]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - fs/btrfs/ioctl.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: make get_extent_allocation_hint take btrfs_inode
[mirror_ubuntu-hirsute-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"
949964c9 31#include "export.h"
f46b5a66
CH		 32#include "transaction.h"
33#include "btrfs_inode.h"
f46b5a66
CH		 34#include "print-tree.h"
35#include "volumes.h"
925baedd 36#include "locking.h"
581bb050 37#include "inode-map.h"
d7728c96 38#include "backref.h"
606686ee 39#include "rcu-string.h"
31db9f7c 40#include "send.h"
3f6bcfbd 41#include "dev-replace.h"
63541927 42#include "props.h"
3b02a68a 43#include "sysfs.h"
fcebe456 44#include "qgroup.h"
1ec9a1ae 45#include "tree-log.h"
ebb8765b 46#include "compression.h"
8719aaae 47#include "space-info.h"
86736342 48#include "delalloc-space.h"
aac0023c 49#include "block-group.h"
f46b5a66 50
abccd00f
HM		 51#ifdef CONFIG_64BIT
52/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
53 * structures are incorrect, as the timespec structure from userspace
54 * is 4 bytes too small. We define these alternatives here to teach
55 * the kernel about the 32-bit struct packing.
56 */
57struct btrfs_ioctl_timespec_32 {
58 __u64 sec;
59 __u32 nsec;
60} __attribute__ ((__packed__));
61
62struct btrfs_ioctl_received_subvol_args_32 {
63 char uuid[BTRFS_UUID_SIZE]; /* in */
64 __u64 stransid; /* in */
65 __u64 rtransid; /* out */
66 struct btrfs_ioctl_timespec_32 stime; /* in */
67 struct btrfs_ioctl_timespec_32 rtime; /* out */
68 __u64 flags; /* in */
69 __u64 reserved[16]; /* in */
70} __attribute__ ((__packed__));
71
72#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
73 struct btrfs_ioctl_received_subvol_args_32)
74#endif
75
2351f431
JB		 76#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
77struct btrfs_ioctl_send_args_32 {
78 __s64 send_fd; /* in */
79 __u64 clone_sources_count; /* in */
80 compat_uptr_t clone_sources; /* in */
81 __u64 parent_root; /* in */
82 __u64 flags; /* in */
83 __u64 reserved[4]; /* in */
84} __attribute__ ((__packed__));
85
86#define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
87 struct btrfs_ioctl_send_args_32)
88#endif
abccd00f 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 }
ff9fef55
AJ		 312 } else {
313 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
314 0, 0);
315 if (ret && ret != -ENODATA) {
316 btrfs_abort_transaction(trans, ret);
317 goto out_end_trans;
318 }
319 }
320
d2b8fcfe 321 binode->flags = binode_flags;
7b6a221e 322 btrfs_sync_inode_flags_to_i_flags(inode);
0c4d2d95 323 inode_inc_iversion(inode);
c2050a45 324 inode->i_ctime = current_time(inode);
6cbff00f 325 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 326
ff9fef55 327 out_end_trans:
3a45bb20 328 btrfs_end_transaction(trans);
6cbff00f 329 out_unlock:
5955102c 330 inode_unlock(inode);
e7848683 331 mnt_drop_write_file(file);
2d4e6f6a 332 return ret;
6cbff00f
CH		 333}
334
19f93b3c
DS		 335/*
336 * Translate btrfs internal inode flags to xflags as expected by the
337 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
338 * silently dropped.
339 */
340static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
341{
342 unsigned int xflags = 0;
343
344 if (flags & BTRFS_INODE_APPEND)
345 xflags |= FS_XFLAG_APPEND;
346 if (flags & BTRFS_INODE_IMMUTABLE)
347 xflags |= FS_XFLAG_IMMUTABLE;
348 if (flags & BTRFS_INODE_NOATIME)
349 xflags |= FS_XFLAG_NOATIME;
350 if (flags & BTRFS_INODE_NODUMP)
351 xflags |= FS_XFLAG_NODUMP;
352 if (flags & BTRFS_INODE_SYNC)
353 xflags |= FS_XFLAG_SYNC;
354
355 return xflags;
356}
357
358/* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
359static int check_xflags(unsigned int flags)
360{
361 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
362 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
363 return -EOPNOTSUPP;
364 return 0;
365}
366
e4202ac9
DS		 367/*
368 * Set the xflags from the internal inode flags. The remaining items of fsxattr
369 * are zeroed.
370 */
371static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
372{
373 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
374 struct fsxattr fa;
375
7b0e492e 376 simple_fill_fsxattr(&fa, btrfs_inode_flags_to_xflags(binode->flags));
e4202ac9
DS		 377 if (copy_to_user(arg, &fa, sizeof(fa)))
378 return -EFAULT;
379
380 return 0;
381}
382
025f2121
DS		 383static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
384{
385 struct inode *inode = file_inode(file);
386 struct btrfs_inode *binode = BTRFS_I(inode);
387 struct btrfs_root *root = binode->root;
388 struct btrfs_trans_handle *trans;
7b0e492e 389 struct fsxattr fa, old_fa;
025f2121
DS		 390 unsigned old_flags;
391 unsigned old_i_flags;
392 int ret = 0;
393
394 if (!inode_owner_or_capable(inode))
395 return -EPERM;
396
397 if (btrfs_root_readonly(root))
398 return -EROFS;
399
025f2121
DS		 400 if (copy_from_user(&fa, arg, sizeof(fa)))
401 return -EFAULT;
402
403 ret = check_xflags(fa.fsx_xflags);
404 if (ret)
405 return ret;
406
407 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
408 return -EOPNOTSUPP;
409
410 ret = mnt_want_write_file(file);
411 if (ret)
412 return ret;
413
414 inode_lock(inode);
415
416 old_flags = binode->flags;
417 old_i_flags = inode->i_flags;
418
7b0e492e
DW		 419 simple_fill_fsxattr(&old_fa,
420 btrfs_inode_flags_to_xflags(binode->flags));
421 ret = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
422 if (ret)
025f2121 423 goto out_unlock;
025f2121
DS		 424
425 if (fa.fsx_xflags & FS_XFLAG_SYNC)
426 binode->flags |= BTRFS_INODE_SYNC;
427 else
428 binode->flags &= ~BTRFS_INODE_SYNC;
429 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
430 binode->flags |= BTRFS_INODE_IMMUTABLE;
431 else
432 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
433 if (fa.fsx_xflags & FS_XFLAG_APPEND)
434 binode->flags |= BTRFS_INODE_APPEND;
435 else
436 binode->flags &= ~BTRFS_INODE_APPEND;
437 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
438 binode->flags |= BTRFS_INODE_NODUMP;
439 else
440 binode->flags &= ~BTRFS_INODE_NODUMP;
441 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
442 binode->flags |= BTRFS_INODE_NOATIME;
443 else
444 binode->flags &= ~BTRFS_INODE_NOATIME;
445
446 /* 1 item for the inode */
447 trans = btrfs_start_transaction(root, 1);
448 if (IS_ERR(trans)) {
449 ret = PTR_ERR(trans);
450 goto out_unlock;
451 }
452
453 btrfs_sync_inode_flags_to_i_flags(inode);
454 inode_inc_iversion(inode);
455 inode->i_ctime = current_time(inode);
456 ret = btrfs_update_inode(trans, root, inode);
457
458 btrfs_end_transaction(trans);
459
460out_unlock:
461 if (ret) {
462 binode->flags = old_flags;
463 inode->i_flags = old_i_flags;
464 }
465
466 inode_unlock(inode);
467 mnt_drop_write_file(file);
468
469 return ret;
470}
471
6cbff00f
CH		 472static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
473{
496ad9aa 474 struct inode *inode = file_inode(file);
6cbff00f
CH		 475
476 return put_user(inode->i_generation, arg);
477}
f46b5a66 478
b929c1d8
MPS		 479static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
480 void __user *arg)
f7039b1d 481{
f7039b1d
LD		 482 struct btrfs_device *device;
483 struct request_queue *q;
484 struct fstrim_range range;
485 u64 minlen = ULLONG_MAX;
486 u64 num_devices = 0;
487 int ret;
488
489 if (!capable(CAP_SYS_ADMIN))
490 return -EPERM;
491
f35f06c3
FM		 492 /*
493 * If the fs is mounted with nologreplay, which requires it to be
494 * mounted in RO mode as well, we can not allow discard on free space
495 * inside block groups, because log trees refer to extents that are not
496 * pinned in a block group's free space cache (pinning the extents is
497 * precisely the first phase of replaying a log tree).
498 */
499 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
500 return -EROFS;
501
1f78160c
XG		 502 rcu_read_lock();
503 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
504 dev_list) {
f7039b1d
LD		 505 if (!device->bdev)
506 continue;
507 q = bdev_get_queue(device->bdev);
508 if (blk_queue_discard(q)) {
509 num_devices++;
50d0446e 510 minlen = min_t(u64, q->limits.discard_granularity,
f7039b1d
LD		 511 minlen);
512 }
513 }
1f78160c 514 rcu_read_unlock();
f4c697e6 515
f7039b1d
LD		 516 if (!num_devices)
517 return -EOPNOTSUPP;
f7039b1d
LD		 518 if (copy_from_user(&range, arg, sizeof(range)))
519 return -EFAULT;
6ba9fc8e
QW		 520
521 /*
522 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
523 * block group is in the logical address space, which can be any
524 * sectorsize aligned bytenr in the range [0, U64_MAX].
525 */
526 if (range.len < fs_info->sb->s_blocksize)
f4c697e6 527 return -EINVAL;
f7039b1d
LD		 528
529 range.minlen = max(range.minlen, minlen);
2ff7e61e 530 ret = btrfs_trim_fs(fs_info, &range);
f7039b1d
LD		 531 if (ret < 0)
532 return ret;
533
534 if (copy_to_user(arg, &range, sizeof(range)))
535 return -EFAULT;
536
537 return 0;
538}
539
e1f60a65 540int __pure btrfs_is_empty_uuid(u8 *uuid)
dd5f9615 541{
46e0f66a
CM		 542 int i;
543
544 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
545 if (uuid[i])
546 return 0;
547 }
548 return 1;
dd5f9615
SB		 549}
550
d5c12070 551static noinline int create_subvol(struct inode *dir,
cb8e7090 552 struct dentry *dentry,
52f75f4f 553 const char *name, int namelen,
8696c533 554 struct btrfs_qgroup_inherit *inherit)
f46b5a66 555{
0b246afa 556 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
f46b5a66
CH		 557 struct btrfs_trans_handle *trans;
558 struct btrfs_key key;
49a3c4d9 559 struct btrfs_root_item *root_item;
f46b5a66
CH		 560 struct btrfs_inode_item *inode_item;
561 struct extent_buffer *leaf;
d5c12070 562 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 563 struct btrfs_root *new_root;
d5c12070 564 struct btrfs_block_rsv block_rsv;
95582b00 565 struct timespec64 cur_time = current_time(dir);
5662344b 566 struct inode *inode;
f46b5a66
CH		 567 int ret;
568 int err;
569 u64 objectid;
570 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 571 u64 index = 0;
f46b5a66 572
49a3c4d9
DS		 573 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
574 if (!root_item)
575 return -ENOMEM;
576
0b246afa 577 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
2fbe8c8a 578 if (ret)
49a3c4d9 579 goto fail_free;
6a912213 580
e09fe2d2
QW		 581 /*
582 * Don't create subvolume whose level is not zero. Or qgroup will be
01327610 583 * screwed up since it assumes subvolume qgroup's level to be 0.
e09fe2d2 584 */
49a3c4d9
DS		 585 if (btrfs_qgroup_level(objectid)) {
586 ret = -ENOSPC;
587 goto fail_free;
588 }
e09fe2d2 589
d5c12070 590 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 591 /*
d5c12070
MX		 592 * The same as the snapshot creation, please see the comment
593 * of create_snapshot().
9ed74f2d 594 */
c4c129db 595 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
d5c12070 596 if (ret)
49a3c4d9 597 goto fail_free;
d5c12070
MX		 598
599 trans = btrfs_start_transaction(root, 0);
600 if (IS_ERR(trans)) {
601 ret = PTR_ERR(trans);
7775c818 602 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
49a3c4d9 603 goto fail_free;
d5c12070
MX		 604 }
605 trans->block_rsv = &block_rsv;
606 trans->bytes_reserved = block_rsv.size;
f46b5a66 607
a9377422 608 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
6f72c7e2
AJ		 609 if (ret)
610 goto fail;
611
4d75f8a9 612 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 613 if (IS_ERR(leaf)) {
614 ret = PTR_ERR(leaf);
615 goto fail;
616 }
f46b5a66 617
f46b5a66
CH		 618 btrfs_mark_buffer_dirty(leaf);
619
49a3c4d9 620 inode_item = &root_item->inode;
3cae210f
QW		 621 btrfs_set_stack_inode_generation(inode_item, 1);
622 btrfs_set_stack_inode_size(inode_item, 3);
623 btrfs_set_stack_inode_nlink(inode_item, 1);
da17066c 624 btrfs_set_stack_inode_nbytes(inode_item,
0b246afa 625 fs_info->nodesize);
3cae210f 626 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 627
49a3c4d9
DS		 628 btrfs_set_root_flags(root_item, 0);
629 btrfs_set_root_limit(root_item, 0);
3cae210f 630 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 631
49a3c4d9
DS		 632 btrfs_set_root_bytenr(root_item, leaf->start);
633 btrfs_set_root_generation(root_item, trans->transid);
634 btrfs_set_root_level(root_item, 0);
635 btrfs_set_root_refs(root_item, 1);
636 btrfs_set_root_used(root_item, leaf->len);
637 btrfs_set_root_last_snapshot(root_item, 0);
f46b5a66 638
49a3c4d9
DS		 639 btrfs_set_root_generation_v2(root_item,
640 btrfs_root_generation(root_item));
807fc790 641 generate_random_guid(root_item->uuid);
49a3c4d9
DS		 642 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
643 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
644 root_item->ctime = root_item->otime;
645 btrfs_set_root_ctransid(root_item, trans->transid);
646 btrfs_set_root_otransid(root_item, trans->transid);
f46b5a66 647
925baedd 648 btrfs_tree_unlock(leaf);
f46b5a66
CH		 649 free_extent_buffer(leaf);
650 leaf = NULL;
651
49a3c4d9 652 btrfs_set_root_dirid(root_item, new_dirid);
f46b5a66
CH		 653
654 key.objectid = objectid;
5d4f98a2 655 key.offset = 0;
962a298f 656 key.type = BTRFS_ROOT_ITEM_KEY;
0b246afa 657 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
49a3c4d9 658 root_item);
f46b5a66
CH		 659 if (ret)
660 goto fail;
661
76dda93c 662 key.offset = (u64)-1;
56e9357a 663 new_root = btrfs_get_fs_root(fs_info, objectid, true);
79787eaa 664 if (IS_ERR(new_root)) {
79787eaa 665 ret = PTR_ERR(new_root);
66642832 666 btrfs_abort_transaction(trans, ret);
79787eaa
JM		 667 goto fail;
668 }
76dda93c
YZ		 669
670 btrfs_record_root_in_trans(trans, new_root);
671
63541927 672 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
00246528 673 btrfs_put_root(new_root);
ce598979
MF		 674 if (ret) {
675 /* We potentially lose an unused inode item here */
66642832 676 btrfs_abort_transaction(trans, ret);
ce598979
MF		 677 goto fail;
678 }
679
f32e48e9
CR		 680 mutex_lock(&new_root->objectid_mutex);
681 new_root->highest_objectid = new_dirid;
682 mutex_unlock(&new_root->objectid_mutex);
683
f46b5a66
CH		 684 /*
685 * insert the directory item
686 */
877574e2 687 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
79787eaa 688 if (ret) {
66642832 689 btrfs_abort_transaction(trans, ret);
79787eaa
JM		 690 goto fail;
691 }
3de4586c 692
684572df 693 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
3de4586c 694 BTRFS_FT_DIR, index);
79787eaa 695 if (ret) {
66642832 696 btrfs_abort_transaction(trans, ret);
f46b5a66 697 goto fail;
79787eaa 698 }
0660b5af 699
6ef06d27 700 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
52c26179 701 ret = btrfs_update_inode(trans, root, dir);
c7e54b51
JB		 702 if (ret) {
703 btrfs_abort_transaction(trans, ret);
704 goto fail;
705 }
52c26179 706
6025c19f 707 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
4a0cc7ca 708 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
c7e54b51
JB		 709 if (ret) {
710 btrfs_abort_transaction(trans, ret);
711 goto fail;
712 }
f46b5a66 713
cdb345a8 714 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
6bccf3ab 715 BTRFS_UUID_KEY_SUBVOL, objectid);
dd5f9615 716 if (ret)
66642832 717 btrfs_abort_transaction(trans, ret);
dd5f9615 718
f46b5a66 719fail:
49a3c4d9 720 kfree(root_item);
d5c12070
MX		 721 trans->block_rsv = NULL;
722 trans->bytes_reserved = 0;
7775c818 723 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
de6e8200 724
9babda9f 725 err = btrfs_commit_transaction(trans);
f46b5a66
CH		 726 if (err && !ret)
727 ret = err;
1a65e24b 728
5662344b
TI		 729 if (!ret) {
730 inode = btrfs_lookup_dentry(dir, dentry);
de6e8200
LB		 731 if (IS_ERR(inode))
732 return PTR_ERR(inode);
5662344b
TI		 733 d_instantiate(dentry, inode);
734 }
f46b5a66 735 return ret;
49a3c4d9
DS		 736
737fail_free:
738 kfree(root_item);
739 return ret;
f46b5a66
CH		 740}
741
e9662f70 742static int create_snapshot(struct btrfs_root *root, struct inode *dir,
9babda9f 743 struct dentry *dentry, bool readonly,
e9662f70 744 struct btrfs_qgroup_inherit *inherit)
f46b5a66 745{
0b246afa 746 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
2e4bfab9 747 struct inode *inode;
f46b5a66
CH		 748 struct btrfs_pending_snapshot *pending_snapshot;
749 struct btrfs_trans_handle *trans;
2e4bfab9 750 int ret;
f46b5a66 751
92a7cc42 752 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
f46b5a66
CH		 753 return -EINVAL;
754
eede2bf3
OS		 755 if (atomic_read(&root->nr_swapfiles)) {
756 btrfs_warn(fs_info,
757 "cannot snapshot subvolume with active swapfile");
758 return -ETXTBSY;
759 }
760
23269bf5 761 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
a1ee7362
DS		 762 if (!pending_snapshot)
763 return -ENOMEM;
764
b0c0ea63 765 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
23269bf5 766 GFP_KERNEL);
8546b570
DS		 767 pending_snapshot->path = btrfs_alloc_path();
768 if (!pending_snapshot->root_item || !pending_snapshot->path) {
b0c0ea63
DS		 769 ret = -ENOMEM;
770 goto free_pending;
771 }
772
66d8f3dd
MX		 773 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
774 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 775 /*
776 * 1 - parent dir inode
777 * 2 - dir entries
778 * 1 - root item
779 * 2 - root ref/backref
780 * 1 - root of snapshot
dd5f9615 781 * 1 - UUID item
d5c12070
MX		 782 */
783 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 784 &pending_snapshot->block_rsv, 8,
ee3441b4 785 false);
d5c12070 786 if (ret)
c11fbb6e 787 goto free_pending;
d5c12070 788
3de4586c 789 pending_snapshot->dentry = dentry;
f46b5a66 790 pending_snapshot->root = root;
b83cc969 791 pending_snapshot->readonly = readonly;
e9662f70 792 pending_snapshot->dir = dir;
8696c533 793 pending_snapshot->inherit = inherit;
a22285a6 794
d5c12070 795 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 796 if (IS_ERR(trans)) {
797 ret = PTR_ERR(trans);
798 goto fail;
799 }
800
0b246afa 801 spin_lock(&fs_info->trans_lock);
f46b5a66
CH		 802 list_add(&pending_snapshot->list,
803 &trans->transaction->pending_snapshots);
0b246afa 804 spin_unlock(&fs_info->trans_lock);
9babda9f
NB		 805
806 ret = btrfs_commit_transaction(trans);
aec8030a 807 if (ret)
c37b2b62 808 goto fail;
a22285a6
YZ		 809
810 ret = pending_snapshot->error;
811 if (ret)
812 goto fail;
813
d3797308
CM		 814 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
815 if (ret)
816 goto fail;
817
2b0143b5 818 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
2e4bfab9
YZ		 819 if (IS_ERR(inode)) {
820 ret = PTR_ERR(inode);
821 goto fail;
822 }
5662344b 823
2e4bfab9
YZ		 824 d_instantiate(dentry, inode);
825 ret = 0;
826fail:
00246528 827 btrfs_put_root(pending_snapshot->snap);
7775c818 828 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
b0c0ea63
DS		 829free_pending:
830 kfree(pending_snapshot->root_item);
8546b570 831 btrfs_free_path(pending_snapshot->path);
a1ee7362
DS		 832 kfree(pending_snapshot);
833
f46b5a66
CH		 834 return ret;
835}
836
4260f7c7
SW		 837/* copy of may_delete in fs/namei.c()
838 * Check whether we can remove a link victim from directory dir, check
839 * whether the type of victim is right.
