]>
Commit | Line | Data |
---|---|---|
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> |
97fc2977 | 29 | #include <linux/fileattr.h> |
14605409 | 30 | #include <linux/fsverity.h> |
f46b5a66 CH |
31 | #include "ctree.h" |
32 | #include "disk-io.h" | |
949964c9 | 33 | #include "export.h" |
f46b5a66 CH |
34 | #include "transaction.h" |
35 | #include "btrfs_inode.h" | |
f46b5a66 CH |
36 | #include "print-tree.h" |
37 | #include "volumes.h" | |
925baedd | 38 | #include "locking.h" |
d7728c96 | 39 | #include "backref.h" |
606686ee | 40 | #include "rcu-string.h" |
31db9f7c | 41 | #include "send.h" |
3f6bcfbd | 42 | #include "dev-replace.h" |
63541927 | 43 | #include "props.h" |
3b02a68a | 44 | #include "sysfs.h" |
fcebe456 | 45 | #include "qgroup.h" |
1ec9a1ae | 46 | #include "tree-log.h" |
ebb8765b | 47 | #include "compression.h" |
8719aaae | 48 | #include "space-info.h" |
86736342 | 49 | #include "delalloc-space.h" |
aac0023c | 50 | #include "block-group.h" |
22b398ee | 51 | #include "subpage.h" |
f46b5a66 | 52 | |
abccd00f HM |
53 | #ifdef CONFIG_64BIT |
54 | /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI | |
55 | * structures are incorrect, as the timespec structure from userspace | |
56 | * is 4 bytes too small. We define these alternatives here to teach | |
57 | * the kernel about the 32-bit struct packing. | |
58 | */ | |
59 | struct btrfs_ioctl_timespec_32 { | |
60 | __u64 sec; | |
61 | __u32 nsec; | |
62 | } __attribute__ ((__packed__)); | |
63 | ||
64 | struct btrfs_ioctl_received_subvol_args_32 { | |
65 | char uuid[BTRFS_UUID_SIZE]; /* in */ | |
66 | __u64 stransid; /* in */ | |
67 | __u64 rtransid; /* out */ | |
68 | struct btrfs_ioctl_timespec_32 stime; /* in */ | |
69 | struct btrfs_ioctl_timespec_32 rtime; /* out */ | |
70 | __u64 flags; /* in */ | |
71 | __u64 reserved[16]; /* in */ | |
72 | } __attribute__ ((__packed__)); | |
73 | ||
74 | #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ | |
75 | struct btrfs_ioctl_received_subvol_args_32) | |
76 | #endif | |
77 | ||
2351f431 JB |
78 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) |
79 | struct btrfs_ioctl_send_args_32 { | |
80 | __s64 send_fd; /* in */ | |
81 | __u64 clone_sources_count; /* in */ | |
82 | compat_uptr_t clone_sources; /* in */ | |
83 | __u64 parent_root; /* in */ | |
84 | __u64 flags; /* in */ | |
e77fbf99 DS |
85 | __u32 version; /* in */ |
86 | __u8 reserved[28]; /* in */ | |
2351f431 JB |
87 | } __attribute__ ((__packed__)); |
88 | ||
89 | #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ | |
90 | struct btrfs_ioctl_send_args_32) | |
91 | #endif | |
abccd00f | 92 | |
6cbff00f | 93 | /* Mask out flags that are inappropriate for the given type of inode. */ |
1905a0f7 DS |
94 | static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, |
95 | unsigned int flags) | |
6cbff00f | 96 | { |
1905a0f7 | 97 | if (S_ISDIR(inode->i_mode)) |
6cbff00f | 98 | return flags; |
1905a0f7 | 99 | else if (S_ISREG(inode->i_mode)) |
6cbff00f CH |
100 | return flags & ~FS_DIRSYNC_FL; |
101 | else | |
102 | return flags & (FS_NODUMP_FL | FS_NOATIME_FL); | |
103 | } | |
104 | ||
105 | /* | |
a157d4fd DS |
106 | * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS |
107 | * ioctl. | |
6cbff00f | 108 | */ |
77eea05e | 109 | static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode) |
6cbff00f CH |
110 | { |
111 | unsigned int iflags = 0; | |
77eea05e | 112 | u32 flags = binode->flags; |
14605409 | 113 | u32 ro_flags = binode->ro_flags; |
6cbff00f CH |
114 | |
115 | if (flags & BTRFS_INODE_SYNC) | |
116 | iflags |= FS_SYNC_FL; | |
117 | if (flags & BTRFS_INODE_IMMUTABLE) | |
118 | iflags |= FS_IMMUTABLE_FL; | |
119 | if (flags & BTRFS_INODE_APPEND) | |
120 | iflags |= FS_APPEND_FL; | |
121 | if (flags & BTRFS_INODE_NODUMP) | |
122 | iflags |= FS_NODUMP_FL; | |
123 | if (flags & BTRFS_INODE_NOATIME) | |
124 | iflags |= FS_NOATIME_FL; | |
125 | if (flags & BTRFS_INODE_DIRSYNC) | |
126 | iflags |= FS_DIRSYNC_FL; | |
d0092bdd LZ |
127 | if (flags & BTRFS_INODE_NODATACOW) |
128 | iflags |= FS_NOCOW_FL; | |
14605409 BB |
129 | if (ro_flags & BTRFS_INODE_RO_VERITY) |
130 | iflags |= FS_VERITY_FL; | |
d0092bdd | 131 | |
13f48dc9 | 132 | if (flags & BTRFS_INODE_NOCOMPRESS) |
d0092bdd | 133 | iflags |= FS_NOCOMP_FL; |
13f48dc9 ST |
134 | else if (flags & BTRFS_INODE_COMPRESS) |
135 | iflags |= FS_COMPR_FL; | |
6cbff00f CH |
136 | |
137 | return iflags; | |
138 | } | |
139 | ||
140 | /* | |
141 | * Update inode->i_flags based on the btrfs internal flags. | |
142 | */ | |
7b6a221e | 143 | void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) |
6cbff00f | 144 | { |
5c57b8b6 | 145 | struct btrfs_inode *binode = BTRFS_I(inode); |
3cc79392 | 146 | unsigned int new_fl = 0; |
6cbff00f | 147 | |
5c57b8b6 | 148 | if (binode->flags & BTRFS_INODE_SYNC) |
3cc79392 | 149 | new_fl |= S_SYNC; |
5c57b8b6 | 150 | if (binode->flags & BTRFS_INODE_IMMUTABLE) |
3cc79392 | 151 | new_fl |= S_IMMUTABLE; |
5c57b8b6 | 152 | if (binode->flags & BTRFS_INODE_APPEND) |
3cc79392 | 153 | new_fl |= S_APPEND; |
5c57b8b6 | 154 | if (binode->flags & BTRFS_INODE_NOATIME) |
3cc79392 | 155 | new_fl |= S_NOATIME; |
5c57b8b6 | 156 | if (binode->flags & BTRFS_INODE_DIRSYNC) |
3cc79392 | 157 | new_fl |= S_DIRSYNC; |
14605409 BB |
158 | if (binode->ro_flags & BTRFS_INODE_RO_VERITY) |
159 | new_fl |= S_VERITY; | |
3cc79392 FM |
160 | |
161 | set_mask_bits(&inode->i_flags, | |
14605409 BB |
162 | S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC | |
163 | S_VERITY, new_fl); | |
6cbff00f CH |
164 | } |
165 | ||
f37c563b DS |
166 | /* |
167 | * Check if @flags are a supported and valid set of FS_*_FL flags and that | |
168 | * the old and new flags are not conflicting | |
169 | */ | |
170 | static int check_fsflags(unsigned int old_flags, unsigned int flags) | |
75e7cb7f LB |
171 | { |
172 | if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ | |
173 | FS_NOATIME_FL | FS_NODUMP_FL | \ | |
174 | FS_SYNC_FL | FS_DIRSYNC_FL | \ | |
e1e8fb6a LZ |
175 | FS_NOCOMP_FL | FS_COMPR_FL | |
176 | FS_NOCOW_FL)) | |
75e7cb7f LB |
177 | return -EOPNOTSUPP; |
178 | ||
f37c563b | 179 | /* COMPR and NOCOMP on new/old are valid */ |
75e7cb7f LB |
180 | if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) |
181 | return -EINVAL; | |
182 | ||
f37c563b DS |
183 | if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL)) |
184 | return -EINVAL; | |
185 | ||
186 | /* NOCOW and compression options are mutually exclusive */ | |
187 | if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL))) | |
188 | return -EINVAL; | |
189 | if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL))) | |
190 | return -EINVAL; | |
191 | ||
75e7cb7f LB |
192 | return 0; |
193 | } | |
194 | ||
d206e9c9 NA |
195 | static int check_fsflags_compatible(struct btrfs_fs_info *fs_info, |
196 | unsigned int flags) | |
197 | { | |
198 | if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL)) | |
199 | return -EPERM; | |
200 | ||
201 | return 0; | |
202 | } | |
203 | ||
97fc2977 MS |
204 | /* |
205 | * Set flags/xflags from the internal inode flags. The remaining items of | |
206 | * fsxattr are zeroed. | |
207 | */ | |
208 | int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa) | |
6cbff00f | 209 | { |
97fc2977 MS |
210 | struct btrfs_inode *binode = BTRFS_I(d_inode(dentry)); |
211 | ||
77eea05e | 212 | fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode)); |
97fc2977 MS |
213 | return 0; |
214 | } | |
215 | ||
216 | int btrfs_fileattr_set(struct user_namespace *mnt_userns, | |
217 | struct dentry *dentry, struct fileattr *fa) | |
218 | { | |
219 | struct inode *inode = d_inode(dentry); | |
0b246afa | 220 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
5c57b8b6 DS |
221 | struct btrfs_inode *binode = BTRFS_I(inode); |
222 | struct btrfs_root *root = binode->root; | |
6cbff00f | 223 | struct btrfs_trans_handle *trans; |
5aca2842 | 224 | unsigned int fsflags, old_fsflags; |
6cbff00f | 225 | int ret; |
ff9fef55 | 226 | const char *comp = NULL; |
f37c563b | 227 | u32 binode_flags; |
6cbff00f | 228 | |
b83cc969 LZ |
229 | if (btrfs_root_readonly(root)) |
230 | return -EROFS; | |
231 | ||
97fc2977 MS |
232 | if (fileattr_has_fsx(fa)) |
233 | return -EOPNOTSUPP; | |
e7848683 | 234 | |
97fc2977 | 235 | fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags); |
77eea05e | 236 | old_fsflags = btrfs_inode_flags_to_fsflags(binode); |
f37c563b DS |
237 | ret = check_fsflags(old_fsflags, fsflags); |
238 | if (ret) | |
97fc2977 | 239 | return ret; |
f37c563b | 240 | |
d206e9c9 NA |
241 | ret = check_fsflags_compatible(fs_info, fsflags); |
242 | if (ret) | |
97fc2977 | 243 | return ret; |
d206e9c9 | 244 | |
f37c563b | 245 | binode_flags = binode->flags; |
5c57b8b6 | 246 | if (fsflags & FS_SYNC_FL) |
d2b8fcfe | 247 | binode_flags |= BTRFS_INODE_SYNC; |
6cbff00f | 248 | else |
d2b8fcfe | 249 | binode_flags &= ~BTRFS_INODE_SYNC; |
5c57b8b6 | 250 | if (fsflags & FS_IMMUTABLE_FL) |
d2b8fcfe | 251 | binode_flags |= BTRFS_INODE_IMMUTABLE; |
6cbff00f | 252 | else |
d2b8fcfe | 253 | binode_flags &= ~BTRFS_INODE_IMMUTABLE; |
5c57b8b6 | 254 | if (fsflags & FS_APPEND_FL) |
d2b8fcfe | 255 | binode_flags |= BTRFS_INODE_APPEND; |
6cbff00f | 256 | else |
d2b8fcfe | 257 | binode_flags &= ~BTRFS_INODE_APPEND; |
5c57b8b6 | 258 | if (fsflags & FS_NODUMP_FL) |
d2b8fcfe | 259 | binode_flags |= BTRFS_INODE_NODUMP; |
6cbff00f | 260 | else |
d2b8fcfe | 261 | binode_flags &= ~BTRFS_INODE_NODUMP; |
5c57b8b6 | 262 | if (fsflags & FS_NOATIME_FL) |
d2b8fcfe | 263 | binode_flags |= BTRFS_INODE_NOATIME; |
6cbff00f | 264 | else |
d2b8fcfe | 265 | binode_flags &= ~BTRFS_INODE_NOATIME; |
97fc2977 MS |
266 | |
267 | /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */ | |
268 | if (!fa->flags_valid) { | |
269 | /* 1 item for the inode */ | |
270 | trans = btrfs_start_transaction(root, 1); | |
9b8a233b RH |
271 | if (IS_ERR(trans)) |
272 | return PTR_ERR(trans); | |
97fc2977 MS |
273 | goto update_flags; |
274 | } | |
275 | ||
5c57b8b6 | 276 | if (fsflags & FS_DIRSYNC_FL) |
d2b8fcfe | 277 | binode_flags |= BTRFS_INODE_DIRSYNC; |
6cbff00f | 278 | else |
d2b8fcfe | 279 | binode_flags &= ~BTRFS_INODE_DIRSYNC; |
5c57b8b6 | 280 | if (fsflags & FS_NOCOW_FL) { |
44e5194b | 281 | if (S_ISREG(inode->i_mode)) { |
7e97b8da DS |
282 | /* |
283 | * It's safe to turn csums off here, no extents exist. | |
284 | * Otherwise we want the flag to reflect the real COW | |
285 | * status of the file and will not set it. | |
286 | */ | |
287 | if (inode->i_size == 0) | |
d2b8fcfe AJ |
288 | binode_flags |= BTRFS_INODE_NODATACOW | |
289 | BTRFS_INODE_NODATASUM; | |
7e97b8da | 290 | } else { |
d2b8fcfe | 291 | binode_flags |= BTRFS_INODE_NODATACOW; |
7e97b8da DS |
292 | } |
293 | } else { | |
294 | /* | |
01327610 | 295 | * Revert back under same assumptions as above |
7e97b8da | 296 | */ |
44e5194b | 297 | if (S_ISREG(inode->i_mode)) { |
7e97b8da | 298 | if (inode->i_size == 0) |
d2b8fcfe AJ |
299 | binode_flags &= ~(BTRFS_INODE_NODATACOW | |
300 | BTRFS_INODE_NODATASUM); | |
7e97b8da | 301 | } else { |
d2b8fcfe | 302 | binode_flags &= ~BTRFS_INODE_NODATACOW; |
7e97b8da DS |
303 | } |
304 | } | |
6cbff00f | 305 | |
75e7cb7f LB |
306 | /* |
307 | * The COMPRESS flag can only be changed by users, while the NOCOMPRESS | |
308 | * flag may be changed automatically if compression code won't make | |
309 | * things smaller. | |
310 | */ | |
5c57b8b6 | 311 | if (fsflags & FS_NOCOMP_FL) { |
d2b8fcfe AJ |
312 | binode_flags &= ~BTRFS_INODE_COMPRESS; |
313 | binode_flags |= BTRFS_INODE_NOCOMPRESS; | |
5c57b8b6 | 314 | } else if (fsflags & FS_COMPR_FL) { |
63541927 | 315 | |
97fc2977 MS |
316 | if (IS_SWAPFILE(inode)) |
317 | return -ETXTBSY; | |
eede2bf3 | 318 | |
d2b8fcfe AJ |
319 | binode_flags |= BTRFS_INODE_COMPRESS; |
320 | binode_flags &= ~BTRFS_INODE_NOCOMPRESS; | |
63541927 | 321 | |
93370509 DS |
322 | comp = btrfs_compress_type2str(fs_info->compress_type); |
323 | if (!comp || comp[0] == 0) | |
324 | comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); | |
ebcb904d | 325 | } else { |
d2b8fcfe | 326 | binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); |
75e7cb7f | 327 | } |
6cbff00f | 328 | |
ff9fef55 AJ |
329 | /* |
330 | * 1 for inode item | |
331 | * 2 for properties | |
332 | */ | |
333 | trans = btrfs_start_transaction(root, 3); | |
97fc2977 MS |
334 | if (IS_ERR(trans)) |
335 | return PTR_ERR(trans); | |
6cbff00f | 336 | |
ff9fef55 AJ |
337 | if (comp) { |
338 | ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, | |
339 | strlen(comp), 0); | |
340 | if (ret) { | |
341 | btrfs_abort_transaction(trans, ret); | |
342 | goto out_end_trans; | |
343 | } | |
ff9fef55 AJ |
344 | } else { |
345 | ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, | |
346 | 0, 0); | |
347 | if (ret && ret != -ENODATA) { | |
348 | btrfs_abort_transaction(trans, ret); | |
349 | goto out_end_trans; | |
350 | } | |
351 | } | |
352 | ||
97fc2977 | 353 | update_flags: |
d2b8fcfe | 354 | binode->flags = binode_flags; |
7b6a221e | 355 | btrfs_sync_inode_flags_to_i_flags(inode); |
0c4d2d95 | 356 | inode_inc_iversion(inode); |
c2050a45 | 357 | inode->i_ctime = current_time(inode); |
9a56fcd1 | 358 | ret = btrfs_update_inode(trans, root, BTRFS_I(inode)); |
6cbff00f | 359 | |
ff9fef55 | 360 | out_end_trans: |
3a45bb20 | 361 | btrfs_end_transaction(trans); |
2d4e6f6a | 362 | return ret; |
6cbff00f CH |
363 | } |
364 | ||
0d7ed32c DS |
365 | /* |
366 | * Start exclusive operation @type, return true on success | |
367 | */ | |
c3e1f96c GR |
368 | bool btrfs_exclop_start(struct btrfs_fs_info *fs_info, |
369 | enum btrfs_exclusive_operation type) | |
370 | { | |
0d7ed32c DS |
371 | bool ret = false; |
372 | ||
373 | spin_lock(&fs_info->super_lock); | |
374 | if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) { | |
375 | fs_info->exclusive_operation = type; | |
376 | ret = true; | |
377 | } | |
378 | spin_unlock(&fs_info->super_lock); | |
379 | ||
380 | return ret; | |
c3e1f96c GR |
381 | } |
382 | ||
578bda9e DS |
383 | /* |
384 | * Conditionally allow to enter the exclusive operation in case it's compatible | |
385 | * with the running one. This must be paired with btrfs_exclop_start_unlock and | |
386 | * btrfs_exclop_finish. | |
387 | * | |
388 | * Compatibility: | |
389 | * - the same type is already running | |
621a1ee1 | 390 | * - when trying to add a device and balance has been paused |
578bda9e DS |
391 | * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller |
392 | * must check the condition first that would allow none -> @type | |
393 | */ | |
394 | bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info, | |
395 | enum btrfs_exclusive_operation type) | |
396 | { | |
397 | spin_lock(&fs_info->super_lock); | |
621a1ee1 NB |
398 | if (fs_info->exclusive_operation == type || |
399 | (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED && | |
400 | type == BTRFS_EXCLOP_DEV_ADD)) | |
578bda9e DS |
401 | return true; |
402 | ||
403 | spin_unlock(&fs_info->super_lock); | |
404 | return false; | |
405 | } | |
406 | ||
407 | void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info) | |
408 | { | |
409 | spin_unlock(&fs_info->super_lock); | |
410 | } | |
411 | ||
c3e1f96c GR |
412 | void btrfs_exclop_finish(struct btrfs_fs_info *fs_info) |
413 | { | |
0d7ed32c | 414 | spin_lock(&fs_info->super_lock); |
c3e1f96c | 415 | WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE); |
0d7ed32c | 416 | spin_unlock(&fs_info->super_lock); |
66a2823c | 417 | sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation"); |
c3e1f96c GR |
418 | } |
419 | ||
efc0e69c NB |
420 | void btrfs_exclop_balance(struct btrfs_fs_info *fs_info, |
421 | enum btrfs_exclusive_operation op) | |
422 | { | |
423 | switch (op) { | |
424 | case BTRFS_EXCLOP_BALANCE_PAUSED: | |
425 | spin_lock(&fs_info->super_lock); | |
426 | ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE || | |
427 | fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD); | |
428 | fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED; | |
429 | spin_unlock(&fs_info->super_lock); | |
430 | break; | |
431 | case BTRFS_EXCLOP_BALANCE: | |
432 | spin_lock(&fs_info->super_lock); | |
433 | ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED); | |
434 | fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE; | |
435 | spin_unlock(&fs_info->super_lock); | |
436 | break; | |
437 | default: | |
438 | btrfs_warn(fs_info, | |
439 | "invalid exclop balance operation %d requested", op); | |
440 | } | |
441 | } | |
442 | ||
6cbff00f CH |
443 | static int btrfs_ioctl_getversion(struct file *file, int __user *arg) |
444 | { | |
496ad9aa | 445 | struct inode *inode = file_inode(file); |
6cbff00f CH |
446 | |
447 | return put_user(inode->i_generation, arg); | |
448 | } | |
f46b5a66 | 449 | |
b929c1d8 MPS |
450 | static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, |
451 | void __user *arg) | |
f7039b1d | 452 | { |
f7039b1d LD |
453 | struct btrfs_device *device; |
454 | struct request_queue *q; | |
455 | struct fstrim_range range; | |
456 | u64 minlen = ULLONG_MAX; | |
457 | u64 num_devices = 0; | |
458 | int ret; | |
459 | ||
460 | if (!capable(CAP_SYS_ADMIN)) | |
461 | return -EPERM; | |
462 | ||
1cb3dc3f NA |
463 | /* |
464 | * btrfs_trim_block_group() depends on space cache, which is not | |
465 | * available in zoned filesystem. So, disallow fitrim on a zoned | |
466 | * filesystem for now. | |
467 | */ | |
468 | if (btrfs_is_zoned(fs_info)) | |
469 | return -EOPNOTSUPP; | |
470 | ||
f35f06c3 FM |
471 | /* |
472 | * If the fs is mounted with nologreplay, which requires it to be | |
473 | * mounted in RO mode as well, we can not allow discard on free space | |
474 | * inside block groups, because log trees refer to extents that are not | |
475 | * pinned in a block group's free space cache (pinning the extents is | |
476 | * precisely the first phase of replaying a log tree). | |
477 | */ | |
478 | if (btrfs_test_opt(fs_info, NOLOGREPLAY)) | |
479 | return -EROFS; | |
480 | ||
1f78160c XG |
481 | rcu_read_lock(); |
482 | list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, | |
483 | dev_list) { | |
f7039b1d LD |
484 | if (!device->bdev) |
485 | continue; | |
486 | q = bdev_get_queue(device->bdev); | |
487 | if (blk_queue_discard(q)) { | |
488 | num_devices++; | |
50d0446e | 489 | minlen = min_t(u64, q->limits.discard_granularity, |
f7039b1d LD |
490 | minlen); |
491 | } | |
492 | } | |
1f78160c | 493 | rcu_read_unlock(); |
f4c697e6 | 494 | |
f7039b1d LD |
495 | if (!num_devices) |
496 | return -EOPNOTSUPP; | |
f7039b1d LD |
497 | if (copy_from_user(&range, arg, sizeof(range))) |
498 | return -EFAULT; | |
6ba9fc8e QW |
499 | |
500 | /* | |
501 | * NOTE: Don't truncate the range using super->total_bytes. Bytenr of | |
502 | * block group is in the logical address space, which can be any | |
503 | * sectorsize aligned bytenr in the range [0, U64_MAX]. | |
504 | */ | |
505 | if (range.len < fs_info->sb->s_blocksize) | |
f4c697e6 | 506 | return -EINVAL; |
f7039b1d LD |
507 | |
508 | range.minlen = max(range.minlen, minlen); | |
2ff7e61e | 509 | ret = btrfs_trim_fs(fs_info, &range); |
f7039b1d LD |
510 | if (ret < 0) |
511 | return ret; | |
512 | ||
513 | if (copy_to_user(arg, &range, sizeof(range))) | |
514 | return -EFAULT; | |
515 | ||
516 | return 0; | |
517 | } | |
518 | ||
e1f60a65 | 519 | int __pure btrfs_is_empty_uuid(u8 *uuid) |
dd5f9615 | 520 | { |
46e0f66a CM |
521 | int i; |
522 | ||
523 | for (i = 0; i < BTRFS_UUID_SIZE; i++) { | |
524 | if (uuid[i]) | |
525 | return 0; | |
526 | } | |
527 | return 1; | |
dd5f9615 SB |
528 | } |
529 | ||
4d4340c9 CB |
530 | static noinline int create_subvol(struct user_namespace *mnt_userns, |
531 | struct inode *dir, struct dentry *dentry, | |
52f75f4f | 532 | const char *name, int namelen, |
8696c533 | 533 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 534 | { |
0b246afa | 535 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); |
f46b5a66 CH |
536 | struct btrfs_trans_handle *trans; |
537 | struct btrfs_key key; | |
49a3c4d9 | 538 | struct btrfs_root_item *root_item; |
f46b5a66 CH |
539 | struct btrfs_inode_item *inode_item; |
540 | struct extent_buffer *leaf; | |
d5c12070 | 541 | struct btrfs_root *root = BTRFS_I(dir)->root; |
76dda93c | 542 | struct btrfs_root *new_root; |
d5c12070 | 543 | struct btrfs_block_rsv block_rsv; |
95582b00 | 544 | struct timespec64 cur_time = current_time(dir); |
5662344b | 545 | struct inode *inode; |
f46b5a66 | 546 | int ret; |
2dfb1e43 | 547 | dev_t anon_dev = 0; |
f46b5a66 | 548 | u64 objectid; |
3de4586c | 549 | u64 index = 0; |
f46b5a66 | 550 | |
49a3c4d9 DS |
551 | root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); |
552 | if (!root_item) | |
553 | return -ENOMEM; | |
554 | ||
543068a2 | 555 | ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid); |
2fbe8c8a | 556 | if (ret) |
49a3c4d9 | 557 | goto fail_free; |
6a912213 | 558 | |
2dfb1e43 QW |
559 | ret = get_anon_bdev(&anon_dev); |
560 | if (ret < 0) | |
561 | goto fail_free; | |
562 | ||
e09fe2d2 QW |
563 | /* |
564 | * Don't create subvolume whose level is not zero. Or qgroup will be | |
01327610 | 565 | * screwed up since it assumes subvolume qgroup's level to be 0. |
e09fe2d2 | 566 | */ |
49a3c4d9 DS |
567 | if (btrfs_qgroup_level(objectid)) { |
568 | ret = -ENOSPC; | |
569 | goto fail_free; | |
570 | } | |
e09fe2d2 | 571 | |
d5c12070 | 572 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
9ed74f2d | 573 | /* |
d5c12070 MX |
574 | * The same as the snapshot creation, please see the comment |
575 | * of create_snapshot(). | |
9ed74f2d | 576 | */ |
c4c129db | 577 | ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); |
d5c12070 | 578 | if (ret) |
49a3c4d9 | 579 | goto fail_free; |
d5c12070 MX |
580 | |
581 | trans = btrfs_start_transaction(root, 0); | |
582 | if (IS_ERR(trans)) { | |
583 | ret = PTR_ERR(trans); | |
e85fde51 | 584 | btrfs_subvolume_release_metadata(root, &block_rsv); |
49a3c4d9 | 585 | goto fail_free; |
d5c12070 MX |
586 | } |
587 | trans->block_rsv = &block_rsv; | |
588 | trans->bytes_reserved = block_rsv.size; | |
f46b5a66 | 589 | |
a9377422 | 590 | ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); |
6f72c7e2 AJ |
591 | if (ret) |
592 | goto fail; | |
593 | ||
9631e4cc JB |
594 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
595 | BTRFS_NESTING_NORMAL); | |
8e8a1e31 JB |
596 | if (IS_ERR(leaf)) { |
597 | ret = PTR_ERR(leaf); | |
598 | goto fail; | |
599 | } | |
f46b5a66 | 600 | |
f46b5a66 CH |
601 | btrfs_mark_buffer_dirty(leaf); |
602 | ||
49a3c4d9 | 603 | inode_item = &root_item->inode; |
3cae210f QW |
604 | btrfs_set_stack_inode_generation(inode_item, 1); |
605 | btrfs_set_stack_inode_size(inode_item, 3); | |
606 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 607 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 608 | fs_info->nodesize); |
3cae210f | 609 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
f46b5a66 | 610 | |
49a3c4d9 DS |
611 | btrfs_set_root_flags(root_item, 0); |
612 | btrfs_set_root_limit(root_item, 0); | |
3cae210f | 613 | btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); |
08fe4db1 | 614 | |
49a3c4d9 DS |
615 | btrfs_set_root_bytenr(root_item, leaf->start); |
616 | btrfs_set_root_generation(root_item, trans->transid); | |
617 | btrfs_set_root_level(root_item, 0); | |
618 | btrfs_set_root_refs(root_item, 1); | |
619 | btrfs_set_root_used(root_item, leaf->len); | |
620 | btrfs_set_root_last_snapshot(root_item, 0); | |
f46b5a66 | 621 | |
49a3c4d9 DS |
622 | btrfs_set_root_generation_v2(root_item, |
623 | btrfs_root_generation(root_item)); | |
807fc790 | 624 | generate_random_guid(root_item->uuid); |
49a3c4d9 DS |
625 | btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); |
626 | btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); | |
627 | root_item->ctime = root_item->otime; | |
628 | btrfs_set_root_ctransid(root_item, trans->transid); | |
629 | btrfs_set_root_otransid(root_item, trans->transid); | |
f46b5a66 | 630 | |
925baedd | 631 | btrfs_tree_unlock(leaf); |
f46b5a66 | 632 | |
69948022 | 633 | btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID); |
f46b5a66 CH |
634 | |
635 | key.objectid = objectid; | |
5d4f98a2 | 636 | key.offset = 0; |
962a298f | 637 | key.type = BTRFS_ROOT_ITEM_KEY; |
