]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - fs/btrfs/ioctl.c
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vfs: create a generic checking and prep function for FS_IOC_SETFLAGS
[mirror_ubuntu-jammy-kernel.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;
5aca2842 190 unsigned int fsflags, old_fsflags;
6cbff00f 191 int ret;
ff9fef55 192 const char *comp = NULL;
d2b8fcfe 193 u32 binode_flags = binode->flags;
6cbff00f 194
bd60ea0f
DS		 195 if (!inode_owner_or_capable(inode))
196 return -EPERM;
197
b83cc969
LZ		 198 if (btrfs_root_readonly(root))
199 return -EROFS;
200
5c57b8b6 201 if (copy_from_user(&fsflags, arg, sizeof(fsflags)))
6cbff00f
CH		 202 return -EFAULT;
203
5c57b8b6 204 ret = check_fsflags(fsflags);
75e7cb7f
LB		 205 if (ret)
206 return ret;
f46b5a66 207
e7848683
JK		 208 ret = mnt_want_write_file(file);
209 if (ret)
210 return ret;
211
5955102c 212 inode_lock(inode);
6cbff00f 213
5c57b8b6 214 fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
5aca2842
DW		 215 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
216 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
217 if (ret)
218 goto out_unlock;
6cbff00f 219
5c57b8b6 220 if (fsflags & FS_SYNC_FL)
d2b8fcfe 221 binode_flags |= BTRFS_INODE_SYNC;
6cbff00f 222 else
d2b8fcfe 223 binode_flags &= ~BTRFS_INODE_SYNC;
5c57b8b6 224 if (fsflags & FS_IMMUTABLE_FL)
d2b8fcfe 225 binode_flags |= BTRFS_INODE_IMMUTABLE;
6cbff00f 226 else
d2b8fcfe 227 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
5c57b8b6 228 if (fsflags & FS_APPEND_FL)
d2b8fcfe 229 binode_flags |= BTRFS_INODE_APPEND;
6cbff00f 230 else
d2b8fcfe 231 binode_flags &= ~BTRFS_INODE_APPEND;
5c57b8b6 232 if (fsflags & FS_NODUMP_FL)
d2b8fcfe 233 binode_flags |= BTRFS_INODE_NODUMP;
6cbff00f 234 else
d2b8fcfe 235 binode_flags &= ~BTRFS_INODE_NODUMP;
5c57b8b6 236 if (fsflags & FS_NOATIME_FL)
d2b8fcfe 237 binode_flags |= BTRFS_INODE_NOATIME;
6cbff00f 238 else
d2b8fcfe 239 binode_flags &= ~BTRFS_INODE_NOATIME;
5c57b8b6 240 if (fsflags & FS_DIRSYNC_FL)
d2b8fcfe 241 binode_flags |= BTRFS_INODE_DIRSYNC;
6cbff00f 242 else
d2b8fcfe 243 binode_flags &= ~BTRFS_INODE_DIRSYNC;
5c57b8b6 244 if (fsflags & FS_NOCOW_FL) {
44e5194b 245 if (S_ISREG(inode->i_mode)) {
7e97b8da
DS		 246 /*
247 * It's safe to turn csums off here, no extents exist.
248 * Otherwise we want the flag to reflect the real COW
249 * status of the file and will not set it.
250 */
251 if (inode->i_size == 0)
d2b8fcfe
AJ		 252 binode_flags |= BTRFS_INODE_NODATACOW |
253 BTRFS_INODE_NODATASUM;
7e97b8da 254 } else {
d2b8fcfe 255 binode_flags |= BTRFS_INODE_NODATACOW;
7e97b8da
DS		 256 }
257 } else {
258 /*
01327610 259 * Revert back under same assumptions as above
7e97b8da 260 */
44e5194b 261 if (S_ISREG(inode->i_mode)) {
7e97b8da 262 if (inode->i_size == 0)
d2b8fcfe
AJ		 263 binode_flags &= ~(BTRFS_INODE_NODATACOW |
264 BTRFS_INODE_NODATASUM);
7e97b8da 265 } else {
d2b8fcfe 266 binode_flags &= ~BTRFS_INODE_NODATACOW;
7e97b8da
DS		 267 }
268 }
6cbff00f 269
75e7cb7f
LB		 270 /*
271 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
272 * flag may be changed automatically if compression code won't make
273 * things smaller.
274 */
5c57b8b6 275 if (fsflags & FS_NOCOMP_FL) {
d2b8fcfe
AJ		 276 binode_flags &= ~BTRFS_INODE_COMPRESS;
277 binode_flags |= BTRFS_INODE_NOCOMPRESS;
5c57b8b6 278 } else if (fsflags & FS_COMPR_FL) {
63541927 279
eede2bf3
OS		 280 if (IS_SWAPFILE(inode)) {
281 ret = -ETXTBSY;
282 goto out_unlock;
283 }
284
d2b8fcfe
AJ		 285 binode_flags |= BTRFS_INODE_COMPRESS;
286 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
63541927 287
93370509
DS		 288 comp = btrfs_compress_type2str(fs_info->compress_type);
289 if (!comp || comp[0] == 0)
290 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
ebcb904d 291 } else {
d2b8fcfe 292 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 293 }
6cbff00f 294
ff9fef55
AJ		 295 /*
296 * 1 for inode item
297 * 2 for properties
298 */
299 trans = btrfs_start_transaction(root, 3);
f062abf0
LZ		 300 if (IS_ERR(trans)) {
301 ret = PTR_ERR(trans);
d2b8fcfe 302 goto out_unlock;
f062abf0 303 }
6cbff00f 304
ff9fef55
AJ		 305 if (comp) {
306 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
307 strlen(comp), 0);
308 if (ret) {
309 btrfs_abort_transaction(trans, ret);
310 goto out_end_trans;
311 }
312 set_bit(BTRFS_INODE_COPY_EVERYTHING,
313 &BTRFS_I(inode)->runtime_flags);
314 } else {
315 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
316 0, 0);
317 if (ret && ret != -ENODATA) {
318 btrfs_abort_transaction(trans, ret);
319 goto out_end_trans;
320 }
321 }
322
d2b8fcfe 323 binode->flags = binode_flags;
7b6a221e 324 btrfs_sync_inode_flags_to_i_flags(inode);
0c4d2d95 325 inode_inc_iversion(inode);
c2050a45 326 inode->i_ctime = current_time(inode);
6cbff00f 327 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 328
ff9fef55 329 out_end_trans:
3a45bb20 330 btrfs_end_transaction(trans);
6cbff00f 331 out_unlock:
5955102c 332 inode_unlock(inode);
e7848683 333 mnt_drop_write_file(file);
2d4e6f6a 334 return ret;
6cbff00f
CH		 335}
336
19f93b3c
DS		 337/*
338 * Translate btrfs internal inode flags to xflags as expected by the
339 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
340 * silently dropped.
341 */
342static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
343{
344 unsigned int xflags = 0;
345
346 if (flags & BTRFS_INODE_APPEND)
347 xflags |= FS_XFLAG_APPEND;
348 if (flags & BTRFS_INODE_IMMUTABLE)
349 xflags |= FS_XFLAG_IMMUTABLE;
350 if (flags & BTRFS_INODE_NOATIME)
351 xflags |= FS_XFLAG_NOATIME;
352 if (flags & BTRFS_INODE_NODUMP)
353 xflags |= FS_XFLAG_NODUMP;
354 if (flags & BTRFS_INODE_SYNC)
355 xflags |= FS_XFLAG_SYNC;
356
357 return xflags;
358}
359
360/* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
361static int check_xflags(unsigned int flags)
362{
363 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
364 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
365 return -EOPNOTSUPP;
366 return 0;
367}
368
e4202ac9
DS		 369/*
370 * Set the xflags from the internal inode flags. The remaining items of fsxattr
371 * are zeroed.
372 */
373static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
374{
375 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
376 struct fsxattr fa;
377
378 memset(&fa, 0, sizeof(fa));
379 fa.fsx_xflags = btrfs_inode_flags_to_xflags(binode->flags);
380
381 if (copy_to_user(arg, &fa, sizeof(fa)))
382 return -EFAULT;
383
384 return 0;
385}
386
025f2121
DS		 387static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
388{
389 struct inode *inode = file_inode(file);
390 struct btrfs_inode *binode = BTRFS_I(inode);
391 struct btrfs_root *root = binode->root;
392 struct btrfs_trans_handle *trans;
393 struct fsxattr fa;
394 unsigned old_flags;
395 unsigned old_i_flags;
396 int ret = 0;
397
398 if (!inode_owner_or_capable(inode))
399 return -EPERM;
400
401 if (btrfs_root_readonly(root))
402 return -EROFS;
403
404 memset(&fa, 0, sizeof(fa));
405 if (copy_from_user(&fa, arg, sizeof(fa)))
406 return -EFAULT;
407
408 ret = check_xflags(fa.fsx_xflags);
409 if (ret)
410 return ret;
411
412 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
413 return -EOPNOTSUPP;
414
415 ret = mnt_want_write_file(file);
416 if (ret)
417 return ret;
418
419 inode_lock(inode);
420
421 old_flags = binode->flags;
422 old_i_flags = inode->i_flags;
423
424 /* We need the capabilities to change append-only or immutable inode */
425 if (((old_flags & (BTRFS_INODE_APPEND | BTRFS_INODE_IMMUTABLE)) ||
426 (fa.fsx_xflags & (FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE))) &&
427 !capable(CAP_LINUX_IMMUTABLE)) {
428 ret = -EPERM;
429 goto out_unlock;
430 }
431
432 if (fa.fsx_xflags & FS_XFLAG_SYNC)
433 binode->flags |= BTRFS_INODE_SYNC;
434 else
435 binode->flags &= ~BTRFS_INODE_SYNC;
436 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
437 binode->flags |= BTRFS_INODE_IMMUTABLE;
438 else
439 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
440 if (fa.fsx_xflags & FS_XFLAG_APPEND)
441 binode->flags |= BTRFS_INODE_APPEND;
442 else
443 binode->flags &= ~BTRFS_INODE_APPEND;
444 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
445 binode->flags |= BTRFS_INODE_NODUMP;
446 else
447 binode->flags &= ~BTRFS_INODE_NODUMP;
448 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
449 binode->flags |= BTRFS_INODE_NOATIME;
450 else
451 binode->flags &= ~BTRFS_INODE_NOATIME;
452
453 /* 1 item for the inode */
454 trans = btrfs_start_transaction(root, 1);
455 if (IS_ERR(trans)) {
456 ret = PTR_ERR(trans);
457 goto out_unlock;
458 }
459
460 btrfs_sync_inode_flags_to_i_flags(inode);
461 inode_inc_iversion(inode);
462 inode->i_ctime = current_time(inode);
463 ret = btrfs_update_inode(trans, root, inode);
464
465 btrfs_end_transaction(trans);
466
467out_unlock:
468 if (ret) {
469 binode->flags = old_flags;
470 inode->i_flags = old_i_flags;
471 }
472
473 inode_unlock(inode);
474 mnt_drop_write_file(file);
475
476 return ret;
477}
478
6cbff00f
CH		 479static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
480{
496ad9aa 481 struct inode *inode = file_inode(file);
6cbff00f
CH		 482
483 return put_user(inode->i_generation, arg);
484}
f46b5a66 485
f7039b1d
LD		 486static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
487{
0b246afa
JM		 488 struct inode *inode = file_inode(file);
489 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f7039b1d
LD		 490 struct btrfs_device *device;
491 struct request_queue *q;
492 struct fstrim_range range;
493 u64 minlen = ULLONG_MAX;
494 u64 num_devices = 0;
495 int ret;
496
497 if (!capable(CAP_SYS_ADMIN))
498 return -EPERM;
499
f35f06c3
FM		 500 /*
501 * If the fs is mounted with nologreplay, which requires it to be
502 * mounted in RO mode as well, we can not allow discard on free space
503 * inside block groups, because log trees refer to extents that are not
504 * pinned in a block group's free space cache (pinning the extents is
505 * precisely the first phase of replaying a log tree).
506 */
507 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
508 return -EROFS;
509
1f78160c
XG		 510 rcu_read_lock();
511 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
512 dev_list) {
f7039b1d
LD		 513 if (!device->bdev)
514 continue;
515 q = bdev_get_queue(device->bdev);
516 if (blk_queue_discard(q)) {
517 num_devices++;
50d0446e 518 minlen = min_t(u64, q->limits.discard_granularity,
f7039b1d
LD		 519 minlen);
520 }
521 }
1f78160c 522 rcu_read_unlock();
f4c697e6 523
f7039b1d
LD		 524 if (!num_devices)
525 return -EOPNOTSUPP;
f7039b1d
LD		 526 if (copy_from_user(&range, arg, sizeof(range)))
527 return -EFAULT;
6ba9fc8e
QW		 528
529 /*
530 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
531 * block group is in the logical address space, which can be any
532 * sectorsize aligned bytenr in the range [0, U64_MAX].
533 */
534 if (range.len < fs_info->sb->s_blocksize)
f4c697e6 535 return -EINVAL;
f7039b1d
LD		 536
537 range.minlen = max(range.minlen, minlen);
2ff7e61e 538 ret = btrfs_trim_fs(fs_info, &range);
f7039b1d
LD		 539 if (ret < 0)
540 return ret;
541
542 if (copy_to_user(arg, &range, sizeof(range)))
543 return -EFAULT;
544
545 return 0;
546}
547
dd5f9615
SB		 548int btrfs_is_empty_uuid(u8 *uuid)
549{
46e0f66a
CM		 550 int i;
551
552 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
553 if (uuid[i])
554 return 0;
555 }
556 return 1;
dd5f9615
SB		 557}
558
d5c12070 559static noinline int create_subvol(struct inode *dir,
cb8e7090 560 struct dentry *dentry,
52f75f4f 561 const char *name, int namelen,
6f72c7e2 562 u64 *async_transid,
8696c533 563 struct btrfs_qgroup_inherit *inherit)
f46b5a66 564{
0b246afa 565 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
f46b5a66
CH		 566 struct btrfs_trans_handle *trans;
567 struct btrfs_key key;
49a3c4d9 568 struct btrfs_root_item *root_item;
f46b5a66
CH		 569 struct btrfs_inode_item *inode_item;
570 struct extent_buffer *leaf;
d5c12070 571 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 572 struct btrfs_root *new_root;
d5c12070 573 struct btrfs_block_rsv block_rsv;
95582b00 574 struct timespec64 cur_time = current_time(dir);
5662344b 575 struct inode *inode;
f46b5a66
CH		 576 int ret;
577 int err;
578 u64 objectid;
579 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 580 u64 index = 0;
8ea05e3a 581 uuid_le new_uuid;
f46b5a66 582
49a3c4d9
DS		 583 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
584 if (!root_item)
585 return -ENOMEM;
586
0b246afa 587 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
2fbe8c8a 588 if (ret)
49a3c4d9 589 goto fail_free;
6a912213 590
e09fe2d2
QW		 591 /*
592 * Don't create subvolume whose level is not zero. Or qgroup will be
01327610 593 * screwed up since it assumes subvolume qgroup's level to be 0.
e09fe2d2 594 */
49a3c4d9
DS		 595 if (btrfs_qgroup_level(objectid)) {
596 ret = -ENOSPC;
597 goto fail_free;
598 }
e09fe2d2 599
d5c12070 600 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 601 /*
d5c12070
MX		 602 * The same as the snapshot creation, please see the comment
603 * of create_snapshot().
9ed74f2d 604 */
c4c129db 605 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
d5c12070 606 if (ret)
49a3c4d9 607 goto fail_free;
d5c12070
MX		 608
609 trans = btrfs_start_transaction(root, 0);
610 if (IS_ERR(trans)) {
611 ret = PTR_ERR(trans);
7775c818 612 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
49a3c4d9 613 goto fail_free;
d5c12070
MX		 614 }
615 trans->block_rsv = &block_rsv;
616 trans->bytes_reserved = block_rsv.size;
f46b5a66 617
a9377422 618 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
6f72c7e2
AJ		 619 if (ret)
620 goto fail;
621
4d75f8a9 622 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 623 if (IS_ERR(leaf)) {
624 ret = PTR_ERR(leaf);
625 goto fail;
626 }
f46b5a66 627
f46b5a66
CH		 628 btrfs_mark_buffer_dirty(leaf);
629
49a3c4d9 630 inode_item = &root_item->inode;
3cae210f
QW		 631 btrfs_set_stack_inode_generation(inode_item, 1);
632 btrfs_set_stack_inode_size(inode_item, 3);
633 btrfs_set_stack_inode_nlink(inode_item, 1);
da17066c 634 btrfs_set_stack_inode_nbytes(inode_item,
0b246afa 635 fs_info->nodesize);
3cae210f 636 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 637
49a3c4d9
DS		 638 btrfs_set_root_flags(root_item, 0);
639 btrfs_set_root_limit(root_item, 0);
3cae210f 640 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 641
49a3c4d9
DS		 642 btrfs_set_root_bytenr(root_item, leaf->start);
643 btrfs_set_root_generation(root_item, trans->transid);
644 btrfs_set_root_level(root_item, 0);
645 btrfs_set_root_refs(root_item, 1);
646 btrfs_set_root_used(root_item, leaf->len);
647 btrfs_set_root_last_snapshot(root_item, 0);
f46b5a66 648
49a3c4d9
DS		 649 btrfs_set_root_generation_v2(root_item,
650 btrfs_root_generation(root_item));
8ea05e3a 651 uuid_le_gen(&new_uuid);
49a3c4d9
DS		 652 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
653 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
654 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
655 root_item->ctime = root_item->otime;
656 btrfs_set_root_ctransid(root_item, trans->transid);
657 btrfs_set_root_otransid(root_item, trans->transid);
f46b5a66 658
925baedd 659 btrfs_tree_unlock(leaf);
f46b5a66
CH		 660 free_extent_buffer(leaf);
661 leaf = NULL;
662
49a3c4d9 663 btrfs_set_root_dirid(root_item, new_dirid);
f46b5a66
CH		 664
665 key.objectid = objectid;
5d4f98a2 666 key.offset = 0;
962a298f 667 key.type = BTRFS_ROOT_ITEM_KEY;
0b246afa 668 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
49a3c4d9 669 root_item);
f46b5a66
CH		 670 if (ret)
671 goto fail;
672
76dda93c 673 key.offset = (u64)-1;
0b246afa 674 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
79787eaa 675 if (IS_ERR(new_root)) {
79787eaa 676 ret = PTR_ERR(new_root);
66642832 677 btrfs_abort_transaction(trans, ret);
79787eaa
JM		 678 goto fail;
679 }
76dda93c
YZ		 680
681 btrfs_record_root_in_trans(trans, new_root);
682
63541927 683 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
ce598979
MF		 684 if (ret) {
685 /* We potentially lose an unused inode item here */
66642832 686 btrfs_abort_transaction(trans, ret);
ce598979
MF		 687 goto fail;
688 }
689
f32e48e9
CR		 690 mutex_lock(&new_root->objectid_mutex);
691 new_root->highest_objectid = new_dirid;
692 mutex_unlock(&new_root->objectid_mutex);
693
f46b5a66
CH		 694 /*
695 * insert the directory item
696 */
877574e2 697 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
79787eaa 698 if (ret) {
66642832 699 btrfs_abort_transaction(trans, ret);
79787eaa
JM		 700 goto fail;
701 }
3de4586c 702
684572df 703 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
3de4586c 704 BTRFS_FT_DIR, index);
79787eaa 705 if (ret) {
66642832 706 btrfs_abort_transaction(trans, ret);
f46b5a66 707 goto fail;
79787eaa 708 }
0660b5af 709
6ef06d27 710 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
52c26179
YZ		 711 ret = btrfs_update_inode(trans, root, dir);
712 BUG_ON(ret);
713
6025c19f 714 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
4a0cc7ca 715 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
76dda93c 716 BUG_ON(ret);
f46b5a66 717
cdb345a8 718 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
6bccf3ab 719 BTRFS_UUID_KEY_SUBVOL, objectid);
dd5f9615 720 if (ret)
66642832 721 btrfs_abort_transaction(trans, ret);
dd5f9615 722
f46b5a66 723fail:
49a3c4d9 724 kfree(root_item);
d5c12070
MX		 725 trans->block_rsv = NULL;
726 trans->bytes_reserved = 0;
7775c818 727 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
de6e8200 728
72fd032e
SW		 729 if (async_transid) {
730 *async_transid = trans->transid;
3a45bb20 731 err = btrfs_commit_transaction_async(trans, 1);
00d71c9c 732 if (err)
3a45bb20 733 err = btrfs_commit_transaction(trans);
72fd032e 734 } else {
3a45bb20 735 err = btrfs_commit_transaction(trans);
72fd032e 736 }
f46b5a66
CH		 737 if (err && !ret)
738 ret = err;
1a65e24b 739
5662344b
TI		 740 if (!ret) {
741 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 742 if (IS_ERR(inode))
743 return PTR_ERR(inode);
5662344b
TI		 744 d_instantiate(dentry, inode);
745 }
f46b5a66 746 return ret;
49a3c4d9
DS		 747
748fail_free:
749 kfree(root_item);
750 return ret;
f46b5a66
CH		 751}
752
e9662f70 753static int create_snapshot(struct btrfs_root *root, struct inode *dir,
61d7e4cb 754 struct dentry *dentry,
e9662f70
MX		 755 u64 *async_transid, bool readonly,
756 struct btrfs_qgroup_inherit *inherit)
f46b5a66 757{
0b246afa 758 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
2e4bfab9 759 struct inode *inode;
f46b5a66
CH		 760 struct btrfs_pending_snapshot *pending_snapshot;
761 struct btrfs_trans_handle *trans;
2e4bfab9 762 int ret;
8ecebf4d 763 bool snapshot_force_cow = false;
f46b5a66 764
27cdeb70 765 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
f46b5a66
CH		 766 return -EINVAL;
767
eede2bf3
OS		 768 if (atomic_read(&root->nr_swapfiles)) {
769 btrfs_warn(fs_info,
770 "cannot snapshot subvolume with active swapfile");
771 return -ETXTBSY;
772 }
773
23269bf5 774 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
a1ee7362
DS		 775 if (!pending_snapshot)
776 return -ENOMEM;
777
b0c0ea63 778 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
23269bf5 779 GFP_KERNEL);
8546b570
DS		 780 pending_snapshot->path = btrfs_alloc_path();
781 if (!pending_snapshot->root_item || !pending_snapshot->path) {
b0c0ea63
DS		 782 ret = -ENOMEM;
783 goto free_pending;
784 }
785
8ecebf4d
RK		 786 /*
787 * Force new buffered writes to reserve space even when NOCOW is
788 * possible. This is to avoid later writeback (running dealloc) to
789 * fallback to COW mode and unexpectedly fail with ENOSPC.
790 */
ea14b57f 791 atomic_inc(&root->will_be_snapshotted);
4e857c58 792 smp_mb__after_atomic();
45bac0f3
LB		 793 /* wait for no snapshot writes */
794 wait_event(root->subv_writers->wait,
795 percpu_counter_sum(&root->subv_writers->counter) == 0);
8257b2dc 796
3cd24c69 797 ret = btrfs_start_delalloc_snapshot(root);
6a03843d 798 if (ret)
a1ee7362 799 goto dec_and_free;
6a03843d 800
8ecebf4d
RK		 801 /*
802 * All previous writes have started writeback in NOCOW mode, so now
803 * we force future writes to fallback to COW mode during snapshot
804 * creation.