840 * 1. We can't do it if dir is read-only (done in permission())
841 * 2. We should have write and exec permissions on dir
842 * 3. We can't remove anything from append-only dir
843 * 4. We can't do anything with immutable dir (done in permission())
844 * 5. If the sticky bit on dir is set we should either
845 * a. be owner of dir, or
846 * b. be owner of victim, or
847 * c. have CAP_FOWNER capability
01327610 848 * 6. If the victim is append-only or immutable we can't do anything with
4260f7c7
SW		 849 * links pointing to it.
850 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
851 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
852 * 9. We can't remove a root or mountpoint.
853 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
854 * nfs_async_unlink().
855 */
856
67871254 857static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
4260f7c7
SW		 858{
859 int error;
860
2b0143b5 861 if (d_really_is_negative(victim))
4260f7c7
SW		 862 return -ENOENT;
863
2b0143b5 864 BUG_ON(d_inode(victim->d_parent) != dir);
4fa6b5ec 865 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 866
867 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
868 if (error)
869 return error;
870 if (IS_APPEND(dir))
871 return -EPERM;
2b0143b5
DH		 872 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
873 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
4260f7c7
SW		 874 return -EPERM;
875 if (isdir) {
e36cb0b8 876 if (!d_is_dir(victim))
4260f7c7
SW		 877 return -ENOTDIR;
878 if (IS_ROOT(victim))
879 return -EBUSY;
e36cb0b8 880 } else if (d_is_dir(victim))
4260f7c7
SW		 881 return -EISDIR;
882 if (IS_DEADDIR(dir))
883 return -ENOENT;
884 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
885 return -EBUSY;
886 return 0;
887}
888
cb8e7090
CH		 889/* copy of may_create in fs/namei.c() */
890static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
891{
2b0143b5 892 if (d_really_is_positive(child))
cb8e7090
CH		 893 return -EEXIST;
894 if (IS_DEADDIR(dir))
895 return -ENOENT;
896 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
897}
898
899/*
900 * Create a new subvolume below @parent. This is largely modeled after
901 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
902 * inside this filesystem so it's quite a bit simpler.
903 */
92872094 904static noinline int btrfs_mksubvol(const struct path *parent,
52f75f4f 905 const char *name, int namelen,
72fd032e 906 struct btrfs_root *snap_src,
9babda9f 907 bool readonly,
8696c533 908 struct btrfs_qgroup_inherit *inherit)
cb8e7090 909{
0b246afa
JM		 910 struct inode *dir = d_inode(parent->dentry);
911 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
cb8e7090
CH		 912 struct dentry *dentry;
913 int error;
914
00235411
AV		 915 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
916 if (error == -EINTR)
917 return error;
cb8e7090
CH		 918
919 dentry = lookup_one_len(name, parent->dentry, namelen);
920 error = PTR_ERR(dentry);
921 if (IS_ERR(dentry))
922 goto out_unlock;
923
76dda93c 924 error = btrfs_may_create(dir, dentry);
cb8e7090 925 if (error)
a874a63e 926 goto out_dput;
cb8e7090 927
9c52057c
CM		 928 /*
929 * even if this name doesn't exist, we may get hash collisions.
930 * check for them now when we can safely fail
931 */
932 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
933 dir->i_ino, name,
934 namelen);
935 if (error)
936 goto out_dput;
937
0b246afa 938 down_read(&fs_info->subvol_sem);
76dda93c
YZ		 939
940 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
941 goto out_up_read;
942
9babda9f
NB		 943 if (snap_src)
944 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
945 else
946 error = create_subvol(dir, dentry, name, namelen, inherit);
947
76dda93c
YZ		 948 if (!error)
949 fsnotify_mkdir(dir, dentry);
950out_up_read:
0b246afa 951 up_read(&fs_info->subvol_sem);
cb8e7090
CH		 952out_dput:
953 dput(dentry);
954out_unlock:
5955102c 955 inode_unlock(dir);
cb8e7090
CH		 956 return error;
957}
958
c11fbb6e
RK		 959static noinline int btrfs_mksnapshot(const struct path *parent,
960 const char *name, int namelen,
961 struct btrfs_root *root,
962 bool readonly,
963 struct btrfs_qgroup_inherit *inherit)
964{
965 int ret;
966 bool snapshot_force_cow = false;
967
968 /*
969 * Force new buffered writes to reserve space even when NOCOW is
970 * possible. This is to avoid later writeback (running dealloc) to
971 * fallback to COW mode and unexpectedly fail with ENOSPC.
972 */
973 btrfs_drew_read_lock(&root->snapshot_lock);
974
975 ret = btrfs_start_delalloc_snapshot(root);
976 if (ret)
977 goto out;
978
979 /*
980 * All previous writes have started writeback in NOCOW mode, so now
981 * we force future writes to fallback to COW mode during snapshot
982 * creation.
983 */
984 atomic_inc(&root->snapshot_force_cow);
985 snapshot_force_cow = true;
986
987 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
988
989 ret = btrfs_mksubvol(parent, name, namelen,
990 root, readonly, inherit);
991out:
992 if (snapshot_force_cow)
993 atomic_dec(&root->snapshot_force_cow);
994 btrfs_drew_read_unlock(&root->snapshot_lock);
995 return ret;
996}
997
4cb5300b
CM		 998/*
999 * When we're defragging a range, we don't want to kick it off again
1000 * if it is really just waiting for delalloc to send it down.
1001 * If we find a nice big extent or delalloc range for the bytes in the
1002 * file you want to defrag, we return 0 to let you know to skip this
1003 * part of the file
1004 */
aab110ab 1005static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
4cb5300b
CM		 1006{
1007 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1008 struct extent_map *em = NULL;
1009 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1010 u64 end;
1011
1012 read_lock(&em_tree->lock);
09cbfeaf 1013 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
4cb5300b
CM		 1014 read_unlock(&em_tree->lock);
1015
1016 if (em) {
1017 end = extent_map_end(em);
1018 free_extent_map(em);
1019 if (end - offset > thresh)
1020 return 0;
1021 }
1022 /* if we already have a nice delalloc here, just stop */
1023 thresh /= 2;
1024 end = count_range_bits(io_tree, &offset, offset + thresh,
1025 thresh, EXTENT_DELALLOC, 1);
1026 if (end >= thresh)
1027 return 0;
1028 return 1;
1029}
1030
1031/*
1032 * helper function to walk through a file and find extents
1033 * newer than a specific transid, and smaller than thresh.
1034 *
1035 * This is used by the defragging code to find new and small
1036 * extents
1037 */
1038static int find_new_extents(struct btrfs_root *root,
1039 struct inode *inode, u64 newer_than,
aab110ab 1040 u64 *off, u32 thresh)
4cb5300b
CM		 1041{
1042 struct btrfs_path *path;
1043 struct btrfs_key min_key;
4cb5300b
CM		 1044 struct extent_buffer *leaf;
1045 struct btrfs_file_extent_item *extent;
1046 int type;
1047 int ret;
4a0cc7ca 1048 u64 ino = btrfs_ino(BTRFS_I(inode));
4cb5300b
CM		 1049
1050 path = btrfs_alloc_path();
1051 if (!path)
1052 return -ENOMEM;
1053
a4689d2b 1054 min_key.objectid = ino;
4cb5300b
CM		 1055 min_key.type = BTRFS_EXTENT_DATA_KEY;
1056 min_key.offset = *off;
1057
67871254 1058 while (1) {
6174d3cb 1059 ret = btrfs_search_forward(root, &min_key, path, newer_than);
4cb5300b
CM		 1060 if (ret != 0)
1061 goto none;
f094c9bd 1062process_slot:
a4689d2b 1063 if (min_key.objectid != ino)
4cb5300b
CM		 1064 goto none;
1065 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1066 goto none;
1067
1068 leaf = path->nodes[0];
1069 extent = btrfs_item_ptr(leaf, path->slots[0],
1070 struct btrfs_file_extent_item);
1071
1072 type = btrfs_file_extent_type(leaf, extent);
1073 if (type == BTRFS_FILE_EXTENT_REG &&
1074 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1075 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1076 *off = min_key.offset;
1077 btrfs_free_path(path);
1078 return 0;
1079 }
1080
f094c9bd
FM		 1081 path->slots[0]++;
1082 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1083 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1084 goto process_slot;
1085 }
1086
4cb5300b
CM		 1087 if (min_key.offset == (u64)-1)
1088 goto none;
1089
1090 min_key.offset++;
1091 btrfs_release_path(path);
1092 }
1093none:
1094 btrfs_free_path(path);
1095 return -ENOENT;
1096}
1097
6c282eb4 1098static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 1099{
1100 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 1101 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1102 struct extent_map *em;
09cbfeaf 1103 u64 len = PAGE_SIZE;
17ce6ef8 1104
6c282eb4
LZ		 1105 /*
1106 * hopefully we have this extent in the tree already, try without
1107 * the full extent lock
1108 */
17ce6ef8 1109 read_lock(&em_tree->lock);
6c282eb4 1110 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 1111 read_unlock(&em_tree->lock);
1112
6c282eb4 1113 if (!em) {
308d9800
FM		 1114 struct extent_state *cached = NULL;
1115 u64 end = start + len - 1;
1116
6c282eb4 1117 /* get the big lock and read metadata off disk */
ff13db41 1118 lock_extent_bits(io_tree, start, end, &cached);
39b07b5d 1119 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
e43bbe5e 1120 unlock_extent_cached(io_tree, start, end, &cached);
6c282eb4
LZ		 1121
1122 if (IS_ERR(em))
1123 return NULL;
1124 }
1125
1126 return em;
1127}
17ce6ef8 1128
6c282eb4
LZ		 1129static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1130{
1131 struct extent_map *next;
1132 bool ret = true;
1133
1134 /* this is the last extent */
1135 if (em->start + em->len >= i_size_read(inode))
1136 return false;
1137
1138 next = defrag_lookup_extent(inode, em->start + em->len);
e9512d72
CM		 1139 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1140 ret = false;
1141 else if ((em->block_start + em->block_len == next->block_start) &&
ee22184b 1142 (em->block_len > SZ_128K && next->block_len > SZ_128K))
6c282eb4
LZ		 1143 ret = false;
1144
1145 free_extent_map(next);
17ce6ef8
LB		 1146 return ret;
1147}
1148
aab110ab 1149static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
a43a2111
AM		 1150 u64 *last_len, u64 *skip, u64 *defrag_end,
1151 int compress)
940100a4 1152{
6c282eb4 1153 struct extent_map *em;
940100a4 1154 int ret = 1;
6c282eb4 1155 bool next_mergeable = true;
4a3560c4 1156 bool prev_mergeable = true;
940100a4
CM		 1157
1158 /*
008873ea 1159 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 1160 * defragging it
1161 */
1162 if (start < *defrag_end)
1163 return 1;
1164
1165 *skip = 0;
1166
6c282eb4
LZ		 1167 em = defrag_lookup_extent(inode, start);
1168 if (!em)
1169 return 0;
940100a4
CM		 1170
1171 /* this will cover holes, and inline extents */
17ce6ef8 1172 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 1173 ret = 0;
17ce6ef8
LB		 1174 goto out;
1175 }
1176
4a3560c4
LB		 1177 if (!*defrag_end)
1178 prev_mergeable = false;
1179
6c282eb4 1180 next_mergeable = defrag_check_next_extent(inode, em);
940100a4 1181 /*
6c282eb4
LZ		 1182 * we hit a real extent, if it is big or the next extent is not a
1183 * real extent, don't bother defragging it
940100a4 1184 */
a43a2111 1185 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
4a3560c4 1186 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
940100a4 1187 ret = 0;
17ce6ef8 1188out:
940100a4
CM		 1189 /*
1190 * last_len ends up being a counter of how many bytes we've defragged.
1191 * every time we choose not to defrag an extent, we reset *last_len
1192 * so that the next tiny extent will force a defrag.
1193 *
1194 * The end result of this is that tiny extents before a single big
1195 * extent will force at least part of that big extent to be defragged.
1196 */
1197 if (ret) {
940100a4
CM		 1198 *defrag_end = extent_map_end(em);
1199 } else {
1200 *last_len = 0;
1201 *skip = extent_map_end(em);
1202 *defrag_end = 0;
1203 }
1204
1205 free_extent_map(em);
1206 return ret;
1207}
1208
4cb5300b
CM		 1209/*
1210 * it doesn't do much good to defrag one or two pages
1211 * at a time. This pulls in a nice chunk of pages
1212 * to COW and defrag.
1213 *
1214 * It also makes sure the delalloc code has enough
1215 * dirty data to avoid making new small extents as part
1216 * of the defrag
1217 *
1218 * It's a good idea to start RA on this range
1219 * before calling this.
1220 */
1221static int cluster_pages_for_defrag(struct inode *inode,
1222 struct page **pages,
1223 unsigned long start_index,
c41570c9 1224 unsigned long num_pages)
f46b5a66 1225{
4cb5300b
CM		 1226 unsigned long file_end;
1227 u64 isize = i_size_read(inode);
1228 u64 page_start;
1229 u64 page_end;
1f12bd06 1230 u64 page_cnt;
4cb5300b
CM		 1231 int ret;
1232 int i;
1233 int i_done;
3eaa2885 1234 struct btrfs_ordered_extent *ordered;
4cb5300b 1235 struct extent_state *cached_state = NULL;
600a45e1 1236 struct extent_io_tree *tree;
364ecf36 1237 struct extent_changeset *data_reserved = NULL;
3b16a4e3 1238 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1239
09cbfeaf 1240 file_end = (isize - 1) >> PAGE_SHIFT;
1f12bd06
LB		 1241 if (!isize || start_index > file_end)
1242 return 0;
1243
1244 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b 1245
364ecf36 1246 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
09cbfeaf
KS		 1247 start_index << PAGE_SHIFT,
1248 page_cnt << PAGE_SHIFT);
4cb5300b
CM		 1249 if (ret)
1250 return ret;
4cb5300b 1251 i_done = 0;
600a45e1 1252 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1253
1254 /* step one, lock all the pages */
1f12bd06 1255 for (i = 0; i < page_cnt; i++) {
4cb5300b 1256 struct page *page;
600a45e1 1257again:
a94733d0 1258 page = find_or_create_page(inode->i_mapping,
600a45e1 1259 start_index + i, mask);
4cb5300b
CM		 1260 if (!page)
1261 break;
1262
600a45e1 1263 page_start = page_offset(page);
09cbfeaf 1264 page_end = page_start + PAGE_SIZE - 1;
600a45e1 1265 while (1) {
308d9800 1266 lock_extent_bits(tree, page_start, page_end,
ff13db41 1267 &cached_state);
600a45e1
MX		 1268 ordered = btrfs_lookup_ordered_extent(inode,
1269 page_start);
308d9800 1270 unlock_extent_cached(tree, page_start, page_end,
e43bbe5e 1271 &cached_state);
600a45e1
MX		 1272 if (!ordered)
1273 break;
1274
1275 unlock_page(page);
1276 btrfs_start_ordered_extent(inode, ordered, 1);
1277 btrfs_put_ordered_extent(ordered);
1278 lock_page(page);
1f12bd06
LB		 1279 /*
1280 * we unlocked the page above, so we need check if
1281 * it was released or not.
1282 */
1283 if (page->mapping != inode->i_mapping) {
1284 unlock_page(page);
09cbfeaf 1285 put_page(page);
1f12bd06
LB		 1286 goto again;
1287 }
600a45e1
MX		 1288 }
1289
4cb5300b
CM		 1290 if (!PageUptodate(page)) {
1291 btrfs_readpage(NULL, page);
1292 lock_page(page);
1293 if (!PageUptodate(page)) {
1294 unlock_page(page);
09cbfeaf 1295 put_page(page);
4cb5300b
CM		 1296 ret = -EIO;
1297 break;
1298 }
1299 }
600a45e1 1300
600a45e1
MX		 1301 if (page->mapping != inode->i_mapping) {
1302 unlock_page(page);
09cbfeaf 1303 put_page(page);
600a45e1
MX		 1304 goto again;
1305 }
1306
4cb5300b
CM		 1307 pages[i] = page;
1308 i_done++;
1309 }
1310 if (!i_done || ret)
1311 goto out;
1312
1751e8a6 1313 if (!(inode->i_sb->s_flags & SB_ACTIVE))
4cb5300b
CM		 1314 goto out;
1315
1316 /*
1317 * so now we have a nice long stream of locked
1318 * and up to date pages, lets wait on them
1319 */
1320 for (i = 0; i < i_done; i++)
1321 wait_on_page_writeback(pages[i]);
1322
1323 page_start = page_offset(pages[0]);
09cbfeaf 1324 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
4cb5300b
CM		 1325
1326 lock_extent_bits(&BTRFS_I(inode)->io_tree,
ff13db41 1327 page_start, page_end - 1, &cached_state);
4cb5300b 1328 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
e182163d
OS		 1329 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1330 EXTENT_DEFRAG, 0, 0, &cached_state);
4cb5300b 1331
1f12bd06 1332 if (i_done != page_cnt) {
9e0baf60 1333 spin_lock(&BTRFS_I(inode)->lock);
28c4a3e2 1334 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
9e0baf60 1335 spin_unlock(&BTRFS_I(inode)->lock);
bc42bda2 1336 btrfs_delalloc_release_space(inode, data_reserved,
09cbfeaf 1337 start_index << PAGE_SHIFT,
43b18595 1338 (page_cnt - i_done) << PAGE_SHIFT, true);
4cb5300b
CM		 1339 }
1340
1341
9e8a4a8b 1342 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
018ed4f7 1343 &cached_state);
4cb5300b
CM		 1344
1345 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
e43bbe5e 1346 page_start, page_end - 1, &cached_state);
4cb5300b
CM		 1347
1348 for (i = 0; i < i_done; i++) {
1349 clear_page_dirty_for_io(pages[i]);
1350 ClearPageChecked(pages[i]);
1351 set_page_extent_mapped(pages[i]);
1352 set_page_dirty(pages[i]);
1353 unlock_page(pages[i]);
09cbfeaf 1354 put_page(pages[i]);
4cb5300b 1355 }
8702ba93 1356 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
364ecf36 1357 extent_changeset_free(data_reserved);
4cb5300b
CM		 1358 return i_done;
1359out:
1360 for (i = 0; i < i_done; i++) {
1361 unlock_page(pages[i]);
09cbfeaf 1362 put_page(pages[i]);
4cb5300b 1363 }
bc42bda2 1364 btrfs_delalloc_release_space(inode, data_reserved,
09cbfeaf 1365 start_index << PAGE_SHIFT,
43b18595 1366 page_cnt << PAGE_SHIFT, true);
8702ba93 1367 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
364ecf36 1368 extent_changeset_free(data_reserved);
4cb5300b
CM		 1369 return ret;
1370
1371}
1372
1373int btrfs_defrag_file(struct inode *inode, struct file *file,
1374 struct btrfs_ioctl_defrag_range_args *range,
1375 u64 newer_than, unsigned long max_to_defrag)
1376{
0b246afa 1377 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4cb5300b 1378 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1379 struct file_ra_state *ra = NULL;
f46b5a66 1380 unsigned long last_index;
151a31b2 1381 u64 isize = i_size_read(inode);
940100a4
CM		 1382 u64 last_len = 0;
1383 u64 skip = 0;
1384 u64 defrag_end = 0;
4cb5300b 1385 u64 newer_off = range->start;
f46b5a66 1386 unsigned long i;
008873ea 1387 unsigned long ra_index = 0;
f46b5a66 1388 int ret;
4cb5300b 1389 int defrag_count = 0;
1a419d85 1390 int compress_type = BTRFS_COMPRESS_ZLIB;
aab110ab 1391 u32 extent_thresh = range->extent_thresh;
09cbfeaf 1392 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
c41570c9 1393 unsigned long cluster = max_cluster;
ee22184b 1394 u64 new_align = ~((u64)SZ_128K - 1);
4cb5300b 1395 struct page **pages = NULL;
1e2ef46d 1396 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
4cb5300b 1397
0abd5b17
LB		 1398 if (isize == 0)
1399 return 0;
1400
1401 if (range->start >= isize)
1402 return -EINVAL;
1a419d85 1403
1e2ef46d 1404 if (do_compress) {
ce96b7ff 1405 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1a419d85
LZ		 1406 return -EINVAL;
1407 if (range->compress_type)
1408 compress_type = range->compress_type;
1409 }
f46b5a66 1410
0abd5b17 1411 if (extent_thresh == 0)
ee22184b 1412 extent_thresh = SZ_256K;
940100a4 1413
4cb5300b 1414 /*
0a52d108
DS		 1415 * If we were not given a file, allocate a readahead context. As
1416 * readahead is just an optimization, defrag will work without it so
1417 * we don't error out.