0b246afa | 638 | ret = btrfs_insert_root(trans, fs_info->tree_root, &key, |
49a3c4d9 | 639 | root_item); |
67addf29 FM |
640 | if (ret) { |
641 | /* | |
642 | * Since we don't abort the transaction in this case, free the | |
643 | * tree block so that we don't leak space and leave the | |
644 | * filesystem in an inconsistent state (an extent item in the | |
7a163608 | 645 | * extent tree with a backreference for a root that does not |
212a58fd | 646 | * exists). |
67addf29 | 647 | */ |
212a58fd FM |
648 | btrfs_tree_lock(leaf); |
649 | btrfs_clean_tree_block(leaf); | |
650 | btrfs_tree_unlock(leaf); | |
7a163608 | 651 | btrfs_free_tree_block(trans, objectid, leaf, 0, 1); |
67addf29 | 652 | free_extent_buffer(leaf); |
f46b5a66 | 653 | goto fail; |
67addf29 FM |
654 | } |
655 | ||
656 | free_extent_buffer(leaf); | |
657 | leaf = NULL; | |
f46b5a66 | 658 | |
76dda93c | 659 | key.offset = (u64)-1; |
2dfb1e43 | 660 | new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev); |
79787eaa | 661 | if (IS_ERR(new_root)) { |
2dfb1e43 | 662 | free_anon_bdev(anon_dev); |
79787eaa | 663 | ret = PTR_ERR(new_root); |
66642832 | 664 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
665 | goto fail; |
666 | } | |
2dfb1e43 QW |
667 | /* Freeing will be done in btrfs_put_root() of new_root */ |
668 | anon_dev = 0; | |
76dda93c | 669 | |
221581e4 JB |
670 | ret = btrfs_record_root_in_trans(trans, new_root); |
671 | if (ret) { | |
672 | btrfs_put_root(new_root); | |
673 | btrfs_abort_transaction(trans, ret); | |
674 | goto fail; | |
675 | } | |
76dda93c | 676 | |
4d4340c9 | 677 | ret = btrfs_create_subvol_root(trans, new_root, root, mnt_userns); |
00246528 | 678 | btrfs_put_root(new_root); |
ce598979 MF |
679 | if (ret) { |
680 | /* We potentially lose an unused inode item here */ | |
66642832 | 681 | btrfs_abort_transaction(trans, ret); |
ce598979 MF |
682 | goto fail; |
683 | } | |
684 | ||
f46b5a66 CH |
685 | /* |
686 | * insert the directory item | |
687 | */ | |
877574e2 | 688 | ret = btrfs_set_inode_index(BTRFS_I(dir), &index); |
79787eaa | 689 | if (ret) { |
66642832 | 690 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
691 | goto fail; |
692 | } | |
3de4586c | 693 | |
684572df | 694 | ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, |
3de4586c | 695 | BTRFS_FT_DIR, index); |
79787eaa | 696 | if (ret) { |
66642832 | 697 | btrfs_abort_transaction(trans, ret); |
f46b5a66 | 698 | goto fail; |
79787eaa | 699 | } |
0660b5af | 700 | |
6ef06d27 | 701 | btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); |
9a56fcd1 | 702 | ret = btrfs_update_inode(trans, root, BTRFS_I(dir)); |
c7e54b51 JB |
703 | if (ret) { |
704 | btrfs_abort_transaction(trans, ret); | |
705 | goto fail; | |
706 | } | |
52c26179 | 707 | |
6025c19f | 708 | ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, |
4a0cc7ca | 709 | btrfs_ino(BTRFS_I(dir)), index, name, namelen); |
c7e54b51 JB |
710 | if (ret) { |
711 | btrfs_abort_transaction(trans, ret); | |
712 | goto fail; | |
713 | } | |
f46b5a66 | 714 | |
cdb345a8 | 715 | ret = btrfs_uuid_tree_add(trans, root_item->uuid, |
6bccf3ab | 716 | BTRFS_UUID_KEY_SUBVOL, objectid); |
dd5f9615 | 717 | if (ret) |
66642832 | 718 | btrfs_abort_transaction(trans, ret); |
dd5f9615 | 719 | |
f46b5a66 | 720 | fail: |
49a3c4d9 | 721 | kfree(root_item); |
d5c12070 MX |
722 | trans->block_rsv = NULL; |
723 | trans->bytes_reserved = 0; | |
e85fde51 | 724 | btrfs_subvolume_release_metadata(root, &block_rsv); |
de6e8200 | 725 | |
1b58ae0e FM |
726 | if (ret) |
727 | btrfs_end_transaction(trans); | |
728 | else | |
729 | ret = btrfs_commit_transaction(trans); | |
1a65e24b | 730 | |
5662344b TI |
731 | if (!ret) { |
732 | inode = btrfs_lookup_dentry(dir, dentry); | |
de6e8200 LB |
733 | if (IS_ERR(inode)) |
734 | return PTR_ERR(inode); | |
5662344b TI |
735 | d_instantiate(dentry, inode); |
736 | } | |
f46b5a66 | 737 | return ret; |
49a3c4d9 DS |
738 | |
739 | fail_free: | |
2dfb1e43 QW |
740 | if (anon_dev) |
741 | free_anon_bdev(anon_dev); | |
49a3c4d9 DS |
742 | kfree(root_item); |
743 | return ret; | |
f46b5a66 CH |
744 | } |
745 | ||
e9662f70 | 746 | static int create_snapshot(struct btrfs_root *root, struct inode *dir, |
9babda9f | 747 | struct dentry *dentry, bool readonly, |
e9662f70 | 748 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 749 | { |
0b246afa | 750 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); |
2e4bfab9 | 751 | struct inode *inode; |
f46b5a66 CH |
752 | struct btrfs_pending_snapshot *pending_snapshot; |
753 | struct btrfs_trans_handle *trans; | |
2e4bfab9 | 754 | int ret; |
f46b5a66 | 755 | |
92a7cc42 | 756 | if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) |
f46b5a66 CH |
757 | return -EINVAL; |
758 | ||
eede2bf3 OS |
759 | if (atomic_read(&root->nr_swapfiles)) { |
760 | btrfs_warn(fs_info, | |
761 | "cannot snapshot subvolume with active swapfile"); | |
762 | return -ETXTBSY; | |
763 | } | |
764 | ||
23269bf5 | 765 | pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); |
a1ee7362 DS |
766 | if (!pending_snapshot) |
767 | return -ENOMEM; | |
768 | ||
2dfb1e43 QW |
769 | ret = get_anon_bdev(&pending_snapshot->anon_dev); |
770 | if (ret < 0) | |
771 | goto free_pending; | |
b0c0ea63 | 772 | pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), |
23269bf5 | 773 | GFP_KERNEL); |
8546b570 DS |
774 | pending_snapshot->path = btrfs_alloc_path(); |
775 | if (!pending_snapshot->root_item || !pending_snapshot->path) { | |
b0c0ea63 DS |
776 | ret = -ENOMEM; |
777 | goto free_pending; | |
778 | } | |
779 | ||
66d8f3dd MX |
780 | btrfs_init_block_rsv(&pending_snapshot->block_rsv, |
781 | BTRFS_BLOCK_RSV_TEMP); | |
d5c12070 MX |
782 | /* |
783 | * 1 - parent dir inode | |
784 | * 2 - dir entries | |
785 | * 1 - root item | |
786 | * 2 - root ref/backref | |
787 | * 1 - root of snapshot | |
dd5f9615 | 788 | * 1 - UUID item |
d5c12070 MX |
789 | */ |
790 | ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, | |
dd5f9615 | 791 | &pending_snapshot->block_rsv, 8, |
ee3441b4 | 792 | false); |
d5c12070 | 793 | if (ret) |
c11fbb6e | 794 | goto free_pending; |
d5c12070 | 795 | |
3de4586c | 796 | pending_snapshot->dentry = dentry; |
f46b5a66 | 797 | pending_snapshot->root = root; |
b83cc969 | 798 | pending_snapshot->readonly = readonly; |
e9662f70 | 799 | pending_snapshot->dir = dir; |
8696c533 | 800 | pending_snapshot->inherit = inherit; |
a22285a6 | 801 | |
d5c12070 | 802 | trans = btrfs_start_transaction(root, 0); |
a22285a6 YZ |
803 | if (IS_ERR(trans)) { |
804 | ret = PTR_ERR(trans); | |
805 | goto fail; | |
806 | } | |
807 | ||
28b21c55 | 808 | trans->pending_snapshot = pending_snapshot; |
9babda9f NB |
809 | |
810 | ret = btrfs_commit_transaction(trans); | |
aec8030a | 811 | if (ret) |
c37b2b62 | 812 | goto fail; |
a22285a6 YZ |
813 | |
814 | ret = pending_snapshot->error; | |
815 | if (ret) | |
816 | goto fail; | |
817 | ||
d3797308 CM |
818 | ret = btrfs_orphan_cleanup(pending_snapshot->snap); |
819 | if (ret) | |
820 | goto fail; | |
821 | ||
2b0143b5 | 822 | inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); |
2e4bfab9 YZ |
823 | if (IS_ERR(inode)) { |
824 | ret = PTR_ERR(inode); | |
825 | goto fail; | |
826 | } | |
5662344b | 827 | |
2e4bfab9 YZ |
828 | d_instantiate(dentry, inode); |
829 | ret = 0; | |
2dfb1e43 | 830 | pending_snapshot->anon_dev = 0; |
2e4bfab9 | 831 | fail: |
2dfb1e43 QW |
832 | /* Prevent double freeing of anon_dev */ |
833 | if (ret && pending_snapshot->snap) | |
834 | pending_snapshot->snap->anon_dev = 0; | |
00246528 | 835 | btrfs_put_root(pending_snapshot->snap); |
e85fde51 | 836 | btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv); |
b0c0ea63 | 837 | free_pending: |
2dfb1e43 QW |
838 | if (pending_snapshot->anon_dev) |
839 | free_anon_bdev(pending_snapshot->anon_dev); | |
b0c0ea63 | 840 | kfree(pending_snapshot->root_item); |
8546b570 | 841 | btrfs_free_path(pending_snapshot->path); |
a1ee7362 DS |
842 | kfree(pending_snapshot); |
843 | ||
f46b5a66 CH |
844 | return ret; |
845 | } | |
846 | ||
4260f7c7 SW |
847 | /* copy of may_delete in fs/namei.c() |
848 | * Check whether we can remove a link victim from directory dir, check | |
849 | * whether the type of victim is right. | |
850 | * 1. We can't do it if dir is read-only (done in permission()) | |
851 | * 2. We should have write and exec permissions on dir | |
852 | * 3. We can't remove anything from append-only dir | |
853 | * 4. We can't do anything with immutable dir (done in permission()) | |
854 | * 5. If the sticky bit on dir is set we should either | |
855 | * a. be owner of dir, or | |
856 | * b. be owner of victim, or | |
857 | * c. have CAP_FOWNER capability | |
01327610 | 858 | * 6. If the victim is append-only or immutable we can't do anything with |
4260f7c7 SW |
859 | * links pointing to it. |
860 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
861 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
862 | * 9. We can't remove a root or mountpoint. | |
863 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
864 | * nfs_async_unlink(). | |
865 | */ | |
866 | ||
c4ed533b CB |
867 | static int btrfs_may_delete(struct user_namespace *mnt_userns, |
868 | struct inode *dir, struct dentry *victim, int isdir) | |
4260f7c7 SW |
869 | { |
870 | int error; | |
871 | ||
2b0143b5 | 872 | if (d_really_is_negative(victim)) |
4260f7c7 SW |
873 | return -ENOENT; |
874 | ||
2b0143b5 | 875 | BUG_ON(d_inode(victim->d_parent) != dir); |
4fa6b5ec | 876 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
4260f7c7 | 877 | |
c4ed533b | 878 | error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); |
4260f7c7 SW |
879 | if (error) |
880 | return error; | |
881 | if (IS_APPEND(dir)) | |
882 | return -EPERM; | |
c4ed533b | 883 | if (check_sticky(mnt_userns, dir, d_inode(victim)) || |
ba73d987 CB |
884 | IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) || |
885 | IS_SWAPFILE(d_inode(victim))) | |
4260f7c7 SW |
886 | return -EPERM; |
887 | if (isdir) { | |
e36cb0b8 | 888 | if (!d_is_dir(victim)) |
4260f7c7 SW |
889 | return -ENOTDIR; |
890 | if (IS_ROOT(victim)) | |
891 | return -EBUSY; | |
e36cb0b8 | 892 | } else if (d_is_dir(victim)) |
4260f7c7 SW |
893 | return -EISDIR; |
894 | if (IS_DEADDIR(dir)) | |
895 | return -ENOENT; | |
896 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
897 | return -EBUSY; | |
898 | return 0; | |
899 | } | |
900 | ||
cb8e7090 | 901 | /* copy of may_create in fs/namei.c() */ |
4d4340c9 CB |
902 | static inline int btrfs_may_create(struct user_namespace *mnt_userns, |
903 | struct inode *dir, struct dentry *child) | |
cb8e7090 | 904 | { |
2b0143b5 | 905 | if (d_really_is_positive(child)) |
cb8e7090 CH |
906 | return -EEXIST; |
907 | if (IS_DEADDIR(dir)) | |
908 | return -ENOENT; | |
4d4340c9 | 909 | if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns)) |
5474bf40 | 910 | return -EOVERFLOW; |
4d4340c9 | 911 | return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); |
cb8e7090 CH |
912 | } |
913 | ||
914 | /* | |
915 | * Create a new subvolume below @parent. This is largely modeled after | |
916 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
917 | * inside this filesystem so it's quite a bit simpler. | |
918 | */ | |
92872094 | 919 | static noinline int btrfs_mksubvol(const struct path *parent, |
4d4340c9 | 920 | struct user_namespace *mnt_userns, |
52f75f4f | 921 | const char *name, int namelen, |
72fd032e | 922 | struct btrfs_root *snap_src, |
9babda9f | 923 | bool readonly, |
8696c533 | 924 | struct btrfs_qgroup_inherit *inherit) |
cb8e7090 | 925 | { |
0b246afa JM |
926 | struct inode *dir = d_inode(parent->dentry); |
927 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); | |
cb8e7090 CH |
928 | struct dentry *dentry; |
929 | int error; | |
930 | ||
00235411 AV |
931 | error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); |
932 | if (error == -EINTR) | |
933 | return error; | |
cb8e7090 | 934 | |
4d4340c9 | 935 | dentry = lookup_one(mnt_userns, name, parent->dentry, namelen); |
cb8e7090 CH |
936 | error = PTR_ERR(dentry); |
937 | if (IS_ERR(dentry)) | |
938 | goto out_unlock; | |
939 | ||
4d4340c9 | 940 | error = btrfs_may_create(mnt_userns, dir, dentry); |
cb8e7090 | 941 | if (error) |
a874a63e | 942 | goto out_dput; |
cb8e7090 | 943 | |
9c52057c CM |
944 | /* |
945 | * even if this name doesn't exist, we may get hash collisions. | |
946 | * check for them now when we can safely fail | |
947 | */ | |
948 | error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, | |
949 | dir->i_ino, name, | |
950 | namelen); | |
951 | if (error) | |
952 | goto out_dput; | |
953 | ||
0b246afa | 954 | down_read(&fs_info->subvol_sem); |
76dda93c YZ |
955 | |
956 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
957 | goto out_up_read; | |
958 | ||
9babda9f NB |
959 | if (snap_src) |
960 | error = create_snapshot(snap_src, dir, dentry, readonly, inherit); | |
961 | else | |
4d4340c9 | 962 | error = create_subvol(mnt_userns, dir, dentry, name, namelen, inherit); |
9babda9f | 963 | |
76dda93c YZ |
964 | if (!error) |
965 | fsnotify_mkdir(dir, dentry); | |
966 | out_up_read: | |
0b246afa | 967 | up_read(&fs_info->subvol_sem); |
cb8e7090 CH |
968 | out_dput: |
969 | dput(dentry); | |
970 | out_unlock: | |
64708539 | 971 | btrfs_inode_unlock(dir, 0); |
cb8e7090 CH |
972 | return error; |
973 | } | |
974 | ||
c11fbb6e | 975 | static noinline int btrfs_mksnapshot(const struct path *parent, |
4d4340c9 | 976 | struct user_namespace *mnt_userns, |
c11fbb6e RK |
977 | const char *name, int namelen, |
978 | struct btrfs_root *root, | |
979 | bool readonly, | |
980 | struct btrfs_qgroup_inherit *inherit) | |
981 | { | |
982 | int ret; | |
983 | bool snapshot_force_cow = false; | |
984 | ||
985 | /* | |
986 | * Force new buffered writes to reserve space even when NOCOW is | |
987 | * possible. This is to avoid later writeback (running dealloc) to | |
988 | * fallback to COW mode and unexpectedly fail with ENOSPC. | |
989 | */ | |
990 | btrfs_drew_read_lock(&root->snapshot_lock); | |
991 | ||
f9baa501 | 992 | ret = btrfs_start_delalloc_snapshot(root, false); |
c11fbb6e RK |
993 | if (ret) |
994 | goto out; | |
995 | ||
996 | /* | |
997 | * All previous writes have started writeback in NOCOW mode, so now | |
998 | * we force future writes to fallback to COW mode during snapshot | |
999 | * creation. | |
1000 | */ | |
1001 | atomic_inc(&root->snapshot_force_cow); | |
1002 | snapshot_force_cow = true; | |
1003 | ||
1004 | btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); | |
1005 | ||
4d4340c9 | 1006 | ret = btrfs_mksubvol(parent, mnt_userns, name, namelen, |
c11fbb6e RK |
1007 | root, readonly, inherit); |
1008 | out: | |
1009 | if (snapshot_force_cow) | |
1010 | atomic_dec(&root->snapshot_force_cow); | |
1011 | btrfs_drew_read_unlock(&root->snapshot_lock); | |
1012 | return ret; | |
1013 | } | |
1014 | ||
e9eec721 QW |
1015 | static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start, |
1016 | bool locked) | |
17ce6ef8 LB |
1017 | { |
1018 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
6c282eb4 LZ |
1019 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
1020 | struct extent_map *em; | |
76068cae | 1021 | const u32 sectorsize = BTRFS_I(inode)->root->fs_info->sectorsize; |
17ce6ef8 | 1022 | |
6c282eb4 LZ |
1023 | /* |
1024 | * hopefully we have this extent in the tree already, try without | |
1025 | * the full extent lock | |
1026 | */ | |
17ce6ef8 | 1027 | read_lock(&em_tree->lock); |
76068cae | 1028 | em = lookup_extent_mapping(em_tree, start, sectorsize); |
17ce6ef8 LB |
1029 | read_unlock(&em_tree->lock); |
1030 | ||
6c282eb4 | 1031 | if (!em) { |
308d9800 | 1032 | struct extent_state *cached = NULL; |
76068cae | 1033 | u64 end = start + sectorsize - 1; |
308d9800 | 1034 | |
6c282eb4 | 1035 | /* get the big lock and read metadata off disk */ |
e9eec721 QW |
1036 | if (!locked) |
1037 | lock_extent_bits(io_tree, start, end, &cached); | |
76068cae | 1038 | em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, sectorsize); |
e9eec721 QW |
1039 | if (!locked) |
1040 | unlock_extent_cached(io_tree, start, end, &cached); | |
6c282eb4 LZ |
1041 | |
1042 | if (IS_ERR(em)) | |
1043 | return NULL; | |
1044 | } | |
1045 | ||
1046 | return em; | |
1047 | } | |
17ce6ef8 | 1048 | |
e9eec721 QW |
1049 | static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em, |
1050 | bool locked) | |
6c282eb4 LZ |
1051 | { |
1052 | struct extent_map *next; | |
1053 | bool ret = true; | |
1054 | ||
1055 | /* this is the last extent */ | |
1056 | if (em->start + em->len >= i_size_read(inode)) | |
1057 | return false; | |
1058 | ||
e9eec721 | 1059 | next = defrag_lookup_extent(inode, em->start + em->len, locked); |
e9512d72 CM |
1060 | if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) |
1061 | ret = false; | |
1062 | else if ((em->block_start + em->block_len == next->block_start) && | |
ee22184b | 1063 | (em->block_len > SZ_128K && next->block_len > SZ_128K)) |
6c282eb4 LZ |
1064 | ret = false; |
1065 | ||
1066 | free_extent_map(next); | |
17ce6ef8 LB |
1067 | return ret; |
1068 | } | |
1069 | ||
5767b50c QW |
1070 | /* |
1071 | * Prepare one page to be defragged. | |
1072 | * | |
1073 | * This will ensure: | |
1074 | * | |
1075 | * - Returned page is locked and has been set up properly. | |
1076 | * - No ordered extent exists in the page. | |
1077 | * - The page is uptodate. | |
1078 | * | |
1079 | * NOTE: Caller should also wait for page writeback after the cluster is | |
1080 | * prepared, here we don't do writeback wait for each page. | |
1081 | */ | |
1082 | static struct page *defrag_prepare_one_page(struct btrfs_inode *inode, | |
1083 | pgoff_t index) | |
940100a4 | 1084 | { |
5767b50c QW |
1085 | struct address_space *mapping = inode->vfs_inode.i_mapping; |
1086 | gfp_t mask = btrfs_alloc_write_mask(mapping); | |
1087 | u64 page_start = (u64)index << PAGE_SHIFT; | |
1088 | u64 page_end = page_start + PAGE_SIZE - 1; | |
1089 | struct extent_state *cached_state = NULL; | |
1090 | struct page *page; | |
1091 | int ret; | |
1092 | ||
1093 | again: | |
1094 | page = find_or_create_page(mapping, index, mask); | |
1095 | if (!page) | |
1096 | return ERR_PTR(-ENOMEM); | |
940100a4 CM |
1097 | |
1098 | /* | |
24bcb454 OS |
1099 | * Since we can defragment files opened read-only, we can encounter |
1100 | * transparent huge pages here (see CONFIG_READ_ONLY_THP_FOR_FS). We | |
1101 | * can't do I/O using huge pages yet, so return an error for now. | |
1102 | * Filesystem transparent huge pages are typically only used for | |
1103 | * executables that explicitly enable them, so this isn't very | |
1104 | * restrictive. | |
940100a4 | 1105 | */ |
24bcb454 OS |
1106 | if (PageCompound(page)) { |
1107 | unlock_page(page); | |
1108 | put_page(page); | |
1109 | return ERR_PTR(-ETXTBSY); | |
1110 | } | |
940100a4 | 1111 | |
5767b50c QW |
1112 | ret = set_page_extent_mapped(page); |
1113 | if (ret < 0) { | |
1114 | unlock_page(page); | |
1115 | put_page(page); | |
1116 | return ERR_PTR(ret); | |
1117 | } | |
940100a4 | 1118 | |
5767b50c QW |
1119 | /* Wait for any existing ordered extent in the range */ |
1120 | while (1) { | |
1121 | struct btrfs_ordered_extent *ordered; | |
940100a4 | 1122 | |
5767b50c QW |
1123 | lock_extent_bits(&inode->io_tree, page_start, page_end, &cached_state); |
1124 | ordered = btrfs_lookup_ordered_range(inode, page_start, PAGE_SIZE); | |
1125 | unlock_extent_cached(&inode->io_tree, page_start, page_end, | |
1126 | &cached_state); | |
1127 | if (!ordered) | |
1128 | break; | |
17ce6ef8 | 1129 | |
5767b50c QW |
1130 | unlock_page(page); |
1131 | btrfs_start_ordered_extent(ordered, 1); | |
1132 | btrfs_put_ordered_extent(ordered); | |
1133 | lock_page(page); | |
1134 | /* | |
1135 | * We unlocked the page above, so we need check if it was | |
1136 | * released or not. | |
1137 | */ | |
1138 | if (page->mapping != mapping || !PagePrivate(page)) { | |
1139 | unlock_page(page); | |
1140 | put_page(page); | |
1141 | goto again; | |
1142 | } | |
1143 | } | |
4a3560c4 | 1144 | |
940100a4 | 1145 | /* |
5767b50c QW |
1146 | * Now the page range has no ordered extent any more. Read the page to |
1147 | * make it uptodate. | |
940100a4 | 1148 | */ |
5767b50c QW |
1149 | if (!PageUptodate(page)) { |
1150 | btrfs_readpage(NULL, page); | |
1151 | lock_page(page); | |
1152 | if (page->mapping != mapping || !PagePrivate(page)) { | |
1153 | unlock_page(page); | |
1154 | put_page(page); | |
1155 | goto again; | |
1156 | } | |
1157 | if (!PageUptodate(page)) { | |
1158 | unlock_page(page); | |
1159 | put_page(page); | |
1160 | return ERR_PTR(-EIO); | |
1161 | } | |
940100a4 | 1162 | } |
5767b50c | 1163 | return page; |
940100a4 CM |
1164 | } |
1165 | ||
eb793cf8 QW |
1166 | struct defrag_target_range { |
1167 | struct list_head list; | |
1168 | u64 start; | |
1169 | u64 len; | |
1170 | }; | |
1171 | ||
4cb5300b | 1172 | /* |
eb793cf8 | 1173 | * Collect all valid target extents. |
4cb5300b | 1174 | * |
eb793cf8 QW |
1175 | * @start: file offset to lookup |
1176 | * @len: length to lookup | |
1177 | * @extent_thresh: file extent size threshold, any extent size >= this value | |
1178 | * will be ignored | |
1179 | * @newer_than: only defrag extents newer than this value | |
1180 | * @do_compress: whether the defrag is doing compression | |
1181 | * if true, @extent_thresh will be ignored and all regular | |
1182 | * file extents meeting @newer_than will be targets. | |
e9eec721 | 1183 | * @locked: if the range has already held extent lock |
eb793cf8 | 1184 | * @target_list: list of targets file extents |
4cb5300b | 1185 | */ |
eb793cf8 QW |
1186 | static int defrag_collect_targets(struct btrfs_inode *inode, |
1187 | u64 start, u64 len, u32 extent_thresh, | |
1188 | u64 newer_than, bool do_compress, | |
e9eec721 | 1189 | bool locked, struct list_head *target_list) |
f46b5a66 | 1190 | { |
eb793cf8 QW |
1191 | u64 cur = start; |
1192 | int ret = 0; | |
4cb5300b | 1193 | |
eb793cf8 QW |
1194 | while (cur < start + len) { |
1195 | struct extent_map *em; | |
1196 | struct defrag_target_range *new; | |
1197 | bool next_mergeable = true; | |
1198 | u64 range_len; | |
1f12bd06 | 1199 | |
e9eec721 | 1200 | em = defrag_lookup_extent(&inode->vfs_inode, cur, locked); |
eb793cf8 QW |
1201 | if (!em) |
1202 | break; | |
4cb5300b | 1203 | |
eb793cf8 QW |
1204 | /* Skip hole/inline/preallocated extents */ |
1205 | if (em->block_start >= EXTENT_MAP_LAST_BYTE || | |
1206 | test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) | |
1207 | goto next; | |
4cb5300b | 1208 | |
eb793cf8 QW |
1209 | /* Skip older extent */ |
1210 | if (em->generation < newer_than) | |
1211 | goto next; | |
4cb5300b | 1212 | |
0cb5950f FM |
1213 | /* |
1214 | * Our start offset might be in the middle of an existing extent | |
1215 | * map, so take that into account. | |
1216 | */ | |
1217 | range_len = em->len - (cur - em->start); | |
1218 | /* | |
1219 | * If this range of the extent map is already flagged for delalloc, | |
1220 | * skip it, because: | |
1221 | * | |
1222 | * 1) We could deadlock later, when trying to reserve space for | |
1223 | * delalloc, because in case we can't immediately reserve space | |
1224 | * the flusher can start delalloc and wait for the respective | |
1225 | * ordered extents to complete. The deadlock would happen | |
1226 | * because we do the space reservation while holding the range | |
1227 | * locked, and starting writeback, or finishing an ordered | |
1228 | * extent, requires locking the range; | |
1229 | * | |
1230 | * 2) If there's delalloc there, it means there's dirty pages for | |
1231 | * which writeback has not started yet (we clean the delalloc | |
1232 | * flag when starting writeback and after creating an ordered | |
1233 | * extent). If we mark pages in an adjacent range for defrag, | |
1234 | * then we will have a larger contiguous range for delalloc, | |
1235 | * very likely resulting in a larger extent after writeback is | |
1236 | * triggered (except in a case of free space fragmentation). | |
1237 | */ | |
1238 | if (test_range_bit(&inode->io_tree, cur, cur + range_len - 1, | |
1239 | EXTENT_DELALLOC, 0, NULL)) | |
1240 | goto next; | |
1241 | ||
eb793cf8 QW |
1242 | /* |
1243 | * For do_compress case, we want to compress all valid file | |
1244 | * extents, thus no @extent_thresh or mergeable check. | |
