2869 *
2870 * Users who want to delete empty subvols should try
2871 * rmdir(2).
2872 */
2873 err = -EPERM;
0b246afa 2874 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
4260f7c7
SW		 2875 goto out_dput;
2876
2877 /*
2878 * Do not allow deletion if the parent dir is the same
2879 * as the dir to be deleted. That means the ioctl
2880 * must be called on the dentry referencing the root
2881 * of the subvol, not a random directory contained
2882 * within it.
2883 */
2884 err = -EINVAL;
2885 if (root == dest)
2886 goto out_dput;
2887
2888 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2889 if (err)
2890 goto out_dput;
4260f7c7
SW		 2891 }
2892
5c39da5b
MX		 2893 /* check if subvolume may be deleted by a user */
2894 err = btrfs_may_delete(dir, dentry, 1);
2895 if (err)
2896 goto out_dput;
2897
4a0cc7ca 2898 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2899 err = -EINVAL;
2900 goto out_dput;
2901 }
2902
5955102c 2903 inode_lock(inode);
f60a2364 2904 err = btrfs_delete_subvolume(dir, dentry);
5955102c 2905 inode_unlock(inode);
f60a2364 2906 if (!err)
76dda93c 2907 d_delete(dentry);
fa6ac876 2908
76dda93c
YZ		 2909out_dput:
2910 dput(dentry);
2911out_unlock_dir:
5955102c 2912 inode_unlock(dir);
00235411 2913out_drop_write:
2a79f17e 2914 mnt_drop_write_file(file);
76dda93c
YZ		 2915out:
2916 kfree(vol_args);
2917 return err;
2918}
2919
1e701a32 2920static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2921{
496ad9aa 2922 struct inode *inode = file_inode(file);
f46b5a66 2923 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2924 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2925 int ret;
2926
25122d15
ID		 2927 ret = mnt_want_write_file(file);
2928 if (ret)
2929 return ret;
b83cc969 2930
25122d15
ID		 2931 if (btrfs_root_readonly(root)) {
2932 ret = -EROFS;
2933 goto out;
5ac00add 2934 }
f46b5a66
CH		 2935
2936 switch (inode->i_mode & S_IFMT) {
2937 case S_IFDIR:
e441d54d
CM		 2938 if (!capable(CAP_SYS_ADMIN)) {
2939 ret = -EPERM;
2940 goto out;
2941 }
de78b51a 2942 ret = btrfs_defrag_root(root);
f46b5a66
CH		 2943 break;
2944 case S_IFREG:
616d374e
AB		 2945 /*
2946 * Note that this does not check the file descriptor for write
2947 * access. This prevents defragmenting executables that are
2948 * running and allows defrag on files open in read-only mode.
2949 */
2950 if (!capable(CAP_SYS_ADMIN) &&
2951 inode_permission(inode, MAY_WRITE)) {
2952 ret = -EPERM;
e441d54d
CM		 2953 goto out;
2954 }
1e701a32
CM		 2955
2956 range = kzalloc(sizeof(*range), GFP_KERNEL);
2957 if (!range) {
2958 ret = -ENOMEM;
2959 goto out;
2960 }
2961
2962 if (argp) {
2963 if (copy_from_user(range, argp,
2964 sizeof(*range))) {
2965 ret = -EFAULT;
2966 kfree(range);
683be16e 2967 goto out;
1e701a32
CM		 2968 }
2969 /* compression requires us to start the IO */
2970 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2971 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2972 range->extent_thresh = (u32)-1;
2973 }
2974 } else {
2975 /* the rest are all set to zero by kzalloc */
2976 range->len = (u64)-1;
2977 }
496ad9aa 2978 ret = btrfs_defrag_file(file_inode(file), file,
7c829b72 2979 range, BTRFS_OLDEST_GENERATION, 0);
4cb5300b
CM		 2980 if (ret > 0)
2981 ret = 0;
1e701a32 2982 kfree(range);
f46b5a66 2983 break;
8929ecfa
YZ		 2984 default:
2985 ret = -EINVAL;
f46b5a66 2986 }
e441d54d 2987out:
25122d15 2988 mnt_drop_write_file(file);
e441d54d 2989 return ret;
f46b5a66
CH		 2990}
2991
2ff7e61e 2992static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
f46b5a66
CH		 2993{
2994 struct btrfs_ioctl_vol_args *vol_args;
2995 int ret;
2996
e441d54d
CM		 2997 if (!capable(CAP_SYS_ADMIN))
2998 return -EPERM;
2999
171938e5 3000 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
e57138b3 3001 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b 3002
dae7b665 3003 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 3004 if (IS_ERR(vol_args)) {
3005 ret = PTR_ERR(vol_args);
3006 goto out;
3007 }
f46b5a66 3008
5516e595 3009 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2ff7e61e 3010 ret = btrfs_init_new_device(fs_info, vol_args->name);
f46b5a66 3011
43d20761 3012 if (!ret)
0b246afa 3013 btrfs_info(fs_info, "disk added %s", vol_args->name);
43d20761 3014
f46b5a66 3015 kfree(vol_args);
c9e9f97b 3016out:
171938e5 3017 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
f46b5a66
CH		 3018 return ret;
3019}
3020
6b526ed7 3021static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
f46b5a66 3022{
0b246afa
JM		 3023 struct inode *inode = file_inode(file);
3024 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6b526ed7 3025 struct btrfs_ioctl_vol_args_v2 *vol_args;
f46b5a66
CH		 3026 int ret;
3027
e441d54d
CM		 3028 if (!capable(CAP_SYS_ADMIN))
3029 return -EPERM;
3030
da24927b
MX		 3031 ret = mnt_want_write_file(file);
3032 if (ret)
3033 return ret;
c146afad 3034
dae7b665 3035 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 3036 if (IS_ERR(vol_args)) {
3037 ret = PTR_ERR(vol_args);
c47ca32d 3038 goto err_drop;
c9e9f97b 3039 }
f46b5a66 3040
6b526ed7 3041 /* Check for compatibility reject unknown flags */
fd4e994b
OS		 3042 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) {
3043 ret = -EOPNOTSUPP;
3044 goto out;
3045 }
f46b5a66 3046
171938e5 3047 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
183860f6
AJ		 3048 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3049 goto out;
3050 }
3051
735654ea 3052 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2ff7e61e 3053 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
6b526ed7
AJ		 3054 } else {
3055 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2ff7e61e 3056 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
6b526ed7 3057 }
171938e5 3058 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
183860f6 3059
6b526ed7 3060 if (!ret) {
735654ea 3061 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
0b246afa 3062 btrfs_info(fs_info, "device deleted: id %llu",
6b526ed7
AJ		 3063 vol_args->devid);
3064 else
0b246afa 3065 btrfs_info(fs_info, "device deleted: %s",
6b526ed7
AJ		 3066 vol_args->name);
3067 }
183860f6
AJ		 3068out:
3069 kfree(vol_args);
c47ca32d 3070err_drop:
4ac20c70 3071 mnt_drop_write_file(file);
f46b5a66
CH		 3072 return ret;
3073}
3074
da24927b 3075static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 3076{
0b246afa
JM		 3077 struct inode *inode = file_inode(file);
3078 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66
CH		 3079 struct btrfs_ioctl_vol_args *vol_args;
3080 int ret;
3081
e441d54d
CM		 3082 if (!capable(CAP_SYS_ADMIN))
3083 return -EPERM;
3084
da24927b
MX		 3085 ret = mnt_want_write_file(file);
3086 if (ret)
3087 return ret;
c146afad 3088
171938e5 3089 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
183860f6 3090 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
58d7bbf8
DS		 3091 goto out_drop_write;
3092 }
3093
3094 vol_args = memdup_user(arg, sizeof(*vol_args));
3095 if (IS_ERR(vol_args)) {
3096 ret = PTR_ERR(vol_args);
183860f6
AJ		 3097 goto out;
3098 }
3099
58d7bbf8 3100 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2ff7e61e 3101 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
183860f6 3102
ec95d491 3103 if (!ret)
0b246afa 3104 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
183860f6 3105 kfree(vol_args);
58d7bbf8 3106out:
171938e5 3107 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
58d7bbf8 3108out_drop_write:
4ac20c70 3109 mnt_drop_write_file(file);
58d7bbf8 3110
f46b5a66
CH		 3111 return ret;
3112}
3113
2ff7e61e
JM		 3114static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3115 void __user *arg)
475f6387 3116{
027ed2f0 3117 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387 3118 struct btrfs_device *device;
0b246afa 3119 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
027ed2f0 3120 int ret = 0;
475f6387 3121
027ed2f0
LZ		 3122 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
3123 if (!fi_args)
3124 return -ENOMEM;
3125
d03262c7 3126 rcu_read_lock();
027ed2f0 3127 fi_args->num_devices = fs_devices->num_devices;
475f6387 3128
d03262c7 3129 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 3130 if (device->devid > fi_args->max_id)
3131 fi_args->max_id = device->devid;
475f6387 3132 }
d03262c7 3133 rcu_read_unlock();
475f6387 3134
d03262c7 3135 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
bea7eafd
OS		 3136 fi_args->nodesize = fs_info->nodesize;
3137 fi_args->sectorsize = fs_info->sectorsize;
3138 fi_args->clone_alignment = fs_info->sectorsize;
80a773fb 3139
027ed2f0
LZ		 3140 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3141 ret = -EFAULT;
475f6387 3142
027ed2f0
LZ		 3143 kfree(fi_args);
3144 return ret;
475f6387
JS		 3145}
3146
2ff7e61e
JM		 3147static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3148 void __user *arg)
475f6387
JS		 3149{
3150 struct btrfs_ioctl_dev_info_args *di_args;
3151 struct btrfs_device *dev;
475f6387
JS		 3152 int ret = 0;
3153 char *s_uuid = NULL;
475f6387 3154
475f6387
JS		 3155 di_args = memdup_user(arg, sizeof(*di_args));
3156 if (IS_ERR(di_args))
3157 return PTR_ERR(di_args);
3158
dd5f9615 3159 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 3160 s_uuid = di_args->uuid;
3161
c5593ca3 3162 rcu_read_lock();
0b246afa 3163 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 3164
3165 if (!dev) {
3166 ret = -ENODEV;
3167 goto out;
3168 }
3169
3170 di_args->devid = dev->devid;
7cc8e58d
MX		 3171 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3172 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
475f6387 3173 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 3174 if (dev->name) {
672d5990
MT		 3175 strncpy(di_args->path, rcu_str_deref(dev->name),
3176 sizeof(di_args->path) - 1);
a27202fb
JM		 3177 di_args->path[sizeof(di_args->path) - 1] = 0;
3178 } else {
99ba55ad 3179 di_args->path[0] = '\0';
a27202fb 3180 }
475f6387
JS		 3181
3182out:
c5593ca3 3183 rcu_read_unlock();
475f6387
JS		 3184 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3185 ret = -EFAULT;
3186
3187 kfree(di_args);
3188 return ret;
3189}
3190
f4414602 3191static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
416161db
MF		 3192{
3193 struct page *page;
416161db 3194
416161db
MF		 3195 page = grab_cache_page(inode->i_mapping, index);
3196 if (!page)
31314002 3197 return ERR_PTR(-ENOMEM);
416161db
MF		 3198
3199 if (!PageUptodate(page)) {
31314002
FM		 3200 int ret;
3201
3202 ret = btrfs_readpage(NULL, page);
3203 if (ret)
3204 return ERR_PTR(ret);
416161db
MF		 3205 lock_page(page);
3206 if (!PageUptodate(page)) {
3207 unlock_page(page);
09cbfeaf 3208 put_page(page);
31314002
FM		 3209 return ERR_PTR(-EIO);
3210 }
3211 if (page->mapping != inode->i_mapping) {
3212 unlock_page(page);
09cbfeaf 3213 put_page(page);
31314002 3214 return ERR_PTR(-EAGAIN);
416161db
MF		 3215 }
3216 }
416161db
MF		 3217
3218 return page;
3219}
3220
f4414602
MF		 3221static int gather_extent_pages(struct inode *inode, struct page **pages,
3222 int num_pages, u64 off)
3223{
3224 int i;
09cbfeaf 3225 pgoff_t index = off >> PAGE_SHIFT;
f4414602
MF		 3226
3227 for (i = 0; i < num_pages; i++) {
31314002 3228again:
f4414602 3229 pages[i] = extent_same_get_page(inode, index + i);
31314002
FM		 3230 if (IS_ERR(pages[i])) {
3231 int err = PTR_ERR(pages[i]);
3232
3233 if (err == -EAGAIN)
3234 goto again;
3235 pages[i] = NULL;
3236 return err;
3237 }
f4414602
MF		 3238 }
3239 return 0;
3240}
3241
e0bd70c6
FM		 3242static int lock_extent_range(struct inode *inode, u64 off, u64 len,
3243 bool retry_range_locking)
77fe20dc 3244{
e0bd70c6
FM		 3245 /*
3246 * Do any pending delalloc/csum calculations on inode, one way or
3247 * another, and lock file content.