4cb5300b
CM		 1418 */
1419 if (!file) {
63e727ec 1420 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
0a52d108
DS		 1421 if (ra)
1422 file_ra_state_init(ra, inode->i_mapping);
4cb5300b
CM		 1423 } else {
1424 ra = &file->f_ra;
1425 }
1426
63e727ec 1427 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
4cb5300b
CM		 1428 if (!pages) {
1429 ret = -ENOMEM;
1430 goto out_ra;
1431 }
1432
1433 /* find the last page to defrag */
1e701a32 1434 if (range->start + range->len > range->start) {
151a31b2 1435 last_index = min_t(u64, isize - 1,
09cbfeaf 1436 range->start + range->len - 1) >> PAGE_SHIFT;
1e701a32 1437 } else {
09cbfeaf 1438 last_index = (isize - 1) >> PAGE_SHIFT;
1e701a32
CM		 1439 }
1440
4cb5300b
CM		 1441 if (newer_than) {
1442 ret = find_new_extents(root, inode, newer_than,
ee22184b 1443 &newer_off, SZ_64K);
4cb5300b
CM		 1444 if (!ret) {
1445 range->start = newer_off;
1446 /*
1447 * we always align our defrag to help keep
1448 * the extents in the file evenly spaced
1449 */
09cbfeaf 1450 i = (newer_off & new_align) >> PAGE_SHIFT;
4cb5300b
CM		 1451 } else
1452 goto out_ra;
1453 } else {
09cbfeaf 1454 i = range->start >> PAGE_SHIFT;
4cb5300b
CM		 1455 }
1456 if (!max_to_defrag)
070034bd 1457 max_to_defrag = last_index - i + 1;
4cb5300b 1458
2a0f7f57
LZ		 1459 /*
1460 * make writeback starts from i, so the defrag range can be
1461 * written sequentially.
1462 */
1463 if (i < inode->i_mapping->writeback_index)
1464 inode->i_mapping->writeback_index = i;
1465
f7f43cc8 1466 while (i <= last_index && defrag_count < max_to_defrag &&
09cbfeaf 1467 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
4cb5300b
CM		 1468 /*
1469 * make sure we stop running if someone unmounts
1470 * the FS
1471 */
1751e8a6 1472 if (!(inode->i_sb->s_flags & SB_ACTIVE))
4cb5300b
CM		 1473 break;
1474
0b246afa
JM		 1475 if (btrfs_defrag_cancelled(fs_info)) {
1476 btrfs_debug(fs_info, "defrag_file cancelled");
210549eb
DS		 1477 ret = -EAGAIN;
1478 break;
1479 }
1480
09cbfeaf 1481 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
6c282eb4 1482 extent_thresh, &last_len, &skip,
1e2ef46d 1483 &defrag_end, do_compress)){
940100a4
CM		 1484 unsigned long next;
1485 /*
1486 * the should_defrag function tells us how much to skip
1487 * bump our counter by the suggested amount
1488 */
09cbfeaf 1489 next = DIV_ROUND_UP(skip, PAGE_SIZE);
940100a4
CM		 1490 i = max(i + 1, next);
1491 continue;
1492 }
008873ea
LZ		 1493
1494 if (!newer_than) {
09cbfeaf
KS		 1495 cluster = (PAGE_ALIGN(defrag_end) >>
1496 PAGE_SHIFT) - i;
008873ea
LZ		 1497 cluster = min(cluster, max_cluster);
1498 } else {
1499 cluster = max_cluster;
1500 }
1501
008873ea
LZ		 1502 if (i + cluster > ra_index) {
1503 ra_index = max(i, ra_index);
0a52d108 1504 if (ra)
d3c0bab5
DS		 1505 page_cache_sync_readahead(inode->i_mapping, ra,
1506 file, ra_index, cluster);
e4826a5b 1507 ra_index += cluster;
008873ea 1508 }
940100a4 1509
5955102c 1510 inode_lock(inode);
eede2bf3
OS		 1511 if (IS_SWAPFILE(inode)) {
1512 ret = -ETXTBSY;
1513 } else {
1514 if (do_compress)
1515 BTRFS_I(inode)->defrag_compress = compress_type;
1516 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1517 }
ecb8bea8 1518 if (ret < 0) {
5955102c 1519 inode_unlock(inode);
4cb5300b 1520 goto out_ra;
ecb8bea8 1521 }
4cb5300b
CM		 1522
1523 defrag_count += ret;
d0e1d66b 1524 balance_dirty_pages_ratelimited(inode->i_mapping);
5955102c 1525 inode_unlock(inode);
4cb5300b
CM		 1526
1527 if (newer_than) {
1528 if (newer_off == (u64)-1)
1529 break;
1530
e1f041e1
LB		 1531 if (ret > 0)
1532 i += ret;
1533
4cb5300b 1534 newer_off = max(newer_off + 1,
09cbfeaf 1535 (u64)i << PAGE_SHIFT);
4cb5300b 1536
ee22184b
BL		 1537 ret = find_new_extents(root, inode, newer_than,
1538 &newer_off, SZ_64K);
4cb5300b
CM		 1539 if (!ret) {
1540 range->start = newer_off;
09cbfeaf 1541 i = (newer_off & new_align) >> PAGE_SHIFT;
4cb5300b
CM		 1542 } else {
1543 break;
f46b5a66 1544 }
4cb5300b 1545 } else {
008873ea 1546 if (ret > 0) {
cbcc8326 1547 i += ret;
09cbfeaf 1548 last_len += ret << PAGE_SHIFT;
008873ea 1549 } else {
cbcc8326 1550 i++;
008873ea
LZ		 1551 last_len = 0;
1552 }
f46b5a66 1553 }
f46b5a66
CH		 1554 }
1555
dec8ef90 1556 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1e701a32 1557 filemap_flush(inode->i_mapping);
dec8ef90
FM		 1558 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1559 &BTRFS_I(inode)->runtime_flags))
1560 filemap_flush(inode->i_mapping);
1561 }
1e701a32 1562
1a419d85 1563 if (range->compress_type == BTRFS_COMPRESS_LZO) {
0b246afa 1564 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
5c1aab1d
NT		 1565 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1566 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1a419d85
LZ		 1567 }
1568
60ccf82f 1569 ret = defrag_count;
940100a4 1570
4cb5300b 1571out_ra:
1e2ef46d 1572 if (do_compress) {
5955102c 1573 inode_lock(inode);
eec63c65 1574 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
5955102c 1575 inode_unlock(inode);
633085c7 1576 }
4cb5300b
CM		 1577 if (!file)
1578 kfree(ra);
1579 kfree(pages);
940100a4 1580 return ret;
f46b5a66
CH		 1581}
1582
198605a8 1583static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1584 void __user *arg)
f46b5a66 1585{
0b246afa
JM		 1586 struct inode *inode = file_inode(file);
1587 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66
CH		 1588 u64 new_size;
1589 u64 old_size;
1590 u64 devid = 1;
0b246afa 1591 struct btrfs_root *root = BTRFS_I(inode)->root;
f46b5a66
CH		 1592 struct btrfs_ioctl_vol_args *vol_args;
1593 struct btrfs_trans_handle *trans;
1594 struct btrfs_device *device = NULL;
1595 char *sizestr;
9a40f122 1596 char *retptr;
f46b5a66
CH		 1597 char *devstr = NULL;
1598 int ret = 0;
f46b5a66
CH		 1599 int mod = 0;
1600
e441d54d
CM		 1601 if (!capable(CAP_SYS_ADMIN))
1602 return -EPERM;
1603
198605a8
MX		 1604 ret = mnt_want_write_file(file);
1605 if (ret)
1606 return ret;
1607
171938e5 1608 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
97547676 1609 mnt_drop_write_file(file);
e57138b3 1610 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b
ID		 1611 }
1612
dae7b665 1613 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1614 if (IS_ERR(vol_args)) {
1615 ret = PTR_ERR(vol_args);
1616 goto out;
1617 }
5516e595
MF		 1618
1619 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1620
f46b5a66
CH		 1621 sizestr = vol_args->name;
1622 devstr = strchr(sizestr, ':');
1623 if (devstr) {
f46b5a66
CH		 1624 sizestr = devstr + 1;
1625 *devstr = '\0';
1626 devstr = vol_args->name;
58dfae63
Z		 1627 ret = kstrtoull(devstr, 10, &devid);
1628 if (ret)
1629 goto out_free;
dfd79829
MX		 1630 if (!devid) {
1631 ret = -EINVAL;
1632 goto out_free;
1633 }
0b246afa 1634 btrfs_info(fs_info, "resizing devid %llu", devid);
f46b5a66 1635 }
dba60f3f 1636
09ba3bc9 1637 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true);
f46b5a66 1638 if (!device) {
0b246afa
JM		 1639 btrfs_info(fs_info, "resizer unable to find device %llu",
1640 devid);
dfd79829 1641 ret = -ENODEV;
c9e9f97b 1642 goto out_free;
f46b5a66 1643 }
dba60f3f 1644
ebbede42 1645 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
0b246afa 1646 btrfs_info(fs_info,
efe120a0 1647 "resizer unable to apply on readonly device %llu",
c1c9ff7c 1648 devid);
dfd79829 1649 ret = -EPERM;
4e42ae1b
LB		 1650 goto out_free;
1651 }
1652
f46b5a66
CH		 1653 if (!strcmp(sizestr, "max"))
1654 new_size = device->bdev->bd_inode->i_size;
1655 else {
1656 if (sizestr[0] == '-') {
1657 mod = -1;
1658 sizestr++;
1659 } else if (sizestr[0] == '+') {
1660 mod = 1;
1661 sizestr++;
1662 }
9a40f122
GH		 1663 new_size = memparse(sizestr, &retptr);
1664 if (*retptr != '\0' || new_size == 0) {
f46b5a66 1665 ret = -EINVAL;
c9e9f97b 1666 goto out_free;
f46b5a66
CH		 1667 }
1668 }
1669
401e29c1 1670 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
dfd79829 1671 ret = -EPERM;
63a212ab
SB		 1672 goto out_free;
1673 }
1674
7cc8e58d 1675 old_size = btrfs_device_get_total_bytes(device);
f46b5a66
CH		 1676
1677 if (mod < 0) {
1678 if (new_size > old_size) {
1679 ret = -EINVAL;
c9e9f97b 1680 goto out_free;
f46b5a66
CH		 1681 }
1682 new_size = old_size - new_size;
1683 } else if (mod > 0) {
eb8052e0 1684 if (new_size > ULLONG_MAX - old_size) {
902c68a4 1685 ret = -ERANGE;
eb8052e0
WF		 1686 goto out_free;
1687 }
f46b5a66
CH		 1688 new_size = old_size + new_size;
1689 }
1690
ee22184b 1691 if (new_size < SZ_256M) {
f46b5a66 1692 ret = -EINVAL;
c9e9f97b 1693 goto out_free;
f46b5a66
CH		 1694 }
1695 if (new_size > device->bdev->bd_inode->i_size) {
1696 ret = -EFBIG;
c9e9f97b 1697 goto out_free;
f46b5a66
CH		 1698 }
1699
47f08b96 1700 new_size = round_down(new_size, fs_info->sectorsize);
f46b5a66 1701
f46b5a66 1702 if (new_size > old_size) {
a22285a6 1703 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1704 if (IS_ERR(trans)) {
1705 ret = PTR_ERR(trans);
c9e9f97b 1706 goto out_free;
98d5dc13 1707 }
f46b5a66 1708 ret = btrfs_grow_device(trans, device, new_size);
3a45bb20 1709 btrfs_commit_transaction(trans);
ece7d20e 1710 } else if (new_size < old_size) {
f46b5a66 1711 ret = btrfs_shrink_device(device, new_size);
0253f40e 1712 } /* equal, nothing need to do */
f46b5a66 1713
faf8f7b9
MPS		 1714 if (ret == 0 && new_size != old_size)
1715 btrfs_info_in_rcu(fs_info,
1716 "resize device %s (devid %llu) from %llu to %llu",
1717 rcu_str_deref(device->name), device->devid,
1718 old_size, new_size);
c9e9f97b 1719out_free:
f46b5a66 1720 kfree(vol_args);
c9e9f97b 1721out:
171938e5 1722 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
18f39c41 1723 mnt_drop_write_file(file);
f46b5a66
CH		 1724 return ret;
1725}
1726
5d54c67e 1727static noinline int __btrfs_ioctl_snap_create(struct file *file,
52f75f4f 1728 const char *name, unsigned long fd, int subvol,
5d54c67e 1729 bool readonly,
8696c533 1730 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1731{
f46b5a66 1732 int namelen;
3de4586c 1733 int ret = 0;
f46b5a66 1734
325c50e3
JM		 1735 if (!S_ISDIR(file_inode(file)->i_mode))
1736 return -ENOTDIR;
1737
a874a63e
LB		 1738 ret = mnt_want_write_file(file);
1739 if (ret)
1740 goto out;
1741
72fd032e
SW		 1742 namelen = strlen(name);
1743 if (strchr(name, '/')) {
f46b5a66 1744 ret = -EINVAL;
a874a63e 1745 goto out_drop_write;
f46b5a66
CH		 1746 }
1747
16780cab
CM		 1748 if (name[0] == '.' &&
1749 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1750 ret = -EEXIST;
a874a63e 1751 goto out_drop_write;
16780cab
CM		 1752 }
1753
3de4586c 1754 if (subvol) {
72fd032e 1755 ret = btrfs_mksubvol(&file->f_path, name, namelen,
5d54c67e 1756 NULL, readonly, inherit);
cb8e7090 1757 } else {
2903ff01 1758 struct fd src = fdget(fd);
3de4586c 1759 struct inode *src_inode;
2903ff01 1760 if (!src.file) {
3de4586c 1761 ret = -EINVAL;
a874a63e 1762 goto out_drop_write;
3de4586c
CM		 1763 }
1764
496ad9aa
AV		 1765 src_inode = file_inode(src.file);
1766 if (src_inode->i_sb != file_inode(file)->i_sb) {
c79b4713 1767 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
efe120a0 1768 "Snapshot src from another FS");
23ad5b17 1769 ret = -EXDEV;
d0242061
DS		 1770 } else if (!inode_owner_or_capable(src_inode)) {
1771 /*
1772 * Subvolume creation is not restricted, but snapshots
1773 * are limited to own subvolumes only
1774 */
1775 ret = -EPERM;
ecd18815 1776 } else {
c11fbb6e 1777 ret = btrfs_mksnapshot(&file->f_path, name, namelen,
ecd18815 1778 BTRFS_I(src_inode)->root,
5d54c67e 1779 readonly, inherit);
3de4586c 1780 }
2903ff01 1781 fdput(src);
cb8e7090 1782 }
a874a63e
LB		 1783out_drop_write:
1784 mnt_drop_write_file(file);
f46b5a66 1785out:
72fd032e
SW		 1786 return ret;
1787}
1788
1789static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1790 void __user *arg, int subvol)
72fd032e 1791{
fa0d2b9b 1792 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1793 int ret;
1794
325c50e3
JM		 1795 if (!S_ISDIR(file_inode(file)->i_mode))
1796 return -ENOTDIR;
1797
fa0d2b9b
LZ		 1798 vol_args = memdup_user(arg, sizeof(*vol_args));
1799 if (IS_ERR(vol_args))
1800 return PTR_ERR(vol_args);
1801 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1802
5d54c67e
NB		 1803 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1804 subvol, false, NULL);
fdfb1e4f 1805
fa0d2b9b
LZ		 1806 kfree(vol_args);
1807 return ret;
1808}
fdfb1e4f 1809
fa0d2b9b
LZ		 1810static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1811 void __user *arg, int subvol)
1812{
1813 struct btrfs_ioctl_vol_args_v2 *vol_args;
1814 int ret;
b83cc969 1815 bool readonly = false;
6f72c7e2 1816 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1817
325c50e3
JM		 1818 if (!S_ISDIR(file_inode(file)->i_mode))
1819 return -ENOTDIR;
1820
fa0d2b9b
LZ		 1821 vol_args = memdup_user(arg, sizeof(*vol_args));
1822 if (IS_ERR(vol_args))
1823 return PTR_ERR(vol_args);
1824 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1825
673990db 1826 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
b83cc969 1827 ret = -EOPNOTSUPP;
c47ca32d 1828 goto free_args;
72fd032e 1829 }
fa0d2b9b 1830
b83cc969
LZ		 1831 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1832 readonly = true;
6f72c7e2 1833 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
09cbfeaf 1834 if (vol_args->size > PAGE_SIZE) {
6f72c7e2 1835 ret = -EINVAL;
c47ca32d 1836 goto free_args;
6f72c7e2
AJ		 1837 }
1838 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1839 if (IS_ERR(inherit)) {
1840 ret = PTR_ERR(inherit);
c47ca32d 1841 goto free_args;
6f72c7e2
AJ		 1842 }
1843 }
fa0d2b9b 1844
5d54c67e
NB		 1845 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1846 subvol, readonly, inherit);
c47ca32d
DC		 1847 if (ret)
1848 goto free_inherit;
c47ca32d 1849free_inherit:
6f72c7e2 1850 kfree(inherit);
c47ca32d
DC		 1851free_args:
1852 kfree(vol_args);
f46b5a66
CH		 1853 return ret;
1854}
1855
0caa102d
LZ		 1856static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1857 void __user *arg)
1858{
496ad9aa 1859 struct inode *inode = file_inode(file);
0b246afa 1860 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
0caa102d
LZ		 1861 struct btrfs_root *root = BTRFS_I(inode)->root;
1862 int ret = 0;
1863 u64 flags = 0;
1864
4a0cc7ca 1865 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1866 return -EINVAL;
1867
0b246afa 1868 down_read(&fs_info->subvol_sem);
0caa102d
LZ		 1869 if (btrfs_root_readonly(root))
1870 flags |= BTRFS_SUBVOL_RDONLY;
0b246afa 1871 up_read(&fs_info->subvol_sem);
0caa102d
LZ		 1872
1873 if (copy_to_user(arg, &flags, sizeof(flags)))
1874 ret = -EFAULT;
1875
1876 return ret;
1877}
1878
1879static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1880 void __user *arg)
1881{
496ad9aa 1882 struct inode *inode = file_inode(file);
0b246afa 1883 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
0caa102d
LZ		 1884 struct btrfs_root *root = BTRFS_I(inode)->root;
1885 struct btrfs_trans_handle *trans;
1886 u64 root_flags;
1887 u64 flags;
1888 int ret = 0;
1889
bd60ea0f
DS		 1890 if (!inode_owner_or_capable(inode))
1891 return -EPERM;
1892
b9ca0664
LB		 1893 ret = mnt_want_write_file(file);
1894 if (ret)
1895 goto out;
0caa102d 1896
4a0cc7ca 1897 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
b9ca0664
LB		 1898 ret = -EINVAL;
1899 goto out_drop_write;
1900 }
0caa102d 1901
b9ca0664
LB		 1902 if (copy_from_user(&flags, arg, sizeof(flags))) {
1903 ret = -EFAULT;
1904 goto out_drop_write;
1905 }
0caa102d 1906
b9ca0664
LB		 1907 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1908 ret = -EOPNOTSUPP;
1909 goto out_drop_write;
1910 }
0caa102d 1911
0b246afa 1912 down_write(&fs_info->subvol_sem);
0caa102d
LZ		 1913
1914 /* nothing to do */
1915 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1916 goto out_drop_sem;
0caa102d
LZ		 1917
1918 root_flags = btrfs_root_flags(&root->root_item);
2c686537 1919 if (flags & BTRFS_SUBVOL_RDONLY) {
0caa102d
LZ		 1920 btrfs_set_root_flags(&root->root_item,
1921 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1922 } else {
1923 /*
1924 * Block RO -> RW transition if this subvolume is involved in
1925 * send
1926 */
1927 spin_lock(&root->root_item_lock);
1928 if (root->send_in_progress == 0) {
1929 btrfs_set_root_flags(&root->root_item,
0caa102d 1930 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2c686537
DS		 1931 spin_unlock(&root->root_item_lock);
1932 } else {
1933 spin_unlock(&root->root_item_lock);
0b246afa
JM		 1934 btrfs_warn(fs_info,
1935 "Attempt to set subvolume %llu read-write during send",
1936 root->root_key.objectid);
2c686537
DS		 1937 ret = -EPERM;
1938 goto out_drop_sem;
1939 }
1940 }
0caa102d
LZ		 1941
1942 trans = btrfs_start_transaction(root, 1);
1943 if (IS_ERR(trans)) {
1944 ret = PTR_ERR(trans);
1945 goto out_reset;
1946 }
1947
0b246afa 1948 ret = btrfs_update_root(trans, fs_info->tree_root,
0caa102d 1949 &root->root_key, &root->root_item);
9417ebc8
NB		 1950 if (ret < 0) {
1951 btrfs_end_transaction(trans);
1952 goto out_reset;
1953 }
1954
1955 ret = btrfs_commit_transaction(trans);
0caa102d 1956
0caa102d
LZ		 1957out_reset:
1958 if (ret)
1959 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1960out_drop_sem:
0b246afa 1961 up_write(&fs_info->subvol_sem);
b9ca0664
LB		 1962out_drop_write:
1963 mnt_drop_write_file(file);
1964out:
0caa102d
LZ		 1965 return ret;
1966}
1967
ac8e9819
CM		 1968static noinline int key_in_sk(struct btrfs_key *key,
1969 struct btrfs_ioctl_search_key *sk)
1970{
abc6e134
CM		 1971 struct btrfs_key test;
1972 int ret;
1973
1974 test.objectid = sk->min_objectid;
1975 test.type = sk->min_type;
1976 test.offset = sk->min_offset;
1977
1978 ret = btrfs_comp_cpu_keys(key, &test);
1979 if (ret < 0)
ac8e9819 1980 return 0;
abc6e134
CM		 1981
1982 test.objectid = sk->max_objectid;
1983 test.type = sk->max_type;
1984 test.offset = sk->max_offset;
1985
1986 ret = btrfs_comp_cpu_keys(key, &test);
1987 if (ret > 0)
ac8e9819
CM		 1988 return 0;
1989 return 1;
1990}
1991
df397565 1992static noinline int copy_to_sk(struct btrfs_path *path,
ac8e9819
CM		 1993 struct btrfs_key *key,
1994 struct btrfs_ioctl_search_key *sk,
9b6e817d 1995 size_t *buf_size,
ba346b35 1996 char __user *ubuf,
ac8e9819
CM		 1997 unsigned long *sk_offset,
1998 int *num_found)
1999{
2000 u64 found_transid;
2001 struct extent_buffer *leaf;
2002 struct btrfs_ioctl_search_header sh;
dd81d459 2003 struct btrfs_key test;
ac8e9819
CM		 2004 unsigned long item_off;
2005 unsigned long item_len;
2006 int nritems;
2007 int i;
2008 int slot;
ac8e9819
CM		 2009 int ret = 0;
2010
2011 leaf = path->nodes[0];
2012 slot = path->slots[0];
2013 nritems = btrfs_header_nritems(leaf);
2014
2015 if (btrfs_header_generation(leaf) > sk->max_transid) {
2016 i = nritems;
2017 goto advance_key;
2018 }
2019 found_transid = btrfs_header_generation(leaf);
2020
2021 for (i = slot; i < nritems; i++) {
2022 item_off = btrfs_item_ptr_offset(leaf, i);
2023 item_len = btrfs_item_size_nr(leaf, i);
2024
03b71c6c
GP		 2025 btrfs_item_key_to_cpu(leaf, key, i);
2026 if (!key_in_sk(key, sk))
2027 continue;
2028
9b6e817d 2029 if (sizeof(sh) + item_len > *buf_size) {
8f5f6178
GH		 2030 if (*num_found) {
2031 ret = 1;
2032 goto out;
2033 }
2034
2035 /*
2036 * return one empty item back for v1, which does not
2037 * handle -EOVERFLOW
2038 */
2039
9b6e817d 2040 *buf_size = sizeof(sh) + item_len;
ac8e9819 2041 item_len = 0;
8f5f6178
GH		 2042 ret = -EOVERFLOW;
2043 }
ac8e9819 2044
9b6e817d 2045 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
ac8e9819 2046 ret = 1;
25c9bc2e 2047 goto out;
ac8e9819
CM		 2048 }
2049
ac8e9819
CM		 2050 sh.objectid = key->objectid;
2051 sh.offset = key->offset;
2052 sh.type = key->type;
2053 sh.len = item_len;
2054 sh.transid = found_transid;
2055
2056 /* copy search result header */
ba346b35
GH		 2057 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2058 ret = -EFAULT;
2059 goto out;
2060 }
2061
ac8e9819
CM		 2062 *sk_offset += sizeof(sh);
2063
2064 if (item_len) {
ba346b35 2065 char __user *up = ubuf + *sk_offset;
ac8e9819 2066 /* copy the item */
ba346b35
GH		 2067 if (read_extent_buffer_to_user(leaf, up,
2068 item_off, item_len)) {
2069 ret = -EFAULT;
2070 goto out;
2071 }
2072
ac8e9819 2073 *sk_offset += item_len;
ac8e9819 2074 }
e2156867 2075 (*num_found)++;
ac8e9819 2076
8f5f6178
GH		 2077 if (ret) /* -EOVERFLOW from above */
2078 goto out;
2079
25c9bc2e
GH		 2080 if (*num_found >= sk->nr_items) {
2081 ret = 1;
2082 goto out;
2083 }
ac8e9819
CM		 2084 }
2085advance_key:
abc6e134 2086 ret = 0;
dd81d459
NA		 2087 test.objectid = sk->max_objectid;
2088 test.type = sk->max_type;
2089 test.offset = sk->max_offset;
2090 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2091 ret = 1;
2092 else if (key->offset < (u64)-1)
ac8e9819 2093 key->offset++;
dd81d459 2094 else if (key->type < (u8)-1) {
abc6e134 2095 key->offset = 0;
ac8e9819 2096 key->type++;
dd81d459 2097 } else if (key->objectid < (u64)-1) {
abc6e134
CM		 2098 key->offset = 0;
2099 key->type = 0;
ac8e9819 2100 key->objectid++;
abc6e134
CM		 2101 } else
2102 ret = 1;
25c9bc2e 2103out:
ba346b35
GH		 2104 /*
2105 * 0: all items from this leaf copied, continue with next
2106 * 1: * more items can be copied, but unused buffer is too small
2107 * * all items were found
2108 * Either way, it will stops the loop which iterates to the next
2109 * leaf
2110 * -EOVERFLOW: item was to large for buffer
2111 * -EFAULT: could not copy extent buffer back to userspace
2112 */
ac8e9819
CM		 2113 return ret;
2114}
2115
2116static noinline int search_ioctl(struct inode *inode,
12544442 2117 struct btrfs_ioctl_search_key *sk,
9b6e817d 2118 size_t *buf_size,
ba346b35 2119 char __user *ubuf)
ac8e9819 2120{
0b246afa 2121 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
ac8e9819
CM		 2122 struct btrfs_root *root;
2123 struct btrfs_key key;
ac8e9819 2124 struct btrfs_path *path;
ac8e9819
CM		 2125 int ret;
2126 int num_found = 0;
2127 unsigned long sk_offset = 0;
2128
9b6e817d
GH		 2129 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2130 *buf_size = sizeof(struct btrfs_ioctl_search_header);
12544442 2131 return -EOVERFLOW;
9b6e817d 2132 }
12544442 2133
ac8e9819
CM		 2134 path = btrfs_alloc_path();
2135 if (!path)
2136 return -ENOMEM;
2137
2138 if (sk->tree_id == 0) {
2139 /* search the root of the inode that was passed */
00246528 2140 root = btrfs_grab_root(BTRFS_I(inode)->root);
ac8e9819 2141 } else {
56e9357a 2142 root = btrfs_get_fs_root(info, sk->tree_id, true);
ac8e9819 2143 if (IS_ERR(root)) {
ac8e9819 2144 btrfs_free_path(path);
ad1e3d56 2145 return PTR_ERR(root);
ac8e9819
CM		 2146 }
2147 }
2148
2149 key.objectid = sk->min_objectid;
2150 key.type = sk->min_type;
2151 key.offset = sk->min_offset;
2152
67871254 2153 while (1) {
6174d3cb 2154 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
ac8e9819
CM		 2155 if (ret != 0) {
2156 if (ret > 0)
2157 ret = 0;
2158 goto err;
2159 }
df397565 2160 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
ac8e9819 2161 &sk_offset, &num_found);
b3b4aa74 2162 btrfs_release_path(path);
25c9bc2e 2163 if (ret)
ac8e9819
CM		 2164 break;
2165
2166 }
8f5f6178
GH		 2167 if (ret > 0)
2168 ret = 0;
ac8e9819
CM		 2169err:
2170 sk->nr_items = num_found;
00246528 2171 btrfs_put_root(root);
ac8e9819
CM		 2172 btrfs_free_path(path);
2173 return ret;
2174}
2175
2176static noinline int btrfs_ioctl_tree_search(struct file *file,
2177 void __user *argp)
2178{
ba346b35
GH		 2179 struct btrfs_ioctl_search_args __user *uargs;
2180 struct btrfs_ioctl_search_key sk;
9b6e817d
GH		 2181 struct inode *inode;
2182 int ret;
2183 size_t buf_size;
ac8e9819
CM		 2184
2185 if (!capable(CAP_SYS_ADMIN))
2186 return -EPERM;
2187
ba346b35
GH		 2188 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2189
2190 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2191 return -EFAULT;
ac8e9819 2192
ba346b35 2193 buf_size = sizeof(uargs->buf);
ac8e9819 2194
496ad9aa 2195 inode = file_inode(file);
ba346b35 2196 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
8f5f6178
GH		 2197
2198 /*
2199 * In the origin implementation an overflow is handled by returning a
2200 * search header with a len of zero, so reset ret.
2201 */
2202 if (ret == -EOVERFLOW)
2203 ret = 0;
2204
ba346b35 2205 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
ac8e9819 2206 ret = -EFAULT;
ac8e9819
CM		 2207 return ret;
2208}
2209
cc68a8a5
GH		 2210static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2211 void __user *argp)
2212{
2213 struct btrfs_ioctl_search_args_v2 __user *uarg;
2214 struct btrfs_ioctl_search_args_v2 args;
2215 struct inode *inode;
2216 int ret;
2217 size_t buf_size;
ee22184b 2218 const size_t buf_limit = SZ_16M;
cc68a8a5
GH		 2219
2220 if (!capable(CAP_SYS_ADMIN))
2221 return -EPERM;
2222
2223 /* copy search header and buffer size */
2224 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2225 if (copy_from_user(&args, uarg, sizeof(args)))
2226 return -EFAULT;
2227
2228 buf_size = args.buf_size;
2229
cc68a8a5
GH		 2230 /* limit result size to 16MB */
2231 if (buf_size > buf_limit)
2232 buf_size = buf_limit;
2233
2234 inode = file_inode(file);
2235 ret = search_ioctl(inode, &args.key, &buf_size,
718dc5fa 2236 (char __user *)(&uarg->buf[0]));
cc68a8a5
GH		 2237 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2238 ret = -EFAULT;
2239 else if (ret == -EOVERFLOW &&
2240 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2241 ret = -EFAULT;
2242
ac8e9819
CM		 2243 return ret;
2244}
2245
98d377a0 2246/*
ac8e9819
CM		 2247 * Search INODE_REFs to identify path name of 'dirid' directory
2248 * in a 'tree_id' tree. and sets path name to 'name'.
2249 */
98d377a0
TH		 2250static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2251 u64 tree_id, u64 dirid, char *name)
2252{
2253 struct btrfs_root *root;
2254 struct btrfs_key key;
ac8e9819 2255 char *ptr;
98d377a0
TH		 2256 int ret = -1;
2257 int slot;
2258 int len;
2259 int total_len = 0;
2260 struct btrfs_inode_ref *iref;
2261 struct extent_buffer *l;
2262 struct btrfs_path *path;
2263
2264 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2265 name[0]='\0';
2266 return 0;
2267 }
2268
2269 path = btrfs_alloc_path();
2270 if (!path)
2271 return -ENOMEM;
2272
c8bcbfbd 2273 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
98d377a0 2274
56e9357a 2275 root = btrfs_get_fs_root(info, tree_id, true);
98d377a0 2276 if (IS_ERR(root)) {
ad1e3d56 2277 ret = PTR_ERR(root);
88234012
JB		 2278 root = NULL;
2279 goto out;
2280 }
98d377a0
TH		 2281
2282 key.objectid = dirid;
2283 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2284 key.offset = (u64)-1;
98d377a0 2285
67871254 2286 while (1) {
98d377a0
TH		 2287 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2288 if (ret < 0)
2289 goto out;
18674c6c
FDBM		 2290 else if (ret > 0) {
2291 ret = btrfs_previous_item(root, path, dirid,
2292 BTRFS_INODE_REF_KEY);
2293 if (ret < 0)
2294 goto out;
2295 else if (ret > 0) {
2296 ret = -ENOENT;
2297 goto out;
2298 }
2299 }
98d377a0
TH		 2300
2301 l = path->nodes[0];
2302 slot = path->slots[0];
2303 btrfs_item_key_to_cpu(l, &key, slot);
2304
98d377a0
TH		 2305 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2306 len = btrfs_inode_ref_name_len(l, iref);
2307 ptr -= len + 1;
2308 total_len += len + 1;
a696cf35
FDBM		 2309 if (ptr < name) {
2310 ret = -ENAMETOOLONG;
98d377a0 2311 goto out;
a696cf35 2312 }
98d377a0
TH		 2313
2314 *(ptr + len) = '/';
67871254 2315 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
98d377a0
TH		 2316
2317 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2318 break;
2319
b3b4aa74 2320 btrfs_release_path(path);
98d377a0 2321 key.objectid = key.offset;
8ad6fcab 2322 key.offset = (u64)-1;
98d377a0 2323 dirid = key.objectid;
98d377a0 2324 }
77906a50 2325 memmove(name, ptr, total_len);
67871254 2326 name[total_len] = '\0';
98d377a0
TH		 2327 ret = 0;
2328out:
00246528 2329 btrfs_put_root(root);
98d377a0 2330 btrfs_free_path(path);
ac8e9819
CM		 2331 return ret;
2332}
2333
23d0b79d
TM		 2334static int btrfs_search_path_in_tree_user(struct inode *inode,
2335 struct btrfs_ioctl_ino_lookup_user_args *args)
2336{
2337 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2338 struct super_block *sb = inode->i_sb;
2339 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2340 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2341 u64 dirid = args->dirid;
2342 unsigned long item_off;
2343 unsigned long item_len;
2344 struct btrfs_inode_ref *iref;
2345 struct btrfs_root_ref *rref;
b8a49ae1 2346 struct btrfs_root *root = NULL;
23d0b79d
TM		 2347 struct btrfs_path *path;
2348 struct btrfs_key key, key2;
2349 struct extent_buffer *leaf;
2350 struct inode *temp_inode;
2351 char *ptr;
2352 int slot;
2353 int len;
2354 int total_len = 0;
2355 int ret;
2356
2357 path = btrfs_alloc_path();
2358 if (!path)
2359 return -ENOMEM;
2360
2361 /*
2362 * If the bottom subvolume does not exist directly under upper_limit,
2363 * construct the path in from the bottom up.
2364 */
2365 if (dirid != upper_limit.objectid) {
2366 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2367
56e9357a 2368 root = btrfs_get_fs_root(fs_info, treeid, true);
23d0b79d
TM		 2369 if (IS_ERR(root)) {
2370 ret = PTR_ERR(root);
2371 goto out;
2372 }
2373
2374 key.objectid = dirid;
2375 key.type = BTRFS_INODE_REF_KEY;
2376 key.offset = (u64)-1;
2377 while (1) {
2378 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2379 if (ret < 0) {
b8a49ae1 2380 goto out_put;
23d0b79d
TM		 2381 } else if (ret > 0) {
2382 ret = btrfs_previous_item(root, path, dirid,
2383 BTRFS_INODE_REF_KEY);
2384 if (ret < 0) {
b8a49ae1 2385 goto out_put;
23d0b79d
TM		 2386 } else if (ret > 0) {
2387 ret = -ENOENT;
b8a49ae1 2388 goto out_put;
23d0b79d
TM		 2389 }
2390 }
2391
2392 leaf = path->nodes[0];
2393 slot = path->slots[0];
2394 btrfs_item_key_to_cpu(leaf, &key, slot);
2395
2396 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2397 len = btrfs_inode_ref_name_len(leaf, iref);
2398 ptr -= len + 1;
2399 total_len += len + 1;
2400 if (ptr < args->path) {
2401 ret = -ENAMETOOLONG;
b8a49ae1 2402 goto out_put;
23d0b79d
TM		 2403 }
2404
2405 *(ptr + len) = '/';
2406 read_extent_buffer(leaf, ptr,
2407 (unsigned long)(iref + 1), len);
2408
2409 /* Check the read+exec permission of this directory */
2410 ret = btrfs_previous_item(root, path, dirid,
2411 BTRFS_INODE_ITEM_KEY);
2412 if (ret < 0) {
b8a49ae1 2413 goto out_put;
23d0b79d
TM		 2414 } else if (ret > 0) {
2415 ret = -ENOENT;
b8a49ae1 2416 goto out_put;
23d0b79d
TM		 2417 }
2418
2419 leaf = path->nodes[0];
2420 slot = path->slots[0];
2421 btrfs_item_key_to_cpu(leaf, &key2, slot);
2422 if (key2.objectid != dirid) {
2423 ret = -ENOENT;
b8a49ae1 2424 goto out_put;
23d0b79d
TM		 2425 }
2426
0202e83f 2427 temp_inode = btrfs_iget(sb, key2.objectid, root);
3ca57bd6
MT		 2428 if (IS_ERR(temp_inode)) {
2429 ret = PTR_ERR(temp_inode);
b8a49ae1 2430 goto out_put;
3ca57bd6 2431 }
23d0b79d
TM		 2432 ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2433 iput(temp_inode);
2434 if (ret) {
2435 ret = -EACCES;
b8a49ae1 2436 goto out_put;
23d0b79d
TM		 2437 }
2438
2439 if (key.offset == upper_limit.objectid)
2440 break;
2441 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2442 ret = -EACCES;
b8a49ae1 2443 goto out_put;
23d0b79d
TM		 2444 }
2445
2446 btrfs_release_path(path);
2447 key.objectid = key.offset;
2448 key.offset = (u64)-1;
2449 dirid = key.objectid;
2450 }
2451
2452 memmove(args->path, ptr, total_len);
2453 args->path[total_len] = '\0';
00246528 2454 btrfs_put_root(root);
b8a49ae1 2455 root = NULL;
23d0b79d
TM		 2456 btrfs_release_path(path);
2457 }
2458
2459 /* Get the bottom subvolume's name from ROOT_REF */
23d0b79d
TM		 2460 key.objectid = treeid;
2461 key.type = BTRFS_ROOT_REF_KEY;
2462 key.offset = args->treeid;
b8a49ae1 2463 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
23d0b79d
TM		 2464 if (ret < 0) {
2465 goto out;
2466 } else if (ret > 0) {
2467 ret = -ENOENT;
2468 goto out;
2469 }
2470
2471 leaf = path->nodes[0];
2472 slot = path->slots[0];
2473 btrfs_item_key_to_cpu(leaf, &key, slot);
2474
2475 item_off = btrfs_item_ptr_offset(leaf, slot);
2476 item_len = btrfs_item_size_nr(leaf, slot);
2477 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2478 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2479 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2480 ret = -EINVAL;
2481 goto out;
2482 }
2483
2484 /* Copy subvolume's name */
2485 item_off += sizeof(struct btrfs_root_ref);
2486 item_len -= sizeof(struct btrfs_root_ref);
2487 read_extent_buffer(leaf, args->name, item_off, item_len);
2488 args->name[item_len] = 0;
2489
b8a49ae1 2490out_put:
00246528 2491 btrfs_put_root(root);
23d0b79d
TM		 2492out:
2493 btrfs_free_path(path);
2494 return ret;
2495}
2496
ac8e9819
CM		 2497static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2498 void __user *argp)
2499{
bece2e82
BVA		 2500 struct btrfs_ioctl_ino_lookup_args *args;
2501 struct inode *inode;
01b810b8 2502 int ret = 0;
ac8e9819 2503
2354d08f
JL		 2504 args = memdup_user(argp, sizeof(*args));
2505 if (IS_ERR(args))
2506 return PTR_ERR(args);
c2b96929 2507
496ad9aa 2508 inode = file_inode(file);
ac8e9819 2509
01b810b8
DS		 2510 /*
2511 * Unprivileged query to obtain the containing subvolume root id. The
2512 * path is reset so it's consistent with btrfs_search_path_in_tree.
2513 */
1b53ac4d
CM		 2514 if (args->treeid == 0)
2515 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2516
01b810b8
DS		 2517 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2518 args->name[0] = 0;
2519 goto out;
2520 }
2521
2522 if (!capable(CAP_SYS_ADMIN)) {
2523 ret = -EPERM;
2524 goto out;
2525 }
2526
ac8e9819
CM		 2527 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2528 args->treeid, args->objectid,
2529 args->name);
2530
01b810b8 2531out:
ac8e9819
CM		 2532 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2533 ret = -EFAULT;
2534
2535 kfree(args);
98d377a0
TH		 2536 return ret;
2537}
2538
23d0b79d
TM		 2539/*
2540 * Version of ino_lookup ioctl (unprivileged)
2541 *
2542 * The main differences from ino_lookup ioctl are:
2543 *
2544 * 1. Read + Exec permission will be checked using inode_permission() during
2545 * path construction. -EACCES will be returned in case of failure.