1245 | */ | |
1246 | if (do_compress) | |
1247 | goto add; | |
1248 | ||
1249 | /* Skip too large extent */ | |
0cb5950f | 1250 | if (range_len >= extent_thresh) |
eb793cf8 QW |
1251 | goto next; |
1252 | ||
e9eec721 QW |
1253 | next_mergeable = defrag_check_next_extent(&inode->vfs_inode, em, |
1254 | locked); | |
eb793cf8 QW |
1255 | if (!next_mergeable) { |
1256 | struct defrag_target_range *last; | |
1257 | ||
1258 | /* Empty target list, no way to merge with last entry */ | |
1259 | if (list_empty(target_list)) | |
1260 | goto next; | |
1261 | last = list_entry(target_list->prev, | |
1262 | struct defrag_target_range, list); | |
1263 | /* Not mergeable with last entry */ | |
1264 | if (last->start + last->len != cur) | |
1265 | goto next; | |
1266 | ||
1267 | /* Mergeable, fall through to add it to @target_list. */ | |
32443de3 QW |
1268 | } |
1269 | ||
eb793cf8 QW |
1270 | add: |
1271 | range_len = min(extent_map_end(em), start + len) - cur; | |
1272 | /* | |
1273 | * This one is a good target, check if it can be merged into | |
1274 | * last range of the target list. | |
1275 | */ | |
1276 | if (!list_empty(target_list)) { | |
1277 | struct defrag_target_range *last; | |
1278 | ||
1279 | last = list_entry(target_list->prev, | |
1280 | struct defrag_target_range, list); | |
1281 | ASSERT(last->start + last->len <= cur); | |
1282 | if (last->start + last->len == cur) { | |
1283 | /* Mergeable, enlarge the last entry */ | |
1284 | last->len += range_len; | |
1285 | goto next; | |
1f12bd06 | 1286 | } |
eb793cf8 | 1287 | /* Fall through to allocate a new entry */ |
600a45e1 MX |
1288 | } |
1289 | ||
eb793cf8 QW |
1290 | /* Allocate new defrag_target_range */ |
1291 | new = kmalloc(sizeof(*new), GFP_NOFS); | |
1292 | if (!new) { | |
1293 | free_extent_map(em); | |
1294 | ret = -ENOMEM; | |
1295 | break; | |
4cb5300b | 1296 | } |
eb793cf8 QW |
1297 | new->start = cur; |
1298 | new->len = range_len; | |
1299 | list_add_tail(&new->list, target_list); | |
600a45e1 | 1300 | |
eb793cf8 QW |
1301 | next: |
1302 | cur = extent_map_end(em); | |
1303 | free_extent_map(em); | |
1304 | } | |
1305 | if (ret < 0) { | |
1306 | struct defrag_target_range *entry; | |
1307 | struct defrag_target_range *tmp; | |
1308 | ||
1309 | list_for_each_entry_safe(entry, tmp, target_list, list) { | |
1310 | list_del_init(&entry->list); | |
1311 | kfree(entry); | |
600a45e1 | 1312 | } |
eb793cf8 QW |
1313 | } |
1314 | return ret; | |
1315 | } | |
1316 | ||
22b398ee QW |
1317 | #define CLUSTER_SIZE (SZ_256K) |
1318 | ||
1319 | /* | |
1320 | * Defrag one contiguous target range. | |
1321 | * | |
1322 | * @inode: target inode | |
1323 | * @target: target range to defrag | |
1324 | * @pages: locked pages covering the defrag range | |
1325 | * @nr_pages: number of locked pages | |
1326 | * | |
1327 | * Caller should ensure: | |
1328 | * | |
1329 | * - Pages are prepared | |
1330 | * Pages should be locked, no ordered extent in the pages range, | |
1331 | * no writeback. | |
1332 | * | |
1333 | * - Extent bits are locked | |
1334 | */ | |
1335 | static int defrag_one_locked_target(struct btrfs_inode *inode, | |
1336 | struct defrag_target_range *target, | |
1337 | struct page **pages, int nr_pages, | |
1338 | struct extent_state **cached_state) | |
1339 | { | |
1340 | struct btrfs_fs_info *fs_info = inode->root->fs_info; | |
1341 | struct extent_changeset *data_reserved = NULL; | |
1342 | const u64 start = target->start; | |
1343 | const u64 len = target->len; | |
1344 | unsigned long last_index = (start + len - 1) >> PAGE_SHIFT; | |
1345 | unsigned long start_index = start >> PAGE_SHIFT; | |
1346 | unsigned long first_index = page_index(pages[0]); | |
1347 | int ret = 0; | |
1348 | int i; | |
1349 | ||
1350 | ASSERT(last_index - first_index + 1 <= nr_pages); | |
1351 | ||
1352 | ret = btrfs_delalloc_reserve_space(inode, &data_reserved, start, len); | |
1353 | if (ret < 0) | |
1354 | return ret; | |
1355 | clear_extent_bit(&inode->io_tree, start, start + len - 1, | |
1356 | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | | |
1357 | EXTENT_DEFRAG, 0, 0, cached_state); | |
1358 | set_extent_defrag(&inode->io_tree, start, start + len - 1, cached_state); | |
600a45e1 | 1359 | |
22b398ee QW |
1360 | /* Update the page status */ |
1361 | for (i = start_index - first_index; i <= last_index - first_index; i++) { | |
1362 | ClearPageChecked(pages[i]); | |
1363 | btrfs_page_clamp_set_dirty(fs_info, pages[i], start, len); | |
4cb5300b | 1364 | } |
22b398ee QW |
1365 | btrfs_delalloc_release_extents(inode, len); |
1366 | extent_changeset_free(data_reserved); | |
4cb5300b | 1367 | |
22b398ee QW |
1368 | return ret; |
1369 | } | |
4cb5300b | 1370 | |
e9eec721 QW |
1371 | static int defrag_one_range(struct btrfs_inode *inode, u64 start, u32 len, |
1372 | u32 extent_thresh, u64 newer_than, bool do_compress) | |
1373 | { | |
1374 | struct extent_state *cached_state = NULL; | |
1375 | struct defrag_target_range *entry; | |
1376 | struct defrag_target_range *tmp; | |
1377 | LIST_HEAD(target_list); | |
1378 | struct page **pages; | |
1379 | const u32 sectorsize = inode->root->fs_info->sectorsize; | |
1380 | u64 last_index = (start + len - 1) >> PAGE_SHIFT; | |
1381 | u64 start_index = start >> PAGE_SHIFT; | |
1382 | unsigned int nr_pages = last_index - start_index + 1; | |
1383 | int ret = 0; | |
1384 | int i; | |
4cb5300b | 1385 | |
e9eec721 QW |
1386 | ASSERT(nr_pages <= CLUSTER_SIZE / PAGE_SIZE); |
1387 | ASSERT(IS_ALIGNED(start, sectorsize) && IS_ALIGNED(len, sectorsize)); | |
4cb5300b | 1388 | |
e9eec721 QW |
1389 | pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS); |
1390 | if (!pages) | |
1391 | return -ENOMEM; | |
7f458a38 | 1392 | |
e9eec721 QW |
1393 | /* Prepare all pages */ |
1394 | for (i = 0; i < nr_pages; i++) { | |
1395 | pages[i] = defrag_prepare_one_page(inode, start_index + i); | |
1396 | if (IS_ERR(pages[i])) { | |
1397 | ret = PTR_ERR(pages[i]); | |
1398 | pages[i] = NULL; | |
1399 | goto free_pages; | |
1400 | } | |
1401 | } | |
1402 | for (i = 0; i < nr_pages; i++) | |
1403 | wait_on_page_writeback(pages[i]); | |
1404 | ||
1405 | /* Lock the pages range */ | |
1406 | lock_extent_bits(&inode->io_tree, start_index << PAGE_SHIFT, | |
1407 | (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, | |
1408 | &cached_state); | |
7f458a38 | 1409 | /* |
e9eec721 QW |
1410 | * Now we have a consistent view about the extent map, re-check |
1411 | * which range really needs to be defragged. | |
1412 | * | |
1413 | * And this time we have extent locked already, pass @locked = true | |
1414 | * so that we won't relock the extent range and cause deadlock. | |
7f458a38 | 1415 | */ |
e9eec721 QW |
1416 | ret = defrag_collect_targets(inode, start, len, extent_thresh, |
1417 | newer_than, do_compress, true, | |
1418 | &target_list); | |
1419 | if (ret < 0) | |
1420 | goto unlock_extent; | |
7f458a38 | 1421 | |
e9eec721 QW |
1422 | list_for_each_entry(entry, &target_list, list) { |
1423 | ret = defrag_one_locked_target(inode, entry, pages, nr_pages, | |
1424 | &cached_state); | |
1425 | if (ret < 0) | |
1426 | break; | |
1427 | } | |
1428 | ||
1429 | list_for_each_entry_safe(entry, tmp, &target_list, list) { | |
1430 | list_del_init(&entry->list); | |
1431 | kfree(entry); | |
1432 | } | |
1433 | unlock_extent: | |
1434 | unlock_extent_cached(&inode->io_tree, start_index << PAGE_SHIFT, | |
1435 | (last_index << PAGE_SHIFT) + PAGE_SIZE - 1, | |
1436 | &cached_state); | |
1437 | free_pages: | |
1438 | for (i = 0; i < nr_pages; i++) { | |
1439 | if (pages[i]) { | |
1440 | unlock_page(pages[i]); | |
1441 | put_page(pages[i]); | |
7f458a38 | 1442 | } |
7f458a38 | 1443 | } |
e9eec721 QW |
1444 | kfree(pages); |
1445 | return ret; | |
1446 | } | |
7f458a38 | 1447 | |
b18c3ab2 QW |
1448 | static int defrag_one_cluster(struct btrfs_inode *inode, |
1449 | struct file_ra_state *ra, | |
1450 | u64 start, u32 len, u32 extent_thresh, | |
1451 | u64 newer_than, bool do_compress, | |
1452 | unsigned long *sectors_defragged, | |
1453 | unsigned long max_sectors) | |
1454 | { | |
1455 | const u32 sectorsize = inode->root->fs_info->sectorsize; | |
1456 | struct defrag_target_range *entry; | |
1457 | struct defrag_target_range *tmp; | |
1458 | LIST_HEAD(target_list); | |
1459 | int ret; | |
4cb5300b | 1460 | |
b18c3ab2 QW |
1461 | BUILD_BUG_ON(!IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE)); |
1462 | ret = defrag_collect_targets(inode, start, len, extent_thresh, | |
1463 | newer_than, do_compress, false, | |
1464 | &target_list); | |
1465 | if (ret < 0) | |
1466 | goto out; | |
4cb5300b | 1467 | |
b18c3ab2 QW |
1468 | list_for_each_entry(entry, &target_list, list) { |
1469 | u32 range_len = entry->len; | |
4cb5300b | 1470 | |
484167da | 1471 | /* Reached or beyond the limit */ |
c080b414 QW |
1472 | if (max_sectors && *sectors_defragged >= max_sectors) { |
1473 | ret = 1; | |
b18c3ab2 | 1474 | break; |
c080b414 | 1475 | } |
4cb5300b | 1476 | |
b18c3ab2 QW |
1477 | if (max_sectors) |
1478 | range_len = min_t(u32, range_len, | |
1479 | (max_sectors - *sectors_defragged) * sectorsize); | |
4cb5300b | 1480 | |
b18c3ab2 QW |
1481 | if (ra) |
1482 | page_cache_sync_readahead(inode->vfs_inode.i_mapping, | |
1483 | ra, NULL, entry->start >> PAGE_SHIFT, | |
1484 | ((entry->start + range_len - 1) >> PAGE_SHIFT) - | |
1485 | (entry->start >> PAGE_SHIFT) + 1); | |
1486 | /* | |
1487 | * Here we may not defrag any range if holes are punched before | |
1488 | * we locked the pages. | |
1489 | * But that's fine, it only affects the @sectors_defragged | |
1490 | * accounting. | |
1491 | */ | |
1492 | ret = defrag_one_range(inode, entry->start, range_len, | |
1493 | extent_thresh, newer_than, do_compress); | |
1494 | if (ret < 0) | |
1495 | break; | |
484167da QW |
1496 | *sectors_defragged += range_len >> |
1497 | inode->root->fs_info->sectorsize_bits; | |
4cb5300b | 1498 | } |
4cb5300b | 1499 | out: |
b18c3ab2 QW |
1500 | list_for_each_entry_safe(entry, tmp, &target_list, list) { |
1501 | list_del_init(&entry->list); | |
1502 | kfree(entry); | |
4cb5300b | 1503 | } |
4cb5300b | 1504 | return ret; |
4cb5300b CM |
1505 | } |
1506 | ||
1ccc2e8a QW |
1507 | /* |
1508 | * Entry point to file defragmentation. | |
1509 | * | |
1510 | * @inode: inode to be defragged | |
1511 | * @ra: readahead state (can be NUL) | |
1512 | * @range: defrag options including range and flags | |
1513 | * @newer_than: minimum transid to defrag | |
1514 | * @max_to_defrag: max number of sectors to be defragged, if 0, the whole inode | |
1515 | * will be defragged. | |
484167da QW |
1516 | * |
1517 | * Return <0 for error. | |
c080b414 QW |
1518 | * Return >=0 for the number of sectors defragged, and range->start will be updated |
1519 | * to indicate the file offset where next defrag should be started at. | |
1520 | * (Mostly for autodefrag, which sets @max_to_defrag thus we may exit early without | |
1521 | * defragging all the range). | |
1ccc2e8a QW |
1522 | */ |
1523 | int btrfs_defrag_file(struct inode *inode, struct file_ra_state *ra, | |
4cb5300b CM |
1524 | struct btrfs_ioctl_defrag_range_args *range, |
1525 | u64 newer_than, unsigned long max_to_defrag) | |
1526 | { | |
0b246afa | 1527 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
7b508037 | 1528 | unsigned long sectors_defragged = 0; |
151a31b2 | 1529 | u64 isize = i_size_read(inode); |
7b508037 QW |
1530 | u64 cur; |
1531 | u64 last_byte; | |
1e2ef46d | 1532 | bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; |
1ccc2e8a | 1533 | bool ra_allocated = false; |
1a419d85 | 1534 | int compress_type = BTRFS_COMPRESS_ZLIB; |
7b508037 | 1535 | int ret = 0; |
aab110ab | 1536 | u32 extent_thresh = range->extent_thresh; |
27cdfde1 | 1537 | pgoff_t start_index; |
4cb5300b | 1538 | |
0abd5b17 LB |
1539 | if (isize == 0) |
1540 | return 0; | |
1541 | ||
1542 | if (range->start >= isize) | |
1543 | return -EINVAL; | |
1a419d85 | 1544 | |
1e2ef46d | 1545 | if (do_compress) { |
ce96b7ff | 1546 | if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES) |
1a419d85 LZ |
1547 | return -EINVAL; |
1548 | if (range->compress_type) | |
1549 | compress_type = range->compress_type; | |
1550 | } | |
f46b5a66 | 1551 | |
0abd5b17 | 1552 | if (extent_thresh == 0) |
ee22184b | 1553 | extent_thresh = SZ_256K; |
940100a4 | 1554 | |
7b508037 QW |
1555 | if (range->start + range->len > range->start) { |
1556 | /* Got a specific range */ | |
6b34cd8e | 1557 | last_byte = min(isize, range->start + range->len); |
7b508037 QW |
1558 | } else { |
1559 | /* Defrag until file end */ | |
6b34cd8e | 1560 | last_byte = isize; |
7b508037 QW |
1561 | } |
1562 | ||
6b34cd8e FM |
1563 | /* Align the range */ |
1564 | cur = round_down(range->start, fs_info->sectorsize); | |
1565 | last_byte = round_up(last_byte, fs_info->sectorsize) - 1; | |
1566 | ||
4cb5300b | 1567 | /* |
1ccc2e8a | 1568 | * If we were not given a ra, allocate a readahead context. As |
0a52d108 DS |
1569 | * readahead is just an optimization, defrag will work without it so |
1570 | * we don't error out. | |
4cb5300b | 1571 | */ |
1ccc2e8a QW |
1572 | if (!ra) { |
1573 | ra_allocated = true; | |
63e727ec | 1574 | ra = kzalloc(sizeof(*ra), GFP_KERNEL); |
0a52d108 DS |
1575 | if (ra) |
1576 | file_ra_state_init(ra, inode->i_mapping); | |
4cb5300b | 1577 | } |
4cb5300b | 1578 | |
27cdfde1 FM |
1579 | /* |
1580 | * Make writeback start from the beginning of the range, so that the | |
1581 | * defrag range can be written sequentially. | |
1582 | */ | |
1583 | start_index = cur >> PAGE_SHIFT; | |
1584 | if (start_index < inode->i_mapping->writeback_index) | |
1585 | inode->i_mapping->writeback_index = start_index; | |
1586 | ||
7b508037 | 1587 | while (cur < last_byte) { |
3c9d31c7 | 1588 | const unsigned long prev_sectors_defragged = sectors_defragged; |
7b508037 | 1589 | u64 cluster_end; |
008873ea | 1590 | |
7b508037 QW |
1591 | /* The cluster size 256K should always be page aligned */ |
1592 | BUILD_BUG_ON(!IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE)); | |
008873ea | 1593 | |
b767c2fc FM |
1594 | if (btrfs_defrag_cancelled(fs_info)) { |
1595 | ret = -EAGAIN; | |
1596 | break; | |
1597 | } | |
1598 | ||
7b508037 QW |
1599 | /* We want the cluster end at page boundary when possible */ |
1600 | cluster_end = (((cur >> PAGE_SHIFT) + | |
1601 | (SZ_256K >> PAGE_SHIFT)) << PAGE_SHIFT) - 1; | |
1602 | cluster_end = min(cluster_end, last_byte); | |
940100a4 | 1603 | |
64708539 | 1604 | btrfs_inode_lock(inode, 0); |
eede2bf3 OS |
1605 | if (IS_SWAPFILE(inode)) { |
1606 | ret = -ETXTBSY; | |
7b508037 QW |
1607 | btrfs_inode_unlock(inode, 0); |
1608 | break; | |
eede2bf3 | 1609 | } |
7b508037 | 1610 | if (!(inode->i_sb->s_flags & SB_ACTIVE)) { |
64708539 | 1611 | btrfs_inode_unlock(inode, 0); |
7b508037 | 1612 | break; |
ecb8bea8 | 1613 | } |
7b508037 QW |
1614 | if (do_compress) |
1615 | BTRFS_I(inode)->defrag_compress = compress_type; | |
1616 | ret = defrag_one_cluster(BTRFS_I(inode), ra, cur, | |
1617 | cluster_end + 1 - cur, extent_thresh, | |
1618 | newer_than, do_compress, | |
1619 | §ors_defragged, max_to_defrag); | |
3c9d31c7 FM |
1620 | |
1621 | if (sectors_defragged > prev_sectors_defragged) | |
1622 | balance_dirty_pages_ratelimited(inode->i_mapping); | |
1623 | ||
64708539 | 1624 | btrfs_inode_unlock(inode, 0); |
7b508037 QW |
1625 | if (ret < 0) |
1626 | break; | |
1627 | cur = cluster_end + 1; | |
c080b414 QW |
1628 | if (ret > 0) { |
1629 | ret = 0; | |
1630 | break; | |
1631 | } | |
ea0eba69 | 1632 | cond_resched(); |
f46b5a66 CH |
1633 | } |
1634 | ||
7b508037 QW |
1635 | if (ra_allocated) |
1636 | kfree(ra); | |
c080b414 QW |
1637 | /* |
1638 | * Update range.start for autodefrag, this will indicate where to start | |
1639 | * in next run. | |
1640 | */ | |
1641 | range->start = cur; | |
7b508037 QW |
1642 | if (sectors_defragged) { |
1643 | /* | |
1644 | * We have defragged some sectors, for compression case they | |
1645 | * need to be written back immediately. | |
1646 | */ | |
1647 | if (range->flags & BTRFS_DEFRAG_RANGE_START_IO) { | |
dec8ef90 | 1648 | filemap_flush(inode->i_mapping); |
7b508037 QW |
1649 | if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, |
1650 | &BTRFS_I(inode)->runtime_flags)) | |
1651 | filemap_flush(inode->i_mapping); | |
1652 | } | |
1653 | if (range->compress_type == BTRFS_COMPRESS_LZO) | |
1654 | btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); | |
1655 | else if (range->compress_type == BTRFS_COMPRESS_ZSTD) | |
1656 | btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); | |
1657 | ret = sectors_defragged; | |
dec8ef90 | 1658 | } |
1e2ef46d | 1659 | if (do_compress) { |
64708539 | 1660 | btrfs_inode_lock(inode, 0); |
eec63c65 | 1661 | BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; |
64708539 | 1662 | btrfs_inode_unlock(inode, 0); |
633085c7 | 1663 | } |
940100a4 | 1664 | return ret; |
f46b5a66 CH |
1665 | } |
1666 | ||
17aaa434 DS |
1667 | /* |
1668 | * Try to start exclusive operation @type or cancel it if it's running. | |
1669 | * | |
1670 | * Return: | |
1671 | * 0 - normal mode, newly claimed op started | |
1672 | * >0 - normal mode, something else is running, | |
1673 | * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space | |
1674 | * ECANCELED - cancel mode, successful cancel | |
1675 | * ENOTCONN - cancel mode, operation not running anymore | |
1676 | */ | |
1677 | static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info, | |
1678 | enum btrfs_exclusive_operation type, bool cancel) | |
1679 | { | |
1680 | if (!cancel) { | |
1681 | /* Start normal op */ | |
1682 | if (!btrfs_exclop_start(fs_info, type)) | |
1683 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; | |
1684 | /* Exclusive operation is now claimed */ | |
1685 | return 0; | |
1686 | } | |
1687 | ||
1688 | /* Cancel running op */ | |
1689 | if (btrfs_exclop_start_try_lock(fs_info, type)) { | |
1690 | /* | |
1691 | * This blocks any exclop finish from setting it to NONE, so we | |
1692 | * request cancellation. Either it runs and we will wait for it, | |
1693 | * or it has finished and no waiting will happen. | |
1694 | */ | |
1695 | atomic_inc(&fs_info->reloc_cancel_req); | |
1696 | btrfs_exclop_start_unlock(fs_info); | |
1697 | ||
1698 | if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) | |
1699 | wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING, | |
1700 | TASK_INTERRUPTIBLE); | |
1701 | ||
1702 | return -ECANCELED; | |
1703 | } | |
1704 | ||
1705 | /* Something else is running or none */ | |
1706 | return -ENOTCONN; | |
1707 | } | |
1708 | ||
198605a8 | 1709 | static noinline int btrfs_ioctl_resize(struct file *file, |
76dda93c | 1710 | void __user *arg) |
f46b5a66 | 1711 | { |
562d7b15 | 1712 | BTRFS_DEV_LOOKUP_ARGS(args); |
0b246afa JM |
1713 | struct inode *inode = file_inode(file); |
1714 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f46b5a66 CH |
1715 | u64 new_size; |
1716 | u64 old_size; | |
1717 | u64 devid = 1; | |
0b246afa | 1718 | struct btrfs_root *root = BTRFS_I(inode)->root; |
f46b5a66 CH |
1719 | struct btrfs_ioctl_vol_args *vol_args; |
1720 | struct btrfs_trans_handle *trans; | |
1721 | struct btrfs_device *device = NULL; | |
1722 | char *sizestr; | |
9a40f122 | 1723 | char *retptr; |
f46b5a66 CH |
1724 | char *devstr = NULL; |
1725 | int ret = 0; | |
f46b5a66 | 1726 | int mod = 0; |
bb059a37 | 1727 | bool cancel; |
f46b5a66 | 1728 | |
e441d54d CM |
1729 | if (!capable(CAP_SYS_ADMIN)) |
1730 | return -EPERM; | |
1731 | ||
198605a8 MX |
1732 | ret = mnt_want_write_file(file); |
1733 | if (ret) | |
1734 | return ret; | |
1735 | ||
bb059a37 DS |
1736 | /* |
1737 | * Read the arguments before checking exclusivity to be able to | |
1738 | * distinguish regular resize and cancel | |
1739 | */ | |
dae7b665 | 1740 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
1741 | if (IS_ERR(vol_args)) { |
1742 | ret = PTR_ERR(vol_args); | |
bb059a37 | 1743 | goto out_drop; |
c9e9f97b | 1744 | } |
5516e595 | 1745 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 | 1746 | sizestr = vol_args->name; |
bb059a37 DS |
1747 | cancel = (strcmp("cancel", sizestr) == 0); |
1748 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel); | |
1749 | if (ret) | |
1750 | goto out_free; | |
1751 | /* Exclusive operation is now claimed */ | |
1752 | ||
f46b5a66 CH |
1753 | devstr = strchr(sizestr, ':'); |
1754 | if (devstr) { | |
f46b5a66 CH |
1755 | sizestr = devstr + 1; |
1756 | *devstr = '\0'; | |
1757 | devstr = vol_args->name; | |
58dfae63 Z |
1758 | ret = kstrtoull(devstr, 10, &devid); |
1759 | if (ret) | |
bb059a37 | 1760 | goto out_finish; |
dfd79829 MX |
1761 | if (!devid) { |
1762 | ret = -EINVAL; | |
bb059a37 | 1763 | goto out_finish; |
dfd79829 | 1764 | } |
0b246afa | 1765 | btrfs_info(fs_info, "resizing devid %llu", devid); |
f46b5a66 | 1766 | } |
dba60f3f | 1767 | |
562d7b15 JB |
1768 | args.devid = devid; |
1769 | device = btrfs_find_device(fs_info->fs_devices, &args); | |
f46b5a66 | 1770 | if (!device) { |
0b246afa JM |
1771 | btrfs_info(fs_info, "resizer unable to find device %llu", |
1772 | devid); | |
dfd79829 | 1773 | ret = -ENODEV; |
bb059a37 | 1774 | goto out_finish; |
f46b5a66 | 1775 | } |
dba60f3f | 1776 | |
ebbede42 | 1777 | if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { |
0b246afa | 1778 | btrfs_info(fs_info, |
efe120a0 | 1779 | "resizer unable to apply on readonly device %llu", |
c1c9ff7c | 1780 | devid); |
dfd79829 | 1781 | ret = -EPERM; |
bb059a37 | 1782 | goto out_finish; |
4e42ae1b LB |
1783 | } |
1784 | ||
f46b5a66 | 1785 | if (!strcmp(sizestr, "max")) |
cda00eba | 1786 | new_size = bdev_nr_bytes(device->bdev); |
f46b5a66 CH |
1787 | else { |
1788 | if (sizestr[0] == '-') { | |
1789 | mod = -1; | |
1790 | sizestr++; | |
1791 | } else if (sizestr[0] == '+') { | |
1792 | mod = 1; | |
1793 | sizestr++; | |
1794 | } | |
9a40f122 GH |
1795 | new_size = memparse(sizestr, &retptr); |
1796 | if (*retptr != '\0' || new_size == 0) { | |
f46b5a66 | 1797 | ret = -EINVAL; |
bb059a37 | 1798 | goto out_finish; |
f46b5a66 CH |
1799 | } |
1800 | } | |
1801 | ||
401e29c1 | 1802 | if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { |
dfd79829 | 1803 | ret = -EPERM; |
bb059a37 | 1804 | goto out_finish; |
63a212ab SB |
1805 | } |
1806 | ||
7cc8e58d | 1807 | old_size = btrfs_device_get_total_bytes(device); |
f46b5a66 CH |
1808 | |
1809 | if (mod < 0) { | |
1810 | if (new_size > old_size) { | |
1811 | ret = -EINVAL; | |
bb059a37 | 1812 | goto out_finish; |
f46b5a66 CH |
1813 | } |
1814 | new_size = old_size - new_size; | |
1815 | } else if (mod > 0) { | |
eb8052e0 | 1816 | if (new_size > ULLONG_MAX - old_size) { |
902c68a4 | 1817 | ret = -ERANGE; |
bb059a37 | 1818 | goto out_finish; |
eb8052e0 | 1819 | } |
f46b5a66 CH |
1820 | new_size = old_size + new_size; |
1821 | } | |
1822 | ||
ee22184b | 1823 | if (new_size < SZ_256M) { |
f46b5a66 | 1824 | ret = -EINVAL; |
bb059a37 | 1825 | goto out_finish; |
f46b5a66 | 1826 | } |
cda00eba | 1827 | if (new_size > bdev_nr_bytes(device->bdev)) { |
f46b5a66 | 1828 | ret = -EFBIG; |
bb059a37 | 1829 | goto out_finish; |
f46b5a66 CH |
1830 | } |
1831 | ||
47f08b96 | 1832 | new_size = round_down(new_size, fs_info->sectorsize); |
f46b5a66 | 1833 | |
f46b5a66 | 1834 | if (new_size > old_size) { |
a22285a6 | 1835 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
1836 | if (IS_ERR(trans)) { |
1837 | ret = PTR_ERR(trans); | |
bb059a37 | 1838 | goto out_finish; |
98d5dc13 | 1839 | } |
f46b5a66 | 1840 | ret = btrfs_grow_device(trans, device, new_size); |
3a45bb20 | 1841 | btrfs_commit_transaction(trans); |
ece7d20e | 1842 | } else if (new_size < old_size) { |
f46b5a66 | 1843 | ret = btrfs_shrink_device(device, new_size); |
0253f40e | 1844 | } /* equal, nothing need to do */ |
f46b5a66 | 1845 | |
faf8f7b9 MPS |
1846 | if (ret == 0 && new_size != old_size) |
1847 | btrfs_info_in_rcu(fs_info, | |
1848 | "resize device %s (devid %llu) from %llu to %llu", | |
1849 | rcu_str_deref(device->name), device->devid, | |
1850 | old_size, new_size); | |
bb059a37 DS |
1851 | out_finish: |
1852 | btrfs_exclop_finish(fs_info); | |
c9e9f97b | 1853 | out_free: |
f46b5a66 | 1854 | kfree(vol_args); |
bb059a37 | 1855 | out_drop: |
18f39c41 | 1856 | mnt_drop_write_file(file); |
f46b5a66 CH |
1857 | return ret; |
1858 | } | |
1859 | ||