805 */
806 atomic_inc(&root->snapshot_force_cow);
807 snapshot_force_cow = true;
808
6374e57a 809 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
6a03843d 810
66d8f3dd
MX		 811 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
812 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 813 /*
814 * 1 - parent dir inode
815 * 2 - dir entries
816 * 1 - root item
817 * 2 - root ref/backref
818 * 1 - root of snapshot
dd5f9615 819 * 1 - UUID item
d5c12070
MX		 820 */
821 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 822 &pending_snapshot->block_rsv, 8,
ee3441b4 823 false);
d5c12070 824 if (ret)
a1ee7362 825 goto dec_and_free;
d5c12070 826
3de4586c 827 pending_snapshot->dentry = dentry;
f46b5a66 828 pending_snapshot->root = root;
b83cc969 829 pending_snapshot->readonly = readonly;
e9662f70 830 pending_snapshot->dir = dir;
8696c533 831 pending_snapshot->inherit = inherit;
a22285a6 832
d5c12070 833 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 834 if (IS_ERR(trans)) {
835 ret = PTR_ERR(trans);
836 goto fail;
837 }
838
0b246afa 839 spin_lock(&fs_info->trans_lock);
f46b5a66
CH		 840 list_add(&pending_snapshot->list,
841 &trans->transaction->pending_snapshots);
0b246afa 842 spin_unlock(&fs_info->trans_lock);
72fd032e
SW		 843 if (async_transid) {
844 *async_transid = trans->transid;
3a45bb20 845 ret = btrfs_commit_transaction_async(trans, 1);
00d71c9c 846 if (ret)
3a45bb20 847 ret = btrfs_commit_transaction(trans);
72fd032e 848 } else {
3a45bb20 849 ret = btrfs_commit_transaction(trans);
72fd032e 850 }
aec8030a 851 if (ret)
c37b2b62 852 goto fail;
a22285a6
YZ		 853
854 ret = pending_snapshot->error;
855 if (ret)
856 goto fail;
857
d3797308
CM		 858 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
859 if (ret)
860 goto fail;
861
2b0143b5 862 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
2e4bfab9
YZ		 863 if (IS_ERR(inode)) {
864 ret = PTR_ERR(inode);
865 goto fail;
866 }
5662344b 867
2e4bfab9
YZ		 868 d_instantiate(dentry, inode);
869 ret = 0;
870fail:
7775c818 871 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
a1ee7362 872dec_and_free:
8ecebf4d
RK		 873 if (snapshot_force_cow)
874 atomic_dec(&root->snapshot_force_cow);
ea14b57f 875 if (atomic_dec_and_test(&root->will_be_snapshotted))
4625956a 876 wake_up_var(&root->will_be_snapshotted);
b0c0ea63
DS		 877free_pending:
878 kfree(pending_snapshot->root_item);
8546b570 879 btrfs_free_path(pending_snapshot->path);
a1ee7362
DS		 880 kfree(pending_snapshot);
881
f46b5a66
CH		 882 return ret;
883}
884
4260f7c7
SW		 885/* copy of may_delete in fs/namei.c()
886 * Check whether we can remove a link victim from directory dir, check
887 * whether the type of victim is right.
888 * 1. We can't do it if dir is read-only (done in permission())
889 * 2. We should have write and exec permissions on dir
890 * 3. We can't remove anything from append-only dir
891 * 4. We can't do anything with immutable dir (done in permission())
892 * 5. If the sticky bit on dir is set we should either
893 * a. be owner of dir, or
894 * b. be owner of victim, or
895 * c. have CAP_FOWNER capability
01327610 896 * 6. If the victim is append-only or immutable we can't do anything with
4260f7c7
SW		 897 * links pointing to it.
898 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
899 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
900 * 9. We can't remove a root or mountpoint.
901 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
902 * nfs_async_unlink().
903 */
904
67871254 905static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 906{
907 int error;
908
2b0143b5 909 if (d_really_is_negative(victim))
4260f7c7
SW		 910 return -ENOENT;
911
2b0143b5 912 BUG_ON(d_inode(victim->d_parent) != dir);
4fa6b5ec 913 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 914
915 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
916 if (error)
917 return error;
918 if (IS_APPEND(dir))
919 return -EPERM;
2b0143b5
DH		 920 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
921 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
4260f7c7
SW		 922 return -EPERM;
923 if (isdir) {
e36cb0b8 924 if (!d_is_dir(victim))
4260f7c7
SW		 925 return -ENOTDIR;
926 if (IS_ROOT(victim))
927 return -EBUSY;
e36cb0b8 928 } else if (d_is_dir(victim))
4260f7c7
SW		 929 return -EISDIR;
930 if (IS_DEADDIR(dir))
931 return -ENOENT;
932 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
933 return -EBUSY;
934 return 0;
935}
936
cb8e7090
CH		 937/* copy of may_create in fs/namei.c() */
938static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
939{
2b0143b5 940 if (d_really_is_positive(child))
cb8e7090
CH		 941 return -EEXIST;
942 if (IS_DEADDIR(dir))
943 return -ENOENT;
944 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
945}
946
947/*
948 * Create a new subvolume below @parent. This is largely modeled after
949 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
950 * inside this filesystem so it's quite a bit simpler.
951 */
92872094 952static noinline int btrfs_mksubvol(const struct path *parent,
52f75f4f 953 const char *name, int namelen,
72fd032e 954 struct btrfs_root *snap_src,
6f72c7e2 955 u64 *async_transid, bool readonly,
8696c533 956 struct btrfs_qgroup_inherit *inherit)
cb8e7090 957{
0b246afa
JM		 958 struct inode *dir = d_inode(parent->dentry);
959 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
cb8e7090
CH		 960 struct dentry *dentry;
961 int error;
962
00235411
AV		 963 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
964 if (error == -EINTR)
965 return error;
cb8e7090
CH		 966
967 dentry = lookup_one_len(name, parent->dentry, namelen);
968 error = PTR_ERR(dentry);
969 if (IS_ERR(dentry))
970 goto out_unlock;
971
76dda93c 972 error = btrfs_may_create(dir, dentry);
cb8e7090 973 if (error)
a874a63e 974 goto out_dput;
cb8e7090 975
9c52057c
CM		 976 /*
977 * even if this name doesn't exist, we may get hash collisions.
978 * check for them now when we can safely fail
979 */
980 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
981 dir->i_ino, name,
982 namelen);
983 if (error)
984 goto out_dput;
985
0b246afa 986 down_read(&fs_info->subvol_sem);
76dda93c
YZ		 987
988 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
989 goto out_up_read;
990
3de4586c 991 if (snap_src) {
61d7e4cb 992 error = create_snapshot(snap_src, dir, dentry,
6f72c7e2 993 async_transid, readonly, inherit);
3de4586c 994 } else {
d5c12070
MX		 995 error = create_subvol(dir, dentry, name, namelen,
996 async_transid, inherit);
3de4586c 997 }
76dda93c
YZ		 998 if (!error)
999 fsnotify_mkdir(dir, dentry);
1000out_up_read:
0b246afa 1001 up_read(&fs_info->subvol_sem);
cb8e7090
CH		 1002out_dput:
1003 dput(dentry);
1004out_unlock:
5955102c 1005 inode_unlock(dir);
cb8e7090
CH		 1006 return error;
1007}
1008
4cb5300b
CM		 1009/*
1010 * When we're defragging a range, we don't want to kick it off again
1011 * if it is really just waiting for delalloc to send it down.
1012 * If we find a nice big extent or delalloc range for the bytes in the
1013 * file you want to defrag, we return 0 to let you know to skip this
1014 * part of the file
1015 */
aab110ab 1016static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
4cb5300b
CM		 1017{
1018 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1019 struct extent_map *em = NULL;
1020 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1021 u64 end;
1022
1023 read_lock(&em_tree->lock);
09cbfeaf 1024 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
4cb5300b
CM		 1025 read_unlock(&em_tree->lock);
1026
1027 if (em) {
1028 end = extent_map_end(em);
1029 free_extent_map(em);
1030 if (end - offset > thresh)
1031 return 0;
1032 }
1033 /* if we already have a nice delalloc here, just stop */
1034 thresh /= 2;
1035 end = count_range_bits(io_tree, &offset, offset + thresh,
1036 thresh, EXTENT_DELALLOC, 1);
1037 if (end >= thresh)
1038 return 0;
1039 return 1;
1040}
1041
1042/*
1043 * helper function to walk through a file and find extents
1044 * newer than a specific transid, and smaller than thresh.
1045 *
1046 * This is used by the defragging code to find new and small
1047 * extents
1048 */
1049static int find_new_extents(struct btrfs_root *root,
1050 struct inode *inode, u64 newer_than,
aab110ab 1051 u64 *off, u32 thresh)
4cb5300b
CM		 1052{
1053 struct btrfs_path *path;
1054 struct btrfs_key min_key;
4cb5300b
CM		 1055 struct extent_buffer *leaf;
1056 struct btrfs_file_extent_item *extent;
1057 int type;
1058 int ret;
4a0cc7ca 1059 u64 ino = btrfs_ino(BTRFS_I(inode));
4cb5300b
CM		 1060
1061 path = btrfs_alloc_path();
1062 if (!path)
1063 return -ENOMEM;
1064
a4689d2b 1065 min_key.objectid = ino;
4cb5300b
CM		 1066 min_key.type = BTRFS_EXTENT_DATA_KEY;
1067 min_key.offset = *off;
1068
67871254 1069 while (1) {
6174d3cb 1070 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 1071 if (ret != 0)
1072 goto none;
f094c9bd 1073process_slot:
a4689d2b 1074 if (min_key.objectid != ino)
4cb5300b
CM		 1075 goto none;
1076 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1077 goto none;
1078
1079 leaf = path->nodes[0];
1080 extent = btrfs_item_ptr(leaf, path->slots[0],
1081 struct btrfs_file_extent_item);
1082
1083 type = btrfs_file_extent_type(leaf, extent);
1084 if (type == BTRFS_FILE_EXTENT_REG &&
1085 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1086 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1087 *off = min_key.offset;
1088 btrfs_free_path(path);
1089 return 0;
1090 }
1091
f094c9bd
FM		 1092 path->slots[0]++;
1093 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1094 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1095 goto process_slot;
1096 }
1097
4cb5300b
CM		 1098 if (min_key.offset == (u64)-1)
1099 goto none;
1100
1101 min_key.offset++;
1102 btrfs_release_path(path);
1103 }
1104none:
1105 btrfs_free_path(path);
1106 return -ENOENT;
1107}
1108
6c282eb4 1109static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 1110{
1111 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 1112 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1113 struct extent_map *em;
09cbfeaf 1114 u64 len = PAGE_SIZE;
17ce6ef8 1115
6c282eb4
LZ		 1116 /*
1117 * hopefully we have this extent in the tree already, try without
1118 * the full extent lock
1119 */
17ce6ef8 1120 read_lock(&em_tree->lock);
6c282eb4 1121 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 1122 read_unlock(&em_tree->lock);
1123
6c282eb4 1124 if (!em) {
308d9800
FM		 1125 struct extent_state *cached = NULL;
1126 u64 end = start + len - 1;
1127
6c282eb4 1128 /* get the big lock and read metadata off disk */
ff13db41 1129 lock_extent_bits(io_tree, start, end, &cached);
fc4f21b1 1130 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
e43bbe5e 1131 unlock_extent_cached(io_tree, start, end, &cached);
6c282eb4
LZ		 1132
1133 if (IS_ERR(em))
1134 return NULL;
1135 }
1136
1137 return em;
1138}
17ce6ef8 1139
6c282eb4
LZ		 1140static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1141{
1142 struct extent_map *next;
1143 bool ret = true;
1144
1145 /* this is the last extent */
1146 if (em->start + em->len >= i_size_read(inode))
1147 return false;
1148
1149 next = defrag_lookup_extent(inode, em->start + em->len);
e9512d72
CM		 1150 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1151 ret = false;
1152 else if ((em->block_start + em->block_len == next->block_start) &&
ee22184b 1153 (em->block_len > SZ_128K && next->block_len > SZ_128K))
6c282eb4
LZ		 1154 ret = false;
1155
1156 free_extent_map(next);
17ce6ef8
LB		 1157 return ret;
1158}
1159
aab110ab 1160static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
a43a2111
AM		 1161 u64 *last_len, u64 *skip, u64 *defrag_end,
1162 int compress)
940100a4 1163{
6c282eb4 1164 struct extent_map *em;
940100a4 1165 int ret = 1;
6c282eb4 1166 bool next_mergeable = true;
4a3560c4 1167 bool prev_mergeable = true;
940100a4
CM		 1168
1169 /*
008873ea 1170 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1171 * defragging it
1172 */
1173 if (start < *defrag_end)
1174 return 1;
1175
1176 *skip = 0;
1177
6c282eb4
LZ		 1178 em = defrag_lookup_extent(inode, start);
1179 if (!em)
1180 return 0;
940100a4
CM		 1181
1182 /* this will cover holes, and inline extents */
17ce6ef8 1183 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1184 ret = 0;
17ce6ef8
LB		 1185 goto out;
1186 }
1187
4a3560c4
LB		 1188 if (!*defrag_end)
1189 prev_mergeable = false;
1190
6c282eb4 1191 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1192 /*
6c282eb4
LZ		 1193 * we hit a real extent, if it is big or the next extent is not a
1194 * real extent, don't bother defragging it
940100a4 1195 */
a43a2111 1196 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
4a3560c4 1197 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
940100a4 1198 ret = 0;
17ce6ef8 1199out:
940100a4
CM		 1200 /*
1201 * last_len ends up being a counter of how many bytes we've defragged.
1202 * every time we choose not to defrag an extent, we reset *last_len
1203 * so that the next tiny extent will force a defrag.
1204 *
1205 * The end result of this is that tiny extents before a single big
1206 * extent will force at least part of that big extent to be defragged.
1207 */
1208 if (ret) {
940100a4
CM		 1209 *defrag_end = extent_map_end(em);
1210 } else {
1211 *last_len = 0;
1212 *skip = extent_map_end(em);
1213 *defrag_end = 0;
1214 }
1215
1216 free_extent_map(em);
1217 return ret;
1218}
1219
4cb5300b
CM		 1220/*
1221 * it doesn't do much good to defrag one or two pages
1222 * at a time. This pulls in a nice chunk of pages
1223 * to COW and defrag.
1224 *
1225 * It also makes sure the delalloc code has enough
1226 * dirty data to avoid making new small extents as part
1227 * of the defrag
1228 *
1229 * It's a good idea to start RA on this range
1230 * before calling this.
1231 */
1232static int cluster_pages_for_defrag(struct inode *inode,
1233 struct page **pages,
1234 unsigned long start_index,
c41570c9 1235 unsigned long num_pages)
f46b5a66 1236{
4cb5300b
CM		 1237 unsigned long file_end;
1238 u64 isize = i_size_read(inode);
1239 u64 page_start;
1240 u64 page_end;
1f12bd06 1241 u64 page_cnt;
4cb5300b
CM		 1242 int ret;
1243 int i;
1244 int i_done;
3eaa2885 1245 struct btrfs_ordered_extent *ordered;
4cb5300b 1246 struct extent_state *cached_state = NULL;
600a45e1 1247 struct extent_io_tree *tree;
364ecf36 1248 struct extent_changeset *data_reserved = NULL;
3b16a4e3 1249 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1250
09cbfeaf 1251 file_end = (isize - 1) >> PAGE_SHIFT;
1f12bd06
LB		 1252 if (!isize || start_index > file_end)
1253 return 0;
1254
1255 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b 1256
364ecf36 1257 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
09cbfeaf
KS		 1258 start_index << PAGE_SHIFT,
1259 page_cnt << PAGE_SHIFT);
4cb5300b
CM		 1260 if (ret)
1261 return ret;
4cb5300b 1262 i_done = 0;
600a45e1 1263 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1264
1265 /* step one, lock all the pages */
1f12bd06 1266 for (i = 0; i < page_cnt; i++) {
4cb5300b 1267 struct page *page;
600a45e1 1268again:
a94733d0 1269 page = find_or_create_page(inode->i_mapping,
600a45e1 1270 start_index + i, mask);
4cb5300b
CM		 1271 if (!page)
1272 break;
1273
600a45e1 1274 page_start = page_offset(page);
09cbfeaf 1275 page_end = page_start + PAGE_SIZE - 1;
600a45e1 1276 while (1) {
308d9800 1277 lock_extent_bits(tree, page_start, page_end,
ff13db41 1278 &cached_state);
600a45e1
MX		 1279 ordered = btrfs_lookup_ordered_extent(inode,
1280 page_start);
308d9800 1281 unlock_extent_cached(tree, page_start, page_end,
e43bbe5e 1282 &cached_state);
600a45e1
MX		 1283 if (!ordered)
1284 break;
1285
1286 unlock_page(page);
1287 btrfs_start_ordered_extent(inode, ordered, 1);
1288 btrfs_put_ordered_extent(ordered);
1289 lock_page(page);
1f12bd06
LB		 1290 /*
1291 * we unlocked the page above, so we need check if
1292 * it was released or not.
1293 */
1294 if (page->mapping != inode->i_mapping) {
1295 unlock_page(page);
09cbfeaf 1296 put_page(page);
1f12bd06
LB		 1297 goto again;
1298 }
600a45e1
MX		 1299 }
1300
4cb5300b
CM		 1301 if (!PageUptodate(page)) {
1302 btrfs_readpage(NULL, page);
1303 lock_page(page);
1304 if (!PageUptodate(page)) {
1305 unlock_page(page);
09cbfeaf 1306 put_page(page);
4cb5300b
CM		 1307 ret = -EIO;
1308 break;
1309 }
1310 }
600a45e1 1311
600a45e1
MX		 1312 if (page->mapping != inode->i_mapping) {
1313 unlock_page(page);
09cbfeaf 1314 put_page(page);
600a45e1
MX		 1315 goto again;
1316 }
1317
4cb5300b
CM		 1318 pages[i] = page;
1319 i_done++;
1320 }
1321 if (!i_done || ret)
1322 goto out;
1323
1751e8a6 1324 if (!(inode->i_sb->s_flags & SB_ACTIVE))
4cb5300b
CM		 1325 goto out;
1326
1327 /*
1328 * so now we have a nice long stream of locked
1329 * and up to date pages, lets wait on them
1330 */
1331 for (i = 0; i < i_done; i++)
1332 wait_on_page_writeback(pages[i]);
1333
1334 page_start = page_offset(pages[0]);
09cbfeaf 1335 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
4cb5300b
CM		 1336
1337 lock_extent_bits(&BTRFS_I(inode)->io_tree,
ff13db41 1338 page_start, page_end - 1, &cached_state);
4cb5300b
CM		 1339 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1340 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b 1341 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
ae0f1625 1342 &cached_state);
4cb5300b 1343
1f12bd06 1344 if (i_done != page_cnt) {
9e0baf60 1345 spin_lock(&BTRFS_I(inode)->lock);
28c4a3e2 1346 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
9e0baf60 1347 spin_unlock(&BTRFS_I(inode)->lock);
bc42bda2 1348 btrfs_delalloc_release_space(inode, data_reserved,
09cbfeaf 1349 start_index << PAGE_SHIFT,
43b18595 1350 (page_cnt - i_done) << PAGE_SHIFT, true);
4cb5300b
CM		 1351 }
1352
1353
9e8a4a8b 1354 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
018ed4f7 1355 &cached_state);
4cb5300b
CM		 1356
1357 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
e43bbe5e 1358 page_start, page_end - 1, &cached_state);
4cb5300b
CM		 1359
1360 for (i = 0; i < i_done; i++) {
1361 clear_page_dirty_for_io(pages[i]);
1362 ClearPageChecked(pages[i]);
1363 set_page_extent_mapped(pages[i]);
1364 set_page_dirty(pages[i]);
1365 unlock_page(pages[i]);
09cbfeaf 1366 put_page(pages[i]);
4cb5300b 1367 }
43b18595
QW		 1368 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1369 false);
364ecf36 1370 extent_changeset_free(data_reserved);
4cb5300b
CM		 1371 return i_done;
1372out:
1373 for (i = 0; i < i_done; i++) {
1374 unlock_page(pages[i]);
09cbfeaf 1375 put_page(pages[i]);
4cb5300b 1376 }
bc42bda2 1377 btrfs_delalloc_release_space(inode, data_reserved,
09cbfeaf 1378 start_index << PAGE_SHIFT,
43b18595
QW		 1379 page_cnt << PAGE_SHIFT, true);
1380 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT,
1381 true);
364ecf36 1382 extent_changeset_free(data_reserved);
4cb5300b
CM		 1383 return ret;
1384
1385}
1386
1387int btrfs_defrag_file(struct inode *inode, struct file *file,
1388 struct btrfs_ioctl_defrag_range_args *range,
1389 u64 newer_than, unsigned long max_to_defrag)
1390{
0b246afa 1391 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4cb5300b 1392 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1393 struct file_ra_state *ra = NULL;
f46b5a66 1394 unsigned long last_index;
151a31b2 1395 u64 isize = i_size_read(inode);
940100a4
CM		 1396 u64 last_len = 0;
1397 u64 skip = 0;
1398 u64 defrag_end = 0;
4cb5300b 1399 u64 newer_off = range->start;
f46b5a66 1400 unsigned long i;
008873ea 1401 unsigned long ra_index = 0;
f46b5a66 1402 int ret;
4cb5300b 1403 int defrag_count = 0;
1a419d85 1404 int compress_type = BTRFS_COMPRESS_ZLIB;
aab110ab 1405 u32 extent_thresh = range->extent_thresh;
09cbfeaf 1406 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
c41570c9 1407 unsigned long cluster = max_cluster;
ee22184b 1408 u64 new_align = ~((u64)SZ_128K - 1);
4cb5300b 1409 struct page **pages = NULL;
1e2ef46d 1410 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
4cb5300b 1411
0abd5b17
LB		 1412 if (isize == 0)
1413 return 0;
1414
1415 if (range->start >= isize)
1416 return -EINVAL;
1a419d85 1417
1e2ef46d 1418 if (do_compress) {
1a419d85
LZ		 1419 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1420 return -EINVAL;
1421 if (range->compress_type)
1422 compress_type = range->compress_type;
1423 }
f46b5a66 1424
0abd5b17 1425 if (extent_thresh == 0)
ee22184b 1426 extent_thresh = SZ_256K;
940100a4 1427
4cb5300b 1428 /*
0a52d108
DS		 1429 * If we were not given a file, allocate a readahead context. As
1430 * readahead is just an optimization, defrag will work without it so
1431 * we don't error out.
4cb5300b
CM		 1432 */
1433 if (!file) {
63e727ec 1434 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
0a52d108
DS		 1435 if (ra)
1436 file_ra_state_init(ra, inode->i_mapping);
4cb5300b
CM		 1437 } else {
1438 ra = &file->f_ra;
1439 }
1440
63e727ec 1441 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
4cb5300b
CM		 1442 if (!pages) {
1443 ret = -ENOMEM;
1444 goto out_ra;
1445 }
1446
1447 /* find the last page to defrag */
1e701a32 1448 if (range->start + range->len > range->start) {
151a31b2 1449 last_index = min_t(u64, isize - 1,
09cbfeaf 1450 range->start + range->len - 1) >> PAGE_SHIFT;
1e701a32 1451 } else {
09cbfeaf 1452 last_index = (isize - 1) >> PAGE_SHIFT;
1e701a32
CM		 1453 }
1454
4cb5300b
CM		 1455 if (newer_than) {
1456 ret = find_new_extents(root, inode, newer_than,
ee22184b 1457 &newer_off, SZ_64K);
4cb5300b
CM		 1458 if (!ret) {
1459 range->start = newer_off;
1460 /*
1461 * we always align our defrag to help keep
1462 * the extents in the file evenly spaced
1463 */
09cbfeaf 1464 i = (newer_off & new_align) >> PAGE_SHIFT;
4cb5300b
CM		 1465 } else
1466 goto out_ra;
1467 } else {
09cbfeaf 1468 i = range->start >> PAGE_SHIFT;
4cb5300b
CM		 1469 }
1470 if (!max_to_defrag)
070034bd 1471 max_to_defrag = last_index - i + 1;
4cb5300b 1472
2a0f7f57
LZ		 1473 /*
1474 * make writeback starts from i, so the defrag range can be
1475 * written sequentially.
