]>
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" |
f46b5a66 | 51 | |
abccd00f HM |
52 | #ifdef CONFIG_64BIT |
53 | /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI | |
54 | * structures are incorrect, as the timespec structure from userspace | |
55 | * is 4 bytes too small. We define these alternatives here to teach | |
56 | * the kernel about the 32-bit struct packing. | |
57 | */ | |
58 | struct btrfs_ioctl_timespec_32 { | |
59 | __u64 sec; | |
60 | __u32 nsec; | |
61 | } __attribute__ ((__packed__)); | |
62 | ||
63 | struct btrfs_ioctl_received_subvol_args_32 { | |
64 | char uuid[BTRFS_UUID_SIZE]; /* in */ | |
65 | __u64 stransid; /* in */ | |
66 | __u64 rtransid; /* out */ | |
67 | struct btrfs_ioctl_timespec_32 stime; /* in */ | |
68 | struct btrfs_ioctl_timespec_32 rtime; /* out */ | |
69 | __u64 flags; /* in */ | |
70 | __u64 reserved[16]; /* in */ | |
71 | } __attribute__ ((__packed__)); | |
72 | ||
73 | #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ | |
74 | struct btrfs_ioctl_received_subvol_args_32) | |
75 | #endif | |
76 | ||
2351f431 JB |
77 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) |
78 | struct btrfs_ioctl_send_args_32 { | |
79 | __s64 send_fd; /* in */ | |
80 | __u64 clone_sources_count; /* in */ | |
81 | compat_uptr_t clone_sources; /* in */ | |
82 | __u64 parent_root; /* in */ | |
83 | __u64 flags; /* in */ | |
84 | __u64 reserved[4]; /* in */ | |
85 | } __attribute__ ((__packed__)); | |
86 | ||
87 | #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \ | |
88 | struct btrfs_ioctl_send_args_32) | |
89 | #endif | |
abccd00f | 90 | |
6cbff00f | 91 | /* Mask out flags that are inappropriate for the given type of inode. */ |
1905a0f7 DS |
92 | static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode, |
93 | unsigned int flags) | |
6cbff00f | 94 | { |
1905a0f7 | 95 | if (S_ISDIR(inode->i_mode)) |
6cbff00f | 96 | return flags; |
1905a0f7 | 97 | else if (S_ISREG(inode->i_mode)) |
6cbff00f CH |
98 | return flags & ~FS_DIRSYNC_FL; |
99 | else | |
100 | return flags & (FS_NODUMP_FL | FS_NOATIME_FL); | |
101 | } | |
102 | ||
103 | /* | |
a157d4fd DS |
104 | * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS |
105 | * ioctl. | |
6cbff00f | 106 | */ |
77eea05e | 107 | static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode) |
6cbff00f CH |
108 | { |
109 | unsigned int iflags = 0; | |
77eea05e | 110 | u32 flags = binode->flags; |
14605409 | 111 | u32 ro_flags = binode->ro_flags; |
6cbff00f CH |
112 | |
113 | if (flags & BTRFS_INODE_SYNC) | |
114 | iflags |= FS_SYNC_FL; | |
115 | if (flags & BTRFS_INODE_IMMUTABLE) | |
116 | iflags |= FS_IMMUTABLE_FL; | |
117 | if (flags & BTRFS_INODE_APPEND) | |
118 | iflags |= FS_APPEND_FL; | |
119 | if (flags & BTRFS_INODE_NODUMP) | |
120 | iflags |= FS_NODUMP_FL; | |
121 | if (flags & BTRFS_INODE_NOATIME) | |
122 | iflags |= FS_NOATIME_FL; | |
123 | if (flags & BTRFS_INODE_DIRSYNC) | |
124 | iflags |= FS_DIRSYNC_FL; | |
d0092bdd LZ |
125 | if (flags & BTRFS_INODE_NODATACOW) |
126 | iflags |= FS_NOCOW_FL; | |
14605409 BB |
127 | if (ro_flags & BTRFS_INODE_RO_VERITY) |
128 | iflags |= FS_VERITY_FL; | |
d0092bdd | 129 | |
13f48dc9 | 130 | if (flags & BTRFS_INODE_NOCOMPRESS) |
d0092bdd | 131 | iflags |= FS_NOCOMP_FL; |
13f48dc9 ST |
132 | else if (flags & BTRFS_INODE_COMPRESS) |
133 | iflags |= FS_COMPR_FL; | |
6cbff00f CH |
134 | |
135 | return iflags; | |
136 | } | |
137 | ||
138 | /* | |
139 | * Update inode->i_flags based on the btrfs internal flags. | |
140 | */ | |
7b6a221e | 141 | void btrfs_sync_inode_flags_to_i_flags(struct inode *inode) |
6cbff00f | 142 | { |
5c57b8b6 | 143 | struct btrfs_inode *binode = BTRFS_I(inode); |
3cc79392 | 144 | unsigned int new_fl = 0; |
6cbff00f | 145 | |
5c57b8b6 | 146 | if (binode->flags & BTRFS_INODE_SYNC) |
3cc79392 | 147 | new_fl |= S_SYNC; |
5c57b8b6 | 148 | if (binode->flags & BTRFS_INODE_IMMUTABLE) |
3cc79392 | 149 | new_fl |= S_IMMUTABLE; |
5c57b8b6 | 150 | if (binode->flags & BTRFS_INODE_APPEND) |
3cc79392 | 151 | new_fl |= S_APPEND; |
5c57b8b6 | 152 | if (binode->flags & BTRFS_INODE_NOATIME) |
3cc79392 | 153 | new_fl |= S_NOATIME; |
5c57b8b6 | 154 | if (binode->flags & BTRFS_INODE_DIRSYNC) |
3cc79392 | 155 | new_fl |= S_DIRSYNC; |
14605409 BB |
156 | if (binode->ro_flags & BTRFS_INODE_RO_VERITY) |
157 | new_fl |= S_VERITY; | |
3cc79392 FM |
158 | |
159 | set_mask_bits(&inode->i_flags, | |
14605409 BB |
160 | S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC | |
161 | S_VERITY, new_fl); | |
6cbff00f CH |
162 | } |
163 | ||
f37c563b DS |
164 | /* |
165 | * Check if @flags are a supported and valid set of FS_*_FL flags and that | |
166 | * the old and new flags are not conflicting | |
167 | */ | |
168 | static int check_fsflags(unsigned int old_flags, unsigned int flags) | |
75e7cb7f LB |
169 | { |
170 | if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ | |
171 | FS_NOATIME_FL | FS_NODUMP_FL | \ | |
172 | FS_SYNC_FL | FS_DIRSYNC_FL | \ | |
e1e8fb6a LZ |
173 | FS_NOCOMP_FL | FS_COMPR_FL | |
174 | FS_NOCOW_FL)) | |
75e7cb7f LB |
175 | return -EOPNOTSUPP; |
176 | ||
f37c563b | 177 | /* COMPR and NOCOMP on new/old are valid */ |
75e7cb7f LB |
178 | if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) |
179 | return -EINVAL; | |
180 | ||
f37c563b DS |
181 | if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL)) |
182 | return -EINVAL; | |
183 | ||
184 | /* NOCOW and compression options are mutually exclusive */ | |
185 | if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL))) | |
186 | return -EINVAL; | |
187 | if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL))) | |
188 | return -EINVAL; | |
189 | ||
75e7cb7f LB |
190 | return 0; |
191 | } | |
192 | ||
d206e9c9 NA |
193 | static int check_fsflags_compatible(struct btrfs_fs_info *fs_info, |
194 | unsigned int flags) | |
195 | { | |
196 | if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL)) | |
197 | return -EPERM; | |
198 | ||
199 | return 0; | |
200 | } | |
201 | ||
97fc2977 MS |
202 | /* |
203 | * Set flags/xflags from the internal inode flags. The remaining items of | |
204 | * fsxattr are zeroed. | |
205 | */ | |
206 | int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa) | |
6cbff00f | 207 | { |
97fc2977 MS |
208 | struct btrfs_inode *binode = BTRFS_I(d_inode(dentry)); |
209 | ||
77eea05e | 210 | fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode)); |
97fc2977 MS |
211 | return 0; |
212 | } | |
213 | ||
214 | int btrfs_fileattr_set(struct user_namespace *mnt_userns, | |
215 | struct dentry *dentry, struct fileattr *fa) | |
216 | { | |
217 | struct inode *inode = d_inode(dentry); | |
0b246afa | 218 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
5c57b8b6 DS |
219 | struct btrfs_inode *binode = BTRFS_I(inode); |
220 | struct btrfs_root *root = binode->root; | |
6cbff00f | 221 | struct btrfs_trans_handle *trans; |
5aca2842 | 222 | unsigned int fsflags, old_fsflags; |
6cbff00f | 223 | int ret; |
ff9fef55 | 224 | const char *comp = NULL; |
f37c563b | 225 | u32 binode_flags; |
6cbff00f | 226 | |
b83cc969 LZ |
227 | if (btrfs_root_readonly(root)) |
228 | return -EROFS; | |
229 | ||
97fc2977 MS |
230 | if (fileattr_has_fsx(fa)) |
231 | return -EOPNOTSUPP; | |
e7848683 | 232 | |
97fc2977 | 233 | fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags); |
77eea05e | 234 | old_fsflags = btrfs_inode_flags_to_fsflags(binode); |
f37c563b DS |
235 | ret = check_fsflags(old_fsflags, fsflags); |
236 | if (ret) | |
97fc2977 | 237 | return ret; |
f37c563b | 238 | |
d206e9c9 NA |
239 | ret = check_fsflags_compatible(fs_info, fsflags); |
240 | if (ret) | |
97fc2977 | 241 | return ret; |
d206e9c9 | 242 | |
f37c563b | 243 | binode_flags = binode->flags; |
5c57b8b6 | 244 | if (fsflags & FS_SYNC_FL) |
d2b8fcfe | 245 | binode_flags |= BTRFS_INODE_SYNC; |
6cbff00f | 246 | else |
d2b8fcfe | 247 | binode_flags &= ~BTRFS_INODE_SYNC; |
5c57b8b6 | 248 | if (fsflags & FS_IMMUTABLE_FL) |
d2b8fcfe | 249 | binode_flags |= BTRFS_INODE_IMMUTABLE; |
6cbff00f | 250 | else |
d2b8fcfe | 251 | binode_flags &= ~BTRFS_INODE_IMMUTABLE; |
5c57b8b6 | 252 | if (fsflags & FS_APPEND_FL) |
d2b8fcfe | 253 | binode_flags |= BTRFS_INODE_APPEND; |
6cbff00f | 254 | else |
d2b8fcfe | 255 | binode_flags &= ~BTRFS_INODE_APPEND; |
5c57b8b6 | 256 | if (fsflags & FS_NODUMP_FL) |
d2b8fcfe | 257 | binode_flags |= BTRFS_INODE_NODUMP; |
6cbff00f | 258 | else |
d2b8fcfe | 259 | binode_flags &= ~BTRFS_INODE_NODUMP; |
5c57b8b6 | 260 | if (fsflags & FS_NOATIME_FL) |
d2b8fcfe | 261 | binode_flags |= BTRFS_INODE_NOATIME; |
6cbff00f | 262 | else |
d2b8fcfe | 263 | binode_flags &= ~BTRFS_INODE_NOATIME; |
97fc2977 MS |
264 | |
265 | /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */ | |
266 | if (!fa->flags_valid) { | |
267 | /* 1 item for the inode */ | |
268 | trans = btrfs_start_transaction(root, 1); | |
9b8a233b RH |
269 | if (IS_ERR(trans)) |
270 | return PTR_ERR(trans); | |
97fc2977 MS |
271 | goto update_flags; |
272 | } | |
273 | ||
5c57b8b6 | 274 | if (fsflags & FS_DIRSYNC_FL) |
d2b8fcfe | 275 | binode_flags |= BTRFS_INODE_DIRSYNC; |
6cbff00f | 276 | else |
d2b8fcfe | 277 | binode_flags &= ~BTRFS_INODE_DIRSYNC; |
5c57b8b6 | 278 | if (fsflags & FS_NOCOW_FL) { |
44e5194b | 279 | if (S_ISREG(inode->i_mode)) { |
7e97b8da DS |
280 | /* |
281 | * It's safe to turn csums off here, no extents exist. | |
282 | * Otherwise we want the flag to reflect the real COW | |
283 | * status of the file and will not set it. | |
284 | */ | |
285 | if (inode->i_size == 0) | |
d2b8fcfe AJ |
286 | binode_flags |= BTRFS_INODE_NODATACOW | |
287 | BTRFS_INODE_NODATASUM; | |
7e97b8da | 288 | } else { |
d2b8fcfe | 289 | binode_flags |= BTRFS_INODE_NODATACOW; |
7e97b8da DS |
290 | } |
291 | } else { | |
292 | /* | |
01327610 | 293 | * Revert back under same assumptions as above |
7e97b8da | 294 | */ |
44e5194b | 295 | if (S_ISREG(inode->i_mode)) { |
7e97b8da | 296 | if (inode->i_size == 0) |
d2b8fcfe AJ |
297 | binode_flags &= ~(BTRFS_INODE_NODATACOW | |
298 | BTRFS_INODE_NODATASUM); | |
7e97b8da | 299 | } else { |
d2b8fcfe | 300 | binode_flags &= ~BTRFS_INODE_NODATACOW; |
7e97b8da DS |
301 | } |
302 | } | |
6cbff00f | 303 | |
75e7cb7f LB |
304 | /* |
305 | * The COMPRESS flag can only be changed by users, while the NOCOMPRESS | |
306 | * flag may be changed automatically if compression code won't make | |
307 | * things smaller. | |
308 | */ | |
5c57b8b6 | 309 | if (fsflags & FS_NOCOMP_FL) { |
d2b8fcfe AJ |
310 | binode_flags &= ~BTRFS_INODE_COMPRESS; |
311 | binode_flags |= BTRFS_INODE_NOCOMPRESS; | |
5c57b8b6 | 312 | } else if (fsflags & FS_COMPR_FL) { |
63541927 | 313 | |
97fc2977 MS |
314 | if (IS_SWAPFILE(inode)) |
315 | return -ETXTBSY; | |
eede2bf3 | 316 | |
d2b8fcfe AJ |
317 | binode_flags |= BTRFS_INODE_COMPRESS; |
318 | binode_flags &= ~BTRFS_INODE_NOCOMPRESS; | |
63541927 | 319 | |
93370509 DS |
320 | comp = btrfs_compress_type2str(fs_info->compress_type); |
321 | if (!comp || comp[0] == 0) | |
322 | comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB); | |
ebcb904d | 323 | } else { |
d2b8fcfe | 324 | binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); |
75e7cb7f | 325 | } |
6cbff00f | 326 | |
ff9fef55 AJ |
327 | /* |
328 | * 1 for inode item | |
329 | * 2 for properties | |
330 | */ | |
331 | trans = btrfs_start_transaction(root, 3); | |
97fc2977 MS |
332 | if (IS_ERR(trans)) |
333 | return PTR_ERR(trans); | |
6cbff00f | 334 | |
ff9fef55 AJ |
335 | if (comp) { |
336 | ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp, | |
337 | strlen(comp), 0); | |
338 | if (ret) { | |
339 | btrfs_abort_transaction(trans, ret); | |
340 | goto out_end_trans; | |
341 | } | |
ff9fef55 AJ |
342 | } else { |
343 | ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL, | |
344 | 0, 0); | |
345 | if (ret && ret != -ENODATA) { | |
346 | btrfs_abort_transaction(trans, ret); | |
347 | goto out_end_trans; | |
348 | } | |
349 | } | |
350 | ||
97fc2977 | 351 | update_flags: |
d2b8fcfe | 352 | binode->flags = binode_flags; |
7b6a221e | 353 | btrfs_sync_inode_flags_to_i_flags(inode); |
0c4d2d95 | 354 | inode_inc_iversion(inode); |
c2050a45 | 355 | inode->i_ctime = current_time(inode); |
9a56fcd1 | 356 | ret = btrfs_update_inode(trans, root, BTRFS_I(inode)); |
6cbff00f | 357 | |
ff9fef55 | 358 | out_end_trans: |
3a45bb20 | 359 | btrfs_end_transaction(trans); |
2d4e6f6a | 360 | return ret; |
6cbff00f CH |
361 | } |
362 | ||
0d7ed32c DS |
363 | /* |
364 | * Start exclusive operation @type, return true on success | |
365 | */ | |
c3e1f96c GR |
366 | bool btrfs_exclop_start(struct btrfs_fs_info *fs_info, |
367 | enum btrfs_exclusive_operation type) | |
368 | { | |
0d7ed32c DS |
369 | bool ret = false; |
370 | ||
371 | spin_lock(&fs_info->super_lock); | |
372 | if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) { | |
373 | fs_info->exclusive_operation = type; | |
374 | ret = true; | |
375 | } | |
376 | spin_unlock(&fs_info->super_lock); | |
377 | ||
378 | return ret; | |
c3e1f96c GR |
379 | } |
380 | ||
578bda9e DS |
381 | /* |
382 | * Conditionally allow to enter the exclusive operation in case it's compatible | |
383 | * with the running one. This must be paired with btrfs_exclop_start_unlock and | |
384 | * btrfs_exclop_finish. | |
385 | * | |
386 | * Compatibility: | |
387 | * - the same type is already running | |
388 | * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller | |
389 | * must check the condition first that would allow none -> @type | |
390 | */ | |
391 | bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info, | |
392 | enum btrfs_exclusive_operation type) | |
393 | { | |
394 | spin_lock(&fs_info->super_lock); | |
395 | if (fs_info->exclusive_operation == type) | |
396 | return true; | |
397 | ||
398 | spin_unlock(&fs_info->super_lock); | |
399 | return false; | |
400 | } | |
401 | ||
402 | void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info) | |
403 | { | |
404 | spin_unlock(&fs_info->super_lock); | |
405 | } | |
406 | ||
c3e1f96c GR |
407 | void btrfs_exclop_finish(struct btrfs_fs_info *fs_info) |
408 | { | |
0d7ed32c | 409 | spin_lock(&fs_info->super_lock); |
c3e1f96c | 410 | WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE); |
0d7ed32c | 411 | spin_unlock(&fs_info->super_lock); |
66a2823c | 412 | sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation"); |
c3e1f96c GR |
413 | } |
414 | ||
6cbff00f CH |
415 | static int btrfs_ioctl_getversion(struct file *file, int __user *arg) |
416 | { | |
496ad9aa | 417 | struct inode *inode = file_inode(file); |
6cbff00f CH |
418 | |
419 | return put_user(inode->i_generation, arg); | |
420 | } | |
f46b5a66 | 421 | |
b929c1d8 MPS |
422 | static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info, |
423 | void __user *arg) | |
f7039b1d | 424 | { |
f7039b1d LD |
425 | struct btrfs_device *device; |
426 | struct request_queue *q; | |
427 | struct fstrim_range range; | |
428 | u64 minlen = ULLONG_MAX; | |
429 | u64 num_devices = 0; | |
430 | int ret; | |
431 | ||
432 | if (!capable(CAP_SYS_ADMIN)) | |
433 | return -EPERM; | |
434 | ||
1cb3dc3f NA |
435 | /* |
436 | * btrfs_trim_block_group() depends on space cache, which is not | |
437 | * available in zoned filesystem. So, disallow fitrim on a zoned | |
438 | * filesystem for now. | |
439 | */ | |
440 | if (btrfs_is_zoned(fs_info)) | |
441 | return -EOPNOTSUPP; | |
442 | ||
f35f06c3 FM |
443 | /* |
444 | * If the fs is mounted with nologreplay, which requires it to be | |
445 | * mounted in RO mode as well, we can not allow discard on free space | |
446 | * inside block groups, because log trees refer to extents that are not | |