2546 * 2. Path construction will be stopped at the inode number which corresponds
2547 * to the fd with which this ioctl is called. If constructed path does not
2548 * exist under fd's inode, -EACCES will be returned.
2549 * 3. The name of bottom subvolume is also searched and filled.
2550 */
2551static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2552{
2553 struct btrfs_ioctl_ino_lookup_user_args *args;
2554 struct inode *inode;
2555 int ret;
2556
2557 args = memdup_user(argp, sizeof(*args));
2558 if (IS_ERR(args))
2559 return PTR_ERR(args);
2560
2561 inode = file_inode(file);
2562
2563 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2564 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2565 /*
2566 * The subvolume does not exist under fd with which this is
2567 * called
2568 */
2569 kfree(args);
2570 return -EACCES;
2571 }
2572
2573 ret = btrfs_search_path_in_tree_user(inode, args);
2574
2575 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2576 ret = -EFAULT;
2577
2578 kfree(args);
2579 return ret;
2580}
2581
b64ec075
TM		 2582/* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2583static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2584{
2585 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2586 struct btrfs_fs_info *fs_info;
2587 struct btrfs_root *root;
2588 struct btrfs_path *path;
2589 struct btrfs_key key;
2590 struct btrfs_root_item *root_item;
2591 struct btrfs_root_ref *rref;
2592 struct extent_buffer *leaf;
2593 unsigned long item_off;
2594 unsigned long item_len;
2595 struct inode *inode;
2596 int slot;
2597 int ret = 0;
2598
2599 path = btrfs_alloc_path();
2600 if (!path)
2601 return -ENOMEM;
2602
2603 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2604 if (!subvol_info) {
2605 btrfs_free_path(path);
2606 return -ENOMEM;
2607 }
2608
2609 inode = file_inode(file);
2610 fs_info = BTRFS_I(inode)->root->fs_info;
2611
2612 /* Get root_item of inode's subvolume */
2613 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
56e9357a 2614 root = btrfs_get_fs_root(fs_info, key.objectid, true);
b64ec075
TM		 2615 if (IS_ERR(root)) {
2616 ret = PTR_ERR(root);
04734e84
JB		 2617 goto out_free;
2618 }
b64ec075
TM		 2619 root_item = &root->root_item;
2620
2621 subvol_info->treeid = key.objectid;
2622
2623 subvol_info->generation = btrfs_root_generation(root_item);
2624 subvol_info->flags = btrfs_root_flags(root_item);
2625
2626 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2627 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2628 BTRFS_UUID_SIZE);
2629 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2630 BTRFS_UUID_SIZE);
2631
2632 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2633 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2634 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2635
2636 subvol_info->otransid = btrfs_root_otransid(root_item);
2637 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2638 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2639
2640 subvol_info->stransid = btrfs_root_stransid(root_item);
2641 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2642 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2643
2644 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2645 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2646 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2647
2648 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2649 /* Search root tree for ROOT_BACKREF of this subvolume */
b64ec075
TM		 2650 key.type = BTRFS_ROOT_BACKREF_KEY;
2651 key.offset = 0;
04734e84 2652 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
b64ec075
TM		 2653 if (ret < 0) {
2654 goto out;
2655 } else if (path->slots[0] >=
2656 btrfs_header_nritems(path->nodes[0])) {
04734e84 2657 ret = btrfs_next_leaf(fs_info->tree_root, path);
b64ec075
TM		 2658 if (ret < 0) {
2659 goto out;
2660 } else if (ret > 0) {
2661 ret = -EUCLEAN;
2662 goto out;
2663 }
2664 }
2665
2666 leaf = path->nodes[0];
2667 slot = path->slots[0];
2668 btrfs_item_key_to_cpu(leaf, &key, slot);
2669 if (key.objectid == subvol_info->treeid &&
2670 key.type == BTRFS_ROOT_BACKREF_KEY) {
2671 subvol_info->parent_id = key.offset;
2672
2673 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2674 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2675
2676 item_off = btrfs_item_ptr_offset(leaf, slot)
2677 + sizeof(struct btrfs_root_ref);
2678 item_len = btrfs_item_size_nr(leaf, slot)
2679 - sizeof(struct btrfs_root_ref);
2680 read_extent_buffer(leaf, subvol_info->name,
2681 item_off, item_len);
2682 } else {
2683 ret = -ENOENT;
2684 goto out;
2685 }
2686 }
2687
2688 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2689 ret = -EFAULT;
2690
2691out:
00246528 2692 btrfs_put_root(root);
04734e84 2693out_free:
b64ec075 2694 btrfs_free_path(path);
b091f7fe 2695 kfree(subvol_info);
b64ec075
TM		 2696 return ret;
2697}
2698
42e4b520
TM		 2699/*
2700 * Return ROOT_REF information of the subvolume containing this inode
2701 * except the subvolume name.
2702 */
2703static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2704{
2705 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2706 struct btrfs_root_ref *rref;
2707 struct btrfs_root *root;
2708 struct btrfs_path *path;
2709 struct btrfs_key key;
2710 struct extent_buffer *leaf;
2711 struct inode *inode;
2712 u64 objectid;
2713 int slot;
2714 int ret;
2715 u8 found;
2716
2717 path = btrfs_alloc_path();
2718 if (!path)
2719 return -ENOMEM;
2720
2721 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2722 if (IS_ERR(rootrefs)) {
2723 btrfs_free_path(path);
2724 return PTR_ERR(rootrefs);
2725 }
2726
2727 inode = file_inode(file);
2728 root = BTRFS_I(inode)->root->fs_info->tree_root;
2729 objectid = BTRFS_I(inode)->root->root_key.objectid;
2730
2731 key.objectid = objectid;
2732 key.type = BTRFS_ROOT_REF_KEY;
2733 key.offset = rootrefs->min_treeid;
2734 found = 0;
2735
2736 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2737 if (ret < 0) {
2738 goto out;
2739 } else if (path->slots[0] >=
2740 btrfs_header_nritems(path->nodes[0])) {
2741 ret = btrfs_next_leaf(root, path);
2742 if (ret < 0) {
2743 goto out;
2744 } else if (ret > 0) {
2745 ret = -EUCLEAN;
2746 goto out;
2747 }
2748 }
2749 while (1) {
2750 leaf = path->nodes[0];
2751 slot = path->slots[0];
2752
2753 btrfs_item_key_to_cpu(leaf, &key, slot);
2754 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2755 ret = 0;
2756 goto out;
2757 }
2758
2759 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2760 ret = -EOVERFLOW;
2761 goto out;
2762 }
2763
2764 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2765 rootrefs->rootref[found].treeid = key.offset;
2766 rootrefs->rootref[found].dirid =
2767 btrfs_root_ref_dirid(leaf, rref);
2768 found++;
2769
2770 ret = btrfs_next_item(root, path);
2771 if (ret < 0) {
2772 goto out;
2773 } else if (ret > 0) {
2774 ret = -EUCLEAN;
2775 goto out;
2776 }
2777 }
2778
2779out:
2780 if (!ret || ret == -EOVERFLOW) {
2781 rootrefs->num_items = found;
2782 /* update min_treeid for next search */
2783 if (found)
2784 rootrefs->min_treeid =
2785 rootrefs->rootref[found - 1].treeid + 1;
2786 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2787 ret = -EFAULT;
2788 }
2789
2790 kfree(rootrefs);
2791 btrfs_free_path(path);
2792
2793 return ret;
2794}
2795
76dda93c 2796static noinline int btrfs_ioctl_snap_destroy(struct file *file,
949964c9
MPS		 2797 void __user *arg,
2798 bool destroy_v2)
76dda93c 2799{
54563d41 2800 struct dentry *parent = file->f_path.dentry;
0b246afa 2801 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
76dda93c 2802 struct dentry *dentry;
2b0143b5 2803 struct inode *dir = d_inode(parent);
76dda93c
YZ		 2804 struct inode *inode;
2805 struct btrfs_root *root = BTRFS_I(dir)->root;
2806 struct btrfs_root *dest = NULL;
949964c9
MPS		 2807 struct btrfs_ioctl_vol_args *vol_args = NULL;
2808 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2809 char *subvol_name, *subvol_name_ptr = NULL;
2810 int subvol_namelen;
76dda93c 2811 int err = 0;
949964c9 2812 bool destroy_parent = false;
76dda93c 2813
949964c9
MPS		 2814 if (destroy_v2) {
2815 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2816 if (IS_ERR(vol_args2))
2817 return PTR_ERR(vol_args2);
325c50e3 2818
949964c9
MPS		 2819 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2820 err = -EOPNOTSUPP;
2821 goto out;
2822 }
76dda93c 2823
949964c9
MPS		 2824 /*
2825 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2826 * name, same as v1 currently does.
2827 */
2828 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2829 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2830 subvol_name = vol_args2->name;
2831
2832 err = mnt_want_write_file(file);
2833 if (err)
2834 goto out;
2835 } else {
2836 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2837 err = -EINVAL;
2838 goto out;
2839 }
2840
2841 err = mnt_want_write_file(file);
2842 if (err)
2843 goto out;
2844
2845 dentry = btrfs_get_dentry(fs_info->sb,
2846 BTRFS_FIRST_FREE_OBJECTID,
2847 vol_args2->subvolid, 0, 0);
2848 if (IS_ERR(dentry)) {
2849 err = PTR_ERR(dentry);
2850 goto out_drop_write;
2851 }
2852
2853 /*
2854 * Change the default parent since the subvolume being
2855 * deleted can be outside of the current mount point.
2856 */
2857 parent = btrfs_get_parent(dentry);
2858
2859 /*
2860 * At this point dentry->d_name can point to '/' if the
2861 * subvolume we want to destroy is outsite of the
2862 * current mount point, so we need to release the
2863 * current dentry and execute the lookup to return a new
2864 * one with ->d_name pointing to the
2865 * <mount point>/subvol_name.
2866 */
2867 dput(dentry);
2868 if (IS_ERR(parent)) {
2869 err = PTR_ERR(parent);
2870 goto out_drop_write;
2871 }
2872 dir = d_inode(parent);
2873
2874 /*
2875 * If v2 was used with SPEC_BY_ID, a new parent was
2876 * allocated since the subvolume can be outside of the
2877 * current mount point. Later on we need to release this
2878 * new parent dentry.
2879 */
2880 destroy_parent = true;
2881
2882 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2883 fs_info, vol_args2->subvolid);
2884 if (IS_ERR(subvol_name_ptr)) {
2885 err = PTR_ERR(subvol_name_ptr);
2886 goto free_parent;
2887 }
2888 /* subvol_name_ptr is already NULL termined */
2889 subvol_name = (char *)kbasename(subvol_name_ptr);
2890 }
2891 } else {
2892 vol_args = memdup_user(arg, sizeof(*vol_args));
2893 if (IS_ERR(vol_args))
2894 return PTR_ERR(vol_args);
2895
2896 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2897 subvol_name = vol_args->name;
2898
2899 err = mnt_want_write_file(file);
2900 if (err)
2901 goto out;
76dda93c
YZ		 2902 }
2903
949964c9 2904 subvol_namelen = strlen(subvol_name);
76dda93c 2905
949964c9
MPS		 2906 if (strchr(subvol_name, '/') ||
2907 strncmp(subvol_name, "..", subvol_namelen) == 0) {
2908 err = -EINVAL;
2909 goto free_subvol_name;
2910 }
2911
2912 if (!S_ISDIR(dir->i_mode)) {
2913 err = -ENOTDIR;
2914 goto free_subvol_name;
2915 }
521e0546 2916
00235411
AV		 2917 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2918 if (err == -EINTR)
949964c9
MPS		 2919 goto free_subvol_name;
2920 dentry = lookup_one_len(subvol_name, parent, subvol_namelen);
76dda93c
YZ		 2921 if (IS_ERR(dentry)) {
2922 err = PTR_ERR(dentry);
2923 goto out_unlock_dir;
2924 }
2925
2b0143b5 2926 if (d_really_is_negative(dentry)) {
76dda93c
YZ		 2927 err = -ENOENT;
2928 goto out_dput;
2929 }
2930
2b0143b5 2931 inode = d_inode(dentry);
4260f7c7 2932 dest = BTRFS_I(inode)->root;
67871254 2933 if (!capable(CAP_SYS_ADMIN)) {
4260f7c7
SW		 2934 /*
2935 * Regular user. Only allow this with a special mount
2936 * option, when the user has write+exec access to the
2937 * subvol root, and when rmdir(2) would have been
2938 * allowed.
2939 *
2940 * Note that this is _not_ check that the subvol is
2941 * empty or doesn't contain data that we wouldn't
2942 * otherwise be able to delete.
2943 *
2944 * Users who want to delete empty subvols should try
2945 * rmdir(2).
2946 */
2947 err = -EPERM;
0b246afa 2948 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
4260f7c7
SW		 2949 goto out_dput;
2950
2951 /*
2952 * Do not allow deletion if the parent dir is the same
2953 * as the dir to be deleted. That means the ioctl
2954 * must be called on the dentry referencing the root
2955 * of the subvol, not a random directory contained
2956 * within it.
2957 */
2958 err = -EINVAL;
2959 if (root == dest)
2960 goto out_dput;
2961
2962 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2963 if (err)
2964 goto out_dput;
4260f7c7
SW		 2965 }
2966
5c39da5b
MX		 2967 /* check if subvolume may be deleted by a user */
2968 err = btrfs_may_delete(dir, dentry, 1);
2969 if (err)
2970 goto out_dput;
2971
4a0cc7ca 2972 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2973 err = -EINVAL;
2974 goto out_dput;
2975 }
2976
5955102c 2977 inode_lock(inode);
f60a2364 2978 err = btrfs_delete_subvolume(dir, dentry);
5955102c 2979 inode_unlock(inode);
46008d9d
AG		 2980 if (!err) {
2981 fsnotify_rmdir(dir, dentry);
76dda93c 2982 d_delete(dentry);
46008d9d 2983 }
fa6ac876 2984
76dda93c
YZ		 2985out_dput:
2986 dput(dentry);
2987out_unlock_dir:
5955102c 2988 inode_unlock(dir);
949964c9
MPS		 2989free_subvol_name:
2990 kfree(subvol_name_ptr);
2991free_parent:
2992 if (destroy_parent)
2993 dput(parent);
00235411 2994out_drop_write:
2a79f17e 2995 mnt_drop_write_file(file);
76dda93c 2996out:
949964c9 2997 kfree(vol_args2);
76dda93c
YZ		 2998 kfree(vol_args);
2999 return err;
3000}
3001
1e701a32 3002static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 3003{
496ad9aa 3004 struct inode *inode = file_inode(file);
f46b5a66 3005 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 3006 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 3007 int ret;
3008
25122d15
ID		 3009 ret = mnt_want_write_file(file);
3010 if (ret)
3011 return ret;
b83cc969 3012
25122d15
ID		 3013 if (btrfs_root_readonly(root)) {
3014 ret = -EROFS;
3015 goto out;
5ac00add 3016 }
f46b5a66
CH		 3017
3018 switch (inode->i_mode & S_IFMT) {
3019 case S_IFDIR:
e441d54d
CM		 3020 if (!capable(CAP_SYS_ADMIN)) {
3021 ret = -EPERM;
3022 goto out;
3023 }
de78b51a 3024 ret = btrfs_defrag_root(root);
f46b5a66
CH		 3025 break;
3026 case S_IFREG:
616d374e
AB		 3027 /*
3028 * Note that this does not check the file descriptor for write
3029 * access. This prevents defragmenting executables that are
3030 * running and allows defrag on files open in read-only mode.