5d54c67e | 1860 | static noinline int __btrfs_ioctl_snap_create(struct file *file, |
4d4340c9 | 1861 | struct user_namespace *mnt_userns, |
52f75f4f | 1862 | const char *name, unsigned long fd, int subvol, |
5d54c67e | 1863 | bool readonly, |
8696c533 | 1864 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 1865 | { |
f46b5a66 | 1866 | int namelen; |
3de4586c | 1867 | int ret = 0; |
f46b5a66 | 1868 | |
325c50e3 JM |
1869 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1870 | return -ENOTDIR; | |
1871 | ||
a874a63e LB |
1872 | ret = mnt_want_write_file(file); |
1873 | if (ret) | |
1874 | goto out; | |
1875 | ||
72fd032e SW |
1876 | namelen = strlen(name); |
1877 | if (strchr(name, '/')) { | |
f46b5a66 | 1878 | ret = -EINVAL; |
a874a63e | 1879 | goto out_drop_write; |
f46b5a66 CH |
1880 | } |
1881 | ||
16780cab CM |
1882 | if (name[0] == '.' && |
1883 | (namelen == 1 || (name[1] == '.' && namelen == 2))) { | |
1884 | ret = -EEXIST; | |
a874a63e | 1885 | goto out_drop_write; |
16780cab CM |
1886 | } |
1887 | ||
3de4586c | 1888 | if (subvol) { |
4d4340c9 CB |
1889 | ret = btrfs_mksubvol(&file->f_path, mnt_userns, name, |
1890 | namelen, NULL, readonly, inherit); | |
cb8e7090 | 1891 | } else { |
2903ff01 | 1892 | struct fd src = fdget(fd); |
3de4586c | 1893 | struct inode *src_inode; |
2903ff01 | 1894 | if (!src.file) { |
3de4586c | 1895 | ret = -EINVAL; |
a874a63e | 1896 | goto out_drop_write; |
3de4586c CM |
1897 | } |
1898 | ||
496ad9aa AV |
1899 | src_inode = file_inode(src.file); |
1900 | if (src_inode->i_sb != file_inode(file)->i_sb) { | |
c79b4713 | 1901 | btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, |
efe120a0 | 1902 | "Snapshot src from another FS"); |
23ad5b17 | 1903 | ret = -EXDEV; |
4d4340c9 | 1904 | } else if (!inode_owner_or_capable(mnt_userns, src_inode)) { |
d0242061 DS |
1905 | /* |
1906 | * Subvolume creation is not restricted, but snapshots | |
1907 | * are limited to own subvolumes only | |
1908 | */ | |
1909 | ret = -EPERM; | |
ecd18815 | 1910 | } else { |
4d4340c9 CB |
1911 | ret = btrfs_mksnapshot(&file->f_path, mnt_userns, |
1912 | name, namelen, | |
1913 | BTRFS_I(src_inode)->root, | |
1914 | readonly, inherit); | |
3de4586c | 1915 | } |
2903ff01 | 1916 | fdput(src); |
cb8e7090 | 1917 | } |
a874a63e LB |
1918 | out_drop_write: |
1919 | mnt_drop_write_file(file); | |
f46b5a66 | 1920 | out: |
72fd032e SW |
1921 | return ret; |
1922 | } | |
1923 | ||
1924 | static noinline int btrfs_ioctl_snap_create(struct file *file, | |
fa0d2b9b | 1925 | void __user *arg, int subvol) |
72fd032e | 1926 | { |
fa0d2b9b | 1927 | struct btrfs_ioctl_vol_args *vol_args; |
72fd032e SW |
1928 | int ret; |
1929 | ||
325c50e3 JM |
1930 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1931 | return -ENOTDIR; | |
1932 | ||
fa0d2b9b LZ |
1933 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1934 | if (IS_ERR(vol_args)) | |
1935 | return PTR_ERR(vol_args); | |
1936 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
72fd032e | 1937 | |
4d4340c9 CB |
1938 | ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), |
1939 | vol_args->name, vol_args->fd, subvol, | |
1940 | false, NULL); | |
fdfb1e4f | 1941 | |
fa0d2b9b LZ |
1942 | kfree(vol_args); |
1943 | return ret; | |
1944 | } | |
fdfb1e4f | 1945 | |
fa0d2b9b LZ |
1946 | static noinline int btrfs_ioctl_snap_create_v2(struct file *file, |
1947 | void __user *arg, int subvol) | |
1948 | { | |
1949 | struct btrfs_ioctl_vol_args_v2 *vol_args; | |
1950 | int ret; | |
b83cc969 | 1951 | bool readonly = false; |
6f72c7e2 | 1952 | struct btrfs_qgroup_inherit *inherit = NULL; |
75eaa0e2 | 1953 | |
325c50e3 JM |
1954 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1955 | return -ENOTDIR; | |
1956 | ||
fa0d2b9b LZ |
1957 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1958 | if (IS_ERR(vol_args)) | |
1959 | return PTR_ERR(vol_args); | |
1960 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
75eaa0e2 | 1961 | |
673990db | 1962 | if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) { |
b83cc969 | 1963 | ret = -EOPNOTSUPP; |
c47ca32d | 1964 | goto free_args; |
72fd032e | 1965 | } |
fa0d2b9b | 1966 | |
b83cc969 LZ |
1967 | if (vol_args->flags & BTRFS_SUBVOL_RDONLY) |
1968 | readonly = true; | |
6f72c7e2 | 1969 | if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { |
5011c5a6 DC |
1970 | u64 nums; |
1971 | ||
1972 | if (vol_args->size < sizeof(*inherit) || | |
1973 | vol_args->size > PAGE_SIZE) { | |
6f72c7e2 | 1974 | ret = -EINVAL; |
c47ca32d | 1975 | goto free_args; |
6f72c7e2 AJ |
1976 | } |
1977 | inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); | |
1978 | if (IS_ERR(inherit)) { | |
1979 | ret = PTR_ERR(inherit); | |
c47ca32d | 1980 | goto free_args; |
6f72c7e2 | 1981 | } |
5011c5a6 DC |
1982 | |
1983 | if (inherit->num_qgroups > PAGE_SIZE || | |
1984 | inherit->num_ref_copies > PAGE_SIZE || | |
1985 | inherit->num_excl_copies > PAGE_SIZE) { | |
1986 | ret = -EINVAL; | |
1987 | goto free_inherit; | |
1988 | } | |
1989 | ||
1990 | nums = inherit->num_qgroups + 2 * inherit->num_ref_copies + | |
1991 | 2 * inherit->num_excl_copies; | |
1992 | if (vol_args->size != struct_size(inherit, qgroups, nums)) { | |
1993 | ret = -EINVAL; | |
1994 | goto free_inherit; | |
1995 | } | |
6f72c7e2 | 1996 | } |
fa0d2b9b | 1997 | |
4d4340c9 CB |
1998 | ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), |
1999 | vol_args->name, vol_args->fd, subvol, | |
2000 | readonly, inherit); | |
c47ca32d DC |
2001 | if (ret) |
2002 | goto free_inherit; | |
c47ca32d | 2003 | free_inherit: |
6f72c7e2 | 2004 | kfree(inherit); |
c47ca32d DC |
2005 | free_args: |
2006 | kfree(vol_args); | |
f46b5a66 CH |
2007 | return ret; |
2008 | } | |
2009 | ||
0caa102d LZ |
2010 | static noinline int btrfs_ioctl_subvol_getflags(struct file *file, |
2011 | void __user *arg) | |
2012 | { | |
496ad9aa | 2013 | struct inode *inode = file_inode(file); |
0b246afa | 2014 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
0caa102d LZ |
2015 | struct btrfs_root *root = BTRFS_I(inode)->root; |
2016 | int ret = 0; | |
2017 | u64 flags = 0; | |
2018 | ||
4a0cc7ca | 2019 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
2020 | return -EINVAL; |
2021 | ||
0b246afa | 2022 | down_read(&fs_info->subvol_sem); |
0caa102d LZ |
2023 | if (btrfs_root_readonly(root)) |
2024 | flags |= BTRFS_SUBVOL_RDONLY; | |
0b246afa | 2025 | up_read(&fs_info->subvol_sem); |
0caa102d LZ |
2026 | |
2027 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
2028 | ret = -EFAULT; | |
2029 | ||
2030 | return ret; | |
2031 | } | |
2032 | ||
2033 | static noinline int btrfs_ioctl_subvol_setflags(struct file *file, | |
2034 | void __user *arg) | |
2035 | { | |
496ad9aa | 2036 | struct inode *inode = file_inode(file); |
0b246afa | 2037 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
0caa102d LZ |
2038 | struct btrfs_root *root = BTRFS_I(inode)->root; |
2039 | struct btrfs_trans_handle *trans; | |
2040 | u64 root_flags; | |
2041 | u64 flags; | |
2042 | int ret = 0; | |
2043 | ||
39e1674f | 2044 | if (!inode_owner_or_capable(file_mnt_user_ns(file), inode)) |
bd60ea0f DS |
2045 | return -EPERM; |
2046 | ||
b9ca0664 LB |
2047 | ret = mnt_want_write_file(file); |
2048 | if (ret) | |
2049 | goto out; | |
0caa102d | 2050 | |
4a0cc7ca | 2051 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
b9ca0664 LB |
2052 | ret = -EINVAL; |
2053 | goto out_drop_write; | |
2054 | } | |
0caa102d | 2055 | |
b9ca0664 LB |
2056 | if (copy_from_user(&flags, arg, sizeof(flags))) { |
2057 | ret = -EFAULT; | |
2058 | goto out_drop_write; | |
2059 | } | |
0caa102d | 2060 | |
b9ca0664 LB |
2061 | if (flags & ~BTRFS_SUBVOL_RDONLY) { |
2062 | ret = -EOPNOTSUPP; | |
2063 | goto out_drop_write; | |
2064 | } | |
0caa102d | 2065 | |
0b246afa | 2066 | down_write(&fs_info->subvol_sem); |
0caa102d LZ |
2067 | |
2068 | /* nothing to do */ | |
2069 | if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) | |
b9ca0664 | 2070 | goto out_drop_sem; |
0caa102d LZ |
2071 | |
2072 | root_flags = btrfs_root_flags(&root->root_item); | |
2c686537 | 2073 | if (flags & BTRFS_SUBVOL_RDONLY) { |
0caa102d LZ |
2074 | btrfs_set_root_flags(&root->root_item, |
2075 | root_flags | BTRFS_ROOT_SUBVOL_RDONLY); | |
2c686537 DS |
2076 | } else { |
2077 | /* | |
2078 | * Block RO -> RW transition if this subvolume is involved in | |
2079 | * send | |
2080 | */ | |
2081 | spin_lock(&root->root_item_lock); | |
2082 | if (root->send_in_progress == 0) { | |
2083 | btrfs_set_root_flags(&root->root_item, | |
0caa102d | 2084 | root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); |
2c686537 DS |
2085 | spin_unlock(&root->root_item_lock); |
2086 | } else { | |
2087 | spin_unlock(&root->root_item_lock); | |
0b246afa JM |
2088 | btrfs_warn(fs_info, |
2089 | "Attempt to set subvolume %llu read-write during send", | |
2090 | root->root_key.objectid); | |
2c686537 DS |
2091 | ret = -EPERM; |
2092 | goto out_drop_sem; | |
2093 | } | |
2094 | } | |
0caa102d LZ |
2095 | |
2096 | trans = btrfs_start_transaction(root, 1); | |
2097 | if (IS_ERR(trans)) { | |
2098 | ret = PTR_ERR(trans); | |
2099 | goto out_reset; | |
2100 | } | |
2101 | ||
0b246afa | 2102 | ret = btrfs_update_root(trans, fs_info->tree_root, |
0caa102d | 2103 | &root->root_key, &root->root_item); |
9417ebc8 NB |
2104 | if (ret < 0) { |
2105 | btrfs_end_transaction(trans); | |
2106 | goto out_reset; | |
2107 | } | |
2108 | ||
2109 | ret = btrfs_commit_transaction(trans); | |
0caa102d | 2110 | |
0caa102d LZ |
2111 | out_reset: |
2112 | if (ret) | |
2113 | btrfs_set_root_flags(&root->root_item, root_flags); | |
b9ca0664 | 2114 | out_drop_sem: |
0b246afa | 2115 | up_write(&fs_info->subvol_sem); |
b9ca0664 LB |
2116 | out_drop_write: |
2117 | mnt_drop_write_file(file); | |
2118 | out: | |
0caa102d LZ |
2119 | return ret; |
2120 | } | |
2121 | ||
ac8e9819 CM |
2122 | static noinline int key_in_sk(struct btrfs_key *key, |
2123 | struct btrfs_ioctl_search_key *sk) | |
2124 | { | |
abc6e134 CM |
2125 | struct btrfs_key test; |
2126 | int ret; | |
2127 | ||
2128 | test.objectid = sk->min_objectid; | |
2129 | test.type = sk->min_type; | |
2130 | test.offset = sk->min_offset; | |
2131 | ||
2132 | ret = btrfs_comp_cpu_keys(key, &test); | |
2133 | if (ret < 0) | |
ac8e9819 | 2134 | return 0; |
abc6e134 CM |
2135 | |
2136 | test.objectid = sk->max_objectid; | |
2137 | test.type = sk->max_type; | |
2138 | test.offset = sk->max_offset; | |
2139 | ||
2140 | ret = btrfs_comp_cpu_keys(key, &test); | |
2141 | if (ret > 0) | |
ac8e9819 CM |
2142 | return 0; |
2143 | return 1; | |
2144 | } | |
2145 | ||
df397565 | 2146 | static noinline int copy_to_sk(struct btrfs_path *path, |
ac8e9819 CM |
2147 | struct btrfs_key *key, |
2148 | struct btrfs_ioctl_search_key *sk, | |
9b6e817d | 2149 | size_t *buf_size, |
ba346b35 | 2150 | char __user *ubuf, |
ac8e9819 CM |
2151 | unsigned long *sk_offset, |
2152 | int *num_found) | |
2153 | { | |
2154 | u64 found_transid; | |
2155 | struct extent_buffer *leaf; | |
2156 | struct btrfs_ioctl_search_header sh; | |
dd81d459 | 2157 | struct btrfs_key test; |
ac8e9819 CM |
2158 | unsigned long item_off; |
2159 | unsigned long item_len; | |
2160 | int nritems; | |
2161 | int i; | |
2162 | int slot; | |
ac8e9819 CM |
2163 | int ret = 0; |
2164 | ||
2165 | leaf = path->nodes[0]; | |
2166 | slot = path->slots[0]; | |
2167 | nritems = btrfs_header_nritems(leaf); | |
2168 | ||
2169 | if (btrfs_header_generation(leaf) > sk->max_transid) { | |
2170 | i = nritems; | |
2171 | goto advance_key; | |
2172 | } | |
2173 | found_transid = btrfs_header_generation(leaf); | |
2174 | ||
2175 | for (i = slot; i < nritems; i++) { | |
2176 | item_off = btrfs_item_ptr_offset(leaf, i); | |
3212fa14 | 2177 | item_len = btrfs_item_size(leaf, i); |
ac8e9819 | 2178 | |
03b71c6c GP |
2179 | btrfs_item_key_to_cpu(leaf, key, i); |
2180 | if (!key_in_sk(key, sk)) | |
2181 | continue; | |
2182 | ||
9b6e817d | 2183 | if (sizeof(sh) + item_len > *buf_size) { |
8f5f6178 GH |
2184 | if (*num_found) { |
2185 | ret = 1; | |
2186 | goto out; | |
2187 | } | |
2188 | ||
2189 | /* | |
2190 | * return one empty item back for v1, which does not | |
2191 | * handle -EOVERFLOW | |
2192 | */ | |
2193 | ||
9b6e817d | 2194 | *buf_size = sizeof(sh) + item_len; |
ac8e9819 | 2195 | item_len = 0; |
8f5f6178 GH |
2196 | ret = -EOVERFLOW; |
2197 | } | |
ac8e9819 | 2198 | |
9b6e817d | 2199 | if (sizeof(sh) + item_len + *sk_offset > *buf_size) { |
ac8e9819 | 2200 | ret = 1; |
25c9bc2e | 2201 | goto out; |
ac8e9819 CM |
2202 | } |
2203 | ||
ac8e9819 CM |
2204 | sh.objectid = key->objectid; |
2205 | sh.offset = key->offset; | |
2206 | sh.type = key->type; | |
2207 | sh.len = item_len; | |
2208 | sh.transid = found_transid; | |
2209 | ||
a48b73ec JB |
2210 | /* |
2211 | * Copy search result header. If we fault then loop again so we | |
2212 | * can fault in the pages and -EFAULT there if there's a | |
2213 | * problem. Otherwise we'll fault and then copy the buffer in | |
2214 | * properly this next time through | |
2215 | */ | |
2216 | if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) { | |
2217 | ret = 0; | |
ba346b35 GH |
2218 | goto out; |
2219 | } | |
2220 | ||
ac8e9819 CM |
2221 | *sk_offset += sizeof(sh); |
2222 | ||
2223 | if (item_len) { | |
ba346b35 | 2224 | char __user *up = ubuf + *sk_offset; |
a48b73ec JB |
2225 | /* |
2226 | * Copy the item, same behavior as above, but reset the | |
2227 | * * sk_offset so we copy the full thing again. | |
2228 | */ | |
2229 | if (read_extent_buffer_to_user_nofault(leaf, up, | |
2230 | item_off, item_len)) { | |
2231 | ret = 0; | |
2232 | *sk_offset -= sizeof(sh); | |
ba346b35 GH |
2233 | goto out; |
2234 | } | |
2235 | ||
ac8e9819 | 2236 | *sk_offset += item_len; |
ac8e9819 | 2237 | } |
e2156867 | 2238 | (*num_found)++; |
ac8e9819 | 2239 | |
8f5f6178 GH |
2240 | if (ret) /* -EOVERFLOW from above */ |
2241 | goto out; | |
2242 | ||
25c9bc2e GH |
2243 | if (*num_found >= sk->nr_items) { |
2244 | ret = 1; | |
2245 | goto out; | |
2246 | } | |
ac8e9819 CM |
2247 | } |
2248 | advance_key: | |
abc6e134 | 2249 | ret = 0; |
dd81d459 NA |
2250 | test.objectid = sk->max_objectid; |
2251 | test.type = sk->max_type; | |
2252 | test.offset = sk->max_offset; | |
2253 | if (btrfs_comp_cpu_keys(key, &test) >= 0) | |
2254 | ret = 1; | |
2255 | else if (key->offset < (u64)-1) | |
ac8e9819 | 2256 | key->offset++; |
dd81d459 | 2257 | else if (key->type < (u8)-1) { |
abc6e134 | 2258 | key->offset = 0; |
ac8e9819 | 2259 | key->type++; |
dd81d459 | 2260 | } else if (key->objectid < (u64)-1) { |
abc6e134 CM |
2261 | key->offset = 0; |
2262 | key->type = 0; | |
ac8e9819 | 2263 | key->objectid++; |
abc6e134 CM |
2264 | } else |
2265 | ret = 1; | |
25c9bc2e | 2266 | out: |
ba346b35 GH |
2267 | /* |
2268 | * 0: all items from this leaf copied, continue with next | |
2269 | * 1: * more items can be copied, but unused buffer is too small | |
2270 | * * all items were found | |
2271 | * Either way, it will stops the loop which iterates to the next | |
2272 | * leaf | |
2273 | * -EOVERFLOW: item was to large for buffer | |
2274 | * -EFAULT: could not copy extent buffer back to userspace | |
2275 | */ | |
ac8e9819 CM |
2276 | return ret; |
2277 | } | |
2278 | ||
2279 | static noinline int search_ioctl(struct inode *inode, | |
12544442 | 2280 | struct btrfs_ioctl_search_key *sk, |
9b6e817d | 2281 | size_t *buf_size, |
ba346b35 | 2282 | char __user *ubuf) |
ac8e9819 | 2283 | { |
0b246afa | 2284 | struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); |
ac8e9819 CM |
2285 | struct btrfs_root *root; |
2286 | struct btrfs_key key; | |
ac8e9819 | 2287 | struct btrfs_path *path; |
ac8e9819 CM |
2288 | int ret; |
2289 | int num_found = 0; | |
2290 | unsigned long sk_offset = 0; | |
2291 | ||
9b6e817d GH |
2292 | if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { |
2293 | *buf_size = sizeof(struct btrfs_ioctl_search_header); | |
12544442 | 2294 | return -EOVERFLOW; |
9b6e817d | 2295 | } |
12544442 | 2296 | |
ac8e9819 CM |
2297 | path = btrfs_alloc_path(); |
2298 | if (!path) | |
2299 | return -ENOMEM; | |
2300 | ||
2301 | if (sk->tree_id == 0) { | |
2302 | /* search the root of the inode that was passed */ | |
00246528 | 2303 | root = btrfs_grab_root(BTRFS_I(inode)->root); |
ac8e9819 | 2304 | } else { |
56e9357a | 2305 | root = btrfs_get_fs_root(info, sk->tree_id, true); |
ac8e9819 | 2306 | if (IS_ERR(root)) { |
ac8e9819 | 2307 | btrfs_free_path(path); |
ad1e3d56 | 2308 | return PTR_ERR(root); |
ac8e9819 CM |
2309 | } |
2310 | } | |
2311 | ||
2312 | key.objectid = sk->min_objectid; | |
2313 | key.type = sk->min_type; | |
2314 | key.offset = sk->min_offset; | |
2315 | ||
67871254 | 2316 | while (1) { |
bb523b40 AG |
2317 | ret = -EFAULT; |
2318 | if (fault_in_writeable(ubuf + sk_offset, *buf_size - sk_offset)) | |
a48b73ec JB |
2319 | break; |
2320 | ||
6174d3cb | 2321 | ret = btrfs_search_forward(root, &key, path, sk->min_transid); |
ac8e9819 CM |
2322 | if (ret != 0) { |
2323 | if (ret > 0) | |
2324 | ret = 0; | |
2325 | goto err; | |
2326 | } | |
df397565 | 2327 | ret = copy_to_sk(path, &key, sk, buf_size, ubuf, |
ac8e9819 | 2328 | &sk_offset, &num_found); |
b3b4aa74 | 2329 | btrfs_release_path(path); |
25c9bc2e | 2330 | if (ret) |
ac8e9819 CM |
2331 | break; |
2332 | ||
2333 | } | |
8f5f6178 GH |
2334 | if (ret > 0) |
2335 | ret = 0; | |
ac8e9819 CM |
2336 | err: |
2337 | sk->nr_items = num_found; | |
00246528 | 2338 | btrfs_put_root(root); |
ac8e9819 CM |
2339 | btrfs_free_path(path); |
2340 | return ret; | |
2341 | } | |
2342 | ||
2343 | static noinline int btrfs_ioctl_tree_search(struct file *file, | |
2344 | void __user *argp) | |
2345 | { | |
ba346b35 GH |
2346 | struct btrfs_ioctl_search_args __user *uargs; |
2347 | struct btrfs_ioctl_search_key sk; | |
9b6e817d GH |
2348 | struct inode *inode; |
2349 | int ret; | |
2350 | size_t buf_size; | |
ac8e9819 CM |
2351 | |
2352 | if (!capable(CAP_SYS_ADMIN)) | |
2353 | return -EPERM; | |
2354 | ||
ba346b35 GH |
2355 | uargs = (struct btrfs_ioctl_search_args __user *)argp; |
2356 | ||
2357 | if (copy_from_user(&sk, &uargs->key, sizeof(sk))) | |
2358 | return -EFAULT; | |
ac8e9819 | 2359 | |
ba346b35 | 2360 | buf_size = sizeof(uargs->buf); |
ac8e9819 | 2361 | |
496ad9aa | 2362 | inode = file_inode(file); |
ba346b35 | 2363 | ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); |
8f5f6178 GH |
2364 | |
2365 | /* | |
2366 | * In the origin implementation an overflow is handled by returning a | |
2367 | * search header with a len of zero, so reset ret. | |
2368 | */ | |
2369 | if (ret == -EOVERFLOW) | |
2370 | ret = 0; | |
2371 | ||
ba346b35 | 2372 | if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) |
ac8e9819 | 2373 | ret = -EFAULT; |
ac8e9819 CM |
2374 | return ret; |
2375 | } | |
2376 | ||
cc68a8a5 GH |
2377 | static noinline int btrfs_ioctl_tree_search_v2(struct file *file, |
2378 | void __user *argp) | |
2379 | { | |
2380 | struct btrfs_ioctl_search_args_v2 __user *uarg; | |
2381 | struct btrfs_ioctl_search_args_v2 args; | |
2382 | struct inode *inode; | |
2383 | int ret; | |
2384 | size_t buf_size; | |
ee22184b | 2385 | const size_t buf_limit = SZ_16M; |
cc68a8a5 GH |
2386 | |
2387 | if (!capable(CAP_SYS_ADMIN)) | |
2388 | return -EPERM; | |
2389 | ||
2390 | /* copy search header and buffer size */ | |
2391 | uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; | |
2392 | if (copy_from_user(&args, uarg, sizeof(args))) | |
2393 | return -EFAULT; | |
2394 | ||
2395 | buf_size = args.buf_size; | |
2396 | ||
cc68a8a5 GH |
2397 | /* limit result size to 16MB */ |
2398 | if (buf_size > buf_limit) | |
2399 | buf_size = buf_limit; | |
2400 | ||
2401 | inode = file_inode(file); | |
2402 | ret = search_ioctl(inode, &args.key, &buf_size, | |
718dc5fa | 2403 | (char __user *)(&uarg->buf[0])); |
cc68a8a5 GH |
2404 | if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) |
2405 | ret = -EFAULT; | |
2406 | else if (ret == -EOVERFLOW && | |
2407 | copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) | |
2408 | ret = -EFAULT; | |
2409 | ||
ac8e9819 CM |
2410 | return ret; |
2411 | } | |
2412 | ||
98d377a0 | 2413 | /* |
ac8e9819 CM |
2414 | * Search INODE_REFs to identify path name of 'dirid' directory |
2415 | * in a 'tree_id' tree. and sets path name to 'name'. | |
2416 | */ | |
98d377a0 TH |
2417 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
2418 | u64 tree_id, u64 dirid, char *name) | |
2419 | { | |
2420 | struct btrfs_root *root; | |
2421 | struct btrfs_key key; | |
ac8e9819 | 2422 | char *ptr; |
98d377a0 TH |
2423 | int ret = -1; |
2424 | int slot; | |
2425 | int len; | |
2426 | int total_len = 0; | |
2427 | struct btrfs_inode_ref *iref; | |
2428 | struct extent_buffer *l; | |
2429 | struct btrfs_path *path; | |
2430 | ||
2431 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
2432 | name[0]='\0'; | |
2433 | return 0; | |
2434 | } | |
2435 | ||
2436 | path = btrfs_alloc_path(); | |
2437 | if (!path) | |
2438 | return -ENOMEM; | |
2439 | ||
c8bcbfbd | 2440 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; |
98d377a0 | 2441 | |
56e9357a | 2442 | root = btrfs_get_fs_root(info, tree_id, true); |
98d377a0 | 2443 | if (IS_ERR(root)) { |
ad1e3d56 | 2444 | ret = PTR_ERR(root); |
88234012 JB |
2445 | root = NULL; |
2446 | goto out; | |
2447 | } | |
98d377a0 TH |
2448 | |
2449 | key.objectid = dirid; | |
2450 | key.type = BTRFS_INODE_REF_KEY; | |
8ad6fcab | 2451 | key.offset = (u64)-1; |
98d377a0 | 2452 | |
67871254 | 2453 | while (1) { |
0ff40a91 | 2454 | ret = btrfs_search_backwards(root, &key, path); |
98d377a0 TH |
2455 | if (ret < 0) |
2456 | goto out; | |
18674c6c | 2457 | else if (ret > 0) { |
0ff40a91 MPS |
2458 | ret = -ENOENT; |
2459 | goto out; | |
18674c6c | 2460 | } |
98d377a0 TH |
2461 | |
2462 | l = path->nodes[0]; | |
2463 | slot = path->slots[0]; | |
98d377a0 | 2464 | |
98d377a0 TH |
2465 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); |
2466 | len = btrfs_inode_ref_name_len(l, iref); | |
2467 | ptr -= len + 1; | |
2468 | total_len += len + 1; | |
a696cf35 FDBM |
2469 | if (ptr < name) { |
2470 | ret = -ENAMETOOLONG; | |
98d377a0 | 2471 | goto out; |
a696cf35 | 2472 | } |
98d377a0 TH |
2473 | |
2474 | *(ptr + len) = '/'; | |
67871254 | 2475 | read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); |
98d377a0 TH |
2476 | |
2477 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
2478 | break; | |
2479 | ||
b3b4aa74 | 2480 | btrfs_release_path(path); |
98d377a0 | 2481 | key.objectid = key.offset; |
8ad6fcab | 2482 | key.offset = (u64)-1; |
98d377a0 | 2483 | dirid = key.objectid; |
98d377a0 | 2484 | } |
77906a50 | 2485 | memmove(name, ptr, total_len); |
67871254 | 2486 | name[total_len] = '\0'; |
98d377a0 TH |
2487 | ret = 0; |
2488 | out: | |
00246528 | 2489 | btrfs_put_root(root); |
98d377a0 | 2490 | btrfs_free_path(path); |
ac8e9819 CM |
2491 | return ret; |
2492 | } | |
2493 | ||
6623d9a0 CB |
2494 | static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns, |
2495 | struct inode *inode, | |
23d0b79d TM |
2496 | struct btrfs_ioctl_ino_lookup_user_args *args) |
2497 | { | |
2498 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; | |
2499 | struct super_block *sb = inode->i_sb; | |
2500 | struct btrfs_key upper_limit = BTRFS_I(inode)->location; | |
2501 | u64 treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2502 | u64 dirid = args->dirid; | |
2503 | unsigned long item_off; | |
2504 | unsigned long item_len; | |
2505 | struct btrfs_inode_ref *iref; | |
2506 | struct btrfs_root_ref *rref; | |
b8a49ae1 | 2507 | struct btrfs_root *root = NULL; |
23d0b79d TM |
2508 | struct btrfs_path *path; |
2509 | struct btrfs_key key, key2; | |
2510 | struct extent_buffer *leaf; | |
2511 | struct inode *temp_inode; | |
2512 | char *ptr; | |
2513 | int slot; | |
2514 | int len; | |
2515 | int total_len = 0; | |
2516 | int ret; | |
2517 | ||
2518 | path = btrfs_alloc_path(); | |
2519 | if (!path) | |
2520 | return -ENOMEM; | |
2521 | ||
2522 | /* | |
2523 | * If the bottom subvolume does not exist directly under upper_limit, | |
2524 | * construct the path in from the bottom up. | |
2525 | */ | |
2526 | if (dirid != upper_limit.objectid) { | |
2527 | ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; | |
2528 | ||
56e9357a | 2529 | root = btrfs_get_fs_root(fs_info, treeid, true); |
23d0b79d TM |
2530 | if (IS_ERR(root)) { |