1476 */
1477 if (i < inode->i_mapping->writeback_index)
1478 inode->i_mapping->writeback_index = i;
1479
f7f43cc8 1480 while (i <= last_index && defrag_count < max_to_defrag &&
09cbfeaf 1481 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
4cb5300b
CM		 1482 /*
1483 * make sure we stop running if someone unmounts
1484 * the FS
1485 */
1751e8a6 1486 if (!(inode->i_sb->s_flags & SB_ACTIVE))
4cb5300b
CM		 1487 break;
1488
0b246afa
JM		 1489 if (btrfs_defrag_cancelled(fs_info)) {
1490 btrfs_debug(fs_info, "defrag_file cancelled");
210549eb
DS		 1491 ret = -EAGAIN;
1492 break;
1493 }
1494
09cbfeaf 1495 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
6c282eb4 1496 extent_thresh, &last_len, &skip,
1e2ef46d 1497 &defrag_end, do_compress)){
940100a4
CM		 1498 unsigned long next;
1499 /*
1500 * the should_defrag function tells us how much to skip
1501 * bump our counter by the suggested amount
1502 */
09cbfeaf 1503 next = DIV_ROUND_UP(skip, PAGE_SIZE);
940100a4
CM		 1504 i = max(i + 1, next);
1505 continue;
1506 }
008873ea
LZ		 1507
1508 if (!newer_than) {
09cbfeaf
KS		 1509 cluster = (PAGE_ALIGN(defrag_end) >>
1510 PAGE_SHIFT) - i;
008873ea
LZ		 1511 cluster = min(cluster, max_cluster);
1512 } else {
1513 cluster = max_cluster;
1514 }
1515
008873ea
LZ		 1516 if (i + cluster > ra_index) {
1517 ra_index = max(i, ra_index);
0a52d108 1518 if (ra)
d3c0bab5
DS		 1519 page_cache_sync_readahead(inode->i_mapping, ra,
1520 file, ra_index, cluster);
e4826a5b 1521 ra_index += cluster;
008873ea 1522 }
940100a4 1523
5955102c 1524 inode_lock(inode);
eede2bf3
OS		 1525 if (IS_SWAPFILE(inode)) {
1526 ret = -ETXTBSY;
1527 } else {
1528 if (do_compress)
1529 BTRFS_I(inode)->defrag_compress = compress_type;
1530 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1531 }
ecb8bea8 1532 if (ret < 0) {
5955102c 1533 inode_unlock(inode);
4cb5300b 1534 goto out_ra;
ecb8bea8 1535 }
4cb5300b
CM		 1536
1537 defrag_count += ret;
d0e1d66b 1538 balance_dirty_pages_ratelimited(inode->i_mapping);
5955102c 1539 inode_unlock(inode);
4cb5300b
CM		 1540
1541 if (newer_than) {
1542 if (newer_off == (u64)-1)
1543 break;
1544
e1f041e1
LB		 1545 if (ret > 0)
1546 i += ret;
1547
4cb5300b 1548 newer_off = max(newer_off + 1,
09cbfeaf 1549 (u64)i << PAGE_SHIFT);
4cb5300b 1550
ee22184b
BL		 1551 ret = find_new_extents(root, inode, newer_than,
1552 &newer_off, SZ_64K);
4cb5300b
CM		 1553 if (!ret) {
1554 range->start = newer_off;
09cbfeaf 1555 i = (newer_off & new_align) >> PAGE_SHIFT;
4cb5300b
CM		 1556 } else {
1557 break;
f46b5a66 1558 }
4cb5300b 1559 } else {
008873ea 1560 if (ret > 0) {
cbcc8326 1561 i += ret;
09cbfeaf 1562 last_len += ret << PAGE_SHIFT;
008873ea 1563 } else {
cbcc8326 1564 i++;
008873ea
LZ		 1565 last_len = 0;
1566 }
f46b5a66 1567 }
f46b5a66
CH		 1568 }
1569
dec8ef90 1570 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1571 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1572 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1573 &BTRFS_I(inode)->runtime_flags))
1574 filemap_flush(inode->i_mapping);
1575 }
1e701a32 1576
1a419d85 1577 if (range->compress_type == BTRFS_COMPRESS_LZO) {
0b246afa 1578 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
5c1aab1d
NT		 1579 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1580 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1a419d85
LZ		 1581 }
1582
60ccf82f 1583 ret = defrag_count;
940100a4 1584
4cb5300b 1585out_ra:
1e2ef46d 1586 if (do_compress) {
5955102c 1587 inode_lock(inode);
eec63c65 1588 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
5955102c 1589 inode_unlock(inode);
633085c7 1590 }
4cb5300b
CM		 1591 if (!file)
1592 kfree(ra);
1593 kfree(pages);
940100a4 1594 return ret;
f46b5a66
CH		 1595}
1596
198605a8 1597static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1598 void __user *arg)
f46b5a66 1599{
0b246afa
JM		 1600 struct inode *inode = file_inode(file);
1601 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66
CH		 1602 u64 new_size;
1603 u64 old_size;
1604 u64 devid = 1;
0b246afa 1605 struct btrfs_root *root = BTRFS_I(inode)->root;
f46b5a66
CH		 1606 struct btrfs_ioctl_vol_args *vol_args;
1607 struct btrfs_trans_handle *trans;
1608 struct btrfs_device *device = NULL;
1609 char *sizestr;
9a40f122 1610 char *retptr;
f46b5a66
CH		 1611 char *devstr = NULL;
1612 int ret = 0;
f46b5a66
CH		 1613 int mod = 0;
1614
e441d54d
CM		 1615 if (!capable(CAP_SYS_ADMIN))
1616 return -EPERM;
1617
198605a8
MX		 1618 ret = mnt_want_write_file(file);
1619 if (ret)
1620 return ret;
1621
171938e5 1622 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
97547676 1623 mnt_drop_write_file(file);
e57138b3 1624 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1625 }
1626
dae7b665 1627 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1628 if (IS_ERR(vol_args)) {
1629 ret = PTR_ERR(vol_args);
1630 goto out;
1631 }
5516e595
MF		 1632
1633 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1634
f46b5a66
CH		 1635 sizestr = vol_args->name;
1636 devstr = strchr(sizestr, ':');
1637 if (devstr) {
f46b5a66
CH		 1638 sizestr = devstr + 1;
1639 *devstr = '\0';
1640 devstr = vol_args->name;
58dfae63
Z		 1641 ret = kstrtoull(devstr, 10, &devid);
1642 if (ret)
1643 goto out_free;
dfd79829
MX		 1644 if (!devid) {
1645 ret = -EINVAL;
1646 goto out_free;
1647 }
0b246afa 1648 btrfs_info(fs_info, "resizing devid %llu", devid);
f46b5a66 1649 }
dba60f3f 1650
09ba3bc9 1651 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true);
f46b5a66 1652 if (!device) {
0b246afa
JM		 1653 btrfs_info(fs_info, "resizer unable to find device %llu",
1654 devid);
dfd79829 1655 ret = -ENODEV;
c9e9f97b 1656 goto out_free;
f46b5a66 1657 }
dba60f3f 1658
ebbede42 1659 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
0b246afa 1660 btrfs_info(fs_info,
efe120a0 1661 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1662 devid);
dfd79829 1663 ret = -EPERM;
4e42ae1b
LB		 1664 goto out_free;
1665 }
1666
f46b5a66
CH		 1667 if (!strcmp(sizestr, "max"))
1668 new_size = device->bdev->bd_inode->i_size;
1669 else {
1670 if (sizestr[0] == '-') {
1671 mod = -1;
1672 sizestr++;
1673 } else if (sizestr[0] == '+') {
1674 mod = 1;
1675 sizestr++;
1676 }
9a40f122
GH		 1677 new_size = memparse(sizestr, &retptr);
1678 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1679 ret = -EINVAL;
c9e9f97b 1680 goto out_free;
f46b5a66
CH		 1681 }
1682 }
1683
401e29c1 1684 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
dfd79829 1685 ret = -EPERM;
63a212ab
SB		 1686 goto out_free;
1687 }
1688
7cc8e58d 1689 old_size = btrfs_device_get_total_bytes(device);
f46b5a66
CH		 1690
1691 if (mod < 0) {
1692 if (new_size > old_size) {
1693 ret = -EINVAL;
c9e9f97b 1694 goto out_free;
f46b5a66
CH		 1695 }
1696 new_size = old_size - new_size;
1697 } else if (mod > 0) {
eb8052e0 1698 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1699 ret = -ERANGE;
eb8052e0
WF		 1700 goto out_free;
1701 }
f46b5a66
CH		 1702 new_size = old_size + new_size;
1703 }
1704
ee22184b 1705 if (new_size < SZ_256M) {
f46b5a66 1706 ret = -EINVAL;
c9e9f97b 1707 goto out_free;
f46b5a66
CH		 1708 }
1709 if (new_size > device->bdev->bd_inode->i_size) {
1710 ret = -EFBIG;
c9e9f97b 1711 goto out_free;
f46b5a66
CH		 1712 }
1713
47f08b96 1714 new_size = round_down(new_size, fs_info->sectorsize);
f46b5a66 1715
0b246afa
JM		 1716 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1717 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1718
1719 if (new_size > old_size) {
a22285a6 1720 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1721 if (IS_ERR(trans)) {
1722 ret = PTR_ERR(trans);
c9e9f97b 1723 goto out_free;
98d5dc13 1724 }
f46b5a66 1725 ret = btrfs_grow_device(trans, device, new_size);
3a45bb20 1726 btrfs_commit_transaction(trans);
ece7d20e 1727 } else if (new_size < old_size) {
f46b5a66 1728 ret = btrfs_shrink_device(device, new_size);
0253f40e 1729 } /* equal, nothing need to do */
f46b5a66 1730
c9e9f97b 1731out_free:
f46b5a66 1732 kfree(vol_args);
c9e9f97b 1733out:
171938e5 1734 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
18f39c41 1735 mnt_drop_write_file(file);
f46b5a66
CH		 1736 return ret;
1737}
1738
72fd032e 1739static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
52f75f4f 1740 const char *name, unsigned long fd, int subvol,
6f72c7e2 1741 u64 *transid, bool readonly,
8696c533 1742 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1743{
f46b5a66 1744 int namelen;
3de4586c 1745 int ret = 0;
f46b5a66 1746
325c50e3
JM		 1747 if (!S_ISDIR(file_inode(file)->i_mode))
1748 return -ENOTDIR;
1749
a874a63e
LB		 1750 ret = mnt_want_write_file(file);
1751 if (ret)
1752 goto out;
1753
72fd032e
SW		 1754 namelen = strlen(name);
1755 if (strchr(name, '/')) {
f46b5a66 1756 ret = -EINVAL;
a874a63e 1757 goto out_drop_write;
f46b5a66
CH		 1758 }
1759
16780cab
CM		 1760 if (name[0] == '.' &&
1761 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1762 ret = -EEXIST;
a874a63e 1763 goto out_drop_write;
16780cab
CM		 1764 }
1765
3de4586c 1766 if (subvol) {
72fd032e 1767 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1768 NULL, transid, readonly, inherit);
cb8e7090 1769 } else {
2903ff01 1770 struct fd src = fdget(fd);
3de4586c 1771 struct inode *src_inode;
2903ff01 1772 if (!src.file) {
3de4586c 1773 ret = -EINVAL;
a874a63e 1774 goto out_drop_write;
3de4586c
CM		 1775 }
1776
496ad9aa
AV		 1777 src_inode = file_inode(src.file);
1778 if (src_inode->i_sb != file_inode(file)->i_sb) {
c79b4713 1779 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
efe120a0 1780 "Snapshot src from another FS");
23ad5b17 1781 ret = -EXDEV;
d0242061
DS		 1782 } else if (!inode_owner_or_capable(src_inode)) {
1783 /*
1784 * Subvolume creation is not restricted, but snapshots
1785 * are limited to own subvolumes only
1786 */
1787 ret = -EPERM;
ecd18815
AV		 1788 } else {
1789 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1790 BTRFS_I(src_inode)->root,
1791 transid, readonly, inherit);
3de4586c 1792 }
2903ff01 1793 fdput(src);
cb8e7090 1794 }
a874a63e
LB		 1795out_drop_write:
1796 mnt_drop_write_file(file);
f46b5a66 1797out:
72fd032e
SW		 1798 return ret;
1799}
1800
1801static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1802 void __user *arg, int subvol)
72fd032e 1803{
fa0d2b9b 1804 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1805 int ret;
1806
325c50e3
JM		 1807 if (!S_ISDIR(file_inode(file)->i_mode))
1808 return -ENOTDIR;
1809
fa0d2b9b
LZ		 1810 vol_args = memdup_user(arg, sizeof(*vol_args));
1811 if (IS_ERR(vol_args))
1812 return PTR_ERR(vol_args);
1813 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1814
fa0d2b9b 1815 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1816 vol_args->fd, subvol,
6f72c7e2 1817 NULL, false, NULL);
fdfb1e4f 1818
fa0d2b9b
LZ		 1819 kfree(vol_args);
1820 return ret;
1821}
fdfb1e4f 1822
fa0d2b9b
LZ		 1823static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1824 void __user *arg, int subvol)
1825{
1826 struct btrfs_ioctl_vol_args_v2 *vol_args;
1827 int ret;
1828 u64 transid = 0;
1829 u64 *ptr = NULL;
b83cc969 1830 bool readonly = false;
6f72c7e2 1831 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1832
325c50e3
JM		 1833 if (!S_ISDIR(file_inode(file)->i_mode))
1834 return -ENOTDIR;
1835
fa0d2b9b
LZ		 1836 vol_args = memdup_user(arg, sizeof(*vol_args));
1837 if (IS_ERR(vol_args))
1838 return PTR_ERR(vol_args);
1839 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1840
b83cc969 1841 if (vol_args->flags &
6f72c7e2
AJ		 1842 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1843 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1844 ret = -EOPNOTSUPP;
c47ca32d 1845 goto free_args;
72fd032e 1846 }
fa0d2b9b
LZ		 1847
1848 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1849 ptr = &transid;
b83cc969
LZ		 1850 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1851 readonly = true;
6f72c7e2 1852 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
09cbfeaf 1853 if (vol_args->size > PAGE_SIZE) {
6f72c7e2 1854 ret = -EINVAL;
c47ca32d 1855 goto free_args;
6f72c7e2
AJ		 1856 }
1857 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1858 if (IS_ERR(inherit)) {
1859 ret = PTR_ERR(inherit);
c47ca32d 1860 goto free_args;
6f72c7e2
AJ		 1861 }
1862 }
fa0d2b9b
LZ		 1863
1864 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1865 vol_args->fd, subvol, ptr,
8696c533 1866 readonly, inherit);
c47ca32d
DC		 1867 if (ret)
1868 goto free_inherit;
fa0d2b9b 1869
c47ca32d
DC		 1870 if (ptr && copy_to_user(arg +
1871 offsetof(struct btrfs_ioctl_vol_args_v2,
1872 transid),
1873 ptr, sizeof(*ptr)))
fa0d2b9b 1874 ret = -EFAULT;
c47ca32d
DC		 1875
1876free_inherit:
6f72c7e2 1877 kfree(inherit);
c47ca32d
DC		 1878free_args:
1879 kfree(vol_args);
f46b5a66
CH		 1880 return ret;
1881}
1882
0caa102d
LZ		 1883static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1884 void __user *arg)
1885{
496ad9aa 1886 struct inode *inode = file_inode(file);
0b246afa 1887 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
0caa102d
LZ		 1888 struct btrfs_root *root = BTRFS_I(inode)->root;
1889 int ret = 0;
1890 u64 flags = 0;
1891
4a0cc7ca 1892 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1893 return -EINVAL;
1894
0b246afa 1895 down_read(&fs_info->subvol_sem);
0caa102d
LZ		 1896 if (btrfs_root_readonly(root))
1897 flags |= BTRFS_SUBVOL_RDONLY;
0b246afa 1898 up_read(&fs_info->subvol_sem);
0caa102d
LZ		 1899
1900 if (copy_to_user(arg, &flags, sizeof(flags)))
1901 ret = -EFAULT;
1902
1903 return ret;
1904}
1905
1906static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1907 void __user *arg)
1908{
496ad9aa 1909 struct inode *inode = file_inode(file);
0b246afa 1910 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
0caa102d
LZ		 1911 struct btrfs_root *root = BTRFS_I(inode)->root;
1912 struct btrfs_trans_handle *trans;
1913 u64 root_flags;
1914 u64 flags;
1915 int ret = 0;
1916
bd60ea0f
DS		 1917 if (!inode_owner_or_capable(inode))
1918 return -EPERM;
1919
b9ca0664
LB		 1920 ret = mnt_want_write_file(file);
1921 if (ret)
1922 goto out;
0caa102d 1923
4a0cc7ca 1924 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
b9ca0664
LB		 1925 ret = -EINVAL;
1926 goto out_drop_write;
1927 }
0caa102d 1928
b9ca0664
LB		 1929 if (copy_from_user(&flags, arg, sizeof(flags))) {
1930 ret = -EFAULT;
1931 goto out_drop_write;
1932 }
0caa102d 1933
b9ca0664
LB		 1934 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1935 ret = -EINVAL;
1936 goto out_drop_write;
1937 }
0caa102d 1938
b9ca0664
LB		 1939 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1940 ret = -EOPNOTSUPP;
1941 goto out_drop_write;
1942 }
0caa102d 1943
0b246afa 1944 down_write(&fs_info->subvol_sem);
0caa102d
LZ		 1945
1946 /* nothing to do */
1947 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1948 goto out_drop_sem;
0caa102d
LZ		 1949
1950 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1951 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1952 btrfs_set_root_flags(&root->root_item,
1953 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1954 } else {
1955 /*
1956 * Block RO -> RW transition if this subvolume is involved in
1957 * send
1958 */
1959 spin_lock(&root->root_item_lock);
1960 if (root->send_in_progress == 0) {
1961 btrfs_set_root_flags(&root->root_item,
0caa102d 1962 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1963 spin_unlock(&root->root_item_lock);
1964 } else {
1965 spin_unlock(&root->root_item_lock);
0b246afa
JM		 1966 btrfs_warn(fs_info,
1967 "Attempt to set subvolume %llu read-write during send",
1968 root->root_key.objectid);
2c686537
DS		 1969 ret = -EPERM;
1970 goto out_drop_sem;
1971 }
1972 }
0caa102d
LZ		 1973
1974 trans = btrfs_start_transaction(root, 1);
1975 if (IS_ERR(trans)) {
1976 ret = PTR_ERR(trans);
1977 goto out_reset;
1978 }
1979
0b246afa 1980 ret = btrfs_update_root(trans, fs_info->tree_root,
0caa102d 1981 &root->root_key, &root->root_item);
9417ebc8
NB		 1982 if (ret < 0) {
1983 btrfs_end_transaction(trans);
1984 goto out_reset;
1985 }
1986
1987 ret = btrfs_commit_transaction(trans);
0caa102d 1988
0caa102d
LZ		 1989out_reset:
1990 if (ret)
1991 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1992out_drop_sem:
0b246afa 1993 up_write(&fs_info->subvol_sem);
b9ca0664
LB		 1994out_drop_write:
1995 mnt_drop_write_file(file);
1996out:
0caa102d
LZ		 1997 return ret;
1998}
1999
ac8e9819
CM		 2000static noinline int key_in_sk(struct btrfs_key *key,
2001 struct btrfs_ioctl_search_key *sk)
2002{
abc6e134
CM		 2003 struct btrfs_key test;
2004 int ret;
2005
2006 test.objectid = sk->min_objectid;
2007 test.type = sk->min_type;
2008 test.offset = sk->min_offset;
2009
2010 ret = btrfs_comp_cpu_keys(key, &test);
2011 if (ret < 0)
ac8e9819 2012 return 0;
abc6e134
CM		 2013
2014 test.objectid = sk->max_objectid;
2015 test.type = sk->max_type;
2016 test.offset = sk->max_offset;
2017
2018 ret = btrfs_comp_cpu_keys(key, &test);
2019 if (ret > 0)
ac8e9819
CM		 2020 return 0;
2021 return 1;
2022}
2023
df397565 2024static noinline int copy_to_sk(struct btrfs_path *path,
ac8e9819
CM		 2025 struct btrfs_key *key,
2026 struct btrfs_ioctl_search_key *sk,
9b6e817d 2027 size_t *buf_size,
ba346b35 2028 char __user *ubuf,
ac8e9819
CM		 2029 unsigned long *sk_offset,
2030 int *num_found)
2031{
2032 u64 found_transid;
2033 struct extent_buffer *leaf;
2034 struct btrfs_ioctl_search_header sh;
dd81d459 2035 struct btrfs_key test;
ac8e9819
CM		 2036 unsigned long item_off;
2037 unsigned long item_len;
2038 int nritems;
2039 int i;
2040 int slot;
ac8e9819
CM		 2041 int ret = 0;
2042
2043 leaf = path->nodes[0];
2044 slot = path->slots[0];
2045 nritems = btrfs_header_nritems(leaf);
2046
2047 if (btrfs_header_generation(leaf) > sk->max_transid) {
2048 i = nritems;
2049 goto advance_key;
2050 }
2051 found_transid = btrfs_header_generation(leaf);
2052
2053 for (i = slot; i < nritems; i++) {
2054 item_off = btrfs_item_ptr_offset(leaf, i);
2055 item_len = btrfs_item_size_nr(leaf, i);
2056
03b71c6c
GP		 2057 btrfs_item_key_to_cpu(leaf, key, i);
2058 if (!key_in_sk(key, sk))
2059 continue;
2060
9b6e817d 2061 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 2062 if (*num_found) {
2063 ret = 1;
2064 goto out;
2065 }
2066
2067 /*
2068 * return one empty item back for v1, which does not
2069 * handle -EOVERFLOW
2070 */
2071
9b6e817d 2072 *buf_size = sizeof(sh) + item_len;
ac8e9819 2073 item_len = 0;
8f5f6178
GH		 2074 ret = -EOVERFLOW;
2075 }
ac8e9819 2076
9b6e817d 2077 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 2078 ret = 1;
25c9bc2e 2079 goto out;
ac8e9819
CM		 2080 }
2081
ac8e9819
CM		 2082 sh.objectid = key->objectid;
2083 sh.offset = key->offset;
2084 sh.type = key->type;
2085 sh.len = item_len;
2086 sh.transid = found_transid;
2087
2088 /* copy search result header */
ba346b35
GH		 2089 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2090 ret = -EFAULT;
2091 goto out;
2092 }
2093
ac8e9819
CM		 2094 *sk_offset += sizeof(sh);
2095
2096 if (item_len) {
ba346b35 2097 char __user *up = ubuf + *sk_offset;
ac8e9819 2098 /* copy the item */
ba346b35
GH		 2099 if (read_extent_buffer_to_user(leaf, up,
2100 item_off, item_len)) {
2101 ret = -EFAULT;
2102 goto out;
2103 }
2104
ac8e9819 2105 *sk_offset += item_len;
ac8e9819 2106 }
e2156867 2107 (*num_found)++;
ac8e9819 2108
8f5f6178
GH		 2109 if (ret) /* -EOVERFLOW from above */
2110 goto out;
2111
25c9bc2e
GH		 2112 if (*num_found >= sk->nr_items) {
2113 ret = 1;
2114 goto out;
2115 }
ac8e9819
CM		 2116 }
2117advance_key:
abc6e134 2118 ret = 0;
dd81d459
NA		 2119 test.objectid = sk->max_objectid;
2120 test.type = sk->max_type;
2121 test.offset = sk->max_offset;
2122 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2123 ret = 1;
2124 else if (key->offset < (u64)-1)
ac8e9819 2125 key->offset++;
dd81d459 2126 else if (key->type < (u8)-1) {
abc6e134 2127 key->offset = 0;
ac8e9819 2128 key->type++;
dd81d459 2129 } else if (key->objectid < (u64)-1) {
abc6e134
CM		 2130 key->offset = 0;
2131 key->type = 0;
ac8e9819 2132 key->objectid++;
abc6e134
CM		 2133 } else
2134 ret = 1;
25c9bc2e 2135out:
ba346b35
GH		 2136 /*
2137 * 0: all items from this leaf copied, continue with next
2138 * 1: * more items can be copied, but unused buffer is too small
2139 * * all items were found
2140 * Either way, it will stops the loop which iterates to the next
2141 * leaf
2142 * -EOVERFLOW: item was to large for buffer
2143 * -EFAULT: could not copy extent buffer back to userspace
2144 */
ac8e9819
CM		 2145 return ret;
2146}
2147
2148static noinline int search_ioctl(struct inode *inode,
12544442 2149 struct btrfs_ioctl_search_key *sk,
9b6e817d 2150 size_t *buf_size,
ba346b35 2151 char __user *ubuf)
ac8e9819 2152{
0b246afa 2153 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
ac8e9819
CM		 2154 struct btrfs_root *root;
2155 struct btrfs_key key;
ac8e9819 2156 struct btrfs_path *path;
ac8e9819
CM		 2157 int ret;
2158 int num_found = 0;
2159 unsigned long sk_offset = 0;
2160
9b6e817d
GH		 2161 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2162 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2163 return -EOVERFLOW;
9b6e817d 2164 }
12544442 2165
ac8e9819
CM		 2166 path = btrfs_alloc_path();
2167 if (!path)
2168 return -ENOMEM;
2169
2170 if (sk->tree_id == 0) {
2171 /* search the root of the inode that was passed */
2172 root = BTRFS_I(inode)->root;
2173 } else {
2174 key.objectid = sk->tree_id;
2175 key.type = BTRFS_ROOT_ITEM_KEY;
2176 key.offset = (u64)-1;
2177 root = btrfs_read_fs_root_no_name(info, &key);
2178 if (IS_ERR(root)) {
ac8e9819 2179 btrfs_free_path(path);
ad1e3d56 2180 return PTR_ERR(root);
ac8e9819
CM		 2181 }
2182 }
2183
2184 key.objectid = sk->min_objectid;
2185 key.type = sk->min_type;
2186 key.offset = sk->min_offset;
2187
67871254 2188 while (1) {
6174d3cb 2189 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2190 if (ret != 0) {
2191 if (ret > 0)
2192 ret = 0;
2193 goto err;
2194 }
df397565 2195 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
ac8e9819 2196 &sk_offset, &num_found);
b3b4aa74 2197 btrfs_release_path(path);
25c9bc2e 2198 if (ret)
ac8e9819
CM		 2199 break;
2200
2201 }
8f5f6178
GH		 2202 if (ret > 0)
2203 ret = 0;
ac8e9819
CM		 2204err:
2205 sk->nr_items = num_found;
2206 btrfs_free_path(path);
2207 return ret;
2208}
2209
2210static noinline int btrfs_ioctl_tree_search(struct file *file,
2211 void __user *argp)
2212{
ba346b35
GH		 2213 struct btrfs_ioctl_search_args __user *uargs;
2214 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2215 struct inode *inode;
2216 int ret;
2217 size_t buf_size;
ac8e9819
CM		 2218
2219 if (!capable(CAP_SYS_ADMIN))
2220 return -EPERM;
2221
ba346b35
GH		 2222 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2223
2224 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2225 return -EFAULT;
ac8e9819 2226
ba346b35 2227 buf_size = sizeof(uargs->buf);
ac8e9819 2228
496ad9aa 2229 inode = file_inode(file);
ba346b35 2230 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2231
2232 /*
2233 * In the origin implementation an overflow is handled by returning a
2234 * search header with a len of zero, so reset ret.