3248 * The locking order is:
3249 *
3250 * 1) pages
3251 * 2) range in the inode's io tree
3252 */
77fe20dc
MF		 3253 while (1) {
3254 struct btrfs_ordered_extent *ordered;
3255 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
3256 ordered = btrfs_lookup_first_ordered_extent(inode,
3257 off + len - 1);
ff5df9b8
FM		 3258 if ((!ordered ||
3259 ordered->file_offset + ordered->len <= off ||
3260 ordered->file_offset >= off + len) &&
77fe20dc 3261 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
ff5df9b8
FM		 3262 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
3263 if (ordered)
3264 btrfs_put_ordered_extent(ordered);
77fe20dc 3265 break;
ff5df9b8 3266 }
77fe20dc
MF		 3267 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
3268 if (ordered)
3269 btrfs_put_ordered_extent(ordered);
e0bd70c6
FM		 3270 if (!retry_range_locking)
3271 return -EAGAIN;
77fe20dc
MF		 3272 btrfs_wait_ordered_range(inode, off, len);
3273 }
e0bd70c6 3274 return 0;
77fe20dc
MF		 3275}
3276
f4414602 3277static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
416161db 3278{
5955102c
AV		 3279 inode_unlock(inode1);
3280 inode_unlock(inode2);
416161db
MF		 3281}
3282
f4414602
MF		 3283static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
3284{
3285 if (inode1 < inode2)
3286 swap(inode1, inode2);
3287
5955102c
AV		 3288 inode_lock_nested(inode1, I_MUTEX_PARENT);
3289 inode_lock_nested(inode2, I_MUTEX_CHILD);
f4414602
MF		 3290}
3291
3292static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
3293 struct inode *inode2, u64 loff2, u64 len)
3294{
3295 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3296 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3297}
3298
e0bd70c6
FM		 3299static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
3300 struct inode *inode2, u64 loff2, u64 len,
3301 bool retry_range_locking)
416161db 3302{
e0bd70c6
FM		 3303 int ret;
3304
416161db
MF		 3305 if (inode1 < inode2) {
3306 swap(inode1, inode2);
3307 swap(loff1, loff2);
3308 }
e0bd70c6
FM		 3309 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
3310 if (ret)
3311 return ret;
3312 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
3313 if (ret)
3314 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
3315 loff1 + len - 1);
3316 return ret;
f4414602
MF		 3317}
3318
3319struct cmp_pages {
3320 int num_pages;
3321 struct page **src_pages;
3322 struct page **dst_pages;
3323};
3324
3325static void btrfs_cmp_data_free(struct cmp_pages *cmp)
3326{
3327 int i;
3328 struct page *pg;
3329
3330 for (i = 0; i < cmp->num_pages; i++) {
3331 pg = cmp->src_pages[i];
e0bd70c6
FM		 3332 if (pg) {
3333 unlock_page(pg);
09cbfeaf 3334 put_page(pg);
97b19170 3335 cmp->src_pages[i] = NULL;
e0bd70c6 3336 }
f4414602 3337 pg = cmp->dst_pages[i];
e0bd70c6
FM		 3338 if (pg) {
3339 unlock_page(pg);
09cbfeaf 3340 put_page(pg);
97b19170 3341 cmp->dst_pages[i] = NULL;
e0bd70c6 3342 }
f4414602 3343 }
f4414602
MF		 3344}
3345
3346static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
3347 struct inode *dst, u64 dst_loff,
3348 u64 len, struct cmp_pages *cmp)
3349{
3350 int ret;
09cbfeaf 3351 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
f4414602 3352
f4414602 3353 cmp->num_pages = num_pages;
b1517622 3354
67b07bd4 3355 ret = gather_extent_pages(src, cmp->src_pages, num_pages, loff);
f4414602
MF		 3356 if (ret)
3357 goto out;
3358
67b07bd4 3359 ret = gather_extent_pages(dst, cmp->dst_pages, num_pages, dst_loff);
f4414602
MF		 3360
3361out:
3362 if (ret)
3363 btrfs_cmp_data_free(cmp);
78ad4ce0 3364 return ret;
416161db
MF		 3365}
3366
1a287cfe 3367static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
416161db
MF		 3368{
3369 int ret = 0;
f4414602 3370 int i;
416161db 3371 struct page *src_page, *dst_page;
09cbfeaf 3372 unsigned int cmp_len = PAGE_SIZE;
416161db
MF		 3373 void *addr, *dst_addr;
3374
f4414602 3375 i = 0;
416161db 3376 while (len) {
09cbfeaf 3377 if (len < PAGE_SIZE)
416161db
MF		 3378 cmp_len = len;
3379
f4414602
MF		 3380 BUG_ON(i >= cmp->num_pages);
3381
3382 src_page = cmp->src_pages[i];
3383 dst_page = cmp->dst_pages[i];
e0bd70c6
FM		 3384 ASSERT(PageLocked(src_page));
3385 ASSERT(PageLocked(dst_page));
f4414602 3386
416161db
MF		 3387 addr = kmap_atomic(src_page);
3388 dst_addr = kmap_atomic(dst_page);
3389
3390 flush_dcache_page(src_page);
3391 flush_dcache_page(dst_page);
3392
3393 if (memcmp(addr, dst_addr, cmp_len))
2b3909f8 3394 ret = -EBADE;
416161db
MF		 3395
3396 kunmap_atomic(addr);
3397 kunmap_atomic(dst_addr);
416161db
MF		 3398
3399 if (ret)
3400 break;
3401
416161db 3402 len -= cmp_len;
f4414602 3403 i++;
416161db
MF		 3404 }
3405
3406 return ret;
3407}
3408
e1d227a4
MF		 3409static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3410 u64 olen)
416161db 3411{
e1d227a4 3412 u64 len = *plen;
416161db
MF		 3413 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3414
e1d227a4 3415 if (off + olen > inode->i_size || off + olen < off)
416161db 3416 return -EINVAL;
e1d227a4
MF		 3417
3418 /* if we extend to eof, continue to block boundary */
3419 if (off + len == inode->i_size)
3420 *plen = len = ALIGN(inode->i_size, bs) - off;
3421
416161db
MF		 3422 /* Check that we are block aligned - btrfs_clone() requires this */
3423 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3424 return -EINVAL;
3425
3426 return 0;
3427}
3428
3973909d 3429static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 olen,
67b07bd4
TT		 3430 struct inode *dst, u64 dst_loff,
3431 struct cmp_pages *cmp)
416161db
MF		 3432{
3433 int ret;
e1d227a4 3434 u64 len = olen;
fc4badd9 3435 bool same_inode = (src == dst);
0efa9f48
MF		 3436 u64 same_lock_start = 0;
3437 u64 same_lock_len = 0;
416161db 3438
fc4badd9
OS		 3439 ret = extent_same_check_offsets(src, loff, &len, olen);
3440 if (ret)
3973909d 3441 return ret;
416161db 3442
fc4badd9
OS		 3443 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3444 if (ret)
3973909d 3445 return ret;
fc4badd9
OS		 3446
3447 if (same_inode) {
0efa9f48
MF		 3448 /*
3449 * Single inode case wants the same checks, except we
3450 * don't want our length pushed out past i_size as
3451 * comparing that data range makes no sense.
3452 *
3453 * extent_same_check_offsets() will do this for an
3454 * unaligned length at i_size, so catch it here and
3455 * reject the request.
3456 *
3457 * This effectively means we require aligned extents
3458 * for the single-inode case, whereas the other cases
3459 * allow an unaligned length so long as it ends at
3460 * i_size.
3461 */
3973909d
TT		 3462 if (len != olen)
3463 return -EINVAL;
0efa9f48
MF		 3464
3465 /* Check for overlapping ranges */
3973909d
TT		 3466 if (dst_loff + len > loff && dst_loff < loff + len)
3467 return -EINVAL;
0efa9f48
MF		 3468
3469 same_lock_start = min_t(u64, loff, dst_loff);
3470 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
de02b9f6
FM		 3471 } else {
3472 /*
3473 * If the source and destination inodes are different, the
3474 * source's range end offset matches the source's i_size, that
3475 * i_size is not a multiple of the sector size, and the
3476 * destination range does not go past the destination's i_size,
3477 * we must round down the length to the nearest sector size
3478 * multiple. If we don't do this adjustment we end replacing
3479 * with zeroes the bytes in the range that starts at the
3480 * deduplication range's end offset and ends at the next sector
3481 * size multiple.
3482 */
3483 if (loff + olen == i_size_read(src) &&
3484 dst_loff + len < i_size_read(dst)) {
3485 const u64 sz = BTRFS_I(src)->root->fs_info->sectorsize;
3486
3487 len = round_down(i_size_read(src), sz) - loff;
3488 olen = len;
3489 }
0efa9f48 3490 }
416161db 3491
e0bd70c6 3492again:
67b07bd4 3493 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, cmp);
f4414602 3494 if (ret)
3973909d 3495 return ret;
f4414602 3496
0efa9f48 3497 if (same_inode)
e0bd70c6
FM		 3498 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3499 false);
0efa9f48 3500 else
e0bd70c6
FM		 3501 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3502 false);
3503 /*
3504 * If one of the inodes has dirty pages in the respective range or
3505 * ordered extents, we need to flush dellaloc and wait for all ordered
3506 * extents in the range. We must unlock the pages and the ranges in the
3507 * io trees to avoid deadlocks when flushing delalloc (requires locking
3508 * pages) and when waiting for ordered extents to complete (they require
3509 * range locking).
3510 */
3511 if (ret == -EAGAIN) {
3512 /*
3513 * Ranges in the io trees already unlocked. Now unlock all
3514 * pages before waiting for all IO to complete.
3515 */
67b07bd4 3516 btrfs_cmp_data_free(cmp);
e0bd70c6
FM		 3517 if (same_inode) {
3518 btrfs_wait_ordered_range(src, same_lock_start,
3519 same_lock_len);
3520 } else {
3521 btrfs_wait_ordered_range(src, loff, len);
3522 btrfs_wait_ordered_range(dst, dst_loff, len);
3523 }
3524 goto again;
3525 }
3526 ASSERT(ret == 0);
3527 if (WARN_ON(ret)) {
3528 /* ranges in the io trees already unlocked */
67b07bd4 3529 btrfs_cmp_data_free(cmp);
e0bd70c6
FM		 3530 return ret;
3531 }
f4414602 3532
207910dd 3533 /* pass original length for comparison so we stay within i_size */
67b07bd4 3534 ret = btrfs_cmp_data(olen, cmp);
416161db 3535 if (ret == 0)
1c919a5e 3536 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
416161db 3537
0efa9f48
MF		 3538 if (same_inode)
3539 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3540 same_lock_start + same_lock_len - 1);
3541 else
3542 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
f4414602 3543
67b07bd4 3544 btrfs_cmp_data_free(cmp);
3973909d
TT		 3545
3546 return ret;
3547}
3548
b6728768
TT		 3549#define BTRFS_MAX_DEDUPE_LEN SZ_16M
3550
3973909d
TT		 3551static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3552 struct inode *dst, u64 dst_loff)
3553{
3554 int ret;
67b07bd4
TT		 3555 struct cmp_pages cmp;
3556 int num_pages = PAGE_ALIGN(BTRFS_MAX_DEDUPE_LEN) >> PAGE_SHIFT;
3973909d 3557 bool same_inode = (src == dst);
b6728768 3558 u64 i, tail_len, chunk_count;
3973909d
TT		 3559
3560 if (olen == 0)
3561 return 0;
3562
3563 if (same_inode)
3564 inode_lock(src);
3565 else
3566 btrfs_double_inode_lock(src, dst);
3567
3568 /* don't make the dst file partly checksummed */
3569 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3570 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3571 ret = -EINVAL;
3572 goto out_unlock;
3573 }
3574
b6728768
TT		 3575 tail_len = olen % BTRFS_MAX_DEDUPE_LEN;
3576 chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN);
67b07bd4
TT		 3577 if (chunk_count == 0)
3578 num_pages = PAGE_ALIGN(tail_len) >> PAGE_SHIFT;
3579
3580 /*
3581 * If deduping ranges in the same inode, locking rules make it
3582 * mandatory to always lock pages in ascending order to avoid deadlocks
3583 * with concurrent tasks (such as starting writeback/delalloc).
3584 */
3585 if (same_inode && dst_loff < loff)
3586 swap(loff, dst_loff);
3587
3588 /*
3589 * We must gather up all the pages before we initiate our extent
3590 * locking. We use an array for the page pointers. Size of the array is
3591 * bounded by len, which is in turn bounded by BTRFS_MAX_DEDUPE_LEN.
3592 */
bf5091c8
DS		 3593 cmp.src_pages = kvmalloc_array(num_pages, sizeof(struct page *),
3594 GFP_KERNEL | __GFP_ZERO);
3595 cmp.dst_pages = kvmalloc_array(num_pages, sizeof(struct page *),
3596 GFP_KERNEL | __GFP_ZERO);
67b07bd4 3597 if (!cmp.src_pages || !cmp.dst_pages) {
bf5091c8
DS		 3598 ret = -ENOMEM;
3599 goto out_free;
67b07bd4 3600 }
b6728768
TT		 3601
3602 for (i = 0; i < chunk_count; i++) {
3603 ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN,
67b07bd4 3604 dst, dst_loff, &cmp);
b6728768 3605 if (ret)
22883ddc 3606 goto out_free;
b6728768
TT		 3607
3608 loff += BTRFS_MAX_DEDUPE_LEN;
3609 dst_loff += BTRFS_MAX_DEDUPE_LEN;
3610 }
3611
3612 if (tail_len > 0)
67b07bd4
TT		 3613 ret = btrfs_extent_same_range(src, loff, tail_len, dst,
3614 dst_loff, &cmp);
3973909d 3615
22883ddc
LF		 3616out_free:
3617 kvfree(cmp.src_pages);
3618 kvfree(cmp.dst_pages);
3619
416161db 3620out_unlock:
0efa9f48 3621 if (same_inode)
5955102c 3622 inode_unlock(src);
0efa9f48
MF		 3623 else
3624 btrfs_double_inode_unlock(src, dst);
416161db
MF		 3625
3626 return ret;
3627}
3628
87eb5eb2
MS		 3629int btrfs_dedupe_file_range(struct file *src_file, loff_t src_loff,
3630 struct file *dst_file, loff_t dst_loff,
3631 u64 olen)
416161db 3632{
2b3909f8
DW		 3633 struct inode *src = file_inode(src_file);
3634 struct inode *dst = file_inode(dst_file);
416161db 3635 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
416161db 3636
09cbfeaf 3637 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
416161db
MF		 3638 /*
3639 * Btrfs does not support blocksize < page_size. As a
3640 * result, btrfs_cmp_data() won't correctly handle
3641 * this situation without an update.
3642 */
2b3909f8 3643 return -EINVAL;
416161db
MF		 3644 }
3645
87eb5eb2 3646 return btrfs_extent_same(src, src_loff, olen, dst, dst_loff);
416161db
MF		 3647}
3648
f82a9901
FM		 3649static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3650 struct inode *inode,
3651 u64 endoff,
3652 const u64 destoff,
1c919a5e
MF		 3653 const u64 olen,
3654 int no_time_update)
f82a9901
FM		 3655{
3656 struct btrfs_root *root = BTRFS_I(inode)->root;
3657 int ret;
3658
3659 inode_inc_iversion(inode);
1c919a5e 3660 if (!no_time_update)
c2050a45 3661 inode->i_mtime = inode->i_ctime = current_time(inode);
f82a9901
FM		 3662 /*
3663 * We round up to the block size at eof when determining which
3664 * extents to clone above, but shouldn't round up the file size.