2531 | ret = PTR_ERR(root); | |
2532 | goto out; | |
2533 | } | |
2534 | ||
2535 | key.objectid = dirid; | |
2536 | key.type = BTRFS_INODE_REF_KEY; | |
2537 | key.offset = (u64)-1; | |
2538 | while (1) { | |
0ff40a91 MPS |
2539 | ret = btrfs_search_backwards(root, &key, path); |
2540 | if (ret < 0) | |
2541 | goto out_put; | |
2542 | else if (ret > 0) { | |
2543 | ret = -ENOENT; | |
b8a49ae1 | 2544 | goto out_put; |
23d0b79d TM |
2545 | } |
2546 | ||
2547 | leaf = path->nodes[0]; | |
2548 | slot = path->slots[0]; | |
23d0b79d TM |
2549 | |
2550 | iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); | |
2551 | len = btrfs_inode_ref_name_len(leaf, iref); | |
2552 | ptr -= len + 1; | |
2553 | total_len += len + 1; | |
2554 | if (ptr < args->path) { | |
2555 | ret = -ENAMETOOLONG; | |
b8a49ae1 | 2556 | goto out_put; |
23d0b79d TM |
2557 | } |
2558 | ||
2559 | *(ptr + len) = '/'; | |
2560 | read_extent_buffer(leaf, ptr, | |
2561 | (unsigned long)(iref + 1), len); | |
2562 | ||
2563 | /* Check the read+exec permission of this directory */ | |
2564 | ret = btrfs_previous_item(root, path, dirid, | |
2565 | BTRFS_INODE_ITEM_KEY); | |
2566 | if (ret < 0) { | |
b8a49ae1 | 2567 | goto out_put; |
23d0b79d TM |
2568 | } else if (ret > 0) { |
2569 | ret = -ENOENT; | |
b8a49ae1 | 2570 | goto out_put; |
23d0b79d TM |
2571 | } |
2572 | ||
2573 | leaf = path->nodes[0]; | |
2574 | slot = path->slots[0]; | |
2575 | btrfs_item_key_to_cpu(leaf, &key2, slot); | |
2576 | if (key2.objectid != dirid) { | |
2577 | ret = -ENOENT; | |
b8a49ae1 | 2578 | goto out_put; |
23d0b79d TM |
2579 | } |
2580 | ||
0202e83f | 2581 | temp_inode = btrfs_iget(sb, key2.objectid, root); |
3ca57bd6 MT |
2582 | if (IS_ERR(temp_inode)) { |
2583 | ret = PTR_ERR(temp_inode); | |
b8a49ae1 | 2584 | goto out_put; |
3ca57bd6 | 2585 | } |
6623d9a0 | 2586 | ret = inode_permission(mnt_userns, temp_inode, |
47291baa | 2587 | MAY_READ | MAY_EXEC); |
23d0b79d TM |
2588 | iput(temp_inode); |
2589 | if (ret) { | |
2590 | ret = -EACCES; | |
b8a49ae1 | 2591 | goto out_put; |
23d0b79d TM |
2592 | } |
2593 | ||
2594 | if (key.offset == upper_limit.objectid) | |
2595 | break; | |
2596 | if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { | |
2597 | ret = -EACCES; | |
b8a49ae1 | 2598 | goto out_put; |
23d0b79d TM |
2599 | } |
2600 | ||
2601 | btrfs_release_path(path); | |
2602 | key.objectid = key.offset; | |
2603 | key.offset = (u64)-1; | |
2604 | dirid = key.objectid; | |
2605 | } | |
2606 | ||
2607 | memmove(args->path, ptr, total_len); | |
2608 | args->path[total_len] = '\0'; | |
00246528 | 2609 | btrfs_put_root(root); |
b8a49ae1 | 2610 | root = NULL; |
23d0b79d TM |
2611 | btrfs_release_path(path); |
2612 | } | |
2613 | ||
2614 | /* Get the bottom subvolume's name from ROOT_REF */ | |
23d0b79d TM |
2615 | key.objectid = treeid; |
2616 | key.type = BTRFS_ROOT_REF_KEY; | |
2617 | key.offset = args->treeid; | |
b8a49ae1 | 2618 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); |
23d0b79d TM |
2619 | if (ret < 0) { |
2620 | goto out; | |
2621 | } else if (ret > 0) { | |
2622 | ret = -ENOENT; | |
2623 | goto out; | |
2624 | } | |
2625 | ||
2626 | leaf = path->nodes[0]; | |
2627 | slot = path->slots[0]; | |
2628 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
2629 | ||
2630 | item_off = btrfs_item_ptr_offset(leaf, slot); | |
3212fa14 | 2631 | item_len = btrfs_item_size(leaf, slot); |
23d0b79d TM |
2632 | /* Check if dirid in ROOT_REF corresponds to passed dirid */ |
2633 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
2634 | if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { | |
2635 | ret = -EINVAL; | |
2636 | goto out; | |
2637 | } | |
2638 | ||
2639 | /* Copy subvolume's name */ | |
2640 | item_off += sizeof(struct btrfs_root_ref); | |
2641 | item_len -= sizeof(struct btrfs_root_ref); | |
2642 | read_extent_buffer(leaf, args->name, item_off, item_len); | |
2643 | args->name[item_len] = 0; | |
2644 | ||
b8a49ae1 | 2645 | out_put: |
00246528 | 2646 | btrfs_put_root(root); |
23d0b79d TM |
2647 | out: |
2648 | btrfs_free_path(path); | |
2649 | return ret; | |
2650 | } | |
2651 | ||
ac8e9819 CM |
2652 | static noinline int btrfs_ioctl_ino_lookup(struct file *file, |
2653 | void __user *argp) | |
2654 | { | |
bece2e82 BVA |
2655 | struct btrfs_ioctl_ino_lookup_args *args; |
2656 | struct inode *inode; | |
01b810b8 | 2657 | int ret = 0; |
ac8e9819 | 2658 | |
2354d08f JL |
2659 | args = memdup_user(argp, sizeof(*args)); |
2660 | if (IS_ERR(args)) | |
2661 | return PTR_ERR(args); | |
c2b96929 | 2662 | |
496ad9aa | 2663 | inode = file_inode(file); |
ac8e9819 | 2664 | |
01b810b8 DS |
2665 | /* |
2666 | * Unprivileged query to obtain the containing subvolume root id. The | |
2667 | * path is reset so it's consistent with btrfs_search_path_in_tree. | |
2668 | */ | |
1b53ac4d CM |
2669 | if (args->treeid == 0) |
2670 | args->treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2671 | ||
01b810b8 DS |
2672 | if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { |
2673 | args->name[0] = 0; | |
2674 | goto out; | |
2675 | } | |
2676 | ||
2677 | if (!capable(CAP_SYS_ADMIN)) { | |
2678 | ret = -EPERM; | |
2679 | goto out; | |
2680 | } | |
2681 | ||
ac8e9819 CM |
2682 | ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, |
2683 | args->treeid, args->objectid, | |
2684 | args->name); | |
2685 | ||
01b810b8 | 2686 | out: |
ac8e9819 CM |
2687 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) |
2688 | ret = -EFAULT; | |
2689 | ||
2690 | kfree(args); | |
98d377a0 TH |
2691 | return ret; |
2692 | } | |
2693 | ||
23d0b79d TM |
2694 | /* |
2695 | * Version of ino_lookup ioctl (unprivileged) | |
2696 | * | |
2697 | * The main differences from ino_lookup ioctl are: | |
2698 | * | |
2699 | * 1. Read + Exec permission will be checked using inode_permission() during | |
2700 | * path construction. -EACCES will be returned in case of failure. | |
2701 | * 2. Path construction will be stopped at the inode number which corresponds | |
2702 | * to the fd with which this ioctl is called. If constructed path does not | |
2703 | * exist under fd's inode, -EACCES will be returned. | |
2704 | * 3. The name of bottom subvolume is also searched and filled. | |
2705 | */ | |
2706 | static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) | |
2707 | { | |
2708 | struct btrfs_ioctl_ino_lookup_user_args *args; | |
2709 | struct inode *inode; | |
2710 | int ret; | |
2711 | ||
2712 | args = memdup_user(argp, sizeof(*args)); | |
2713 | if (IS_ERR(args)) | |
2714 | return PTR_ERR(args); | |
2715 | ||
2716 | inode = file_inode(file); | |
2717 | ||
2718 | if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && | |
2719 | BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { | |
2720 | /* | |
2721 | * The subvolume does not exist under fd with which this is | |
2722 | * called | |
2723 | */ | |
2724 | kfree(args); | |
2725 | return -EACCES; | |
2726 | } | |
2727 | ||
6623d9a0 | 2728 | ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args); |
23d0b79d TM |
2729 | |
2730 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
2731 | ret = -EFAULT; | |
2732 | ||
2733 | kfree(args); | |
2734 | return ret; | |
2735 | } | |
2736 | ||
b64ec075 TM |
2737 | /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ |
2738 | static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) | |
2739 | { | |
2740 | struct btrfs_ioctl_get_subvol_info_args *subvol_info; | |
2741 | struct btrfs_fs_info *fs_info; | |
2742 | struct btrfs_root *root; | |
2743 | struct btrfs_path *path; | |
2744 | struct btrfs_key key; | |
2745 | struct btrfs_root_item *root_item; | |
2746 | struct btrfs_root_ref *rref; | |
2747 | struct extent_buffer *leaf; | |
2748 | unsigned long item_off; | |
2749 | unsigned long item_len; | |
2750 | struct inode *inode; | |
2751 | int slot; | |
2752 | int ret = 0; | |
2753 | ||
2754 | path = btrfs_alloc_path(); | |
2755 | if (!path) | |
2756 | return -ENOMEM; | |
2757 | ||
2758 | subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); | |
2759 | if (!subvol_info) { | |
2760 | btrfs_free_path(path); | |
2761 | return -ENOMEM; | |
2762 | } | |
2763 | ||
2764 | inode = file_inode(file); | |
2765 | fs_info = BTRFS_I(inode)->root->fs_info; | |
2766 | ||
2767 | /* Get root_item of inode's subvolume */ | |
2768 | key.objectid = BTRFS_I(inode)->root->root_key.objectid; | |
56e9357a | 2769 | root = btrfs_get_fs_root(fs_info, key.objectid, true); |
b64ec075 TM |
2770 | if (IS_ERR(root)) { |
2771 | ret = PTR_ERR(root); | |
04734e84 JB |
2772 | goto out_free; |
2773 | } | |
b64ec075 TM |
2774 | root_item = &root->root_item; |
2775 | ||
2776 | subvol_info->treeid = key.objectid; | |
2777 | ||
2778 | subvol_info->generation = btrfs_root_generation(root_item); | |
2779 | subvol_info->flags = btrfs_root_flags(root_item); | |
2780 | ||
2781 | memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); | |
2782 | memcpy(subvol_info->parent_uuid, root_item->parent_uuid, | |
2783 | BTRFS_UUID_SIZE); | |
2784 | memcpy(subvol_info->received_uuid, root_item->received_uuid, | |
2785 | BTRFS_UUID_SIZE); | |
2786 | ||
2787 | subvol_info->ctransid = btrfs_root_ctransid(root_item); | |
2788 | subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); | |
2789 | subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); | |
2790 | ||
2791 | subvol_info->otransid = btrfs_root_otransid(root_item); | |
2792 | subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); | |
2793 | subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); | |
2794 | ||
2795 | subvol_info->stransid = btrfs_root_stransid(root_item); | |
2796 | subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); | |
2797 | subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); | |
2798 | ||
2799 | subvol_info->rtransid = btrfs_root_rtransid(root_item); | |
2800 | subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); | |
2801 | subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); | |
2802 | ||
2803 | if (key.objectid != BTRFS_FS_TREE_OBJECTID) { | |
2804 | /* Search root tree for ROOT_BACKREF of this subvolume */ | |
b64ec075 TM |
2805 | key.type = BTRFS_ROOT_BACKREF_KEY; |
2806 | key.offset = 0; | |
04734e84 | 2807 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); |
b64ec075 TM |
2808 | if (ret < 0) { |
2809 | goto out; | |
2810 | } else if (path->slots[0] >= | |
2811 | btrfs_header_nritems(path->nodes[0])) { | |
04734e84 | 2812 | ret = btrfs_next_leaf(fs_info->tree_root, path); |
b64ec075 TM |
2813 | if (ret < 0) { |
2814 | goto out; | |
2815 | } else if (ret > 0) { | |
2816 | ret = -EUCLEAN; | |
2817 | goto out; | |
2818 | } | |
2819 | } | |
2820 | ||
2821 | leaf = path->nodes[0]; | |
2822 | slot = path->slots[0]; | |
2823 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
2824 | if (key.objectid == subvol_info->treeid && | |
2825 | key.type == BTRFS_ROOT_BACKREF_KEY) { | |
2826 | subvol_info->parent_id = key.offset; | |
2827 | ||
2828 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
2829 | subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); | |
2830 | ||
2831 | item_off = btrfs_item_ptr_offset(leaf, slot) | |
2832 | + sizeof(struct btrfs_root_ref); | |
3212fa14 | 2833 | item_len = btrfs_item_size(leaf, slot) |
b64ec075 TM |
2834 | - sizeof(struct btrfs_root_ref); |
2835 | read_extent_buffer(leaf, subvol_info->name, | |
2836 | item_off, item_len); | |
2837 | } else { | |
2838 | ret = -ENOENT; | |
2839 | goto out; | |
2840 | } | |
2841 | } | |
2842 | ||
2843 | if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) | |
2844 | ret = -EFAULT; | |
2845 | ||
2846 | out: | |
00246528 | 2847 | btrfs_put_root(root); |
04734e84 | 2848 | out_free: |
b64ec075 | 2849 | btrfs_free_path(path); |
b091f7fe | 2850 | kfree(subvol_info); |
b64ec075 TM |
2851 | return ret; |
2852 | } | |
2853 | ||
42e4b520 TM |
2854 | /* |
2855 | * Return ROOT_REF information of the subvolume containing this inode | |
2856 | * except the subvolume name. | |
2857 | */ | |
2858 | static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) | |
2859 | { | |
2860 | struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; | |
2861 | struct btrfs_root_ref *rref; | |
2862 | struct btrfs_root *root; | |
2863 | struct btrfs_path *path; | |
2864 | struct btrfs_key key; | |
2865 | struct extent_buffer *leaf; | |
2866 | struct inode *inode; | |
2867 | u64 objectid; | |
2868 | int slot; | |
2869 | int ret; | |
2870 | u8 found; | |
2871 | ||
2872 | path = btrfs_alloc_path(); | |
2873 | if (!path) | |
2874 | return -ENOMEM; | |
2875 | ||
2876 | rootrefs = memdup_user(argp, sizeof(*rootrefs)); | |
2877 | if (IS_ERR(rootrefs)) { | |
2878 | btrfs_free_path(path); | |
2879 | return PTR_ERR(rootrefs); | |
2880 | } | |
2881 | ||
2882 | inode = file_inode(file); | |
2883 | root = BTRFS_I(inode)->root->fs_info->tree_root; | |
2884 | objectid = BTRFS_I(inode)->root->root_key.objectid; | |
2885 | ||
2886 | key.objectid = objectid; | |
2887 | key.type = BTRFS_ROOT_REF_KEY; | |
2888 | key.offset = rootrefs->min_treeid; | |
2889 | found = 0; | |
2890 | ||
2891 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
2892 | if (ret < 0) { | |
2893 | goto out; | |
2894 | } else if (path->slots[0] >= | |
2895 | btrfs_header_nritems(path->nodes[0])) { | |
2896 | ret = btrfs_next_leaf(root, path); | |
2897 | if (ret < 0) { | |
2898 | goto out; | |
2899 | } else if (ret > 0) { | |
2900 | ret = -EUCLEAN; | |
2901 | goto out; | |
2902 | } | |
2903 | } | |
2904 | while (1) { | |
2905 | leaf = path->nodes[0]; | |
2906 | slot = path->slots[0]; | |
2907 | ||
2908 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
2909 | if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { | |
2910 | ret = 0; | |
2911 | goto out; | |
2912 | } | |
2913 | ||
2914 | if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { | |
2915 | ret = -EOVERFLOW; | |
2916 | goto out; | |
2917 | } | |
2918 | ||
2919 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
2920 | rootrefs->rootref[found].treeid = key.offset; | |
2921 | rootrefs->rootref[found].dirid = | |
2922 | btrfs_root_ref_dirid(leaf, rref); | |
2923 | found++; | |
2924 | ||
2925 | ret = btrfs_next_item(root, path); | |
2926 | if (ret < 0) { | |
2927 | goto out; | |
2928 | } else if (ret > 0) { | |
2929 | ret = -EUCLEAN; | |
2930 | goto out; | |
2931 | } | |
2932 | } | |
2933 | ||
2934 | out: | |
2935 | if (!ret || ret == -EOVERFLOW) { | |
2936 | rootrefs->num_items = found; | |
2937 | /* update min_treeid for next search */ | |
2938 | if (found) | |
2939 | rootrefs->min_treeid = | |
2940 | rootrefs->rootref[found - 1].treeid + 1; | |
2941 | if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) | |
2942 | ret = -EFAULT; | |
2943 | } | |
2944 | ||
2945 | kfree(rootrefs); | |
2946 | btrfs_free_path(path); | |
2947 | ||
2948 | return ret; | |
2949 | } | |
2950 | ||
76dda93c | 2951 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
949964c9 MPS |
2952 | void __user *arg, |
2953 | bool destroy_v2) | |
76dda93c | 2954 | { |
54563d41 | 2955 | struct dentry *parent = file->f_path.dentry; |
0b246afa | 2956 | struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); |
76dda93c | 2957 | struct dentry *dentry; |
2b0143b5 | 2958 | struct inode *dir = d_inode(parent); |
76dda93c YZ |
2959 | struct inode *inode; |
2960 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
2961 | struct btrfs_root *dest = NULL; | |
949964c9 MPS |
2962 | struct btrfs_ioctl_vol_args *vol_args = NULL; |
2963 | struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL; | |
c4ed533b | 2964 | struct user_namespace *mnt_userns = file_mnt_user_ns(file); |
949964c9 MPS |
2965 | char *subvol_name, *subvol_name_ptr = NULL; |
2966 | int subvol_namelen; | |
76dda93c | 2967 | int err = 0; |
949964c9 | 2968 | bool destroy_parent = false; |
76dda93c | 2969 | |
949964c9 MPS |
2970 | if (destroy_v2) { |
2971 | vol_args2 = memdup_user(arg, sizeof(*vol_args2)); | |
2972 | if (IS_ERR(vol_args2)) | |
2973 | return PTR_ERR(vol_args2); | |
325c50e3 | 2974 | |
949964c9 MPS |
2975 | if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) { |
2976 | err = -EOPNOTSUPP; | |
2977 | goto out; | |
2978 | } | |
76dda93c | 2979 | |
949964c9 MPS |
2980 | /* |
2981 | * If SPEC_BY_ID is not set, we are looking for the subvolume by | |
2982 | * name, same as v1 currently does. | |
2983 | */ | |
2984 | if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { | |
2985 | vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0; | |
2986 | subvol_name = vol_args2->name; | |
2987 | ||
2988 | err = mnt_want_write_file(file); | |
2989 | if (err) | |
2990 | goto out; | |
2991 | } else { | |
aabb34e7 | 2992 | struct inode *old_dir; |
c4ed533b | 2993 | |
949964c9 MPS |
2994 | if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) { |
2995 | err = -EINVAL; | |
2996 | goto out; | |
2997 | } | |
2998 | ||
2999 | err = mnt_want_write_file(file); | |
3000 | if (err) | |
3001 | goto out; | |
3002 | ||
3003 | dentry = btrfs_get_dentry(fs_info->sb, | |
3004 | BTRFS_FIRST_FREE_OBJECTID, | |
3005 | vol_args2->subvolid, 0, 0); | |
3006 | if (IS_ERR(dentry)) { | |
3007 | err = PTR_ERR(dentry); | |
3008 | goto out_drop_write; | |
3009 | } | |
3010 | ||
3011 | /* | |
3012 | * Change the default parent since the subvolume being | |
3013 | * deleted can be outside of the current mount point. | |
3014 | */ | |
3015 | parent = btrfs_get_parent(dentry); | |
3016 | ||
3017 | /* | |
3018 | * At this point dentry->d_name can point to '/' if the | |
3019 | * subvolume we want to destroy is outsite of the | |
3020 | * current mount point, so we need to release the | |
3021 | * current dentry and execute the lookup to return a new | |
3022 | * one with ->d_name pointing to the | |
3023 | * <mount point>/subvol_name. | |
3024 | */ | |
3025 | dput(dentry); | |
3026 | if (IS_ERR(parent)) { | |
3027 | err = PTR_ERR(parent); | |
3028 | goto out_drop_write; | |
3029 | } | |
aabb34e7 | 3030 | old_dir = dir; |
949964c9 MPS |
3031 | dir = d_inode(parent); |
3032 | ||
3033 | /* | |
3034 | * If v2 was used with SPEC_BY_ID, a new parent was | |
3035 | * allocated since the subvolume can be outside of the | |
3036 | * current mount point. Later on we need to release this | |
3037 | * new parent dentry. | |
3038 | */ | |
3039 | destroy_parent = true; | |
3040 | ||
aabb34e7 CB |
3041 | /* |
3042 | * On idmapped mounts, deletion via subvolid is | |
3043 | * restricted to subvolumes that are immediate | |
3044 | * ancestors of the inode referenced by the file | |
3045 | * descriptor in the ioctl. Otherwise the idmapping | |
3046 | * could potentially be abused to delete subvolumes | |
3047 | * anywhere in the filesystem the user wouldn't be able | |
3048 | * to delete without an idmapped mount. | |
3049 | */ | |
3050 | if (old_dir != dir && mnt_userns != &init_user_ns) { | |
3051 | err = -EOPNOTSUPP; | |
3052 | goto free_parent; | |
3053 | } | |
3054 | ||
949964c9 MPS |
3055 | subvol_name_ptr = btrfs_get_subvol_name_from_objectid( |
3056 | fs_info, vol_args2->subvolid); | |
3057 | if (IS_ERR(subvol_name_ptr)) { | |
3058 | err = PTR_ERR(subvol_name_ptr); | |
3059 | goto free_parent; | |
3060 | } | |
1a9fd417 | 3061 | /* subvol_name_ptr is already nul terminated */ |
949964c9 MPS |
3062 | subvol_name = (char *)kbasename(subvol_name_ptr); |
3063 | } | |
3064 | } else { | |
3065 | vol_args = memdup_user(arg, sizeof(*vol_args)); | |
3066 | if (IS_ERR(vol_args)) | |
3067 | return PTR_ERR(vol_args); | |
3068 | ||
3069 | vol_args->name[BTRFS_PATH_NAME_MAX] = 0; | |
3070 | subvol_name = vol_args->name; | |
3071 | ||
3072 | err = mnt_want_write_file(file); | |
3073 | if (err) | |
3074 | goto out; | |
76dda93c YZ |
3075 | } |
3076 | ||
949964c9 | 3077 | subvol_namelen = strlen(subvol_name); |
76dda93c | 3078 | |
949964c9 MPS |
3079 | if (strchr(subvol_name, '/') || |
3080 | strncmp(subvol_name, "..", subvol_namelen) == 0) { | |
3081 | err = -EINVAL; | |
3082 | goto free_subvol_name; | |
3083 | } | |
3084 | ||
3085 | if (!S_ISDIR(dir->i_mode)) { | |
3086 | err = -ENOTDIR; | |
3087 | goto free_subvol_name; | |
3088 | } | |
521e0546 | 3089 | |
00235411 AV |
3090 | err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); |
3091 | if (err == -EINTR) | |
949964c9 | 3092 | goto free_subvol_name; |
c4ed533b | 3093 | dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen); |
76dda93c YZ |
3094 | if (IS_ERR(dentry)) { |
3095 | err = PTR_ERR(dentry); | |
3096 | goto out_unlock_dir; | |
3097 | } | |
3098 | ||
2b0143b5 | 3099 | if (d_really_is_negative(dentry)) { |
76dda93c YZ |
3100 | err = -ENOENT; |
3101 | goto out_dput; | |
3102 | } | |
3103 | ||
2b0143b5 | 3104 | inode = d_inode(dentry); |
4260f7c7 | 3105 | dest = BTRFS_I(inode)->root; |
67871254 | 3106 | if (!capable(CAP_SYS_ADMIN)) { |
4260f7c7 SW |
3107 | /* |
3108 | * Regular user. Only allow this with a special mount | |
3109 | * option, when the user has write+exec access to the | |
3110 | * subvol root, and when rmdir(2) would have been | |
3111 | * allowed. | |
3112 | * | |
3113 | * Note that this is _not_ check that the subvol is | |
3114 | * empty or doesn't contain data that we wouldn't | |
3115 | * otherwise be able to delete. | |
3116 | * | |
3117 | * Users who want to delete empty subvols should try | |
3118 | * rmdir(2). | |
3119 | */ | |
3120 | err = -EPERM; | |
0b246afa | 3121 | if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) |
4260f7c7 SW |
3122 | goto out_dput; |
3123 | ||
3124 | /* | |
3125 | * Do not allow deletion if the parent dir is the same | |
3126 | * as the dir to be deleted. That means the ioctl | |
3127 | * must be called on the dentry referencing the root | |
3128 | * of the subvol, not a random directory contained | |
3129 | * within it. | |
3130 | */ | |
3131 | err = -EINVAL; | |
3132 | if (root == dest) | |
3133 | goto out_dput; | |
3134 | ||
c4ed533b | 3135 | err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC); |
4260f7c7 SW |
3136 | if (err) |
3137 | goto out_dput; | |
4260f7c7 SW |
3138 | } |
3139 | ||
5c39da5b | 3140 | /* check if subvolume may be deleted by a user */ |
c4ed533b | 3141 | err = btrfs_may_delete(mnt_userns, dir, dentry, 1); |
5c39da5b MX |
3142 | if (err) |
3143 | goto out_dput; | |
3144 | ||
4a0cc7ca | 3145 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
76dda93c YZ |
3146 | err = -EINVAL; |
3147 | goto out_dput; | |
3148 | } | |
3149 | ||
64708539 | 3150 | btrfs_inode_lock(inode, 0); |
f60a2364 | 3151 | err = btrfs_delete_subvolume(dir, dentry); |
64708539 | 3152 | btrfs_inode_unlock(inode, 0); |
46008d9d AG |
3153 | if (!err) { |
3154 | fsnotify_rmdir(dir, dentry); | |
76dda93c | 3155 | d_delete(dentry); |
46008d9d | 3156 | } |
fa6ac876 | 3157 | |
76dda93c YZ |
3158 | out_dput: |
3159 | dput(dentry); | |
3160 | out_unlock_dir: | |
64708539 | 3161 | btrfs_inode_unlock(dir, 0); |
949964c9 MPS |
3162 | free_subvol_name: |
3163 | kfree(subvol_name_ptr); | |
3164 | free_parent: | |
3165 | if (destroy_parent) | |
3166 | dput(parent); | |
00235411 | 3167 | out_drop_write: |
2a79f17e | 3168 | mnt_drop_write_file(file); |
76dda93c | 3169 | out: |
949964c9 | 3170 | kfree(vol_args2); |
76dda93c YZ |
3171 | kfree(vol_args); |
3172 | return err; | |
3173 | } | |
3174 | ||
1e701a32 | 3175 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 | 3176 | { |
496ad9aa | 3177 | struct inode *inode = file_inode(file); |
f46b5a66 | 3178 | struct btrfs_root *root = BTRFS_I(inode)->root; |
c853a578 | 3179 | struct btrfs_ioctl_defrag_range_args range = {0}; |
c146afad YZ |
3180 | int ret; |
3181 | ||
25122d15 ID |
3182 | ret = mnt_want_write_file(file); |
3183 | if (ret) | |
3184 | return ret; | |
b83cc969 | 3185 | |
25122d15 ID |
3186 | if (btrfs_root_readonly(root)) { |
3187 | ret = -EROFS; | |
3188 | goto out; | |
5ac00add | 3189 | } |
f46b5a66 CH |
3190 | |
3191 | switch (inode->i_mode & S_IFMT) { | |
3192 | case S_IFDIR: | |
e441d54d CM |
3193 | if (!capable(CAP_SYS_ADMIN)) { |
3194 | ret = -EPERM; | |
3195 | goto out; | |
3196 | } | |
de78b51a | 3197 | ret = btrfs_defrag_root(root); |
f46b5a66 CH |
3198 | break; |
3199 | case S_IFREG: | |
616d374e AB |
3200 | /* |
3201 | * Note that this does not check the file descriptor for write | |
3202 | * access. This prevents defragmenting executables that are | |
3203 | * running and allows defrag on files open in read-only mode. | |
3204 | */ | |