2235 */
2236 if (ret == -EOVERFLOW)
2237 ret = 0;
2238
ba346b35 2239 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2240 ret = -EFAULT;
ac8e9819
CM		 2241 return ret;
2242}
2243
cc68a8a5
GH		 2244static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2245 void __user *argp)
2246{
2247 struct btrfs_ioctl_search_args_v2 __user *uarg;
2248 struct btrfs_ioctl_search_args_v2 args;
2249 struct inode *inode;
2250 int ret;
2251 size_t buf_size;
ee22184b 2252 const size_t buf_limit = SZ_16M;
cc68a8a5
GH		 2253
2254 if (!capable(CAP_SYS_ADMIN))
2255 return -EPERM;
2256
2257 /* copy search header and buffer size */
2258 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2259 if (copy_from_user(&args, uarg, sizeof(args)))
2260 return -EFAULT;
2261
2262 buf_size = args.buf_size;
2263
cc68a8a5
GH		 2264 /* limit result size to 16MB */
2265 if (buf_size > buf_limit)
2266 buf_size = buf_limit;
2267
2268 inode = file_inode(file);
2269 ret = search_ioctl(inode, &args.key, &buf_size,
718dc5fa 2270 (char __user *)(&uarg->buf[0]));
cc68a8a5
GH		 2271 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2272 ret = -EFAULT;
2273 else if (ret == -EOVERFLOW &&
2274 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2275 ret = -EFAULT;
2276
ac8e9819
CM		 2277 return ret;
2278}
2279
98d377a0 2280/*
ac8e9819
CM		 2281 * Search INODE_REFs to identify path name of 'dirid' directory
2282 * in a 'tree_id' tree. and sets path name to 'name'.
2283 */
98d377a0
TH		 2284static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2285 u64 tree_id, u64 dirid, char *name)
2286{
2287 struct btrfs_root *root;
2288 struct btrfs_key key;
ac8e9819 2289 char *ptr;
98d377a0
TH		 2290 int ret = -1;
2291 int slot;
2292 int len;
2293 int total_len = 0;
2294 struct btrfs_inode_ref *iref;
2295 struct extent_buffer *l;
2296 struct btrfs_path *path;
2297
2298 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2299 name[0]='\0';
2300 return 0;
2301 }
2302
2303 path = btrfs_alloc_path();
2304 if (!path)
2305 return -ENOMEM;
2306
c8bcbfbd 2307 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
98d377a0
TH		 2308
2309 key.objectid = tree_id;
2310 key.type = BTRFS_ROOT_ITEM_KEY;
2311 key.offset = (u64)-1;
2312 root = btrfs_read_fs_root_no_name(info, &key);
2313 if (IS_ERR(root)) {
ad1e3d56 2314 ret = PTR_ERR(root);
8ad6fcab 2315 goto out;
98d377a0
TH		 2316 }
2317
2318 key.objectid = dirid;
2319 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2320 key.offset = (u64)-1;
98d377a0 2321
67871254 2322 while (1) {
98d377a0
TH		 2323 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2324 if (ret < 0)
2325 goto out;
18674c6c
FDBM		 2326 else if (ret > 0) {
2327 ret = btrfs_previous_item(root, path, dirid,
2328 BTRFS_INODE_REF_KEY);
2329 if (ret < 0)
2330 goto out;
2331 else if (ret > 0) {
2332 ret = -ENOENT;
2333 goto out;
2334 }
2335 }
98d377a0
TH		 2336
2337 l = path->nodes[0];
2338 slot = path->slots[0];
2339 btrfs_item_key_to_cpu(l, &key, slot);
2340
98d377a0
TH		 2341 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2342 len = btrfs_inode_ref_name_len(l, iref);
2343 ptr -= len + 1;
2344 total_len += len + 1;
a696cf35
FDBM		 2345 if (ptr < name) {
2346 ret = -ENAMETOOLONG;
98d377a0 2347 goto out;
a696cf35 2348 }
98d377a0
TH		 2349
2350 *(ptr + len) = '/';
67871254 2351 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2352
2353 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2354 break;
2355
b3b4aa74 2356 btrfs_release_path(path);
98d377a0 2357 key.objectid = key.offset;
8ad6fcab 2358 key.offset = (u64)-1;
98d377a0 2359 dirid = key.objectid;
98d377a0 2360 }
77906a50 2361 memmove(name, ptr, total_len);
67871254 2362 name[total_len] = '\0';
98d377a0
TH		 2363 ret = 0;
2364out:
2365 btrfs_free_path(path);
ac8e9819
CM		 2366 return ret;
2367}
2368
23d0b79d
TM		 2369static int btrfs_search_path_in_tree_user(struct inode *inode,
2370 struct btrfs_ioctl_ino_lookup_user_args *args)
2371{
2372 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2373 struct super_block *sb = inode->i_sb;
2374 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2375 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2376 u64 dirid = args->dirid;
2377 unsigned long item_off;
2378 unsigned long item_len;
2379 struct btrfs_inode_ref *iref;
2380 struct btrfs_root_ref *rref;
2381 struct btrfs_root *root;
2382 struct btrfs_path *path;
2383 struct btrfs_key key, key2;
2384 struct extent_buffer *leaf;
2385 struct inode *temp_inode;
2386 char *ptr;
2387 int slot;
2388 int len;
2389 int total_len = 0;
2390 int ret;
2391
2392 path = btrfs_alloc_path();
2393 if (!path)
2394 return -ENOMEM;
2395
2396 /*
2397 * If the bottom subvolume does not exist directly under upper_limit,
2398 * construct the path in from the bottom up.
2399 */
2400 if (dirid != upper_limit.objectid) {
2401 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2402
2403 key.objectid = treeid;
2404 key.type = BTRFS_ROOT_ITEM_KEY;
2405 key.offset = (u64)-1;
2406 root = btrfs_read_fs_root_no_name(fs_info, &key);
2407 if (IS_ERR(root)) {
2408 ret = PTR_ERR(root);
2409 goto out;
2410 }
2411
2412 key.objectid = dirid;
2413 key.type = BTRFS_INODE_REF_KEY;
2414 key.offset = (u64)-1;
2415 while (1) {
2416 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2417 if (ret < 0) {
2418 goto out;
2419 } else if (ret > 0) {
2420 ret = btrfs_previous_item(root, path, dirid,
2421 BTRFS_INODE_REF_KEY);
2422 if (ret < 0) {
2423 goto out;
2424 } else if (ret > 0) {
2425 ret = -ENOENT;
2426 goto out;
2427 }
2428 }
2429
2430 leaf = path->nodes[0];
2431 slot = path->slots[0];
2432 btrfs_item_key_to_cpu(leaf, &key, slot);
2433
2434 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2435 len = btrfs_inode_ref_name_len(leaf, iref);
2436 ptr -= len + 1;
2437 total_len += len + 1;
2438 if (ptr < args->path) {
2439 ret = -ENAMETOOLONG;
2440 goto out;
2441 }
2442
2443 *(ptr + len) = '/';
2444 read_extent_buffer(leaf, ptr,
2445 (unsigned long)(iref + 1), len);
2446
2447 /* Check the read+exec permission of this directory */
2448 ret = btrfs_previous_item(root, path, dirid,
2449 BTRFS_INODE_ITEM_KEY);
2450 if (ret < 0) {
2451 goto out;
2452 } else if (ret > 0) {
2453 ret = -ENOENT;
2454 goto out;
2455 }
2456
2457 leaf = path->nodes[0];
2458 slot = path->slots[0];
2459 btrfs_item_key_to_cpu(leaf, &key2, slot);
2460 if (key2.objectid != dirid) {
2461 ret = -ENOENT;
2462 goto out;
2463 }
2464
2465 temp_inode = btrfs_iget(sb, &key2, root, NULL);
3ca57bd6
MT		 2466 if (IS_ERR(temp_inode)) {
2467 ret = PTR_ERR(temp_inode);
2468 goto out;
2469 }
23d0b79d
TM		 2470 ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2471 iput(temp_inode);
2472 if (ret) {
2473 ret = -EACCES;
2474 goto out;
2475 }
2476
2477 if (key.offset == upper_limit.objectid)
2478 break;
2479 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2480 ret = -EACCES;
2481 goto out;
2482 }
2483
2484 btrfs_release_path(path);
2485 key.objectid = key.offset;
2486 key.offset = (u64)-1;
2487 dirid = key.objectid;
2488 }
2489
2490 memmove(args->path, ptr, total_len);
2491 args->path[total_len] = '\0';
2492 btrfs_release_path(path);
2493 }
2494
2495 /* Get the bottom subvolume's name from ROOT_REF */
2496 root = fs_info->tree_root;
2497 key.objectid = treeid;
2498 key.type = BTRFS_ROOT_REF_KEY;
2499 key.offset = args->treeid;
2500 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2501 if (ret < 0) {
2502 goto out;
2503 } else if (ret > 0) {
2504 ret = -ENOENT;
2505 goto out;
2506 }
2507
2508 leaf = path->nodes[0];
2509 slot = path->slots[0];
2510 btrfs_item_key_to_cpu(leaf, &key, slot);
2511
2512 item_off = btrfs_item_ptr_offset(leaf, slot);
2513 item_len = btrfs_item_size_nr(leaf, slot);
2514 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2515 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2516 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2517 ret = -EINVAL;
2518 goto out;
2519 }
2520
2521 /* Copy subvolume's name */
2522 item_off += sizeof(struct btrfs_root_ref);
2523 item_len -= sizeof(struct btrfs_root_ref);
2524 read_extent_buffer(leaf, args->name, item_off, item_len);
2525 args->name[item_len] = 0;
2526
2527out:
2528 btrfs_free_path(path);
2529 return ret;
2530}
2531
ac8e9819
CM		 2532static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2533 void __user *argp)
2534{
bece2e82
BVA		 2535 struct btrfs_ioctl_ino_lookup_args *args;
2536 struct inode *inode;
01b810b8 2537 int ret = 0;
ac8e9819 2538
2354d08f
JL		 2539 args = memdup_user(argp, sizeof(*args));
2540 if (IS_ERR(args))
2541 return PTR_ERR(args);
c2b96929 2542
496ad9aa 2543 inode = file_inode(file);
ac8e9819 2544
01b810b8
DS		 2545 /*
2546 * Unprivileged query to obtain the containing subvolume root id. The
2547 * path is reset so it's consistent with btrfs_search_path_in_tree.
2548 */
1b53ac4d
CM		 2549 if (args->treeid == 0)
2550 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2551
01b810b8
DS		 2552 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2553 args->name[0] = 0;
2554 goto out;
2555 }
2556
2557 if (!capable(CAP_SYS_ADMIN)) {
2558 ret = -EPERM;
2559 goto out;
2560 }
2561
ac8e9819
CM		 2562 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2563 args->treeid, args->objectid,
2564 args->name);
2565
01b810b8 2566out:
ac8e9819
CM		 2567 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2568 ret = -EFAULT;
2569
2570 kfree(args);
98d377a0
TH		 2571 return ret;
2572}
2573
23d0b79d
TM		 2574/*
2575 * Version of ino_lookup ioctl (unprivileged)
2576 *
2577 * The main differences from ino_lookup ioctl are:
2578 *
2579 * 1. Read + Exec permission will be checked using inode_permission() during
2580 * path construction. -EACCES will be returned in case of failure.
2581 * 2. Path construction will be stopped at the inode number which corresponds
2582 * to the fd with which this ioctl is called. If constructed path does not
2583 * exist under fd's inode, -EACCES will be returned.
2584 * 3. The name of bottom subvolume is also searched and filled.
2585 */
2586static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2587{
2588 struct btrfs_ioctl_ino_lookup_user_args *args;
2589 struct inode *inode;
2590 int ret;
2591
2592 args = memdup_user(argp, sizeof(*args));
2593 if (IS_ERR(args))
2594 return PTR_ERR(args);
2595
2596 inode = file_inode(file);
2597
2598 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2599 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2600 /*
2601 * The subvolume does not exist under fd with which this is
2602 * called
2603 */
2604 kfree(args);
2605 return -EACCES;
2606 }
2607
2608 ret = btrfs_search_path_in_tree_user(inode, args);
2609
2610 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2611 ret = -EFAULT;
2612
2613 kfree(args);
2614 return ret;
2615}
2616
b64ec075
TM		 2617/* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2618static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2619{
2620 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2621 struct btrfs_fs_info *fs_info;
2622 struct btrfs_root *root;
2623 struct btrfs_path *path;
2624 struct btrfs_key key;
2625 struct btrfs_root_item *root_item;
2626 struct btrfs_root_ref *rref;
2627 struct extent_buffer *leaf;
2628 unsigned long item_off;
2629 unsigned long item_len;
2630 struct inode *inode;
2631 int slot;
2632 int ret = 0;
2633
2634 path = btrfs_alloc_path();
2635 if (!path)
2636 return -ENOMEM;
2637
2638 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2639 if (!subvol_info) {
2640 btrfs_free_path(path);
2641 return -ENOMEM;
2642 }
2643
2644 inode = file_inode(file);
2645 fs_info = BTRFS_I(inode)->root->fs_info;
2646
2647 /* Get root_item of inode's subvolume */
2648 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2649 key.type = BTRFS_ROOT_ITEM_KEY;
2650 key.offset = (u64)-1;
2651 root = btrfs_read_fs_root_no_name(fs_info, &key);
2652 if (IS_ERR(root)) {
2653 ret = PTR_ERR(root);
2654 goto out;
2655 }
2656 root_item = &root->root_item;
2657
2658 subvol_info->treeid = key.objectid;
2659
2660 subvol_info->generation = btrfs_root_generation(root_item);
2661 subvol_info->flags = btrfs_root_flags(root_item);
2662
2663 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2664 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2665 BTRFS_UUID_SIZE);
2666 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2667 BTRFS_UUID_SIZE);
2668
2669 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2670 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2671 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2672
2673 subvol_info->otransid = btrfs_root_otransid(root_item);
2674 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2675 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2676
2677 subvol_info->stransid = btrfs_root_stransid(root_item);
2678 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2679 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2680
2681 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2682 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2683 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2684
2685 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2686 /* Search root tree for ROOT_BACKREF of this subvolume */
2687 root = fs_info->tree_root;
2688
2689 key.type = BTRFS_ROOT_BACKREF_KEY;
2690 key.offset = 0;
2691 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2692 if (ret < 0) {
2693 goto out;
2694 } else if (path->slots[0] >=
2695 btrfs_header_nritems(path->nodes[0])) {
2696 ret = btrfs_next_leaf(root, path);
2697 if (ret < 0) {
2698 goto out;
2699 } else if (ret > 0) {
2700 ret = -EUCLEAN;
2701 goto out;
2702 }
2703 }
2704
2705 leaf = path->nodes[0];
2706 slot = path->slots[0];
2707 btrfs_item_key_to_cpu(leaf, &key, slot);
2708 if (key.objectid == subvol_info->treeid &&
2709 key.type == BTRFS_ROOT_BACKREF_KEY) {
2710 subvol_info->parent_id = key.offset;
2711
2712 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2713 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2714
2715 item_off = btrfs_item_ptr_offset(leaf, slot)
2716 + sizeof(struct btrfs_root_ref);
2717 item_len = btrfs_item_size_nr(leaf, slot)
2718 - sizeof(struct btrfs_root_ref);
2719 read_extent_buffer(leaf, subvol_info->name,
2720 item_off, item_len);
2721 } else {
2722 ret = -ENOENT;
2723 goto out;
2724 }
2725 }
2726
2727 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2728 ret = -EFAULT;
2729
2730out:
2731 btrfs_free_path(path);
2732 kzfree(subvol_info);
2733 return ret;
2734}
2735
42e4b520
TM		 2736/*
2737 * Return ROOT_REF information of the subvolume containing this inode
2738 * except the subvolume name.
2739 */
2740static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2741{
2742 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2743 struct btrfs_root_ref *rref;
2744 struct btrfs_root *root;
2745 struct btrfs_path *path;
2746 struct btrfs_key key;
2747 struct extent_buffer *leaf;
2748 struct inode *inode;
2749 u64 objectid;
2750 int slot;
2751 int ret;
2752 u8 found;
2753
2754 path = btrfs_alloc_path();
2755 if (!path)
2756 return -ENOMEM;
2757
2758 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2759 if (IS_ERR(rootrefs)) {
2760 btrfs_free_path(path);
2761 return PTR_ERR(rootrefs);
2762 }
2763
2764 inode = file_inode(file);
2765 root = BTRFS_I(inode)->root->fs_info->tree_root;
2766 objectid = BTRFS_I(inode)->root->root_key.objectid;
2767
2768 key.objectid = objectid;
2769 key.type = BTRFS_ROOT_REF_KEY;
2770 key.offset = rootrefs->min_treeid;
2771 found = 0;
2772
2773 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2774 if (ret < 0) {
2775 goto out;
2776 } else if (path->slots[0] >=
2777 btrfs_header_nritems(path->nodes[0])) {
2778 ret = btrfs_next_leaf(root, path);
2779 if (ret < 0) {
2780 goto out;
2781 } else if (ret > 0) {
2782 ret = -EUCLEAN;
2783 goto out;
2784 }
2785 }
2786 while (1) {
2787 leaf = path->nodes[0];
2788 slot = path->slots[0];
2789
2790 btrfs_item_key_to_cpu(leaf, &key, slot);
2791 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2792 ret = 0;
2793 goto out;
2794 }
2795
2796 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2797 ret = -EOVERFLOW;
2798 goto out;
2799 }
2800
2801 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2802 rootrefs->rootref[found].treeid = key.offset;
2803 rootrefs->rootref[found].dirid =
2804 btrfs_root_ref_dirid(leaf, rref);
2805 found++;
2806
2807 ret = btrfs_next_item(root, path);
2808 if (ret < 0) {
2809 goto out;
2810 } else if (ret > 0) {
2811 ret = -EUCLEAN;
2812 goto out;
2813 }
2814 }
2815
2816out:
2817 if (!ret || ret == -EOVERFLOW) {
2818 rootrefs->num_items = found;
2819 /* update min_treeid for next search */
2820 if (found)
2821 rootrefs->min_treeid =
2822 rootrefs->rootref[found - 1].treeid + 1;
2823 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2824 ret = -EFAULT;
2825 }
2826
2827 kfree(rootrefs);
2828 btrfs_free_path(path);
2829
2830 return ret;
2831}
2832
76dda93c
YZ		 2833static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2834 void __user *arg)
2835{
54563d41 2836 struct dentry *parent = file->f_path.dentry;
0b246afa 2837 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
76dda93c 2838 struct dentry *dentry;
2b0143b5 2839 struct inode *dir = d_inode(parent);
76dda93c
YZ		 2840 struct inode *inode;
2841 struct btrfs_root *root = BTRFS_I(dir)->root;
2842 struct btrfs_root *dest = NULL;
2843 struct btrfs_ioctl_vol_args *vol_args;
76dda93c 2844 int namelen;
76dda93c
YZ		 2845 int err = 0;
2846
325c50e3
JM		 2847 if (!S_ISDIR(dir->i_mode))
2848 return -ENOTDIR;
2849
76dda93c
YZ		 2850 vol_args = memdup_user(arg, sizeof(*vol_args));
2851 if (IS_ERR(vol_args))
2852 return PTR_ERR(vol_args);
2853
2854 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2855 namelen = strlen(vol_args->name);
2856 if (strchr(vol_args->name, '/') ||
2857 strncmp(vol_args->name, "..", namelen) == 0) {
2858 err = -EINVAL;
2859 goto out;
2860 }
2861
a561be71 2862 err = mnt_want_write_file(file);
76dda93c
YZ		 2863 if (err)
2864 goto out;
2865
521e0546 2866
00235411
AV		 2867 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2868 if (err == -EINTR)
2869 goto out_drop_write;
76dda93c
YZ		 2870 dentry = lookup_one_len(vol_args->name, parent, namelen);
2871 if (IS_ERR(dentry)) {
2872 err = PTR_ERR(dentry);
2873 goto out_unlock_dir;
2874 }
2875
2b0143b5 2876 if (d_really_is_negative(dentry)) {
76dda93c
YZ		 2877 err = -ENOENT;
2878 goto out_dput;
2879 }
2880
2b0143b5 2881 inode = d_inode(dentry);
4260f7c7 2882 dest = BTRFS_I(inode)->root;
67871254 2883 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2884 /*
2885 * Regular user. Only allow this with a special mount
2886 * option, when the user has write+exec access to the
2887 * subvol root, and when rmdir(2) would have been
2888 * allowed.
2889 *
2890 * Note that this is _not_ check that the subvol is
2891 * empty or doesn't contain data that we wouldn't
2892 * otherwise be able to delete.
2893 *
2894 * Users who want to delete empty subvols should try
2895 * rmdir(2).
2896 */
2897 err = -EPERM;
0b246afa 2898 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
4260f7c7
SW		 2899 goto out_dput;
2900
2901 /*
2902 * Do not allow deletion if the parent dir is the same
2903 * as the dir to be deleted. That means the ioctl
2904 * must be called on the dentry referencing the root
2905 * of the subvol, not a random directory contained
2906 * within it.
2907 */
2908 err = -EINVAL;
2909 if (root == dest)
2910 goto out_dput;
2911
2912 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2913 if (err)
2914 goto out_dput;
4260f7c7
SW		 2915 }
2916
5c39da5b
MX		 2917 /* check if subvolume may be deleted by a user */
2918 err = btrfs_may_delete(dir, dentry, 1);
2919 if (err)
2920 goto out_dput;
2921
4a0cc7ca 2922 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2923 err = -EINVAL;
2924 goto out_dput;
2925 }
2926
5955102c 2927 inode_lock(inode);
f60a2364 2928 err = btrfs_delete_subvolume(dir, dentry);
5955102c 2929 inode_unlock(inode);
f60a2364 2930 if (!err)
76dda93c 2931 d_delete(dentry);
fa6ac876 2932
76dda93c
YZ		 2933out_dput:
2934 dput(dentry);
2935out_unlock_dir:
5955102c 2936 inode_unlock(dir);
00235411 2937out_drop_write:
2a79f17e 2938 mnt_drop_write_file(file);
76dda93c
YZ		 2939out:
2940 kfree(vol_args);
2941 return err;
2942}
2943
1e701a32 2944static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2945{
496ad9aa 2946 struct inode *inode = file_inode(file);
f46b5a66 2947 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2948 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2949 int ret;
2950
25122d15
ID		 2951 ret = mnt_want_write_file(file);
2952 if (ret)
2953 return ret;
b83cc969 2954
25122d15
ID		 2955 if (btrfs_root_readonly(root)) {
2956 ret = -EROFS;
2957 goto out;
5ac00add 2958 }
f46b5a66
CH		 2959
2960 switch (inode->i_mode & S_IFMT) {
2961 case S_IFDIR:
e441d54d
CM		 2962 if (!capable(CAP_SYS_ADMIN)) {
2963 ret = -EPERM;
2964 goto out;
2965 }
de78b51a 2966 ret = btrfs_defrag_root(root);
f46b5a66
CH		 2967 break;
2968 case S_IFREG:
616d374e
AB		 2969 /*
2970 * Note that this does not check the file descriptor for write
2971 * access. This prevents defragmenting executables that are
2972 * running and allows defrag on files open in read-only mode.