3665 */
3666 if (endoff > destoff + olen)
3667 endoff = destoff + olen;
3668 if (endoff > inode->i_size)
6ef06d27 3669 btrfs_i_size_write(BTRFS_I(inode), endoff);
f82a9901
FM		 3670
3671 ret = btrfs_update_inode(trans, root, inode);
3672 if (ret) {
66642832 3673 btrfs_abort_transaction(trans, ret);
3a45bb20 3674 btrfs_end_transaction(trans);
f82a9901
FM		 3675 goto out;
3676 }
3a45bb20 3677 ret = btrfs_end_transaction(trans);
f82a9901
FM		 3678out:
3679 return ret;
3680}
3681
a2f392e4 3682static void clone_update_extent_map(struct btrfs_inode *inode,
7ffbb598
FM		 3683 const struct btrfs_trans_handle *trans,
3684 const struct btrfs_path *path,
7ffbb598
FM		 3685 const u64 hole_offset,
3686 const u64 hole_len)
3687{
a2f392e4 3688 struct extent_map_tree *em_tree = &inode->extent_tree;
7ffbb598
FM		 3689 struct extent_map *em;
3690 int ret;
3691
3692 em = alloc_extent_map();
3693 if (!em) {
a2f392e4 3694 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
7ffbb598
FM		 3695 return;
3696 }
3697
14f59796
FM		 3698 if (path) {
3699 struct btrfs_file_extent_item *fi;
3700
3701 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3702 struct btrfs_file_extent_item);
a2f392e4 3703 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
7ffbb598
FM		 3704 em->generation = -1;
3705 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3706 BTRFS_FILE_EXTENT_INLINE)
3707 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
a2f392e4 3708 &inode->runtime_flags);
7ffbb598
FM		 3709 } else {
3710 em->start = hole_offset;
3711 em->len = hole_len;
3712 em->ram_bytes = em->len;
3713 em->orig_start = hole_offset;
3714 em->block_start = EXTENT_MAP_HOLE;
3715 em->block_len = 0;
3716 em->orig_block_len = 0;
3717 em->compress_type = BTRFS_COMPRESS_NONE;
3718 em->generation = trans->transid;
3719 }
3720
3721 while (1) {
3722 write_lock(&em_tree->lock);
3723 ret = add_extent_mapping(em_tree, em, 1);
3724 write_unlock(&em_tree->lock);
3725 if (ret != -EEXIST) {
3726 free_extent_map(em);
3727 break;
3728 }
a2f392e4 3729 btrfs_drop_extent_cache(inode, em->start,
7ffbb598
FM		 3730 em->start + em->len - 1, 0);
3731 }
3732
ee39b432 3733 if (ret)
a2f392e4 3734 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
7ffbb598
FM		 3735}
3736
8039d87d
FM		 3737/*
3738 * Make sure we do not end up inserting an inline extent into a file that has
3739 * already other (non-inline) extents. If a file has an inline extent it can
3740 * not have any other extents and the (single) inline extent must start at the
3741 * file offset 0. Failing to respect these rules will lead to file corruption,
3742 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3743 *
3744 * We can have extents that have been already written to disk or we can have
3745 * dirty ranges still in delalloc, in which case the extent maps and items are
3746 * created only when we run delalloc, and the delalloc ranges might fall outside
3747 * the range we are currently locking in the inode's io tree. So we check the
3748 * inode's i_size because of that (i_size updates are done while holding the
3749 * i_mutex, which we are holding here).
3750 * We also check to see if the inode has a size not greater than "datal" but has
3751 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3752 * protected against such concurrent fallocate calls by the i_mutex).
3753 *
3754 * If the file has no extents but a size greater than datal, do not allow the
3755 * copy because we would need turn the inline extent into a non-inline one (even
3756 * with NO_HOLES enabled). If we find our destination inode only has one inline
3757 * extent, just overwrite it with the source inline extent if its size is less
3758 * than the source extent's size, or we could copy the source inline extent's
3759 * data into the destination inode's inline extent if the later is greater then
3760 * the former.
3761 */
4a0ab9d7 3762static int clone_copy_inline_extent(struct inode *dst,
8039d87d
FM		 3763 struct btrfs_trans_handle *trans,
3764 struct btrfs_path *path,
3765 struct btrfs_key *new_key,
3766 const u64 drop_start,
3767 const u64 datal,
3768 const u64 skip,
3769 const u64 size,
3770 char *inline_data)
3771{
0b246afa 3772 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
8039d87d
FM		 3773 struct btrfs_root *root = BTRFS_I(dst)->root;
3774 const u64 aligned_end = ALIGN(new_key->offset + datal,
0b246afa 3775 fs_info->sectorsize);
8039d87d
FM		 3776 int ret;
3777 struct btrfs_key key;
3778
3779 if (new_key->offset > 0)
3780 return -EOPNOTSUPP;
3781
4a0cc7ca 3782 key.objectid = btrfs_ino(BTRFS_I(dst));
8039d87d
FM		 3783 key.type = BTRFS_EXTENT_DATA_KEY;
3784 key.offset = 0;
3785 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3786 if (ret < 0) {
3787 return ret;
3788 } else if (ret > 0) {
3789 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3790 ret = btrfs_next_leaf(root, path);
3791 if (ret < 0)
3792 return ret;
3793 else if (ret > 0)
3794 goto copy_inline_extent;
3795 }
3796 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
4a0cc7ca 3797 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
8039d87d
FM		 3798 key.type == BTRFS_EXTENT_DATA_KEY) {
3799 ASSERT(key.offset > 0);
3800 return -EOPNOTSUPP;
3801 }
3802 } else if (i_size_read(dst) <= datal) {
3803 struct btrfs_file_extent_item *ei;
3804 u64 ext_len;
3805
3806 /*
3807 * If the file size is <= datal, make sure there are no other
3808 * extents following (can happen do to an fallocate call with
3809 * the flag FALLOC_FL_KEEP_SIZE).
3810 */
3811 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3812 struct btrfs_file_extent_item);
3813 /*
3814 * If it's an inline extent, it can not have other extents
3815 * following it.
3816 */
3817 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3818 BTRFS_FILE_EXTENT_INLINE)
3819 goto copy_inline_extent;
3820
3821 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3822 if (ext_len > aligned_end)
3823 return -EOPNOTSUPP;
3824
3825 ret = btrfs_next_item(root, path);
3826 if (ret < 0) {
3827 return ret;
3828 } else if (ret == 0) {
3829 btrfs_item_key_to_cpu(path->nodes[0], &key,
3830 path->slots[0]);
4a0cc7ca 3831 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
8039d87d
FM		 3832 key.type == BTRFS_EXTENT_DATA_KEY)
3833 return -EOPNOTSUPP;
3834 }
3835 }
3836
3837copy_inline_extent:
3838 /*
3839 * We have no extent items, or we have an extent at offset 0 which may
3840 * or may not be inlined. All these cases are dealt the same way.
3841 */
3842 if (i_size_read(dst) > datal) {
3843 /*
3844 * If the destination inode has an inline extent...
3845 * This would require copying the data from the source inline
3846 * extent into the beginning of the destination's inline extent.
3847 * But this is really complex, both extents can be compressed
3848 * or just one of them, which would require decompressing and
3849 * re-compressing data (which could increase the new compressed
3850 * size, not allowing the compressed data to fit anymore in an
3851 * inline extent).
3852 * So just don't support this case for now (it should be rare,
3853 * we are not really saving space when cloning inline extents).
3854 */
3855 return -EOPNOTSUPP;
3856 }
3857
3858 btrfs_release_path(path);
3859 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3860 if (ret)
3861 return ret;
3862 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3863 if (ret)
3864 return ret;
3865
3866 if (skip) {
3867 const u32 start = btrfs_file_extent_calc_inline_size(0);
3868
3869 memmove(inline_data + start, inline_data + start + skip, datal);
3870 }
3871
3872 write_extent_buffer(path->nodes[0], inline_data,
3873 btrfs_item_ptr_offset(path->nodes[0],
3874 path->slots[0]),
3875 size);
3876 inode_add_bytes(dst, datal);
3877
3878 return 0;
3879}
3880
32b7c687
MF		 3881/**
3882 * btrfs_clone() - clone a range from inode file to another
3883 *
3884 * @src: Inode to clone from
3885 * @inode: Inode to clone to
3886 * @off: Offset within source to start clone from
3887 * @olen: Original length, passed by user, of range to clone
1c919a5e 3888 * @olen_aligned: Block-aligned value of olen
32b7c687 3889 * @destoff: Offset within @inode to start clone
1c919a5e 3890 * @no_time_update: Whether to update mtime/ctime on the target inode
32b7c687
MF		 3891 */
3892static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901 3893 const u64 off, const u64 olen, const u64 olen_aligned,
1c919a5e 3894 const u64 destoff, int no_time_update)
f46b5a66 3895{
0b246afa 3896 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66 3897 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3898 struct btrfs_path *path = NULL;
f46b5a66 3899 struct extent_buffer *leaf;
32b7c687
MF		 3900 struct btrfs_trans_handle *trans;
3901 char *buf = NULL;
ae01a0ab 3902 struct btrfs_key key;
f46b5a66
CH		 3903 u32 nritems;
3904 int slot;
ae01a0ab 3905 int ret;
f82a9901 3906 const u64 len = olen_aligned;
f82a9901 3907 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3908
3909 ret = -ENOMEM;
752ade68
MH		 3910 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3911 if (!buf)
3912 return ret;
ae01a0ab
YZ		 3913
3914 path = btrfs_alloc_path();
3915 if (!path) {
15351955 3916 kvfree(buf);
32b7c687 3917 return ret;
d525e8ab
LZ		 3918 }
3919
e4058b54 3920 path->reada = READA_FORWARD;
c5c9cd4d 3921 /* clone data */
4a0cc7ca 3922 key.objectid = btrfs_ino(BTRFS_I(src));
ae01a0ab 3923 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3924 key.offset = off;
f46b5a66
CH		 3925
3926 while (1) {
de249e66 3927 u64 next_key_min_offset = key.offset + 1;
df858e76 3928
f46b5a66
CH		 3929 /*
3930 * note the key will change type as we walk through the
3931 * tree.
3932 */
e4355f34 3933 path->leave_spinning = 1;
362a20c5
DS		 3934 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3935 0, 0);
f46b5a66
CH		 3936 if (ret < 0)
3937 goto out;
2c463823
FM		 3938 /*
3939 * First search, if no extent item that starts at offset off was
3940 * found but the previous item is an extent item, it's possible
3941 * it might overlap our target range, therefore process it.
3942 */
3943 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3944 btrfs_item_key_to_cpu(path->nodes[0], &key,
3945 path->slots[0] - 1);
3946 if (key.type == BTRFS_EXTENT_DATA_KEY)
3947 path->slots[0]--;
3948 }
f46b5a66 3949
ae01a0ab 3950 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3951process_slot:
ae01a0ab 3952 if (path->slots[0] >= nritems) {
362a20c5 3953 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3954 if (ret < 0)
3955 goto out;
3956 if (ret > 0)
3957 break;
ae01a0ab 3958 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3959 }
3960 leaf = path->nodes[0];
3961 slot = path->slots[0];
f46b5a66 3962
ae01a0ab 3963 btrfs_item_key_to_cpu(leaf, &key, slot);
962a298f 3964 if (key.type > BTRFS_EXTENT_DATA_KEY ||
4a0cc7ca 3965 key.objectid != btrfs_ino(BTRFS_I(src)))
f46b5a66
CH		 3966 break;
3967
962a298f 3968 if (key.type == BTRFS_EXTENT_DATA_KEY) {
c5c9cd4d
SW		 3969 struct btrfs_file_extent_item *extent;
3970 int type;
31840ae1
ZY		 3971 u32 size;
3972 struct btrfs_key new_key;
c5c9cd4d
SW		 3973 u64 disko = 0, diskl = 0;
3974 u64 datao = 0, datal = 0;
3975 u8 comp;
f82a9901 3976 u64 drop_start;
31840ae1 3977
c5c9cd4d
SW		 3978 extent = btrfs_item_ptr(leaf, slot,
3979 struct btrfs_file_extent_item);
3980 comp = btrfs_file_extent_compression(leaf, extent);
3981 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3982 if (type == BTRFS_FILE_EXTENT_REG ||
3983 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3984 disko = btrfs_file_extent_disk_bytenr(leaf,
3985 extent);
3986 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3987 extent);
c5c9cd4d 3988 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3989 datal = btrfs_file_extent_num_bytes(leaf,
3990 extent);
c5c9cd4d
SW		 3991 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3992 /* take upper bound, may be compressed */
3993 datal = btrfs_file_extent_ram_bytes(leaf,
3994 extent);
3995 }
31840ae1 3996
2c463823
FM		 3997 /*
3998 * The first search might have left us at an extent
3999 * item that ends before our target range's start, can
4000 * happen if we have holes and NO_HOLES feature enabled.
4001 */
4002 if (key.offset + datal <= off) {
e4355f34
FDBM		 4003 path->slots[0]++;
4004 goto process_slot;
2c463823
FM		 4005 } else if (key.offset >= off + len) {
4006 break;
e4355f34 4007 }
df858e76 4008 next_key_min_offset = key.offset + datal;
e4355f34
FDBM		 4009 size = btrfs_item_size_nr(leaf, slot);
4010 read_extent_buffer(leaf, buf,
4011 btrfs_item_ptr_offset(leaf, slot),
4012 size);
4013
4014 btrfs_release_path(path);
4015 path->leave_spinning = 0;
c5c9cd4d 4016
31840ae1 4017 memcpy(&new_key, &key, sizeof(new_key));
4a0cc7ca 4018 new_key.objectid = btrfs_ino(BTRFS_I(inode));
4d728ec7
LZ		 4019 if (off <= key.offset)
4020 new_key.offset = key.offset + destoff - off;
4021 else
4022 new_key.offset = destoff;
31840ae1 4023
f82a9901
FM		 4024 /*
4025 * Deal with a hole that doesn't have an extent item
4026 * that represents it (NO_HOLES feature enabled).