3031 */
3032 if (!capable(CAP_SYS_ADMIN) &&
3033 inode_permission(inode, MAY_WRITE)) {
3034 ret = -EPERM;
e441d54d
CM		 3035 goto out;
3036 }
1e701a32
CM		 3037
3038 range = kzalloc(sizeof(*range), GFP_KERNEL);
3039 if (!range) {
3040 ret = -ENOMEM;
3041 goto out;
3042 }
3043
3044 if (argp) {
3045 if (copy_from_user(range, argp,
3046 sizeof(*range))) {
3047 ret = -EFAULT;
3048 kfree(range);
683be16e 3049 goto out;
1e701a32
CM		 3050 }
3051 /* compression requires us to start the IO */
3052 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3053 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3054 range->extent_thresh = (u32)-1;
3055 }
3056 } else {
3057 /* the rest are all set to zero by kzalloc */
3058 range->len = (u64)-1;
3059 }
496ad9aa 3060 ret = btrfs_defrag_file(file_inode(file), file,
7c829b72 3061 range, BTRFS_OLDEST_GENERATION, 0);
4cb5300b
CM		 3062 if (ret > 0)
3063 ret = 0;
1e701a32 3064 kfree(range);
f46b5a66 3065 break;
8929ecfa
YZ		 3066 default:
3067 ret = -EINVAL;
f46b5a66 3068 }
e441d54d 3069out:
25122d15 3070 mnt_drop_write_file(file);
e441d54d 3071 return ret;
f46b5a66
CH		 3072}
3073
2ff7e61e 3074static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
f46b5a66
CH		 3075{
3076 struct btrfs_ioctl_vol_args *vol_args;
3077 int ret;
3078
e441d54d
CM		 3079 if (!capable(CAP_SYS_ADMIN))
3080 return -EPERM;
3081
171938e5 3082 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
e57138b3 3083 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
c9e9f97b 3084
dae7b665 3085 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 3086 if (IS_ERR(vol_args)) {
3087 ret = PTR_ERR(vol_args);
3088 goto out;
3089 }
f46b5a66 3090
5516e595 3091 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2ff7e61e 3092 ret = btrfs_init_new_device(fs_info, vol_args->name);
f46b5a66 3093
43d20761 3094 if (!ret)
0b246afa 3095 btrfs_info(fs_info, "disk added %s", vol_args->name);
43d20761 3096
f46b5a66 3097 kfree(vol_args);
c9e9f97b 3098out:
171938e5 3099 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
f46b5a66
CH		 3100 return ret;
3101}
3102
6b526ed7 3103static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
f46b5a66 3104{
0b246afa
JM		 3105 struct inode *inode = file_inode(file);
3106 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6b526ed7 3107 struct btrfs_ioctl_vol_args_v2 *vol_args;
f46b5a66
CH		 3108 int ret;
3109
e441d54d
CM		 3110 if (!capable(CAP_SYS_ADMIN))
3111 return -EPERM;
3112
da24927b
MX		 3113 ret = mnt_want_write_file(file);
3114 if (ret)
3115 return ret;
c146afad 3116
dae7b665 3117 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 3118 if (IS_ERR(vol_args)) {
3119 ret = PTR_ERR(vol_args);
c47ca32d 3120 goto err_drop;
c9e9f97b 3121 }
f46b5a66 3122
748449cd 3123 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
fd4e994b
OS		 3124 ret = -EOPNOTSUPP;
3125 goto out;
3126 }
f46b5a66 3127
171938e5 3128 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
183860f6
AJ		 3129 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3130 goto out;
3131 }
3132
735654ea 3133 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2ff7e61e 3134 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
6b526ed7
AJ		 3135 } else {
3136 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2ff7e61e 3137 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
6b526ed7 3138 }
171938e5 3139 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
183860f6 3140
6b526ed7 3141 if (!ret) {
735654ea 3142 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
0b246afa 3143 btrfs_info(fs_info, "device deleted: id %llu",
6b526ed7
AJ		 3144 vol_args->devid);
3145 else
0b246afa 3146 btrfs_info(fs_info, "device deleted: %s",
6b526ed7
AJ		 3147 vol_args->name);
3148 }
183860f6
AJ		 3149out:
3150 kfree(vol_args);
c47ca32d 3151err_drop:
4ac20c70 3152 mnt_drop_write_file(file);
f46b5a66
CH		 3153 return ret;
3154}
3155
da24927b 3156static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 3157{
0b246afa
JM		 3158 struct inode *inode = file_inode(file);
3159 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
f46b5a66
CH		 3160 struct btrfs_ioctl_vol_args *vol_args;
3161 int ret;
3162
e441d54d
CM		 3163 if (!capable(CAP_SYS_ADMIN))
3164 return -EPERM;
3165
da24927b
MX		 3166 ret = mnt_want_write_file(file);
3167 if (ret)
3168 return ret;
c146afad 3169
171938e5 3170 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
183860f6 3171 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
58d7bbf8
DS		 3172 goto out_drop_write;
3173 }
3174
3175 vol_args = memdup_user(arg, sizeof(*vol_args));
3176 if (IS_ERR(vol_args)) {
3177 ret = PTR_ERR(vol_args);
183860f6
AJ		 3178 goto out;
3179 }
3180
58d7bbf8 3181 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2ff7e61e 3182 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
183860f6 3183
ec95d491 3184 if (!ret)
0b246afa 3185 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
183860f6 3186 kfree(vol_args);
58d7bbf8 3187out:
171938e5 3188 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
58d7bbf8 3189out_drop_write:
4ac20c70 3190 mnt_drop_write_file(file);
58d7bbf8 3191
f46b5a66
CH		 3192 return ret;
3193}
3194
2ff7e61e
JM		 3195static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3196 void __user *arg)
475f6387 3197{
027ed2f0 3198 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387 3199 struct btrfs_device *device;
0b246afa 3200 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
027ed2f0 3201 int ret = 0;
475f6387 3202
027ed2f0
LZ		 3203 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
3204 if (!fi_args)
3205 return -ENOMEM;
3206
d03262c7 3207 rcu_read_lock();
027ed2f0 3208 fi_args->num_devices = fs_devices->num_devices;
475f6387 3209
d03262c7 3210 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 3211 if (device->devid > fi_args->max_id)
3212 fi_args->max_id = device->devid;
475f6387 3213 }
d03262c7 3214 rcu_read_unlock();
475f6387 3215
de37aa51 3216 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
bea7eafd
OS		 3217 fi_args->nodesize = fs_info->nodesize;
3218 fi_args->sectorsize = fs_info->sectorsize;
3219 fi_args->clone_alignment = fs_info->sectorsize;
80a773fb 3220
027ed2f0
LZ		 3221 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3222 ret = -EFAULT;
475f6387 3223
027ed2f0
LZ		 3224 kfree(fi_args);
3225 return ret;
475f6387
JS		 3226}
3227
2ff7e61e
JM		 3228static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3229 void __user *arg)
475f6387
JS		 3230{
3231 struct btrfs_ioctl_dev_info_args *di_args;
3232 struct btrfs_device *dev;
475f6387
JS		 3233 int ret = 0;
3234 char *s_uuid = NULL;
475f6387 3235
475f6387
JS		 3236 di_args = memdup_user(arg, sizeof(*di_args));
3237 if (IS_ERR(di_args))
3238 return PTR_ERR(di_args);
3239
dd5f9615 3240 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 3241 s_uuid = di_args->uuid;
3242
c5593ca3 3243 rcu_read_lock();
e4319cd9 3244 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
09ba3bc9 3245 NULL, true);
475f6387
JS		 3246
3247 if (!dev) {
3248 ret = -ENODEV;
3249 goto out;
3250 }
3251
3252 di_args->devid = dev->devid;
7cc8e58d
MX		 3253 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3254 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
475f6387 3255 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 3256 if (dev->name) {
672d5990
MT		 3257 strncpy(di_args->path, rcu_str_deref(dev->name),
3258 sizeof(di_args->path) - 1);
a27202fb
JM		 3259 di_args->path[sizeof(di_args->path) - 1] = 0;
3260 } else {
99ba55ad 3261 di_args->path[0] = '\0';
a27202fb 3262 }
475f6387
JS		 3263
3264out:
c5593ca3 3265 rcu_read_unlock();
475f6387
JS		 3266 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3267 ret = -EFAULT;
3268
3269 kfree(di_args);
3270 return ret;
3271}
3272
6ef5ed0d
JB		 3273static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3274{
496ad9aa 3275 struct inode *inode = file_inode(file);
0b246afa 3276 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
6ef5ed0d
JB		 3277 struct btrfs_root *root = BTRFS_I(inode)->root;
3278 struct btrfs_root *new_root;
3279 struct btrfs_dir_item *di;
3280 struct btrfs_trans_handle *trans;
2a2b5d62 3281 struct btrfs_path *path = NULL;
6ef5ed0d 3282 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3283 u64 objectid = 0;
3284 u64 dir_id;
3c04ce01 3285 int ret;
6ef5ed0d
JB		 3286
3287 if (!capable(CAP_SYS_ADMIN))
3288 return -EPERM;
3289
3c04ce01
MX		 3290 ret = mnt_want_write_file(file);
3291 if (ret)
3292 return ret;
3293
3294 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3295 ret = -EFAULT;
3296 goto out;
3297 }
6ef5ed0d
JB		 3298
3299 if (!objectid)
1cecf579 3300 objectid = BTRFS_FS_TREE_OBJECTID;
6ef5ed0d 3301
56e9357a 3302 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3c04ce01
MX		 3303 if (IS_ERR(new_root)) {
3304 ret = PTR_ERR(new_root);
3305 goto out;
3306 }
2a2b5d62
JB		 3307 if (!is_fstree(new_root->root_key.objectid)) {
3308 ret = -ENOENT;
3309 goto out_free;
3310 }
6ef5ed0d 3311
6ef5ed0d 3312 path = btrfs_alloc_path();
3c04ce01
MX		 3313 if (!path) {
3314 ret = -ENOMEM;
2a2b5d62 3315 goto out_free;
3c04ce01 3316 }
6ef5ed0d
JB		 3317 path->leave_spinning = 1;
3318
3319 trans = btrfs_start_transaction(root, 1);
98d5dc13 3320 if (IS_ERR(trans)) {
3c04ce01 3321 ret = PTR_ERR(trans);
2a2b5d62 3322 goto out_free;
6ef5ed0d
JB		 3323 }
3324
0b246afa
JM		 3325 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3326 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
6ef5ed0d 3327 dir_id, "default", 7, 1);
cf1e99a4 3328 if (IS_ERR_OR_NULL(di)) {
2a2b5d62 3329 btrfs_release_path(path);
3a45bb20 3330 btrfs_end_transaction(trans);
0b246afa 3331 btrfs_err(fs_info,
5d163e0e 3332 "Umm, you don't have the default diritem, this isn't going to work");
3c04ce01 3333 ret = -ENOENT;
2a2b5d62 3334 goto out_free;
6ef5ed0d
JB		 3335 }
3336
3337 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3338 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3339 btrfs_mark_buffer_dirty(path->nodes[0]);
2a2b5d62 3340 btrfs_release_path(path);
6ef5ed0d 3341
0b246afa 3342 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3a45bb20 3343 btrfs_end_transaction(trans);
2a2b5d62 3344out_free:
00246528 3345 btrfs_put_root(new_root);
2a2b5d62 3346 btrfs_free_path(path);
3c04ce01
MX		 3347out:
3348 mnt_drop_write_file(file);
3349 return ret;
6ef5ed0d
JB		 3350}
3351
c065f5b1
SY		 3352static void get_block_group_info(struct list_head *groups_list,
3353 struct btrfs_ioctl_space_info *space)
bf5fc093 3354{
32da5386 3355 struct btrfs_block_group *block_group;
bf5fc093
JB		 3356
3357 space->total_bytes = 0;
3358 space->used_bytes = 0;
3359 space->flags = 0;
3360 list_for_each_entry(block_group, groups_list, list) {
3361 space->flags = block_group->flags;
b3470b5d 3362 space->total_bytes += block_group->length;
bf38be65 3363 space->used_bytes += block_group->used;
bf5fc093
JB		 3364 }
3365}
3366
2ff7e61e
JM		 3367static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3368 void __user *arg)
1406e432
JB		 3369{
3370 struct btrfs_ioctl_space_args space_args;
3371 struct btrfs_ioctl_space_info space;
3372 struct btrfs_ioctl_space_info *dest;
7fde62bf 3373 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3374 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3375 struct btrfs_space_info *info;
315d8e98
CIK		 3376 static const u64 types[] = {
3377 BTRFS_BLOCK_GROUP_DATA,
3378 BTRFS_BLOCK_GROUP_SYSTEM,
3379 BTRFS_BLOCK_GROUP_METADATA,
3380 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3381 };
bf5fc093 3382 int num_types = 4;
7fde62bf 3383 int alloc_size;
1406e432 3384 int ret = 0;
51788b1b 3385 u64 slot_count = 0;
bf5fc093 3386 int i, c;
1406e432
JB		 3387
3388 if (copy_from_user(&space_args,
3389 (struct btrfs_ioctl_space_args __user *)arg,
3390 sizeof(space_args)))
3391 return -EFAULT;
3392
bf5fc093
JB		 3393 for (i = 0; i < num_types; i++) {
3394 struct btrfs_space_info *tmp;
3395
3396 info = NULL;
3397 rcu_read_lock();
0b246afa 3398 list_for_each_entry_rcu(tmp, &fs_info->space_info,
bf5fc093
JB		 3399 list) {
3400 if (tmp->flags == types[i]) {
3401 info = tmp;
3402 break;
3403 }
3404 }
3405 rcu_read_unlock();
3406
3407 if (!info)
3408 continue;
3409
3410 down_read(&info->groups_sem);
3411 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3412 if (!list_empty(&info->block_groups[c]))
3413 slot_count++;
3414 }
3415 up_read(&info->groups_sem);
3416 }
7fde62bf 3417
36523e95
DS		 3418 /*
3419 * Global block reserve, exported as a space_info
3420 */
3421 slot_count++;
3422
7fde62bf
CM		 3423 /* space_slots == 0 means they are asking for a count */
3424 if (space_args.space_slots == 0) {
3425 space_args.total_spaces = slot_count;
3426 goto out;
3427 }
bf5fc093 3428
51788b1b 3429 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3430
7fde62bf 3431 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3432
7fde62bf
CM		 3433 /* we generally have at most 6 or so space infos, one for each raid
3434 * level. So, a whole page should be more than enough for everyone
3435 */
09cbfeaf 3436 if (alloc_size > PAGE_SIZE)
7fde62bf
CM		 3437 return -ENOMEM;
3438
1406e432 3439 space_args.total_spaces = 0;
8d2db785 3440 dest = kmalloc(alloc_size, GFP_KERNEL);
7fde62bf
CM		 3441 if (!dest)
3442 return -ENOMEM;
3443 dest_orig = dest;
1406e432 3444
7fde62bf 3445 /* now we have a buffer to copy into */
bf5fc093
JB		 3446 for (i = 0; i < num_types; i++) {
3447 struct btrfs_space_info *tmp;
3448
51788b1b
DR		 3449 if (!slot_count)
3450 break;
3451
bf5fc093
JB		 3452 info = NULL;
3453 rcu_read_lock();
0b246afa 3454 list_for_each_entry_rcu(tmp, &fs_info->space_info,
bf5fc093
JB		 3455 list) {
3456 if (tmp->flags == types[i]) {
3457 info = tmp;
3458 break;
3459 }
3460 }
3461 rcu_read_unlock();
7fde62bf 3462
bf5fc093
JB		 3463 if (!info)
3464 continue;
3465 down_read(&info->groups_sem);
3466 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3467 if (!list_empty(&info->block_groups[c])) {
c065f5b1
SY		 3468 get_block_group_info(&info->block_groups[c],
3469 &space);
bf5fc093
JB		 3470 memcpy(dest, &space, sizeof(space));
3471 dest++;
3472 space_args.total_spaces++;
51788b1b 3473 slot_count--;
bf5fc093 3474 }
51788b1b
DR		 3475 if (!slot_count)
3476 break;
bf5fc093
JB		 3477 }
3478 up_read(&info->groups_sem);
1406e432 3479 }
1406e432 3480
36523e95
DS		 3481 /*
3482 * Add global block reserve
3483 */
3484 if (slot_count) {
0b246afa 3485 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
36523e95
DS		 3486
3487 spin_lock(&block_rsv->lock);
3488 space.total_bytes = block_rsv->size;
3489 space.used_bytes = block_rsv->size - block_rsv->reserved;
3490 spin_unlock(&block_rsv->lock);
3491 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3492 memcpy(dest, &space, sizeof(space));
3493 space_args.total_spaces++;
3494 }
3495
2eec6c81 3496 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3497 (arg + sizeof(struct btrfs_ioctl_space_args));
3498
3499 if (copy_to_user(user_dest, dest_orig, alloc_size))
3500 ret = -EFAULT;
3501
3502 kfree(dest_orig);
3503out:
3504 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 3505 ret = -EFAULT;
3506
3507 return ret;
3508}
3509
9a8c28be
MX		 3510static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3511 void __user *argp)
46204592 3512{
46204592
SW		 3513 struct btrfs_trans_handle *trans;
3514 u64 transid;
db5b493a 3515 int ret;
46204592 3516
d4edf39b 3517 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 3518 if (IS_ERR(trans)) {
3519 if (PTR_ERR(trans) != -ENOENT)
3520 return PTR_ERR(trans);
3521
3522 /* No running transaction, don't bother */
3523 transid = root->fs_info->last_trans_committed;
3524 goto out;
3525 }
46204592 3526 transid = trans->transid;
3a45bb20 3527 ret = btrfs_commit_transaction_async(trans, 0);
8b2b2d3c 3528 if (ret) {
3a45bb20 3529 btrfs_end_transaction(trans);
db5b493a 3530 return ret;
8b2b2d3c 3531 }
ff7c1d33 3532out:
46204592
SW		 3533 if (argp)
3534 if (copy_to_user(argp, &transid, sizeof(transid)))
3535 return -EFAULT;
3536 return 0;
3537}
3538
2ff7e61e 3539static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
9a8c28be 3540 void __user *argp)
46204592 3541{
46204592
SW		 3542 u64 transid;
3543
3544 if (argp) {
3545 if (copy_from_user(&transid, argp, sizeof(transid)))
3546 return -EFAULT;
3547 } else {
3548 transid = 0; /* current trans */
3549 }
2ff7e61e 3550 return btrfs_wait_for_commit(fs_info, transid);
46204592
SW		 3551}
3552
b8e95489 3553static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 3554{
0b246afa 3555 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
475f6387 3556 struct btrfs_ioctl_scrub_args *sa;
b8e95489 3557 int ret;
475f6387
JS		 3558
3559 if (!capable(CAP_SYS_ADMIN))
3560 return -EPERM;
3561
3562 sa = memdup_user(arg, sizeof(*sa));
3563 if (IS_ERR(sa))
3564 return PTR_ERR(sa);
3565
b8e95489
MX		 3566 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3567 ret = mnt_want_write_file(file);
3568 if (ret)
3569 goto out;
3570 }
3571
0b246afa 3572 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 3573 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3574 0);
475f6387 3575
5afe6ce7
FM		 3576 /*
3577 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3578 * error. This is important as it allows user space to know how much
3579 * progress scrub has done. For example, if scrub is canceled we get
3580 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3581 * space. Later user space can inspect the progress from the structure
3582 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3583 * previously (btrfs-progs does this).
3584 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3585 * then return -EFAULT to signal the structure was not copied or it may
3586 * be corrupt and unreliable due to a partial copy.
3587 */
3588 if (copy_to_user(arg, sa, sizeof(*sa)))
475f6387
JS		 3589 ret = -EFAULT;
3590
b8e95489
MX		 3591 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3592 mnt_drop_write_file(file);
3593out:
475f6387
JS		 3594 kfree(sa);
3595 return ret;
3596}
3597
2ff7e61e 3598static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
475f6387
JS		 3599{
3600 if (!capable(CAP_SYS_ADMIN))
3601 return -EPERM;
3602
2ff7e61e 3603 return btrfs_scrub_cancel(fs_info);
475f6387
JS		 3604}
3605
2ff7e61e 3606static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
475f6387
JS		 3607 void __user *arg)
3608{
3609 struct btrfs_ioctl_scrub_args *sa;
3610 int ret;
3611
3612 if (!capable(CAP_SYS_ADMIN))
3613 return -EPERM;
3614
3615 sa = memdup_user(arg, sizeof(*sa));
3616 if (IS_ERR(sa))
3617 return PTR_ERR(sa);
3618
2ff7e61e 3619 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
475f6387 3620
4fa99b00 3621 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
475f6387
JS		 3622 ret = -EFAULT;
3623
3624 kfree(sa);
3625 return ret;
3626}
3627
2ff7e61e 3628static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
b27f7c0c 3629 void __user *arg)
c11d2c23
SB		 3630{
3631 struct btrfs_ioctl_get_dev_stats *sa;
3632 int ret;
3633
c11d2c23
SB		 3634 sa = memdup_user(arg, sizeof(*sa));
3635 if (IS_ERR(sa))
3636 return PTR_ERR(sa);
3637
b27f7c0c
DS		 3638 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3639 kfree(sa);
3640 return -EPERM;
3641 }
3642
2ff7e61e 3643 ret = btrfs_get_dev_stats(fs_info, sa);
c11d2c23 3644
eee99577 3645 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
c11d2c23
SB		 3646 ret = -EFAULT;
3647
3648 kfree(sa);
3649 return ret;
3650}
3651
2ff7e61e
JM		 3652static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3653 void __user *arg)
3f6bcfbd
SB		 3654{
3655 struct btrfs_ioctl_dev_replace_args *p;
3656 int ret;
3657
3658 if (!capable(CAP_SYS_ADMIN))
3659 return -EPERM;
3660
3661 p = memdup_user(arg, sizeof(*p));
3662 if (IS_ERR(p))
3663 return PTR_ERR(p);
3664
3665 switch (p->cmd) {
3666 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
bc98a42c 3667 if (sb_rdonly(fs_info->sb)) {
adfa97cb
ID		 3668 ret = -EROFS;
3669 goto out;
3670 }
171938e5 3671 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
e57138b3 3672 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3f6bcfbd 3673 } else {
2ff7e61e 3674 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
171938e5 3675 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3f6bcfbd
SB		 3676 }
3677 break;
3678 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
0b246afa 3679 btrfs_dev_replace_status(fs_info, p);
3f6bcfbd
SB		 3680 ret = 0;
3681 break;
3682 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
17d202b9 3683 p->result = btrfs_dev_replace_cancel(fs_info);
97282031 3684 ret = 0;
3f6bcfbd
SB		 3685 break;
3686 default:
3687 ret = -EINVAL;
3688 break;
3689 }
3690
d3a53286 3691 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3f6bcfbd 3692 ret = -EFAULT;
adfa97cb 3693out:
3f6bcfbd
SB		 3694 kfree(p);
3695 return ret;
3696}
3697
d7728c96
JS		 3698static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3699{
3700 int ret = 0;
3701 int i;
740c3d22 3702 u64 rel_ptr;
d7728c96 3703 int size;
806468f8 3704 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3705 struct inode_fs_paths *ipath = NULL;
3706 struct btrfs_path *path;
3707
82b22ac8 3708 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 3709 return -EPERM;
3710
3711 path = btrfs_alloc_path();
3712 if (!path) {
3713 ret = -ENOMEM;
3714 goto out;
3715 }
3716
3717 ipa = memdup_user(arg, sizeof(*ipa));
3718 if (IS_ERR(ipa)) {
3719 ret = PTR_ERR(ipa);
3720 ipa = NULL;
3721 goto out;
3722 }
3723
3724 size = min_t(u32, ipa->size, 4096);
3725 ipath = init_ipath(size, root, path);
3726 if (IS_ERR(ipath)) {
3727 ret = PTR_ERR(ipath);
3728 ipath = NULL;
3729 goto out;
3730 }
3731
3732 ret = paths_from_inode(ipa->inum, ipath);
3733 if (ret < 0)
3734 goto out;
3735
3736 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3737 rel_ptr = ipath->fspath->val[i] -
3738 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3739 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3740 }
3741
718dc5fa
OS		 3742 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3743 ipath->fspath, size);
d7728c96
JS		 3744 if (ret) {
3745 ret = -EFAULT;
3746 goto out;
3747 }
3748
3749out:
3750 btrfs_free_path(path);
3751 free_ipath(ipath);
3752 kfree(ipa);
3753
3754 return ret;
3755}
3756
3757static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3758{
3759 struct btrfs_data_container *inodes = ctx;
3760 const size_t c = 3 * sizeof(u64);
3761
3762 if (inodes->bytes_left >= c) {
3763 inodes->bytes_left -= c;
3764 inodes->val[inodes->elem_cnt] = inum;
3765 inodes->val[inodes->elem_cnt + 1] = offset;
3766 inodes->val[inodes->elem_cnt + 2] = root;
3767 inodes->elem_cnt += 3;
3768 } else {
3769 inodes->bytes_missing += c - inodes->bytes_left;
3770 inodes->bytes_left = 0;
3771 inodes->elem_missed += 3;
3772 }
3773
3774 return 0;
3775}
3776
2ff7e61e 3777static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
d24a67b2 3778 void __user *arg, int version)
d7728c96
JS		 3779{
3780 int ret = 0;
3781 int size;
d7728c96
JS		 3782 struct btrfs_ioctl_logical_ino_args *loi;
3783 struct btrfs_data_container *inodes = NULL;
3784 struct btrfs_path *path = NULL;
d24a67b2 3785 bool ignore_offset;
d7728c96
JS		 3786
3787 if (!capable(CAP_SYS_ADMIN))
3788 return -EPERM;
3789
3790 loi = memdup_user(arg, sizeof(*loi));
7b9ea627
SV		 3791 if (IS_ERR(loi))
3792 return PTR_ERR(loi);
d7728c96 3793
d24a67b2
ZB		 3794 if (version == 1) {
3795 ignore_offset = false;
b115e3bc 3796 size = min_t(u32, loi->size, SZ_64K);
d24a67b2
ZB		 3797 } else {
3798 /* All reserved bits must be 0 for now */
3799 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3800 ret = -EINVAL;
3801 goto out_loi;
3802 }
3803 /* Only accept flags we have defined so far */
3804 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3805 ret = -EINVAL;
3806 goto out_loi;
3807 }
3808 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
b115e3bc 3809 size = min_t(u32, loi->size, SZ_16M);
d24a67b2
ZB		 3810 }
3811
d7728c96
JS		 3812 path = btrfs_alloc_path();
3813 if (!path) {
3814 ret = -ENOMEM;
3815 goto out;
3816 }
3817
d7728c96
JS		 3818 inodes = init_data_container(size);
3819 if (IS_ERR(inodes)) {
3820 ret = PTR_ERR(inodes);
3821 inodes = NULL;
3822 goto out;
3823 }
3824
2ff7e61e 3825 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
d24a67b2 3826 build_ino_list, inodes, ignore_offset);
df031f07 3827 if (ret == -EINVAL)
d7728c96
JS		 3828 ret = -ENOENT;
3829 if (ret < 0)
3830 goto out;
3831
718dc5fa
OS		 3832 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3833 size);
d7728c96
JS		 3834 if (ret)
3835 ret = -EFAULT;
3836
3837out:
3838 btrfs_free_path(path);
f54de068 3839 kvfree(inodes);
d24a67b2 3840out_loi:
d7728c96
JS		 3841 kfree(loi);
3842
3843 return ret;
3844}
3845
008ef096 3846void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
c9e9f97b
ID		 3847 struct btrfs_ioctl_balance_args *bargs)
3848{
3849 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3850
3851 bargs->flags = bctl->flags;
3852
3009a62f 3853 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
837d5b6e
ID		 3854 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3855 if (atomic_read(&fs_info->balance_pause_req))
3856 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3857 if (atomic_read(&fs_info->balance_cancel_req))
3858 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3859
c9e9f97b
ID		 3860 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3861 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3862 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce 3863
008ef096
DS		 3864 spin_lock(&fs_info->balance_lock);
3865 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3866 spin_unlock(&fs_info->balance_lock);
c9e9f97b
ID		 3867}
3868
9ba1f6e4 3869static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3870{
496ad9aa 3871 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 3872 struct btrfs_fs_info *fs_info = root->fs_info;
3873 struct btrfs_ioctl_balance_args *bargs;
3874 struct btrfs_balance_control *bctl;
ed0fb78f 3875 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 3876 int ret;
3877
3878 if (!capable(CAP_SYS_ADMIN))
3879 return -EPERM;
3880
e54bfa31 3881 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 3882 if (ret)
3883 return ret;
3884
ed0fb78f 3885again:
171938e5 3886 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
ed0fb78f
ID		 3887 mutex_lock(&fs_info->balance_mutex);
3888 need_unlock = true;
3889 goto locked;
3890 }
3891
3892 /*
01327610 3893 * mut. excl. ops lock is locked. Three possibilities:
ed0fb78f
ID		 3894 * (1) some other op is running
3895 * (2) balance is running
3896 * (3) balance is paused -- special case (think resume)
3897 */
c9e9f97b 3898 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 3899 if (fs_info->balance_ctl) {
3900 /* this is either (2) or (3) */
3009a62f 3901 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
ed0fb78f 3902 mutex_unlock(&fs_info->balance_mutex);
dccdb07b
DS		 3903 /*
3904 * Lock released to allow other waiters to continue,
3905 * we'll reexamine the status again.