2973 */
2974 if (!capable(CAP_SYS_ADMIN) &&
2975 inode_permission(inode, MAY_WRITE)) {
2976 ret = -EPERM;
e441d54d
CM		 2977 goto out;
2978 }
1e701a32
CM		 2979
2980 range = kzalloc(sizeof(*range), GFP_KERNEL);
2981 if (!range) {
2982 ret = -ENOMEM;
2983 goto out;
2984 }
2985
2986 if (argp) {
2987 if (copy_from_user(range, argp,
2988 sizeof(*range))) {
2989 ret = -EFAULT;
2990 kfree(range);
683be16e 2991 goto out;
1e701a32
CM		 2992 }
2993 /* compression requires us to start the IO */
2994 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2995 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2996 range->extent_thresh = (u32)-1;
2997 }
2998 } else {
2999 /* the rest are all set to zero by kzalloc */
3000 range->len = (u64)-1;
3001 }
496ad9aa 3002 ret = btrfs_defrag_file(file_inode(file), file,
7c829b72 3003 range, BTRFS_OLDEST_GENERATION, 0);
4cb5300b
CM		 3004 if (ret > 0)
3005 ret = 0;
1e701a32 3006 kfree(range);
f46b5a66 3007 break;
8929ecfa
YZ		 3008 default:
3009 ret = -EINVAL;
f46b5a66 3010 }
e441d54d 3011out:
25122d15 3012 mnt_drop_write_file(file);
e441d54d 3013 return ret;
f46b5a66
CH		 3014}
3015
2ff7e61e 3016static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
f46b5a66
CH		 3017{
3018 struct btrfs_ioctl_vol_args *vol_args;
3019 int ret;
3020
e441d54d
CM		 3021 if (!capable(CAP_SYS_ADMIN))
3022 return -EPERM;
3023
171938e5 3024 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
e57138b3 3025 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b 3026
dae7b665 3027 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 3028 if (IS_ERR(vol_args)) {
3029 ret = PTR_ERR(vol_args);
3030 goto out;
3031 }
f46b5a66 3032
5516e595 3033 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2ff7e61e 3034 ret = btrfs_init_new_device(fs_info, vol_args->name);
f46b5a66 3035
43d20761 3036 if (!ret)
0b246afa 3037 btrfs_info(fs_info, "disk added %s", vol_args->name);
43d20761 3038
f46b5a66 3039 kfree(vol_args);
c9e9f97b 3040out:
171938e5 3041 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
f46b5a66
CH		 3042 return ret;
3043}
3044
6b526ed7 3045static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
f46b5a66 3046{
0b246afa
JM		 3047 struct inode *inode = file_inode(file);
3048 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6b526ed7 3049 struct btrfs_ioctl_vol_args_v2 *vol_args;
f46b5a66
CH		 3050 int ret;
3051
e441d54d
CM		 3052 if (!capable(CAP_SYS_ADMIN))
3053 return -EPERM;
3054
da24927b
MX		 3055 ret = mnt_want_write_file(file);
3056 if (ret)
3057 return ret;
c146afad 3058
dae7b665 3059 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 3060 if (IS_ERR(vol_args)) {
3061 ret = PTR_ERR(vol_args);
c47ca32d 3062 goto err_drop;
c9e9f97b 3063 }
f46b5a66 3064
6b526ed7 3065 /* Check for compatibility reject unknown flags */
fd4e994b
OS		 3066 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) {
3067 ret = -EOPNOTSUPP;
3068 goto out;
3069 }
f46b5a66 3070
171938e5 3071 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
183860f6
AJ		 3072 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3073 goto out;
3074 }
3075
735654ea 3076 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2ff7e61e 3077 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
6b526ed7
AJ		 3078 } else {
3079 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2ff7e61e 3080 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
6b526ed7 3081 }
171938e5 3082 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
183860f6 3083
6b526ed7 3084 if (!ret) {
735654ea 3085 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
0b246afa 3086 btrfs_info(fs_info, "device deleted: id %llu",
6b526ed7
AJ		 3087 vol_args->devid);
3088 else
0b246afa 3089 btrfs_info(fs_info, "device deleted: %s",
6b526ed7
AJ		 3090 vol_args->name);
3091 }
183860f6
AJ		 3092out:
3093 kfree(vol_args);
c47ca32d 3094err_drop:
4ac20c70 3095 mnt_drop_write_file(file);
f46b5a66
CH		 3096 return ret;
3097}
3098
da24927b 3099static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 3100{
0b246afa
JM		 3101 struct inode *inode = file_inode(file);
3102 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66
CH		 3103 struct btrfs_ioctl_vol_args *vol_args;
3104 int ret;
3105
e441d54d
CM		 3106 if (!capable(CAP_SYS_ADMIN))
3107 return -EPERM;
3108
da24927b
MX		 3109 ret = mnt_want_write_file(file);
3110 if (ret)
3111 return ret;
c146afad 3112
171938e5 3113 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
183860f6 3114 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
58d7bbf8
DS		 3115 goto out_drop_write;
3116 }
3117
3118 vol_args = memdup_user(arg, sizeof(*vol_args));
3119 if (IS_ERR(vol_args)) {
3120 ret = PTR_ERR(vol_args);
183860f6
AJ		 3121 goto out;
3122 }
3123
58d7bbf8 3124 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2ff7e61e 3125 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
183860f6 3126
ec95d491 3127 if (!ret)
0b246afa 3128 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
183860f6 3129 kfree(vol_args);
58d7bbf8 3130out:
171938e5 3131 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
58d7bbf8 3132out_drop_write:
4ac20c70 3133 mnt_drop_write_file(file);
58d7bbf8 3134
f46b5a66
CH		 3135 return ret;
3136}
3137
2ff7e61e
JM		 3138static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3139 void __user *arg)
475f6387 3140{
027ed2f0 3141 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387 3142 struct btrfs_device *device;
0b246afa 3143 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
027ed2f0 3144 int ret = 0;
475f6387 3145
027ed2f0
LZ		 3146 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
3147 if (!fi_args)
3148 return -ENOMEM;
3149
d03262c7 3150 rcu_read_lock();
027ed2f0 3151 fi_args->num_devices = fs_devices->num_devices;
475f6387 3152
d03262c7 3153 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 3154 if (device->devid > fi_args->max_id)
3155 fi_args->max_id = device->devid;
475f6387 3156 }
d03262c7 3157 rcu_read_unlock();
475f6387 3158
de37aa51 3159 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
bea7eafd
OS		 3160 fi_args->nodesize = fs_info->nodesize;
3161 fi_args->sectorsize = fs_info->sectorsize;
3162 fi_args->clone_alignment = fs_info->sectorsize;
80a773fb 3163
027ed2f0
LZ		 3164 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3165 ret = -EFAULT;
475f6387 3166
027ed2f0
LZ		 3167 kfree(fi_args);
3168 return ret;
475f6387
JS		 3169}
3170
2ff7e61e
JM		 3171static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3172 void __user *arg)
475f6387
JS		 3173{
3174 struct btrfs_ioctl_dev_info_args *di_args;
3175 struct btrfs_device *dev;
475f6387
JS		 3176 int ret = 0;
3177 char *s_uuid = NULL;
475f6387 3178
475f6387
JS		 3179 di_args = memdup_user(arg, sizeof(*di_args));
3180 if (IS_ERR(di_args))
3181 return PTR_ERR(di_args);
3182
dd5f9615 3183 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 3184 s_uuid = di_args->uuid;
3185
c5593ca3 3186 rcu_read_lock();
e4319cd9 3187 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
09ba3bc9 3188 NULL, true);
475f6387
JS		 3189
3190 if (!dev) {
3191 ret = -ENODEV;
3192 goto out;
3193 }
3194
3195 di_args->devid = dev->devid;
7cc8e58d
MX		 3196 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3197 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
475f6387 3198 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 3199 if (dev->name) {
672d5990
MT		 3200 strncpy(di_args->path, rcu_str_deref(dev->name),
3201 sizeof(di_args->path) - 1);
a27202fb
JM		 3202 di_args->path[sizeof(di_args->path) - 1] = 0;
3203 } else {
99ba55ad 3204 di_args->path[0] = '\0';
a27202fb 3205 }
475f6387
JS		 3206
3207out:
c5593ca3 3208 rcu_read_unlock();
475f6387
JS		 3209 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3210 ret = -EFAULT;
3211
3212 kfree(di_args);
3213 return ret;
3214}
3215
d8b55242
FM		 3216static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
3217 struct inode *inode2, u64 loff2, u64 len)
3218{
3219 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3220 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3221}
3222
3223static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
3224 struct inode *inode2, u64 loff2, u64 len)
3225{
3226 if (inode1 < inode2) {
3227 swap(inode1, inode2);
3228 swap(loff1, loff2);
3229 } else if (inode1 == inode2 && loff2 < loff1) {
3230 swap(loff1, loff2);
3231 }
3232 lock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3233 lock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3234}
3235
57a50e25 3236static int btrfs_extent_same_range(struct inode *src, u64 loff, u64 len,
34a28e3d 3237 struct inode *dst, u64 dst_loff)
416161db
MF		 3238{
3239 int ret;
f4414602 3240
34a28e3d 3241 /*
d8b55242
FM		 3242 * Lock destination range to serialize with concurrent readpages() and
3243 * source range to serialize with relocation.
34a28e3d 3244 */
d8b55242 3245 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
57a50e25 3246 ret = btrfs_clone(src, dst, loff, len, len, dst_loff, 1);
d8b55242 3247 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3973909d
TT		 3248
3249 return ret;
3250}
3251
b6728768
TT		 3252#define BTRFS_MAX_DEDUPE_LEN SZ_16M
3253
3973909d
TT		 3254static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3255 struct inode *dst, u64 dst_loff)
3256{
3257 int ret;
b6728768 3258 u64 i, tail_len, chunk_count;
62d54f3a
FM		 3259 struct btrfs_root *root_dst = BTRFS_I(dst)->root;
3260
3261 spin_lock(&root_dst->root_item_lock);
3262 if (root_dst->send_in_progress) {
3263 btrfs_warn_rl(root_dst->fs_info,
3264"cannot deduplicate to root %llu while send operations are using it (%d in progress)",
3265 root_dst->root_key.objectid,
3266 root_dst->send_in_progress);
3267 spin_unlock(&root_dst->root_item_lock);
3268 return -EAGAIN;
3269 }
3270 root_dst->dedupe_in_progress++;
3271 spin_unlock(&root_dst->root_item_lock);
3973909d 3272
b6728768
TT		 3273 tail_len = olen % BTRFS_MAX_DEDUPE_LEN;
3274 chunk_count = div_u64(olen, BTRFS_MAX_DEDUPE_LEN);
67b07bd4 3275
b6728768
TT		 3276 for (i = 0; i < chunk_count; i++) {
3277 ret = btrfs_extent_same_range(src, loff, BTRFS_MAX_DEDUPE_LEN,
34a28e3d 3278 dst, dst_loff);
b6728768 3279 if (ret)
62d54f3a 3280 goto out;
b6728768
TT		 3281
3282 loff += BTRFS_MAX_DEDUPE_LEN;
3283 dst_loff += BTRFS_MAX_DEDUPE_LEN;
3284 }
3285
3286 if (tail_len > 0)
67b07bd4 3287 ret = btrfs_extent_same_range(src, loff, tail_len, dst,
34a28e3d 3288 dst_loff);
62d54f3a
FM		 3289out:
3290 spin_lock(&root_dst->root_item_lock);
3291 root_dst->dedupe_in_progress--;
3292 spin_unlock(&root_dst->root_item_lock);
416161db
MF		 3293
3294 return ret;
3295}
3296
f82a9901
FM		 3297static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3298 struct inode *inode,
3299 u64 endoff,
3300 const u64 destoff,
1c919a5e
MF		 3301 const u64 olen,
3302 int no_time_update)
f82a9901
FM		 3303{
3304 struct btrfs_root *root = BTRFS_I(inode)->root;
3305 int ret;
3306
3307 inode_inc_iversion(inode);
1c919a5e 3308 if (!no_time_update)
c2050a45 3309 inode->i_mtime = inode->i_ctime = current_time(inode);
f82a9901
FM		 3310 /*
3311 * We round up to the block size at eof when determining which
3312 * extents to clone above, but shouldn't round up the file size.
3313 */
3314 if (endoff > destoff + olen)
3315 endoff = destoff + olen;
3316 if (endoff > inode->i_size)
6ef06d27 3317 btrfs_i_size_write(BTRFS_I(inode), endoff);
f82a9901
FM		 3318
3319 ret = btrfs_update_inode(trans, root, inode);
3320 if (ret) {
66642832 3321 btrfs_abort_transaction(trans, ret);
3a45bb20 3322 btrfs_end_transaction(trans);
f82a9901
FM		 3323 goto out;
3324 }
3a45bb20 3325 ret = btrfs_end_transaction(trans);
f82a9901
FM		 3326out:
3327 return ret;
3328}
3329
a2f392e4 3330static void clone_update_extent_map(struct btrfs_inode *inode,
7ffbb598
FM		 3331 const struct btrfs_trans_handle *trans,
3332 const struct btrfs_path *path,
7ffbb598
FM		 3333 const u64 hole_offset,
3334 const u64 hole_len)
3335{
a2f392e4 3336 struct extent_map_tree *em_tree = &inode->extent_tree;
7ffbb598
FM		 3337 struct extent_map *em;
3338 int ret;
3339
3340 em = alloc_extent_map();
3341 if (!em) {
a2f392e4 3342 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
7ffbb598
FM		 3343 return;
3344 }
3345
14f59796
FM		 3346 if (path) {
3347 struct btrfs_file_extent_item *fi;
3348
3349 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3350 struct btrfs_file_extent_item);
a2f392e4 3351 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
7ffbb598
FM		 3352 em->generation = -1;
3353 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3354 BTRFS_FILE_EXTENT_INLINE)
3355 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
a2f392e4 3356 &inode->runtime_flags);
7ffbb598
FM		 3357 } else {
3358 em->start = hole_offset;
3359 em->len = hole_len;
3360 em->ram_bytes = em->len;
3361 em->orig_start = hole_offset;
3362 em->block_start = EXTENT_MAP_HOLE;
3363 em->block_len = 0;
3364 em->orig_block_len = 0;
3365 em->compress_type = BTRFS_COMPRESS_NONE;
3366 em->generation = trans->transid;
3367 }
3368
3369 while (1) {
3370 write_lock(&em_tree->lock);
3371 ret = add_extent_mapping(em_tree, em, 1);
3372 write_unlock(&em_tree->lock);
3373 if (ret != -EEXIST) {
3374 free_extent_map(em);
3375 break;
3376 }
a2f392e4 3377 btrfs_drop_extent_cache(inode, em->start,
7ffbb598
FM		 3378 em->start + em->len - 1, 0);
3379 }
3380
ee39b432 3381 if (ret)
a2f392e4 3382 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
7ffbb598
FM		 3383}
3384
8039d87d
FM		 3385/*
3386 * Make sure we do not end up inserting an inline extent into a file that has
3387 * already other (non-inline) extents. If a file has an inline extent it can
3388 * not have any other extents and the (single) inline extent must start at the
3389 * file offset 0. Failing to respect these rules will lead to file corruption,
3390 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3391 *
3392 * We can have extents that have been already written to disk or we can have
3393 * dirty ranges still in delalloc, in which case the extent maps and items are
3394 * created only when we run delalloc, and the delalloc ranges might fall outside
3395 * the range we are currently locking in the inode's io tree. So we check the
3396 * inode's i_size because of that (i_size updates are done while holding the
3397 * i_mutex, which we are holding here).
3398 * We also check to see if the inode has a size not greater than "datal" but has
3399 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3400 * protected against such concurrent fallocate calls by the i_mutex).
3401 *
3402 * If the file has no extents but a size greater than datal, do not allow the
3403 * copy because we would need turn the inline extent into a non-inline one (even
3404 * with NO_HOLES enabled). If we find our destination inode only has one inline
3405 * extent, just overwrite it with the source inline extent if its size is less
3406 * than the source extent's size, or we could copy the source inline extent's
3407 * data into the destination inode's inline extent if the later is greater then
3408 * the former.
3409 */
4a0ab9d7 3410static int clone_copy_inline_extent(struct inode *dst,
8039d87d
FM		 3411 struct btrfs_trans_handle *trans,
3412 struct btrfs_path *path,
3413 struct btrfs_key *new_key,
3414 const u64 drop_start,
3415 const u64 datal,
3416 const u64 skip,
3417 const u64 size,
3418 char *inline_data)
3419{
0b246afa 3420 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
8039d87d
FM		 3421 struct btrfs_root *root = BTRFS_I(dst)->root;
3422 const u64 aligned_end = ALIGN(new_key->offset + datal,
0b246afa 3423 fs_info->sectorsize);
8039d87d
FM		 3424 int ret;
3425 struct btrfs_key key;
3426
3427 if (new_key->offset > 0)
3428 return -EOPNOTSUPP;
3429
4a0cc7ca 3430 key.objectid = btrfs_ino(BTRFS_I(dst));
8039d87d
FM		 3431 key.type = BTRFS_EXTENT_DATA_KEY;
3432 key.offset = 0;
3433 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3434 if (ret < 0) {
3435 return ret;
3436 } else if (ret > 0) {
3437 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3438 ret = btrfs_next_leaf(root, path);
3439 if (ret < 0)
3440 return ret;
3441 else if (ret > 0)
3442 goto copy_inline_extent;
3443 }
3444 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
4a0cc7ca 3445 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
8039d87d
FM		 3446 key.type == BTRFS_EXTENT_DATA_KEY) {
3447 ASSERT(key.offset > 0);
3448 return -EOPNOTSUPP;
3449 }
3450 } else if (i_size_read(dst) <= datal) {
3451 struct btrfs_file_extent_item *ei;
3452 u64 ext_len;
3453
3454 /*
3455 * If the file size is <= datal, make sure there are no other
3456 * extents following (can happen do to an fallocate call with
3457 * the flag FALLOC_FL_KEEP_SIZE).
3458 */
3459 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3460 struct btrfs_file_extent_item);
3461 /*
3462 * If it's an inline extent, it can not have other extents
3463 * following it.
3464 */
3465 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3466 BTRFS_FILE_EXTENT_INLINE)
3467 goto copy_inline_extent;
3468
3469 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3470 if (ext_len > aligned_end)
3471 return -EOPNOTSUPP;
3472
3473 ret = btrfs_next_item(root, path);
3474 if (ret < 0) {
3475 return ret;
3476 } else if (ret == 0) {
3477 btrfs_item_key_to_cpu(path->nodes[0], &key,
3478 path->slots[0]);
4a0cc7ca 3479 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
8039d87d
FM		 3480 key.type == BTRFS_EXTENT_DATA_KEY)
3481 return -EOPNOTSUPP;
3482 }
3483 }
3484
3485copy_inline_extent:
3486 /*
3487 * We have no extent items, or we have an extent at offset 0 which may
3488 * or may not be inlined. All these cases are dealt the same way.
3489 */
3490 if (i_size_read(dst) > datal) {
3491 /*
3492 * If the destination inode has an inline extent...
3493 * This would require copying the data from the source inline
3494 * extent into the beginning of the destination's inline extent.
3495 * But this is really complex, both extents can be compressed
3496 * or just one of them, which would require decompressing and
3497 * re-compressing data (which could increase the new compressed
3498 * size, not allowing the compressed data to fit anymore in an
3499 * inline extent).
3500 * So just don't support this case for now (it should be rare,
3501 * we are not really saving space when cloning inline extents).
3502 */
3503 return -EOPNOTSUPP;
3504 }
3505
3506 btrfs_release_path(path);
3507 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3508 if (ret)
3509 return ret;
3510 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3511 if (ret)
3512 return ret;
3513
3514 if (skip) {
3515 const u32 start = btrfs_file_extent_calc_inline_size(0);
3516
3517 memmove(inline_data + start, inline_data + start + skip, datal);
3518 }
3519
3520 write_extent_buffer(path->nodes[0], inline_data,
3521 btrfs_item_ptr_offset(path->nodes[0],
3522 path->slots[0]),
3523 size);
3524 inode_add_bytes(dst, datal);
3525
3526 return 0;
3527}
3528
32b7c687
MF		 3529/**
3530 * btrfs_clone() - clone a range from inode file to another
3531 *
3532 * @src: Inode to clone from
3533 * @inode: Inode to clone to
3534 * @off: Offset within source to start clone from
3535 * @olen: Original length, passed by user, of range to clone
1c919a5e 3536 * @olen_aligned: Block-aligned value of olen
32b7c687 3537 * @destoff: Offset within @inode to start clone
1c919a5e 3538 * @no_time_update: Whether to update mtime/ctime on the target inode
32b7c687
MF		 3539 */
3540static int btrfs_clone(struct inode *src, struct inode *inode,
f82a9901 3541 const u64 off, const u64 olen, const u64 olen_aligned,
1c919a5e 3542 const u64 destoff, int no_time_update)
f46b5a66 3543{
0b246afa 3544 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66 3545 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 3546 struct btrfs_path *path = NULL;
f46b5a66 3547 struct extent_buffer *leaf;
32b7c687
MF		 3548 struct btrfs_trans_handle *trans;
3549 char *buf = NULL;
ae01a0ab 3550 struct btrfs_key key;
f46b5a66
CH		 3551 u32 nritems;
3552 int slot;
ae01a0ab 3553 int ret;
f82a9901 3554 const u64 len = olen_aligned;
f82a9901 3555 u64 last_dest_end = destoff;
ae01a0ab
YZ		 3556
3557 ret = -ENOMEM;
752ade68
MH		 3558 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3559 if (!buf)
3560 return ret;
ae01a0ab
YZ		 3561
3562 path = btrfs_alloc_path();
3563 if (!path) {
15351955 3564 kvfree(buf);
32b7c687 3565 return ret;
d525e8ab
LZ		 3566 }
3567
e4058b54 3568 path->reada = READA_FORWARD;
c5c9cd4d 3569 /* clone data */
4a0cc7ca 3570 key.objectid = btrfs_ino(BTRFS_I(src));
ae01a0ab 3571 key.type = BTRFS_EXTENT_DATA_KEY;
2c463823 3572 key.offset = off;
f46b5a66
CH		 3573
3574 while (1) {
de249e66 3575 u64 next_key_min_offset = key.offset + 1;
df858e76 3576
f46b5a66
CH		 3577 /*
3578 * note the key will change type as we walk through the
3579 * tree.
3580 */
e4355f34 3581 path->leave_spinning = 1;
362a20c5
DS		 3582 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3583 0, 0);
f46b5a66
CH		 3584 if (ret < 0)
3585 goto out;
2c463823
FM		 3586 /*
3587 * First search, if no extent item that starts at offset off was
3588 * found but the previous item is an extent item, it's possible
3589 * it might overlap our target range, therefore process it.
3590 */
3591 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3592 btrfs_item_key_to_cpu(path->nodes[0], &key,
3593 path->slots[0] - 1);
3594 if (key.type == BTRFS_EXTENT_DATA_KEY)
3595 path->slots[0]--;
3596 }
f46b5a66 3597
ae01a0ab 3598 nritems = btrfs_header_nritems(path->nodes[0]);
e4355f34 3599process_slot:
ae01a0ab 3600 if (path->slots[0] >= nritems) {
362a20c5 3601 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 3602 if (ret < 0)
3603 goto out;
3604 if (ret > 0)
3605 break;
ae01a0ab 3606 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 3607 }
3608 leaf = path->nodes[0];
3609 slot = path->slots[0];
f46b5a66 3610
ae01a0ab 3611 btrfs_item_key_to_cpu(leaf, &key, slot);
962a298f 3612 if (key.type > BTRFS_EXTENT_DATA_KEY ||
4a0cc7ca 3613 key.objectid != btrfs_ino(BTRFS_I(src)))
f46b5a66
CH		 3614 break;
3615
962a298f 3616 if (key.type == BTRFS_EXTENT_DATA_KEY) {
c5c9cd4d
SW		 3617 struct btrfs_file_extent_item *extent;
3618 int type;
31840ae1
ZY		 3619 u32 size;
3620 struct btrfs_key new_key;
c5c9cd4d
SW		 3621 u64 disko = 0, diskl = 0;
3622 u64 datao = 0, datal = 0;
3623 u8 comp;
f82a9901 3624 u64 drop_start;
31840ae1 3625
c5c9cd4d
SW		 3626 extent = btrfs_item_ptr(leaf, slot,
3627 struct btrfs_file_extent_item);
3628 comp = btrfs_file_extent_compression(leaf, extent);
3629 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 3630 if (type == BTRFS_FILE_EXTENT_REG ||
3631 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 3632 disko = btrfs_file_extent_disk_bytenr(leaf,
3633 extent);
3634 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3635 extent);
c5c9cd4d 3636 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 3637 datal = btrfs_file_extent_num_bytes(leaf,
3638 extent);
c5c9cd4d
SW		 3639 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3640 /* take upper bound, may be compressed */
3641 datal = btrfs_file_extent_ram_bytes(leaf,
3642 extent);
3643 }
31840ae1 3644
2c463823
FM		 3645 /*
3646 * The first search might have left us at an extent
3647 * item that ends before our target range's start, can
3648 * happen if we have holes and NO_HOLES feature enabled.