4027 * This hole is either in the middle of the cloning
4028 * range or at the beginning (fully overlaps it or
4029 * partially overlaps it).
4030 */
4031 if (new_key.offset != last_dest_end)
4032 drop_start = last_dest_end;
4033 else
4034 drop_start = new_key.offset;
4035
b6f3409b
SW		 4036 /*
4037 * 1 - adjusting old extent (we may have to split it)
4038 * 1 - add new extent
4039 * 1 - inode update
4040 */
4041 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 4042 if (IS_ERR(trans)) {
4043 ret = PTR_ERR(trans);
4044 goto out;
4045 }
4046
c8a894d7
CM		 4047 if (type == BTRFS_FILE_EXTENT_REG ||
4048 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 4049 /*
4050 * a | --- range to clone ---| b
4051 * | ------------- extent ------------- |
4052 */
4053
93915584 4054 /* subtract range b */
d72c0842
LZ		 4055 if (key.offset + datal > off + len)
4056 datal = off + len - key.offset;
4057
93915584 4058 /* subtract range a */
a22285a6
YZ		 4059 if (off > key.offset) {
4060 datao += off - key.offset;
4061 datal -= off - key.offset;
4062 }
4063
5dc562c5 4064 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 4065 drop_start,
a22285a6 4066 new_key.offset + datal,
2671485d 4067 1);
79787eaa 4068 if (ret) {
3f9e3df8 4069 if (ret != -EOPNOTSUPP)
00fdf13a 4070 btrfs_abort_transaction(trans,
66642832 4071 ret);
3a45bb20 4072 btrfs_end_transaction(trans);
79787eaa
JM		 4073 goto out;
4074 }
a22285a6 4075
c5c9cd4d
SW		 4076 ret = btrfs_insert_empty_item(trans, root, path,
4077 &new_key, size);
79787eaa 4078 if (ret) {
66642832 4079 btrfs_abort_transaction(trans, ret);
3a45bb20 4080 btrfs_end_transaction(trans);
79787eaa
JM		 4081 goto out;
4082 }
c5c9cd4d
SW		 4083
4084 leaf = path->nodes[0];
4085 slot = path->slots[0];
4086 write_extent_buffer(leaf, buf,
31840ae1
ZY		 4087 btrfs_item_ptr_offset(leaf, slot),
4088 size);
ae01a0ab 4089
c5c9cd4d 4090 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 4091 struct btrfs_file_extent_item);
c5c9cd4d 4092
c5c9cd4d
SW		 4093 /* disko == 0 means it's a hole */
4094 if (!disko)
4095 datao = 0;
c5c9cd4d
SW		 4096
4097 btrfs_set_file_extent_offset(leaf, extent,
4098 datao);
4099 btrfs_set_file_extent_num_bytes(leaf, extent,
4100 datal);
fcebe456 4101
c5c9cd4d
SW		 4102 if (disko) {
4103 inode_add_bytes(inode, datal);
2ff7e61e 4104 ret = btrfs_inc_extent_ref(trans,
84f7d8e6 4105 root,
5d4f98a2
YZ		 4106 disko, diskl, 0,
4107 root->root_key.objectid,
4a0cc7ca 4108 btrfs_ino(BTRFS_I(inode)),
b06c4bf5 4109 new_key.offset - datao);
79787eaa
JM		 4110 if (ret) {
4111 btrfs_abort_transaction(trans,
79787eaa 4112 ret);
3a45bb20 4113 btrfs_end_transaction(trans);
79787eaa
JM		 4114 goto out;
4115
4116 }
f46b5a66 4117 }
c5c9cd4d
SW		 4118 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
4119 u64 skip = 0;
4120 u64 trim = 0;
ed958762 4121
c5c9cd4d
SW		 4122 if (off > key.offset) {
4123 skip = off - key.offset;
4124 new_key.offset += skip;
4125 }
d397712b 4126
aa42ffd9
LB		 4127 if (key.offset + datal > off + len)
4128 trim = key.offset + datal - (off + len);
d397712b 4129
c5c9cd4d 4130 if (comp && (skip || trim)) {
c5c9cd4d 4131 ret = -EINVAL;
3a45bb20 4132 btrfs_end_transaction(trans);
c5c9cd4d
SW		 4133 goto out;
4134 }
4135 size -= skip + trim;
4136 datal -= skip + trim;
a22285a6 4137
4a0ab9d7 4138 ret = clone_copy_inline_extent(inode,
8039d87d
FM		 4139 trans, path,
4140 &new_key,
4141 drop_start,
4142 datal,
4143 skip, size, buf);
79787eaa 4144 if (ret) {
3f9e3df8 4145 if (ret != -EOPNOTSUPP)
3a29bc09 4146 btrfs_abort_transaction(trans,
8039d87d 4147 ret);
3a45bb20 4148 btrfs_end_transaction(trans);
79787eaa
JM		 4149 goto out;
4150 }
c5c9cd4d
SW		 4151 leaf = path->nodes[0];
4152 slot = path->slots[0];
f46b5a66 4153 }
c5c9cd4d 4154
7ffbb598
FM		 4155 /* If we have an implicit hole (NO_HOLES feature). */
4156 if (drop_start < new_key.offset)
a2f392e4 4157 clone_update_extent_map(BTRFS_I(inode), trans,
14f59796 4158 NULL, drop_start,
7ffbb598
FM		 4159 new_key.offset - drop_start);
4160
a2f392e4
NB		 4161 clone_update_extent_map(BTRFS_I(inode), trans,
4162 path, 0, 0);
7ffbb598 4163
c5c9cd4d 4164 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 4165 btrfs_release_path(path);
c5c9cd4d 4166
62e2390e 4167 last_dest_end = ALIGN(new_key.offset + datal,
0b246afa 4168 fs_info->sectorsize);
f82a9901
FM		 4169 ret = clone_finish_inode_update(trans, inode,
4170 last_dest_end,
1c919a5e
MF		 4171 destoff, olen,
4172 no_time_update);
f82a9901 4173 if (ret)
79787eaa 4174 goto out;
2c463823
FM		 4175 if (new_key.offset + datal >= destoff + len)
4176 break;
a22285a6 4177 }
b3b4aa74 4178 btrfs_release_path(path);
df858e76 4179 key.offset = next_key_min_offset;
69ae5e44
WX		 4180
4181 if (fatal_signal_pending(current)) {
4182 ret = -EINTR;
4183 goto out;
4184 }
f46b5a66 4185 }
f46b5a66 4186 ret = 0;
32b7c687 4187
f82a9901
FM		 4188 if (last_dest_end < destoff + len) {
4189 /*
4190 * We have an implicit hole (NO_HOLES feature is enabled) that
4191 * fully or partially overlaps our cloning range at its end.
4192 */
4193 btrfs_release_path(path);
4194
4195 /*
4196 * 1 - remove extent(s)
4197 * 1 - inode update
4198 */
4199 trans = btrfs_start_transaction(root, 2);
4200 if (IS_ERR(trans)) {
4201 ret = PTR_ERR(trans);
4202 goto out;
4203 }
4204 ret = btrfs_drop_extents(trans, root, inode,
4205 last_dest_end, destoff + len, 1);
4206 if (ret) {
4207 if (ret != -EOPNOTSUPP)
66642832 4208 btrfs_abort_transaction(trans, ret);
3a45bb20 4209 btrfs_end_transaction(trans);
f82a9901
FM		 4210 goto out;
4211 }
a2f392e4
NB		 4212 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
4213 last_dest_end,
4214 destoff + len - last_dest_end);
f82a9901 4215 ret = clone_finish_inode_update(trans, inode, destoff + len,
1c919a5e 4216 destoff, olen, no_time_update);
f82a9901
FM		 4217 }
4218
f46b5a66 4219out:
32b7c687 4220 btrfs_free_path(path);
15351955 4221 kvfree(buf);
32b7c687
MF		 4222 return ret;
4223}
4224
3db11b2e
ZB		 4225static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
4226 u64 off, u64 olen, u64 destoff)
32b7c687 4227{
54563d41 4228 struct inode *inode = file_inode(file);
3db11b2e 4229 struct inode *src = file_inode(file_src);
0b246afa 4230 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
32b7c687 4231 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687
MF		 4232 int ret;
4233 u64 len = olen;
0b246afa 4234 u64 bs = fs_info->sb->s_blocksize;
3db11b2e 4235 int same_inode = src == inode;
32b7c687
MF		 4236
4237 /*
4238 * TODO:
4239 * - split compressed inline extents. annoying: we need to
4240 * decompress into destination's address_space (the file offset
4241 * may change, so source mapping won't do), then recompress (or
4242 * otherwise reinsert) a subrange.
00fdf13a
LB		 4243 *
4244 * - split destination inode's inline extents. The inline extents can
4245 * be either compressed or non-compressed.
32b7c687
MF		 4246 */
4247
32b7c687
MF		 4248 if (btrfs_root_readonly(root))
4249 return -EROFS;
4250
3db11b2e
ZB		 4251 if (file_src->f_path.mnt != file->f_path.mnt ||
4252 src->i_sb != inode->i_sb)
4253 return -EXDEV;
32b7c687 4254
32b7c687 4255 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3db11b2e 4256 return -EISDIR;
32b7c687
MF		 4257
4258 if (!same_inode) {
293a8489 4259 btrfs_double_inode_lock(src, inode);
32b7c687 4260 } else {
5955102c 4261 inode_lock(src);
32b7c687
MF		 4262 }
4263
b5c40d59
OS		 4264 /* don't make the dst file partly checksummed */
4265 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
4266 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
4267 ret = -EINVAL;
4268 goto out_unlock;
4269 }
4270
32b7c687
MF		 4271 /* determine range to clone */
4272 ret = -EINVAL;
4273 if (off + len > src->i_size || off + len < off)
4274 goto out_unlock;
4275 if (len == 0)
4276 olen = len = src->i_size - off;
4277 /* if we extend to eof, continue to block boundary */
4278 if (off + len == src->i_size)
4279 len = ALIGN(src->i_size, bs) - off;
4280
ccccf3d6
FM		 4281 if (len == 0) {
4282 ret = 0;
4283 goto out_unlock;
4284 }
4285
32b7c687
MF		 4286 /* verify the end result is block aligned */
4287 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
4288 !IS_ALIGNED(destoff, bs))
4289 goto out_unlock;
4290
4291 /* verify if ranges are overlapped within the same file */
4292 if (same_inode) {
4293 if (destoff + len > off && destoff < off + len)
4294 goto out_unlock;
4295 }
4296
4297 if (destoff > inode->i_size) {
4298 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
4299 if (ret)
4300 goto out_unlock;
4301 }
4302
c125b8bf
FM		 4303 /*
4304 * Lock the target range too. Right after we replace the file extent
4305 * items in the fs tree (which now point to the cloned data), we might
4306 * have a worker replace them with extent items relative to a write
4307 * operation that was issued before this clone operation (i.e. confront
4308 * with inode.c:btrfs_finish_ordered_io).
4309 */
4310 if (same_inode) {
4311 u64 lock_start = min_t(u64, off, destoff);
4312 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
32b7c687 4313
e0bd70c6 4314 ret = lock_extent_range(src, lock_start, lock_len, true);
c125b8bf 4315 } else {
e0bd70c6
FM		 4316 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
4317 true);
4318 }
4319 ASSERT(ret == 0);
4320 if (WARN_ON(ret)) {
4321 /* ranges in the io trees already unlocked */
4322 goto out_unlock;
c125b8bf 4323 }
32b7c687 4324
1c919a5e 4325 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
32b7c687 4326
c125b8bf
FM		 4327 if (same_inode) {
4328 u64 lock_start = min_t(u64, off, destoff);
4329 u64 lock_end = max_t(u64, off, destoff) + len - 1;
4330
4331 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
4332 } else {
293a8489 4333 btrfs_double_extent_unlock(src, off, inode, destoff, len);
c125b8bf
FM		 4334 }
4335 /*
4336 * Truncate page cache pages so that future reads will see the cloned
4337 * data immediately and not the previous data.