3906 */
ed0fb78f
ID		 3907 mutex_lock(&fs_info->balance_mutex);
3908
3909 if (fs_info->balance_ctl &&
3009a62f 3910 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
ed0fb78f
ID		 3911 /* this is (3) */
3912 need_unlock = false;
3913 goto locked;
3914 }
3915
3916 mutex_unlock(&fs_info->balance_mutex);
ed0fb78f
ID		 3917 goto again;
3918 } else {
3919 /* this is (2) */
3920 mutex_unlock(&fs_info->balance_mutex);
3921 ret = -EINPROGRESS;
3922 goto out;
3923 }
3924 } else {
3925 /* this is (1) */
3926 mutex_unlock(&fs_info->balance_mutex);
e57138b3 3927 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
ed0fb78f
ID		 3928 goto out;
3929 }
3930
3931locked:
171938e5 3932 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
c9e9f97b
ID		 3933
3934 if (arg) {
3935 bargs = memdup_user(arg, sizeof(*bargs));
3936 if (IS_ERR(bargs)) {
3937 ret = PTR_ERR(bargs);
ed0fb78f 3938 goto out_unlock;
c9e9f97b 3939 }
de322263
ID		 3940
3941 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3942 if (!fs_info->balance_ctl) {
3943 ret = -ENOTCONN;
3944 goto out_bargs;
3945 }
3946
3947 bctl = fs_info->balance_ctl;
3948 spin_lock(&fs_info->balance_lock);
3949 bctl->flags |= BTRFS_BALANCE_RESUME;
3950 spin_unlock(&fs_info->balance_lock);
3951
3952 goto do_balance;
3953 }
c9e9f97b
ID		 3954 } else {
3955 bargs = NULL;
3956 }
3957
ed0fb78f 3958 if (fs_info->balance_ctl) {
837d5b6e
ID		 3959 ret = -EINPROGRESS;
3960 goto out_bargs;
3961 }
3962
8d2db785 3963 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
c9e9f97b
ID		 3964 if (!bctl) {
3965 ret = -ENOMEM;
3966 goto out_bargs;
3967 }
3968
c9e9f97b
ID		 3969 if (arg) {
3970 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3971 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3972 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3973
3974 bctl->flags = bargs->flags;
f43ffb60
ID		 3975 } else {
3976 /* balance everything - no filters */
3977 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3978 }
3979
8eb93459
DS		 3980 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3981 ret = -EINVAL;
0f89abf5 3982 goto out_bctl;
8eb93459
DS		 3983 }
3984
de322263 3985do_balance:
c9e9f97b 3986 /*
149196a2
DS		 3987 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
3988 * btrfs_balance. bctl is freed in reset_balance_state, or, if
3989 * restriper was paused all the way until unmount, in free_fs_info.
3990 * The flag should be cleared after reset_balance_state.
c9e9f97b 3991 */
ed0fb78f
ID		 3992 need_unlock = false;
3993
6fcf6e2b 3994 ret = btrfs_balance(fs_info, bctl, bargs);
0f89abf5 3995 bctl = NULL;
ed0fb78f 3996
d00c2d9c 3997 if ((ret == 0 || ret == -ECANCELED) && arg) {
c9e9f97b
ID		 3998 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3999 ret = -EFAULT;
4000 }
4001
0f89abf5
CE		 4002out_bctl:
4003 kfree(bctl);
c9e9f97b
ID		 4004out_bargs:
4005 kfree(bargs);
ed0fb78f 4006out_unlock:
c9e9f97b 4007 mutex_unlock(&fs_info->balance_mutex);
ed0fb78f 4008 if (need_unlock)
171938e5 4009 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
ed0fb78f 4010out:
e54bfa31 4011 mnt_drop_write_file(file);
c9e9f97b
ID		 4012 return ret;
4013}
4014
2ff7e61e 4015static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
837d5b6e
ID		 4016{
4017 if (!capable(CAP_SYS_ADMIN))
4018 return -EPERM;
4019
4020 switch (cmd) {
4021 case BTRFS_BALANCE_CTL_PAUSE:
0b246afa 4022 return btrfs_pause_balance(fs_info);
a7e99c69 4023 case BTRFS_BALANCE_CTL_CANCEL:
0b246afa 4024 return btrfs_cancel_balance(fs_info);
837d5b6e
ID		 4025 }
4026
4027 return -EINVAL;
4028}
4029
2ff7e61e 4030static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
19a39dce
ID		 4031 void __user *arg)
4032{
19a39dce
ID		 4033 struct btrfs_ioctl_balance_args *bargs;
4034 int ret = 0;
4035
4036 if (!capable(CAP_SYS_ADMIN))
4037 return -EPERM;
4038
4039 mutex_lock(&fs_info->balance_mutex);
4040 if (!fs_info->balance_ctl) {
4041 ret = -ENOTCONN;
4042 goto out;
4043 }
4044
8d2db785 4045 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
19a39dce
ID		 4046 if (!bargs) {
4047 ret = -ENOMEM;
4048 goto out;
4049 }
4050
008ef096 4051 btrfs_update_ioctl_balance_args(fs_info, bargs);
19a39dce
ID		 4052
4053 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4054 ret = -EFAULT;
4055
4056 kfree(bargs);
4057out:
4058 mutex_unlock(&fs_info->balance_mutex);
4059 return ret;
4060}
4061
905b0dda 4062static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4063{
0b246afa
JM		 4064 struct inode *inode = file_inode(file);
4065 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5d13a37b 4066 struct btrfs_ioctl_quota_ctl_args *sa;
5d13a37b 4067 int ret;
5d13a37b
AJ		 4068
4069 if (!capable(CAP_SYS_ADMIN))
4070 return -EPERM;
4071
905b0dda
MX		 4072 ret = mnt_want_write_file(file);
4073 if (ret)
4074 return ret;
5d13a37b
AJ		 4075
4076 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4077 if (IS_ERR(sa)) {
4078 ret = PTR_ERR(sa);
4079 goto drop_write;
4080 }
5d13a37b 4081
0b246afa 4082 down_write(&fs_info->subvol_sem);
5d13a37b
AJ		 4083
4084 switch (sa->cmd) {
4085 case BTRFS_QUOTA_CTL_ENABLE:
340f1aa2 4086 ret = btrfs_quota_enable(fs_info);
5d13a37b
AJ		 4087 break;
4088 case BTRFS_QUOTA_CTL_DISABLE:
340f1aa2 4089 ret = btrfs_quota_disable(fs_info);
5d13a37b 4090 break;
5d13a37b
AJ		 4091 default:
4092 ret = -EINVAL;
4093 break;
4094 }
4095
5d13a37b 4096 kfree(sa);
0b246afa 4097 up_write(&fs_info->subvol_sem);
905b0dda
MX		 4098drop_write:
4099 mnt_drop_write_file(file);
5d13a37b
AJ		 4100 return ret;
4101}
4102
905b0dda 4103static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4104{
0b246afa
JM		 4105 struct inode *inode = file_inode(file);
4106 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4107 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4108 struct btrfs_ioctl_qgroup_assign_args *sa;
4109 struct btrfs_trans_handle *trans;
4110 int ret;
4111 int err;
4112
4113 if (!capable(CAP_SYS_ADMIN))
4114 return -EPERM;
4115
905b0dda
MX		 4116 ret = mnt_want_write_file(file);
4117 if (ret)
4118 return ret;
5d13a37b
AJ		 4119
4120 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4121 if (IS_ERR(sa)) {
4122 ret = PTR_ERR(sa);
4123 goto drop_write;
4124 }
5d13a37b
AJ		 4125
4126 trans = btrfs_join_transaction(root);
4127 if (IS_ERR(trans)) {
4128 ret = PTR_ERR(trans);
4129 goto out;
4130 }
4131
5d13a37b 4132 if (sa->assign) {
9f8a6ce6 4133 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
5d13a37b 4134 } else {
39616c27 4135 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
5d13a37b
AJ		 4136 }
4137
e082f563 4138 /* update qgroup status and info */
280f8bd2 4139 err = btrfs_run_qgroups(trans);
e082f563 4140 if (err < 0)
0b246afa
JM		 4141 btrfs_handle_fs_error(fs_info, err,
4142 "failed to update qgroup status and info");
3a45bb20 4143 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 4144 if (err && !ret)
4145 ret = err;
4146
4147out:
4148 kfree(sa);
905b0dda
MX		 4149drop_write:
4150 mnt_drop_write_file(file);
5d13a37b
AJ		 4151 return ret;
4152}
4153
905b0dda 4154static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4155{
0b246afa 4156 struct inode *inode = file_inode(file);
0b246afa 4157 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4158 struct btrfs_ioctl_qgroup_create_args *sa;
4159 struct btrfs_trans_handle *trans;
4160 int ret;
4161 int err;
4162
4163 if (!capable(CAP_SYS_ADMIN))
4164 return -EPERM;
4165
905b0dda
MX		 4166 ret = mnt_want_write_file(file);
4167 if (ret)
4168 return ret;
5d13a37b
AJ		 4169
4170 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4171 if (IS_ERR(sa)) {
4172 ret = PTR_ERR(sa);
4173 goto drop_write;
4174 }
5d13a37b 4175
d86e56cf
MX		 4176 if (!sa->qgroupid) {
4177 ret = -EINVAL;
4178 goto out;
4179 }
4180
5d13a37b
AJ		 4181 trans = btrfs_join_transaction(root);
4182 if (IS_ERR(trans)) {
4183 ret = PTR_ERR(trans);
4184 goto out;
4185 }
4186
5d13a37b 4187 if (sa->create) {
49a05ecd 4188 ret = btrfs_create_qgroup(trans, sa->qgroupid);
5d13a37b 4189 } else {
3efbee1d 4190 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
5d13a37b
AJ		 4191 }
4192
3a45bb20 4193 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 4194 if (err && !ret)
4195 ret = err;
4196
4197out:
4198 kfree(sa);
905b0dda
MX		 4199drop_write:
4200 mnt_drop_write_file(file);
5d13a37b
AJ		 4201 return ret;
4202}
4203
905b0dda 4204static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4205{
0b246afa 4206 struct inode *inode = file_inode(file);
0b246afa 4207 struct btrfs_root *root = BTRFS_I(inode)->root;
5d13a37b
AJ		 4208 struct btrfs_ioctl_qgroup_limit_args *sa;
4209 struct btrfs_trans_handle *trans;
4210 int ret;
4211 int err;
4212 u64 qgroupid;
4213
4214 if (!capable(CAP_SYS_ADMIN))
4215 return -EPERM;
4216
905b0dda
MX		 4217 ret = mnt_want_write_file(file);
4218 if (ret)
4219 return ret;
5d13a37b
AJ		 4220
4221 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4222 if (IS_ERR(sa)) {
4223 ret = PTR_ERR(sa);
4224 goto drop_write;
4225 }
5d13a37b
AJ		 4226
4227 trans = btrfs_join_transaction(root);
4228 if (IS_ERR(trans)) {
4229 ret = PTR_ERR(trans);
4230 goto out;
4231 }
4232
4233 qgroupid = sa->qgroupid;
4234 if (!qgroupid) {
4235 /* take the current subvol as qgroup */
4236 qgroupid = root->root_key.objectid;
4237 }
4238
f0042d5e 4239 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
5d13a37b 4240
3a45bb20 4241 err = btrfs_end_transaction(trans);
5d13a37b
AJ		 4242 if (err && !ret)
4243 ret = err;
4244
4245out:
4246 kfree(sa);
905b0dda
MX		 4247drop_write:
4248 mnt_drop_write_file(file);
5d13a37b
AJ		 4249 return ret;
4250}
4251
2f232036
JS		 4252static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4253{
0b246afa
JM		 4254 struct inode *inode = file_inode(file);
4255 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2f232036
JS		 4256 struct btrfs_ioctl_quota_rescan_args *qsa;
4257 int ret;
4258
4259 if (!capable(CAP_SYS_ADMIN))
4260 return -EPERM;
4261
4262 ret = mnt_want_write_file(file);
4263 if (ret)
4264 return ret;
4265
4266 qsa = memdup_user(arg, sizeof(*qsa));
4267 if (IS_ERR(qsa)) {
4268 ret = PTR_ERR(qsa);
4269 goto drop_write;
4270 }
4271
4272 if (qsa->flags) {
4273 ret = -EINVAL;
4274 goto out;
4275 }
4276
0b246afa 4277 ret = btrfs_qgroup_rescan(fs_info);
2f232036
JS		 4278
4279out:
4280 kfree(qsa);
4281drop_write:
4282 mnt_drop_write_file(file);
4283 return ret;
4284}
4285
b929c1d8
MPS		 4286static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4287 void __user *arg)
2f232036 4288{
2f232036
JS		 4289 struct btrfs_ioctl_quota_rescan_args *qsa;
4290 int ret = 0;
4291
4292 if (!capable(CAP_SYS_ADMIN))
4293 return -EPERM;
4294
8d2db785 4295 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
2f232036
JS		 4296 if (!qsa)
4297 return -ENOMEM;
4298
0b246afa 4299 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
2f232036 4300 qsa->flags = 1;
0b246afa 4301 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
2f232036
JS		 4302 }
4303
4304 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4305 ret = -EFAULT;
4306
4307 kfree(qsa);
4308 return ret;
4309}
4310
b929c1d8
MPS		 4311static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4312 void __user *arg)
57254b6e 4313{
57254b6e
JS		 4314 if (!capable(CAP_SYS_ADMIN))
4315 return -EPERM;
4316
0b246afa 4317 return btrfs_qgroup_wait_for_completion(fs_info, true);
57254b6e
JS		 4318}
4319
abccd00f
HM		 4320static long _btrfs_ioctl_set_received_subvol(struct file *file,
4321 struct btrfs_ioctl_received_subvol_args *sa)
8ea05e3a 4322{
496ad9aa 4323 struct inode *inode = file_inode(file);
0b246afa 4324 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
8ea05e3a
AB		 4325 struct btrfs_root *root = BTRFS_I(inode)->root;
4326 struct btrfs_root_item *root_item = &root->root_item;
4327 struct btrfs_trans_handle *trans;
95582b00 4328 struct timespec64 ct = current_time(inode);
8ea05e3a 4329 int ret = 0;
dd5f9615 4330 int received_uuid_changed;
8ea05e3a 4331
bd60ea0f
DS		 4332 if (!inode_owner_or_capable(inode))
4333 return -EPERM;
4334
8ea05e3a
AB		 4335 ret = mnt_want_write_file(file);
4336 if (ret < 0)
4337 return ret;
4338
0b246afa 4339 down_write(&fs_info->subvol_sem);
8ea05e3a 4340
4a0cc7ca 4341 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
8ea05e3a
AB		 4342 ret = -EINVAL;
4343 goto out;
4344 }
4345
4346 if (btrfs_root_readonly(root)) {
4347 ret = -EROFS;
4348 goto out;
4349 }
4350
dd5f9615
SB		 4351 /*
4352 * 1 - root item
4353 * 2 - uuid items (received uuid + subvol uuid)
4354 */
4355 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4356 if (IS_ERR(trans)) {
4357 ret = PTR_ERR(trans);
4358 trans = NULL;
4359 goto out;
4360 }
4361
4362 sa->rtransid = trans->transid;
4363 sa->rtime.sec = ct.tv_sec;
4364 sa->rtime.nsec = ct.tv_nsec;
4365
dd5f9615
SB		 4366 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4367 BTRFS_UUID_SIZE);
4368 if (received_uuid_changed &&
d87ff758 4369 !btrfs_is_empty_uuid(root_item->received_uuid)) {
d1957791 4370 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
d87ff758
NB		 4371 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4372 root->root_key.objectid);
4373 if (ret && ret != -ENOENT) {
4374 btrfs_abort_transaction(trans, ret);
4375 btrfs_end_transaction(trans);
4376 goto out;
4377 }
4378 }
8ea05e3a
AB		 4379 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4380 btrfs_set_root_stransid(root_item, sa->stransid);
4381 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4382 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4383 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4384 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4385 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a 4386
0b246afa 4387 ret = btrfs_update_root(trans, fs_info->tree_root,
8ea05e3a
AB		 4388 &root->root_key, &root->root_item);
4389 if (ret < 0) {
3a45bb20 4390 btrfs_end_transaction(trans);
8ea05e3a 4391 goto out;
dd5f9615
SB		 4392 }
4393 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
cdb345a8 4394 ret = btrfs_uuid_tree_add(trans, sa->uuid,
dd5f9615
SB		 4395 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4396 root->root_key.objectid);
4397 if (ret < 0 && ret != -EEXIST) {
66642832 4398 btrfs_abort_transaction(trans, ret);
efd38150 4399 btrfs_end_transaction(trans);
8ea05e3a 4400 goto out;
dd5f9615
SB		 4401 }
4402 }
3a45bb20 4403 ret = btrfs_commit_transaction(trans);
abccd00f 4404out:
0b246afa 4405 up_write(&fs_info->subvol_sem);
abccd00f
HM		 4406 mnt_drop_write_file(file);
4407 return ret;
4408}
4409
4410#ifdef CONFIG_64BIT
4411static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4412 void __user *arg)
4413{
4414 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4415 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4416 int ret = 0;
4417
4418 args32 = memdup_user(arg, sizeof(*args32));
7b9ea627
SV		 4419 if (IS_ERR(args32))
4420 return PTR_ERR(args32);
abccd00f 4421
8d2db785 4422 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
84dbeb87
DC		 4423 if (!args64) {
4424 ret = -ENOMEM;
abccd00f
HM		 4425 goto out;
4426 }
4427
4428 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4429 args64->stransid = args32->stransid;
4430 args64->rtransid = args32->rtransid;
4431 args64->stime.sec = args32->stime.sec;
4432 args64->stime.nsec = args32->stime.nsec;
4433 args64->rtime.sec = args32->rtime.sec;
4434 args64->rtime.nsec = args32->rtime.nsec;
4435 args64->flags = args32->flags;
4436
4437 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4438 if (ret)
4439 goto out;
4440
4441 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4442 args32->stransid = args64->stransid;
4443 args32->rtransid = args64->rtransid;
4444 args32->stime.sec = args64->stime.sec;
4445 args32->stime.nsec = args64->stime.nsec;
4446 args32->rtime.sec = args64->rtime.sec;
4447 args32->rtime.nsec = args64->rtime.nsec;
4448 args32->flags = args64->flags;
4449
4450 ret = copy_to_user(arg, args32, sizeof(*args32));
4451 if (ret)
4452 ret = -EFAULT;
4453
4454out:
4455 kfree(args32);
4456 kfree(args64);
4457 return ret;
4458}
4459#endif
4460
4461static long btrfs_ioctl_set_received_subvol(struct file *file,
4462 void __user *arg)
4463{
4464 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4465 int ret = 0;
4466
4467 sa = memdup_user(arg, sizeof(*sa));
7b9ea627
SV		 4468 if (IS_ERR(sa))
4469 return PTR_ERR(sa);
abccd00f
HM		 4470
4471 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4472
4473 if (ret)
4474 goto out;
4475
8ea05e3a
AB		 4476 ret = copy_to_user(arg, sa, sizeof(*sa));
4477 if (ret)
4478 ret = -EFAULT;
4479
4480out:
4481 kfree(sa);
8ea05e3a
AB		 4482 return ret;
4483}
4484
b929c1d8
MPS		 4485static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4486 void __user *arg)