447 | * pinned in a block group's free space cache (pinning the extents is | |
448 | * precisely the first phase of replaying a log tree). | |
449 | */ | |
450 | if (btrfs_test_opt(fs_info, NOLOGREPLAY)) | |
451 | return -EROFS; | |
452 | ||
1f78160c XG |
453 | rcu_read_lock(); |
454 | list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, | |
455 | dev_list) { | |
f7039b1d LD |
456 | if (!device->bdev) |
457 | continue; | |
458 | q = bdev_get_queue(device->bdev); | |
459 | if (blk_queue_discard(q)) { | |
460 | num_devices++; | |
50d0446e | 461 | minlen = min_t(u64, q->limits.discard_granularity, |
f7039b1d LD |
462 | minlen); |
463 | } | |
464 | } | |
1f78160c | 465 | rcu_read_unlock(); |
f4c697e6 | 466 | |
f7039b1d LD |
467 | if (!num_devices) |
468 | return -EOPNOTSUPP; | |
f7039b1d LD |
469 | if (copy_from_user(&range, arg, sizeof(range))) |
470 | return -EFAULT; | |
6ba9fc8e QW |
471 | |
472 | /* | |
473 | * NOTE: Don't truncate the range using super->total_bytes. Bytenr of | |
474 | * block group is in the logical address space, which can be any | |
475 | * sectorsize aligned bytenr in the range [0, U64_MAX]. | |
476 | */ | |
477 | if (range.len < fs_info->sb->s_blocksize) | |
f4c697e6 | 478 | return -EINVAL; |
f7039b1d LD |
479 | |
480 | range.minlen = max(range.minlen, minlen); | |
2ff7e61e | 481 | ret = btrfs_trim_fs(fs_info, &range); |
f7039b1d LD |
482 | if (ret < 0) |
483 | return ret; | |
484 | ||
485 | if (copy_to_user(arg, &range, sizeof(range))) | |
486 | return -EFAULT; | |
487 | ||
488 | return 0; | |
489 | } | |
490 | ||
e1f60a65 | 491 | int __pure btrfs_is_empty_uuid(u8 *uuid) |
dd5f9615 | 492 | { |
46e0f66a CM |
493 | int i; |
494 | ||
495 | for (i = 0; i < BTRFS_UUID_SIZE; i++) { | |
496 | if (uuid[i]) | |
497 | return 0; | |
498 | } | |
499 | return 1; | |
dd5f9615 SB |
500 | } |
501 | ||
4d4340c9 CB |
502 | static noinline int create_subvol(struct user_namespace *mnt_userns, |
503 | struct inode *dir, struct dentry *dentry, | |
52f75f4f | 504 | const char *name, int namelen, |
8696c533 | 505 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 506 | { |
0b246afa | 507 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); |
f46b5a66 CH |
508 | struct btrfs_trans_handle *trans; |
509 | struct btrfs_key key; | |
49a3c4d9 | 510 | struct btrfs_root_item *root_item; |
f46b5a66 CH |
511 | struct btrfs_inode_item *inode_item; |
512 | struct extent_buffer *leaf; | |
d5c12070 | 513 | struct btrfs_root *root = BTRFS_I(dir)->root; |
76dda93c | 514 | struct btrfs_root *new_root; |
d5c12070 | 515 | struct btrfs_block_rsv block_rsv; |
95582b00 | 516 | struct timespec64 cur_time = current_time(dir); |
5662344b | 517 | struct inode *inode; |
f46b5a66 CH |
518 | int ret; |
519 | int err; | |
2dfb1e43 | 520 | dev_t anon_dev = 0; |
f46b5a66 | 521 | u64 objectid; |
3de4586c | 522 | u64 index = 0; |
f46b5a66 | 523 | |
49a3c4d9 DS |
524 | root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); |
525 | if (!root_item) | |
526 | return -ENOMEM; | |
527 | ||
543068a2 | 528 | ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid); |
2fbe8c8a | 529 | if (ret) |
49a3c4d9 | 530 | goto fail_free; |
6a912213 | 531 | |
2dfb1e43 QW |
532 | ret = get_anon_bdev(&anon_dev); |
533 | if (ret < 0) | |
534 | goto fail_free; | |
535 | ||
e09fe2d2 QW |
536 | /* |
537 | * Don't create subvolume whose level is not zero. Or qgroup will be | |
01327610 | 538 | * screwed up since it assumes subvolume qgroup's level to be 0. |
e09fe2d2 | 539 | */ |
49a3c4d9 DS |
540 | if (btrfs_qgroup_level(objectid)) { |
541 | ret = -ENOSPC; | |
542 | goto fail_free; | |
543 | } | |
e09fe2d2 | 544 | |
d5c12070 | 545 | btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); |
9ed74f2d | 546 | /* |
d5c12070 MX |
547 | * The same as the snapshot creation, please see the comment |
548 | * of create_snapshot(). | |
9ed74f2d | 549 | */ |
c4c129db | 550 | ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false); |
d5c12070 | 551 | if (ret) |
49a3c4d9 | 552 | goto fail_free; |
d5c12070 MX |
553 | |
554 | trans = btrfs_start_transaction(root, 0); | |
555 | if (IS_ERR(trans)) { | |
556 | ret = PTR_ERR(trans); | |
e85fde51 | 557 | btrfs_subvolume_release_metadata(root, &block_rsv); |
49a3c4d9 | 558 | goto fail_free; |
d5c12070 MX |
559 | } |
560 | trans->block_rsv = &block_rsv; | |
561 | trans->bytes_reserved = block_rsv.size; | |
f46b5a66 | 562 | |
a9377422 | 563 | ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit); |
6f72c7e2 AJ |
564 | if (ret) |
565 | goto fail; | |
566 | ||
9631e4cc JB |
567 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
568 | BTRFS_NESTING_NORMAL); | |
8e8a1e31 JB |
569 | if (IS_ERR(leaf)) { |
570 | ret = PTR_ERR(leaf); | |
571 | goto fail; | |
572 | } | |
f46b5a66 | 573 | |
f46b5a66 CH |
574 | btrfs_mark_buffer_dirty(leaf); |
575 | ||
49a3c4d9 | 576 | inode_item = &root_item->inode; |
3cae210f QW |
577 | btrfs_set_stack_inode_generation(inode_item, 1); |
578 | btrfs_set_stack_inode_size(inode_item, 3); | |
579 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 580 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 581 | fs_info->nodesize); |
3cae210f | 582 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
f46b5a66 | 583 | |
49a3c4d9 DS |
584 | btrfs_set_root_flags(root_item, 0); |
585 | btrfs_set_root_limit(root_item, 0); | |
3cae210f | 586 | btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); |
08fe4db1 | 587 | |
49a3c4d9 DS |
588 | btrfs_set_root_bytenr(root_item, leaf->start); |
589 | btrfs_set_root_generation(root_item, trans->transid); | |
590 | btrfs_set_root_level(root_item, 0); | |
591 | btrfs_set_root_refs(root_item, 1); | |
592 | btrfs_set_root_used(root_item, leaf->len); | |
593 | btrfs_set_root_last_snapshot(root_item, 0); | |
f46b5a66 | 594 | |
49a3c4d9 DS |
595 | btrfs_set_root_generation_v2(root_item, |
596 | btrfs_root_generation(root_item)); | |
807fc790 | 597 | generate_random_guid(root_item->uuid); |
49a3c4d9 DS |
598 | btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); |
599 | btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); | |
600 | root_item->ctime = root_item->otime; | |
601 | btrfs_set_root_ctransid(root_item, trans->transid); | |
602 | btrfs_set_root_otransid(root_item, trans->transid); | |
f46b5a66 | 603 | |
925baedd | 604 | btrfs_tree_unlock(leaf); |
f46b5a66 | 605 | |
69948022 | 606 | btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID); |
f46b5a66 CH |
607 | |
608 | key.objectid = objectid; | |
5d4f98a2 | 609 | key.offset = 0; |
962a298f | 610 | key.type = BTRFS_ROOT_ITEM_KEY; |
0b246afa | 611 | ret = btrfs_insert_root(trans, fs_info->tree_root, &key, |
49a3c4d9 | 612 | root_item); |
67addf29 FM |
613 | if (ret) { |
614 | /* | |
615 | * Since we don't abort the transaction in this case, free the | |
616 | * tree block so that we don't leak space and leave the | |
617 | * filesystem in an inconsistent state (an extent item in the | |
618 | * extent tree without backreferences). Also no need to have | |
619 | * the tree block locked since it is not in any tree at this | |
620 | * point, so no other task can find it and use it. | |
621 | */ | |
622 | btrfs_free_tree_block(trans, root, leaf, 0, 1); | |
623 | free_extent_buffer(leaf); | |
f46b5a66 | 624 | goto fail; |
67addf29 FM |
625 | } |
626 | ||
627 | free_extent_buffer(leaf); | |
628 | leaf = NULL; | |
f46b5a66 | 629 | |
76dda93c | 630 | key.offset = (u64)-1; |
2dfb1e43 | 631 | new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev); |
79787eaa | 632 | if (IS_ERR(new_root)) { |
2dfb1e43 | 633 | free_anon_bdev(anon_dev); |
79787eaa | 634 | ret = PTR_ERR(new_root); |
66642832 | 635 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
636 | goto fail; |
637 | } | |
2dfb1e43 QW |
638 | /* Freeing will be done in btrfs_put_root() of new_root */ |
639 | anon_dev = 0; | |
76dda93c | 640 | |
221581e4 JB |
641 | ret = btrfs_record_root_in_trans(trans, new_root); |
642 | if (ret) { | |
643 | btrfs_put_root(new_root); | |
644 | btrfs_abort_transaction(trans, ret); | |
645 | goto fail; | |
646 | } | |
76dda93c | 647 | |
4d4340c9 | 648 | ret = btrfs_create_subvol_root(trans, new_root, root, mnt_userns); |
00246528 | 649 | btrfs_put_root(new_root); |
ce598979 MF |
650 | if (ret) { |
651 | /* We potentially lose an unused inode item here */ | |
66642832 | 652 | btrfs_abort_transaction(trans, ret); |
ce598979 MF |
653 | goto fail; |
654 | } | |
655 | ||
f46b5a66 CH |
656 | /* |
657 | * insert the directory item | |
658 | */ | |
877574e2 | 659 | ret = btrfs_set_inode_index(BTRFS_I(dir), &index); |
79787eaa | 660 | if (ret) { |
66642832 | 661 | btrfs_abort_transaction(trans, ret); |
79787eaa JM |
662 | goto fail; |
663 | } | |
3de4586c | 664 | |
684572df | 665 | ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key, |
3de4586c | 666 | BTRFS_FT_DIR, index); |
79787eaa | 667 | if (ret) { |
66642832 | 668 | btrfs_abort_transaction(trans, ret); |
f46b5a66 | 669 | goto fail; |
79787eaa | 670 | } |
0660b5af | 671 | |
6ef06d27 | 672 | btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); |
9a56fcd1 | 673 | ret = btrfs_update_inode(trans, root, BTRFS_I(dir)); |
c7e54b51 JB |
674 | if (ret) { |
675 | btrfs_abort_transaction(trans, ret); | |
676 | goto fail; | |
677 | } | |
52c26179 | 678 | |
6025c19f | 679 | ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid, |
4a0cc7ca | 680 | btrfs_ino(BTRFS_I(dir)), index, name, namelen); |
c7e54b51 JB |
681 | if (ret) { |
682 | btrfs_abort_transaction(trans, ret); | |
683 | goto fail; | |
684 | } | |
f46b5a66 | 685 | |
cdb345a8 | 686 | ret = btrfs_uuid_tree_add(trans, root_item->uuid, |
6bccf3ab | 687 | BTRFS_UUID_KEY_SUBVOL, objectid); |
dd5f9615 | 688 | if (ret) |
66642832 | 689 | btrfs_abort_transaction(trans, ret); |
dd5f9615 | 690 | |
f46b5a66 | 691 | fail: |
49a3c4d9 | 692 | kfree(root_item); |
d5c12070 MX |
693 | trans->block_rsv = NULL; |
694 | trans->bytes_reserved = 0; | |
e85fde51 | 695 | btrfs_subvolume_release_metadata(root, &block_rsv); |
de6e8200 | 696 | |
9babda9f | 697 | err = btrfs_commit_transaction(trans); |
f46b5a66 CH |
698 | if (err && !ret) |
699 | ret = err; | |
1a65e24b | 700 | |
5662344b TI |
701 | if (!ret) { |
702 | inode = btrfs_lookup_dentry(dir, dentry); | |
de6e8200 LB |
703 | if (IS_ERR(inode)) |
704 | return PTR_ERR(inode); | |
5662344b TI |
705 | d_instantiate(dentry, inode); |
706 | } | |
f46b5a66 | 707 | return ret; |
49a3c4d9 DS |
708 | |
709 | fail_free: | |
2dfb1e43 QW |
710 | if (anon_dev) |
711 | free_anon_bdev(anon_dev); | |
49a3c4d9 DS |
712 | kfree(root_item); |
713 | return ret; | |
f46b5a66 CH |
714 | } |
715 | ||
e9662f70 | 716 | static int create_snapshot(struct btrfs_root *root, struct inode *dir, |
9babda9f | 717 | struct dentry *dentry, bool readonly, |
e9662f70 | 718 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 719 | { |
0b246afa | 720 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); |
2e4bfab9 | 721 | struct inode *inode; |
f46b5a66 CH |
722 | struct btrfs_pending_snapshot *pending_snapshot; |
723 | struct btrfs_trans_handle *trans; | |
2e4bfab9 | 724 | int ret; |
f46b5a66 | 725 | |
92a7cc42 | 726 | if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) |
f46b5a66 CH |
727 | return -EINVAL; |
728 | ||
eede2bf3 OS |
729 | if (atomic_read(&root->nr_swapfiles)) { |
730 | btrfs_warn(fs_info, | |
731 | "cannot snapshot subvolume with active swapfile"); | |
732 | return -ETXTBSY; | |
733 | } | |
734 | ||
23269bf5 | 735 | pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); |
a1ee7362 DS |
736 | if (!pending_snapshot) |
737 | return -ENOMEM; | |
738 | ||
2dfb1e43 QW |
739 | ret = get_anon_bdev(&pending_snapshot->anon_dev); |
740 | if (ret < 0) | |
741 | goto free_pending; | |
b0c0ea63 | 742 | pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), |
23269bf5 | 743 | GFP_KERNEL); |
8546b570 DS |
744 | pending_snapshot->path = btrfs_alloc_path(); |
745 | if (!pending_snapshot->root_item || !pending_snapshot->path) { | |
b0c0ea63 DS |
746 | ret = -ENOMEM; |
747 | goto free_pending; | |
748 | } | |
749 | ||
66d8f3dd MX |
750 | btrfs_init_block_rsv(&pending_snapshot->block_rsv, |
751 | BTRFS_BLOCK_RSV_TEMP); | |
d5c12070 MX |
752 | /* |
753 | * 1 - parent dir inode | |
754 | * 2 - dir entries | |
755 | * 1 - root item | |
756 | * 2 - root ref/backref | |
757 | * 1 - root of snapshot | |
dd5f9615 | 758 | * 1 - UUID item |
d5c12070 MX |
759 | */ |
760 | ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, | |
dd5f9615 | 761 | &pending_snapshot->block_rsv, 8, |
ee3441b4 | 762 | false); |
d5c12070 | 763 | if (ret) |
c11fbb6e | 764 | goto free_pending; |
d5c12070 | 765 | |
3de4586c | 766 | pending_snapshot->dentry = dentry; |
f46b5a66 | 767 | pending_snapshot->root = root; |
b83cc969 | 768 | pending_snapshot->readonly = readonly; |
e9662f70 | 769 | pending_snapshot->dir = dir; |
8696c533 | 770 | pending_snapshot->inherit = inherit; |
a22285a6 | 771 | |
d5c12070 | 772 | trans = btrfs_start_transaction(root, 0); |
a22285a6 YZ |
773 | if (IS_ERR(trans)) { |
774 | ret = PTR_ERR(trans); | |
775 | goto fail; | |
776 | } | |
777 | ||
0b246afa | 778 | spin_lock(&fs_info->trans_lock); |
f46b5a66 CH |
779 | list_add(&pending_snapshot->list, |
780 | &trans->transaction->pending_snapshots); | |
0b246afa | 781 | spin_unlock(&fs_info->trans_lock); |
9babda9f NB |
782 | |
783 | ret = btrfs_commit_transaction(trans); | |
aec8030a | 784 | if (ret) |
c37b2b62 | 785 | goto fail; |
a22285a6 YZ |
786 | |
787 | ret = pending_snapshot->error; | |
788 | if (ret) | |
789 | goto fail; | |
790 | ||
d3797308 CM |
791 | ret = btrfs_orphan_cleanup(pending_snapshot->snap); |
792 | if (ret) | |
793 | goto fail; | |
794 | ||
2b0143b5 | 795 | inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); |
2e4bfab9 YZ |
796 | if (IS_ERR(inode)) { |
797 | ret = PTR_ERR(inode); | |
798 | goto fail; | |
799 | } | |
5662344b | 800 | |
2e4bfab9 YZ |
801 | d_instantiate(dentry, inode); |
802 | ret = 0; | |
2dfb1e43 | 803 | pending_snapshot->anon_dev = 0; |
2e4bfab9 | 804 | fail: |
2dfb1e43 QW |
805 | /* Prevent double freeing of anon_dev */ |
806 | if (ret && pending_snapshot->snap) | |
807 | pending_snapshot->snap->anon_dev = 0; | |
00246528 | 808 | btrfs_put_root(pending_snapshot->snap); |
e85fde51 | 809 | btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv); |
b0c0ea63 | 810 | free_pending: |
2dfb1e43 QW |
811 | if (pending_snapshot->anon_dev) |
812 | free_anon_bdev(pending_snapshot->anon_dev); | |
b0c0ea63 | 813 | kfree(pending_snapshot->root_item); |
8546b570 | 814 | btrfs_free_path(pending_snapshot->path); |
a1ee7362 DS |
815 | kfree(pending_snapshot); |
816 | ||
f46b5a66 CH |
817 | return ret; |
818 | } | |
819 | ||
4260f7c7 SW |
820 | /* copy of may_delete in fs/namei.c() |
821 | * Check whether we can remove a link victim from directory dir, check | |
822 | * whether the type of victim is right. | |
823 | * 1. We can't do it if dir is read-only (done in permission()) | |
824 | * 2. We should have write and exec permissions on dir | |
825 | * 3. We can't remove anything from append-only dir | |
826 | * 4. We can't do anything with immutable dir (done in permission()) | |
827 | * 5. If the sticky bit on dir is set we should either | |
828 | * a. be owner of dir, or | |
829 | * b. be owner of victim, or | |
830 | * c. have CAP_FOWNER capability | |
01327610 | 831 | * 6. If the victim is append-only or immutable we can't do anything with |
4260f7c7 SW |
832 | * links pointing to it. |
833 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
834 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
835 | * 9. We can't remove a root or mountpoint. | |
836 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