4338 */
65bfa658 4339 truncate_inode_pages_range(&inode->i_data,
09cbfeaf
KS		 4340 round_down(destoff, PAGE_SIZE),
4341 round_up(destoff + len, PAGE_SIZE) - 1);
f46b5a66 4342out_unlock:
293a8489
MF		 4343 if (!same_inode)
4344 btrfs_double_inode_unlock(src, inode);
4345 else
5955102c 4346 inode_unlock(src);
3db11b2e
ZB		 4347 return ret;
4348}
4349
04b38d60
CH		 4350int btrfs_clone_file_range(struct file *src_file, loff_t off,
4351 struct file *dst_file, loff_t destoff, u64 len)
3db11b2e 4352{
04b38d60 4353 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
c5c9cd4d
SW		 4354}
4355
6ef5ed0d
JB		 4356static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4357{
496ad9aa 4358 struct inode *inode = file_inode(file);
0b246afa 4359 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6ef5ed0d
JB		 4360 struct btrfs_root *root = BTRFS_I(inode)->root;
4361 struct btrfs_root *new_root;
4362 struct btrfs_dir_item *di;
4363 struct btrfs_trans_handle *trans;
4364 struct btrfs_path *path;
4365 struct btrfs_key location;
4366 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 4367 u64 objectid = 0;
4368 u64 dir_id;
3c04ce01 4369 int ret;
6ef5ed0d
JB		 4370
4371 if (!capable(CAP_SYS_ADMIN))
4372 return -EPERM;
4373
3c04ce01
MX		 4374 ret = mnt_want_write_file(file);
4375 if (ret)
4376 return ret;
4377
4378 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4379 ret = -EFAULT;
4380 goto out;
4381 }
6ef5ed0d
JB		 4382
4383 if (!objectid)
1cecf579 4384 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 4385
4386 location.objectid = objectid;
4387 location.type = BTRFS_ROOT_ITEM_KEY;
4388 location.offset = (u64)-1;
4389
0b246afa 4390 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
3c04ce01
MX		 4391 if (IS_ERR(new_root)) {
4392 ret = PTR_ERR(new_root);
4393 goto out;
4394 }
4fd786e6 4395 if (!is_fstree(new_root->root_key.objectid)) {
6d6d2829 4396 ret = -ENOENT;
4397 goto out;
4398 }
6ef5ed0d 4399
6ef5ed0d 4400 path = btrfs_alloc_path();
3c04ce01
MX		 4401 if (!path) {
4402 ret = -ENOMEM;
4403 goto out;
4404 }
6ef5ed0d
JB		 4405 path->leave_spinning = 1;
4406
4407 trans = btrfs_start_transaction(root, 1);
98d5dc13 4408 if (IS_ERR(trans)) {
6ef5ed0d 4409 btrfs_free_path(path);
3c04ce01
MX		 4410 ret = PTR_ERR(trans);
4411 goto out;
6ef5ed0d
JB		 4412 }
4413
0b246afa
JM		 4414 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4415 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
6ef5ed0d 4416 dir_id, "default", 7, 1);
cf1e99a4 4417 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d 4418 btrfs_free_path(path);
3a45bb20 4419 btrfs_end_transaction(trans);
0b246afa 4420 btrfs_err(fs_info,
5d163e0e 4421 "Umm, you don't have the default diritem, this isn't going to work");
3c04ce01
MX		 4422 ret = -ENOENT;
4423 goto out;
6ef5ed0d
JB		 4424 }
4425
4426 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4427 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4428 btrfs_mark_buffer_dirty(path->nodes[0]);
4429 btrfs_free_path(path);
4430
0b246afa 4431 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3a45bb20 4432 btrfs_end_transaction(trans);
3c04ce01
MX		 4433out:
4434 mnt_drop_write_file(file);
4435 return ret;
6ef5ed0d
JB		 4436}
4437
c065f5b1
SY		 4438static void get_block_group_info(struct list_head *groups_list,
4439 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 4440{
4441 struct btrfs_block_group_cache *block_group;
4442
4443 space->total_bytes = 0;
4444 space->used_bytes = 0;
4445 space->flags = 0;
4446 list_for_each_entry(block_group, groups_list, list) {
4447 space->flags = block_group->flags;
4448 space->total_bytes += block_group->key.offset;
4449 space->used_bytes +=
4450 btrfs_block_group_used(&block_group->item);
4451 }
4452}
4453
2ff7e61e
JM		 4454static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4455 void __user *arg)
1406e432
JB		 4456{
4457 struct btrfs_ioctl_space_args space_args;
4458 struct btrfs_ioctl_space_info space;
4459 struct btrfs_ioctl_space_info *dest;
7fde62bf 4460 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 4461 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 4462 struct btrfs_space_info *info;
315d8e98
CIK		 4463 static const u64 types[] = {
4464 BTRFS_BLOCK_GROUP_DATA,
4465 BTRFS_BLOCK_GROUP_SYSTEM,
4466 BTRFS_BLOCK_GROUP_METADATA,
4467 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4468 };
bf5fc093 4469 int num_types = 4;
7fde62bf 4470 int alloc_size;
1406e432 4471 int ret = 0;
51788b1b 4472 u64 slot_count = 0;
bf5fc093 4473 int i, c;
1406e432
JB		 4474
4475 if (copy_from_user(&space_args,
4476 (struct btrfs_ioctl_space_args __user *)arg,
4477 sizeof(space_args)))
4478 return -EFAULT;
4479
bf5fc093
JB		 4480 for (i = 0; i < num_types; i++) {
4481 struct btrfs_space_info *tmp;
4482
4483 info = NULL;
4484 rcu_read_lock();
0b246afa 4485 list_for_each_entry_rcu(tmp, &fs_info->space_info,
bf5fc093
JB		 4486 list) {
4487 if (tmp->flags == types[i]) {
4488 info = tmp;
4489 break;
4490 }
4491 }
4492 rcu_read_unlock();
4493
4494 if (!info)
4495 continue;
4496
4497 down_read(&info->groups_sem);
4498 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4499 if (!list_empty(&info->block_groups[c]))
4500 slot_count++;
4501 }
4502 up_read(&info->groups_sem);
4503 }
7fde62bf 4504
36523e95
DS		 4505 /*
4506 * Global block reserve, exported as a space_info
4507 */
4508 slot_count++;
4509
7fde62bf
CM		 4510 /* space_slots == 0 means they are asking for a count */
4511 if (space_args.space_slots == 0) {
4512 space_args.total_spaces = slot_count;
4513 goto out;
4514 }
bf5fc093 4515
51788b1b 4516 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 4517
7fde62bf 4518 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 4519
7fde62bf
CM		 4520 /* we generally have at most 6 or so space infos, one for each raid
4521 * level. So, a whole page should be more than enough for everyone
4522 */
09cbfeaf 4523 if (alloc_size > PAGE_SIZE)
7fde62bf
CM		 4524 return -ENOMEM;
4525
1406e432 4526 space_args.total_spaces = 0;
8d2db785 4527 dest = kmalloc(alloc_size, GFP_KERNEL);
7fde62bf
CM		 4528 if (!dest)
4529 return -ENOMEM;
4530 dest_orig = dest;
1406e432 4531
7fde62bf 4532 /* now we have a buffer to copy into */
bf5fc093
JB		 4533 for (i = 0; i < num_types; i++) {
4534 struct btrfs_space_info *tmp;
4535
51788b1b
DR		 4536 if (!slot_count)
4537 break;
4538
bf5fc093
JB		 4539 info = NULL;
4540 rcu_read_lock();
0b246afa 4541 list_for_each_entry_rcu(tmp, &fs_info->space_info,
bf5fc093
JB		 4542 list) {
4543 if (tmp->flags == types[i]) {
4544 info = tmp;
4545 break;
4546 }
4547 }
4548 rcu_read_unlock();
7fde62bf 4549
bf5fc093
JB		 4550 if (!info)
4551 continue;
4552 down_read(&info->groups_sem);
4553 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4554 if (!list_empty(&info->block_groups[c])) {
c065f5b1
SY		 4555 get_block_group_info(&info->block_groups[c],
4556 &space);
bf5fc093
JB		 4557 memcpy(dest, &space, sizeof(space));
4558 dest++;
4559 space_args.total_spaces++;
51788b1b 4560 slot_count--;
bf5fc093 4561 }
51788b1b
DR		 4562 if (!slot_count)
4563 break;
bf5fc093
JB		 4564 }
4565 up_read(&info->groups_sem);
1406e432 4566 }
1406e432 4567
36523e95
DS		 4568 /*
4569 * Add global block reserve
4570 */
4571 if (slot_count) {
0b246afa 4572 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
36523e95
DS		 4573
4574 spin_lock(&block_rsv->lock);
4575 space.total_bytes = block_rsv->size;
4576 space.used_bytes = block_rsv->size - block_rsv->reserved;
4577 spin_unlock(&block_rsv->lock);
4578 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4579 memcpy(dest, &space, sizeof(space));
4580 space_args.total_spaces++;
4581 }
4582
2eec6c81 4583 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 4584 (arg + sizeof(struct btrfs_ioctl_space_args));
4585
4586 if (copy_to_user(user_dest, dest_orig, alloc_size))
4587 ret = -EFAULT;
4588
4589 kfree(dest_orig);
4590out:
4591 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 4592 ret = -EFAULT;
4593
4594 return ret;
4595}
4596
9a8c28be
MX		 4597static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4598 void __user *argp)
46204592 4599{
46204592
SW		 4600 struct btrfs_trans_handle *trans;
4601 u64 transid;
db5b493a 4602 int ret;
46204592 4603
d4edf39b 4604 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4605 if (IS_ERR(trans)) {
4606 if (PTR_ERR(trans) != -ENOENT)
4607 return PTR_ERR(trans);
4608
4609 /* No running transaction, don't bother */
4610 transid = root->fs_info->last_trans_committed;
4611 goto out;
4612 }
46204592 4613 transid = trans->transid;
3a45bb20 4614 ret = btrfs_commit_transaction_async(trans, 0);
8b2b2d3c 4615 if (ret) {
3a45bb20 4616 btrfs_end_transaction(trans);
db5b493a 4617 return ret;
8b2b2d3c 4618 }
ff7c1d33 4619out:
46204592
SW		 4620 if (argp)
4621 if (copy_to_user(argp, &transid, sizeof(transid)))
4622 return -EFAULT;
4623 return 0;
4624}
4625
2ff7e61e 4626static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
9a8c28be 4627 void __user *argp)
46204592 4628{
46204592
SW		 4629 u64 transid;
4630
4631 if (argp) {
4632 if (copy_from_user(&transid, argp, sizeof(transid)))
4633 return -EFAULT;
4634 } else {
4635 transid = 0; /* current trans */
4636 }
2ff7e61e 4637 return btrfs_wait_for_commit(fs_info, transid);
46204592
SW		 4638}
4639
b8e95489 4640static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4641{
0b246afa 4642 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
475f6387 4643 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4644 int ret;
475f6387
JS		 4645
4646 if (!capable(CAP_SYS_ADMIN))
4647 return -EPERM;
4648
4649 sa = memdup_user(arg, sizeof(*sa));
4650 if (IS_ERR(sa))
4651 return PTR_ERR(sa);
4652
b8e95489
MX		 4653 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4654 ret = mnt_want_write_file(file);
4655 if (ret)
4656 goto out;
4657 }
4658
0b246afa 4659 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4660 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4661 0);
475f6387
JS		 4662
4663 if (copy_to_user(arg, sa, sizeof(*sa)))
4664 ret = -EFAULT;
4665
b8e95489
MX		 4666 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4667 mnt_drop_write_file(file);
4668out:
475f6387
JS		 4669 kfree(sa);
4670 return ret;
4671}
4672
2ff7e61e 4673static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
475f6387
JS		 4674{
4675 if (!capable(CAP_SYS_ADMIN))
4676 return -EPERM;
4677
2ff7e61e 4678 return btrfs_scrub_cancel(fs_info);
475f6387
JS		 4679}
4680
2ff7e61e 4681static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
475f6387
JS		 4682 void __user *arg)
4683{
4684 struct btrfs_ioctl_scrub_args *sa;
4685 int ret;
4686
4687 if (!capable(CAP_SYS_ADMIN))
4688 return -EPERM;
4689
4690 sa = memdup_user(arg, sizeof(*sa));
4691 if (IS_ERR(sa))
4692 return PTR_ERR(sa);
4693
2ff7e61e 4694 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
475f6387
JS		 4695
4696 if (copy_to_user(arg, sa, sizeof(*sa)))
4697 ret = -EFAULT;
4698
4699 kfree(sa);
4700 return ret;
4701}
4702
2ff7e61e 4703static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
b27f7c0c 4704 void __user *arg)
c11d2c23
SB		 4705{
4706 struct btrfs_ioctl_get_dev_stats *sa;
4707 int ret;
4708
c11d2c23
SB		 4709 sa = memdup_user(arg, sizeof(*sa));
4710 if (IS_ERR(sa))
4711 return PTR_ERR(sa);
4712
b27f7c0c
DS		 4713 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4714 kfree(sa);
4715 return -EPERM;
4716 }
4717
2ff7e61e 4718 ret = btrfs_get_dev_stats(fs_info, sa);
c11d2c23
SB		 4719
4720 if (copy_to_user(arg, sa, sizeof(*sa)))
4721 ret = -EFAULT;
4722
4723 kfree(sa);
4724 return ret;
4725}
4726
2ff7e61e
JM		 4727static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4728 void __user *arg)
3f6bcfbd
SB		 4729{
4730 struct btrfs_ioctl_dev_replace_args *p;
4731 int ret;
4732
4733 if (!capable(CAP_SYS_ADMIN))
4734 return -EPERM;
4735
4736 p = memdup_user(arg, sizeof(*p));
4737 if (IS_ERR(p))
4738 return PTR_ERR(p);
4739
4740 switch (p->cmd) {
4741 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
bc98a42c 4742 if (sb_rdonly(fs_info->sb)) {
adfa97cb
ID		 4743 ret = -EROFS;
4744 goto out;
4745 }
171938e5 4746 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
e57138b3 4747 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd 4748 } else {
2ff7e61e 4749 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
171938e5 4750 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3f6bcfbd
SB		 4751 }
4752 break;
4753 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
0b246afa 4754 btrfs_dev_replace_status(fs_info, p);
3f6bcfbd
SB		 4755 ret = 0;
4756 break;
4757 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
17d202b9 4758 p->result = btrfs_dev_replace_cancel(fs_info);
97282031 4759 ret = 0;
3f6bcfbd
SB		 4760 break;
4761 default:
4762 ret = -EINVAL;
4763 break;
4764 }
4765
4766 if (copy_to_user(arg, p, sizeof(*p)))
4767 ret = -EFAULT;
adfa97cb 4768out:
3f6bcfbd
SB		 4769 kfree(p);
4770 return ret;
4771}
4772
d7728c96
JS		 4773static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4774{
4775 int ret = 0;
4776 int i;
740c3d22 4777 u64 rel_ptr;
d7728c96 4778 int size;