3649 */
3650 if (key.offset + datal <= off) {
e4355f34
FDBM		 3651 path->slots[0]++;
3652 goto process_slot;
2c463823
FM		 3653 } else if (key.offset >= off + len) {
3654 break;
e4355f34 3655 }
df858e76 3656 next_key_min_offset = key.offset + datal;
e4355f34
FDBM		 3657 size = btrfs_item_size_nr(leaf, slot);
3658 read_extent_buffer(leaf, buf,
3659 btrfs_item_ptr_offset(leaf, slot),
3660 size);
3661
3662 btrfs_release_path(path);
3663 path->leave_spinning = 0;
c5c9cd4d 3664
31840ae1 3665 memcpy(&new_key, &key, sizeof(new_key));
4a0cc7ca 3666 new_key.objectid = btrfs_ino(BTRFS_I(inode));
4d728ec7
LZ		 3667 if (off <= key.offset)
3668 new_key.offset = key.offset + destoff - off;
3669 else
3670 new_key.offset = destoff;
31840ae1 3671
f82a9901
FM		 3672 /*
3673 * Deal with a hole that doesn't have an extent item
3674 * that represents it (NO_HOLES feature enabled).
3675 * This hole is either in the middle of the cloning
3676 * range or at the beginning (fully overlaps it or
3677 * partially overlaps it).
3678 */
3679 if (new_key.offset != last_dest_end)
3680 drop_start = last_dest_end;
3681 else
3682 drop_start = new_key.offset;
3683
b6f3409b
SW		 3684 /*
3685 * 1 - adjusting old extent (we may have to split it)
3686 * 1 - add new extent
3687 * 1 - inode update
3688 */
3689 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 3690 if (IS_ERR(trans)) {
3691 ret = PTR_ERR(trans);
3692 goto out;
3693 }
3694
c8a894d7
CM		 3695 if (type == BTRFS_FILE_EXTENT_REG ||
3696 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 3697 /*
3698 * a | --- range to clone ---| b
3699 * | ------------- extent ------------- |
3700 */
3701
93915584 3702 /* subtract range b */
d72c0842
LZ		 3703 if (key.offset + datal > off + len)
3704 datal = off + len - key.offset;
3705
93915584 3706 /* subtract range a */
a22285a6
YZ		 3707 if (off > key.offset) {
3708 datao += off - key.offset;
3709 datal -= off - key.offset;
3710 }
3711
5dc562c5 3712 ret = btrfs_drop_extents(trans, root, inode,
f82a9901 3713 drop_start,
a22285a6 3714 new_key.offset + datal,
2671485d 3715 1);
79787eaa 3716 if (ret) {
3f9e3df8 3717 if (ret != -EOPNOTSUPP)
00fdf13a 3718 btrfs_abort_transaction(trans,
66642832 3719 ret);
3a45bb20 3720 btrfs_end_transaction(trans);
79787eaa
JM		 3721 goto out;
3722 }
a22285a6 3723
c5c9cd4d
SW		 3724 ret = btrfs_insert_empty_item(trans, root, path,
3725 &new_key, size);
79787eaa 3726 if (ret) {
66642832 3727 btrfs_abort_transaction(trans, ret);
3a45bb20 3728 btrfs_end_transaction(trans);
79787eaa
JM		 3729 goto out;
3730 }
c5c9cd4d
SW		 3731
3732 leaf = path->nodes[0];
3733 slot = path->slots[0];
3734 write_extent_buffer(leaf, buf,
31840ae1
ZY		 3735 btrfs_item_ptr_offset(leaf, slot),
3736 size);
ae01a0ab 3737
c5c9cd4d 3738 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 3739 struct btrfs_file_extent_item);
c5c9cd4d 3740
c5c9cd4d
SW		 3741 /* disko == 0 means it's a hole */
3742 if (!disko)
3743 datao = 0;
c5c9cd4d
SW		 3744
3745 btrfs_set_file_extent_offset(leaf, extent,
3746 datao);
3747 btrfs_set_file_extent_num_bytes(leaf, extent,
3748 datal);
fcebe456 3749
c5c9cd4d 3750 if (disko) {
82fa113f 3751 struct btrfs_ref ref = { 0 };
c5c9cd4d 3752 inode_add_bytes(inode, datal);
82fa113f
QW		 3753 btrfs_init_generic_ref(&ref,
3754 BTRFS_ADD_DELAYED_REF, disko,
3755 diskl, 0);
3756 btrfs_init_data_ref(&ref,
3757 root->root_key.objectid,
3758 btrfs_ino(BTRFS_I(inode)),
3759 new_key.offset - datao);
3760 ret = btrfs_inc_extent_ref(trans, &ref);
79787eaa
JM		 3761 if (ret) {
3762 btrfs_abort_transaction(trans,
79787eaa 3763 ret);
3a45bb20 3764 btrfs_end_transaction(trans);
79787eaa
JM		 3765 goto out;
3766
3767 }
f46b5a66 3768 }
c5c9cd4d
SW		 3769 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3770 u64 skip = 0;
3771 u64 trim = 0;
ed958762 3772
c5c9cd4d
SW		 3773 if (off > key.offset) {
3774 skip = off - key.offset;
3775 new_key.offset += skip;
3776 }
d397712b 3777
aa42ffd9
LB		 3778 if (key.offset + datal > off + len)
3779 trim = key.offset + datal - (off + len);
d397712b 3780
c5c9cd4d 3781 if (comp && (skip || trim)) {
c5c9cd4d 3782 ret = -EINVAL;
3a45bb20 3783 btrfs_end_transaction(trans);
c5c9cd4d
SW		 3784 goto out;
3785 }
3786 size -= skip + trim;
3787 datal -= skip + trim;
a22285a6 3788
4a0ab9d7 3789 ret = clone_copy_inline_extent(inode,
8039d87d
FM		 3790 trans, path,
3791 &new_key,
3792 drop_start,
3793 datal,
3794 skip, size, buf);
79787eaa 3795 if (ret) {
3f9e3df8 3796 if (ret != -EOPNOTSUPP)
3a29bc09 3797 btrfs_abort_transaction(trans,
8039d87d 3798 ret);
3a45bb20 3799 btrfs_end_transaction(trans);
79787eaa
JM		 3800 goto out;
3801 }
c5c9cd4d
SW		 3802 leaf = path->nodes[0];
3803 slot = path->slots[0];
f46b5a66 3804 }
c5c9cd4d 3805
7ffbb598
FM		 3806 /* If we have an implicit hole (NO_HOLES feature). */
3807 if (drop_start < new_key.offset)
a2f392e4 3808 clone_update_extent_map(BTRFS_I(inode), trans,
14f59796 3809 NULL, drop_start,
7ffbb598
FM		 3810 new_key.offset - drop_start);
3811
a2f392e4
NB		 3812 clone_update_extent_map(BTRFS_I(inode), trans,
3813 path, 0, 0);
7ffbb598 3814
c5c9cd4d 3815 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3816 btrfs_release_path(path);
c5c9cd4d 3817
62e2390e 3818 last_dest_end = ALIGN(new_key.offset + datal,
0b246afa 3819 fs_info->sectorsize);
f82a9901
FM		 3820 ret = clone_finish_inode_update(trans, inode,
3821 last_dest_end,
1c919a5e
MF		 3822 destoff, olen,
3823 no_time_update);
f82a9901 3824 if (ret)
79787eaa 3825 goto out;
2c463823
FM		 3826 if (new_key.offset + datal >= destoff + len)
3827 break;
a22285a6 3828 }
b3b4aa74 3829 btrfs_release_path(path);
df858e76 3830 key.offset = next_key_min_offset;
69ae5e44
WX		 3831
3832 if (fatal_signal_pending(current)) {
3833 ret = -EINTR;
3834 goto out;
3835 }
f46b5a66 3836 }
f46b5a66 3837 ret = 0;
32b7c687 3838
f82a9901
FM		 3839 if (last_dest_end < destoff + len) {
3840 /*
3841 * We have an implicit hole (NO_HOLES feature is enabled) that
3842 * fully or partially overlaps our cloning range at its end.
3843 */
3844 btrfs_release_path(path);
3845
3846 /*
3847 * 1 - remove extent(s)
3848 * 1 - inode update
3849 */
3850 trans = btrfs_start_transaction(root, 2);
3851 if (IS_ERR(trans)) {
3852 ret = PTR_ERR(trans);
3853 goto out;
3854 }
3855 ret = btrfs_drop_extents(trans, root, inode,
3856 last_dest_end, destoff + len, 1);
3857 if (ret) {
3858 if (ret != -EOPNOTSUPP)
66642832 3859 btrfs_abort_transaction(trans, ret);
3a45bb20 3860 btrfs_end_transaction(trans);
f82a9901
FM		 3861 goto out;
3862 }
a2f392e4
NB		 3863 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3864 last_dest_end,
3865 destoff + len - last_dest_end);
f82a9901 3866 ret = clone_finish_inode_update(trans, inode, destoff + len,
1c919a5e 3867 destoff, olen, no_time_update);
f82a9901
FM		 3868 }
3869
f46b5a66 3870out:
32b7c687 3871 btrfs_free_path(path);
15351955 3872 kvfree(buf);
32b7c687
MF		 3873 return ret;
3874}
3875
3db11b2e
ZB		 3876static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3877 u64 off, u64 olen, u64 destoff)
32b7c687 3878{
54563d41 3879 struct inode *inode = file_inode(file);
3db11b2e 3880 struct inode *src = file_inode(file_src);
0b246afa 3881 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
32b7c687
MF		 3882 int ret;
3883 u64 len = olen;
0b246afa 3884 u64 bs = fs_info->sb->s_blocksize;
32b7c687
MF		 3885
3886 /*
3887 * TODO:
3888 * - split compressed inline extents. annoying: we need to
3889 * decompress into destination's address_space (the file offset
3890 * may change, so source mapping won't do), then recompress (or
3891 * otherwise reinsert) a subrange.
00fdf13a
LB		 3892 *
3893 * - split destination inode's inline extents. The inline extents can
3894 * be either compressed or non-compressed.
32b7c687
MF		 3895 */
3896
ac765f83 3897 /*
34a28e3d
FM		 3898 * VFS's generic_remap_file_range_prep() protects us from cloning the
3899 * eof block into the middle of a file, which would result in corruption
3900 * if the file size is not blocksize aligned. So we don't need to check
3901 * for that case here.
ac765f83 3902 */
34a28e3d 3903 if (off + len == src->i_size)
32b7c687 3904 len = ALIGN(src->i_size, bs) - off;
32b7c687
MF		 3905
3906 if (destoff > inode->i_size) {
f7fa1107
FM		 3907 const u64 wb_start = ALIGN_DOWN(inode->i_size, bs);
3908
32b7c687
MF		 3909 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3910 if (ret)
34a28e3d 3911 return ret;
f7fa1107
FM		 3912 /*
3913 * We may have truncated the last block if the inode's size is
3914 * not sector size aligned, so we need to wait for writeback to
3915 * complete before proceeding further, otherwise we can race
3916 * with cloning and attempt to increment a reference to an
3917 * extent that no longer exists (writeback completed right after
3918 * we found the previous extent covering eof and before we
3919 * attempted to increment its reference count).
3920 */
3921 ret = btrfs_wait_ordered_range(inode, wb_start,
3922 destoff - wb_start);
3923 if (ret)
3924 return ret;
32b7c687
MF		 3925 }
3926
c125b8bf 3927 /*
d8b55242
FM		 3928 * Lock destination range to serialize with concurrent readpages() and
3929 * source range to serialize with relocation.
c125b8bf 3930 */
d8b55242 3931 btrfs_double_extent_lock(src, off, inode, destoff, len);
1c919a5e 3932 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
d8b55242 3933 btrfs_double_extent_unlock(src, off, inode, destoff, len);
c125b8bf
FM		 3934 /*
3935 * Truncate page cache pages so that future reads will see the cloned
3936 * data immediately and not the previous data.
3937 */
65bfa658 3938 truncate_inode_pages_range(&inode->i_data,
09cbfeaf
KS		 3939 round_down(destoff, PAGE_SIZE),
3940 round_up(destoff + len, PAGE_SIZE) - 1);
34a28e3d
FM		 3941
3942 return ret;
3943}
3944
3945static int btrfs_remap_file_range_prep(struct file *file_in, loff_t pos_in,
3946 struct file *file_out, loff_t pos_out,
3947 loff_t *len, unsigned int remap_flags)
3948{
3949 struct inode *inode_in = file_inode(file_in);
3950 struct inode *inode_out = file_inode(file_out);
3951 u64 bs = BTRFS_I(inode_out)->root->fs_info->sb->s_blocksize;
3952 bool same_inode = inode_out == inode_in;
3953 u64 wb_len;
3954 int ret;
3955
3956 if (!(remap_flags & REMAP_FILE_DEDUP)) {
3957 struct btrfs_root *root_out = BTRFS_I(inode_out)->root;
3958
3959 if (btrfs_root_readonly(root_out))
3960 return -EROFS;
3961
3962 if (file_in->f_path.mnt != file_out->f_path.mnt ||
3963 inode_in->i_sb != inode_out->i_sb)
3964 return -EXDEV;
3965 }
3966
500710d3
FM		 3967 /* don't make the dst file partly checksummed */
3968 if ((BTRFS_I(inode_in)->flags & BTRFS_INODE_NODATASUM) !=
3969 (BTRFS_I(inode_out)->flags & BTRFS_INODE_NODATASUM)) {
7984ae52 3970 return -EINVAL;
500710d3
FM		 3971 }
3972
34a28e3d
FM		 3973 /*
3974 * Now that the inodes are locked, we need to start writeback ourselves
3975 * and can not rely on the writeback from the VFS's generic helper
3976 * generic_remap_file_range_prep() because:
3977 *
3978 * 1) For compression we must call filemap_fdatawrite_range() range
3979 * twice (btrfs_fdatawrite_range() does it for us), and the generic
3980 * helper only calls it once;
3981 *
3982 * 2) filemap_fdatawrite_range(), called by the generic helper only
3983 * waits for the writeback to complete, i.e. for IO to be done, and
3984 * not for the ordered extents to complete. We need to wait for them
3985 * to complete so that new file extent items are in the fs tree.
3986 */
3987 if (*len == 0 && !(remap_flags & REMAP_FILE_DEDUP))
3988 wb_len = ALIGN(inode_in->i_size, bs) - ALIGN_DOWN(pos_in, bs);
3989 else
3990 wb_len = ALIGN(*len, bs);
3991
3992 /*
3993 * Since we don't lock ranges, wait for ongoing lockless dio writes (as
3994 * any in progress could create its ordered extents after we wait for
3995 * existing ordered extents below).
3996 */
3997 inode_dio_wait(inode_in);
293a8489 3998 if (!same_inode)
34a28e3d
FM		 3999 inode_dio_wait(inode_out);
4000
4001 ret = btrfs_wait_ordered_range(inode_in, ALIGN_DOWN(pos_in, bs),
4002 wb_len);
4003 if (ret < 0)
7984ae52 4004 return ret;
34a28e3d
FM		 4005 ret = btrfs_wait_ordered_range(inode_out, ALIGN_DOWN(pos_out, bs),
4006 wb_len);
4007 if (ret < 0)
7984ae52 4008 return ret;
34a28e3d 4009
7984ae52 4010 return generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out,
34a28e3d 4011 len, remap_flags);
3db11b2e
ZB		 4012}
4013
42ec3d4c
DW		 4014loff_t btrfs_remap_file_range(struct file *src_file, loff_t off,
4015 struct file *dst_file, loff_t destoff, loff_t len,
2e5dfc99 4016 unsigned int remap_flags)
3db11b2e 4017{
34a28e3d
FM		 4018 struct inode *src_inode = file_inode(src_file);
4019 struct inode *dst_inode = file_inode(dst_file);
4020 bool same_inode = dst_inode == src_inode;
42ec3d4c
DW		 4021 int ret;
4022
2e5dfc99
DW		 4023 if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY))
4024 return -EINVAL;
4025
7984ae52
GR		 4026 if (same_inode)
4027 inode_lock(src_inode);
4028 else
4029 lock_two_nondirectories(src_inode, dst_inode);
4030
34a28e3d
FM		 4031 ret = btrfs_remap_file_range_prep(src_file, off, dst_file, destoff,
4032 &len, remap_flags);
4033 if (ret < 0 || len == 0)
7984ae52 4034 goto out_unlock;
2e5dfc99 4035
34a28e3d
FM		 4036 if (remap_flags & REMAP_FILE_DEDUP)
4037 ret = btrfs_extent_same(src_inode, off, len, dst_inode, destoff);
4038 else
42ec3d4c 4039 ret = btrfs_clone_files(dst_file, src_file, off, len, destoff);
34a28e3d 4040
7984ae52 4041out_unlock:
34a28e3d
FM		 4042 if (same_inode)
4043 inode_unlock(src_inode);
4044 else
4ea748e1 4045 unlock_two_nondirectories(src_inode, dst_inode);
34a28e3d 4046
42ec3d4c 4047 return ret < 0 ? ret : len;
c5c9cd4d
SW		 4048}
4049
6ef5ed0d
JB		 4050static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4051{
496ad9aa 4052 struct inode *inode = file_inode(file);
0b246afa 4053 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6ef5ed0d
JB		 4054 struct btrfs_root *root = BTRFS_I(inode)->root;
4055 struct btrfs_root *new_root;
4056 struct btrfs_dir_item *di;
4057 struct btrfs_trans_handle *trans;
4058 struct btrfs_path *path;
4059 struct btrfs_key location;
4060 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 4061 u64 objectid = 0;
4062 u64 dir_id;
3c04ce01 4063 int ret;
6ef5ed0d
JB		 4064
4065 if (!capable(CAP_SYS_ADMIN))
4066 return -EPERM;
4067
3c04ce01
MX		 4068 ret = mnt_want_write_file(file);
4069 if (ret)
4070 return ret;
4071
4072 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4073 ret = -EFAULT;
4074 goto out;
4075 }
6ef5ed0d
JB		 4076
4077 if (!objectid)
1cecf579 4078 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d
JB		 4079
4080 location.objectid = objectid;
4081 location.type = BTRFS_ROOT_ITEM_KEY;
4082 location.offset = (u64)-1;
4083
0b246afa 4084 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
3c04ce01
MX		 4085 if (IS_ERR(new_root)) {
4086 ret = PTR_ERR(new_root);
4087 goto out;
4088 }
4fd786e6 4089 if (!is_fstree(new_root->root_key.objectid)) {
6d6d2829 4090 ret = -ENOENT;
4091 goto out;
4092 }
6ef5ed0d 4093
6ef5ed0d 4094 path = btrfs_alloc_path();
3c04ce01
MX		 4095 if (!path) {
4096 ret = -ENOMEM;
4097 goto out;
4098 }
6ef5ed0d
JB		 4099 path->leave_spinning = 1;
4100
4101 trans = btrfs_start_transaction(root, 1);
98d5dc13 4102 if (IS_ERR(trans)) {
6ef5ed0d 4103 btrfs_free_path(path);
3c04ce01
MX		 4104 ret = PTR_ERR(trans);
4105 goto out;
6ef5ed0d
JB		 4106 }
4107
0b246afa
JM		 4108 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4109 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
6ef5ed0d 4110 dir_id, "default", 7, 1);
cf1e99a4 4111 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d 4112 btrfs_free_path(path);
3a45bb20 4113 btrfs_end_transaction(trans);
0b246afa 4114 btrfs_err(fs_info,
5d163e0e 4115 "Umm, you don't have the default diritem, this isn't going to work");
3c04ce01
MX		 4116 ret = -ENOENT;
4117 goto out;
6ef5ed0d
JB		 4118 }
4119
4120 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4121 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4122 btrfs_mark_buffer_dirty(path->nodes[0]);
4123 btrfs_free_path(path);
4124
0b246afa 4125 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3a45bb20 4126 btrfs_end_transaction(trans);
3c04ce01
MX		 4127out:
4128 mnt_drop_write_file(file);
4129 return ret;
6ef5ed0d
JB		 4130}
4131
c065f5b1
SY		 4132static void get_block_group_info(struct list_head *groups_list,
4133 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 4134{
4135 struct btrfs_block_group_cache *block_group;
4136
4137 space->total_bytes = 0;
4138 space->used_bytes = 0;
4139 space->flags = 0;
4140 list_for_each_entry(block_group, groups_list, list) {
4141 space->flags = block_group->flags;
4142 space->total_bytes += block_group->key.offset;
4143 space->used_bytes +=
4144 btrfs_block_group_used(&block_group->item);
4145 }
4146}
4147
2ff7e61e
JM		 4148static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4149 void __user *arg)
1406e432
JB		 4150{
4151 struct btrfs_ioctl_space_args space_args;
4152 struct btrfs_ioctl_space_info space;
4153 struct btrfs_ioctl_space_info *dest;
7fde62bf 4154 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 4155 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 4156 struct btrfs_space_info *info;
315d8e98
CIK		 4157 static const u64 types[] = {
4158 BTRFS_BLOCK_GROUP_DATA,
4159 BTRFS_BLOCK_GROUP_SYSTEM,
4160 BTRFS_BLOCK_GROUP_METADATA,
4161 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4162 };
bf5fc093 4163 int num_types = 4;
7fde62bf 4164 int alloc_size;
1406e432 4165 int ret = 0;
51788b1b 4166 u64 slot_count = 0;
bf5fc093 4167 int i, c;
1406e432
JB		 4168
4169 if (copy_from_user(&space_args,
4170 (struct btrfs_ioctl_space_args __user *)arg,
4171 sizeof(space_args)))
4172 return -EFAULT;
4173
bf5fc093
JB		 4174 for (i = 0; i < num_types; i++) {
4175 struct btrfs_space_info *tmp;
4176
4177 info = NULL;
4178 rcu_read_lock();
0b246afa 4179 list_for_each_entry_rcu(tmp, &fs_info->space_info,
bf5fc093
JB		 4180 list) {
4181 if (tmp->flags == types[i]) {
4182 info = tmp;
4183 break;
4184 }
4185 }
4186 rcu_read_unlock();
4187
4188 if (!info)
4189 continue;
4190
4191 down_read(&info->groups_sem);
4192 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4193 if (!list_empty(&info->block_groups[c]))
4194 slot_count++;
4195 }
4196 up_read(&info->groups_sem);
4197 }
7fde62bf 4198
36523e95
DS		 4199 /*
4200 * Global block reserve, exported as a space_info
4201 */
4202 slot_count++;
4203
7fde62bf
CM		 4204 /* space_slots == 0 means they are asking for a count */
4205 if (space_args.space_slots == 0) {
4206 space_args.total_spaces = slot_count;
4207 goto out;
4208 }
bf5fc093 4209
51788b1b 4210 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 4211
7fde62bf 4212 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 4213
7fde62bf
CM		 4214 /* we generally have at most 6 or so space infos, one for each raid
4215 * level. So, a whole page should be more than enough for everyone
4216 */
09cbfeaf 4217 if (alloc_size > PAGE_SIZE)
7fde62bf
CM		 4218 return -ENOMEM;
4219
1406e432 4220 space_args.total_spaces = 0;
8d2db785 4221 dest = kmalloc(alloc_size, GFP_KERNEL);
7fde62bf
CM		 4222 if (!dest)
4223 return -ENOMEM;
4224 dest_orig = dest;
1406e432 4225
7fde62bf 4226 /* now we have a buffer to copy into */
bf5fc093
JB		 4227 for (i = 0; i < num_types; i++) {
4228 struct btrfs_space_info *tmp;
4229
51788b1b
DR		 4230 if (!slot_count)
4231 break;
4232
bf5fc093
JB		 4233 info = NULL;
4234 rcu_read_lock();
0b246afa 4235 list_for_each_entry_rcu(tmp, &fs_info->space_info,
bf5fc093
JB		 4236 list) {
4237 if (tmp->flags == types[i]) {
4238 info = tmp;
4239 break;
4240 }
4241 }
4242 rcu_read_unlock();
7fde62bf 4243
bf5fc093
JB		 4244 if (!info)
4245 continue;
4246 down_read(&info->groups_sem);
4247 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4248 if (!list_empty(&info->block_groups[c])) {
c065f5b1
SY		 4249 get_block_group_info(&info->block_groups[c],
4250 &space);
bf5fc093
JB		 4251 memcpy(dest, &space, sizeof(space));
4252 dest++;
4253 space_args.total_spaces++;
51788b1b 4254 slot_count--;
bf5fc093 4255 }
51788b1b
DR		 4256 if (!slot_count)
4257 break;
bf5fc093
JB		 4258 }
4259 up_read(&info->groups_sem);
1406e432 4260 }
1406e432 4261
36523e95
DS		 4262 /*
4263 * Add global block reserve
4264 */
4265 if (slot_count) {
0b246afa 4266 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
36523e95
DS		 4267
4268 spin_lock(&block_rsv->lock);
4269 space.total_bytes = block_rsv->size;
4270 space.used_bytes = block_rsv->size - block_rsv->reserved;
4271 spin_unlock(&block_rsv->lock);
4272 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4273 memcpy(dest, &space, sizeof(space));
4274 space_args.total_spaces++;
4275 }
4276
2eec6c81 4277 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 4278 (arg + sizeof(struct btrfs_ioctl_space_args));
4279
4280 if (copy_to_user(user_dest, dest_orig, alloc_size))
4281 ret = -EFAULT;
4282
4283 kfree(dest_orig);
4284out:
4285 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 4286 ret = -EFAULT;
4287
4288 return ret;
4289}
4290
9a8c28be
MX		 4291static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4292 void __user *argp)
46204592 4293{
46204592
SW		 4294 struct btrfs_trans_handle *trans;
4295 u64 transid;
db5b493a 4296 int ret;
46204592 4297
d4edf39b 4298 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 4299 if (IS_ERR(trans)) {
4300 if (PTR_ERR(trans) != -ENOENT)
4301 return PTR_ERR(trans);
4302
4303 /* No running transaction, don't bother */
4304 transid = root->fs_info->last_trans_committed;
4305 goto out;
4306 }
46204592 4307 transid = trans->transid;
3a45bb20 4308 ret = btrfs_commit_transaction_async(trans, 0);
8b2b2d3c 4309 if (ret) {
3a45bb20 4310 btrfs_end_transaction(trans);
db5b493a 4311 return ret;
8b2b2d3c 4312 }
ff7c1d33 4313out:
46204592
SW		 4314 if (argp)
4315 if (copy_to_user(argp, &transid, sizeof(transid)))
4316 return -EFAULT;
4317 return 0;
4318}
4319