837 | * nfs_async_unlink(). | |
838 | */ | |
839 | ||
c4ed533b CB |
840 | static int btrfs_may_delete(struct user_namespace *mnt_userns, |
841 | struct inode *dir, struct dentry *victim, int isdir) | |
4260f7c7 SW |
842 | { |
843 | int error; | |
844 | ||
2b0143b5 | 845 | if (d_really_is_negative(victim)) |
4260f7c7 SW |
846 | return -ENOENT; |
847 | ||
2b0143b5 | 848 | BUG_ON(d_inode(victim->d_parent) != dir); |
4fa6b5ec | 849 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
4260f7c7 | 850 | |
c4ed533b | 851 | error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); |
4260f7c7 SW |
852 | if (error) |
853 | return error; | |
854 | if (IS_APPEND(dir)) | |
855 | return -EPERM; | |
c4ed533b | 856 | if (check_sticky(mnt_userns, dir, d_inode(victim)) || |
ba73d987 CB |
857 | IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) || |
858 | IS_SWAPFILE(d_inode(victim))) | |
4260f7c7 SW |
859 | return -EPERM; |
860 | if (isdir) { | |
e36cb0b8 | 861 | if (!d_is_dir(victim)) |
4260f7c7 SW |
862 | return -ENOTDIR; |
863 | if (IS_ROOT(victim)) | |
864 | return -EBUSY; | |
e36cb0b8 | 865 | } else if (d_is_dir(victim)) |
4260f7c7 SW |
866 | return -EISDIR; |
867 | if (IS_DEADDIR(dir)) | |
868 | return -ENOENT; | |
869 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
870 | return -EBUSY; | |
871 | return 0; | |
872 | } | |
873 | ||
cb8e7090 | 874 | /* copy of may_create in fs/namei.c() */ |
4d4340c9 CB |
875 | static inline int btrfs_may_create(struct user_namespace *mnt_userns, |
876 | struct inode *dir, struct dentry *child) | |
cb8e7090 | 877 | { |
2b0143b5 | 878 | if (d_really_is_positive(child)) |
cb8e7090 CH |
879 | return -EEXIST; |
880 | if (IS_DEADDIR(dir)) | |
881 | return -ENOENT; | |
4d4340c9 | 882 | if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns)) |
5474bf40 | 883 | return -EOVERFLOW; |
4d4340c9 | 884 | return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); |
cb8e7090 CH |
885 | } |
886 | ||
887 | /* | |
888 | * Create a new subvolume below @parent. This is largely modeled after | |
889 | * sys_mkdirat and vfs_mkdir, but we only do a single component lookup | |
890 | * inside this filesystem so it's quite a bit simpler. | |
891 | */ | |
92872094 | 892 | static noinline int btrfs_mksubvol(const struct path *parent, |
4d4340c9 | 893 | struct user_namespace *mnt_userns, |
52f75f4f | 894 | const char *name, int namelen, |
72fd032e | 895 | struct btrfs_root *snap_src, |
9babda9f | 896 | bool readonly, |
8696c533 | 897 | struct btrfs_qgroup_inherit *inherit) |
cb8e7090 | 898 | { |
0b246afa JM |
899 | struct inode *dir = d_inode(parent->dentry); |
900 | struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); | |
cb8e7090 CH |
901 | struct dentry *dentry; |
902 | int error; | |
903 | ||
00235411 AV |
904 | error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); |
905 | if (error == -EINTR) | |
906 | return error; | |
cb8e7090 | 907 | |
4d4340c9 | 908 | dentry = lookup_one(mnt_userns, name, parent->dentry, namelen); |
cb8e7090 CH |
909 | error = PTR_ERR(dentry); |
910 | if (IS_ERR(dentry)) | |
911 | goto out_unlock; | |
912 | ||
4d4340c9 | 913 | error = btrfs_may_create(mnt_userns, dir, dentry); |
cb8e7090 | 914 | if (error) |
a874a63e | 915 | goto out_dput; |
cb8e7090 | 916 | |
9c52057c CM |
917 | /* |
918 | * even if this name doesn't exist, we may get hash collisions. | |
919 | * check for them now when we can safely fail | |
920 | */ | |
921 | error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, | |
922 | dir->i_ino, name, | |
923 | namelen); | |
924 | if (error) | |
925 | goto out_dput; | |
926 | ||
0b246afa | 927 | down_read(&fs_info->subvol_sem); |
76dda93c YZ |
928 | |
929 | if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) | |
930 | goto out_up_read; | |
931 | ||
9babda9f NB |
932 | if (snap_src) |
933 | error = create_snapshot(snap_src, dir, dentry, readonly, inherit); | |
934 | else | |
4d4340c9 | 935 | error = create_subvol(mnt_userns, dir, dentry, name, namelen, inherit); |
9babda9f | 936 | |
76dda93c YZ |
937 | if (!error) |
938 | fsnotify_mkdir(dir, dentry); | |
939 | out_up_read: | |
0b246afa | 940 | up_read(&fs_info->subvol_sem); |
cb8e7090 CH |
941 | out_dput: |
942 | dput(dentry); | |
943 | out_unlock: | |
64708539 | 944 | btrfs_inode_unlock(dir, 0); |
cb8e7090 CH |
945 | return error; |
946 | } | |
947 | ||
c11fbb6e | 948 | static noinline int btrfs_mksnapshot(const struct path *parent, |
4d4340c9 | 949 | struct user_namespace *mnt_userns, |
c11fbb6e RK |
950 | const char *name, int namelen, |
951 | struct btrfs_root *root, | |
952 | bool readonly, | |
953 | struct btrfs_qgroup_inherit *inherit) | |
954 | { | |
955 | int ret; | |
956 | bool snapshot_force_cow = false; | |
957 | ||
958 | /* | |
959 | * Force new buffered writes to reserve space even when NOCOW is | |
960 | * possible. This is to avoid later writeback (running dealloc) to | |
961 | * fallback to COW mode and unexpectedly fail with ENOSPC. | |
962 | */ | |
963 | btrfs_drew_read_lock(&root->snapshot_lock); | |
964 | ||
f9baa501 | 965 | ret = btrfs_start_delalloc_snapshot(root, false); |
c11fbb6e RK |
966 | if (ret) |
967 | goto out; | |
968 | ||
969 | /* | |
970 | * All previous writes have started writeback in NOCOW mode, so now | |
971 | * we force future writes to fallback to COW mode during snapshot | |
972 | * creation. | |
973 | */ | |
974 | atomic_inc(&root->snapshot_force_cow); | |
975 | snapshot_force_cow = true; | |
976 | ||
977 | btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); | |
978 | ||
4d4340c9 | 979 | ret = btrfs_mksubvol(parent, mnt_userns, name, namelen, |
c11fbb6e RK |
980 | root, readonly, inherit); |
981 | out: | |
982 | if (snapshot_force_cow) | |
983 | atomic_dec(&root->snapshot_force_cow); | |
984 | btrfs_drew_read_unlock(&root->snapshot_lock); | |
985 | return ret; | |
986 | } | |
987 | ||
4cb5300b CM |
988 | /* |
989 | * When we're defragging a range, we don't want to kick it off again | |
990 | * if it is really just waiting for delalloc to send it down. | |
991 | * If we find a nice big extent or delalloc range for the bytes in the | |
992 | * file you want to defrag, we return 0 to let you know to skip this | |
993 | * part of the file | |
994 | */ | |
aab110ab | 995 | static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) |
4cb5300b CM |
996 | { |
997 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
998 | struct extent_map *em = NULL; | |
999 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
1000 | u64 end; | |
1001 | ||
1002 | read_lock(&em_tree->lock); | |
09cbfeaf | 1003 | em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); |
4cb5300b CM |
1004 | read_unlock(&em_tree->lock); |
1005 | ||
1006 | if (em) { | |
1007 | end = extent_map_end(em); | |
1008 | free_extent_map(em); | |
1009 | if (end - offset > thresh) | |
1010 | return 0; | |
1011 | } | |
1012 | /* if we already have a nice delalloc here, just stop */ | |
1013 | thresh /= 2; | |
1014 | end = count_range_bits(io_tree, &offset, offset + thresh, | |
1015 | thresh, EXTENT_DELALLOC, 1); | |
1016 | if (end >= thresh) | |
1017 | return 0; | |
1018 | return 1; | |
1019 | } | |
1020 | ||
1021 | /* | |
1022 | * helper function to walk through a file and find extents | |
1023 | * newer than a specific transid, and smaller than thresh. | |
1024 | * | |
1025 | * This is used by the defragging code to find new and small | |
1026 | * extents | |
1027 | */ | |
1028 | static int find_new_extents(struct btrfs_root *root, | |
1029 | struct inode *inode, u64 newer_than, | |
aab110ab | 1030 | u64 *off, u32 thresh) |
4cb5300b CM |
1031 | { |
1032 | struct btrfs_path *path; | |
1033 | struct btrfs_key min_key; | |
4cb5300b CM |
1034 | struct extent_buffer *leaf; |
1035 | struct btrfs_file_extent_item *extent; | |
1036 | int type; | |
1037 | int ret; | |
4a0cc7ca | 1038 | u64 ino = btrfs_ino(BTRFS_I(inode)); |
4cb5300b CM |
1039 | |
1040 | path = btrfs_alloc_path(); | |
1041 | if (!path) | |
1042 | return -ENOMEM; | |
1043 | ||
a4689d2b | 1044 | min_key.objectid = ino; |
4cb5300b CM |
1045 | min_key.type = BTRFS_EXTENT_DATA_KEY; |
1046 | min_key.offset = *off; | |
1047 | ||
67871254 | 1048 | while (1) { |
6174d3cb | 1049 | ret = btrfs_search_forward(root, &min_key, path, newer_than); |
4cb5300b CM |
1050 | if (ret != 0) |
1051 | goto none; | |
f094c9bd | 1052 | process_slot: |
a4689d2b | 1053 | if (min_key.objectid != ino) |
4cb5300b CM |
1054 | goto none; |
1055 | if (min_key.type != BTRFS_EXTENT_DATA_KEY) | |
1056 | goto none; | |
1057 | ||
1058 | leaf = path->nodes[0]; | |
1059 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
1060 | struct btrfs_file_extent_item); | |
1061 | ||
1062 | type = btrfs_file_extent_type(leaf, extent); | |
1063 | if (type == BTRFS_FILE_EXTENT_REG && | |
1064 | btrfs_file_extent_num_bytes(leaf, extent) < thresh && | |
1065 | check_defrag_in_cache(inode, min_key.offset, thresh)) { | |
1066 | *off = min_key.offset; | |
1067 | btrfs_free_path(path); | |
1068 | return 0; | |
1069 | } | |
1070 | ||
f094c9bd FM |
1071 | path->slots[0]++; |
1072 | if (path->slots[0] < btrfs_header_nritems(leaf)) { | |
1073 | btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); | |
1074 | goto process_slot; | |
1075 | } | |
1076 | ||
4cb5300b CM |
1077 | if (min_key.offset == (u64)-1) |
1078 | goto none; | |
1079 | ||
1080 | min_key.offset++; | |
1081 | btrfs_release_path(path); | |
1082 | } | |
1083 | none: | |
1084 | btrfs_free_path(path); | |
1085 | return -ENOENT; | |
1086 | } | |
1087 | ||
6c282eb4 | 1088 | static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) |
17ce6ef8 LB |
1089 | { |
1090 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
6c282eb4 LZ |
1091 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
1092 | struct extent_map *em; | |
09cbfeaf | 1093 | u64 len = PAGE_SIZE; |
17ce6ef8 | 1094 | |
6c282eb4 LZ |
1095 | /* |
1096 | * hopefully we have this extent in the tree already, try without | |
1097 | * the full extent lock | |
1098 | */ | |
17ce6ef8 | 1099 | read_lock(&em_tree->lock); |
6c282eb4 | 1100 | em = lookup_extent_mapping(em_tree, start, len); |
17ce6ef8 LB |
1101 | read_unlock(&em_tree->lock); |
1102 | ||
6c282eb4 | 1103 | if (!em) { |
308d9800 FM |
1104 | struct extent_state *cached = NULL; |
1105 | u64 end = start + len - 1; | |
1106 | ||
6c282eb4 | 1107 | /* get the big lock and read metadata off disk */ |
ff13db41 | 1108 | lock_extent_bits(io_tree, start, end, &cached); |
39b07b5d | 1109 | em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len); |
e43bbe5e | 1110 | unlock_extent_cached(io_tree, start, end, &cached); |
6c282eb4 LZ |
1111 | |
1112 | if (IS_ERR(em)) | |
1113 | return NULL; | |
1114 | } | |
1115 | ||
1116 | return em; | |
1117 | } | |
17ce6ef8 | 1118 | |
6c282eb4 LZ |
1119 | static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) |
1120 | { | |
1121 | struct extent_map *next; | |
1122 | bool ret = true; | |
1123 | ||
1124 | /* this is the last extent */ | |
1125 | if (em->start + em->len >= i_size_read(inode)) | |
1126 | return false; | |
1127 | ||
1128 | next = defrag_lookup_extent(inode, em->start + em->len); | |
e9512d72 CM |
1129 | if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) |
1130 | ret = false; | |
1131 | else if ((em->block_start + em->block_len == next->block_start) && | |
ee22184b | 1132 | (em->block_len > SZ_128K && next->block_len > SZ_128K)) |
6c282eb4 LZ |
1133 | ret = false; |
1134 | ||
1135 | free_extent_map(next); | |
17ce6ef8 LB |
1136 | return ret; |
1137 | } | |
1138 | ||
aab110ab | 1139 | static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, |
a43a2111 AM |
1140 | u64 *last_len, u64 *skip, u64 *defrag_end, |
1141 | int compress) | |
940100a4 | 1142 | { |
6c282eb4 | 1143 | struct extent_map *em; |
940100a4 | 1144 | int ret = 1; |
6c282eb4 | 1145 | bool next_mergeable = true; |
4a3560c4 | 1146 | bool prev_mergeable = true; |
940100a4 CM |
1147 | |
1148 | /* | |
008873ea | 1149 | * make sure that once we start defragging an extent, we keep on |
940100a4 CM |
1150 | * defragging it |
1151 | */ | |
1152 | if (start < *defrag_end) | |
1153 | return 1; | |
1154 | ||
1155 | *skip = 0; | |
1156 | ||
6c282eb4 LZ |
1157 | em = defrag_lookup_extent(inode, start); |
1158 | if (!em) | |
1159 | return 0; | |
940100a4 CM |
1160 | |
1161 | /* this will cover holes, and inline extents */ | |
17ce6ef8 | 1162 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
940100a4 | 1163 | ret = 0; |
17ce6ef8 LB |
1164 | goto out; |
1165 | } | |
1166 | ||
4a3560c4 LB |
1167 | if (!*defrag_end) |
1168 | prev_mergeable = false; | |
1169 | ||
6c282eb4 | 1170 | next_mergeable = defrag_check_next_extent(inode, em); |
940100a4 | 1171 | /* |
6c282eb4 LZ |
1172 | * we hit a real extent, if it is big or the next extent is not a |
1173 | * real extent, don't bother defragging it | |
940100a4 | 1174 | */ |
a43a2111 | 1175 | if (!compress && (*last_len == 0 || *last_len >= thresh) && |
4a3560c4 | 1176 | (em->len >= thresh || (!next_mergeable && !prev_mergeable))) |
940100a4 | 1177 | ret = 0; |
17ce6ef8 | 1178 | out: |
940100a4 CM |
1179 | /* |
1180 | * last_len ends up being a counter of how many bytes we've defragged. | |
1181 | * every time we choose not to defrag an extent, we reset *last_len | |
1182 | * so that the next tiny extent will force a defrag. | |
1183 | * | |
1184 | * The end result of this is that tiny extents before a single big | |
1185 | * extent will force at least part of that big extent to be defragged. | |
1186 | */ | |
1187 | if (ret) { | |
940100a4 CM |
1188 | *defrag_end = extent_map_end(em); |
1189 | } else { | |
1190 | *last_len = 0; | |
1191 | *skip = extent_map_end(em); | |
1192 | *defrag_end = 0; | |
1193 | } | |
1194 | ||
1195 | free_extent_map(em); | |
1196 | return ret; | |
1197 | } | |
1198 | ||
4cb5300b CM |
1199 | /* |
1200 | * it doesn't do much good to defrag one or two pages | |
1201 | * at a time. This pulls in a nice chunk of pages | |
1202 | * to COW and defrag. | |
1203 | * | |
1204 | * It also makes sure the delalloc code has enough | |
1205 | * dirty data to avoid making new small extents as part | |
1206 | * of the defrag | |
1207 | * | |
1208 | * It's a good idea to start RA on this range | |
1209 | * before calling this. | |
1210 | */ | |
1211 | static int cluster_pages_for_defrag(struct inode *inode, | |
1212 | struct page **pages, | |
1213 | unsigned long start_index, | |
c41570c9 | 1214 | unsigned long num_pages) |
f46b5a66 | 1215 | { |
4cb5300b CM |
1216 | unsigned long file_end; |
1217 | u64 isize = i_size_read(inode); | |
1218 | u64 page_start; | |
1219 | u64 page_end; | |
1f12bd06 | 1220 | u64 page_cnt; |
a1fbc675 | 1221 | u64 start = (u64)start_index << PAGE_SHIFT; |
7f458a38 | 1222 | u64 search_start; |
4cb5300b CM |
1223 | int ret; |
1224 | int i; | |
1225 | int i_done; | |
3eaa2885 | 1226 | struct btrfs_ordered_extent *ordered; |
4cb5300b | 1227 | struct extent_state *cached_state = NULL; |
600a45e1 | 1228 | struct extent_io_tree *tree; |
364ecf36 | 1229 | struct extent_changeset *data_reserved = NULL; |
3b16a4e3 | 1230 | gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); |
4cb5300b | 1231 | |
09cbfeaf | 1232 | file_end = (isize - 1) >> PAGE_SHIFT; |
1f12bd06 LB |
1233 | if (!isize || start_index > file_end) |
1234 | return 0; | |
1235 | ||
1236 | page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); | |
4cb5300b | 1237 | |
e5b7231e | 1238 | ret = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved, |
a1fbc675 | 1239 | start, page_cnt << PAGE_SHIFT); |
4cb5300b CM |
1240 | if (ret) |
1241 | return ret; | |
4cb5300b | 1242 | i_done = 0; |
600a45e1 | 1243 | tree = &BTRFS_I(inode)->io_tree; |