806468f8 4779 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4780 struct inode_fs_paths *ipath = NULL;
4781 struct btrfs_path *path;
4782
82b22ac8 4783 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4784 return -EPERM;
4785
4786 path = btrfs_alloc_path();
4787 if (!path) {
4788 ret = -ENOMEM;
4789 goto out;
4790 }
4791
4792 ipa = memdup_user(arg, sizeof(*ipa));
4793 if (IS_ERR(ipa)) {
4794 ret = PTR_ERR(ipa);
4795 ipa = NULL;
4796 goto out;
4797 }
4798
4799 size = min_t(u32, ipa->size, 4096);
4800 ipath = init_ipath(size, root, path);
4801 if (IS_ERR(ipath)) {
4802 ret = PTR_ERR(ipath);
4803 ipath = NULL;
4804 goto out;
4805 }
4806
4807 ret = paths_from_inode(ipa->inum, ipath);
4808 if (ret < 0)
4809 goto out;
4810
4811 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4812 rel_ptr = ipath->fspath->val[i] -
4813 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4814 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4815 }
4816
718dc5fa
OS		 4817 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4818 ipath->fspath, size);
d7728c96
JS		 4819 if (ret) {
4820 ret = -EFAULT;
4821 goto out;
4822 }
4823
4824out:
4825 btrfs_free_path(path);
4826 free_ipath(ipath);
4827 kfree(ipa);
4828
4829 return ret;
4830}
4831
4832static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4833{
4834 struct btrfs_data_container *inodes = ctx;
4835 const size_t c = 3 * sizeof(u64);
4836
4837 if (inodes->bytes_left >= c) {
4838 inodes->bytes_left -= c;
4839 inodes->val[inodes->elem_cnt] = inum;
4840 inodes->val[inodes->elem_cnt + 1] = offset;
4841 inodes->val[inodes->elem_cnt + 2] = root;
4842 inodes->elem_cnt += 3;
4843 } else {
4844 inodes->bytes_missing += c - inodes->bytes_left;
4845 inodes->bytes_left = 0;
4846 inodes->elem_missed += 3;
4847 }
4848
4849 return 0;
4850}
4851
2ff7e61e 4852static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
d24a67b2 4853 void __user *arg, int version)
d7728c96
JS		 4854{
4855 int ret = 0;
4856 int size;
d7728c96
JS		 4857 struct btrfs_ioctl_logical_ino_args *loi;
4858 struct btrfs_data_container *inodes = NULL;
4859 struct btrfs_path *path = NULL;
d24a67b2 4860 bool ignore_offset;
d7728c96
JS		 4861
4862 if (!capable(CAP_SYS_ADMIN))
4863 return -EPERM;
4864
4865 loi = memdup_user(arg, sizeof(*loi));
7b9ea627
SV		 4866 if (IS_ERR(loi))
4867 return PTR_ERR(loi);
d7728c96 4868
d24a67b2
ZB		 4869 if (version == 1) {
4870 ignore_offset = false;
b115e3bc 4871 size = min_t(u32, loi->size, SZ_64K);
d24a67b2
ZB		 4872 } else {
4873 /* All reserved bits must be 0 for now */
4874 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4875 ret = -EINVAL;
4876 goto out_loi;
4877 }
4878 /* Only accept flags we have defined so far */
4879 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4880 ret = -EINVAL;
4881 goto out_loi;
4882 }
4883 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
b115e3bc 4884 size = min_t(u32, loi->size, SZ_16M);
d24a67b2
ZB		 4885 }
4886
d7728c96
JS		 4887 path = btrfs_alloc_path();
4888 if (!path) {
4889 ret = -ENOMEM;
4890 goto out;
4891 }
4892
d7728c96
JS		 4893 inodes = init_data_container(size);
4894 if (IS_ERR(inodes)) {
4895 ret = PTR_ERR(inodes);
4896 inodes = NULL;
4897 goto out;
4898 }
4899
2ff7e61e 4900 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
d24a67b2 4901 build_ino_list, inodes, ignore_offset);
df031f07 4902 if (ret == -EINVAL)
d7728c96
JS		 4903 ret = -ENOENT;
4904 if (ret < 0)
4905 goto out;
4906
718dc5fa
OS		 4907 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4908 size);
d7728c96
JS		 4909 if (ret)
4910 ret = -EFAULT;
4911
4912out:
4913 btrfs_free_path(path);
f54de068 4914 kvfree(inodes);
d24a67b2 4915out_loi:
d7728c96
JS		 4916 kfree(loi);
4917
4918 return ret;
4919}
4920
008ef096 4921void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
c9e9f97b
ID		 4922 struct btrfs_ioctl_balance_args *bargs)
4923{
4924 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4925
4926 bargs->flags = bctl->flags;
4927
3009a62f 4928 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
837d5b6e
ID		 4929 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4930 if (atomic_read(&fs_info->balance_pause_req))
4931 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4932 if (atomic_read(&fs_info->balance_cancel_req))
4933 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4934
c9e9f97b
ID		 4935 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4936 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4937 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce 4938
008ef096
DS		 4939 spin_lock(&fs_info->balance_lock);
4940 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4941 spin_unlock(&fs_info->balance_lock);
c9e9f97b
ID		 4942}
4943
9ba1f6e4 4944static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4945{
496ad9aa 4946 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4947 struct btrfs_fs_info *fs_info = root->fs_info;
4948 struct btrfs_ioctl_balance_args *bargs;
4949 struct btrfs_balance_control *bctl;
ed0fb78f 4950 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4951 int ret;
4952
4953 if (!capable(CAP_SYS_ADMIN))
4954 return -EPERM;
4955
e54bfa31 4956 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4957 if (ret)
4958 return ret;
4959
ed0fb78f 4960again:
171938e5 4961 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
ed0fb78f
ID		 4962 mutex_lock(&fs_info->balance_mutex);
4963 need_unlock = true;
4964 goto locked;
4965 }
4966
4967 /*
01327610 4968 * mut. excl. ops lock is locked. Three possibilities:
ed0fb78f
ID		 4969 * (1) some other op is running
4970 * (2) balance is running
4971 * (3) balance is paused -- special case (think resume)
4972 */
c9e9f97b 4973 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4974 if (fs_info->balance_ctl) {
4975 /* this is either (2) or (3) */
3009a62f 4976 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
ed0fb78f 4977 mutex_unlock(&fs_info->balance_mutex);
dccdb07b
DS		 4978 /*
4979 * Lock released to allow other waiters to continue,
4980 * we'll reexamine the status again.
4981 */
ed0fb78f
ID		 4982 mutex_lock(&fs_info->balance_mutex);
4983
4984 if (fs_info->balance_ctl &&
3009a62f 4985 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
ed0fb78f
ID		 4986 /* this is (3) */
4987 need_unlock = false;
4988 goto locked;
4989 }
4990
4991 mutex_unlock(&fs_info->balance_mutex);
ed0fb78f
ID		 4992 goto again;
4993 } else {
4994 /* this is (2) */
4995 mutex_unlock(&fs_info->balance_mutex);
4996 ret = -EINPROGRESS;
4997 goto out;
4998 }
4999 } else {
5000 /* this is (1) */
5001 mutex_unlock(&fs_info->balance_mutex);
e57138b3 5002 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 5003 goto out;
5004 }
5005
5006locked:
171938e5 5007 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
c9e9f97b
ID		 5008
5009 if (arg) {
5010 bargs = memdup_user(arg, sizeof(*bargs));
5011 if (IS_ERR(bargs)) {
5012 ret = PTR_ERR(bargs);
ed0fb78f 5013 goto out_unlock;
c9e9f97b 5014 }
de322263
ID		 5015
5016 if (bargs->flags & BTRFS_BALANCE_RESUME) {
5017 if (!fs_info->balance_ctl) {
5018 ret = -ENOTCONN;
5019 goto out_bargs;
5020 }
5021
5022 bctl = fs_info->balance_ctl;
5023 spin_lock(&fs_info->balance_lock);
5024 bctl->flags |= BTRFS_BALANCE_RESUME;
5025 spin_unlock(&fs_info->balance_lock);
5026
5027 goto do_balance;
5028 }
c9e9f97b
ID		 5029 } else {
5030 bargs = NULL;
5031 }
5032
ed0fb78f 5033 if (fs_info->balance_ctl) {
837d5b6e
ID		 5034 ret = -EINPROGRESS;
5035 goto out_bargs;
5036 }
5037
8d2db785 5038 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
c9e9f97b
ID		 5039 if (!bctl) {
5040 ret = -ENOMEM;
5041 goto out_bargs;
5042 }
5043
c9e9f97b
ID		 5044 if (arg) {
5045 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
5046 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
5047 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
5048
5049 bctl->flags = bargs->flags;
f43ffb60
ID		 5050 } else {
5051 /* balance everything - no filters */
5052 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 5053 }
5054
8eb93459
DS		 5055 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
5056 ret = -EINVAL;
0f89abf5 5057 goto out_bctl;
8eb93459
DS		 5058 }
5059
de322263 5060do_balance:
c9e9f97b 5061 /*
149196a2
DS		 5062 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
5063 * btrfs_balance. bctl is freed in reset_balance_state, or, if
5064 * restriper was paused all the way until unmount, in free_fs_info.
5065 * The flag should be cleared after reset_balance_state.
c9e9f97b 5066 */
ed0fb78f
ID		 5067 need_unlock = false;
5068
6fcf6e2b 5069 ret = btrfs_balance(fs_info, bctl, bargs);
0f89abf5 5070 bctl = NULL;
ed0fb78f 5071
c9e9f97b
ID		 5072 if (arg) {
5073 if (copy_to_user(arg, bargs, sizeof(*bargs)))
5074 ret = -EFAULT;
5075 }
5076
0f89abf5
CE		 5077out_bctl:
5078 kfree(bctl);
c9e9f97b
ID		 5079out_bargs:
5080 kfree(bargs);
ed0fb78f 5081out_unlock:
c9e9f97b 5082 mutex_unlock(&fs_info->balance_mutex);
ed0fb78f 5083 if (need_unlock)
171938e5 5084 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
ed0fb78f 5085out:
e54bfa31 5086 mnt_drop_write_file(file);
c9e9f97b
ID		 5087 return ret;
5088}
5089
2ff7e61e 5090static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
837d5b6e
ID		 5091{
5092 if (!capable(CAP_SYS_ADMIN))
5093 return -EPERM;
5094
5095 switch (cmd) {
5096 case BTRFS_BALANCE_CTL_PAUSE:
0b246afa 5097 return btrfs_pause_balance(fs_info);
a7e99c69 5098 case BTRFS_BALANCE_CTL_CANCEL:
0b246afa 5099 return btrfs_cancel_balance(fs_info);
837d5b6e
ID		 5100 }
5101
5102 return -EINVAL;
5103}
5104
2ff7e61e 5105static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
19a39dce
ID		 5106 void __user *arg)
5107{
19a39dce
ID		 5108 struct btrfs_ioctl_balance_args *bargs;
5109 int ret = 0;
5110
5111 if (!capable(CAP_SYS_ADMIN))
5112 return -EPERM;
5113
5114 mutex_lock(&fs_info->balance_mutex);
5115 if (!fs_info->balance_ctl) {
5116 ret = -ENOTCONN;
5117 goto out;
5118 }
5119
8d2db785 5120 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
19a39dce
ID		 5121 if (!bargs) {
5122 ret = -ENOMEM;
5123 goto out;
5124 }
5125
008ef096 5126 btrfs_update_ioctl_balance_args(fs_info, bargs);
19a39dce
ID		 5127
5128 if (copy_to_user(arg, bargs, sizeof(*bargs)))
5129 ret = -EFAULT;
5130
5131 kfree(bargs);
5132out:
5133 mutex_unlock(&fs_info->balance_mutex);
5134 return ret;
5135}
5136
905b0dda 5137static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 5138{
0b246afa
JM		 5139 struct inode *inode = file_inode(file);
5140 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5d13a37b 5141 struct btrfs_ioctl_quota_ctl_args *sa;
5d13a37b 5142 int ret;
5d13a37b
AJ		 5143
5144 if (!capable(CAP_SYS_ADMIN))
5145 return -EPERM;
5146
905b0dda
MX		 5147 ret = mnt_want_write_file(file);
5148 if (ret)
5149 return ret;
5d13a37b
AJ		 5150
5151 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 5152 if (IS_ERR(sa)) {
5153 ret = PTR_ERR(sa);
5154 goto drop_write;
5155 }
5d13a37b 5156
0b246afa 5157 down_write(&fs_info->subvol_sem);
5d13a37b
AJ		 5158
5159 switch (sa->cmd) {
5160 case BTRFS_QUOTA_CTL_ENABLE:
340f1aa2 5161 ret = btrfs_quota_enable(fs_info);
5d13a37b
AJ		 5162 break;
5163 case BTRFS_QUOTA_CTL_DISABLE:
340f1aa2 5164 ret = btrfs_quota_disable(fs_info);
5d13a37b 5165 break;
5d13a37b
AJ		 5166 default:
5167 ret = -EINVAL;
5168 break;
5169 }
5170
5d13a37b 5171 kfree(sa);
0b246afa 5172 up_write(&fs_info->subvol_sem);
905b0dda
MX		 5173drop_write:
5174 mnt_drop_write_file(file);
5d13a37b
AJ		 5175 return ret;
5176}
5177
905b0dda 5178static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 5179{
0b246afa
JM		 5180 struct inode *inode = file_inode(file);
5181 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5182 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 5183 struct btrfs_ioctl_qgroup_assign_args *sa;
5184 struct btrfs_trans_handle *trans;
5185 int ret;
5186 int err;
5187
5188 if (!capable(CAP_SYS_ADMIN))
5189 return -EPERM;
5190
905b0dda
MX		 5191 ret = mnt_want_write_file(file);
5192 if (ret)
5193 return ret;
5d13a37b
AJ		 5194
5195 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 5196 if (IS_ERR(sa)) {
5197 ret = PTR_ERR(sa);
5198 goto drop_write;
5199 }
5d13a37b
AJ		 5200
5201 trans = btrfs_join_transaction(root);
5202 if (IS_ERR(trans)) {
5203 ret = PTR_ERR(trans);
5204 goto out;
5205 }
5206
5d13a37b 5207 if (sa->assign) {
9f8a6ce6 5208 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
5d13a37b 5209 } else {
39616c27 5210 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
5d13a37b
AJ		 5211 }
5212
e082f563 5213 /* update qgroup status and info */
280f8bd2 5214 err = btrfs_run_qgroups(trans);
e082f563 5215 if (err < 0)
0b246afa
JM		 5216 btrfs_handle_fs_error(fs_info, err,
5217 "failed to update qgroup status and info");
3a45bb20 5218 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 5219 if (err && !ret)
5220 ret = err;
5221
5222out:
5223 kfree(sa);
905b0dda
MX		 5224drop_write:
5225 mnt_drop_write_file(file);
5d13a37b
AJ		 5226 return ret;
5227}
5228
905b0dda 5229static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 5230{
0b246afa 5231 struct inode *inode = file_inode(file);