2ff7e61e 4320static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
9a8c28be 4321 void __user *argp)
46204592 4322{
46204592
SW		 4323 u64 transid;
4324
4325 if (argp) {
4326 if (copy_from_user(&transid, argp, sizeof(transid)))
4327 return -EFAULT;
4328 } else {
4329 transid = 0; /* current trans */
4330 }
2ff7e61e 4331 return btrfs_wait_for_commit(fs_info, transid);
46204592
SW		 4332}
4333
b8e95489 4334static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 4335{
0b246afa 4336 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
475f6387 4337 struct btrfs_ioctl_scrub_args *sa;
b8e95489 4338 int ret;
475f6387
JS		 4339
4340 if (!capable(CAP_SYS_ADMIN))
4341 return -EPERM;
4342
4343 sa = memdup_user(arg, sizeof(*sa));
4344 if (IS_ERR(sa))
4345 return PTR_ERR(sa);
4346
b8e95489
MX		 4347 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4348 ret = mnt_want_write_file(file);
4349 if (ret)
4350 goto out;
4351 }
4352
0b246afa 4353 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 4354 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4355 0);
475f6387 4356
06fe39ab 4357 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
475f6387
JS		 4358 ret = -EFAULT;
4359
b8e95489
MX		 4360 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4361 mnt_drop_write_file(file);
4362out:
475f6387
JS		 4363 kfree(sa);
4364 return ret;
4365}
4366
2ff7e61e 4367static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
475f6387
JS		 4368{
4369 if (!capable(CAP_SYS_ADMIN))
4370 return -EPERM;
4371
2ff7e61e 4372 return btrfs_scrub_cancel(fs_info);
475f6387
JS		 4373}
4374
2ff7e61e 4375static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
475f6387
JS		 4376 void __user *arg)
4377{
4378 struct btrfs_ioctl_scrub_args *sa;
4379 int ret;
4380
4381 if (!capable(CAP_SYS_ADMIN))
4382 return -EPERM;
4383
4384 sa = memdup_user(arg, sizeof(*sa));
4385 if (IS_ERR(sa))
4386 return PTR_ERR(sa);
4387
2ff7e61e 4388 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
475f6387 4389
4fa99b00 4390 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
475f6387
JS		 4391 ret = -EFAULT;
4392
4393 kfree(sa);
4394 return ret;
4395}
4396
2ff7e61e 4397static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
b27f7c0c 4398 void __user *arg)
c11d2c23
SB		 4399{
4400 struct btrfs_ioctl_get_dev_stats *sa;
4401 int ret;
4402
c11d2c23
SB		 4403 sa = memdup_user(arg, sizeof(*sa));
4404 if (IS_ERR(sa))
4405 return PTR_ERR(sa);
4406
b27f7c0c
DS		 4407 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4408 kfree(sa);
4409 return -EPERM;
4410 }
4411
2ff7e61e 4412 ret = btrfs_get_dev_stats(fs_info, sa);
c11d2c23 4413
eee99577 4414 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
c11d2c23
SB		 4415 ret = -EFAULT;
4416
4417 kfree(sa);
4418 return ret;
4419}
4420
2ff7e61e
JM		 4421static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4422 void __user *arg)
3f6bcfbd
SB		 4423{
4424 struct btrfs_ioctl_dev_replace_args *p;
4425 int ret;
4426
4427 if (!capable(CAP_SYS_ADMIN))
4428 return -EPERM;
4429
4430 p = memdup_user(arg, sizeof(*p));
4431 if (IS_ERR(p))
4432 return PTR_ERR(p);
4433
4434 switch (p->cmd) {
4435 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
bc98a42c 4436 if (sb_rdonly(fs_info->sb)) {
adfa97cb
ID		 4437 ret = -EROFS;
4438 goto out;
4439 }
171938e5 4440 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
e57138b3 4441 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd 4442 } else {
2ff7e61e 4443 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
171938e5 4444 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3f6bcfbd
SB		 4445 }
4446 break;
4447 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
0b246afa 4448 btrfs_dev_replace_status(fs_info, p);
3f6bcfbd
SB		 4449 ret = 0;
4450 break;
4451 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
17d202b9 4452 p->result = btrfs_dev_replace_cancel(fs_info);
97282031 4453 ret = 0;
3f6bcfbd
SB		 4454 break;
4455 default:
4456 ret = -EINVAL;
4457 break;
4458 }
4459
d3a53286 4460 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3f6bcfbd 4461 ret = -EFAULT;
adfa97cb 4462out:
3f6bcfbd
SB		 4463 kfree(p);
4464 return ret;
4465}
4466
d7728c96
JS		 4467static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4468{
4469 int ret = 0;
4470 int i;
740c3d22 4471 u64 rel_ptr;
d7728c96 4472 int size;
806468f8 4473 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 4474 struct inode_fs_paths *ipath = NULL;
4475 struct btrfs_path *path;
4476
82b22ac8 4477 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 4478 return -EPERM;
4479
4480 path = btrfs_alloc_path();
4481 if (!path) {
4482 ret = -ENOMEM;
4483 goto out;
4484 }
4485
4486 ipa = memdup_user(arg, sizeof(*ipa));
4487 if (IS_ERR(ipa)) {
4488 ret = PTR_ERR(ipa);
4489 ipa = NULL;
4490 goto out;
4491 }
4492
4493 size = min_t(u32, ipa->size, 4096);
4494 ipath = init_ipath(size, root, path);
4495 if (IS_ERR(ipath)) {
4496 ret = PTR_ERR(ipath);
4497 ipath = NULL;
4498 goto out;
4499 }
4500
4501 ret = paths_from_inode(ipa->inum, ipath);
4502 if (ret < 0)
4503 goto out;
4504
4505 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 4506 rel_ptr = ipath->fspath->val[i] -
4507 (u64)(unsigned long)ipath->fspath->val;
740c3d22 4508 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 4509 }
4510
718dc5fa
OS		 4511 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4512 ipath->fspath, size);
d7728c96
JS		 4513 if (ret) {
4514 ret = -EFAULT;
4515 goto out;
4516 }
4517
4518out:
4519 btrfs_free_path(path);
4520 free_ipath(ipath);
4521 kfree(ipa);
4522
4523 return ret;
4524}
4525
4526static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4527{
4528 struct btrfs_data_container *inodes = ctx;
4529 const size_t c = 3 * sizeof(u64);
4530
4531 if (inodes->bytes_left >= c) {
4532 inodes->bytes_left -= c;
4533 inodes->val[inodes->elem_cnt] = inum;
4534 inodes->val[inodes->elem_cnt + 1] = offset;
4535 inodes->val[inodes->elem_cnt + 2] = root;
4536 inodes->elem_cnt += 3;
4537 } else {
4538 inodes->bytes_missing += c - inodes->bytes_left;
4539 inodes->bytes_left = 0;
4540 inodes->elem_missed += 3;
4541 }
4542
4543 return 0;
4544}
4545
2ff7e61e 4546static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
d24a67b2 4547 void __user *arg, int version)
d7728c96
JS		 4548{
4549 int ret = 0;
4550 int size;
d7728c96
JS		 4551 struct btrfs_ioctl_logical_ino_args *loi;
4552 struct btrfs_data_container *inodes = NULL;
4553 struct btrfs_path *path = NULL;
d24a67b2 4554 bool ignore_offset;
d7728c96
JS		 4555
4556 if (!capable(CAP_SYS_ADMIN))
4557 return -EPERM;
4558
4559 loi = memdup_user(arg, sizeof(*loi));
7b9ea627
SV		 4560 if (IS_ERR(loi))
4561 return PTR_ERR(loi);
d7728c96 4562
d24a67b2
ZB		 4563 if (version == 1) {
4564 ignore_offset = false;
b115e3bc 4565 size = min_t(u32, loi->size, SZ_64K);
d24a67b2
ZB		 4566 } else {
4567 /* All reserved bits must be 0 for now */
4568 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4569 ret = -EINVAL;
4570 goto out_loi;
4571 }
4572 /* Only accept flags we have defined so far */
4573 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4574 ret = -EINVAL;
4575 goto out_loi;
4576 }
4577 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
b115e3bc 4578 size = min_t(u32, loi->size, SZ_16M);
d24a67b2
ZB		 4579 }
4580
d7728c96
JS		 4581 path = btrfs_alloc_path();
4582 if (!path) {
4583 ret = -ENOMEM;
4584 goto out;
4585 }
4586
d7728c96
JS		 4587 inodes = init_data_container(size);
4588 if (IS_ERR(inodes)) {
4589 ret = PTR_ERR(inodes);
4590 inodes = NULL;
4591 goto out;
4592 }
4593
2ff7e61e 4594 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
d24a67b2 4595 build_ino_list, inodes, ignore_offset);
df031f07 4596 if (ret == -EINVAL)
d7728c96
JS		 4597 ret = -ENOENT;
4598 if (ret < 0)
4599 goto out;
4600
718dc5fa
OS		 4601 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4602 size);
d7728c96
JS		 4603 if (ret)
4604 ret = -EFAULT;
4605
4606out:
4607 btrfs_free_path(path);
f54de068 4608 kvfree(inodes);
d24a67b2 4609out_loi:
d7728c96
JS		 4610 kfree(loi);
4611
4612 return ret;
4613}
4614
008ef096 4615void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
c9e9f97b
ID		 4616 struct btrfs_ioctl_balance_args *bargs)
4617{
4618 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4619
4620 bargs->flags = bctl->flags;
4621
3009a62f 4622 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
837d5b6e
ID		 4623 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4624 if (atomic_read(&fs_info->balance_pause_req))
4625 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 4626 if (atomic_read(&fs_info->balance_cancel_req))
4627 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 4628
c9e9f97b
ID		 4629 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4630 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4631 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce 4632
008ef096
DS		 4633 spin_lock(&fs_info->balance_lock);
4634 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4635 spin_unlock(&fs_info->balance_lock);
c9e9f97b
ID		 4636}
4637
9ba1f6e4 4638static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 4639{
496ad9aa 4640 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 4641 struct btrfs_fs_info *fs_info = root->fs_info;
4642 struct btrfs_ioctl_balance_args *bargs;
4643 struct btrfs_balance_control *bctl;
ed0fb78f 4644 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 4645 int ret;
4646
4647 if (!capable(CAP_SYS_ADMIN))
4648 return -EPERM;
4649
e54bfa31 4650 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 4651 if (ret)
4652 return ret;
4653
ed0fb78f 4654again:
171938e5 4655 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
ed0fb78f
ID		 4656 mutex_lock(&fs_info->balance_mutex);
4657 need_unlock = true;
4658 goto locked;
4659 }
4660
4661 /*
01327610 4662 * mut. excl. ops lock is locked. Three possibilities:
ed0fb78f
ID		 4663 * (1) some other op is running
4664 * (2) balance is running
4665 * (3) balance is paused -- special case (think resume)
4666 */
c9e9f97b 4667 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 4668 if (fs_info->balance_ctl) {
4669 /* this is either (2) or (3) */
3009a62f 4670 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
ed0fb78f 4671 mutex_unlock(&fs_info->balance_mutex);
dccdb07b
DS		 4672 /*
4673 * Lock released to allow other waiters to continue,
4674 * we'll reexamine the status again.
4675 */
ed0fb78f
ID		 4676 mutex_lock(&fs_info->balance_mutex);
4677
4678 if (fs_info->balance_ctl &&
3009a62f 4679 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
ed0fb78f
ID		 4680 /* this is (3) */
4681 need_unlock = false;
4682 goto locked;
4683 }
4684
4685 mutex_unlock(&fs_info->balance_mutex);
ed0fb78f
ID		 4686 goto again;
4687 } else {
4688 /* this is (2) */
4689 mutex_unlock(&fs_info->balance_mutex);
4690 ret = -EINPROGRESS;
4691 goto out;
4692 }
4693 } else {
4694 /* this is (1) */
4695 mutex_unlock(&fs_info->balance_mutex);
e57138b3 4696 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 4697 goto out;
4698 }
4699
4700locked:
171938e5 4701 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
c9e9f97b
ID		 4702
4703 if (arg) {
4704 bargs = memdup_user(arg, sizeof(*bargs));
4705 if (IS_ERR(bargs)) {
4706 ret = PTR_ERR(bargs);
ed0fb78f 4707 goto out_unlock;
c9e9f97b 4708 }
de322263
ID		 4709
4710 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4711 if (!fs_info->balance_ctl) {
4712 ret = -ENOTCONN;
4713 goto out_bargs;
4714 }
4715
4716 bctl = fs_info->balance_ctl;
4717 spin_lock(&fs_info->balance_lock);
4718 bctl->flags |= BTRFS_BALANCE_RESUME;
4719 spin_unlock(&fs_info->balance_lock);
4720
4721 goto do_balance;
4722 }
c9e9f97b
ID		 4723 } else {
4724 bargs = NULL;
4725 }
4726
ed0fb78f 4727 if (fs_info->balance_ctl) {
837d5b6e
ID		 4728 ret = -EINPROGRESS;
4729 goto out_bargs;
4730 }
4731
8d2db785 4732 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
c9e9f97b
ID		 4733 if (!bctl) {
4734 ret = -ENOMEM;
4735 goto out_bargs;
4736 }
4737
c9e9f97b
ID		 4738 if (arg) {
4739 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4740 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4741 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4742
4743 bctl->flags = bargs->flags;
f43ffb60
ID		 4744 } else {
4745 /* balance everything - no filters */
4746 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 4747 }
4748
8eb93459
DS		 4749 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4750 ret = -EINVAL;
0f89abf5 4751 goto out_bctl;
8eb93459
DS		 4752 }
4753
de322263 4754do_balance:
c9e9f97b 4755 /*
149196a2
DS		 4756 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
4757 * btrfs_balance. bctl is freed in reset_balance_state, or, if
4758 * restriper was paused all the way until unmount, in free_fs_info.
4759 * The flag should be cleared after reset_balance_state.
c9e9f97b 4760 */
ed0fb78f
ID		 4761 need_unlock = false;
4762
6fcf6e2b 4763 ret = btrfs_balance(fs_info, bctl, bargs);
0f89abf5 4764 bctl = NULL;
ed0fb78f 4765
d00c2d9c 4766 if ((ret == 0 || ret == -ECANCELED) && arg) {
c9e9f97b
ID		 4767 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4768 ret = -EFAULT;
4769 }
4770
0f89abf5
CE		 4771out_bctl:
4772 kfree(bctl);
c9e9f97b
ID		 4773out_bargs:
4774 kfree(bargs);
ed0fb78f 4775out_unlock:
c9e9f97b 4776 mutex_unlock(&fs_info->balance_mutex);
ed0fb78f 4777 if (need_unlock)
171938e5 4778 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
ed0fb78f 4779out:
e54bfa31 4780 mnt_drop_write_file(file);
c9e9f97b
ID		 4781 return ret;
4782}
4783
2ff7e61e 4784static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
837d5b6e
ID		 4785{
4786 if (!capable(CAP_SYS_ADMIN))
4787 return -EPERM;
4788
4789 switch (cmd) {
4790 case BTRFS_BALANCE_CTL_PAUSE:
0b246afa 4791 return btrfs_pause_balance(fs_info);
a7e99c69 4792 case BTRFS_BALANCE_CTL_CANCEL:
0b246afa 4793 return btrfs_cancel_balance(fs_info);
837d5b6e
ID		 4794 }
4795
4796 return -EINVAL;
4797}
4798
2ff7e61e 4799static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
19a39dce
ID		 4800 void __user *arg)
4801{
19a39dce
ID		 4802 struct btrfs_ioctl_balance_args *bargs;
4803 int ret = 0;
4804
4805 if (!capable(CAP_SYS_ADMIN))
4806 return -EPERM;
4807
4808 mutex_lock(&fs_info->balance_mutex);
4809 if (!fs_info->balance_ctl) {
4810 ret = -ENOTCONN;
4811 goto out;
4812 }
4813
8d2db785 4814 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
19a39dce
ID		 4815 if (!bargs) {
4816 ret = -ENOMEM;
4817 goto out;
4818 }
4819
008ef096 4820 btrfs_update_ioctl_balance_args(fs_info, bargs);
19a39dce
ID		 4821
4822 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4823 ret = -EFAULT;
4824
4825 kfree(bargs);
4826out:
4827 mutex_unlock(&fs_info->balance_mutex);
4828 return ret;
4829}
4830
905b0dda 4831static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4832{
0b246afa
JM		 4833 struct inode *inode = file_inode(file);
4834 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5d13a37b 4835 struct btrfs_ioctl_quota_ctl_args *sa;
5d13a37b 4836 int ret;
5d13a37b
AJ		 4837
4838 if (!capable(CAP_SYS_ADMIN))
4839 return -EPERM;
4840
905b0dda
MX		 4841 ret = mnt_want_write_file(file);
4842 if (ret)
4843 return ret;
5d13a37b
AJ		 4844
4845 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4846 if (IS_ERR(sa)) {
4847 ret = PTR_ERR(sa);
4848 goto drop_write;
4849 }
5d13a37b 4850
0b246afa 4851 down_write(&fs_info->subvol_sem);
5d13a37b
AJ		 4852
4853 switch (sa->cmd) {
4854 case BTRFS_QUOTA_CTL_ENABLE:
340f1aa2 4855 ret = btrfs_quota_enable(fs_info);
5d13a37b
AJ		 4856 break;
4857 case BTRFS_QUOTA_CTL_DISABLE:
340f1aa2 4858 ret = btrfs_quota_disable(fs_info);
5d13a37b 4859 break;
5d13a37b
AJ		 4860 default:
4861 ret = -EINVAL;
4862 break;
4863 }
4864
5d13a37b 4865 kfree(sa);
0b246afa 4866 up_write(&fs_info->subvol_sem);
905b0dda
MX		 4867drop_write:
4868 mnt_drop_write_file(file);
5d13a37b
AJ		 4869 return ret;
4870}
4871
905b0dda 4872static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4873{
0b246afa
JM		 4874 struct inode *inode = file_inode(file);
4875 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4876 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4877 struct btrfs_ioctl_qgroup_assign_args *sa;
4878 struct btrfs_trans_handle *trans;
4879 int ret;
4880 int err;
4881
4882 if (!capable(CAP_SYS_ADMIN))
4883 return -EPERM;
4884
905b0dda
MX		 4885 ret = mnt_want_write_file(file);
4886 if (ret)
4887 return ret;
5d13a37b
AJ		 4888
4889 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4890 if (IS_ERR(sa)) {
4891 ret = PTR_ERR(sa);
4892 goto drop_write;
4893 }
5d13a37b
AJ		 4894
4895 trans = btrfs_join_transaction(root);
4896 if (IS_ERR(trans)) {
4897 ret = PTR_ERR(trans);
4898 goto out;
4899 }
4900
5d13a37b 4901 if (sa->assign) {
9f8a6ce6 4902 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
5d13a37b 4903 } else {
39616c27 4904 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
5d13a37b
AJ		 4905 }
4906
e082f563 4907 /* update qgroup status and info */
280f8bd2 4908 err = btrfs_run_qgroups(trans);
e082f563 4909 if (err < 0)
0b246afa
JM		 4910 btrfs_handle_fs_error(fs_info, err,
4911 "failed to update qgroup status and info");
3a45bb20 4912 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 4913 if (err && !ret)
4914 ret = err;
4915
4916out:
4917 kfree(sa);
905b0dda
MX		 4918drop_write:
4919 mnt_drop_write_file(file);
5d13a37b
AJ		 4920 return ret;
4921}
4922
905b0dda 4923static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4924{
0b246afa 4925 struct inode *inode = file_inode(file);
0b246afa 4926 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4927 struct btrfs_ioctl_qgroup_create_args *sa;
4928 struct btrfs_trans_handle *trans;
4929 int ret;
4930 int err;
4931
4932 if (!capable(CAP_SYS_ADMIN))
4933 return -EPERM;
4934
905b0dda
MX		 4935 ret = mnt_want_write_file(file);
4936 if (ret)
4937 return ret;
5d13a37b
AJ		 4938
4939 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4940 if (IS_ERR(sa)) {
4941 ret = PTR_ERR(sa);
4942 goto drop_write;
4943 }
5d13a37b 4944
d86e56cf
MX		 4945 if (!sa->qgroupid) {
4946 ret = -EINVAL;
4947 goto out;
4948 }
4949
5d13a37b
AJ		 4950 trans = btrfs_join_transaction(root);
4951 if (IS_ERR(trans)) {
4952 ret = PTR_ERR(trans);
4953 goto out;
4954 }
4955
5d13a37b 4956 if (sa->create) {
49a05ecd 4957 ret = btrfs_create_qgroup(trans, sa->qgroupid);
5d13a37b 4958 } else {
3efbee1d 4959 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
5d13a37b
AJ		 4960 }
4961
3a45bb20 4962 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 4963 if (err && !ret)
4964 ret = err;
4965
4966out:
4967 kfree(sa);
905b0dda
MX		 4968drop_write:
4969 mnt_drop_write_file(file);
5d13a37b
AJ		 4970 return ret;
4971}
4972
905b0dda 4973static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4974{
0b246afa 4975 struct inode *inode = file_inode(file);
0b246afa 4976 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4977 struct btrfs_ioctl_qgroup_limit_args *sa;
4978 struct btrfs_trans_handle *trans;
4979 int ret;
4980 int err;
4981 u64 qgroupid;
4982
4983 if (!capable(CAP_SYS_ADMIN))
4984 return -EPERM;
4985
905b0dda
MX		 4986 ret = mnt_want_write_file(file);
4987 if (ret)
4988 return ret;
5d13a37b
AJ		 4989
4990 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4991 if (IS_ERR(sa)) {
4992 ret = PTR_ERR(sa);
4993 goto drop_write;
4994 }
5d13a37b
AJ		 4995
4996 trans = btrfs_join_transaction(root);
4997 if (IS_ERR(trans)) {
4998 ret = PTR_ERR(trans);
4999 goto out;
5000 }
5001
5002 qgroupid = sa->qgroupid;
5003 if (!qgroupid) {
5004 /* take the current subvol as qgroup */
5005 qgroupid = root->root_key.objectid;
5006 }
5007
f0042d5e 5008 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
5d13a37b 5009
3a45bb20 5010 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 5011 if (err && !ret)
5012 ret = err;
5013
5014out:
5015 kfree(sa);
905b0dda
MX		 5016drop_write:
5017 mnt_drop_write_file(file);
5d13a37b
AJ		 5018 return ret;
5019}
5020
2f232036
JS		 5021static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5022{
0b246afa
JM		 5023 struct inode *inode = file_inode(file);
5024 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2f232036
JS		 5025 struct btrfs_ioctl_quota_rescan_args *qsa;
5026 int ret;
5027
5028 if (!capable(CAP_SYS_ADMIN))
5029 return -EPERM;
5030
5031 ret = mnt_want_write_file(file);
5032 if (ret)
5033 return ret;
5034
5035 qsa = memdup_user(arg, sizeof(*qsa));
5036 if (IS_ERR(qsa)) {
5037 ret = PTR_ERR(qsa);
5038 goto drop_write;
5039 }
5040
5041 if (qsa->flags) {
5042 ret = -EINVAL;
5043 goto out;
5044 }
5045
0b246afa 5046 ret = btrfs_qgroup_rescan(fs_info);
2f232036
JS		 5047
5048out:
5049 kfree(qsa);
5050drop_write:
5051 mnt_drop_write_file(file);
5052 return ret;
5053}
5054
5055static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5056{
0b246afa
JM		 5057 struct inode *inode = file_inode(file);
5058 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2f232036
JS		 5059 struct btrfs_ioctl_quota_rescan_args *qsa;
5060 int ret = 0;
5061
5062 if (!capable(CAP_SYS_ADMIN))
5063 return -EPERM;
5064
8d2db785 5065 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
2f232036
JS		 5066 if (!qsa)
5067 return -ENOMEM;
5068
0b246afa 5069 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
2f232036 5070 qsa->flags = 1;
0b246afa 5071 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
2f232036
JS		 5072 }
5073
5074 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5075 ret = -EFAULT;
5076
5077 kfree(qsa);
5078 return ret;
5079}
5080
57254b6e
JS		 5081static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5082{
0b246afa
JM		 5083 struct inode *inode = file_inode(file);
5084 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
57254b6e
JS		 5085
5086 if (!capable(CAP_SYS_ADMIN))
5087 return -EPERM;
5088
0b246afa 5089 return btrfs_qgroup_wait_for_completion(fs_info, true);
57254b6e
JS		 5090}
5091
abccd00f