4cb5300b CM |
1244 | |
1245 | /* step one, lock all the pages */ | |
1f12bd06 | 1246 | for (i = 0; i < page_cnt; i++) { |
4cb5300b | 1247 | struct page *page; |
600a45e1 | 1248 | again: |
a94733d0 | 1249 | page = find_or_create_page(inode->i_mapping, |
600a45e1 | 1250 | start_index + i, mask); |
4cb5300b CM |
1251 | if (!page) |
1252 | break; | |
1253 | ||
32443de3 QW |
1254 | ret = set_page_extent_mapped(page); |
1255 | if (ret < 0) { | |
1256 | unlock_page(page); | |
1257 | put_page(page); | |
1258 | break; | |
1259 | } | |
1260 | ||
600a45e1 | 1261 | page_start = page_offset(page); |
09cbfeaf | 1262 | page_end = page_start + PAGE_SIZE - 1; |
600a45e1 | 1263 | while (1) { |
308d9800 | 1264 | lock_extent_bits(tree, page_start, page_end, |
ff13db41 | 1265 | &cached_state); |
c3504372 | 1266 | ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode), |
600a45e1 | 1267 | page_start); |
308d9800 | 1268 | unlock_extent_cached(tree, page_start, page_end, |
e43bbe5e | 1269 | &cached_state); |
600a45e1 MX |
1270 | if (!ordered) |
1271 | break; | |
1272 | ||
1273 | unlock_page(page); | |
c0a43603 | 1274 | btrfs_start_ordered_extent(ordered, 1); |
600a45e1 MX |
1275 | btrfs_put_ordered_extent(ordered); |
1276 | lock_page(page); | |
1f12bd06 LB |
1277 | /* |
1278 | * we unlocked the page above, so we need check if | |
1279 | * it was released or not. | |
1280 | */ | |
1281 | if (page->mapping != inode->i_mapping) { | |
1282 | unlock_page(page); | |
09cbfeaf | 1283 | put_page(page); |
1f12bd06 LB |
1284 | goto again; |
1285 | } | |
600a45e1 MX |
1286 | } |
1287 | ||
4cb5300b CM |
1288 | if (!PageUptodate(page)) { |
1289 | btrfs_readpage(NULL, page); | |
1290 | lock_page(page); | |
1291 | if (!PageUptodate(page)) { | |
1292 | unlock_page(page); | |
09cbfeaf | 1293 | put_page(page); |
4cb5300b CM |
1294 | ret = -EIO; |
1295 | break; | |
1296 | } | |
1297 | } | |
600a45e1 | 1298 | |
600a45e1 MX |
1299 | if (page->mapping != inode->i_mapping) { |
1300 | unlock_page(page); | |
09cbfeaf | 1301 | put_page(page); |
600a45e1 MX |
1302 | goto again; |
1303 | } | |
1304 | ||
4cb5300b CM |
1305 | pages[i] = page; |
1306 | i_done++; | |
1307 | } | |
1308 | if (!i_done || ret) | |
1309 | goto out; | |
1310 | ||
1751e8a6 | 1311 | if (!(inode->i_sb->s_flags & SB_ACTIVE)) |
4cb5300b CM |
1312 | goto out; |
1313 | ||
1314 | /* | |
1315 | * so now we have a nice long stream of locked | |
1316 | * and up to date pages, lets wait on them | |
1317 | */ | |
1318 | for (i = 0; i < i_done; i++) | |
1319 | wait_on_page_writeback(pages[i]); | |
1320 | ||
1321 | page_start = page_offset(pages[0]); | |
09cbfeaf | 1322 | page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; |
4cb5300b CM |
1323 | |
1324 | lock_extent_bits(&BTRFS_I(inode)->io_tree, | |
ff13db41 | 1325 | page_start, page_end - 1, &cached_state); |
7f458a38 FM |
1326 | |
1327 | /* | |
1328 | * When defragmenting we skip ranges that have holes or inline extents, | |
1329 | * (check should_defrag_range()), to avoid unnecessary IO and wasting | |
1330 | * space. At btrfs_defrag_file(), we check if a range should be defragged | |
1331 | * before locking the inode and then, if it should, we trigger a sync | |
1332 | * page cache readahead - we lock the inode only after that to avoid | |
1333 | * blocking for too long other tasks that possibly want to operate on | |
1334 | * other file ranges. But before we were able to get the inode lock, | |
1335 | * some other task may have punched a hole in the range, or we may have | |
1336 | * now an inline extent, in which case we should not defrag. So check | |
1337 | * for that here, where we have the inode and the range locked, and bail | |
1338 | * out if that happened. | |
1339 | */ | |
1340 | search_start = page_start; | |
1341 | while (search_start < page_end) { | |
1342 | struct extent_map *em; | |
1343 | ||
1344 | em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, search_start, | |
1345 | page_end - search_start); | |
1346 | if (IS_ERR(em)) { | |
1347 | ret = PTR_ERR(em); | |
1348 | goto out_unlock_range; | |
1349 | } | |
1350 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { | |
1351 | free_extent_map(em); | |
1352 | /* Ok, 0 means we did not defrag anything */ | |
1353 | ret = 0; | |
1354 | goto out_unlock_range; | |
1355 | } | |
1356 | search_start = extent_map_end(em); | |
1357 | free_extent_map(em); | |
1358 | } | |
1359 | ||
4cb5300b | 1360 | clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, |
e182163d OS |
1361 | page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | |
1362 | EXTENT_DEFRAG, 0, 0, &cached_state); | |
4cb5300b | 1363 | |
1f12bd06 | 1364 | if (i_done != page_cnt) { |
9e0baf60 | 1365 | spin_lock(&BTRFS_I(inode)->lock); |
28c4a3e2 | 1366 | btrfs_mod_outstanding_extents(BTRFS_I(inode), 1); |
9e0baf60 | 1367 | spin_unlock(&BTRFS_I(inode)->lock); |
86d52921 | 1368 | btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved, |
a1fbc675 | 1369 | start, (page_cnt - i_done) << PAGE_SHIFT, true); |
4cb5300b CM |
1370 | } |
1371 | ||
1372 | ||
9e8a4a8b | 1373 | set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, |
018ed4f7 | 1374 | &cached_state); |
4cb5300b CM |
1375 | |
1376 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, | |
e43bbe5e | 1377 | page_start, page_end - 1, &cached_state); |
4cb5300b CM |
1378 | |
1379 | for (i = 0; i < i_done; i++) { | |
1380 | clear_page_dirty_for_io(pages[i]); | |
1381 | ClearPageChecked(pages[i]); | |
4cb5300b CM |
1382 | set_page_dirty(pages[i]); |
1383 | unlock_page(pages[i]); | |
09cbfeaf | 1384 | put_page(pages[i]); |
4cb5300b | 1385 | } |
8702ba93 | 1386 | btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT); |
364ecf36 | 1387 | extent_changeset_free(data_reserved); |
4cb5300b | 1388 | return i_done; |
7f458a38 FM |
1389 | |
1390 | out_unlock_range: | |
1391 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, | |
1392 | page_start, page_end - 1, &cached_state); | |
4cb5300b CM |
1393 | out: |
1394 | for (i = 0; i < i_done; i++) { | |
1395 | unlock_page(pages[i]); | |
09cbfeaf | 1396 | put_page(pages[i]); |
4cb5300b | 1397 | } |
86d52921 | 1398 | btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved, |
a1fbc675 | 1399 | start, page_cnt << PAGE_SHIFT, true); |
8702ba93 | 1400 | btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT); |
364ecf36 | 1401 | extent_changeset_free(data_reserved); |
4cb5300b CM |
1402 | return ret; |
1403 | ||
1404 | } | |
1405 | ||
1406 | int btrfs_defrag_file(struct inode *inode, struct file *file, | |
1407 | struct btrfs_ioctl_defrag_range_args *range, | |
1408 | u64 newer_than, unsigned long max_to_defrag) | |
1409 | { | |
0b246afa | 1410 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
4cb5300b | 1411 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4cb5300b | 1412 | struct file_ra_state *ra = NULL; |
f46b5a66 | 1413 | unsigned long last_index; |
151a31b2 | 1414 | u64 isize = i_size_read(inode); |
940100a4 CM |
1415 | u64 last_len = 0; |
1416 | u64 skip = 0; | |
1417 | u64 defrag_end = 0; | |
4cb5300b | 1418 | u64 newer_off = range->start; |
f46b5a66 | 1419 | unsigned long i; |
008873ea | 1420 | unsigned long ra_index = 0; |
f46b5a66 | 1421 | int ret; |
4cb5300b | 1422 | int defrag_count = 0; |
1a419d85 | 1423 | int compress_type = BTRFS_COMPRESS_ZLIB; |
aab110ab | 1424 | u32 extent_thresh = range->extent_thresh; |
09cbfeaf | 1425 | unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; |
c41570c9 | 1426 | unsigned long cluster = max_cluster; |
ee22184b | 1427 | u64 new_align = ~((u64)SZ_128K - 1); |
4cb5300b | 1428 | struct page **pages = NULL; |
1e2ef46d | 1429 | bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; |
4cb5300b | 1430 | |
0abd5b17 LB |
1431 | if (isize == 0) |
1432 | return 0; | |
1433 | ||
1434 | if (range->start >= isize) | |
1435 | return -EINVAL; | |
1a419d85 | 1436 | |
1e2ef46d | 1437 | if (do_compress) { |
ce96b7ff | 1438 | if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES) |
1a419d85 LZ |
1439 | return -EINVAL; |
1440 | if (range->compress_type) | |
1441 | compress_type = range->compress_type; | |
1442 | } | |
f46b5a66 | 1443 | |
0abd5b17 | 1444 | if (extent_thresh == 0) |
ee22184b | 1445 | extent_thresh = SZ_256K; |
940100a4 | 1446 | |
4cb5300b | 1447 | /* |
0a52d108 DS |
1448 | * If we were not given a file, allocate a readahead context. As |
1449 | * readahead is just an optimization, defrag will work without it so | |
1450 | * we don't error out. | |
4cb5300b CM |
1451 | */ |
1452 | if (!file) { | |
63e727ec | 1453 | ra = kzalloc(sizeof(*ra), GFP_KERNEL); |
0a52d108 DS |
1454 | if (ra) |
1455 | file_ra_state_init(ra, inode->i_mapping); | |
4cb5300b CM |
1456 | } else { |
1457 | ra = &file->f_ra; | |
1458 | } | |
1459 | ||
63e727ec | 1460 | pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); |
4cb5300b CM |
1461 | if (!pages) { |
1462 | ret = -ENOMEM; | |
1463 | goto out_ra; | |
1464 | } | |
1465 | ||
1466 | /* find the last page to defrag */ | |
1e701a32 | 1467 | if (range->start + range->len > range->start) { |
151a31b2 | 1468 | last_index = min_t(u64, isize - 1, |
09cbfeaf | 1469 | range->start + range->len - 1) >> PAGE_SHIFT; |
1e701a32 | 1470 | } else { |
09cbfeaf | 1471 | last_index = (isize - 1) >> PAGE_SHIFT; |
1e701a32 CM |
1472 | } |
1473 | ||
4cb5300b CM |
1474 | if (newer_than) { |
1475 | ret = find_new_extents(root, inode, newer_than, | |
ee22184b | 1476 | &newer_off, SZ_64K); |
4cb5300b CM |
1477 | if (!ret) { |
1478 | range->start = newer_off; | |
1479 | /* | |
1480 | * we always align our defrag to help keep | |
1481 | * the extents in the file evenly spaced | |
1482 | */ | |
09cbfeaf | 1483 | i = (newer_off & new_align) >> PAGE_SHIFT; |
4cb5300b CM |
1484 | } else |
1485 | goto out_ra; | |
1486 | } else { | |
09cbfeaf | 1487 | i = range->start >> PAGE_SHIFT; |
4cb5300b CM |
1488 | } |
1489 | if (!max_to_defrag) | |
070034bd | 1490 | max_to_defrag = last_index - i + 1; |
4cb5300b | 1491 | |
2a0f7f57 LZ |
1492 | /* |
1493 | * make writeback starts from i, so the defrag range can be | |
1494 | * written sequentially. | |
1495 | */ | |
1496 | if (i < inode->i_mapping->writeback_index) | |
1497 | inode->i_mapping->writeback_index = i; | |
1498 | ||
f7f43cc8 | 1499 | while (i <= last_index && defrag_count < max_to_defrag && |
09cbfeaf | 1500 | (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { |
4cb5300b CM |
1501 | /* |
1502 | * make sure we stop running if someone unmounts | |
1503 | * the FS | |
1504 | */ | |
1751e8a6 | 1505 | if (!(inode->i_sb->s_flags & SB_ACTIVE)) |
4cb5300b CM |
1506 | break; |
1507 | ||
0b246afa JM |
1508 | if (btrfs_defrag_cancelled(fs_info)) { |
1509 | btrfs_debug(fs_info, "defrag_file cancelled"); | |
210549eb | 1510 | ret = -EAGAIN; |
50535db8 | 1511 | goto error; |
210549eb DS |
1512 | } |
1513 | ||
09cbfeaf | 1514 | if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, |
6c282eb4 | 1515 | extent_thresh, &last_len, &skip, |
1e2ef46d | 1516 | &defrag_end, do_compress)){ |
940100a4 CM |
1517 | unsigned long next; |
1518 | /* | |
1519 | * the should_defrag function tells us how much to skip | |
1520 | * bump our counter by the suggested amount | |
1521 | */ | |
09cbfeaf | 1522 | next = DIV_ROUND_UP(skip, PAGE_SIZE); |
940100a4 CM |
1523 | i = max(i + 1, next); |
1524 | continue; | |
1525 | } | |
008873ea LZ |
1526 | |
1527 | if (!newer_than) { | |
09cbfeaf KS |
1528 | cluster = (PAGE_ALIGN(defrag_end) >> |
1529 | PAGE_SHIFT) - i; | |
008873ea LZ |
1530 | cluster = min(cluster, max_cluster); |
1531 | } else { | |
1532 | cluster = max_cluster; | |
1533 | } | |
1534 | ||
008873ea LZ |
1535 | if (i + cluster > ra_index) { |
1536 | ra_index = max(i, ra_index); | |
0a52d108 | 1537 | if (ra) |
d3c0bab5 DS |
1538 | page_cache_sync_readahead(inode->i_mapping, ra, |
1539 | file, ra_index, cluster); | |
e4826a5b | 1540 | ra_index += cluster; |
008873ea | 1541 | } |
940100a4 | 1542 | |
64708539 | 1543 | btrfs_inode_lock(inode, 0); |
eede2bf3 OS |
1544 | if (IS_SWAPFILE(inode)) { |
1545 | ret = -ETXTBSY; | |
1546 | } else { | |
1547 | if (do_compress) | |
1548 | BTRFS_I(inode)->defrag_compress = compress_type; | |
1549 | ret = cluster_pages_for_defrag(inode, pages, i, cluster); | |
1550 | } | |
ecb8bea8 | 1551 | if (ret < 0) { |
64708539 | 1552 | btrfs_inode_unlock(inode, 0); |
4cb5300b | 1553 | goto out_ra; |
ecb8bea8 | 1554 | } |
4cb5300b CM |
1555 | |
1556 | defrag_count += ret; | |
d0e1d66b | 1557 | balance_dirty_pages_ratelimited(inode->i_mapping); |
64708539 | 1558 | btrfs_inode_unlock(inode, 0); |
4cb5300b CM |
1559 | |
1560 | if (newer_than) { | |
1561 | if (newer_off == (u64)-1) | |
1562 | break; | |
1563 | ||
e1f041e1 LB |
1564 | if (ret > 0) |
1565 | i += ret; | |
1566 | ||
4cb5300b | 1567 | newer_off = max(newer_off + 1, |
09cbfeaf | 1568 | (u64)i << PAGE_SHIFT); |
4cb5300b | 1569 | |
ee22184b BL |
1570 | ret = find_new_extents(root, inode, newer_than, |
1571 | &newer_off, SZ_64K); | |
4cb5300b CM |
1572 | if (!ret) { |
1573 | range->start = newer_off; | |
09cbfeaf | 1574 | i = (newer_off & new_align) >> PAGE_SHIFT; |
4cb5300b CM |
1575 | } else { |
1576 | break; | |
f46b5a66 | 1577 | } |
4cb5300b | 1578 | } else { |
008873ea | 1579 | if (ret > 0) { |
cbcc8326 | 1580 | i += ret; |
09cbfeaf | 1581 | last_len += ret << PAGE_SHIFT; |
008873ea | 1582 | } else { |
cbcc8326 | 1583 | i++; |
008873ea LZ |
1584 | last_len = 0; |
1585 | } | |
f46b5a66 | 1586 | } |
f46b5a66 CH |
1587 | } |
1588 | ||
50535db8 TT |
1589 | ret = defrag_count; |
1590 | error: | |
dec8ef90 | 1591 | if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { |
1e701a32 | 1592 | filemap_flush(inode->i_mapping); |
dec8ef90 FM |
1593 | if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, |
1594 | &BTRFS_I(inode)->runtime_flags)) | |
1595 | filemap_flush(inode->i_mapping); | |
1596 | } | |
1e701a32 | 1597 | |
1a419d85 | 1598 | if (range->compress_type == BTRFS_COMPRESS_LZO) { |
0b246afa | 1599 | btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); |
5c1aab1d NT |
1600 | } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { |
1601 | btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); | |
1a419d85 LZ |
1602 | } |
1603 | ||
4cb5300b | 1604 | out_ra: |
1e2ef46d | 1605 | if (do_compress) { |
64708539 | 1606 | btrfs_inode_lock(inode, 0); |
eec63c65 | 1607 | BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; |
64708539 | 1608 | btrfs_inode_unlock(inode, 0); |
633085c7 | 1609 | } |
4cb5300b CM |
1610 | if (!file) |
1611 | kfree(ra); | |
1612 | kfree(pages); | |
940100a4 | 1613 | return ret; |
f46b5a66 CH |
1614 | } |
1615 | ||
17aaa434 DS |
1616 | /* |
1617 | * Try to start exclusive operation @type or cancel it if it's running. | |
1618 | * | |
1619 | * Return: | |
1620 | * 0 - normal mode, newly claimed op started | |
1621 | * >0 - normal mode, something else is running, | |
1622 | * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space | |
1623 | * ECANCELED - cancel mode, successful cancel | |
1624 | * ENOTCONN - cancel mode, operation not running anymore | |
1625 | */ | |
1626 | static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info, | |
1627 | enum btrfs_exclusive_operation type, bool cancel) | |
1628 | { | |
1629 | if (!cancel) { | |
1630 | /* Start normal op */ | |
1631 | if (!btrfs_exclop_start(fs_info, type)) | |
1632 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; | |
1633 | /* Exclusive operation is now claimed */ | |
1634 | return 0; | |
1635 | } | |
1636 | ||
1637 | /* Cancel running op */ | |
1638 | if (btrfs_exclop_start_try_lock(fs_info, type)) { | |
1639 | /* | |
1640 | * This blocks any exclop finish from setting it to NONE, so we | |