3205 | if (!capable(CAP_SYS_ADMIN) && | |
47291baa | 3206 | inode_permission(&init_user_ns, inode, MAY_WRITE)) { |
616d374e | 3207 | ret = -EPERM; |
e441d54d CM |
3208 | goto out; |
3209 | } | |
1e701a32 | 3210 | |
1e701a32 | 3211 | if (argp) { |
c853a578 | 3212 | if (copy_from_user(&range, argp, sizeof(range))) { |
1e701a32 | 3213 | ret = -EFAULT; |
683be16e | 3214 | goto out; |
1e701a32 CM |
3215 | } |
3216 | /* compression requires us to start the IO */ | |
c853a578 GR |
3217 | if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { |
3218 | range.flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
3219 | range.extent_thresh = (u32)-1; | |
1e701a32 CM |
3220 | } |
3221 | } else { | |
3222 | /* the rest are all set to zero by kzalloc */ | |
c853a578 | 3223 | range.len = (u64)-1; |
1e701a32 | 3224 | } |
1ccc2e8a | 3225 | ret = btrfs_defrag_file(file_inode(file), &file->f_ra, |
c853a578 | 3226 | &range, BTRFS_OLDEST_GENERATION, 0); |
4cb5300b CM |
3227 | if (ret > 0) |
3228 | ret = 0; | |
f46b5a66 | 3229 | break; |
8929ecfa YZ |
3230 | default: |
3231 | ret = -EINVAL; | |
f46b5a66 | 3232 | } |
e441d54d | 3233 | out: |
25122d15 | 3234 | mnt_drop_write_file(file); |
e441d54d | 3235 | return ret; |
f46b5a66 CH |
3236 | } |
3237 | ||
2ff7e61e | 3238 | static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) |
f46b5a66 CH |
3239 | { |
3240 | struct btrfs_ioctl_vol_args *vol_args; | |
a174c0a2 | 3241 | bool restore_op = false; |
f46b5a66 CH |
3242 | int ret; |
3243 | ||
e441d54d CM |
3244 | if (!capable(CAP_SYS_ADMIN)) |
3245 | return -EPERM; | |
3246 | ||
a174c0a2 NB |
3247 | if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) { |
3248 | if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD)) | |
3249 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; | |
3250 | ||
3251 | /* | |
3252 | * We can do the device add because we have a paused balanced, | |
3253 | * change the exclusive op type and remember we should bring | |
3254 | * back the paused balance | |
3255 | */ | |
3256 | fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD; | |
3257 | btrfs_exclop_start_unlock(fs_info); | |
3258 | restore_op = true; | |
3259 | } | |
c9e9f97b | 3260 | |
dae7b665 | 3261 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
3262 | if (IS_ERR(vol_args)) { |
3263 | ret = PTR_ERR(vol_args); | |
3264 | goto out; | |
3265 | } | |
f46b5a66 | 3266 | |
5516e595 | 3267 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
2ff7e61e | 3268 | ret = btrfs_init_new_device(fs_info, vol_args->name); |
f46b5a66 | 3269 | |
43d20761 | 3270 | if (!ret) |
0b246afa | 3271 | btrfs_info(fs_info, "disk added %s", vol_args->name); |
43d20761 | 3272 | |
f46b5a66 | 3273 | kfree(vol_args); |
c9e9f97b | 3274 | out: |
a174c0a2 NB |
3275 | if (restore_op) |
3276 | btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED); | |
3277 | else | |
3278 | btrfs_exclop_finish(fs_info); | |
f46b5a66 CH |
3279 | return ret; |
3280 | } | |
3281 | ||
6b526ed7 | 3282 | static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) |
f46b5a66 | 3283 | { |
1a15eb72 | 3284 | BTRFS_DEV_LOOKUP_ARGS(args); |
0b246afa JM |
3285 | struct inode *inode = file_inode(file); |
3286 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
6b526ed7 | 3287 | struct btrfs_ioctl_vol_args_v2 *vol_args; |
3fa421de JB |
3288 | struct block_device *bdev = NULL; |
3289 | fmode_t mode; | |
f46b5a66 | 3290 | int ret; |
67ae34b6 | 3291 | bool cancel = false; |
f46b5a66 | 3292 | |
e441d54d CM |
3293 | if (!capable(CAP_SYS_ADMIN)) |
3294 | return -EPERM; | |
3295 | ||
dae7b665 | 3296 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
d815b3f2 DC |
3297 | if (IS_ERR(vol_args)) |
3298 | return PTR_ERR(vol_args); | |
f46b5a66 | 3299 | |
748449cd | 3300 | if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { |
fd4e994b OS |
3301 | ret = -EOPNOTSUPP; |
3302 | goto out; | |
3303 | } | |
1a15eb72 | 3304 | |
67ae34b6 | 3305 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; |
1a15eb72 JB |
3306 | if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { |
3307 | args.devid = vol_args->devid; | |
3308 | } else if (!strcmp("cancel", vol_args->name)) { | |
67ae34b6 | 3309 | cancel = true; |
1a15eb72 JB |
3310 | } else { |
3311 | ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); | |
3312 | if (ret) | |
3313 | goto out; | |
3314 | } | |
3315 | ||
3316 | ret = mnt_want_write_file(file); | |
3317 | if (ret) | |
3318 | goto out; | |
f46b5a66 | 3319 | |
67ae34b6 DS |
3320 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, |
3321 | cancel); | |
3322 | if (ret) | |
1a15eb72 | 3323 | goto err_drop; |
183860f6 | 3324 | |
1a15eb72 JB |
3325 | /* Exclusive operation is now claimed */ |
3326 | ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); | |
67ae34b6 | 3327 | |
c3e1f96c | 3328 | btrfs_exclop_finish(fs_info); |
183860f6 | 3329 | |
6b526ed7 | 3330 | if (!ret) { |
735654ea | 3331 | if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) |
0b246afa | 3332 | btrfs_info(fs_info, "device deleted: id %llu", |
6b526ed7 AJ |
3333 | vol_args->devid); |
3334 | else | |
0b246afa | 3335 | btrfs_info(fs_info, "device deleted: %s", |
6b526ed7 AJ |
3336 | vol_args->name); |
3337 | } | |
c47ca32d | 3338 | err_drop: |
4ac20c70 | 3339 | mnt_drop_write_file(file); |
3fa421de JB |
3340 | if (bdev) |
3341 | blkdev_put(bdev, mode); | |
1a15eb72 JB |
3342 | out: |
3343 | btrfs_put_dev_args_from_path(&args); | |
3344 | kfree(vol_args); | |
f46b5a66 CH |
3345 | return ret; |
3346 | } | |
3347 | ||
da24927b | 3348 | static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) |
f46b5a66 | 3349 | { |
1a15eb72 | 3350 | BTRFS_DEV_LOOKUP_ARGS(args); |
0b246afa JM |
3351 | struct inode *inode = file_inode(file); |
3352 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f46b5a66 | 3353 | struct btrfs_ioctl_vol_args *vol_args; |
3fa421de JB |
3354 | struct block_device *bdev = NULL; |
3355 | fmode_t mode; | |
f46b5a66 | 3356 | int ret; |
37b45995 | 3357 | bool cancel = false; |
f46b5a66 | 3358 | |
e441d54d CM |
3359 | if (!capable(CAP_SYS_ADMIN)) |
3360 | return -EPERM; | |
3361 | ||
58d7bbf8 | 3362 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1a15eb72 JB |
3363 | if (IS_ERR(vol_args)) |
3364 | return PTR_ERR(vol_args); | |
3365 | ||
58d7bbf8 | 3366 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
1a15eb72 JB |
3367 | if (!strcmp("cancel", vol_args->name)) { |
3368 | cancel = true; | |
3369 | } else { | |
3370 | ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name); | |
3371 | if (ret) | |
3372 | goto out; | |
3373 | } | |
3374 | ||
3375 | ret = mnt_want_write_file(file); | |
3376 | if (ret) | |
3377 | goto out; | |
67ae34b6 DS |
3378 | |
3379 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, | |
3380 | cancel); | |
3381 | if (ret == 0) { | |
1a15eb72 | 3382 | ret = btrfs_rm_device(fs_info, &args, &bdev, &mode); |
67ae34b6 DS |
3383 | if (!ret) |
3384 | btrfs_info(fs_info, "disk deleted %s", vol_args->name); | |
3385 | btrfs_exclop_finish(fs_info); | |
3386 | } | |
183860f6 | 3387 | |
4ac20c70 | 3388 | mnt_drop_write_file(file); |
3fa421de JB |
3389 | if (bdev) |
3390 | blkdev_put(bdev, mode); | |
1a15eb72 JB |
3391 | out: |
3392 | btrfs_put_dev_args_from_path(&args); | |
3393 | kfree(vol_args); | |
f46b5a66 CH |
3394 | return ret; |
3395 | } | |
3396 | ||
2ff7e61e JM |
3397 | static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, |
3398 | void __user *arg) | |
475f6387 | 3399 | { |
027ed2f0 | 3400 | struct btrfs_ioctl_fs_info_args *fi_args; |
475f6387 | 3401 | struct btrfs_device *device; |
0b246afa | 3402 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
137c5418 | 3403 | u64 flags_in; |
027ed2f0 | 3404 | int ret = 0; |
475f6387 | 3405 | |
137c5418 JT |
3406 | fi_args = memdup_user(arg, sizeof(*fi_args)); |
3407 | if (IS_ERR(fi_args)) | |
3408 | return PTR_ERR(fi_args); | |
3409 | ||
3410 | flags_in = fi_args->flags; | |
3411 | memset(fi_args, 0, sizeof(*fi_args)); | |
027ed2f0 | 3412 | |
d03262c7 | 3413 | rcu_read_lock(); |
027ed2f0 | 3414 | fi_args->num_devices = fs_devices->num_devices; |
475f6387 | 3415 | |
d03262c7 | 3416 | list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { |
027ed2f0 LZ |
3417 | if (device->devid > fi_args->max_id) |
3418 | fi_args->max_id = device->devid; | |
475f6387 | 3419 | } |
d03262c7 | 3420 | rcu_read_unlock(); |
475f6387 | 3421 | |
de37aa51 | 3422 | memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); |
bea7eafd OS |
3423 | fi_args->nodesize = fs_info->nodesize; |
3424 | fi_args->sectorsize = fs_info->sectorsize; | |
3425 | fi_args->clone_alignment = fs_info->sectorsize; | |
80a773fb | 3426 | |
137c5418 JT |
3427 | if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) { |
3428 | fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy); | |
3429 | fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy); | |
3430 | fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO; | |
3431 | } | |
3432 | ||
0fb408a5 JT |
3433 | if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) { |
3434 | fi_args->generation = fs_info->generation; | |
3435 | fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION; | |
3436 | } | |
3437 | ||
49bac897 JT |
3438 | if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) { |
3439 | memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid, | |
3440 | sizeof(fi_args->metadata_uuid)); | |
3441 | fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID; | |
3442 | } | |
3443 | ||
027ed2f0 LZ |
3444 | if (copy_to_user(arg, fi_args, sizeof(*fi_args))) |
3445 | ret = -EFAULT; | |
475f6387 | 3446 | |
027ed2f0 LZ |
3447 | kfree(fi_args); |
3448 | return ret; | |
475f6387 JS |
3449 | } |
3450 | ||
2ff7e61e JM |
3451 | static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, |
3452 | void __user *arg) | |
475f6387 | 3453 | { |
562d7b15 | 3454 | BTRFS_DEV_LOOKUP_ARGS(args); |
475f6387 JS |
3455 | struct btrfs_ioctl_dev_info_args *di_args; |
3456 | struct btrfs_device *dev; | |
475f6387 | 3457 | int ret = 0; |
475f6387 | 3458 | |
475f6387 JS |
3459 | di_args = memdup_user(arg, sizeof(*di_args)); |
3460 | if (IS_ERR(di_args)) | |
3461 | return PTR_ERR(di_args); | |
3462 | ||
562d7b15 | 3463 | args.devid = di_args->devid; |
dd5f9615 | 3464 | if (!btrfs_is_empty_uuid(di_args->uuid)) |
562d7b15 | 3465 | args.uuid = di_args->uuid; |
475f6387 | 3466 | |
c5593ca3 | 3467 | rcu_read_lock(); |
562d7b15 | 3468 | dev = btrfs_find_device(fs_info->fs_devices, &args); |
475f6387 JS |
3469 | if (!dev) { |
3470 | ret = -ENODEV; | |
3471 | goto out; | |
3472 | } | |
3473 | ||
3474 | di_args->devid = dev->devid; | |
7cc8e58d MX |
3475 | di_args->bytes_used = btrfs_device_get_bytes_used(dev); |
3476 | di_args->total_bytes = btrfs_device_get_total_bytes(dev); | |
475f6387 | 3477 | memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); |
a27202fb | 3478 | if (dev->name) { |
672d5990 MT |
3479 | strncpy(di_args->path, rcu_str_deref(dev->name), |
3480 | sizeof(di_args->path) - 1); | |
a27202fb JM |
3481 | di_args->path[sizeof(di_args->path) - 1] = 0; |
3482 | } else { | |
99ba55ad | 3483 | di_args->path[0] = '\0'; |
a27202fb | 3484 | } |
475f6387 JS |
3485 | |
3486 | out: | |
c5593ca3 | 3487 | rcu_read_unlock(); |
475f6387 JS |
3488 | if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) |
3489 | ret = -EFAULT; | |
3490 | ||
3491 | kfree(di_args); | |
3492 | return ret; | |
3493 | } | |
3494 | ||
6ef5ed0d JB |
3495 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
3496 | { | |
496ad9aa | 3497 | struct inode *inode = file_inode(file); |
0b246afa | 3498 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
6ef5ed0d JB |
3499 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3500 | struct btrfs_root *new_root; | |
3501 | struct btrfs_dir_item *di; | |
3502 | struct btrfs_trans_handle *trans; | |
2a2b5d62 | 3503 | struct btrfs_path *path = NULL; |
6ef5ed0d | 3504 | struct btrfs_disk_key disk_key; |
6ef5ed0d JB |
3505 | u64 objectid = 0; |
3506 | u64 dir_id; | |
3c04ce01 | 3507 | int ret; |
6ef5ed0d JB |
3508 | |
3509 | if (!capable(CAP_SYS_ADMIN)) | |
3510 | return -EPERM; | |
3511 | ||
3c04ce01 MX |
3512 | ret = mnt_want_write_file(file); |
3513 | if (ret) | |
3514 | return ret; | |
3515 | ||
3516 | if (copy_from_user(&objectid, argp, sizeof(objectid))) { | |
3517 | ret = -EFAULT; | |
3518 | goto out; | |
3519 | } | |
6ef5ed0d JB |
3520 | |
3521 | if (!objectid) | |
1cecf579 | 3522 | objectid = BTRFS_FS_TREE_OBJECTID; |
6ef5ed0d | 3523 | |
56e9357a | 3524 | new_root = btrfs_get_fs_root(fs_info, objectid, true); |
3c04ce01 MX |
3525 | if (IS_ERR(new_root)) { |
3526 | ret = PTR_ERR(new_root); | |
3527 | goto out; | |
3528 | } | |
2a2b5d62 JB |
3529 | if (!is_fstree(new_root->root_key.objectid)) { |
3530 | ret = -ENOENT; | |
3531 | goto out_free; | |
3532 | } | |
6ef5ed0d | 3533 | |
6ef5ed0d | 3534 | path = btrfs_alloc_path(); |
3c04ce01 MX |
3535 | if (!path) { |
3536 | ret = -ENOMEM; | |
2a2b5d62 | 3537 | goto out_free; |
3c04ce01 | 3538 | } |
6ef5ed0d JB |
3539 | |
3540 | trans = btrfs_start_transaction(root, 1); | |
98d5dc13 | 3541 | if (IS_ERR(trans)) { |
3c04ce01 | 3542 | ret = PTR_ERR(trans); |
2a2b5d62 | 3543 | goto out_free; |
6ef5ed0d JB |
3544 | } |
3545 | ||
0b246afa JM |
3546 | dir_id = btrfs_super_root_dir(fs_info->super_copy); |
3547 | di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, | |
6ef5ed0d | 3548 | dir_id, "default", 7, 1); |
cf1e99a4 | 3549 | if (IS_ERR_OR_NULL(di)) { |
2a2b5d62 | 3550 | btrfs_release_path(path); |
3a45bb20 | 3551 | btrfs_end_transaction(trans); |
0b246afa | 3552 | btrfs_err(fs_info, |
5d163e0e | 3553 | "Umm, you don't have the default diritem, this isn't going to work"); |
3c04ce01 | 3554 | ret = -ENOENT; |
2a2b5d62 | 3555 | goto out_free; |
6ef5ed0d JB |
3556 | } |
3557 | ||
3558 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
3559 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
3560 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
2a2b5d62 | 3561 | btrfs_release_path(path); |
6ef5ed0d | 3562 | |
0b246afa | 3563 | btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); |
3a45bb20 | 3564 | btrfs_end_transaction(trans); |
2a2b5d62 | 3565 | out_free: |
00246528 | 3566 | btrfs_put_root(new_root); |
2a2b5d62 | 3567 | btrfs_free_path(path); |
3c04ce01 MX |
3568 | out: |
3569 | mnt_drop_write_file(file); | |
3570 | return ret; | |
6ef5ed0d JB |
3571 | } |
3572 | ||
c065f5b1 SY |
3573 | static void get_block_group_info(struct list_head *groups_list, |
3574 | struct btrfs_ioctl_space_info *space) | |
bf5fc093 | 3575 | { |
32da5386 | 3576 | struct btrfs_block_group *block_group; |
bf5fc093 JB |
3577 | |
3578 | space->total_bytes = 0; | |
3579 | space->used_bytes = 0; | |
3580 | space->flags = 0; | |
3581 | list_for_each_entry(block_group, groups_list, list) { | |
3582 | space->flags = block_group->flags; | |
b3470b5d | 3583 | space->total_bytes += block_group->length; |
bf38be65 | 3584 | space->used_bytes += block_group->used; |
bf5fc093 JB |
3585 | } |
3586 | } | |
3587 | ||
2ff7e61e JM |
3588 | static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, |
3589 | void __user *arg) | |
1406e432 JB |
3590 | { |
3591 | struct btrfs_ioctl_space_args space_args; | |
3592 | struct btrfs_ioctl_space_info space; | |
3593 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf | 3594 | struct btrfs_ioctl_space_info *dest_orig; |
13f2696f | 3595 | struct btrfs_ioctl_space_info __user *user_dest; |
1406e432 | 3596 | struct btrfs_space_info *info; |
315d8e98 CIK |
3597 | static const u64 types[] = { |
3598 | BTRFS_BLOCK_GROUP_DATA, | |
3599 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3600 | BTRFS_BLOCK_GROUP_METADATA, | |
3601 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA | |
3602 | }; | |
bf5fc093 | 3603 | int num_types = 4; |
7fde62bf | 3604 | int alloc_size; |
1406e432 | 3605 | int ret = 0; |
51788b1b | 3606 | u64 slot_count = 0; |
bf5fc093 | 3607 | int i, c; |
1406e432 JB |
3608 | |
3609 | if (copy_from_user(&space_args, | |
3610 | (struct btrfs_ioctl_space_args __user *)arg, | |
3611 | sizeof(space_args))) | |
3612 | return -EFAULT; | |
3613 | ||
bf5fc093 JB |
3614 | for (i = 0; i < num_types; i++) { |
3615 | struct btrfs_space_info *tmp; | |
3616 | ||
3617 | info = NULL; | |
72804905 | 3618 | list_for_each_entry(tmp, &fs_info->space_info, list) { |
bf5fc093 JB |
3619 | if (tmp->flags == types[i]) { |
3620 | info = tmp; | |
3621 | break; | |
3622 | } | |
3623 | } | |
bf5fc093 JB |
3624 | |
3625 | if (!info) | |
3626 | continue; | |
3627 | ||
3628 | down_read(&info->groups_sem); | |
3629 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3630 | if (!list_empty(&info->block_groups[c])) | |
3631 | slot_count++; | |
3632 | } | |
3633 | up_read(&info->groups_sem); | |
3634 | } | |
7fde62bf | 3635 | |
36523e95 DS |
3636 | /* |
3637 | * Global block reserve, exported as a space_info | |
3638 | */ | |
3639 | slot_count++; | |
3640 | ||
7fde62bf CM |
3641 | /* space_slots == 0 means they are asking for a count */ |
3642 | if (space_args.space_slots == 0) { | |
3643 | space_args.total_spaces = slot_count; | |
3644 | goto out; | |
3645 | } | |
bf5fc093 | 3646 | |
51788b1b | 3647 | slot_count = min_t(u64, space_args.space_slots, slot_count); |
bf5fc093 | 3648 | |
7fde62bf | 3649 | alloc_size = sizeof(*dest) * slot_count; |
bf5fc093 | 3650 | |
7fde62bf CM |
3651 | /* we generally have at most 6 or so space infos, one for each raid |
3652 | * level. So, a whole page should be more than enough for everyone | |
3653 | */ | |
09cbfeaf | 3654 | if (alloc_size > PAGE_SIZE) |
7fde62bf CM |
3655 | return -ENOMEM; |
3656 | ||
1406e432 | 3657 | space_args.total_spaces = 0; |
8d2db785 | 3658 | dest = kmalloc(alloc_size, GFP_KERNEL); |
7fde62bf CM |
3659 | if (!dest) |
3660 | return -ENOMEM; | |
3661 | dest_orig = dest; | |
1406e432 | 3662 | |
7fde62bf | 3663 | /* now we have a buffer to copy into */ |
bf5fc093 JB |
3664 | for (i = 0; i < num_types; i++) { |
3665 | struct btrfs_space_info *tmp; | |
3666 | ||
51788b1b DR |
3667 | if (!slot_count) |
3668 | break; | |
3669 | ||
bf5fc093 | 3670 | info = NULL; |
72804905 | 3671 | list_for_each_entry(tmp, &fs_info->space_info, list) { |
bf5fc093 JB |
3672 | if (tmp->flags == types[i]) { |
3673 | info = tmp; | |
3674 | break; | |
3675 | } | |
3676 | } | |
7fde62bf | 3677 | |
bf5fc093 JB |
3678 | if (!info) |
3679 | continue; | |
3680 | down_read(&info->groups_sem); | |
3681 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3682 | if (!list_empty(&info->block_groups[c])) { | |
c065f5b1 SY |
3683 | get_block_group_info(&info->block_groups[c], |
3684 | &space); | |
bf5fc093 JB |
3685 | memcpy(dest, &space, sizeof(space)); |
3686 | dest++; | |
3687 | space_args.total_spaces++; | |
51788b1b | 3688 | slot_count--; |
bf5fc093 | 3689 | } |
51788b1b DR |
3690 | if (!slot_count) |
3691 | break; | |
bf5fc093 JB |
3692 | } |
3693 | up_read(&info->groups_sem); | |
1406e432 | 3694 | } |
1406e432 | 3695 | |
36523e95 DS |
3696 | /* |
3697 | * Add global block reserve | |
3698 | */ | |
3699 | if (slot_count) { | |
0b246afa | 3700 | struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; |
36523e95 DS |
3701 | |
3702 | spin_lock(&block_rsv->lock); | |
3703 | space.total_bytes = block_rsv->size; | |
3704 | space.used_bytes = block_rsv->size - block_rsv->reserved; | |
3705 | spin_unlock(&block_rsv->lock); | |
3706 | space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; | |
3707 | memcpy(dest, &space, sizeof(space)); | |
3708 | space_args.total_spaces++; | |
3709 | } | |
3710 | ||
2eec6c81 | 3711 | user_dest = (struct btrfs_ioctl_space_info __user *) |
7fde62bf CM |
3712 | (arg + sizeof(struct btrfs_ioctl_space_args)); |
3713 | ||
3714 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
3715 | ret = -EFAULT; | |
3716 | ||
3717 | kfree(dest_orig); | |
3718 | out: | |
3719 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
3720 | ret = -EFAULT; |
3721 | ||
3722 | return ret; | |
3723 | } | |
3724 | ||
9a8c28be MX |
3725 | static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, |
3726 | void __user *argp) | |
46204592 | 3727 | { |
46204592 SW |
3728 | struct btrfs_trans_handle *trans; |
3729 | u64 transid; | |
3730 | ||
d4edf39b | 3731 | trans = btrfs_attach_transaction_barrier(root); |
ff7c1d33 MX |
3732 | if (IS_ERR(trans)) { |
3733 | if (PTR_ERR(trans) != -ENOENT) | |
3734 | return PTR_ERR(trans); | |
3735 | ||
3736 | /* No running transaction, don't bother */ | |
3737 | transid = root->fs_info->last_trans_committed; | |
3738 | goto out; | |
3739 | } | |
46204592 | 3740 | transid = trans->transid; |
fdfbf020 | 3741 | btrfs_commit_transaction_async(trans); |
ff7c1d33 | 3742 | out: |
46204592 SW |
3743 | if (argp) |
3744 | if (copy_to_user(argp, &transid, sizeof(transid))) | |
3745 | return -EFAULT; | |
3746 | return 0; | |
3747 | } | |
3748 | ||
2ff7e61e | 3749 | static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, |
9a8c28be | 3750 | void __user *argp) |
46204592 | 3751 | { |
46204592 SW |
3752 | u64 transid; |
3753 | ||
3754 | if (argp) { | |
3755 | if (copy_from_user(&transid, argp, sizeof(transid))) | |
3756 | return -EFAULT; | |
3757 | } else { | |
3758 | transid = 0; /* current trans */ | |
3759 | } | |
2ff7e61e | 3760 | return btrfs_wait_for_commit(fs_info, transid); |
46204592 SW |
3761 | } |
3762 | ||
b8e95489 | 3763 | static long btrfs_ioctl_scrub(struct file *file, void __user *arg) |
475f6387 | 3764 | { |
0b246afa | 3765 | struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); |
475f6387 | 3766 | struct btrfs_ioctl_scrub_args *sa; |
b8e95489 | 3767 | int ret; |
475f6387 JS |
3768 | |
3769 | if (!capable(CAP_SYS_ADMIN)) | |
3770 | return -EPERM; | |
3771 | ||
3772 | sa = memdup_user(arg, sizeof(*sa)); | |
3773 | if (IS_ERR(sa)) | |
3774 | return PTR_ERR(sa); | |
3775 | ||
b8e95489 MX |
3776 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) { |
3777 | ret = mnt_want_write_file(file); | |
3778 | if (ret) | |
3779 | goto out; | |
3780 | } | |
3781 | ||
0b246afa | 3782 | ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, |
63a212ab SB |
3783 | &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, |
3784 | 0); | |
475f6387 | 3785 | |
5afe6ce7 FM |
3786 | /* |
3787 | * Copy scrub args to user space even if btrfs_scrub_dev() returned an | |
3788 | * error. This is important as it allows user space to know how much | |
3789 | * progress scrub has done. For example, if scrub is canceled we get | |
3790 | * -ECANCELED from btrfs_scrub_dev() and return that error back to user | |
3791 | * space. Later user space can inspect the progress from the structure | |
3792 | * btrfs_ioctl_scrub_args and resume scrub from where it left off | |
3793 | * previously (btrfs-progs does this). | |
3794 | * If we fail to copy the btrfs_ioctl_scrub_args structure to user space | |
3795 | * then return -EFAULT to signal the structure was not copied or it may | |
3796 | * be corrupt and unreliable due to a partial copy. | |
3797 | */ | |
3798 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
475f6387 JS |
3799 | ret = -EFAULT; |
3800 | ||
b8e95489 MX |
3801 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) |
3802 | mnt_drop_write_file(file); | |
3803 | out: | |
475f6387 JS |
3804 | kfree(sa); |
3805 | return ret; | |
3806 | } | |
3807 | ||
2ff7e61e | 3808 | static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) |
475f6387 JS |
3809 | { |
3810 | if (!capable(CAP_SYS_ADMIN)) | |
3811 | return -EPERM; | |
3812 | ||
2ff7e61e | 3813 | return btrfs_scrub_cancel(fs_info); |
475f6387 JS |
3814 | } |
3815 | ||
2ff7e61e | 3816 | static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, |
475f6387 JS |
3817 | void __user *arg) |
3818 | { | |
3819 | struct btrfs_ioctl_scrub_args *sa; | |
3820 | int ret; | |
3821 | ||
3822 | if (!capable(CAP_SYS_ADMIN)) | |
3823 | return -EPERM; | |
3824 | ||
3825 | sa = memdup_user(arg, sizeof(*sa)); | |
3826 | if (IS_ERR(sa)) | |
3827 | return PTR_ERR(sa); | |
3828 | ||
2ff7e61e | 3829 | ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); |
475f6387 | 3830 | |
4fa99b00 | 3831 | if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) |
475f6387 JS |
3832 | ret = -EFAULT; |
3833 | ||
3834 | kfree(sa); | |
3835 | return ret; | |
3836 | } | |
3837 | ||
2ff7e61e | 3838 | static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, |
b27f7c0c | 3839 | void __user *arg) |