0b246afa 5232 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 5233 struct btrfs_ioctl_qgroup_create_args *sa;
5234 struct btrfs_trans_handle *trans;
5235 int ret;
5236 int err;
5237
5238 if (!capable(CAP_SYS_ADMIN))
5239 return -EPERM;
5240
905b0dda
MX		 5241 ret = mnt_want_write_file(file);
5242 if (ret)
5243 return ret;
5d13a37b
AJ		 5244
5245 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 5246 if (IS_ERR(sa)) {
5247 ret = PTR_ERR(sa);
5248 goto drop_write;
5249 }
5d13a37b 5250
d86e56cf
MX		 5251 if (!sa->qgroupid) {
5252 ret = -EINVAL;
5253 goto out;
5254 }
5255
5d13a37b
AJ		 5256 trans = btrfs_join_transaction(root);
5257 if (IS_ERR(trans)) {
5258 ret = PTR_ERR(trans);
5259 goto out;
5260 }
5261
5d13a37b 5262 if (sa->create) {
49a05ecd 5263 ret = btrfs_create_qgroup(trans, sa->qgroupid);
5d13a37b 5264 } else {
3efbee1d 5265 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
5d13a37b
AJ		 5266 }
5267
3a45bb20 5268 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 5269 if (err && !ret)
5270 ret = err;
5271
5272out:
5273 kfree(sa);
905b0dda
MX		 5274drop_write:
5275 mnt_drop_write_file(file);
5d13a37b
AJ		 5276 return ret;
5277}
5278
905b0dda 5279static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 5280{
0b246afa 5281 struct inode *inode = file_inode(file);
0b246afa 5282 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 5283 struct btrfs_ioctl_qgroup_limit_args *sa;
5284 struct btrfs_trans_handle *trans;
5285 int ret;
5286 int err;
5287 u64 qgroupid;
5288
5289 if (!capable(CAP_SYS_ADMIN))
5290 return -EPERM;
5291
905b0dda
MX		 5292 ret = mnt_want_write_file(file);
5293 if (ret)
5294 return ret;
5d13a37b
AJ		 5295
5296 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 5297 if (IS_ERR(sa)) {
5298 ret = PTR_ERR(sa);
5299 goto drop_write;
5300 }
5d13a37b
AJ		 5301
5302 trans = btrfs_join_transaction(root);
5303 if (IS_ERR(trans)) {
5304 ret = PTR_ERR(trans);
5305 goto out;
5306 }
5307
5308 qgroupid = sa->qgroupid;
5309 if (!qgroupid) {
5310 /* take the current subvol as qgroup */
5311 qgroupid = root->root_key.objectid;
5312 }
5313
f0042d5e 5314 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
5d13a37b 5315
3a45bb20 5316 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 5317 if (err && !ret)
5318 ret = err;
5319
5320out:
5321 kfree(sa);
905b0dda
MX		 5322drop_write:
5323 mnt_drop_write_file(file);
5d13a37b
AJ		 5324 return ret;
5325}
5326
2f232036
JS		 5327static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5328{
0b246afa
JM		 5329 struct inode *inode = file_inode(file);
5330 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2f232036
JS		 5331 struct btrfs_ioctl_quota_rescan_args *qsa;
5332 int ret;
5333
5334 if (!capable(CAP_SYS_ADMIN))
5335 return -EPERM;
5336
5337 ret = mnt_want_write_file(file);
5338 if (ret)
5339 return ret;
5340
5341 qsa = memdup_user(arg, sizeof(*qsa));
5342 if (IS_ERR(qsa)) {
5343 ret = PTR_ERR(qsa);
5344 goto drop_write;
5345 }
5346
5347 if (qsa->flags) {
5348 ret = -EINVAL;
5349 goto out;
5350 }
5351
0b246afa 5352 ret = btrfs_qgroup_rescan(fs_info);
2f232036
JS		 5353
5354out:
5355 kfree(qsa);
5356drop_write:
5357 mnt_drop_write_file(file);
5358 return ret;
5359}
5360
5361static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5362{
0b246afa
JM		 5363 struct inode *inode = file_inode(file);
5364 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2f232036
JS		 5365 struct btrfs_ioctl_quota_rescan_args *qsa;
5366 int ret = 0;
5367
5368 if (!capable(CAP_SYS_ADMIN))
5369 return -EPERM;
5370
8d2db785 5371 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
2f232036
JS		 5372 if (!qsa)
5373 return -ENOMEM;
5374
0b246afa 5375 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
2f232036 5376 qsa->flags = 1;
0b246afa 5377 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
2f232036
JS		 5378 }
5379
5380 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5381 ret = -EFAULT;
5382
5383 kfree(qsa);
5384 return ret;
5385}
5386
57254b6e
JS		 5387static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5388{
0b246afa
JM		 5389 struct inode *inode = file_inode(file);
5390 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
57254b6e
JS		 5391
5392 if (!capable(CAP_SYS_ADMIN))
5393 return -EPERM;
5394
0b246afa 5395 return btrfs_qgroup_wait_for_completion(fs_info, true);
57254b6e
JS		 5396}
5397
abccd00f
HM		 5398static long _btrfs_ioctl_set_received_subvol(struct file *file,
5399 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 5400{
496ad9aa 5401 struct inode *inode = file_inode(file);
0b246afa 5402 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
8ea05e3a
AB		 5403 struct btrfs_root *root = BTRFS_I(inode)->root;
5404 struct btrfs_root_item *root_item = &root->root_item;
5405 struct btrfs_trans_handle *trans;
95582b00 5406 struct timespec64 ct = current_time(inode);
8ea05e3a 5407 int ret = 0;
dd5f9615 5408 int received_uuid_changed;
8ea05e3a 5409
bd60ea0f
DS		 5410 if (!inode_owner_or_capable(inode))
5411 return -EPERM;
5412
8ea05e3a
AB		 5413 ret = mnt_want_write_file(file);
5414 if (ret < 0)
5415 return ret;
5416
0b246afa 5417 down_write(&fs_info->subvol_sem);
8ea05e3a 5418
4a0cc7ca 5419 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
8ea05e3a
AB		 5420 ret = -EINVAL;
5421 goto out;
5422 }
5423
5424 if (btrfs_root_readonly(root)) {
5425 ret = -EROFS;
5426 goto out;
5427 }
5428
dd5f9615
SB		 5429 /*
5430 * 1 - root item
5431 * 2 - uuid items (received uuid + subvol uuid)
5432 */
5433 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 5434 if (IS_ERR(trans)) {
5435 ret = PTR_ERR(trans);
5436 trans = NULL;
5437 goto out;
5438 }
5439
5440 sa->rtransid = trans->transid;
5441 sa->rtime.sec = ct.tv_sec;
5442 sa->rtime.nsec = ct.tv_nsec;
5443
dd5f9615
SB		 5444 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5445 BTRFS_UUID_SIZE);
5446 if (received_uuid_changed &&
d87ff758 5447 !btrfs_is_empty_uuid(root_item->received_uuid)) {
d1957791 5448 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
d87ff758
NB		 5449 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5450 root->root_key.objectid);
5451 if (ret && ret != -ENOENT) {
5452 btrfs_abort_transaction(trans, ret);
5453 btrfs_end_transaction(trans);
5454 goto out;
5455 }
5456 }
8ea05e3a
AB		 5457 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5458 btrfs_set_root_stransid(root_item, sa->stransid);
5459 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 5460 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5461 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5462 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5463 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a 5464
0b246afa 5465 ret = btrfs_update_root(trans, fs_info->tree_root,
8ea05e3a
AB		 5466 &root->root_key, &root->root_item);
5467 if (ret < 0) {
3a45bb20 5468 btrfs_end_transaction(trans);
8ea05e3a 5469 goto out;
dd5f9615
SB		 5470 }
5471 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
cdb345a8 5472 ret = btrfs_uuid_tree_add(trans, sa->uuid,
dd5f9615
SB		 5473 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5474 root->root_key.objectid);
5475 if (ret < 0 && ret != -EEXIST) {
66642832 5476 btrfs_abort_transaction(trans, ret);
efd38150 5477 btrfs_end_transaction(trans);
8ea05e3a 5478 goto out;
dd5f9615
SB		 5479 }
5480 }
3a45bb20 5481 ret = btrfs_commit_transaction(trans);
abccd00f 5482out:
0b246afa 5483 up_write(&fs_info->subvol_sem);
abccd00f
HM		 5484 mnt_drop_write_file(file);
5485 return ret;
5486}
5487
5488#ifdef CONFIG_64BIT
5489static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5490 void __user *arg)
5491{
5492 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5493 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5494 int ret = 0;
5495
5496 args32 = memdup_user(arg, sizeof(*args32));
7b9ea627
SV		 5497 if (IS_ERR(args32))
5498 return PTR_ERR(args32);
abccd00f 5499
8d2db785 5500 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
84dbeb87
DC		 5501 if (!args64) {
5502 ret = -ENOMEM;
abccd00f
HM		 5503 goto out;
5504 }
5505
5506 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5507 args64->stransid = args32->stransid;
5508 args64->rtransid = args32->rtransid;
5509 args64->stime.sec = args32->stime.sec;
5510 args64->stime.nsec = args32->stime.nsec;
5511 args64->rtime.sec = args32->rtime.sec;
5512 args64->rtime.nsec = args32->rtime.nsec;
5513 args64->flags = args32->flags;
5514
5515 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5516 if (ret)
5517 goto out;
5518
5519 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5520 args32->stransid = args64->stransid;
5521 args32->rtransid = args64->rtransid;
5522 args32->stime.sec = args64->stime.sec;
5523 args32->stime.nsec = args64->stime.nsec;
5524 args32->rtime.sec = args64->rtime.sec;
5525 args32->rtime.nsec = args64->rtime.nsec;
5526 args32->flags = args64->flags;
5527
5528 ret = copy_to_user(arg, args32, sizeof(*args32));
5529 if (ret)
5530 ret = -EFAULT;
5531
5532out:
5533 kfree(args32);
5534 kfree(args64);
5535 return ret;
5536}
5537#endif
5538
5539static long btrfs_ioctl_set_received_subvol(struct file *file,
5540 void __user *arg)
5541{
5542 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5543 int ret = 0;
5544
5545 sa = memdup_user(arg, sizeof(*sa));
7b9ea627
SV		 5546 if (IS_ERR(sa))
5547 return PTR_ERR(sa);
abccd00f
HM		 5548
5549 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5550
5551 if (ret)
5552 goto out;
5553
8ea05e3a
AB		 5554 ret = copy_to_user(arg, sa, sizeof(*sa));
5555 if (ret)
5556 ret = -EFAULT;
5557
5558out:
5559 kfree(sa);
8ea05e3a
AB		 5560 return ret;
5561}
5562
867ab667 5563static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5564{
0b246afa
JM		 5565 struct inode *inode = file_inode(file);
5566 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
a1b83ac5 5567 size_t len;
867ab667 5568 int ret;
a1b83ac5
AJ		 5569 char label[BTRFS_LABEL_SIZE];
5570
0b246afa
JM		 5571 spin_lock(&fs_info->super_lock);
5572 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5573 spin_unlock(&fs_info->super_lock);
a1b83ac5
AJ		 5574
5575 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5576
5577 if (len == BTRFS_LABEL_SIZE) {
0b246afa
JM		 5578 btrfs_warn(fs_info,
5579 "label is too long, return the first %zu bytes",
5580 --len);
867ab667 5581 }
5582
867ab667 5583 ret = copy_to_user(arg, label, len);
867ab667 5584
5585 return ret ? -EFAULT : 0;
5586}
5587
a8bfd4ab 5588static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5589{
0b246afa
JM		 5590 struct inode *inode = file_inode(file);
5591 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5592 struct btrfs_root *root = BTRFS_I(inode)->root;
5593 struct btrfs_super_block *super_block = fs_info->super_copy;
a8bfd4ab 5594 struct btrfs_trans_handle *trans;
5595 char label[BTRFS_LABEL_SIZE];
5596 int ret;
5597
5598 if (!capable(CAP_SYS_ADMIN))
5599 return -EPERM;
5600
5601 if (copy_from_user(label, arg, sizeof(label)))
5602 return -EFAULT;
5603
5604 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
0b246afa 5605 btrfs_err(fs_info,
5d163e0e
JM		 5606 "unable to set label with more than %d bytes",
5607 BTRFS_LABEL_SIZE - 1);
a8bfd4ab 5608 return -EINVAL;
5609 }
5610
5611 ret = mnt_want_write_file(file);
5612 if (ret)
5613 return ret;
5614
a8bfd4ab 5615 trans = btrfs_start_transaction(root, 0);
5616 if (IS_ERR(trans)) {
5617 ret = PTR_ERR(trans);
5618 goto out_unlock;
5619 }
5620
0b246afa 5621 spin_lock(&fs_info->super_lock);
a8bfd4ab 5622 strcpy(super_block->label, label);
0b246afa 5623 spin_unlock(&fs_info->super_lock);
3a45bb20 5624 ret = btrfs_commit_transaction(trans);
a8bfd4ab 5625
5626out_unlock:
a8bfd4ab 5627 mnt_drop_write_file(file);
5628 return ret;
5629}
5630
2eaa055f
JM		 5631#define INIT_FEATURE_FLAGS(suffix) \
5632 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5633 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5634 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5635
d5131b65 5636int btrfs_ioctl_get_supported_features(void __user *arg)
2eaa055f 5637{
4d4ab6d6 5638 static const struct btrfs_ioctl_feature_flags features[3] = {
2eaa055f
JM		 5639 INIT_FEATURE_FLAGS(SUPP),
5640 INIT_FEATURE_FLAGS(SAFE_SET),
5641 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5642 };
5643
5644 if (copy_to_user(arg, &features, sizeof(features)))
5645 return -EFAULT;
5646
5647 return 0;
5648}
5649
5650static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5651{
0b246afa
JM		 5652 struct inode *inode = file_inode(file);
5653 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5654 struct btrfs_super_block *super_block = fs_info->super_copy;
2eaa055f
JM		 5655 struct btrfs_ioctl_feature_flags features;
5656
5657 features.compat_flags = btrfs_super_compat_flags(super_block);
5658 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5659 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5660
5661 if (copy_to_user(arg, &features, sizeof(features)))
5662 return -EFAULT;
5663
5664 return 0;
5665}
5666
2ff7e61e 5667static int check_feature_bits(struct btrfs_fs_info *fs_info,
3b02a68a 5668 enum btrfs_feature_set set,
2eaa055f
JM		 5669 u64 change_mask, u64 flags, u64 supported_flags,