1641 | * request cancellation. Either it runs and we will wait for it, | |
1642 | * or it has finished and no waiting will happen. | |
1643 | */ | |
1644 | atomic_inc(&fs_info->reloc_cancel_req); | |
1645 | btrfs_exclop_start_unlock(fs_info); | |
1646 | ||
1647 | if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) | |
1648 | wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING, | |
1649 | TASK_INTERRUPTIBLE); | |
1650 | ||
1651 | return -ECANCELED; | |
1652 | } | |
1653 | ||
1654 | /* Something else is running or none */ | |
1655 | return -ENOTCONN; | |
1656 | } | |
1657 | ||
198605a8 | 1658 | static noinline int btrfs_ioctl_resize(struct file *file, |
76dda93c | 1659 | void __user *arg) |
f46b5a66 | 1660 | { |
0b246afa JM |
1661 | struct inode *inode = file_inode(file); |
1662 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f46b5a66 CH |
1663 | u64 new_size; |
1664 | u64 old_size; | |
1665 | u64 devid = 1; | |
0b246afa | 1666 | struct btrfs_root *root = BTRFS_I(inode)->root; |
f46b5a66 CH |
1667 | struct btrfs_ioctl_vol_args *vol_args; |
1668 | struct btrfs_trans_handle *trans; | |
1669 | struct btrfs_device *device = NULL; | |
1670 | char *sizestr; | |
9a40f122 | 1671 | char *retptr; |
f46b5a66 CH |
1672 | char *devstr = NULL; |
1673 | int ret = 0; | |
f46b5a66 | 1674 | int mod = 0; |
bb059a37 | 1675 | bool cancel; |
f46b5a66 | 1676 | |
e441d54d CM |
1677 | if (!capable(CAP_SYS_ADMIN)) |
1678 | return -EPERM; | |
1679 | ||
198605a8 MX |
1680 | ret = mnt_want_write_file(file); |
1681 | if (ret) | |
1682 | return ret; | |
1683 | ||
bb059a37 DS |
1684 | /* |
1685 | * Read the arguments before checking exclusivity to be able to | |
1686 | * distinguish regular resize and cancel | |
1687 | */ | |
dae7b665 | 1688 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
1689 | if (IS_ERR(vol_args)) { |
1690 | ret = PTR_ERR(vol_args); | |
bb059a37 | 1691 | goto out_drop; |
c9e9f97b | 1692 | } |
5516e595 | 1693 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
f46b5a66 | 1694 | sizestr = vol_args->name; |
bb059a37 DS |
1695 | cancel = (strcmp("cancel", sizestr) == 0); |
1696 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel); | |
1697 | if (ret) | |
1698 | goto out_free; | |
1699 | /* Exclusive operation is now claimed */ | |
1700 | ||
f46b5a66 CH |
1701 | devstr = strchr(sizestr, ':'); |
1702 | if (devstr) { | |
f46b5a66 CH |
1703 | sizestr = devstr + 1; |
1704 | *devstr = '\0'; | |
1705 | devstr = vol_args->name; | |
58dfae63 Z |
1706 | ret = kstrtoull(devstr, 10, &devid); |
1707 | if (ret) | |
bb059a37 | 1708 | goto out_finish; |
dfd79829 MX |
1709 | if (!devid) { |
1710 | ret = -EINVAL; | |
bb059a37 | 1711 | goto out_finish; |
dfd79829 | 1712 | } |
0b246afa | 1713 | btrfs_info(fs_info, "resizing devid %llu", devid); |
f46b5a66 | 1714 | } |
dba60f3f | 1715 | |
b2598edf | 1716 | device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL); |
f46b5a66 | 1717 | if (!device) { |
0b246afa JM |
1718 | btrfs_info(fs_info, "resizer unable to find device %llu", |
1719 | devid); | |
dfd79829 | 1720 | ret = -ENODEV; |
bb059a37 | 1721 | goto out_finish; |
f46b5a66 | 1722 | } |
dba60f3f | 1723 | |
ebbede42 | 1724 | if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) { |
0b246afa | 1725 | btrfs_info(fs_info, |
efe120a0 | 1726 | "resizer unable to apply on readonly device %llu", |
c1c9ff7c | 1727 | devid); |
dfd79829 | 1728 | ret = -EPERM; |
bb059a37 | 1729 | goto out_finish; |
4e42ae1b LB |
1730 | } |
1731 | ||
f46b5a66 CH |
1732 | if (!strcmp(sizestr, "max")) |
1733 | new_size = device->bdev->bd_inode->i_size; | |
1734 | else { | |
1735 | if (sizestr[0] == '-') { | |
1736 | mod = -1; | |
1737 | sizestr++; | |
1738 | } else if (sizestr[0] == '+') { | |
1739 | mod = 1; | |
1740 | sizestr++; | |
1741 | } | |
9a40f122 GH |
1742 | new_size = memparse(sizestr, &retptr); |
1743 | if (*retptr != '\0' || new_size == 0) { | |
f46b5a66 | 1744 | ret = -EINVAL; |
bb059a37 | 1745 | goto out_finish; |
f46b5a66 CH |
1746 | } |
1747 | } | |
1748 | ||
401e29c1 | 1749 | if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) { |
dfd79829 | 1750 | ret = -EPERM; |
bb059a37 | 1751 | goto out_finish; |
63a212ab SB |
1752 | } |
1753 | ||
7cc8e58d | 1754 | old_size = btrfs_device_get_total_bytes(device); |
f46b5a66 CH |
1755 | |
1756 | if (mod < 0) { | |
1757 | if (new_size > old_size) { | |
1758 | ret = -EINVAL; | |
bb059a37 | 1759 | goto out_finish; |
f46b5a66 CH |
1760 | } |
1761 | new_size = old_size - new_size; | |
1762 | } else if (mod > 0) { | |
eb8052e0 | 1763 | if (new_size > ULLONG_MAX - old_size) { |
902c68a4 | 1764 | ret = -ERANGE; |
bb059a37 | 1765 | goto out_finish; |
eb8052e0 | 1766 | } |
f46b5a66 CH |
1767 | new_size = old_size + new_size; |
1768 | } | |
1769 | ||
ee22184b | 1770 | if (new_size < SZ_256M) { |
f46b5a66 | 1771 | ret = -EINVAL; |
bb059a37 | 1772 | goto out_finish; |
f46b5a66 CH |
1773 | } |
1774 | if (new_size > device->bdev->bd_inode->i_size) { | |
1775 | ret = -EFBIG; | |
bb059a37 | 1776 | goto out_finish; |
f46b5a66 CH |
1777 | } |
1778 | ||
47f08b96 | 1779 | new_size = round_down(new_size, fs_info->sectorsize); |
f46b5a66 | 1780 | |
f46b5a66 | 1781 | if (new_size > old_size) { |
a22285a6 | 1782 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
1783 | if (IS_ERR(trans)) { |
1784 | ret = PTR_ERR(trans); | |
bb059a37 | 1785 | goto out_finish; |
98d5dc13 | 1786 | } |
f46b5a66 | 1787 | ret = btrfs_grow_device(trans, device, new_size); |
3a45bb20 | 1788 | btrfs_commit_transaction(trans); |
ece7d20e | 1789 | } else if (new_size < old_size) { |
f46b5a66 | 1790 | ret = btrfs_shrink_device(device, new_size); |
0253f40e | 1791 | } /* equal, nothing need to do */ |
f46b5a66 | 1792 | |
faf8f7b9 MPS |
1793 | if (ret == 0 && new_size != old_size) |
1794 | btrfs_info_in_rcu(fs_info, | |
1795 | "resize device %s (devid %llu) from %llu to %llu", | |
1796 | rcu_str_deref(device->name), device->devid, | |
1797 | old_size, new_size); | |
bb059a37 DS |
1798 | out_finish: |
1799 | btrfs_exclop_finish(fs_info); | |
c9e9f97b | 1800 | out_free: |
f46b5a66 | 1801 | kfree(vol_args); |
bb059a37 | 1802 | out_drop: |
18f39c41 | 1803 | mnt_drop_write_file(file); |
f46b5a66 CH |
1804 | return ret; |
1805 | } | |
1806 | ||
5d54c67e | 1807 | static noinline int __btrfs_ioctl_snap_create(struct file *file, |
4d4340c9 | 1808 | struct user_namespace *mnt_userns, |
52f75f4f | 1809 | const char *name, unsigned long fd, int subvol, |
5d54c67e | 1810 | bool readonly, |
8696c533 | 1811 | struct btrfs_qgroup_inherit *inherit) |
f46b5a66 | 1812 | { |
f46b5a66 | 1813 | int namelen; |
3de4586c | 1814 | int ret = 0; |
f46b5a66 | 1815 | |
325c50e3 JM |
1816 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1817 | return -ENOTDIR; | |
1818 | ||
a874a63e LB |
1819 | ret = mnt_want_write_file(file); |
1820 | if (ret) | |
1821 | goto out; | |
1822 | ||
72fd032e SW |
1823 | namelen = strlen(name); |
1824 | if (strchr(name, '/')) { | |
f46b5a66 | 1825 | ret = -EINVAL; |
a874a63e | 1826 | goto out_drop_write; |
f46b5a66 CH |
1827 | } |
1828 | ||
16780cab CM |
1829 | if (name[0] == '.' && |
1830 | (namelen == 1 || (name[1] == '.' && namelen == 2))) { | |
1831 | ret = -EEXIST; | |
a874a63e | 1832 | goto out_drop_write; |
16780cab CM |
1833 | } |
1834 | ||
3de4586c | 1835 | if (subvol) { |
4d4340c9 CB |
1836 | ret = btrfs_mksubvol(&file->f_path, mnt_userns, name, |
1837 | namelen, NULL, readonly, inherit); | |
cb8e7090 | 1838 | } else { |
2903ff01 | 1839 | struct fd src = fdget(fd); |
3de4586c | 1840 | struct inode *src_inode; |
2903ff01 | 1841 | if (!src.file) { |
3de4586c | 1842 | ret = -EINVAL; |
a874a63e | 1843 | goto out_drop_write; |
3de4586c CM |
1844 | } |
1845 | ||
496ad9aa AV |
1846 | src_inode = file_inode(src.file); |
1847 | if (src_inode->i_sb != file_inode(file)->i_sb) { | |
c79b4713 | 1848 | btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, |
efe120a0 | 1849 | "Snapshot src from another FS"); |
23ad5b17 | 1850 | ret = -EXDEV; |
4d4340c9 | 1851 | } else if (!inode_owner_or_capable(mnt_userns, src_inode)) { |
d0242061 DS |
1852 | /* |
1853 | * Subvolume creation is not restricted, but snapshots | |
1854 | * are limited to own subvolumes only | |
1855 | */ | |
1856 | ret = -EPERM; | |
ecd18815 | 1857 | } else { |
4d4340c9 CB |
1858 | ret = btrfs_mksnapshot(&file->f_path, mnt_userns, |
1859 | name, namelen, | |
1860 | BTRFS_I(src_inode)->root, | |
1861 | readonly, inherit); | |
3de4586c | 1862 | } |
2903ff01 | 1863 | fdput(src); |
cb8e7090 | 1864 | } |
a874a63e LB |
1865 | out_drop_write: |
1866 | mnt_drop_write_file(file); | |
f46b5a66 | 1867 | out: |
72fd032e SW |
1868 | return ret; |
1869 | } | |
1870 | ||
1871 | static noinline int btrfs_ioctl_snap_create(struct file *file, | |
fa0d2b9b | 1872 | void __user *arg, int subvol) |
72fd032e | 1873 | { |
fa0d2b9b | 1874 | struct btrfs_ioctl_vol_args *vol_args; |
72fd032e SW |
1875 | int ret; |
1876 | ||
325c50e3 JM |
1877 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1878 | return -ENOTDIR; | |
1879 | ||
fa0d2b9b LZ |
1880 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1881 | if (IS_ERR(vol_args)) | |
1882 | return PTR_ERR(vol_args); | |
1883 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; | |
72fd032e | 1884 | |
4d4340c9 CB |
1885 | ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), |
1886 | vol_args->name, vol_args->fd, subvol, | |
1887 | false, NULL); | |
fdfb1e4f | 1888 | |
fa0d2b9b LZ |
1889 | kfree(vol_args); |
1890 | return ret; | |
1891 | } | |
fdfb1e4f | 1892 | |
fa0d2b9b LZ |
1893 | static noinline int btrfs_ioctl_snap_create_v2(struct file *file, |
1894 | void __user *arg, int subvol) | |
1895 | { | |
1896 | struct btrfs_ioctl_vol_args_v2 *vol_args; | |
1897 | int ret; | |
b83cc969 | 1898 | bool readonly = false; |
6f72c7e2 | 1899 | struct btrfs_qgroup_inherit *inherit = NULL; |
75eaa0e2 | 1900 | |
325c50e3 JM |
1901 | if (!S_ISDIR(file_inode(file)->i_mode)) |
1902 | return -ENOTDIR; | |
1903 | ||
fa0d2b9b LZ |
1904 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
1905 | if (IS_ERR(vol_args)) | |
1906 | return PTR_ERR(vol_args); | |
1907 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; | |
75eaa0e2 | 1908 | |
673990db | 1909 | if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) { |
b83cc969 | 1910 | ret = -EOPNOTSUPP; |
c47ca32d | 1911 | goto free_args; |
72fd032e | 1912 | } |
fa0d2b9b | 1913 | |
b83cc969 LZ |
1914 | if (vol_args->flags & BTRFS_SUBVOL_RDONLY) |
1915 | readonly = true; | |
6f72c7e2 | 1916 | if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { |
5011c5a6 DC |
1917 | u64 nums; |
1918 | ||
1919 | if (vol_args->size < sizeof(*inherit) || | |
1920 | vol_args->size > PAGE_SIZE) { | |
6f72c7e2 | 1921 | ret = -EINVAL; |
c47ca32d | 1922 | goto free_args; |
6f72c7e2 AJ |
1923 | } |
1924 | inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); | |
1925 | if (IS_ERR(inherit)) { | |
1926 | ret = PTR_ERR(inherit); | |
c47ca32d | 1927 | goto free_args; |
6f72c7e2 | 1928 | } |
5011c5a6 DC |
1929 | |
1930 | if (inherit->num_qgroups > PAGE_SIZE || | |
1931 | inherit->num_ref_copies > PAGE_SIZE || | |
1932 | inherit->num_excl_copies > PAGE_SIZE) { | |
1933 | ret = -EINVAL; | |
1934 | goto free_inherit; | |
1935 | } | |
1936 | ||
1937 | nums = inherit->num_qgroups + 2 * inherit->num_ref_copies + | |
1938 | 2 * inherit->num_excl_copies; | |
1939 | if (vol_args->size != struct_size(inherit, qgroups, nums)) { | |
1940 | ret = -EINVAL; | |
1941 | goto free_inherit; | |
1942 | } | |
6f72c7e2 | 1943 | } |
fa0d2b9b | 1944 | |
4d4340c9 CB |
1945 | ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file), |
1946 | vol_args->name, vol_args->fd, subvol, | |
1947 | readonly, inherit); | |
c47ca32d DC |
1948 | if (ret) |
1949 | goto free_inherit; | |
c47ca32d | 1950 | free_inherit: |
6f72c7e2 | 1951 | kfree(inherit); |
c47ca32d DC |
1952 | free_args: |
1953 | kfree(vol_args); | |
f46b5a66 CH |
1954 | return ret; |
1955 | } | |
1956 | ||
0caa102d LZ |
1957 | static noinline int btrfs_ioctl_subvol_getflags(struct file *file, |
1958 | void __user *arg) | |
1959 | { | |
496ad9aa | 1960 | struct inode *inode = file_inode(file); |
0b246afa | 1961 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
0caa102d LZ |
1962 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1963 | int ret = 0; | |
1964 | u64 flags = 0; | |
1965 | ||
4a0cc7ca | 1966 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) |
0caa102d LZ |
1967 | return -EINVAL; |
1968 | ||
0b246afa | 1969 | down_read(&fs_info->subvol_sem); |
0caa102d LZ |
1970 | if (btrfs_root_readonly(root)) |
1971 | flags |= BTRFS_SUBVOL_RDONLY; | |
0b246afa | 1972 | up_read(&fs_info->subvol_sem); |
0caa102d LZ |
1973 | |
1974 | if (copy_to_user(arg, &flags, sizeof(flags))) | |
1975 | ret = -EFAULT; | |
1976 | ||
1977 | return ret; | |
1978 | } | |
1979 | ||
1980 | static noinline int btrfs_ioctl_subvol_setflags(struct file *file, | |
1981 | void __user *arg) | |
1982 | { | |
496ad9aa | 1983 | struct inode *inode = file_inode(file); |
0b246afa | 1984 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
0caa102d LZ |
1985 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1986 | struct btrfs_trans_handle *trans; | |
1987 | u64 root_flags; | |
1988 | u64 flags; | |
1989 | int ret = 0; | |
1990 | ||
39e1674f | 1991 | if (!inode_owner_or_capable(file_mnt_user_ns(file), inode)) |
bd60ea0f DS |
1992 | return -EPERM; |
1993 | ||
b9ca0664 LB |
1994 | ret = mnt_want_write_file(file); |
1995 | if (ret) | |
1996 | goto out; | |
0caa102d | 1997 | |
4a0cc7ca | 1998 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
b9ca0664 LB |
1999 | ret = -EINVAL; |
2000 | goto out_drop_write; | |
2001 | } | |
0caa102d | 2002 | |
b9ca0664 LB |
2003 | if (copy_from_user(&flags, arg, sizeof(flags))) { |
2004 | ret = -EFAULT; | |
2005 | goto out_drop_write; | |
2006 | } | |
0caa102d | 2007 | |
b9ca0664 LB |
2008 | if (flags & ~BTRFS_SUBVOL_RDONLY) { |
2009 | ret = -EOPNOTSUPP; | |
2010 | goto out_drop_write; | |
2011 | } | |
0caa102d | 2012 | |
0b246afa | 2013 | down_write(&fs_info->subvol_sem); |
0caa102d LZ |
2014 | |
2015 | /* nothing to do */ | |
2016 | if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) | |
b9ca0664 | 2017 | goto out_drop_sem; |
0caa102d LZ |
2018 | |
2019 | root_flags = btrfs_root_flags(&root->root_item); | |
2c686537 | 2020 | if (flags & BTRFS_SUBVOL_RDONLY) { |
0caa102d LZ |
2021 | btrfs_set_root_flags(&root->root_item, |
2022 | root_flags | BTRFS_ROOT_SUBVOL_RDONLY); | |
2c686537 DS |
2023 | } else { |
2024 | /* | |
2025 | * Block RO -> RW transition if this subvolume is involved in | |
2026 | * send | |
2027 | */ | |
2028 | spin_lock(&root->root_item_lock); | |
2029 | if (root->send_in_progress == 0) { | |
2030 | btrfs_set_root_flags(&root->root_item, | |
0caa102d | 2031 | root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); |
2c686537 DS |
2032 | spin_unlock(&root->root_item_lock); |
2033 | } else { | |
2034 | spin_unlock(&root->root_item_lock); | |
0b246afa JM |
2035 | btrfs_warn(fs_info, |
2036 | "Attempt to set subvolume %llu read-write during send", | |
2037 | root->root_key.objectid); | |
2c686537 DS |
2038 | ret = -EPERM; |
2039 | goto out_drop_sem; | |
2040 | } | |
2041 | } | |
0caa102d LZ |
2042 | |
2043 | trans = btrfs_start_transaction(root, 1); | |
2044 | if (IS_ERR(trans)) { | |
2045 | ret = PTR_ERR(trans); | |
2046 | goto out_reset; | |
2047 | } | |
2048 | ||
0b246afa | 2049 | ret = btrfs_update_root(trans, fs_info->tree_root, |
0caa102d | 2050 | &root->root_key, &root->root_item); |
9417ebc8 NB |
2051 | if (ret < 0) { |
2052 | btrfs_end_transaction(trans); | |
2053 | goto out_reset; | |