c11d2c23 SB |
3840 | { |
3841 | struct btrfs_ioctl_get_dev_stats *sa; | |
3842 | int ret; | |
3843 | ||
c11d2c23 SB |
3844 | sa = memdup_user(arg, sizeof(*sa)); |
3845 | if (IS_ERR(sa)) | |
3846 | return PTR_ERR(sa); | |
3847 | ||
b27f7c0c DS |
3848 | if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { |
3849 | kfree(sa); | |
3850 | return -EPERM; | |
3851 | } | |
3852 | ||
2ff7e61e | 3853 | ret = btrfs_get_dev_stats(fs_info, sa); |
c11d2c23 | 3854 | |
eee99577 | 3855 | if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) |
c11d2c23 SB |
3856 | ret = -EFAULT; |
3857 | ||
3858 | kfree(sa); | |
3859 | return ret; | |
3860 | } | |
3861 | ||
2ff7e61e JM |
3862 | static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, |
3863 | void __user *arg) | |
3f6bcfbd SB |
3864 | { |
3865 | struct btrfs_ioctl_dev_replace_args *p; | |
3866 | int ret; | |
3867 | ||
3868 | if (!capable(CAP_SYS_ADMIN)) | |
3869 | return -EPERM; | |
3870 | ||
3871 | p = memdup_user(arg, sizeof(*p)); | |
3872 | if (IS_ERR(p)) | |
3873 | return PTR_ERR(p); | |
3874 | ||
3875 | switch (p->cmd) { | |
3876 | case BTRFS_IOCTL_DEV_REPLACE_CMD_START: | |
bc98a42c | 3877 | if (sb_rdonly(fs_info->sb)) { |
adfa97cb ID |
3878 | ret = -EROFS; |
3879 | goto out; | |
3880 | } | |
c3e1f96c | 3881 | if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) { |
e57138b3 | 3882 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
3f6bcfbd | 3883 | } else { |
2ff7e61e | 3884 | ret = btrfs_dev_replace_by_ioctl(fs_info, p); |
c3e1f96c | 3885 | btrfs_exclop_finish(fs_info); |
3f6bcfbd SB |
3886 | } |
3887 | break; | |
3888 | case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: | |
0b246afa | 3889 | btrfs_dev_replace_status(fs_info, p); |
3f6bcfbd SB |
3890 | ret = 0; |
3891 | break; | |
3892 | case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: | |
17d202b9 | 3893 | p->result = btrfs_dev_replace_cancel(fs_info); |
97282031 | 3894 | ret = 0; |
3f6bcfbd SB |
3895 | break; |
3896 | default: | |
3897 | ret = -EINVAL; | |
3898 | break; | |
3899 | } | |
3900 | ||
d3a53286 | 3901 | if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) |
3f6bcfbd | 3902 | ret = -EFAULT; |
adfa97cb | 3903 | out: |
3f6bcfbd SB |
3904 | kfree(p); |
3905 | return ret; | |
3906 | } | |
3907 | ||
d7728c96 JS |
3908 | static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) |
3909 | { | |
3910 | int ret = 0; | |
3911 | int i; | |
740c3d22 | 3912 | u64 rel_ptr; |
d7728c96 | 3913 | int size; |
806468f8 | 3914 | struct btrfs_ioctl_ino_path_args *ipa = NULL; |
d7728c96 JS |
3915 | struct inode_fs_paths *ipath = NULL; |
3916 | struct btrfs_path *path; | |
3917 | ||
82b22ac8 | 3918 | if (!capable(CAP_DAC_READ_SEARCH)) |
d7728c96 JS |
3919 | return -EPERM; |
3920 | ||
3921 | path = btrfs_alloc_path(); | |
3922 | if (!path) { | |
3923 | ret = -ENOMEM; | |
3924 | goto out; | |
3925 | } | |
3926 | ||
3927 | ipa = memdup_user(arg, sizeof(*ipa)); | |
3928 | if (IS_ERR(ipa)) { | |
3929 | ret = PTR_ERR(ipa); | |
3930 | ipa = NULL; | |
3931 | goto out; | |
3932 | } | |
3933 | ||
3934 | size = min_t(u32, ipa->size, 4096); | |
3935 | ipath = init_ipath(size, root, path); | |
3936 | if (IS_ERR(ipath)) { | |
3937 | ret = PTR_ERR(ipath); | |
3938 | ipath = NULL; | |
3939 | goto out; | |
3940 | } | |
3941 | ||
3942 | ret = paths_from_inode(ipa->inum, ipath); | |
3943 | if (ret < 0) | |
3944 | goto out; | |
3945 | ||
3946 | for (i = 0; i < ipath->fspath->elem_cnt; ++i) { | |
745c4d8e JM |
3947 | rel_ptr = ipath->fspath->val[i] - |
3948 | (u64)(unsigned long)ipath->fspath->val; | |
740c3d22 | 3949 | ipath->fspath->val[i] = rel_ptr; |
d7728c96 JS |
3950 | } |
3951 | ||
718dc5fa OS |
3952 | ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, |
3953 | ipath->fspath, size); | |
d7728c96 JS |
3954 | if (ret) { |
3955 | ret = -EFAULT; | |
3956 | goto out; | |
3957 | } | |
3958 | ||
3959 | out: | |
3960 | btrfs_free_path(path); | |
3961 | free_ipath(ipath); | |
3962 | kfree(ipa); | |
3963 | ||
3964 | return ret; | |
3965 | } | |
3966 | ||
3967 | static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) | |
3968 | { | |
3969 | struct btrfs_data_container *inodes = ctx; | |
3970 | const size_t c = 3 * sizeof(u64); | |
3971 | ||
3972 | if (inodes->bytes_left >= c) { | |
3973 | inodes->bytes_left -= c; | |
3974 | inodes->val[inodes->elem_cnt] = inum; | |
3975 | inodes->val[inodes->elem_cnt + 1] = offset; | |
3976 | inodes->val[inodes->elem_cnt + 2] = root; | |
3977 | inodes->elem_cnt += 3; | |
3978 | } else { | |
3979 | inodes->bytes_missing += c - inodes->bytes_left; | |
3980 | inodes->bytes_left = 0; | |
3981 | inodes->elem_missed += 3; | |
3982 | } | |
3983 | ||
3984 | return 0; | |
3985 | } | |
3986 | ||
2ff7e61e | 3987 | static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, |
d24a67b2 | 3988 | void __user *arg, int version) |
d7728c96 JS |
3989 | { |
3990 | int ret = 0; | |
3991 | int size; | |
d7728c96 JS |
3992 | struct btrfs_ioctl_logical_ino_args *loi; |
3993 | struct btrfs_data_container *inodes = NULL; | |
3994 | struct btrfs_path *path = NULL; | |
d24a67b2 | 3995 | bool ignore_offset; |
d7728c96 JS |
3996 | |
3997 | if (!capable(CAP_SYS_ADMIN)) | |
3998 | return -EPERM; | |
3999 | ||
4000 | loi = memdup_user(arg, sizeof(*loi)); | |
7b9ea627 SV |
4001 | if (IS_ERR(loi)) |
4002 | return PTR_ERR(loi); | |
d7728c96 | 4003 | |
d24a67b2 ZB |
4004 | if (version == 1) { |
4005 | ignore_offset = false; | |
b115e3bc | 4006 | size = min_t(u32, loi->size, SZ_64K); |
d24a67b2 ZB |
4007 | } else { |
4008 | /* All reserved bits must be 0 for now */ | |
4009 | if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { | |
4010 | ret = -EINVAL; | |
4011 | goto out_loi; | |
4012 | } | |
4013 | /* Only accept flags we have defined so far */ | |
4014 | if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { | |
4015 | ret = -EINVAL; | |
4016 | goto out_loi; | |
4017 | } | |
4018 | ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; | |
b115e3bc | 4019 | size = min_t(u32, loi->size, SZ_16M); |
d24a67b2 ZB |
4020 | } |
4021 | ||
d7728c96 JS |
4022 | path = btrfs_alloc_path(); |
4023 | if (!path) { | |
4024 | ret = -ENOMEM; | |
4025 | goto out; | |
4026 | } | |
4027 | ||
d7728c96 JS |
4028 | inodes = init_data_container(size); |
4029 | if (IS_ERR(inodes)) { | |
4030 | ret = PTR_ERR(inodes); | |
4031 | inodes = NULL; | |
4032 | goto out; | |
4033 | } | |
4034 | ||
2ff7e61e | 4035 | ret = iterate_inodes_from_logical(loi->logical, fs_info, path, |
d24a67b2 | 4036 | build_ino_list, inodes, ignore_offset); |
df031f07 | 4037 | if (ret == -EINVAL) |
d7728c96 JS |
4038 | ret = -ENOENT; |
4039 | if (ret < 0) | |
4040 | goto out; | |
4041 | ||
718dc5fa OS |
4042 | ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, |
4043 | size); | |
d7728c96 JS |
4044 | if (ret) |
4045 | ret = -EFAULT; | |
4046 | ||
4047 | out: | |
4048 | btrfs_free_path(path); | |
f54de068 | 4049 | kvfree(inodes); |
d24a67b2 | 4050 | out_loi: |
d7728c96 JS |
4051 | kfree(loi); |
4052 | ||
4053 | return ret; | |
4054 | } | |
4055 | ||
008ef096 | 4056 | void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, |
c9e9f97b ID |
4057 | struct btrfs_ioctl_balance_args *bargs) |
4058 | { | |
4059 | struct btrfs_balance_control *bctl = fs_info->balance_ctl; | |
4060 | ||
4061 | bargs->flags = bctl->flags; | |
4062 | ||
3009a62f | 4063 | if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) |
837d5b6e ID |
4064 | bargs->state |= BTRFS_BALANCE_STATE_RUNNING; |
4065 | if (atomic_read(&fs_info->balance_pause_req)) | |
4066 | bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; | |
a7e99c69 ID |
4067 | if (atomic_read(&fs_info->balance_cancel_req)) |
4068 | bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; | |
837d5b6e | 4069 | |
c9e9f97b ID |
4070 | memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); |
4071 | memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); | |
4072 | memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); | |
19a39dce | 4073 | |
008ef096 DS |
4074 | spin_lock(&fs_info->balance_lock); |
4075 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4076 | spin_unlock(&fs_info->balance_lock); | |
c9e9f97b ID |
4077 | } |
4078 | ||
9ba1f6e4 | 4079 | static long btrfs_ioctl_balance(struct file *file, void __user *arg) |
c9e9f97b | 4080 | { |
496ad9aa | 4081 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
c9e9f97b ID |
4082 | struct btrfs_fs_info *fs_info = root->fs_info; |
4083 | struct btrfs_ioctl_balance_args *bargs; | |
4084 | struct btrfs_balance_control *bctl; | |
ed0fb78f | 4085 | bool need_unlock; /* for mut. excl. ops lock */ |
c9e9f97b ID |
4086 | int ret; |
4087 | ||
6c405b24 NB |
4088 | if (!arg) |
4089 | btrfs_warn(fs_info, | |
4090 | "IOC_BALANCE ioctl (v1) is deprecated and will be removed in kernel 5.18"); | |
4091 | ||
c9e9f97b ID |
4092 | if (!capable(CAP_SYS_ADMIN)) |
4093 | return -EPERM; | |
4094 | ||
e54bfa31 | 4095 | ret = mnt_want_write_file(file); |
9ba1f6e4 LB |
4096 | if (ret) |
4097 | return ret; | |
4098 | ||
ed0fb78f | 4099 | again: |
c3e1f96c | 4100 | if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { |
ed0fb78f ID |
4101 | mutex_lock(&fs_info->balance_mutex); |
4102 | need_unlock = true; | |
4103 | goto locked; | |
4104 | } | |
4105 | ||
4106 | /* | |
01327610 | 4107 | * mut. excl. ops lock is locked. Three possibilities: |
ed0fb78f ID |
4108 | * (1) some other op is running |
4109 | * (2) balance is running | |
4110 | * (3) balance is paused -- special case (think resume) | |
4111 | */ | |
c9e9f97b | 4112 | mutex_lock(&fs_info->balance_mutex); |
ed0fb78f ID |
4113 | if (fs_info->balance_ctl) { |
4114 | /* this is either (2) or (3) */ | |
3009a62f | 4115 | if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { |
ed0fb78f | 4116 | mutex_unlock(&fs_info->balance_mutex); |
dccdb07b DS |
4117 | /* |
4118 | * Lock released to allow other waiters to continue, | |
4119 | * we'll reexamine the status again. | |
4120 | */ | |
ed0fb78f ID |
4121 | mutex_lock(&fs_info->balance_mutex); |
4122 | ||
4123 | if (fs_info->balance_ctl && | |
3009a62f | 4124 | !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { |
ed0fb78f ID |
4125 | /* this is (3) */ |
4126 | need_unlock = false; | |
4127 | goto locked; | |
4128 | } | |
4129 | ||
4130 | mutex_unlock(&fs_info->balance_mutex); | |
ed0fb78f ID |
4131 | goto again; |
4132 | } else { | |
4133 | /* this is (2) */ | |
4134 | mutex_unlock(&fs_info->balance_mutex); | |
4135 | ret = -EINPROGRESS; | |
4136 | goto out; | |
4137 | } | |
4138 | } else { | |
4139 | /* this is (1) */ | |
4140 | mutex_unlock(&fs_info->balance_mutex); | |
e57138b3 | 4141 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
ed0fb78f ID |
4142 | goto out; |
4143 | } | |
4144 | ||
4145 | locked: | |
c9e9f97b ID |
4146 | |
4147 | if (arg) { | |
4148 | bargs = memdup_user(arg, sizeof(*bargs)); | |
4149 | if (IS_ERR(bargs)) { | |
4150 | ret = PTR_ERR(bargs); | |
ed0fb78f | 4151 | goto out_unlock; |
c9e9f97b | 4152 | } |
de322263 ID |
4153 | |
4154 | if (bargs->flags & BTRFS_BALANCE_RESUME) { | |
4155 | if (!fs_info->balance_ctl) { | |
4156 | ret = -ENOTCONN; | |
4157 | goto out_bargs; | |
4158 | } | |
4159 | ||
4160 | bctl = fs_info->balance_ctl; | |
4161 | spin_lock(&fs_info->balance_lock); | |
4162 | bctl->flags |= BTRFS_BALANCE_RESUME; | |
4163 | spin_unlock(&fs_info->balance_lock); | |
efc0e69c | 4164 | btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE); |
de322263 ID |
4165 | |
4166 | goto do_balance; | |
4167 | } | |
c9e9f97b ID |
4168 | } else { |
4169 | bargs = NULL; | |
4170 | } | |
4171 | ||
ed0fb78f | 4172 | if (fs_info->balance_ctl) { |
837d5b6e ID |
4173 | ret = -EINPROGRESS; |
4174 | goto out_bargs; | |
4175 | } | |
4176 | ||
8d2db785 | 4177 | bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); |
c9e9f97b ID |
4178 | if (!bctl) { |
4179 | ret = -ENOMEM; | |
4180 | goto out_bargs; | |
4181 | } | |
4182 | ||
c9e9f97b ID |
4183 | if (arg) { |
4184 | memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); | |
4185 | memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); | |
4186 | memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); | |
4187 | ||
4188 | bctl->flags = bargs->flags; | |
f43ffb60 ID |
4189 | } else { |
4190 | /* balance everything - no filters */ | |
4191 | bctl->flags |= BTRFS_BALANCE_TYPE_MASK; | |
c9e9f97b ID |
4192 | } |
4193 | ||
8eb93459 DS |
4194 | if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { |
4195 | ret = -EINVAL; | |
0f89abf5 | 4196 | goto out_bctl; |
8eb93459 DS |
4197 | } |
4198 | ||
de322263 | 4199 | do_balance: |
c9e9f97b | 4200 | /* |
c3e1f96c GR |
4201 | * Ownership of bctl and exclusive operation goes to btrfs_balance. |
4202 | * bctl is freed in reset_balance_state, or, if restriper was paused | |
4203 | * all the way until unmount, in free_fs_info. The flag should be | |
4204 | * cleared after reset_balance_state. | |
c9e9f97b | 4205 | */ |
ed0fb78f ID |
4206 | need_unlock = false; |
4207 | ||
6fcf6e2b | 4208 | ret = btrfs_balance(fs_info, bctl, bargs); |
0f89abf5 | 4209 | bctl = NULL; |
ed0fb78f | 4210 | |
d00c2d9c | 4211 | if ((ret == 0 || ret == -ECANCELED) && arg) { |
c9e9f97b ID |
4212 | if (copy_to_user(arg, bargs, sizeof(*bargs))) |
4213 | ret = -EFAULT; | |
4214 | } | |
4215 | ||
0f89abf5 CE |
4216 | out_bctl: |
4217 | kfree(bctl); | |
c9e9f97b ID |
4218 | out_bargs: |
4219 | kfree(bargs); | |
ed0fb78f | 4220 | out_unlock: |
c9e9f97b | 4221 | mutex_unlock(&fs_info->balance_mutex); |
ed0fb78f | 4222 | if (need_unlock) |
c3e1f96c | 4223 | btrfs_exclop_finish(fs_info); |
ed0fb78f | 4224 | out: |
e54bfa31 | 4225 | mnt_drop_write_file(file); |
c9e9f97b ID |
4226 | return ret; |
4227 | } | |
4228 | ||
2ff7e61e | 4229 | static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) |
837d5b6e ID |
4230 | { |
4231 | if (!capable(CAP_SYS_ADMIN)) | |
4232 | return -EPERM; | |
4233 | ||
4234 | switch (cmd) { | |
4235 | case BTRFS_BALANCE_CTL_PAUSE: | |
0b246afa | 4236 | return btrfs_pause_balance(fs_info); |
a7e99c69 | 4237 | case BTRFS_BALANCE_CTL_CANCEL: |
0b246afa | 4238 | return btrfs_cancel_balance(fs_info); |
837d5b6e ID |
4239 | } |
4240 | ||
4241 | return -EINVAL; | |
4242 | } | |
4243 | ||
2ff7e61e | 4244 | static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, |
19a39dce ID |
4245 | void __user *arg) |
4246 | { | |
19a39dce ID |
4247 | struct btrfs_ioctl_balance_args *bargs; |
4248 | int ret = 0; | |
4249 | ||
4250 | if (!capable(CAP_SYS_ADMIN)) | |
4251 | return -EPERM; | |
4252 | ||
4253 | mutex_lock(&fs_info->balance_mutex); | |
4254 | if (!fs_info->balance_ctl) { | |
4255 | ret = -ENOTCONN; | |
4256 | goto out; | |
4257 | } | |
4258 | ||
8d2db785 | 4259 | bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); |
19a39dce ID |
4260 | if (!bargs) { |
4261 | ret = -ENOMEM; | |
4262 | goto out; | |
4263 | } | |
4264 | ||
008ef096 | 4265 | btrfs_update_ioctl_balance_args(fs_info, bargs); |
19a39dce ID |
4266 | |
4267 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4268 | ret = -EFAULT; | |
4269 | ||
4270 | kfree(bargs); | |
4271 | out: | |
4272 | mutex_unlock(&fs_info->balance_mutex); | |
4273 | return ret; | |
4274 | } | |
4275 | ||
905b0dda | 4276 | static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) |
5d13a37b | 4277 | { |
0b246afa JM |
4278 | struct inode *inode = file_inode(file); |
4279 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5d13a37b | 4280 | struct btrfs_ioctl_quota_ctl_args *sa; |
5d13a37b | 4281 | int ret; |
5d13a37b AJ |
4282 | |
4283 | if (!capable(CAP_SYS_ADMIN)) | |
4284 | return -EPERM; | |
4285 | ||
905b0dda MX |
4286 | ret = mnt_want_write_file(file); |
4287 | if (ret) | |
4288 | return ret; | |
5d13a37b AJ |
4289 | |
4290 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4291 | if (IS_ERR(sa)) { |
4292 | ret = PTR_ERR(sa); | |
4293 | goto drop_write; | |
4294 | } | |
5d13a37b | 4295 | |
0b246afa | 4296 | down_write(&fs_info->subvol_sem); |
5d13a37b AJ |
4297 | |
4298 | switch (sa->cmd) { | |
4299 | case BTRFS_QUOTA_CTL_ENABLE: | |
340f1aa2 | 4300 | ret = btrfs_quota_enable(fs_info); |
5d13a37b AJ |
4301 | break; |
4302 | case BTRFS_QUOTA_CTL_DISABLE: | |
340f1aa2 | 4303 | ret = btrfs_quota_disable(fs_info); |
5d13a37b | 4304 | break; |
5d13a37b AJ |
4305 | default: |
4306 | ret = -EINVAL; | |
4307 | break; | |
4308 | } | |
4309 | ||
5d13a37b | 4310 | kfree(sa); |
0b246afa | 4311 | up_write(&fs_info->subvol_sem); |
905b0dda MX |
4312 | drop_write: |
4313 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4314 | return ret; |
4315 | } | |
4316 | ||
905b0dda | 4317 | static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) |
5d13a37b | 4318 | { |
0b246afa JM |
4319 | struct inode *inode = file_inode(file); |
4320 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4321 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5d13a37b AJ |
4322 | struct btrfs_ioctl_qgroup_assign_args *sa; |
4323 | struct btrfs_trans_handle *trans; | |
4324 | int ret; | |
4325 | int err; | |
4326 | ||
4327 | if (!capable(CAP_SYS_ADMIN)) | |
4328 | return -EPERM; | |
4329 | ||
905b0dda MX |
4330 | ret = mnt_want_write_file(file); |
4331 | if (ret) | |
4332 | return ret; | |
5d13a37b AJ |
4333 | |
4334 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4335 | if (IS_ERR(sa)) { |
4336 | ret = PTR_ERR(sa); | |
4337 | goto drop_write; | |
4338 | } | |
5d13a37b AJ |
4339 | |
4340 | trans = btrfs_join_transaction(root); | |
4341 | if (IS_ERR(trans)) { | |
4342 | ret = PTR_ERR(trans); | |
4343 | goto out; | |
4344 | } | |
4345 | ||
5d13a37b | 4346 | if (sa->assign) { |
9f8a6ce6 | 4347 | ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); |
5d13a37b | 4348 | } else { |
39616c27 | 4349 | ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); |
5d13a37b AJ |
4350 | } |
4351 | ||
e082f563 | 4352 | /* update qgroup status and info */ |
280f8bd2 | 4353 | err = btrfs_run_qgroups(trans); |
e082f563 | 4354 | if (err < 0) |
0b246afa JM |
4355 | btrfs_handle_fs_error(fs_info, err, |
4356 | "failed to update qgroup status and info"); | |
3a45bb20 | 4357 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4358 | if (err && !ret) |
4359 | ret = err; | |
4360 | ||
4361 | out: | |
4362 | kfree(sa); | |
905b0dda MX |
4363 | drop_write: |
4364 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4365 | return ret; |
4366 | } | |
4367 | ||
905b0dda | 4368 | static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) |
5d13a37b | 4369 | { |
0b246afa | 4370 | struct inode *inode = file_inode(file); |
0b246afa | 4371 | struct btrfs_root *root = BTRFS_I(inode)->root; |
5d13a37b AJ |
4372 | struct btrfs_ioctl_qgroup_create_args *sa; |
4373 | struct btrfs_trans_handle *trans; | |
4374 | int ret; | |
4375 | int err; | |
4376 | ||
4377 | if (!capable(CAP_SYS_ADMIN)) | |
4378 | return -EPERM; | |
4379 | ||
905b0dda MX |
4380 | ret = mnt_want_write_file(file); |
4381 | if (ret) | |
4382 | return ret; | |
5d13a37b AJ |
4383 | |
4384 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4385 | if (IS_ERR(sa)) { |
4386 | ret = PTR_ERR(sa); | |
4387 | goto drop_write; | |
4388 | } | |
5d13a37b | 4389 | |
d86e56cf MX |
4390 | if (!sa->qgroupid) { |
4391 | ret = -EINVAL; | |
4392 | goto out; | |
4393 | } | |
4394 | ||
5d13a37b AJ |
4395 | trans = btrfs_join_transaction(root); |
4396 | if (IS_ERR(trans)) { | |
4397 | ret = PTR_ERR(trans); | |
4398 | goto out; | |
4399 | } | |
4400 | ||
5d13a37b | 4401 | if (sa->create) { |
49a05ecd | 4402 | ret = btrfs_create_qgroup(trans, sa->qgroupid); |
5d13a37b | 4403 | } else { |
3efbee1d | 4404 | ret = btrfs_remove_qgroup(trans, sa->qgroupid); |
5d13a37b AJ |
4405 | } |
4406 | ||
3a45bb20 | 4407 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4408 | if (err && !ret) |
4409 | ret = err; | |
4410 | ||
4411 | out: | |
4412 | kfree(sa); | |
905b0dda MX |
4413 | drop_write: |
4414 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4415 | return ret; |
4416 | } | |
4417 | ||
905b0dda | 4418 | static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) |
5d13a37b | 4419 | { |
0b246afa | 4420 | struct inode *inode = file_inode(file); |
0b246afa | 4421 | struct btrfs_root *root = BTRFS_I(inode)->root; |
5d13a37b AJ |
4422 | struct btrfs_ioctl_qgroup_limit_args *sa; |
4423 | struct btrfs_trans_handle *trans; | |
4424 | int ret; | |
4425 | int err; | |
4426 | u64 qgroupid; | |
4427 | ||
4428 | if (!capable(CAP_SYS_ADMIN)) | |
4429 | return -EPERM; | |
4430 | ||
905b0dda MX |
4431 | ret = mnt_want_write_file(file); |
4432 | if (ret) | |
4433 | return ret; | |
5d13a37b AJ |
4434 | |
4435 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4436 | if (IS_ERR(sa)) { |
4437 | ret = PTR_ERR(sa); | |
4438 | goto drop_write; | |
4439 | } | |
5d13a37b AJ |
4440 | |
4441 | trans = btrfs_join_transaction(root); | |
4442 | if (IS_ERR(trans)) { | |
4443 | ret = PTR_ERR(trans); | |
4444 | goto out; | |
4445 | } | |
4446 | ||
4447 | qgroupid = sa->qgroupid; | |
4448 | if (!qgroupid) { | |
4449 | /* take the current subvol as qgroup */ | |
4450 | qgroupid = root->root_key.objectid; | |
4451 | } | |
4452 | ||
f0042d5e | 4453 | ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); |
5d13a37b | 4454 | |
3a45bb20 | 4455 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4456 | if (err && !ret) |
4457 | ret = err; | |
4458 | ||
4459 | out: | |
4460 | kfree(sa); | |
905b0dda MX |
4461 | drop_write: |
4462 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4463 | return ret; |
4464 | } | |
4465 | ||
2f232036 JS |
4466 | static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) |
4467 | { | |
0b246afa JM |
4468 | struct inode *inode = file_inode(file); |
4469 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
2f232036 JS |
4470 | struct btrfs_ioctl_quota_rescan_args *qsa; |
4471 | int ret; | |
4472 | ||
4473 | if (!capable(CAP_SYS_ADMIN)) | |
4474 | return -EPERM; | |
4475 | ||
4476 | ret = mnt_want_write_file(file); | |
4477 | if (ret) | |
4478 | return ret; | |
4479 | ||
4480 | qsa = memdup_user(arg, sizeof(*qsa)); | |
4481 | if (IS_ERR(qsa)) { | |
4482 | ret = PTR_ERR(qsa); | |
4483 | goto drop_write; | |
4484 | } | |
4485 | ||
4486 | if (qsa->flags) { | |
4487 | ret = -EINVAL; | |
4488 | goto out; | |
4489 | } | |
4490 | ||
0b246afa | 4491 | ret = btrfs_qgroup_rescan(fs_info); |
2f232036 JS |
4492 | |
4493 | out: | |
4494 | kfree(qsa); | |
4495 | drop_write: | |
4496 | mnt_drop_write_file(file); | |
4497 | return ret; | |
4498 | } | |
4499 | ||
b929c1d8 MPS |
4500 | static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, |
4501 | void __user *arg) | |
2f232036 | 4502 | { |
0afb603a | 4503 | struct btrfs_ioctl_quota_rescan_args qsa = {0}; |
2f232036 JS |
4504 | |
4505 | if (!capable(CAP_SYS_ADMIN)) | |
4506 | return -EPERM; | |
4507 | ||
0b246afa | 4508 | if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { |
0afb603a GR |
4509 | qsa.flags = 1; |
4510 | qsa.progress = fs_info->qgroup_rescan_progress.objectid; | |
2f232036 JS |
4511 | } |
4512 | ||
0afb603a | 4513 | if (copy_to_user(arg, &qsa, sizeof(qsa))) |
991a3dae | 4514 | return -EFAULT; |
2f232036 | 4515 | |
991a3dae | 4516 | return 0; |
2f232036 JS |
4517 | } |
4518 | ||
b929c1d8 MPS |
4519 | static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info, |
4520 | void __user *arg) | |
57254b6e | 4521 | { |
57254b6e JS |
4522 | if (!capable(CAP_SYS_ADMIN)) |
4523 | return -EPERM; | |
4524 | ||
0b246afa | 4525 | return btrfs_qgroup_wait_for_completion(fs_info, true); |
57254b6e JS |
4526 | } |
4527 | ||
abccd00f | 4528 | static long _btrfs_ioctl_set_received_subvol(struct file *file, |
e4fed17a | 4529 | struct user_namespace *mnt_userns, |
abccd00f | 4530 | struct btrfs_ioctl_received_subvol_args *sa) |
8ea05e3a | 4531 | { |
496ad9aa | 4532 | struct inode *inode = file_inode(file); |
0b246afa | 4533 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
8ea05e3a AB |
4534 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4535 | struct btrfs_root_item *root_item = &root->root_item; | |
4536 | struct btrfs_trans_handle *trans; | |
95582b00 | 4537 | struct timespec64 ct = current_time(inode); |
8ea05e3a | 4538 | int ret = 0; |
dd5f9615 | 4539 | int received_uuid_changed; |
8ea05e3a | 4540 | |
e4fed17a | 4541 | if (!inode_owner_or_capable(mnt_userns, inode)) |
bd60ea0f DS |
4542 | return -EPERM; |
4543 | ||
8ea05e3a AB |