5670 u64 safe_set, u64 safe_clear)
5671{
3b02a68a
JM		 5672 const char *type = btrfs_feature_set_names[set];
5673 char *names;
2eaa055f
JM		 5674 u64 disallowed, unsupported;
5675 u64 set_mask = flags & change_mask;
5676 u64 clear_mask = ~flags & change_mask;
5677
5678 unsupported = set_mask & ~supported_flags;
5679 if (unsupported) {
3b02a68a
JM		 5680 names = btrfs_printable_features(set, unsupported);
5681 if (names) {
0b246afa
JM		 5682 btrfs_warn(fs_info,
5683 "this kernel does not support the %s feature bit%s",
5684 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5685 kfree(names);
5686 } else
0b246afa
JM		 5687 btrfs_warn(fs_info,
5688 "this kernel does not support %s bits 0x%llx",
5689 type, unsupported);
2eaa055f
JM		 5690 return -EOPNOTSUPP;
5691 }
5692
5693 disallowed = set_mask & ~safe_set;
5694 if (disallowed) {
3b02a68a
JM		 5695 names = btrfs_printable_features(set, disallowed);
5696 if (names) {
0b246afa
JM		 5697 btrfs_warn(fs_info,
5698 "can't set the %s feature bit%s while mounted",
5699 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5700 kfree(names);
5701 } else
0b246afa
JM		 5702 btrfs_warn(fs_info,
5703 "can't set %s bits 0x%llx while mounted",
5704 type, disallowed);
2eaa055f
JM		 5705 return -EPERM;
5706 }
5707
5708 disallowed = clear_mask & ~safe_clear;
5709 if (disallowed) {
3b02a68a
JM		 5710 names = btrfs_printable_features(set, disallowed);
5711 if (names) {
0b246afa
JM		 5712 btrfs_warn(fs_info,
5713 "can't clear the %s feature bit%s while mounted",
5714 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5715 kfree(names);
5716 } else
0b246afa
JM		 5717 btrfs_warn(fs_info,
5718 "can't clear %s bits 0x%llx while mounted",
5719 type, disallowed);
2eaa055f
JM		 5720 return -EPERM;
5721 }
5722
5723 return 0;
5724}
5725
2ff7e61e
JM		 5726#define check_feature(fs_info, change_mask, flags, mask_base) \
5727check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5728 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5729 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5730 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5731
5732static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5733{
0b246afa
JM		 5734 struct inode *inode = file_inode(file);
5735 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5736 struct btrfs_root *root = BTRFS_I(inode)->root;
5737 struct btrfs_super_block *super_block = fs_info->super_copy;
2eaa055f
JM		 5738 struct btrfs_ioctl_feature_flags flags[2];
5739 struct btrfs_trans_handle *trans;
5740 u64 newflags;
5741 int ret;
5742
5743 if (!capable(CAP_SYS_ADMIN))
5744 return -EPERM;
5745
5746 if (copy_from_user(flags, arg, sizeof(flags)))
5747 return -EFAULT;
5748
5749 /* Nothing to do */
5750 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5751 !flags[0].incompat_flags)
5752 return 0;
5753
2ff7e61e 5754 ret = check_feature(fs_info, flags[0].compat_flags,
2eaa055f
JM		 5755 flags[1].compat_flags, COMPAT);
5756 if (ret)
5757 return ret;
5758
2ff7e61e 5759 ret = check_feature(fs_info, flags[0].compat_ro_flags,
2eaa055f
JM		 5760 flags[1].compat_ro_flags, COMPAT_RO);
5761 if (ret)
5762 return ret;
5763
2ff7e61e 5764 ret = check_feature(fs_info, flags[0].incompat_flags,
2eaa055f
JM		 5765 flags[1].incompat_flags, INCOMPAT);
5766 if (ret)
5767 return ret;
5768
7ab19625
DS		 5769 ret = mnt_want_write_file(file);
5770 if (ret)
5771 return ret;
5772
8051aa1a 5773 trans = btrfs_start_transaction(root, 0);
7ab19625
DS		 5774 if (IS_ERR(trans)) {
5775 ret = PTR_ERR(trans);
5776 goto out_drop_write;
5777 }
2eaa055f 5778
0b246afa 5779 spin_lock(&fs_info->super_lock);
2eaa055f
JM		 5780 newflags = btrfs_super_compat_flags(super_block);
5781 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5782 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5783 btrfs_set_super_compat_flags(super_block, newflags);
5784
5785 newflags = btrfs_super_compat_ro_flags(super_block);
5786 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5787 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5788 btrfs_set_super_compat_ro_flags(super_block, newflags);
5789
5790 newflags = btrfs_super_incompat_flags(super_block);
5791 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5792 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5793 btrfs_set_super_incompat_flags(super_block, newflags);
0b246afa 5794 spin_unlock(&fs_info->super_lock);
2eaa055f 5795
3a45bb20 5796 ret = btrfs_commit_transaction(trans);
7ab19625
DS		 5797out_drop_write:
5798 mnt_drop_write_file(file);
5799
5800 return ret;
2eaa055f
JM		 5801}
5802
2351f431
JB		 5803static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5804{
5805 struct btrfs_ioctl_send_args *arg;
5806 int ret;
5807
5808 if (compat) {
5809#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5810 struct btrfs_ioctl_send_args_32 args32;
5811
5812 ret = copy_from_user(&args32, argp, sizeof(args32));
5813 if (ret)
5814 return -EFAULT;
5815 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5816 if (!arg)
5817 return -ENOMEM;
5818 arg->send_fd = args32.send_fd;
5819 arg->clone_sources_count = args32.clone_sources_count;
5820 arg->clone_sources = compat_ptr(args32.clone_sources);
5821 arg->parent_root = args32.parent_root;
5822 arg->flags = args32.flags;
5823 memcpy(arg->reserved, args32.reserved,
5824 sizeof(args32.reserved));
5825#else
5826 return -ENOTTY;
5827#endif
5828 } else {
5829 arg = memdup_user(argp, sizeof(*arg));
5830 if (IS_ERR(arg))
5831 return PTR_ERR(arg);
5832 }
5833 ret = btrfs_ioctl_send(file, arg);
5834 kfree(arg);
5835 return ret;
5836}
5837
f46b5a66
CH		 5838long btrfs_ioctl(struct file *file, unsigned int
5839 cmd, unsigned long arg)
5840{
0b246afa
JM		 5841 struct inode *inode = file_inode(file);
5842 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5843 struct btrfs_root *root = BTRFS_I(inode)->root;
4bcabaa3 5844 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5845
5846 switch (cmd) {
6cbff00f
CH		 5847 case FS_IOC_GETFLAGS:
5848 return btrfs_ioctl_getflags(file, argp);
5849 case FS_IOC_SETFLAGS:
5850 return btrfs_ioctl_setflags(file, argp);
5851 case FS_IOC_GETVERSION:
5852 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5853 case FITRIM:
5854 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5855 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5856 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5857 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5858 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5859 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5860 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5861 case BTRFS_IOC_SUBVOL_CREATE_V2:
5862 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5863 case BTRFS_IOC_SNAP_DESTROY:
5864 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5865 case BTRFS_IOC_SUBVOL_GETFLAGS:
5866 return btrfs_ioctl_subvol_getflags(file, argp);
5867 case BTRFS_IOC_SUBVOL_SETFLAGS:
5868 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5869 case BTRFS_IOC_DEFAULT_SUBVOL:
5870 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5871 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5872 return btrfs_ioctl_defrag(file, NULL);
5873 case BTRFS_IOC_DEFRAG_RANGE:
5874 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5875 case BTRFS_IOC_RESIZE:
198605a8 5876 return btrfs_ioctl_resize(file, argp);
f46b5a66 5877 case BTRFS_IOC_ADD_DEV:
2ff7e61e 5878 return btrfs_ioctl_add_dev(fs_info, argp);
f46b5a66 5879 case BTRFS_IOC_RM_DEV:
da24927b 5880 return btrfs_ioctl_rm_dev(file, argp);
6b526ed7
AJ		 5881 case BTRFS_IOC_RM_DEV_V2:
5882 return btrfs_ioctl_rm_dev_v2(file, argp);
475f6387 5883 case BTRFS_IOC_FS_INFO:
2ff7e61e 5884 return btrfs_ioctl_fs_info(fs_info, argp);
475f6387 5885 case BTRFS_IOC_DEV_INFO:
2ff7e61e 5886 return btrfs_ioctl_dev_info(fs_info, argp);
f46b5a66 5887 case BTRFS_IOC_BALANCE:
9ba1f6e4 5888 return btrfs_ioctl_balance(file, NULL);
ac8e9819
CM		 5889 case BTRFS_IOC_TREE_SEARCH:
5890 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5891 case BTRFS_IOC_TREE_SEARCH_V2:
5892 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5893 case BTRFS_IOC_INO_LOOKUP:
5894 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5895 case BTRFS_IOC_INO_PATHS:
5896 return btrfs_ioctl_ino_to_path(root, argp);
5897 case BTRFS_IOC_LOGICAL_INO:
d24a67b2
ZB		 5898 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5899 case BTRFS_IOC_LOGICAL_INO_V2:
5900 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
1406e432 5901 case BTRFS_IOC_SPACE_INFO:
2ff7e61e 5902 return btrfs_ioctl_space_info(fs_info, argp);
9b199859
FDBM		 5903 case BTRFS_IOC_SYNC: {
5904 int ret;
5905
82b3e53b 5906 ret = btrfs_start_delalloc_roots(fs_info, -1);
9b199859
FDBM		 5907 if (ret)
5908 return ret;
0b246afa 5909 ret = btrfs_sync_fs(inode->i_sb, 1);
2fad4e83
DS		 5910 /*
5911 * The transaction thread may want to do more work,
01327610 5912 * namely it pokes the cleaner kthread that will start
2fad4e83
DS		 5913 * processing uncleaned subvols.
5914 */
0b246afa 5915 wake_up_process(fs_info->transaction_kthread);
9b199859
FDBM		 5916 return ret;
5917 }
46204592 5918 case BTRFS_IOC_START_SYNC:
9a8c28be 5919 return btrfs_ioctl_start_sync(root, argp);
46204592 5920 case BTRFS_IOC_WAIT_SYNC:
2ff7e61e 5921 return btrfs_ioctl_wait_sync(fs_info, argp);
475f6387 5922 case BTRFS_IOC_SCRUB:
b8e95489 5923 return btrfs_ioctl_scrub(file, argp);
475f6387 5924 case BTRFS_IOC_SCRUB_CANCEL:
2ff7e61e 5925 return btrfs_ioctl_scrub_cancel(fs_info);
475f6387 5926 case BTRFS_IOC_SCRUB_PROGRESS:
2ff7e61e 5927 return btrfs_ioctl_scrub_progress(fs_info, argp);
c9e9f97b 5928 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5929 return btrfs_ioctl_balance(file, argp);
837d5b6e 5930 case BTRFS_IOC_BALANCE_CTL:
2ff7e61e 5931 return btrfs_ioctl_balance_ctl(fs_info, arg);
19a39dce 5932 case BTRFS_IOC_BALANCE_PROGRESS:
2ff7e61e 5933 return btrfs_ioctl_balance_progress(fs_info, argp);
8ea05e3a
AB		 5934 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5935 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5936#ifdef CONFIG_64BIT
5937 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5938 return btrfs_ioctl_set_received_subvol_32(file, argp);
5939#endif
31db9f7c 5940 case BTRFS_IOC_SEND:
2351f431
JB		 5941 return _btrfs_ioctl_send(file, argp, false);
5942#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5943 case BTRFS_IOC_SEND_32:
5944 return _btrfs_ioctl_send(file, argp, true);
5945#endif
c11d2c23 5946 case BTRFS_IOC_GET_DEV_STATS:
2ff7e61e 5947 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5d13a37b 5948 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5949 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5950 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5951 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5952 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5953 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5954 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5955 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5956 case BTRFS_IOC_QUOTA_RESCAN:
5957 return btrfs_ioctl_quota_rescan(file, argp);
5958 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5959 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5960 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5961 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd 5962 case BTRFS_IOC_DEV_REPLACE:
2ff7e61e 5963 return btrfs_ioctl_dev_replace(fs_info, argp);
867ab667 5964 case BTRFS_IOC_GET_FSLABEL:
5965 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5966 case BTRFS_IOC_SET_FSLABEL:
5967 return btrfs_ioctl_set_fslabel(file, argp);
2eaa055f 5968 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
d5131b65 5969 return btrfs_ioctl_get_supported_features(argp);
2eaa055f
JM		 5970 case BTRFS_IOC_GET_FEATURES:
5971 return btrfs_ioctl_get_features(file, argp);
5972 case BTRFS_IOC_SET_FEATURES:
5973 return btrfs_ioctl_set_features(file, argp);
e4202ac9
DS		 5974 case FS_IOC_FSGETXATTR:
5975 return btrfs_ioctl_fsgetxattr(file, argp);
025f2121
DS		 5976 case FS_IOC_FSSETXATTR:
5977 return btrfs_ioctl_fssetxattr(file, argp);
b64ec075
TM		 5978 case BTRFS_IOC_GET_SUBVOL_INFO:
5979 return btrfs_ioctl_get_subvol_info(file, argp);
42e4b520
TM		 5980 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5981 return btrfs_ioctl_get_subvol_rootref(file, argp);
23d0b79d
TM		 5982 case BTRFS_IOC_INO_LOOKUP_USER:
5983 return btrfs_ioctl_ino_lookup_user(file, argp);
f46b5a66
CH		 5984 }
5985
5986 return -ENOTTY;
5987}
4c63c245
LD		 5988
5989#ifdef CONFIG_COMPAT
5990long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5991{
2a362249
JM		 5992 /*
5993 * These all access 32-bit values anyway so no further
5994 * handling is necessary.
5995 */
4c63c245
LD		 5996 switch (cmd) {
5997 case FS_IOC32_GETFLAGS:
5998 cmd = FS_IOC_GETFLAGS;
5999 break;
6000 case FS_IOC32_SETFLAGS:
6001 cmd = FS_IOC_SETFLAGS;
6002 break;
6003 case FS_IOC32_GETVERSION:
6004 cmd = FS_IOC_GETVERSION;
6005 break;
4c63c245
LD		 6006 }
6007
6008 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
6009}
6010#endif
Michael's Linux tree.