867ab667 4487{
a1b83ac5 4488 size_t len;
867ab667 4489 int ret;
a1b83ac5
AJ		 4490 char label[BTRFS_LABEL_SIZE];
4491
0b246afa
JM		 4492 spin_lock(&fs_info->super_lock);
4493 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4494 spin_unlock(&fs_info->super_lock);
a1b83ac5
AJ		 4495
4496 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 4497
4498 if (len == BTRFS_LABEL_SIZE) {
0b246afa
JM		 4499 btrfs_warn(fs_info,
4500 "label is too long, return the first %zu bytes",
4501 --len);
867ab667 4502 }
4503
867ab667 4504 ret = copy_to_user(arg, label, len);
867ab667 4505
4506 return ret ? -EFAULT : 0;
4507}
4508
a8bfd4ab 4509static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4510{
0b246afa
JM		 4511 struct inode *inode = file_inode(file);
4512 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4513 struct btrfs_root *root = BTRFS_I(inode)->root;
4514 struct btrfs_super_block *super_block = fs_info->super_copy;
a8bfd4ab 4515 struct btrfs_trans_handle *trans;
4516 char label[BTRFS_LABEL_SIZE];
4517 int ret;
4518
4519 if (!capable(CAP_SYS_ADMIN))
4520 return -EPERM;
4521
4522 if (copy_from_user(label, arg, sizeof(label)))
4523 return -EFAULT;
4524
4525 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
0b246afa 4526 btrfs_err(fs_info,
5d163e0e
JM		 4527 "unable to set label with more than %d bytes",
4528 BTRFS_LABEL_SIZE - 1);
a8bfd4ab 4529 return -EINVAL;
4530 }
4531
4532 ret = mnt_want_write_file(file);
4533 if (ret)
4534 return ret;
4535
a8bfd4ab 4536 trans = btrfs_start_transaction(root, 0);
4537 if (IS_ERR(trans)) {
4538 ret = PTR_ERR(trans);
4539 goto out_unlock;
4540 }
4541
0b246afa 4542 spin_lock(&fs_info->super_lock);
a8bfd4ab 4543 strcpy(super_block->label, label);
0b246afa 4544 spin_unlock(&fs_info->super_lock);
3a45bb20 4545 ret = btrfs_commit_transaction(trans);
a8bfd4ab 4546
4547out_unlock:
a8bfd4ab 4548 mnt_drop_write_file(file);
4549 return ret;
4550}
4551
2eaa055f
JM		 4552#define INIT_FEATURE_FLAGS(suffix) \
4553 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4554 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4555 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4556
d5131b65 4557int btrfs_ioctl_get_supported_features(void __user *arg)
2eaa055f 4558{
4d4ab6d6 4559 static const struct btrfs_ioctl_feature_flags features[3] = {
2eaa055f
JM		 4560 INIT_FEATURE_FLAGS(SUPP),
4561 INIT_FEATURE_FLAGS(SAFE_SET),
4562 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4563 };
4564
4565 if (copy_to_user(arg, &features, sizeof(features)))
4566 return -EFAULT;
4567
4568 return 0;
4569}
4570
b929c1d8
MPS		 4571static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4572 void __user *arg)
2eaa055f 4573{
0b246afa 4574 struct btrfs_super_block *super_block = fs_info->super_copy;
2eaa055f
JM		 4575 struct btrfs_ioctl_feature_flags features;
4576
4577 features.compat_flags = btrfs_super_compat_flags(super_block);
4578 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4579 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4580
4581 if (copy_to_user(arg, &features, sizeof(features)))
4582 return -EFAULT;
4583
4584 return 0;
4585}
4586
2ff7e61e 4587static int check_feature_bits(struct btrfs_fs_info *fs_info,
3b02a68a 4588 enum btrfs_feature_set set,
2eaa055f
JM		 4589 u64 change_mask, u64 flags, u64 supported_flags,
4590 u64 safe_set, u64 safe_clear)
4591{
f10152bc 4592 const char *type = btrfs_feature_set_name(set);
3b02a68a 4593 char *names;
2eaa055f
JM		 4594 u64 disallowed, unsupported;
4595 u64 set_mask = flags & change_mask;
4596 u64 clear_mask = ~flags & change_mask;
4597
4598 unsupported = set_mask & ~supported_flags;
4599 if (unsupported) {
3b02a68a
JM		 4600 names = btrfs_printable_features(set, unsupported);
4601 if (names) {
0b246afa
JM		 4602 btrfs_warn(fs_info,
4603 "this kernel does not support the %s feature bit%s",
4604 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 4605 kfree(names);
4606 } else
0b246afa
JM		 4607 btrfs_warn(fs_info,
4608 "this kernel does not support %s bits 0x%llx",
4609 type, unsupported);
2eaa055f
JM		 4610 return -EOPNOTSUPP;
4611 }
4612
4613 disallowed = set_mask & ~safe_set;
4614 if (disallowed) {
3b02a68a
JM		 4615 names = btrfs_printable_features(set, disallowed);
4616 if (names) {
0b246afa
JM		 4617 btrfs_warn(fs_info,
4618 "can't set the %s feature bit%s while mounted",
4619 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 4620 kfree(names);
4621 } else
0b246afa
JM		 4622 btrfs_warn(fs_info,
4623 "can't set %s bits 0x%llx while mounted",
4624 type, disallowed);
2eaa055f
JM		 4625 return -EPERM;
4626 }
4627
4628 disallowed = clear_mask & ~safe_clear;
4629 if (disallowed) {
3b02a68a
JM		 4630 names = btrfs_printable_features(set, disallowed);
4631 if (names) {
0b246afa
JM		 4632 btrfs_warn(fs_info,
4633 "can't clear the %s feature bit%s while mounted",
4634 names, strchr(names, ',') ? "s" : "");
3b02a68a
JM		 4635 kfree(names);
4636 } else
0b246afa
JM		 4637 btrfs_warn(fs_info,
4638 "can't clear %s bits 0x%llx while mounted",
4639 type, disallowed);
2eaa055f
JM		 4640 return -EPERM;
4641 }
4642
4643 return 0;
4644}
4645
2ff7e61e
JM		 4646#define check_feature(fs_info, change_mask, flags, mask_base) \
4647check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
2eaa055f
JM		 4648 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4649 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4650 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4651
4652static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4653{
0b246afa
JM		 4654 struct inode *inode = file_inode(file);
4655 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4656 struct btrfs_root *root = BTRFS_I(inode)->root;
4657 struct btrfs_super_block *super_block = fs_info->super_copy;
2eaa055f
JM		 4658 struct btrfs_ioctl_feature_flags flags[2];
4659 struct btrfs_trans_handle *trans;
4660 u64 newflags;
4661 int ret;
4662
4663 if (!capable(CAP_SYS_ADMIN))
4664 return -EPERM;
4665
4666 if (copy_from_user(flags, arg, sizeof(flags)))
4667 return -EFAULT;
4668
4669 /* Nothing to do */
4670 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4671 !flags[0].incompat_flags)
4672 return 0;
4673
2ff7e61e 4674 ret = check_feature(fs_info, flags[0].compat_flags,
2eaa055f
JM		 4675 flags[1].compat_flags, COMPAT);
4676 if (ret)
4677 return ret;
4678
2ff7e61e 4679 ret = check_feature(fs_info, flags[0].compat_ro_flags,
2eaa055f
JM		 4680 flags[1].compat_ro_flags, COMPAT_RO);
4681 if (ret)
4682 return ret;
4683
2ff7e61e 4684 ret = check_feature(fs_info, flags[0].incompat_flags,
2eaa055f
JM		 4685 flags[1].incompat_flags, INCOMPAT);
4686 if (ret)
4687 return ret;
4688
7ab19625
DS		 4689 ret = mnt_want_write_file(file);
4690 if (ret)
4691 return ret;
4692
8051aa1a 4693 trans = btrfs_start_transaction(root, 0);
7ab19625
DS		 4694 if (IS_ERR(trans)) {
4695 ret = PTR_ERR(trans);
4696 goto out_drop_write;
4697 }
2eaa055f 4698
0b246afa 4699 spin_lock(&fs_info->super_lock);
2eaa055f
JM		 4700 newflags = btrfs_super_compat_flags(super_block);
4701 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4702 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4703 btrfs_set_super_compat_flags(super_block, newflags);
4704
4705 newflags = btrfs_super_compat_ro_flags(super_block);
4706 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4707 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4708 btrfs_set_super_compat_ro_flags(super_block, newflags);
4709
4710 newflags = btrfs_super_incompat_flags(super_block);
4711 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4712 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4713 btrfs_set_super_incompat_flags(super_block, newflags);
0b246afa 4714 spin_unlock(&fs_info->super_lock);
2eaa055f 4715
3a45bb20 4716 ret = btrfs_commit_transaction(trans);
7ab19625
DS		 4717out_drop_write:
4718 mnt_drop_write_file(file);
4719
4720 return ret;
2eaa055f
JM		 4721}
4722
2351f431
JB		 4723static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4724{
4725 struct btrfs_ioctl_send_args *arg;
4726 int ret;
4727
4728 if (compat) {
4729#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4730 struct btrfs_ioctl_send_args_32 args32;
4731
4732 ret = copy_from_user(&args32, argp, sizeof(args32));
4733 if (ret)
4734 return -EFAULT;
4735 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4736 if (!arg)
4737 return -ENOMEM;
4738 arg->send_fd = args32.send_fd;
4739 arg->clone_sources_count = args32.clone_sources_count;
4740 arg->clone_sources = compat_ptr(args32.clone_sources);
4741 arg->parent_root = args32.parent_root;
4742 arg->flags = args32.flags;
4743 memcpy(arg->reserved, args32.reserved,
4744 sizeof(args32.reserved));
4745#else
4746 return -ENOTTY;
4747#endif
4748 } else {
4749 arg = memdup_user(argp, sizeof(*arg));
4750 if (IS_ERR(arg))
4751 return PTR_ERR(arg);
4752 }
4753 ret = btrfs_ioctl_send(file, arg);
4754 kfree(arg);
4755 return ret;
4756}
4757
f46b5a66
CH		 4758long btrfs_ioctl(struct file *file, unsigned int
4759 cmd, unsigned long arg)
4760{
0b246afa
JM		 4761 struct inode *inode = file_inode(file);
4762 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4763 struct btrfs_root *root = BTRFS_I(inode)->root;
4bcabaa3 4764 void __user *argp = (void __user *)arg;
f46b5a66
CH		 4765
4766 switch (cmd) {
6cbff00f
CH		 4767 case FS_IOC_GETFLAGS:
4768 return btrfs_ioctl_getflags(file, argp);
4769 case FS_IOC_SETFLAGS:
4770 return btrfs_ioctl_setflags(file, argp);
4771 case FS_IOC_GETVERSION:
4772 return btrfs_ioctl_getversion(file, argp);
40cf931f 4773 case FS_IOC_GETFSLABEL:
b929c1d8 4774 return btrfs_ioctl_get_fslabel(fs_info, argp);
40cf931f
ES		 4775 case FS_IOC_SETFSLABEL:
4776 return btrfs_ioctl_set_fslabel(file, argp);
f7039b1d 4777 case FITRIM:
b929c1d8 4778 return btrfs_ioctl_fitrim(fs_info, argp);
f46b5a66 4779 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 4780 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 4781 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 4782 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 4783 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 4784 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 4785 case BTRFS_IOC_SUBVOL_CREATE_V2:
4786 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c 4787 case BTRFS_IOC_SNAP_DESTROY:
949964c9
MPS		 4788 return btrfs_ioctl_snap_destroy(file, argp, false);
4789 case BTRFS_IOC_SNAP_DESTROY_V2:
4790 return btrfs_ioctl_snap_destroy(file, argp, true);
0caa102d
LZ		 4791 case BTRFS_IOC_SUBVOL_GETFLAGS:
4792 return btrfs_ioctl_subvol_getflags(file, argp);
4793 case BTRFS_IOC_SUBVOL_SETFLAGS:
4794 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 4795 case BTRFS_IOC_DEFAULT_SUBVOL:
4796 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 4797 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 4798 return btrfs_ioctl_defrag(file, NULL);
4799 case BTRFS_IOC_DEFRAG_RANGE:
4800 return btrfs_ioctl_defrag(file, argp);
f46b5a66 4801 case BTRFS_IOC_RESIZE:
198605a8 4802 return btrfs_ioctl_resize(file, argp);
f46b5a66 4803 case BTRFS_IOC_ADD_DEV:
2ff7e61e 4804 return btrfs_ioctl_add_dev(fs_info, argp);
f46b5a66 4805 case BTRFS_IOC_RM_DEV:
da24927b 4806 return btrfs_ioctl_rm_dev(file, argp);
6b526ed7
AJ		 4807 case BTRFS_IOC_RM_DEV_V2:
4808 return btrfs_ioctl_rm_dev_v2(file, argp);
475f6387 4809 case BTRFS_IOC_FS_INFO:
2ff7e61e 4810 return btrfs_ioctl_fs_info(fs_info, argp);
475f6387 4811 case BTRFS_IOC_DEV_INFO:
2ff7e61e 4812 return btrfs_ioctl_dev_info(fs_info, argp);
f46b5a66 4813 case BTRFS_IOC_BALANCE:
9ba1f6e4 4814 return btrfs_ioctl_balance(file, NULL);
ac8e9819
CM		 4815 case BTRFS_IOC_TREE_SEARCH:
4816 return btrfs_ioctl_tree_search(file, argp);
cc68a8a5
GH		 4817 case BTRFS_IOC_TREE_SEARCH_V2:
4818 return btrfs_ioctl_tree_search_v2(file, argp);
ac8e9819
CM		 4819 case BTRFS_IOC_INO_LOOKUP:
4820 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 4821 case BTRFS_IOC_INO_PATHS:
4822 return btrfs_ioctl_ino_to_path(root, argp);
4823 case BTRFS_IOC_LOGICAL_INO:
d24a67b2
ZB		 4824 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4825 case BTRFS_IOC_LOGICAL_INO_V2:
4826 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
1406e432 4827 case BTRFS_IOC_SPACE_INFO:
2ff7e61e 4828 return btrfs_ioctl_space_info(fs_info, argp);
9b199859
FDBM		 4829 case BTRFS_IOC_SYNC: {
4830 int ret;
4831
82b3e53b 4832 ret = btrfs_start_delalloc_roots(fs_info, -1);
9b199859
FDBM		 4833 if (ret)
4834 return ret;
0b246afa 4835 ret = btrfs_sync_fs(inode->i_sb, 1);
2fad4e83
DS		 4836 /*
4837 * The transaction thread may want to do more work,
01327610 4838 * namely it pokes the cleaner kthread that will start
2fad4e83
DS		 4839 * processing uncleaned subvols.
4840 */
0b246afa 4841 wake_up_process(fs_info->transaction_kthread);
9b199859
FDBM		 4842 return ret;
4843 }
46204592 4844 case BTRFS_IOC_START_SYNC:
9a8c28be 4845 return btrfs_ioctl_start_sync(root, argp);
46204592 4846 case BTRFS_IOC_WAIT_SYNC:
2ff7e61e 4847 return btrfs_ioctl_wait_sync(fs_info, argp);
475f6387 4848 case BTRFS_IOC_SCRUB:
b8e95489 4849 return btrfs_ioctl_scrub(file, argp);
475f6387 4850 case BTRFS_IOC_SCRUB_CANCEL:
2ff7e61e 4851 return btrfs_ioctl_scrub_cancel(fs_info);
475f6387 4852 case BTRFS_IOC_SCRUB_PROGRESS:
2ff7e61e 4853 return btrfs_ioctl_scrub_progress(fs_info, argp);
c9e9f97b 4854 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 4855 return btrfs_ioctl_balance(file, argp);
837d5b6e 4856 case BTRFS_IOC_BALANCE_CTL:
2ff7e61e 4857 return btrfs_ioctl_balance_ctl(fs_info, arg);
19a39dce 4858 case BTRFS_IOC_BALANCE_PROGRESS:
2ff7e61e 4859 return btrfs_ioctl_balance_progress(fs_info, argp);
8ea05e3a
AB		 4860 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4861 return btrfs_ioctl_set_received_subvol(file, argp);
abccd00f
HM		 4862#ifdef CONFIG_64BIT
4863 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4864 return btrfs_ioctl_set_received_subvol_32(file, argp);
4865#endif
31db9f7c 4866 case BTRFS_IOC_SEND:
2351f431
JB		 4867 return _btrfs_ioctl_send(file, argp, false);
4868#if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4869 case BTRFS_IOC_SEND_32:
4870 return _btrfs_ioctl_send(file, argp, true);
4871#endif
c11d2c23 4872 case BTRFS_IOC_GET_DEV_STATS:
2ff7e61e 4873 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5d13a37b 4874 case BTRFS_IOC_QUOTA_CTL:
905b0dda 4875 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 4876 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 4877 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 4878 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 4879 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 4880 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 4881 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 4882 case BTRFS_IOC_QUOTA_RESCAN:
4883 return btrfs_ioctl_quota_rescan(file, argp);
4884 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
b929c1d8 4885 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
57254b6e 4886 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
b929c1d8 4887 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
3f6bcfbd 4888 case BTRFS_IOC_DEV_REPLACE:
2ff7e61e 4889 return btrfs_ioctl_dev_replace(fs_info, argp);
2eaa055f 4890 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
d5131b65 4891 return btrfs_ioctl_get_supported_features(argp);
2eaa055f 4892 case BTRFS_IOC_GET_FEATURES:
b929c1d8 4893 return btrfs_ioctl_get_features(fs_info, argp);
2eaa055f
JM		 4894 case BTRFS_IOC_SET_FEATURES:
4895 return btrfs_ioctl_set_features(file, argp);
e4202ac9
DS		 4896 case FS_IOC_FSGETXATTR:
4897 return btrfs_ioctl_fsgetxattr(file, argp);
025f2121
DS		 4898 case FS_IOC_FSSETXATTR:
4899 return btrfs_ioctl_fssetxattr(file, argp);
b64ec075
TM		 4900 case BTRFS_IOC_GET_SUBVOL_INFO:
4901 return btrfs_ioctl_get_subvol_info(file, argp);
42e4b520
TM		 4902 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4903 return btrfs_ioctl_get_subvol_rootref(file, argp);
23d0b79d
TM		 4904 case BTRFS_IOC_INO_LOOKUP_USER:
4905 return btrfs_ioctl_ino_lookup_user(file, argp);
f46b5a66
CH		 4906 }
4907
4908 return -ENOTTY;
4909}
4c63c245
LD		 4910
4911#ifdef CONFIG_COMPAT
4912long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4913{
2a362249
JM		 4914 /*
4915 * These all access 32-bit values anyway so no further
4916 * handling is necessary.
4917 */
4c63c245
LD		 4918 switch (cmd) {
4919 case FS_IOC32_GETFLAGS:
4920 cmd = FS_IOC_GETFLAGS;
4921 break;
4922 case FS_IOC32_SETFLAGS:
4923 cmd = FS_IOC_SETFLAGS;
4924 break;
4925 case FS_IOC32_GETVERSION:
4926 cmd = FS_IOC_GETVERSION;
4927 break;
4c63c245
LD		 4928 }
4929
4930 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4931}
4932#endif
Ubuntu Hirsute Linux kernel mirror