4544 | ret = mnt_want_write_file(file); |
4545 | if (ret < 0) | |
4546 | return ret; | |
4547 | ||
0b246afa | 4548 | down_write(&fs_info->subvol_sem); |
8ea05e3a | 4549 | |
4a0cc7ca | 4550 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
8ea05e3a AB |
4551 | ret = -EINVAL; |
4552 | goto out; | |
4553 | } | |
4554 | ||
4555 | if (btrfs_root_readonly(root)) { | |
4556 | ret = -EROFS; | |
4557 | goto out; | |
4558 | } | |
4559 | ||
dd5f9615 SB |
4560 | /* |
4561 | * 1 - root item | |
4562 | * 2 - uuid items (received uuid + subvol uuid) | |
4563 | */ | |
4564 | trans = btrfs_start_transaction(root, 3); | |
8ea05e3a AB |
4565 | if (IS_ERR(trans)) { |
4566 | ret = PTR_ERR(trans); | |
4567 | trans = NULL; | |
4568 | goto out; | |
4569 | } | |
4570 | ||
4571 | sa->rtransid = trans->transid; | |
4572 | sa->rtime.sec = ct.tv_sec; | |
4573 | sa->rtime.nsec = ct.tv_nsec; | |
4574 | ||
dd5f9615 SB |
4575 | received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, |
4576 | BTRFS_UUID_SIZE); | |
4577 | if (received_uuid_changed && | |
d87ff758 | 4578 | !btrfs_is_empty_uuid(root_item->received_uuid)) { |
d1957791 | 4579 | ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, |
d87ff758 NB |
4580 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, |
4581 | root->root_key.objectid); | |
4582 | if (ret && ret != -ENOENT) { | |
4583 | btrfs_abort_transaction(trans, ret); | |
4584 | btrfs_end_transaction(trans); | |
4585 | goto out; | |
4586 | } | |
4587 | } | |
8ea05e3a AB |
4588 | memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); |
4589 | btrfs_set_root_stransid(root_item, sa->stransid); | |
4590 | btrfs_set_root_rtransid(root_item, sa->rtransid); | |
3cae210f QW |
4591 | btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); |
4592 | btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); | |
4593 | btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); | |
4594 | btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); | |
8ea05e3a | 4595 | |
0b246afa | 4596 | ret = btrfs_update_root(trans, fs_info->tree_root, |
8ea05e3a AB |
4597 | &root->root_key, &root->root_item); |
4598 | if (ret < 0) { | |
3a45bb20 | 4599 | btrfs_end_transaction(trans); |
8ea05e3a | 4600 | goto out; |
dd5f9615 SB |
4601 | } |
4602 | if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { | |
cdb345a8 | 4603 | ret = btrfs_uuid_tree_add(trans, sa->uuid, |
dd5f9615 SB |
4604 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, |
4605 | root->root_key.objectid); | |
4606 | if (ret < 0 && ret != -EEXIST) { | |
66642832 | 4607 | btrfs_abort_transaction(trans, ret); |
efd38150 | 4608 | btrfs_end_transaction(trans); |
8ea05e3a | 4609 | goto out; |
dd5f9615 SB |
4610 | } |
4611 | } | |
3a45bb20 | 4612 | ret = btrfs_commit_transaction(trans); |
abccd00f | 4613 | out: |
0b246afa | 4614 | up_write(&fs_info->subvol_sem); |
abccd00f HM |
4615 | mnt_drop_write_file(file); |
4616 | return ret; | |
4617 | } | |
4618 | ||
4619 | #ifdef CONFIG_64BIT | |
4620 | static long btrfs_ioctl_set_received_subvol_32(struct file *file, | |
4621 | void __user *arg) | |
4622 | { | |
4623 | struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; | |
4624 | struct btrfs_ioctl_received_subvol_args *args64 = NULL; | |
4625 | int ret = 0; | |
4626 | ||
4627 | args32 = memdup_user(arg, sizeof(*args32)); | |
7b9ea627 SV |
4628 | if (IS_ERR(args32)) |
4629 | return PTR_ERR(args32); | |
abccd00f | 4630 | |
8d2db785 | 4631 | args64 = kmalloc(sizeof(*args64), GFP_KERNEL); |
84dbeb87 DC |
4632 | if (!args64) { |
4633 | ret = -ENOMEM; | |
abccd00f HM |
4634 | goto out; |
4635 | } | |
4636 | ||
4637 | memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); | |
4638 | args64->stransid = args32->stransid; | |
4639 | args64->rtransid = args32->rtransid; | |
4640 | args64->stime.sec = args32->stime.sec; | |
4641 | args64->stime.nsec = args32->stime.nsec; | |
4642 | args64->rtime.sec = args32->rtime.sec; | |
4643 | args64->rtime.nsec = args32->rtime.nsec; | |
4644 | args64->flags = args32->flags; | |
4645 | ||
e4fed17a | 4646 | ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64); |
abccd00f HM |
4647 | if (ret) |
4648 | goto out; | |
4649 | ||
4650 | memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); | |
4651 | args32->stransid = args64->stransid; | |
4652 | args32->rtransid = args64->rtransid; | |
4653 | args32->stime.sec = args64->stime.sec; | |
4654 | args32->stime.nsec = args64->stime.nsec; | |
4655 | args32->rtime.sec = args64->rtime.sec; | |
4656 | args32->rtime.nsec = args64->rtime.nsec; | |
4657 | args32->flags = args64->flags; | |
4658 | ||
4659 | ret = copy_to_user(arg, args32, sizeof(*args32)); | |
4660 | if (ret) | |
4661 | ret = -EFAULT; | |
4662 | ||
4663 | out: | |
4664 | kfree(args32); | |
4665 | kfree(args64); | |
4666 | return ret; | |
4667 | } | |
4668 | #endif | |
4669 | ||
4670 | static long btrfs_ioctl_set_received_subvol(struct file *file, | |
4671 | void __user *arg) | |
4672 | { | |
4673 | struct btrfs_ioctl_received_subvol_args *sa = NULL; | |
4674 | int ret = 0; | |
4675 | ||
4676 | sa = memdup_user(arg, sizeof(*sa)); | |
7b9ea627 SV |
4677 | if (IS_ERR(sa)) |
4678 | return PTR_ERR(sa); | |
abccd00f | 4679 | |
e4fed17a | 4680 | ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa); |
abccd00f HM |
4681 | |
4682 | if (ret) | |
4683 | goto out; | |
4684 | ||
8ea05e3a AB |
4685 | ret = copy_to_user(arg, sa, sizeof(*sa)); |
4686 | if (ret) | |
4687 | ret = -EFAULT; | |
4688 | ||
4689 | out: | |
4690 | kfree(sa); | |
8ea05e3a AB |
4691 | return ret; |
4692 | } | |
4693 | ||
b929c1d8 MPS |
4694 | static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, |
4695 | void __user *arg) | |
867ab667 | 4696 | { |
a1b83ac5 | 4697 | size_t len; |
867ab667 | 4698 | int ret; |
a1b83ac5 AJ |
4699 | char label[BTRFS_LABEL_SIZE]; |
4700 | ||
0b246afa JM |
4701 | spin_lock(&fs_info->super_lock); |
4702 | memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); | |
4703 | spin_unlock(&fs_info->super_lock); | |
a1b83ac5 AJ |
4704 | |
4705 | len = strnlen(label, BTRFS_LABEL_SIZE); | |
867ab667 | 4706 | |
4707 | if (len == BTRFS_LABEL_SIZE) { | |
0b246afa JM |
4708 | btrfs_warn(fs_info, |
4709 | "label is too long, return the first %zu bytes", | |
4710 | --len); | |
867ab667 | 4711 | } |
4712 | ||
867ab667 | 4713 | ret = copy_to_user(arg, label, len); |
867ab667 | 4714 | |
4715 | return ret ? -EFAULT : 0; | |
4716 | } | |
4717 | ||
a8bfd4ab | 4718 | static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) |
4719 | { | |
0b246afa JM |
4720 | struct inode *inode = file_inode(file); |
4721 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4722 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4723 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
a8bfd4ab | 4724 | struct btrfs_trans_handle *trans; |
4725 | char label[BTRFS_LABEL_SIZE]; | |
4726 | int ret; | |
4727 | ||
4728 | if (!capable(CAP_SYS_ADMIN)) | |
4729 | return -EPERM; | |
4730 | ||
4731 | if (copy_from_user(label, arg, sizeof(label))) | |
4732 | return -EFAULT; | |
4733 | ||
4734 | if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { | |
0b246afa | 4735 | btrfs_err(fs_info, |
5d163e0e JM |
4736 | "unable to set label with more than %d bytes", |
4737 | BTRFS_LABEL_SIZE - 1); | |
a8bfd4ab | 4738 | return -EINVAL; |
4739 | } | |
4740 | ||
4741 | ret = mnt_want_write_file(file); | |
4742 | if (ret) | |
4743 | return ret; | |
4744 | ||
a8bfd4ab | 4745 | trans = btrfs_start_transaction(root, 0); |
4746 | if (IS_ERR(trans)) { | |
4747 | ret = PTR_ERR(trans); | |
4748 | goto out_unlock; | |
4749 | } | |
4750 | ||
0b246afa | 4751 | spin_lock(&fs_info->super_lock); |
a8bfd4ab | 4752 | strcpy(super_block->label, label); |
0b246afa | 4753 | spin_unlock(&fs_info->super_lock); |
3a45bb20 | 4754 | ret = btrfs_commit_transaction(trans); |
a8bfd4ab | 4755 | |
4756 | out_unlock: | |
a8bfd4ab | 4757 | mnt_drop_write_file(file); |
4758 | return ret; | |
4759 | } | |
4760 | ||
2eaa055f JM |
4761 | #define INIT_FEATURE_FLAGS(suffix) \ |
4762 | { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ | |
4763 | .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ | |
4764 | .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } | |
4765 | ||
d5131b65 | 4766 | int btrfs_ioctl_get_supported_features(void __user *arg) |
2eaa055f | 4767 | { |
4d4ab6d6 | 4768 | static const struct btrfs_ioctl_feature_flags features[3] = { |
2eaa055f JM |
4769 | INIT_FEATURE_FLAGS(SUPP), |
4770 | INIT_FEATURE_FLAGS(SAFE_SET), | |
4771 | INIT_FEATURE_FLAGS(SAFE_CLEAR) | |
4772 | }; | |
4773 | ||
4774 | if (copy_to_user(arg, &features, sizeof(features))) | |
4775 | return -EFAULT; | |
4776 | ||
4777 | return 0; | |
4778 | } | |
4779 | ||
b929c1d8 MPS |
4780 | static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, |
4781 | void __user *arg) | |
2eaa055f | 4782 | { |
0b246afa | 4783 | struct btrfs_super_block *super_block = fs_info->super_copy; |
2eaa055f JM |
4784 | struct btrfs_ioctl_feature_flags features; |
4785 | ||
4786 | features.compat_flags = btrfs_super_compat_flags(super_block); | |
4787 | features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); | |
4788 | features.incompat_flags = btrfs_super_incompat_flags(super_block); | |
4789 | ||
4790 | if (copy_to_user(arg, &features, sizeof(features))) | |
4791 | return -EFAULT; | |
4792 | ||
4793 | return 0; | |
4794 | } | |
4795 | ||
2ff7e61e | 4796 | static int check_feature_bits(struct btrfs_fs_info *fs_info, |
3b02a68a | 4797 | enum btrfs_feature_set set, |
2eaa055f JM |
4798 | u64 change_mask, u64 flags, u64 supported_flags, |
4799 | u64 safe_set, u64 safe_clear) | |
4800 | { | |
f10152bc | 4801 | const char *type = btrfs_feature_set_name(set); |
3b02a68a | 4802 | char *names; |
2eaa055f JM |
4803 | u64 disallowed, unsupported; |
4804 | u64 set_mask = flags & change_mask; | |
4805 | u64 clear_mask = ~flags & change_mask; | |
4806 | ||
4807 | unsupported = set_mask & ~supported_flags; | |
4808 | if (unsupported) { | |
3b02a68a JM |
4809 | names = btrfs_printable_features(set, unsupported); |
4810 | if (names) { | |
0b246afa JM |
4811 | btrfs_warn(fs_info, |
4812 | "this kernel does not support the %s feature bit%s", | |
4813 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
4814 | kfree(names); |
4815 | } else | |
0b246afa JM |
4816 | btrfs_warn(fs_info, |
4817 | "this kernel does not support %s bits 0x%llx", | |
4818 | type, unsupported); | |
2eaa055f JM |
4819 | return -EOPNOTSUPP; |
4820 | } | |
4821 | ||
4822 | disallowed = set_mask & ~safe_set; | |
4823 | if (disallowed) { | |
3b02a68a JM |
4824 | names = btrfs_printable_features(set, disallowed); |
4825 | if (names) { | |
0b246afa JM |
4826 | btrfs_warn(fs_info, |
4827 | "can't set the %s feature bit%s while mounted", | |
4828 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
4829 | kfree(names); |
4830 | } else | |
0b246afa JM |
4831 | btrfs_warn(fs_info, |
4832 | "can't set %s bits 0x%llx while mounted", | |
4833 | type, disallowed); | |
2eaa055f JM |
4834 | return -EPERM; |
4835 | } | |
4836 | ||
4837 | disallowed = clear_mask & ~safe_clear; | |
4838 | if (disallowed) { | |
3b02a68a JM |
4839 | names = btrfs_printable_features(set, disallowed); |
4840 | if (names) { | |
0b246afa JM |
4841 | btrfs_warn(fs_info, |
4842 | "can't clear the %s feature bit%s while mounted", | |
4843 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
4844 | kfree(names); |
4845 | } else | |
0b246afa JM |
4846 | btrfs_warn(fs_info, |
4847 | "can't clear %s bits 0x%llx while mounted", | |
4848 | type, disallowed); | |
2eaa055f JM |
4849 | return -EPERM; |
4850 | } | |
4851 | ||
4852 | return 0; | |
4853 | } | |
4854 | ||
2ff7e61e JM |
4855 | #define check_feature(fs_info, change_mask, flags, mask_base) \ |
4856 | check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ | |
2eaa055f JM |
4857 | BTRFS_FEATURE_ ## mask_base ## _SUPP, \ |
4858 | BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ | |
4859 | BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) | |
4860 | ||
4861 | static int btrfs_ioctl_set_features(struct file *file, void __user *arg) | |
4862 | { | |
0b246afa JM |
4863 | struct inode *inode = file_inode(file); |
4864 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4865 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4866 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
2eaa055f JM |
4867 | struct btrfs_ioctl_feature_flags flags[2]; |
4868 | struct btrfs_trans_handle *trans; | |
4869 | u64 newflags; | |
4870 | int ret; | |
4871 | ||
4872 | if (!capable(CAP_SYS_ADMIN)) | |
4873 | return -EPERM; | |
4874 | ||
4875 | if (copy_from_user(flags, arg, sizeof(flags))) | |
4876 | return -EFAULT; | |
4877 | ||
4878 | /* Nothing to do */ | |
4879 | if (!flags[0].compat_flags && !flags[0].compat_ro_flags && | |
4880 | !flags[0].incompat_flags) | |
4881 | return 0; | |
4882 | ||
2ff7e61e | 4883 | ret = check_feature(fs_info, flags[0].compat_flags, |
2eaa055f JM |
4884 | flags[1].compat_flags, COMPAT); |
4885 | if (ret) | |
4886 | return ret; | |
4887 | ||
2ff7e61e | 4888 | ret = check_feature(fs_info, flags[0].compat_ro_flags, |
2eaa055f JM |
4889 | flags[1].compat_ro_flags, COMPAT_RO); |
4890 | if (ret) | |
4891 | return ret; | |
4892 | ||
2ff7e61e | 4893 | ret = check_feature(fs_info, flags[0].incompat_flags, |
2eaa055f JM |
4894 | flags[1].incompat_flags, INCOMPAT); |
4895 | if (ret) | |
4896 | return ret; | |
4897 | ||
7ab19625 DS |
4898 | ret = mnt_want_write_file(file); |
4899 | if (ret) | |
4900 | return ret; | |
4901 | ||
8051aa1a | 4902 | trans = btrfs_start_transaction(root, 0); |
7ab19625 DS |
4903 | if (IS_ERR(trans)) { |
4904 | ret = PTR_ERR(trans); | |
4905 | goto out_drop_write; | |
4906 | } | |
2eaa055f | 4907 | |
0b246afa | 4908 | spin_lock(&fs_info->super_lock); |
2eaa055f JM |
4909 | newflags = btrfs_super_compat_flags(super_block); |
4910 | newflags |= flags[0].compat_flags & flags[1].compat_flags; | |
4911 | newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); | |
4912 | btrfs_set_super_compat_flags(super_block, newflags); | |
4913 | ||
4914 | newflags = btrfs_super_compat_ro_flags(super_block); | |
4915 | newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; | |
4916 | newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); | |
4917 | btrfs_set_super_compat_ro_flags(super_block, newflags); | |
4918 | ||
4919 | newflags = btrfs_super_incompat_flags(super_block); | |
4920 | newflags |= flags[0].incompat_flags & flags[1].incompat_flags; | |
4921 | newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); | |
4922 | btrfs_set_super_incompat_flags(super_block, newflags); | |
0b246afa | 4923 | spin_unlock(&fs_info->super_lock); |
2eaa055f | 4924 | |
3a45bb20 | 4925 | ret = btrfs_commit_transaction(trans); |
7ab19625 DS |
4926 | out_drop_write: |
4927 | mnt_drop_write_file(file); | |
4928 | ||
4929 | return ret; | |
2eaa055f JM |
4930 | } |
4931 | ||
2351f431 JB |
4932 | static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) |
4933 | { | |
4934 | struct btrfs_ioctl_send_args *arg; | |
4935 | int ret; | |
4936 | ||
4937 | if (compat) { | |
4938 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) | |
4939 | struct btrfs_ioctl_send_args_32 args32; | |
4940 | ||
4941 | ret = copy_from_user(&args32, argp, sizeof(args32)); | |
4942 | if (ret) | |
4943 | return -EFAULT; | |
4944 | arg = kzalloc(sizeof(*arg), GFP_KERNEL); | |
4945 | if (!arg) | |
4946 | return -ENOMEM; | |
4947 | arg->send_fd = args32.send_fd; | |
4948 | arg->clone_sources_count = args32.clone_sources_count; | |
4949 | arg->clone_sources = compat_ptr(args32.clone_sources); | |
4950 | arg->parent_root = args32.parent_root; | |
4951 | arg->flags = args32.flags; | |
4952 | memcpy(arg->reserved, args32.reserved, | |
4953 | sizeof(args32.reserved)); | |
4954 | #else | |
4955 | return -ENOTTY; | |
4956 | #endif | |
4957 | } else { | |
4958 | arg = memdup_user(argp, sizeof(*arg)); | |
4959 | if (IS_ERR(arg)) | |
4960 | return PTR_ERR(arg); | |
4961 | } | |
4962 | ret = btrfs_ioctl_send(file, arg); | |
4963 | kfree(arg); | |
4964 | return ret; | |
4965 | } | |
4966 | ||
f46b5a66 CH |
4967 | long btrfs_ioctl(struct file *file, unsigned int |
4968 | cmd, unsigned long arg) | |
4969 | { | |
0b246afa JM |
4970 | struct inode *inode = file_inode(file); |
4971 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4972 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4bcabaa3 | 4973 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
4974 | |
4975 | switch (cmd) { | |
6cbff00f CH |
4976 | case FS_IOC_GETVERSION: |
4977 | return btrfs_ioctl_getversion(file, argp); | |
40cf931f | 4978 | case FS_IOC_GETFSLABEL: |
b929c1d8 | 4979 | return btrfs_ioctl_get_fslabel(fs_info, argp); |
40cf931f ES |
4980 | case FS_IOC_SETFSLABEL: |
4981 | return btrfs_ioctl_set_fslabel(file, argp); | |
f7039b1d | 4982 | case FITRIM: |
b929c1d8 | 4983 | return btrfs_ioctl_fitrim(fs_info, argp); |
f46b5a66 | 4984 | case BTRFS_IOC_SNAP_CREATE: |
fa0d2b9b | 4985 | return btrfs_ioctl_snap_create(file, argp, 0); |
fdfb1e4f | 4986 | case BTRFS_IOC_SNAP_CREATE_V2: |
fa0d2b9b | 4987 | return btrfs_ioctl_snap_create_v2(file, argp, 0); |
3de4586c | 4988 | case BTRFS_IOC_SUBVOL_CREATE: |
fa0d2b9b | 4989 | return btrfs_ioctl_snap_create(file, argp, 1); |
6f72c7e2 AJ |
4990 | case BTRFS_IOC_SUBVOL_CREATE_V2: |
4991 | return btrfs_ioctl_snap_create_v2(file, argp, 1); | |
76dda93c | 4992 | case BTRFS_IOC_SNAP_DESTROY: |
949964c9 MPS |
4993 | return btrfs_ioctl_snap_destroy(file, argp, false); |
4994 | case BTRFS_IOC_SNAP_DESTROY_V2: | |
4995 | return btrfs_ioctl_snap_destroy(file, argp, true); | |
0caa102d LZ |
4996 | case BTRFS_IOC_SUBVOL_GETFLAGS: |
4997 | return btrfs_ioctl_subvol_getflags(file, argp); | |
4998 | case BTRFS_IOC_SUBVOL_SETFLAGS: | |
4999 | return btrfs_ioctl_subvol_setflags(file, argp); | |
6ef5ed0d JB |
5000 | case BTRFS_IOC_DEFAULT_SUBVOL: |
5001 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 5002 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
5003 | return btrfs_ioctl_defrag(file, NULL); |
5004 | case BTRFS_IOC_DEFRAG_RANGE: | |
5005 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 5006 | case BTRFS_IOC_RESIZE: |
198605a8 | 5007 | return btrfs_ioctl_resize(file, argp); |
f46b5a66 | 5008 | case BTRFS_IOC_ADD_DEV: |
2ff7e61e | 5009 | return btrfs_ioctl_add_dev(fs_info, argp); |
f46b5a66 | 5010 | case BTRFS_IOC_RM_DEV: |
da24927b | 5011 | return btrfs_ioctl_rm_dev(file, argp); |
6b526ed7 AJ |
5012 | case BTRFS_IOC_RM_DEV_V2: |
5013 | return btrfs_ioctl_rm_dev_v2(file, argp); | |
475f6387 | 5014 | case BTRFS_IOC_FS_INFO: |
2ff7e61e | 5015 | return btrfs_ioctl_fs_info(fs_info, argp); |
475f6387 | 5016 | case BTRFS_IOC_DEV_INFO: |
2ff7e61e | 5017 | return btrfs_ioctl_dev_info(fs_info, argp); |
f46b5a66 | 5018 | case BTRFS_IOC_BALANCE: |
9ba1f6e4 | 5019 | return btrfs_ioctl_balance(file, NULL); |
ac8e9819 CM |
5020 | case BTRFS_IOC_TREE_SEARCH: |
5021 | return btrfs_ioctl_tree_search(file, argp); | |
cc68a8a5 GH |
5022 | case BTRFS_IOC_TREE_SEARCH_V2: |
5023 | return btrfs_ioctl_tree_search_v2(file, argp); | |
ac8e9819 CM |
5024 | case BTRFS_IOC_INO_LOOKUP: |
5025 | return btrfs_ioctl_ino_lookup(file, argp); | |
d7728c96 JS |
5026 | case BTRFS_IOC_INO_PATHS: |
5027 | return btrfs_ioctl_ino_to_path(root, argp); | |
5028 | case BTRFS_IOC_LOGICAL_INO: | |
d24a67b2 ZB |
5029 | return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); |
5030 | case BTRFS_IOC_LOGICAL_INO_V2: | |
5031 | return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); | |
1406e432 | 5032 | case BTRFS_IOC_SPACE_INFO: |
2ff7e61e | 5033 | return btrfs_ioctl_space_info(fs_info, argp); |
9b199859 FDBM |
5034 | case BTRFS_IOC_SYNC: { |
5035 | int ret; | |
5036 | ||
9db4dc24 | 5037 | ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false); |
9b199859 FDBM |
5038 | if (ret) |
5039 | return ret; | |
0b246afa | 5040 | ret = btrfs_sync_fs(inode->i_sb, 1); |
2fad4e83 DS |
5041 | /* |
5042 | * The transaction thread may want to do more work, | |
01327610 | 5043 | * namely it pokes the cleaner kthread that will start |
2fad4e83 DS |
5044 | * processing uncleaned subvols. |
5045 | */ | |
0b246afa | 5046 | wake_up_process(fs_info->transaction_kthread); |
9b199859 FDBM |
5047 | return ret; |
5048 | } | |
46204592 | 5049 | case BTRFS_IOC_START_SYNC: |
9a8c28be | 5050 | return btrfs_ioctl_start_sync(root, argp); |
46204592 | 5051 | case BTRFS_IOC_WAIT_SYNC: |
2ff7e61e | 5052 | return btrfs_ioctl_wait_sync(fs_info, argp); |
475f6387 | 5053 | case BTRFS_IOC_SCRUB: |
b8e95489 | 5054 | return btrfs_ioctl_scrub(file, argp); |
475f6387 | 5055 | case BTRFS_IOC_SCRUB_CANCEL: |
2ff7e61e | 5056 | return btrfs_ioctl_scrub_cancel(fs_info); |
475f6387 | 5057 | case BTRFS_IOC_SCRUB_PROGRESS: |
2ff7e61e | 5058 | return btrfs_ioctl_scrub_progress(fs_info, argp); |
c9e9f97b | 5059 | case BTRFS_IOC_BALANCE_V2: |
9ba1f6e4 | 5060 | return btrfs_ioctl_balance(file, argp); |
837d5b6e | 5061 | case BTRFS_IOC_BALANCE_CTL: |
2ff7e61e | 5062 | return btrfs_ioctl_balance_ctl(fs_info, arg); |
19a39dce | 5063 | case BTRFS_IOC_BALANCE_PROGRESS: |
2ff7e61e | 5064 | return btrfs_ioctl_balance_progress(fs_info, argp); |
8ea05e3a AB |
5065 | case BTRFS_IOC_SET_RECEIVED_SUBVOL: |
5066 | return btrfs_ioctl_set_received_subvol(file, argp); | |
abccd00f HM |
5067 | #ifdef CONFIG_64BIT |
5068 | case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: | |
5069 | return btrfs_ioctl_set_received_subvol_32(file, argp); | |
5070 | #endif | |
31db9f7c | 5071 | case BTRFS_IOC_SEND: |
2351f431 JB |
5072 | return _btrfs_ioctl_send(file, argp, false); |
5073 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) | |
5074 | case BTRFS_IOC_SEND_32: | |
5075 | return _btrfs_ioctl_send(file, argp, true); | |
5076 | #endif | |
c11d2c23 | 5077 | case BTRFS_IOC_GET_DEV_STATS: |
2ff7e61e | 5078 | return btrfs_ioctl_get_dev_stats(fs_info, argp); |
5d13a37b | 5079 | case BTRFS_IOC_QUOTA_CTL: |
905b0dda | 5080 | return btrfs_ioctl_quota_ctl(file, argp); |
5d13a37b | 5081 | case BTRFS_IOC_QGROUP_ASSIGN: |
905b0dda | 5082 | return btrfs_ioctl_qgroup_assign(file, argp); |
5d13a37b | 5083 | case BTRFS_IOC_QGROUP_CREATE: |
905b0dda | 5084 | return btrfs_ioctl_qgroup_create(file, argp); |
5d13a37b | 5085 | case BTRFS_IOC_QGROUP_LIMIT: |
905b0dda | 5086 | return btrfs_ioctl_qgroup_limit(file, argp); |
2f232036 JS |
5087 | case BTRFS_IOC_QUOTA_RESCAN: |
5088 | return btrfs_ioctl_quota_rescan(file, argp); | |
5089 | case BTRFS_IOC_QUOTA_RESCAN_STATUS: | |
b929c1d8 | 5090 | return btrfs_ioctl_quota_rescan_status(fs_info, argp); |
57254b6e | 5091 | case BTRFS_IOC_QUOTA_RESCAN_WAIT: |
b929c1d8 | 5092 | return btrfs_ioctl_quota_rescan_wait(fs_info, argp); |
3f6bcfbd | 5093 | case BTRFS_IOC_DEV_REPLACE: |
2ff7e61e | 5094 | return btrfs_ioctl_dev_replace(fs_info, argp); |
2eaa055f | 5095 | case BTRFS_IOC_GET_SUPPORTED_FEATURES: |
d5131b65 | 5096 | return btrfs_ioctl_get_supported_features(argp); |
2eaa055f | 5097 | case BTRFS_IOC_GET_FEATURES: |
b929c1d8 | 5098 | return btrfs_ioctl_get_features(fs_info, argp); |
2eaa055f JM |
5099 | case BTRFS_IOC_SET_FEATURES: |
5100 | return btrfs_ioctl_set_features(file, argp); | |
b64ec075 TM |
5101 | case BTRFS_IOC_GET_SUBVOL_INFO: |
5102 | return btrfs_ioctl_get_subvol_info(file, argp); | |
42e4b520 TM |
5103 | case BTRFS_IOC_GET_SUBVOL_ROOTREF: |
5104 | return btrfs_ioctl_get_subvol_rootref(file, argp); | |
23d0b79d TM |
5105 | case BTRFS_IOC_INO_LOOKUP_USER: |
5106 | return btrfs_ioctl_ino_lookup_user(file, argp); | |
14605409 BB |
5107 | case FS_IOC_ENABLE_VERITY: |
5108 | return fsverity_ioctl_enable(file, (const void __user *)argp); | |
5109 | case FS_IOC_MEASURE_VERITY: | |
5110 | return fsverity_ioctl_measure(file, argp); | |
f46b5a66 CH |
5111 | } |
5112 | ||
5113 | return -ENOTTY; | |
5114 | } | |
4c63c245 LD |
5115 | |
5116 | #ifdef CONFIG_COMPAT | |
5117 | long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) | |
5118 | { | |
2a362249 JM |
5119 | /* |
5120 | * These all access 32-bit values anyway so no further | |
5121 | * handling is necessary. | |
5122 | */ | |
4c63c245 | 5123 | switch (cmd) { |
4c63c245 LD |
5124 | case FS_IOC32_GETVERSION: |
5125 | cmd = FS_IOC_GETVERSION; | |
5126 | break; | |
4c63c245 LD |
5127 | } |
5128 | ||
5129 | return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); | |
5130 | } | |
5131 | #endif |