2054 | } | |
2055 | ||
2056 | ret = btrfs_commit_transaction(trans); | |
0caa102d | 2057 | |
0caa102d LZ |
2058 | out_reset: |
2059 | if (ret) | |
2060 | btrfs_set_root_flags(&root->root_item, root_flags); | |
b9ca0664 | 2061 | out_drop_sem: |
0b246afa | 2062 | up_write(&fs_info->subvol_sem); |
b9ca0664 LB |
2063 | out_drop_write: |
2064 | mnt_drop_write_file(file); | |
2065 | out: | |
0caa102d LZ |
2066 | return ret; |
2067 | } | |
2068 | ||
ac8e9819 CM |
2069 | static noinline int key_in_sk(struct btrfs_key *key, |
2070 | struct btrfs_ioctl_search_key *sk) | |
2071 | { | |
abc6e134 CM |
2072 | struct btrfs_key test; |
2073 | int ret; | |
2074 | ||
2075 | test.objectid = sk->min_objectid; | |
2076 | test.type = sk->min_type; | |
2077 | test.offset = sk->min_offset; | |
2078 | ||
2079 | ret = btrfs_comp_cpu_keys(key, &test); | |
2080 | if (ret < 0) | |
ac8e9819 | 2081 | return 0; |
abc6e134 CM |
2082 | |
2083 | test.objectid = sk->max_objectid; | |
2084 | test.type = sk->max_type; | |
2085 | test.offset = sk->max_offset; | |
2086 | ||
2087 | ret = btrfs_comp_cpu_keys(key, &test); | |
2088 | if (ret > 0) | |
ac8e9819 CM |
2089 | return 0; |
2090 | return 1; | |
2091 | } | |
2092 | ||
df397565 | 2093 | static noinline int copy_to_sk(struct btrfs_path *path, |
ac8e9819 CM |
2094 | struct btrfs_key *key, |
2095 | struct btrfs_ioctl_search_key *sk, | |
9b6e817d | 2096 | size_t *buf_size, |
ba346b35 | 2097 | char __user *ubuf, |
ac8e9819 CM |
2098 | unsigned long *sk_offset, |
2099 | int *num_found) | |
2100 | { | |
2101 | u64 found_transid; | |
2102 | struct extent_buffer *leaf; | |
2103 | struct btrfs_ioctl_search_header sh; | |
dd81d459 | 2104 | struct btrfs_key test; |
ac8e9819 CM |
2105 | unsigned long item_off; |
2106 | unsigned long item_len; | |
2107 | int nritems; | |
2108 | int i; | |
2109 | int slot; | |
ac8e9819 CM |
2110 | int ret = 0; |
2111 | ||
2112 | leaf = path->nodes[0]; | |
2113 | slot = path->slots[0]; | |
2114 | nritems = btrfs_header_nritems(leaf); | |
2115 | ||
2116 | if (btrfs_header_generation(leaf) > sk->max_transid) { | |
2117 | i = nritems; | |
2118 | goto advance_key; | |
2119 | } | |
2120 | found_transid = btrfs_header_generation(leaf); | |
2121 | ||
2122 | for (i = slot; i < nritems; i++) { | |
2123 | item_off = btrfs_item_ptr_offset(leaf, i); | |
2124 | item_len = btrfs_item_size_nr(leaf, i); | |
2125 | ||
03b71c6c GP |
2126 | btrfs_item_key_to_cpu(leaf, key, i); |
2127 | if (!key_in_sk(key, sk)) | |
2128 | continue; | |
2129 | ||
9b6e817d | 2130 | if (sizeof(sh) + item_len > *buf_size) { |
8f5f6178 GH |
2131 | if (*num_found) { |
2132 | ret = 1; | |
2133 | goto out; | |
2134 | } | |
2135 | ||
2136 | /* | |
2137 | * return one empty item back for v1, which does not | |
2138 | * handle -EOVERFLOW | |
2139 | */ | |
2140 | ||
9b6e817d | 2141 | *buf_size = sizeof(sh) + item_len; |
ac8e9819 | 2142 | item_len = 0; |
8f5f6178 GH |
2143 | ret = -EOVERFLOW; |
2144 | } | |
ac8e9819 | 2145 | |
9b6e817d | 2146 | if (sizeof(sh) + item_len + *sk_offset > *buf_size) { |
ac8e9819 | 2147 | ret = 1; |
25c9bc2e | 2148 | goto out; |
ac8e9819 CM |
2149 | } |
2150 | ||
ac8e9819 CM |
2151 | sh.objectid = key->objectid; |
2152 | sh.offset = key->offset; | |
2153 | sh.type = key->type; | |
2154 | sh.len = item_len; | |
2155 | sh.transid = found_transid; | |
2156 | ||
a48b73ec JB |
2157 | /* |
2158 | * Copy search result header. If we fault then loop again so we | |
2159 | * can fault in the pages and -EFAULT there if there's a | |
2160 | * problem. Otherwise we'll fault and then copy the buffer in | |
2161 | * properly this next time through | |
2162 | */ | |
2163 | if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) { | |
2164 | ret = 0; | |
ba346b35 GH |
2165 | goto out; |
2166 | } | |
2167 | ||
ac8e9819 CM |
2168 | *sk_offset += sizeof(sh); |
2169 | ||
2170 | if (item_len) { | |
ba346b35 | 2171 | char __user *up = ubuf + *sk_offset; |
a48b73ec JB |
2172 | /* |
2173 | * Copy the item, same behavior as above, but reset the | |
2174 | * * sk_offset so we copy the full thing again. | |
2175 | */ | |
2176 | if (read_extent_buffer_to_user_nofault(leaf, up, | |
2177 | item_off, item_len)) { | |
2178 | ret = 0; | |
2179 | *sk_offset -= sizeof(sh); | |
ba346b35 GH |
2180 | goto out; |
2181 | } | |
2182 | ||
ac8e9819 | 2183 | *sk_offset += item_len; |
ac8e9819 | 2184 | } |
e2156867 | 2185 | (*num_found)++; |
ac8e9819 | 2186 | |
8f5f6178 GH |
2187 | if (ret) /* -EOVERFLOW from above */ |
2188 | goto out; | |
2189 | ||
25c9bc2e GH |
2190 | if (*num_found >= sk->nr_items) { |
2191 | ret = 1; | |
2192 | goto out; | |
2193 | } | |
ac8e9819 CM |
2194 | } |
2195 | advance_key: | |
abc6e134 | 2196 | ret = 0; |
dd81d459 NA |
2197 | test.objectid = sk->max_objectid; |
2198 | test.type = sk->max_type; | |
2199 | test.offset = sk->max_offset; | |
2200 | if (btrfs_comp_cpu_keys(key, &test) >= 0) | |
2201 | ret = 1; | |
2202 | else if (key->offset < (u64)-1) | |
ac8e9819 | 2203 | key->offset++; |
dd81d459 | 2204 | else if (key->type < (u8)-1) { |
abc6e134 | 2205 | key->offset = 0; |
ac8e9819 | 2206 | key->type++; |
dd81d459 | 2207 | } else if (key->objectid < (u64)-1) { |
abc6e134 CM |
2208 | key->offset = 0; |
2209 | key->type = 0; | |
ac8e9819 | 2210 | key->objectid++; |
abc6e134 CM |
2211 | } else |
2212 | ret = 1; | |
25c9bc2e | 2213 | out: |
ba346b35 GH |
2214 | /* |
2215 | * 0: all items from this leaf copied, continue with next | |
2216 | * 1: * more items can be copied, but unused buffer is too small | |
2217 | * * all items were found | |
2218 | * Either way, it will stops the loop which iterates to the next | |
2219 | * leaf | |
2220 | * -EOVERFLOW: item was to large for buffer | |
2221 | * -EFAULT: could not copy extent buffer back to userspace | |
2222 | */ | |
ac8e9819 CM |
2223 | return ret; |
2224 | } | |
2225 | ||
2226 | static noinline int search_ioctl(struct inode *inode, | |
12544442 | 2227 | struct btrfs_ioctl_search_key *sk, |
9b6e817d | 2228 | size_t *buf_size, |
ba346b35 | 2229 | char __user *ubuf) |
ac8e9819 | 2230 | { |
0b246afa | 2231 | struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); |
ac8e9819 CM |
2232 | struct btrfs_root *root; |
2233 | struct btrfs_key key; | |
ac8e9819 | 2234 | struct btrfs_path *path; |
ac8e9819 CM |
2235 | int ret; |
2236 | int num_found = 0; | |
2237 | unsigned long sk_offset = 0; | |
2238 | ||
9b6e817d GH |
2239 | if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { |
2240 | *buf_size = sizeof(struct btrfs_ioctl_search_header); | |
12544442 | 2241 | return -EOVERFLOW; |
9b6e817d | 2242 | } |
12544442 | 2243 | |
ac8e9819 CM |
2244 | path = btrfs_alloc_path(); |
2245 | if (!path) | |
2246 | return -ENOMEM; | |
2247 | ||
2248 | if (sk->tree_id == 0) { | |
2249 | /* search the root of the inode that was passed */ | |
00246528 | 2250 | root = btrfs_grab_root(BTRFS_I(inode)->root); |
ac8e9819 | 2251 | } else { |
56e9357a | 2252 | root = btrfs_get_fs_root(info, sk->tree_id, true); |
ac8e9819 | 2253 | if (IS_ERR(root)) { |
ac8e9819 | 2254 | btrfs_free_path(path); |
ad1e3d56 | 2255 | return PTR_ERR(root); |
ac8e9819 CM |
2256 | } |
2257 | } | |
2258 | ||
2259 | key.objectid = sk->min_objectid; | |
2260 | key.type = sk->min_type; | |
2261 | key.offset = sk->min_offset; | |
2262 | ||
67871254 | 2263 | while (1) { |
1c78544e FM |
2264 | ret = fault_in_pages_writeable(ubuf + sk_offset, |
2265 | *buf_size - sk_offset); | |
a48b73ec JB |
2266 | if (ret) |
2267 | break; | |
2268 | ||
6174d3cb | 2269 | ret = btrfs_search_forward(root, &key, path, sk->min_transid); |
ac8e9819 CM |
2270 | if (ret != 0) { |
2271 | if (ret > 0) | |
2272 | ret = 0; | |
2273 | goto err; | |
2274 | } | |
df397565 | 2275 | ret = copy_to_sk(path, &key, sk, buf_size, ubuf, |
ac8e9819 | 2276 | &sk_offset, &num_found); |
b3b4aa74 | 2277 | btrfs_release_path(path); |
25c9bc2e | 2278 | if (ret) |
ac8e9819 CM |
2279 | break; |
2280 | ||
2281 | } | |
8f5f6178 GH |
2282 | if (ret > 0) |
2283 | ret = 0; | |
ac8e9819 CM |
2284 | err: |
2285 | sk->nr_items = num_found; | |
00246528 | 2286 | btrfs_put_root(root); |
ac8e9819 CM |
2287 | btrfs_free_path(path); |
2288 | return ret; | |
2289 | } | |
2290 | ||
2291 | static noinline int btrfs_ioctl_tree_search(struct file *file, | |
2292 | void __user *argp) | |
2293 | { | |
ba346b35 GH |
2294 | struct btrfs_ioctl_search_args __user *uargs; |
2295 | struct btrfs_ioctl_search_key sk; | |
9b6e817d GH |
2296 | struct inode *inode; |
2297 | int ret; | |
2298 | size_t buf_size; | |
ac8e9819 CM |
2299 | |
2300 | if (!capable(CAP_SYS_ADMIN)) | |
2301 | return -EPERM; | |
2302 | ||
ba346b35 GH |
2303 | uargs = (struct btrfs_ioctl_search_args __user *)argp; |
2304 | ||
2305 | if (copy_from_user(&sk, &uargs->key, sizeof(sk))) | |
2306 | return -EFAULT; | |
ac8e9819 | 2307 | |
ba346b35 | 2308 | buf_size = sizeof(uargs->buf); |
ac8e9819 | 2309 | |
496ad9aa | 2310 | inode = file_inode(file); |
ba346b35 | 2311 | ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); |
8f5f6178 GH |
2312 | |
2313 | /* | |
2314 | * In the origin implementation an overflow is handled by returning a | |
2315 | * search header with a len of zero, so reset ret. | |
2316 | */ | |
2317 | if (ret == -EOVERFLOW) | |
2318 | ret = 0; | |
2319 | ||
ba346b35 | 2320 | if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) |
ac8e9819 | 2321 | ret = -EFAULT; |
ac8e9819 CM |
2322 | return ret; |
2323 | } | |
2324 | ||
cc68a8a5 GH |
2325 | static noinline int btrfs_ioctl_tree_search_v2(struct file *file, |
2326 | void __user *argp) | |
2327 | { | |
2328 | struct btrfs_ioctl_search_args_v2 __user *uarg; | |
2329 | struct btrfs_ioctl_search_args_v2 args; | |
2330 | struct inode *inode; | |
2331 | int ret; | |
2332 | size_t buf_size; | |
ee22184b | 2333 | const size_t buf_limit = SZ_16M; |
cc68a8a5 GH |
2334 | |
2335 | if (!capable(CAP_SYS_ADMIN)) | |
2336 | return -EPERM; | |
2337 | ||
2338 | /* copy search header and buffer size */ | |
2339 | uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; | |
2340 | if (copy_from_user(&args, uarg, sizeof(args))) | |
2341 | return -EFAULT; | |
2342 | ||
2343 | buf_size = args.buf_size; | |
2344 | ||
cc68a8a5 GH |
2345 | /* limit result size to 16MB */ |
2346 | if (buf_size > buf_limit) | |
2347 | buf_size = buf_limit; | |
2348 | ||
2349 | inode = file_inode(file); | |
2350 | ret = search_ioctl(inode, &args.key, &buf_size, | |
718dc5fa | 2351 | (char __user *)(&uarg->buf[0])); |
cc68a8a5 GH |
2352 | if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) |
2353 | ret = -EFAULT; | |
2354 | else if (ret == -EOVERFLOW && | |
2355 | copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) | |
2356 | ret = -EFAULT; | |
2357 | ||
ac8e9819 CM |
2358 | return ret; |
2359 | } | |
2360 | ||
98d377a0 | 2361 | /* |
ac8e9819 CM |
2362 | * Search INODE_REFs to identify path name of 'dirid' directory |
2363 | * in a 'tree_id' tree. and sets path name to 'name'. | |
2364 | */ | |
98d377a0 TH |
2365 | static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, |
2366 | u64 tree_id, u64 dirid, char *name) | |
2367 | { | |
2368 | struct btrfs_root *root; | |
2369 | struct btrfs_key key; | |
ac8e9819 | 2370 | char *ptr; |
98d377a0 TH |
2371 | int ret = -1; |
2372 | int slot; | |
2373 | int len; | |
2374 | int total_len = 0; | |
2375 | struct btrfs_inode_ref *iref; | |
2376 | struct extent_buffer *l; | |
2377 | struct btrfs_path *path; | |
2378 | ||
2379 | if (dirid == BTRFS_FIRST_FREE_OBJECTID) { | |
2380 | name[0]='\0'; | |
2381 | return 0; | |
2382 | } | |
2383 | ||
2384 | path = btrfs_alloc_path(); | |
2385 | if (!path) | |
2386 | return -ENOMEM; | |
2387 | ||
c8bcbfbd | 2388 | ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1]; |
98d377a0 | 2389 | |
56e9357a | 2390 | root = btrfs_get_fs_root(info, tree_id, true); |
98d377a0 | 2391 | if (IS_ERR(root)) { |
ad1e3d56 | 2392 | ret = PTR_ERR(root); |
88234012 JB |
2393 | root = NULL; |
2394 | goto out; | |
2395 | } | |
98d377a0 TH |
2396 | |
2397 | key.objectid = dirid; | |
2398 | key.type = BTRFS_INODE_REF_KEY; | |
8ad6fcab | 2399 | key.offset = (u64)-1; |
98d377a0 | 2400 | |
67871254 | 2401 | while (1) { |
0ff40a91 | 2402 | ret = btrfs_search_backwards(root, &key, path); |
98d377a0 TH |
2403 | if (ret < 0) |
2404 | goto out; | |
18674c6c | 2405 | else if (ret > 0) { |
0ff40a91 MPS |
2406 | ret = -ENOENT; |
2407 | goto out; | |
18674c6c | 2408 | } |
98d377a0 TH |
2409 | |
2410 | l = path->nodes[0]; | |
2411 | slot = path->slots[0]; | |
98d377a0 | 2412 | |
98d377a0 TH |
2413 | iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); |
2414 | len = btrfs_inode_ref_name_len(l, iref); | |
2415 | ptr -= len + 1; | |
2416 | total_len += len + 1; | |
a696cf35 FDBM |
2417 | if (ptr < name) { |
2418 | ret = -ENAMETOOLONG; | |
98d377a0 | 2419 | goto out; |
a696cf35 | 2420 | } |
98d377a0 TH |
2421 | |
2422 | *(ptr + len) = '/'; | |
67871254 | 2423 | read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); |
98d377a0 TH |
2424 | |
2425 | if (key.offset == BTRFS_FIRST_FREE_OBJECTID) | |
2426 | break; | |
2427 | ||
b3b4aa74 | 2428 | btrfs_release_path(path); |
98d377a0 | 2429 | key.objectid = key.offset; |
8ad6fcab | 2430 | key.offset = (u64)-1; |
98d377a0 | 2431 | dirid = key.objectid; |
98d377a0 | 2432 | } |
77906a50 | 2433 | memmove(name, ptr, total_len); |
67871254 | 2434 | name[total_len] = '\0'; |
98d377a0 TH |
2435 | ret = 0; |
2436 | out: | |
00246528 | 2437 | btrfs_put_root(root); |
98d377a0 | 2438 | btrfs_free_path(path); |
ac8e9819 CM |
2439 | return ret; |
2440 | } | |
2441 | ||
6623d9a0 CB |
2442 | static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns, |
2443 | struct inode *inode, | |
23d0b79d TM |
2444 | struct btrfs_ioctl_ino_lookup_user_args *args) |
2445 | { | |
2446 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; | |
2447 | struct super_block *sb = inode->i_sb; | |
2448 | struct btrfs_key upper_limit = BTRFS_I(inode)->location; | |
2449 | u64 treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2450 | u64 dirid = args->dirid; | |
2451 | unsigned long item_off; | |
2452 | unsigned long item_len; | |
2453 | struct btrfs_inode_ref *iref; | |
2454 | struct btrfs_root_ref *rref; | |
b8a49ae1 | 2455 | struct btrfs_root *root = NULL; |
23d0b79d TM |
2456 | struct btrfs_path *path; |
2457 | struct btrfs_key key, key2; | |
2458 | struct extent_buffer *leaf; | |
2459 | struct inode *temp_inode; | |
2460 | char *ptr; | |
2461 | int slot; | |
2462 | int len; | |
2463 | int total_len = 0; | |
2464 | int ret; | |
2465 | ||
2466 | path = btrfs_alloc_path(); | |
2467 | if (!path) | |
2468 | return -ENOMEM; | |
2469 | ||
2470 | /* | |
2471 | * If the bottom subvolume does not exist directly under upper_limit, | |
2472 | * construct the path in from the bottom up. | |
2473 | */ | |
2474 | if (dirid != upper_limit.objectid) { | |
2475 | ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1]; | |
2476 | ||
56e9357a | 2477 | root = btrfs_get_fs_root(fs_info, treeid, true); |
23d0b79d TM |
2478 | if (IS_ERR(root)) { |
2479 | ret = PTR_ERR(root); | |
2480 | goto out; | |
2481 | } | |
2482 | ||
2483 | key.objectid = dirid; | |
2484 | key.type = BTRFS_INODE_REF_KEY; | |
2485 | key.offset = (u64)-1; | |
2486 | while (1) { | |
0ff40a91 MPS |
2487 | ret = btrfs_search_backwards(root, &key, path); |
2488 | if (ret < 0) | |
2489 | goto out_put; | |
2490 | else if (ret > 0) { | |
2491 | ret = -ENOENT; | |
b8a49ae1 | 2492 | goto out_put; |
23d0b79d TM |
2493 | } |
2494 | ||
2495 | leaf = path->nodes[0]; | |
2496 | slot = path->slots[0]; | |
23d0b79d TM |
2497 | |
2498 | iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref); | |
2499 | len = btrfs_inode_ref_name_len(leaf, iref); | |