HM		 5092static long _btrfs_ioctl_set_received_subvol(struct file *file,
5093 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 5094{
496ad9aa 5095 struct inode *inode = file_inode(file);
0b246afa 5096 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
8ea05e3a
AB		 5097 struct btrfs_root *root = BTRFS_I(inode)->root;
5098 struct btrfs_root_item *root_item = &root->root_item;
5099 struct btrfs_trans_handle *trans;
95582b00 5100 struct timespec64 ct = current_time(inode);
8ea05e3a 5101 int ret = 0;
dd5f9615 5102 int received_uuid_changed;
8ea05e3a 5103
bd60ea0f
DS		 5104 if (!inode_owner_or_capable(inode))
5105 return -EPERM;
5106
8ea05e3a
AB		 5107 ret = mnt_want_write_file(file);
5108 if (ret < 0)
5109 return ret;
5110
0b246afa 5111 down_write(&fs_info->subvol_sem);
8ea05e3a 5112
4a0cc7ca 5113 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
8ea05e3a
AB		 5114 ret = -EINVAL;
5115 goto out;
5116 }
5117
5118 if (btrfs_root_readonly(root)) {
5119 ret = -EROFS;
5120 goto out;
5121 }
5122
dd5f9615
SB		 5123 /*
5124 * 1 - root item
5125 * 2 - uuid items (received uuid + subvol uuid)
5126 */
5127 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 5128 if (IS_ERR(trans)) {
5129 ret = PTR_ERR(trans);
5130 trans = NULL;
5131 goto out;
5132 }
5133
5134 sa->rtransid = trans->transid;
5135 sa->rtime.sec = ct.tv_sec;
5136 sa->rtime.nsec = ct.tv_nsec;
5137
dd5f9615
SB		 5138 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5139 BTRFS_UUID_SIZE);
5140 if (received_uuid_changed &&
d87ff758 5141 !btrfs_is_empty_uuid(root_item->received_uuid)) {
d1957791 5142 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
d87ff758
NB		 5143 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5144 root->root_key.objectid);
5145 if (ret && ret != -ENOENT) {
5146 btrfs_abort_transaction(trans, ret);
5147 btrfs_end_transaction(trans);
5148 goto out;
5149 }
5150 }
8ea05e3a
AB		 5151 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5152 btrfs_set_root_stransid(root_item, sa->stransid);
5153 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 5154 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5155 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5156 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5157 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a 5158
0b246afa 5159 ret = btrfs_update_root(trans, fs_info->tree_root,
8ea05e3a
AB		 5160 &root->root_key, &root->root_item);
5161 if (ret < 0) {
3a45bb20 5162 btrfs_end_transaction(trans);
8ea05e3a 5163 goto out;
dd5f9615
SB		 5164 }
5165 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
cdb345a8 5166 ret = btrfs_uuid_tree_add(trans, sa->uuid,
dd5f9615
SB		 5167 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5168 root->root_key.objectid);
5169 if (ret < 0 && ret != -EEXIST) {
66642832 5170 btrfs_abort_transaction(trans, ret);
efd38150 5171 btrfs_end_transaction(trans);
8ea05e3a 5172 goto out;
dd5f9615
SB		 5173 }
5174 }
3a45bb20 5175 ret = btrfs_commit_transaction(trans);
abccd00f 5176out:
0b246afa 5177 up_write(&fs_info->subvol_sem);
abccd00f
HM		 5178 mnt_drop_write_file(file);
5179 return ret;
5180}
5181
5182#ifdef CONFIG_64BIT
5183static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5184 void __user *arg)
5185{
5186 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5187 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5188 int ret = 0;
5189
5190 args32 = memdup_user(arg, sizeof(*args32));
7b9ea627
SV		 5191 if (IS_ERR(args32))
5192 return PTR_ERR(args32);
abccd00f 5193
8d2db785 5194 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
84dbeb87
DC		 5195 if (!args64) {
5196 ret = -ENOMEM;
abccd00f
HM		 5197 goto out;
5198 }
5199
5200 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5201 args64->stransid = args32->stransid;
5202 args64->rtransid = args32->rtransid;
5203 args64->stime.sec = args32->stime.sec;
5204 args64->stime.nsec = args32->stime.nsec;
5205 args64->rtime.sec = args32->rtime.sec;
5206 args64->rtime.nsec = args32->rtime.nsec;
5207 args64->flags = args32->flags;
5208
5209 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5210 if (ret)
5211 goto out;
5212
5213 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5214 args32->stransid = args64->stransid;
5215 args32->rtransid = args64->rtransid;
5216 args32->stime.sec = args64->stime.sec;
5217 args32->stime.nsec = args64->stime.nsec;
5218 args32->rtime.sec = args64->rtime.sec;
5219 args32->rtime.nsec = args64->rtime.nsec;
5220 args32->flags = args64->flags;
5221
5222 ret = copy_to_user(arg, args32, sizeof(*args32));
5223 if (ret)
5224 ret = -EFAULT;
5225
5226out:
5227 kfree(args32);
5228 kfree(args64);
5229 return ret;
5230}
5231#endif
5232
5233static long btrfs_ioctl_set_received_subvol(struct file *file,
5234 void __user *arg)
5235{
5236 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5237 int ret = 0;
5238
5239 sa = memdup_user(arg, sizeof(*sa));
7b9ea627
SV		 5240 if (IS_ERR(sa))
5241 return PTR_ERR(sa);
abccd00f
HM		 5242
5243 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5244
5245 if (ret)
5246 goto out;
5247
8ea05e3a
AB		 5248 ret = copy_to_user(arg, sa, sizeof(*sa));
5249 if (ret)
5250 ret = -EFAULT;
5251
5252out:
5253 kfree(sa);
8ea05e3a
AB		 5254 return ret;
5255}
5256
867ab667 5257static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5258{
0b246afa
JM		 5259 struct inode *inode = file_inode(file);
5260 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
a1b83ac5 5261 size_t len;
867ab667 5262 int ret;
a1b83ac5
AJ		 5263 char label[BTRFS_LABEL_SIZE];
5264
0b246afa
JM		 5265 spin_lock(&fs_info->super_lock);
5266 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5267 spin_unlock(&fs_info->super_lock);
a1b83ac5
AJ		 5268
5269 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 5270
5271 if (len == BTRFS_LABEL_SIZE) {
0b246afa
JM		 5272 btrfs_warn(fs_info,
5273 "label is too long, return the first %zu bytes",
5274 --len);
867ab667 5275 }
5276
867ab667 5277 ret = copy_to_user(arg, label, len);
867ab667 5278
5279 return ret ? -EFAULT : 0;
5280}
5281
a8bfd4ab 5282static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5283{
0b246afa
JM		 5284 struct inode *inode = file_inode(file);
5285 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5286 struct btrfs_root *root = BTRFS_I(inode)->root;
5287 struct btrfs_super_block *super_block = fs_info->super_copy;
a8bfd4ab 5288 struct btrfs_trans_handle *trans;
5289 char label[BTRFS_LABEL_SIZE];
5290 int ret;
5291
5292 if (!capable(CAP_SYS_ADMIN))
5293 return -EPERM;
5294
5295 if (copy_from_user(label, arg, sizeof(label)))
5296 return -EFAULT;
5297
5298 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
0b246afa 5299 btrfs_err(fs_info,
5d163e0e
JM		 5300 "unable to set label with more than %d bytes",
5301 BTRFS_LABEL_SIZE - 1);
a8bfd4ab 5302 return -EINVAL;
5303 }
5304
5305 ret = mnt_want_write_file(file);
5306 if (ret)
5307 return ret;
5308
a8bfd4ab 5309 trans = btrfs_start_transaction(root, 0);
5310 if (IS_ERR(trans)) {
5311 ret = PTR_ERR(trans);
5312 goto out_unlock;
5313 }
5314
0b246afa 5315 spin_lock(&fs_info->super_lock);
a8bfd4ab 5316 strcpy(super_block->label, label);
0b246afa 5317 spin_unlock(&fs_info->super_lock);
3a45bb20 5318 ret = btrfs_commit_transaction(trans);
a8bfd4ab 5319
5320out_unlock:
a8bfd4ab 5321 mnt_drop_write_file(file);
5322 return ret;
5323}
5324
2eaa055f
JM		 5325#define INIT_FEATURE_FLAGS(suffix) \
5326 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5327 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5328 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5329
d5131b65 5330int btrfs_ioctl_get_supported_features(void __user *arg)
2eaa055f 5331{
4d4ab6d6 5332 static const struct btrfs_ioctl_feature_flags features[3] = {
2eaa055f
JM		 5333 INIT_FEATURE_FLAGS(SUPP),
5334 INIT_FEATURE_FLAGS(SAFE_SET),
5335 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5336 };
5337
5338 if (copy_to_user(arg, &features, sizeof(features)))
5339 return -EFAULT;
5340
5341 return 0;
5342}
5343
5344static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5345{
0b246afa
JM		 5346 struct inode *inode = file_inode(file);
5347 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5348 struct btrfs_super_block *super_block = fs_info->super_copy;
2eaa055f
JM		 5349 struct btrfs_ioctl_feature_flags features;
5350
5351 features.compat_flags = btrfs_super_compat_flags(super_block);
5352 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5353 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5354
5355 if (copy_to_user(arg, &features, sizeof(features)))
5356 return -EFAULT;
5357
5358 return 0;
5359}
5360
2ff7e61e 5361static int check_feature_bits(struct btrfs_fs_info *fs_info,
3b02a68a 5362 enum btrfs_feature_set set,
2eaa055f
JM		 5363 u64 change_mask, u64 flags, u64 supported_flags,
5364 u64 safe_set, u64 safe_clear)
5365{
3b02a68a
JM		 5366 const char *type = btrfs_feature_set_names[set];
5367 char *names;
2eaa055f
JM		 5368 u64 disallowed, unsupported;
5369 u64 set_mask = flags & change_mask;
5370 u64 clear_mask = ~flags & change_mask;
5371
5372 unsupported = set_mask & ~supported_flags;
5373 if (unsupported) {
3b02a68a
JM		 5374 names = btrfs_printable_features(set, unsupported);
5375 if (names) {
0b246afa
JM		 5376 btrfs_warn(fs_info,
5377 "this kernel does not support the %s feature bit%s",
5378 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5379 kfree(names);
5380 } else
0b246afa
JM		 5381 btrfs_warn(fs_info,
5382 "this kernel does not support %s bits 0x%llx",
5383 type, unsupported);
2eaa055f
JM		 5384 return -EOPNOTSUPP;
5385 }
5386
5387 disallowed = set_mask & ~safe_set;
5388 if (disallowed) {
3b02a68a
JM		 5389 names = btrfs_printable_features(set, disallowed);
5390 if (names) {
0b246afa
JM		 5391 btrfs_warn(fs_info,
5392 "can't set the %s feature bit%s while mounted",
5393 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5394 kfree(names);
5395 } else
0b246afa
JM		 5396 btrfs_warn(fs_info,
5397 "can't set %s bits 0x%llx while mounted",
5398 type, disallowed);
2eaa055f
JM		 5399 return -EPERM;
5400 }
5401
5402 disallowed = clear_mask & ~safe_clear;
5403 if (disallowed) {
3b02a68a
JM		 5404 names = btrfs_printable_features(set, disallowed);
5405 if (names) {
0b246afa
JM		 5406 btrfs_warn(fs_info,
5407 "can't clear the %s feature bit%s while mounted",
5408 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 5409 kfree(names);
5410 } else
0b246afa
JM		 5411 btrfs_warn(fs_info,
5412 "can't clear %s bits 0x%llx while mounted",
5413 type, disallowed);
2eaa055f
JM		 5414 return -EPERM;
5415 }
5416
5417 return 0;
5418}
5419
2ff7e61e
JM		 5420#define check_feature(fs_info, change_mask, flags, mask_base) \
5421check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 5422 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5423 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5424 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5425
5426static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5427{
0b246afa
JM		 5428 struct inode *inode = file_inode(file);
5429 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5430 struct btrfs_root *root = BTRFS_I(inode)->root;
5431 struct btrfs_super_block *super_block = fs_info->super_copy;
2eaa055f
JM		 5432 struct btrfs_ioctl_feature_flags flags[2];
5433 struct btrfs_trans_handle *trans;
5434 u64 newflags;
5435 int ret;
5436
5437 if (!capable(CAP_SYS_ADMIN))
5438 return -EPERM;
5439
5440 if (copy_from_user(flags, arg, sizeof(flags)))
5441 return -EFAULT;
5442
5443 /* Nothing to do */
5444 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5445 !flags[0].incompat_flags)
5446 return 0;
5447
2ff7e61e 5448 ret = check_feature(fs_info, flags[0].compat_flags,
2eaa055f
JM		 5449 flags[1].compat_flags, COMPAT);
5450 if (ret)
5451 return ret;
5452
2ff7e61e 5453 ret = check_feature(fs_info, flags[0].compat_ro_flags,
2eaa055f
JM		 5454 flags[1].compat_ro_flags, COMPAT_RO);
5455 if (ret)
5456 return ret;
5457
2ff7e61e 5458 ret = check_feature(fs_info, flags[0].incompat_flags,
2eaa055f
JM		 5459 flags[1].incompat_flags, INCOMPAT);
5460 if (ret)
5461 return ret;
5462
7ab19625
DS		 5463 ret = mnt_want_write_file(file);
5464 if (ret)
5465 return ret;
5466
8051aa1a 5467 trans = btrfs_start_transaction(root, 0);
7ab19625
DS		 5468 if (IS_ERR(trans)) {
5469 ret = PTR_ERR(trans);
5470 goto out_drop_write;
5471 }
2eaa055f 5472
0b246afa 5473 spin_lock(&fs_info->super_lock);
2eaa055f
JM		 5474 newflags = btrfs_super_compat_flags(super_block);
5475 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5476 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5477 btrfs_set_super_compat_flags(super_block, newflags);
5478
5479 newflags = btrfs_super_compat_ro_flags(super_block);
5480 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5481 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5482 btrfs_set_super_compat_ro_flags(super_block, newflags);
5483
5484 newflags = btrfs_super_incompat_flags(super_block);
5485 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5486 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5487 btrfs_set_super_incompat_flags(super_block, newflags);
0b246afa 5488 spin_unlock(&fs_info->super_lock);
2eaa055f 5489
3a45bb20 5490 ret = btrfs_commit_transaction(trans);
7ab19625
DS		 5491out_drop_write:
5492 mnt_drop_write_file(file);
5493
5494 return ret;
2eaa055f
JM		 5495}
5496
2351f431
JB		 5497static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
5498{
5499 struct btrfs_ioctl_send_args *arg;
5500 int ret;
5501
5502 if (compat) {
5503#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5504 struct btrfs_ioctl_send_args_32 args32;
5505
5506 ret = copy_from_user(&args32, argp, sizeof(args32));
5507 if (ret)
5508 return -EFAULT;
5509 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5510 if (!arg)
5511 return -ENOMEM;
5512 arg->send_fd = args32.send_fd;
5513 arg->clone_sources_count = args32.clone_sources_count;
5514 arg->clone_sources = compat_ptr(args32.clone_sources);
5515 arg->parent_root = args32.parent_root;
5516 arg->flags = args32.flags;
5517 memcpy(arg->reserved, args32.reserved,
5518 sizeof(args32.reserved));
5519#else
5520 return -ENOTTY;
5521#endif
5522 } else {
5523 arg = memdup_user(argp, sizeof(*arg));
5524 if (IS_ERR(arg))
5525 return PTR_ERR(arg);
5526 }
5527 ret = btrfs_ioctl_send(file, arg);
5528 kfree(arg);
5529 return ret;
5530}
5531
f46b5a66
CH		 5532long btrfs_ioctl(struct file *file, unsigned int
5533 cmd, unsigned long arg)
5534{
0b246afa
JM		 5535 struct inode *inode = file_inode(file);
5536 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5537 struct btrfs_root *root = BTRFS_I(inode)->root;
4bcabaa3 5538 void __user *argp = (void __user *)arg;
f46b5a66
CH		 5539
5540 switch (cmd) {
6cbff00f
CH		 5541 case FS_IOC_GETFLAGS:
5542 return btrfs_ioctl_getflags(file, argp);
5543 case FS_IOC_SETFLAGS:
5544 return btrfs_ioctl_setflags(file, argp);
5545 case FS_IOC_GETVERSION:
5546 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 5547 case FITRIM:
5548 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 5549 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 5550 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 5551 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 5552 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 5553 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 5554 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 5555 case BTRFS_IOC_SUBVOL_CREATE_V2:
5556 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 5557 case BTRFS_IOC_SNAP_DESTROY:
5558 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 5559 case BTRFS_IOC_SUBVOL_GETFLAGS:
5560 return btrfs_ioctl_subvol_getflags(file, argp);
5561 case BTRFS_IOC_SUBVOL_SETFLAGS:
5562 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 5563 case BTRFS_IOC_DEFAULT_SUBVOL:
5564 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 5565 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 5566 return btrfs_ioctl_defrag(file, NULL);
5567 case BTRFS_IOC_DEFRAG_RANGE:
5568 return btrfs_ioctl_defrag(file, argp);
f46b5a66 5569 case BTRFS_IOC_RESIZE:
198605a8 5570 return btrfs_ioctl_resize(file, argp);
f46b5a66 5571 case BTRFS_IOC_ADD_DEV:
2ff7e61e 5572 return btrfs_ioctl_add_dev(fs_info, argp);
f46b5a66 5573 case BTRFS_IOC_RM_DEV:
da24927b 5574 return btrfs_ioctl_rm_dev(file, argp);
6b526ed7
AJ		 5575 case BTRFS_IOC_RM_DEV_V2:
5576 return btrfs_ioctl_rm_dev_v2(file, argp);
475f6387 5577 case BTRFS_IOC_FS_INFO:
2ff7e61e 5578 return btrfs_ioctl_fs_info(fs_info, argp);
475f6387 5579 case BTRFS_IOC_DEV_INFO:
2ff7e61e 5580 return btrfs_ioctl_dev_info(fs_info, argp);
f46b5a66 5581 case BTRFS_IOC_BALANCE:
9ba1f6e4 5582 return btrfs_ioctl_balance(file, NULL);
ac8e9819
CM		 5583 case BTRFS_IOC_TREE_SEARCH:
5584 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 5585 case BTRFS_IOC_TREE_SEARCH_V2:
5586 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 5587 case BTRFS_IOC_INO_LOOKUP:
5588 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 5589 case BTRFS_IOC_INO_PATHS:
5590 return btrfs_ioctl_ino_to_path(root, argp);
5591 case BTRFS_IOC_LOGICAL_INO:
d24a67b2
ZB		 5592 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5593 case BTRFS_IOC_LOGICAL_INO_V2:
5594 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
1406e432 5595 case BTRFS_IOC_SPACE_INFO:
2ff7e61e 5596 return btrfs_ioctl_space_info(fs_info, argp);
9b199859
FDBM		 5597 case BTRFS_IOC_SYNC: {
5598 int ret;
5599
82b3e53b 5600 ret = btrfs_start_delalloc_roots(fs_info, -1);
9b199859
FDBM		 5601 if (ret)
5602 return ret;
0b246afa 5603 ret = btrfs_sync_fs(inode->i_sb, 1);
2fad4e83
DS		 5604 /*
5605 * The transaction thread may want to do more work,
01327610 5606 * namely it pokes the cleaner kthread that will start
2fad4e83
DS		 5607 * processing uncleaned subvols.
5608 */
0b246afa 5609 wake_up_process(fs_info->transaction_kthread);
9b199859
FDBM		 5610 return ret;
5611 }
46204592 5612 case BTRFS_IOC_START_SYNC:
9a8c28be 5613 return btrfs_ioctl_start_sync(root, argp);
46204592 5614 case BTRFS_IOC_WAIT_SYNC:
2ff7e61e 5615 return btrfs_ioctl_wait_sync(fs_info, argp);
475f6387 5616 case BTRFS_IOC_SCRUB:
b8e95489 5617 return btrfs_ioctl_scrub(file, argp);
475f6387 5618 case BTRFS_IOC_SCRUB_CANCEL:
2ff7e61e 5619 return btrfs_ioctl_scrub_cancel(fs_info);
475f6387 5620 case BTRFS_IOC_SCRUB_PROGRESS:
2ff7e61e 5621 return btrfs_ioctl_scrub_progress(fs_info, argp);
c9e9f97b 5622 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 5623 return btrfs_ioctl_balance(file, argp);
837d5b6e 5624 case BTRFS_IOC_BALANCE_CTL:
2ff7e61e 5625 return btrfs_ioctl_balance_ctl(fs_info, arg);
19a39dce 5626 case BTRFS_IOC_BALANCE_PROGRESS:
2ff7e61e 5627 return btrfs_ioctl_balance_progress(fs_info, argp);
8ea05e3a
AB		 5628 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5629 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 5630#ifdef CONFIG_64BIT
5631 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5632 return btrfs_ioctl_set_received_subvol_32(file, argp);
5633#endif
31db9f7c 5634 case BTRFS_IOC_SEND:
2351f431
JB		 5635 return _btrfs_ioctl_send(file, argp, false);
5636#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5637 case BTRFS_IOC_SEND_32:
5638 return _btrfs_ioctl_send(file, argp, true);
5639#endif
c11d2c23 5640 case BTRFS_IOC_GET_DEV_STATS:
2ff7e61e 5641 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5d13a37b 5642 case BTRFS_IOC_QUOTA_CTL:
905b0dda 5643 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 5644 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 5645 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 5646 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 5647 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 5648 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 5649 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 5650 case BTRFS_IOC_QUOTA_RESCAN:
5651 return btrfs_ioctl_quota_rescan(file, argp);
5652 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5653 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 5654 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5655 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd 5656 case BTRFS_IOC_DEV_REPLACE:
2ff7e61e 5657 return btrfs_ioctl_dev_replace(fs_info, argp);
867ab667 5658 case BTRFS_IOC_GET_FSLABEL:
5659 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 5660 case BTRFS_IOC_SET_FSLABEL:
5661 return btrfs_ioctl_set_fslabel(file, argp);
2eaa055f 5662 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
d5131b65 5663 return btrfs_ioctl_get_supported_features(argp);
2eaa055f
JM		 5664 case BTRFS_IOC_GET_FEATURES:
5665 return btrfs_ioctl_get_features(file, argp);
5666 case BTRFS_IOC_SET_FEATURES:
5667 return btrfs_ioctl_set_features(file, argp);
e4202ac9
DS		 5668 case FS_IOC_FSGETXATTR:
5669 return btrfs_ioctl_fsgetxattr(file, argp);
025f2121
DS		 5670 case FS_IOC_FSSETXATTR:
5671 return btrfs_ioctl_fssetxattr(file, argp);
b64ec075
TM		 5672 case BTRFS_IOC_GET_SUBVOL_INFO:
5673 return btrfs_ioctl_get_subvol_info(file, argp);
42e4b520
TM		 5674 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5675 return btrfs_ioctl_get_subvol_rootref(file, argp);
23d0b79d
TM		 5676 case BTRFS_IOC_INO_LOOKUP_USER:
5677 return btrfs_ioctl_ino_lookup_user(file, argp);
f46b5a66
CH		 5678 }
5679
5680 return -ENOTTY;
5681}
4c63c245
LD		 5682
5683#ifdef CONFIG_COMPAT
5684long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5685{
2a362249
JM		 5686 /*
5687 * These all access 32-bit values anyway so no further
5688 * handling is necessary.
5689 */
4c63c245
LD		 5690 switch (cmd) {
5691 case FS_IOC32_GETFLAGS:
5692 cmd = FS_IOC_GETFLAGS;
5693 break;
5694 case FS_IOC32_SETFLAGS:
5695 cmd = FS_IOC_SETFLAGS;
5696 break;
5697 case FS_IOC32_GETVERSION:
5698 cmd = FS_IOC_GETVERSION;
5699 break;
4c63c245
LD		 5700 }
5701
5702 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5703}
5704#endif
Ubuntu Jammy Linux kernel mirror