2500 | ptr -= len + 1; | |
2501 | total_len += len + 1; | |
2502 | if (ptr < args->path) { | |
2503 | ret = -ENAMETOOLONG; | |
b8a49ae1 | 2504 | goto out_put; |
23d0b79d TM |
2505 | } |
2506 | ||
2507 | *(ptr + len) = '/'; | |
2508 | read_extent_buffer(leaf, ptr, | |
2509 | (unsigned long)(iref + 1), len); | |
2510 | ||
2511 | /* Check the read+exec permission of this directory */ | |
2512 | ret = btrfs_previous_item(root, path, dirid, | |
2513 | BTRFS_INODE_ITEM_KEY); | |
2514 | if (ret < 0) { | |
b8a49ae1 | 2515 | goto out_put; |
23d0b79d TM |
2516 | } else if (ret > 0) { |
2517 | ret = -ENOENT; | |
b8a49ae1 | 2518 | goto out_put; |
23d0b79d TM |
2519 | } |
2520 | ||
2521 | leaf = path->nodes[0]; | |
2522 | slot = path->slots[0]; | |
2523 | btrfs_item_key_to_cpu(leaf, &key2, slot); | |
2524 | if (key2.objectid != dirid) { | |
2525 | ret = -ENOENT; | |
b8a49ae1 | 2526 | goto out_put; |
23d0b79d TM |
2527 | } |
2528 | ||
0202e83f | 2529 | temp_inode = btrfs_iget(sb, key2.objectid, root); |
3ca57bd6 MT |
2530 | if (IS_ERR(temp_inode)) { |
2531 | ret = PTR_ERR(temp_inode); | |
b8a49ae1 | 2532 | goto out_put; |
3ca57bd6 | 2533 | } |
6623d9a0 | 2534 | ret = inode_permission(mnt_userns, temp_inode, |
47291baa | 2535 | MAY_READ | MAY_EXEC); |
23d0b79d TM |
2536 | iput(temp_inode); |
2537 | if (ret) { | |
2538 | ret = -EACCES; | |
b8a49ae1 | 2539 | goto out_put; |
23d0b79d TM |
2540 | } |
2541 | ||
2542 | if (key.offset == upper_limit.objectid) | |
2543 | break; | |
2544 | if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) { | |
2545 | ret = -EACCES; | |
b8a49ae1 | 2546 | goto out_put; |
23d0b79d TM |
2547 | } |
2548 | ||
2549 | btrfs_release_path(path); | |
2550 | key.objectid = key.offset; | |
2551 | key.offset = (u64)-1; | |
2552 | dirid = key.objectid; | |
2553 | } | |
2554 | ||
2555 | memmove(args->path, ptr, total_len); | |
2556 | args->path[total_len] = '\0'; | |
00246528 | 2557 | btrfs_put_root(root); |
b8a49ae1 | 2558 | root = NULL; |
23d0b79d TM |
2559 | btrfs_release_path(path); |
2560 | } | |
2561 | ||
2562 | /* Get the bottom subvolume's name from ROOT_REF */ | |
23d0b79d TM |
2563 | key.objectid = treeid; |
2564 | key.type = BTRFS_ROOT_REF_KEY; | |
2565 | key.offset = args->treeid; | |
b8a49ae1 | 2566 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); |
23d0b79d TM |
2567 | if (ret < 0) { |
2568 | goto out; | |
2569 | } else if (ret > 0) { | |
2570 | ret = -ENOENT; | |
2571 | goto out; | |
2572 | } | |
2573 | ||
2574 | leaf = path->nodes[0]; | |
2575 | slot = path->slots[0]; | |
2576 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
2577 | ||
2578 | item_off = btrfs_item_ptr_offset(leaf, slot); | |
2579 | item_len = btrfs_item_size_nr(leaf, slot); | |
2580 | /* Check if dirid in ROOT_REF corresponds to passed dirid */ | |
2581 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
2582 | if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) { | |
2583 | ret = -EINVAL; | |
2584 | goto out; | |
2585 | } | |
2586 | ||
2587 | /* Copy subvolume's name */ | |
2588 | item_off += sizeof(struct btrfs_root_ref); | |
2589 | item_len -= sizeof(struct btrfs_root_ref); | |
2590 | read_extent_buffer(leaf, args->name, item_off, item_len); | |
2591 | args->name[item_len] = 0; | |
2592 | ||
b8a49ae1 | 2593 | out_put: |
00246528 | 2594 | btrfs_put_root(root); |
23d0b79d TM |
2595 | out: |
2596 | btrfs_free_path(path); | |
2597 | return ret; | |
2598 | } | |
2599 | ||
ac8e9819 CM |
2600 | static noinline int btrfs_ioctl_ino_lookup(struct file *file, |
2601 | void __user *argp) | |
2602 | { | |
bece2e82 BVA |
2603 | struct btrfs_ioctl_ino_lookup_args *args; |
2604 | struct inode *inode; | |
01b810b8 | 2605 | int ret = 0; |
ac8e9819 | 2606 | |
2354d08f JL |
2607 | args = memdup_user(argp, sizeof(*args)); |
2608 | if (IS_ERR(args)) | |
2609 | return PTR_ERR(args); | |
c2b96929 | 2610 | |
496ad9aa | 2611 | inode = file_inode(file); |
ac8e9819 | 2612 | |
01b810b8 DS |
2613 | /* |
2614 | * Unprivileged query to obtain the containing subvolume root id. The | |
2615 | * path is reset so it's consistent with btrfs_search_path_in_tree. | |
2616 | */ | |
1b53ac4d CM |
2617 | if (args->treeid == 0) |
2618 | args->treeid = BTRFS_I(inode)->root->root_key.objectid; | |
2619 | ||
01b810b8 DS |
2620 | if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { |
2621 | args->name[0] = 0; | |
2622 | goto out; | |
2623 | } | |
2624 | ||
2625 | if (!capable(CAP_SYS_ADMIN)) { | |
2626 | ret = -EPERM; | |
2627 | goto out; | |
2628 | } | |
2629 | ||
ac8e9819 CM |
2630 | ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, |
2631 | args->treeid, args->objectid, | |
2632 | args->name); | |
2633 | ||
01b810b8 | 2634 | out: |
ac8e9819 CM |
2635 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) |
2636 | ret = -EFAULT; | |
2637 | ||
2638 | kfree(args); | |
98d377a0 TH |
2639 | return ret; |
2640 | } | |
2641 | ||
23d0b79d TM |
2642 | /* |
2643 | * Version of ino_lookup ioctl (unprivileged) | |
2644 | * | |
2645 | * The main differences from ino_lookup ioctl are: | |
2646 | * | |
2647 | * 1. Read + Exec permission will be checked using inode_permission() during | |
2648 | * path construction. -EACCES will be returned in case of failure. | |
2649 | * 2. Path construction will be stopped at the inode number which corresponds | |
2650 | * to the fd with which this ioctl is called. If constructed path does not | |
2651 | * exist under fd's inode, -EACCES will be returned. | |
2652 | * 3. The name of bottom subvolume is also searched and filled. | |
2653 | */ | |
2654 | static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp) | |
2655 | { | |
2656 | struct btrfs_ioctl_ino_lookup_user_args *args; | |
2657 | struct inode *inode; | |
2658 | int ret; | |
2659 | ||
2660 | args = memdup_user(argp, sizeof(*args)); | |
2661 | if (IS_ERR(args)) | |
2662 | return PTR_ERR(args); | |
2663 | ||
2664 | inode = file_inode(file); | |
2665 | ||
2666 | if (args->dirid == BTRFS_FIRST_FREE_OBJECTID && | |
2667 | BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) { | |
2668 | /* | |
2669 | * The subvolume does not exist under fd with which this is | |
2670 | * called | |
2671 | */ | |
2672 | kfree(args); | |
2673 | return -EACCES; | |
2674 | } | |
2675 | ||
6623d9a0 | 2676 | ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args); |
23d0b79d TM |
2677 | |
2678 | if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) | |
2679 | ret = -EFAULT; | |
2680 | ||
2681 | kfree(args); | |
2682 | return ret; | |
2683 | } | |
2684 | ||
b64ec075 TM |
2685 | /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */ |
2686 | static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp) | |
2687 | { | |
2688 | struct btrfs_ioctl_get_subvol_info_args *subvol_info; | |
2689 | struct btrfs_fs_info *fs_info; | |
2690 | struct btrfs_root *root; | |
2691 | struct btrfs_path *path; | |
2692 | struct btrfs_key key; | |
2693 | struct btrfs_root_item *root_item; | |
2694 | struct btrfs_root_ref *rref; | |
2695 | struct extent_buffer *leaf; | |
2696 | unsigned long item_off; | |
2697 | unsigned long item_len; | |
2698 | struct inode *inode; | |
2699 | int slot; | |
2700 | int ret = 0; | |
2701 | ||
2702 | path = btrfs_alloc_path(); | |
2703 | if (!path) | |
2704 | return -ENOMEM; | |
2705 | ||
2706 | subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL); | |
2707 | if (!subvol_info) { | |
2708 | btrfs_free_path(path); | |
2709 | return -ENOMEM; | |
2710 | } | |
2711 | ||
2712 | inode = file_inode(file); | |
2713 | fs_info = BTRFS_I(inode)->root->fs_info; | |
2714 | ||
2715 | /* Get root_item of inode's subvolume */ | |
2716 | key.objectid = BTRFS_I(inode)->root->root_key.objectid; | |
56e9357a | 2717 | root = btrfs_get_fs_root(fs_info, key.objectid, true); |
b64ec075 TM |
2718 | if (IS_ERR(root)) { |
2719 | ret = PTR_ERR(root); | |
04734e84 JB |
2720 | goto out_free; |
2721 | } | |
b64ec075 TM |
2722 | root_item = &root->root_item; |
2723 | ||
2724 | subvol_info->treeid = key.objectid; | |
2725 | ||
2726 | subvol_info->generation = btrfs_root_generation(root_item); | |
2727 | subvol_info->flags = btrfs_root_flags(root_item); | |
2728 | ||
2729 | memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE); | |
2730 | memcpy(subvol_info->parent_uuid, root_item->parent_uuid, | |
2731 | BTRFS_UUID_SIZE); | |
2732 | memcpy(subvol_info->received_uuid, root_item->received_uuid, | |
2733 | BTRFS_UUID_SIZE); | |
2734 | ||
2735 | subvol_info->ctransid = btrfs_root_ctransid(root_item); | |
2736 | subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime); | |
2737 | subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime); | |
2738 | ||
2739 | subvol_info->otransid = btrfs_root_otransid(root_item); | |
2740 | subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime); | |
2741 | subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime); | |
2742 | ||
2743 | subvol_info->stransid = btrfs_root_stransid(root_item); | |
2744 | subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime); | |
2745 | subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime); | |
2746 | ||
2747 | subvol_info->rtransid = btrfs_root_rtransid(root_item); | |
2748 | subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime); | |
2749 | subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime); | |
2750 | ||
2751 | if (key.objectid != BTRFS_FS_TREE_OBJECTID) { | |
2752 | /* Search root tree for ROOT_BACKREF of this subvolume */ | |
b64ec075 TM |
2753 | key.type = BTRFS_ROOT_BACKREF_KEY; |
2754 | key.offset = 0; | |
04734e84 | 2755 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); |
b64ec075 TM |
2756 | if (ret < 0) { |
2757 | goto out; | |
2758 | } else if (path->slots[0] >= | |
2759 | btrfs_header_nritems(path->nodes[0])) { | |
04734e84 | 2760 | ret = btrfs_next_leaf(fs_info->tree_root, path); |
b64ec075 TM |
2761 | if (ret < 0) { |
2762 | goto out; | |
2763 | } else if (ret > 0) { | |
2764 | ret = -EUCLEAN; | |
2765 | goto out; | |
2766 | } | |
2767 | } | |
2768 | ||
2769 | leaf = path->nodes[0]; | |
2770 | slot = path->slots[0]; | |
2771 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
2772 | if (key.objectid == subvol_info->treeid && | |
2773 | key.type == BTRFS_ROOT_BACKREF_KEY) { | |
2774 | subvol_info->parent_id = key.offset; | |
2775 | ||
2776 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
2777 | subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref); | |
2778 | ||
2779 | item_off = btrfs_item_ptr_offset(leaf, slot) | |
2780 | + sizeof(struct btrfs_root_ref); | |
2781 | item_len = btrfs_item_size_nr(leaf, slot) | |
2782 | - sizeof(struct btrfs_root_ref); | |
2783 | read_extent_buffer(leaf, subvol_info->name, | |
2784 | item_off, item_len); | |
2785 | } else { | |
2786 | ret = -ENOENT; | |
2787 | goto out; | |
2788 | } | |
2789 | } | |
2790 | ||
2791 | if (copy_to_user(argp, subvol_info, sizeof(*subvol_info))) | |
2792 | ret = -EFAULT; | |
2793 | ||
2794 | out: | |
00246528 | 2795 | btrfs_put_root(root); |
04734e84 | 2796 | out_free: |
b64ec075 | 2797 | btrfs_free_path(path); |
b091f7fe | 2798 | kfree(subvol_info); |
b64ec075 TM |
2799 | return ret; |
2800 | } | |
2801 | ||
42e4b520 TM |
2802 | /* |
2803 | * Return ROOT_REF information of the subvolume containing this inode | |
2804 | * except the subvolume name. | |
2805 | */ | |
2806 | static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp) | |
2807 | { | |
2808 | struct btrfs_ioctl_get_subvol_rootref_args *rootrefs; | |
2809 | struct btrfs_root_ref *rref; | |
2810 | struct btrfs_root *root; | |
2811 | struct btrfs_path *path; | |
2812 | struct btrfs_key key; | |
2813 | struct extent_buffer *leaf; | |
2814 | struct inode *inode; | |
2815 | u64 objectid; | |
2816 | int slot; | |
2817 | int ret; | |
2818 | u8 found; | |
2819 | ||
2820 | path = btrfs_alloc_path(); | |
2821 | if (!path) | |
2822 | return -ENOMEM; | |
2823 | ||
2824 | rootrefs = memdup_user(argp, sizeof(*rootrefs)); | |
2825 | if (IS_ERR(rootrefs)) { | |
2826 | btrfs_free_path(path); | |
2827 | return PTR_ERR(rootrefs); | |
2828 | } | |
2829 | ||
2830 | inode = file_inode(file); | |
2831 | root = BTRFS_I(inode)->root->fs_info->tree_root; | |
2832 | objectid = BTRFS_I(inode)->root->root_key.objectid; | |
2833 | ||
2834 | key.objectid = objectid; | |
2835 | key.type = BTRFS_ROOT_REF_KEY; | |
2836 | key.offset = rootrefs->min_treeid; | |
2837 | found = 0; | |
2838 | ||
2839 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
2840 | if (ret < 0) { | |
2841 | goto out; | |
2842 | } else if (path->slots[0] >= | |
2843 | btrfs_header_nritems(path->nodes[0])) { | |
2844 | ret = btrfs_next_leaf(root, path); | |
2845 | if (ret < 0) { | |
2846 | goto out; | |
2847 | } else if (ret > 0) { | |
2848 | ret = -EUCLEAN; | |
2849 | goto out; | |
2850 | } | |
2851 | } | |
2852 | while (1) { | |
2853 | leaf = path->nodes[0]; | |
2854 | slot = path->slots[0]; | |
2855 | ||
2856 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
2857 | if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) { | |
2858 | ret = 0; | |
2859 | goto out; | |
2860 | } | |
2861 | ||
2862 | if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) { | |
2863 | ret = -EOVERFLOW; | |
2864 | goto out; | |
2865 | } | |
2866 | ||
2867 | rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref); | |
2868 | rootrefs->rootref[found].treeid = key.offset; | |
2869 | rootrefs->rootref[found].dirid = | |
2870 | btrfs_root_ref_dirid(leaf, rref); | |
2871 | found++; | |
2872 | ||
2873 | ret = btrfs_next_item(root, path); | |
2874 | if (ret < 0) { | |
2875 | goto out; | |
2876 | } else if (ret > 0) { | |
2877 | ret = -EUCLEAN; | |
2878 | goto out; | |
2879 | } | |
2880 | } | |
2881 | ||
2882 | out: | |
2883 | if (!ret || ret == -EOVERFLOW) { | |
2884 | rootrefs->num_items = found; | |
2885 | /* update min_treeid for next search */ | |
2886 | if (found) | |
2887 | rootrefs->min_treeid = | |
2888 | rootrefs->rootref[found - 1].treeid + 1; | |
2889 | if (copy_to_user(argp, rootrefs, sizeof(*rootrefs))) | |
2890 | ret = -EFAULT; | |
2891 | } | |
2892 | ||
2893 | kfree(rootrefs); | |
2894 | btrfs_free_path(path); | |
2895 | ||
2896 | return ret; | |
2897 | } | |
2898 | ||
76dda93c | 2899 | static noinline int btrfs_ioctl_snap_destroy(struct file *file, |
949964c9 MPS |
2900 | void __user *arg, |
2901 | bool destroy_v2) | |
76dda93c | 2902 | { |
54563d41 | 2903 | struct dentry *parent = file->f_path.dentry; |
0b246afa | 2904 | struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); |
76dda93c | 2905 | struct dentry *dentry; |
2b0143b5 | 2906 | struct inode *dir = d_inode(parent); |
76dda93c YZ |
2907 | struct inode *inode; |
2908 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
2909 | struct btrfs_root *dest = NULL; | |
949964c9 MPS |
2910 | struct btrfs_ioctl_vol_args *vol_args = NULL; |
2911 | struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL; | |
c4ed533b | 2912 | struct user_namespace *mnt_userns = file_mnt_user_ns(file); |
949964c9 MPS |
2913 | char *subvol_name, *subvol_name_ptr = NULL; |
2914 | int subvol_namelen; | |
76dda93c | 2915 | int err = 0; |
949964c9 | 2916 | bool destroy_parent = false; |
76dda93c | 2917 | |
949964c9 MPS |
2918 | if (destroy_v2) { |
2919 | vol_args2 = memdup_user(arg, sizeof(*vol_args2)); | |
2920 | if (IS_ERR(vol_args2)) | |
2921 | return PTR_ERR(vol_args2); | |
325c50e3 | 2922 | |
949964c9 MPS |
2923 | if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) { |
2924 | err = -EOPNOTSUPP; | |
2925 | goto out; | |
2926 | } | |
76dda93c | 2927 | |
949964c9 MPS |
2928 | /* |
2929 | * If SPEC_BY_ID is not set, we are looking for the subvolume by | |
2930 | * name, same as v1 currently does. | |
2931 | */ | |
2932 | if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) { | |
2933 | vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0; | |
2934 | subvol_name = vol_args2->name; | |
2935 | ||
2936 | err = mnt_want_write_file(file); | |
2937 | if (err) | |
2938 | goto out; | |
2939 | } else { | |
aabb34e7 | 2940 | struct inode *old_dir; |
c4ed533b | 2941 | |
949964c9 MPS |
2942 | if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) { |
2943 | err = -EINVAL; | |
2944 | goto out; | |
2945 | } | |
2946 | ||
2947 | err = mnt_want_write_file(file); | |
2948 | if (err) | |
2949 | goto out; | |
2950 | ||
2951 | dentry = btrfs_get_dentry(fs_info->sb, | |
2952 | BTRFS_FIRST_FREE_OBJECTID, | |
2953 | vol_args2->subvolid, 0, 0); | |
2954 | if (IS_ERR(dentry)) { | |
2955 | err = PTR_ERR(dentry); | |
2956 | goto out_drop_write; | |
2957 | } | |
2958 | ||
2959 | /* | |
2960 | * Change the default parent since the subvolume being | |
2961 | * deleted can be outside of the current mount point. | |
2962 | */ | |
2963 | parent = btrfs_get_parent(dentry); | |
2964 | ||
2965 | /* | |
2966 | * At this point dentry->d_name can point to '/' if the | |
2967 | * subvolume we want to destroy is outsite of the | |
2968 | * current mount point, so we need to release the | |
2969 | * current dentry and execute the lookup to return a new | |
2970 | * one with ->d_name pointing to the | |
2971 | * <mount point>/subvol_name. | |
2972 | */ | |
2973 | dput(dentry); | |
2974 | if (IS_ERR(parent)) { | |
2975 | err = PTR_ERR(parent); | |
2976 | goto out_drop_write; | |
2977 | } | |
aabb34e7 | 2978 | old_dir = dir; |
949964c9 MPS |
2979 | dir = d_inode(parent); |
2980 | ||
2981 | /* | |
2982 | * If v2 was used with SPEC_BY_ID, a new parent was | |
2983 | * allocated since the subvolume can be outside of the | |
2984 | * current mount point. Later on we need to release this | |
2985 | * new parent dentry. | |
2986 | */ | |
2987 | destroy_parent = true; | |
2988 | ||
aabb34e7 CB |
2989 | /* |
2990 | * On idmapped mounts, deletion via subvolid is | |
2991 | * restricted to subvolumes that are immediate | |
2992 | * ancestors of the inode referenced by the file | |
2993 | * descriptor in the ioctl. Otherwise the idmapping | |
2994 | * could potentially be abused to delete subvolumes | |
2995 | * anywhere in the filesystem the user wouldn't be able | |
2996 | * to delete without an idmapped mount. | |
2997 | */ | |
2998 | if (old_dir != dir && mnt_userns != &init_user_ns) { | |
2999 | err = -EOPNOTSUPP; | |
3000 | goto free_parent; | |
3001 | } | |
3002 | ||
949964c9 MPS |
3003 | subvol_name_ptr = btrfs_get_subvol_name_from_objectid( |
3004 | fs_info, vol_args2->subvolid); | |
3005 | if (IS_ERR(subvol_name_ptr)) { | |
3006 | err = PTR_ERR(subvol_name_ptr); | |
3007 | goto free_parent; | |
3008 | } | |
1a9fd417 | 3009 | /* subvol_name_ptr is already nul terminated */ |
949964c9 MPS |
3010 | subvol_name = (char *)kbasename(subvol_name_ptr); |
3011 | } | |
3012 | } else { | |
3013 | vol_args = memdup_user(arg, sizeof(*vol_args)); | |
3014 | if (IS_ERR(vol_args)) | |
3015 | return PTR_ERR(vol_args); | |
3016 | ||
3017 | vol_args->name[BTRFS_PATH_NAME_MAX] = 0; | |
3018 | subvol_name = vol_args->name; | |
3019 | ||
3020 | err = mnt_want_write_file(file); | |
3021 | if (err) | |
3022 | goto out; | |
76dda93c YZ |
3023 | } |
3024 | ||
949964c9 | 3025 | subvol_namelen = strlen(subvol_name); |
76dda93c | 3026 | |
949964c9 MPS |
3027 | if (strchr(subvol_name, '/') || |
3028 | strncmp(subvol_name, "..", subvol_namelen) == 0) { | |
3029 | err = -EINVAL; | |
3030 | goto free_subvol_name; | |
3031 | } | |
3032 | ||
3033 | if (!S_ISDIR(dir->i_mode)) { | |
3034 | err = -ENOTDIR; | |
3035 | goto free_subvol_name; | |
3036 | } | |
521e0546 | 3037 | |
00235411 AV |
3038 | err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); |
3039 | if (err == -EINTR) | |
949964c9 | 3040 | goto free_subvol_name; |
c4ed533b | 3041 | dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen); |
76dda93c YZ |
3042 | if (IS_ERR(dentry)) { |
3043 | err = PTR_ERR(dentry); | |
3044 | goto out_unlock_dir; | |
3045 | } | |
3046 | ||
2b0143b5 | 3047 | if (d_really_is_negative(dentry)) { |
76dda93c YZ |
3048 | err = -ENOENT; |
3049 | goto out_dput; | |
3050 | } | |
3051 | ||
2b0143b5 | 3052 | inode = d_inode(dentry); |
4260f7c7 | 3053 | dest = BTRFS_I(inode)->root; |
67871254 | 3054 | if (!capable(CAP_SYS_ADMIN)) { |
4260f7c7 SW |
3055 | /* |
3056 | * Regular user. Only allow this with a special mount | |
3057 | * option, when the user has write+exec access to the | |
3058 | * subvol root, and when rmdir(2) would have been | |
3059 | * allowed. | |
3060 | * | |
3061 | * Note that this is _not_ check that the subvol is | |
3062 | * empty or doesn't contain data that we wouldn't | |
3063 | * otherwise be able to delete. | |
3064 | * | |
3065 | * Users who want to delete empty subvols should try | |
3066 | * rmdir(2). | |
3067 | */ | |
3068 | err = -EPERM; | |
0b246afa | 3069 | if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) |
4260f7c7 SW |
3070 | goto out_dput; |
3071 | ||
3072 | /* | |
3073 | * Do not allow deletion if the parent dir is the same | |
3074 | * as the dir to be deleted. That means the ioctl | |
3075 | * must be called on the dentry referencing the root | |
3076 | * of the subvol, not a random directory contained | |
3077 | * within it. | |
3078 | */ | |
3079 | err = -EINVAL; | |
3080 | if (root == dest) | |
3081 | goto out_dput; | |
3082 | ||
c4ed533b | 3083 | err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC); |
4260f7c7 SW |
3084 | if (err) |
3085 | goto out_dput; | |
4260f7c7 SW |
3086 | } |
3087 | ||
5c39da5b | 3088 | /* check if subvolume may be deleted by a user */ |
c4ed533b | 3089 | err = btrfs_may_delete(mnt_userns, dir, dentry, 1); |
5c39da5b MX |
3090 | if (err) |
3091 | goto out_dput; | |
3092 | ||
4a0cc7ca | 3093 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
76dda93c YZ |
3094 | err = -EINVAL; |
3095 | goto out_dput; | |
3096 | } | |
3097 | ||
64708539 | 3098 | btrfs_inode_lock(inode, 0); |
f60a2364 | 3099 | err = btrfs_delete_subvolume(dir, dentry); |
64708539 | 3100 | btrfs_inode_unlock(inode, 0); |
c500c68d AG |
3101 | if (!err) |
3102 | d_delete_notify(dir, dentry); | |
fa6ac876 | 3103 | |
76dda93c YZ |
3104 | out_dput: |
3105 | dput(dentry); | |
3106 | out_unlock_dir: | |
64708539 | 3107 | btrfs_inode_unlock(dir, 0); |
949964c9 MPS |
3108 | free_subvol_name: |
3109 | kfree(subvol_name_ptr); | |
3110 | free_parent: | |
3111 | if (destroy_parent) | |
3112 | dput(parent); | |
00235411 | 3113 | out_drop_write: |
2a79f17e | 3114 | mnt_drop_write_file(file); |
76dda93c | 3115 | out: |
949964c9 | 3116 | kfree(vol_args2); |
76dda93c YZ |
3117 | kfree(vol_args); |
3118 | return err; | |
3119 | } | |
3120 | ||
1e701a32 | 3121 | static int btrfs_ioctl_defrag(struct file *file, void __user *argp) |
f46b5a66 | 3122 | { |
496ad9aa | 3123 | struct inode *inode = file_inode(file); |
f46b5a66 | 3124 | struct btrfs_root *root = BTRFS_I(inode)->root; |
c853a578 | 3125 | struct btrfs_ioctl_defrag_range_args range = {0}; |
c146afad YZ |
3126 | int ret; |
3127 | ||
25122d15 ID |
3128 | ret = mnt_want_write_file(file); |
3129 | if (ret) | |
3130 | return ret; | |
b83cc969 | 3131 | |
25122d15 ID |
3132 | if (btrfs_root_readonly(root)) { |
3133 | ret = -EROFS; | |
3134 | goto out; | |
5ac00add | 3135 | } |
f46b5a66 | 3136 | |
95ea0486 QW |
3137 | /* Subpage defrag will be supported in later commits */ |
3138 | if (root->fs_info->sectorsize < PAGE_SIZE) { | |
3139 | ret = -ENOTTY; | |
3140 | goto out; | |
3141 | } | |
3142 | ||
f46b5a66 CH |
3143 | switch (inode->i_mode & S_IFMT) { |
3144 | case S_IFDIR: | |
e441d54d CM |
3145 | if (!capable(CAP_SYS_ADMIN)) { |
3146 | ret = -EPERM; | |
3147 | goto out; | |
3148 | } | |
de78b51a | 3149 | ret = btrfs_defrag_root(root); |
f46b5a66 CH |
3150 | break; |
3151 | case S_IFREG: | |
616d374e AB |
3152 | /* |
3153 | * Note that this does not check the file descriptor for write | |
3154 | * access. This prevents defragmenting executables that are | |
3155 | * running and allows defrag on files open in read-only mode. | |
3156 | */ | |
3157 | if (!capable(CAP_SYS_ADMIN) && | |
47291baa | 3158 | inode_permission(&init_user_ns, inode, MAY_WRITE)) { |
616d374e | 3159 | ret = -EPERM; |
e441d54d CM |
3160 | goto out; |
3161 | } | |
1e701a32 | 3162 | |
1e701a32 | 3163 | if (argp) { |
c853a578 | 3164 | if (copy_from_user(&range, argp, sizeof(range))) { |
1e701a32 | 3165 | ret = -EFAULT; |
683be16e | 3166 | goto out; |
1e701a32 CM |
3167 | } |
3168 | /* compression requires us to start the IO */ | |
c853a578 GR |
3169 | if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { |
3170 | range.flags |= BTRFS_DEFRAG_RANGE_START_IO; | |
3171 | range.extent_thresh = (u32)-1; | |
1e701a32 CM |
3172 | } |
3173 | } else { | |
3174 | /* the rest are all set to zero by kzalloc */ | |
c853a578 | 3175 | range.len = (u64)-1; |
1e701a32 | 3176 | } |
496ad9aa | 3177 | ret = btrfs_defrag_file(file_inode(file), file, |
c853a578 | 3178 | &range, BTRFS_OLDEST_GENERATION, 0); |
4cb5300b CM |
3179 | if (ret > 0) |
3180 | ret = 0; | |
f46b5a66 | 3181 | break; |
8929ecfa YZ |
3182 | default: |
3183 | ret = -EINVAL; | |
f46b5a66 | 3184 | } |
e441d54d | 3185 | out: |
25122d15 | 3186 | mnt_drop_write_file(file); |
e441d54d | 3187 | return ret; |
f46b5a66 CH |
3188 | } |
3189 | ||
2ff7e61e | 3190 | static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) |
f46b5a66 CH |
3191 | { |
3192 | struct btrfs_ioctl_vol_args *vol_args; | |
3193 | int ret; | |
3194 | ||
e441d54d CM |
3195 | if (!capable(CAP_SYS_ADMIN)) |
3196 | return -EPERM; | |
3197 | ||
c3e1f96c | 3198 | if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) |
e57138b3 | 3199 | return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
c9e9f97b | 3200 | |
dae7b665 | 3201 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
3202 | if (IS_ERR(vol_args)) { |
3203 | ret = PTR_ERR(vol_args); | |
3204 | goto out; | |
3205 | } | |
f46b5a66 | 3206 | |
5516e595 | 3207 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
2ff7e61e | 3208 | ret = btrfs_init_new_device(fs_info, vol_args->name); |
f46b5a66 | 3209 | |
43d20761 | 3210 | if (!ret) |
0b246afa | 3211 | btrfs_info(fs_info, "disk added %s", vol_args->name); |
43d20761 | 3212 | |
f46b5a66 | 3213 | kfree(vol_args); |
c9e9f97b | 3214 | out: |
c3e1f96c | 3215 | btrfs_exclop_finish(fs_info); |
f46b5a66 CH |
3216 | return ret; |
3217 | } | |
3218 | ||
6b526ed7 | 3219 | static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) |
f46b5a66 | 3220 | { |
0b246afa JM |
3221 | struct inode *inode = file_inode(file); |
3222 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
6b526ed7 | 3223 | struct btrfs_ioctl_vol_args_v2 *vol_args; |
3fa421de JB |
3224 | struct block_device *bdev = NULL; |
3225 | fmode_t mode; | |
f46b5a66 | 3226 | int ret; |
67ae34b6 | 3227 | bool cancel = false; |
f46b5a66 | 3228 | |
e441d54d CM |
3229 | if (!capable(CAP_SYS_ADMIN)) |
3230 | return -EPERM; | |
3231 | ||
da24927b MX |
3232 | ret = mnt_want_write_file(file); |
3233 | if (ret) | |
3234 | return ret; | |
c146afad | 3235 | |
dae7b665 | 3236 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
c9e9f97b ID |
3237 | if (IS_ERR(vol_args)) { |
3238 | ret = PTR_ERR(vol_args); | |
c47ca32d | 3239 | goto err_drop; |
c9e9f97b | 3240 | } |
f46b5a66 | 3241 | |
748449cd | 3242 | if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) { |
fd4e994b OS |
3243 | ret = -EOPNOTSUPP; |
3244 | goto out; | |
3245 | } | |
67ae34b6 DS |
3246 | vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; |
3247 | if (!(vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) && | |
3248 | strcmp("cancel", vol_args->name) == 0) | |
3249 | cancel = true; | |
f46b5a66 | 3250 | |
67ae34b6 DS |
3251 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, |
3252 | cancel); | |
3253 | if (ret) | |
183860f6 | 3254 | goto out; |
67ae34b6 | 3255 | /* Exclusive operation is now claimed */ |
183860f6 | 3256 | |
67ae34b6 | 3257 | if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) |
3fa421de | 3258 | ret = btrfs_rm_device(fs_info, NULL, vol_args->devid, &bdev, &mode); |
67ae34b6 | 3259 | else |
3fa421de | 3260 | ret = btrfs_rm_device(fs_info, vol_args->name, 0, &bdev, &mode); |
67ae34b6 | 3261 | |
c3e1f96c | 3262 | btrfs_exclop_finish(fs_info); |
183860f6 | 3263 | |
6b526ed7 | 3264 | if (!ret) { |
735654ea | 3265 | if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) |
0b246afa | 3266 | btrfs_info(fs_info, "device deleted: id %llu", |
6b526ed7 AJ |
3267 | vol_args->devid); |
3268 | else | |
0b246afa | 3269 | btrfs_info(fs_info, "device deleted: %s", |
6b526ed7 AJ |
3270 | vol_args->name); |
3271 | } | |
183860f6 AJ |
3272 | out: |
3273 | kfree(vol_args); | |
c47ca32d | 3274 | err_drop: |
4ac20c70 | 3275 | mnt_drop_write_file(file); |
3fa421de JB |
3276 | if (bdev) |
3277 | blkdev_put(bdev, mode); | |
f46b5a66 CH |
3278 | return ret; |
3279 | } | |
3280 | ||
da24927b | 3281 | static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) |
f46b5a66 | 3282 | { |
0b246afa JM |
3283 | struct inode *inode = file_inode(file); |
3284 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
f46b5a66 | 3285 | struct btrfs_ioctl_vol_args *vol_args; |
3fa421de JB |
3286 | struct block_device *bdev = NULL; |
3287 | fmode_t mode; | |
f46b5a66 | 3288 | int ret; |
67ae34b6 | 3289 | bool cancel; |
f46b5a66 | 3290 | |
e441d54d CM |
3291 | if (!capable(CAP_SYS_ADMIN)) |
3292 | return -EPERM; | |
3293 | ||
da24927b MX |
3294 | ret = mnt_want_write_file(file); |
3295 | if (ret) | |
3296 | return ret; | |
c146afad | 3297 | |
58d7bbf8 DS |
3298 | vol_args = memdup_user(arg, sizeof(*vol_args)); |
3299 | if (IS_ERR(vol_args)) { | |
3300 | ret = PTR_ERR(vol_args); | |
67ae34b6 | 3301 | goto out_drop_write; |
183860f6 | 3302 | } |
58d7bbf8 | 3303 | vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; |
67ae34b6 DS |
3304 | cancel = (strcmp("cancel", vol_args->name) == 0); |
3305 | ||
3306 | ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE, | |
3307 | cancel); | |
3308 | if (ret == 0) { | |
3fa421de | 3309 | ret = btrfs_rm_device(fs_info, vol_args->name, 0, &bdev, &mode); |
67ae34b6 DS |
3310 | if (!ret) |
3311 | btrfs_info(fs_info, "disk deleted %s", vol_args->name); | |
3312 | btrfs_exclop_finish(fs_info); | |
3313 | } | |
183860f6 | 3314 | |
183860f6 | 3315 | kfree(vol_args); |
58d7bbf8 | 3316 | out_drop_write: |
4ac20c70 | 3317 | mnt_drop_write_file(file); |
3fa421de JB |
3318 | if (bdev) |
3319 | blkdev_put(bdev, mode); | |
f46b5a66 CH |
3320 | return ret; |
3321 | } | |
3322 | ||
2ff7e61e JM |
3323 | static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, |
3324 | void __user *arg) | |
475f6387 | 3325 | { |
027ed2f0 | 3326 | struct btrfs_ioctl_fs_info_args *fi_args; |
475f6387 | 3327 | struct btrfs_device *device; |
0b246afa | 3328 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
137c5418 | 3329 | u64 flags_in; |
027ed2f0 | 3330 | int ret = 0; |
475f6387 | 3331 | |
137c5418 JT |
3332 | fi_args = memdup_user(arg, sizeof(*fi_args)); |
3333 | if (IS_ERR(fi_args)) | |
3334 | return PTR_ERR(fi_args); | |
3335 | ||
3336 | flags_in = fi_args->flags; | |
3337 | memset(fi_args, 0, sizeof(*fi_args)); | |
027ed2f0 | 3338 | |
d03262c7 | 3339 | rcu_read_lock(); |
027ed2f0 | 3340 | fi_args->num_devices = fs_devices->num_devices; |
475f6387 | 3341 | |
d03262c7 | 3342 | list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { |
027ed2f0 LZ |
3343 | if (device->devid > fi_args->max_id) |
3344 | fi_args->max_id = device->devid; | |
475f6387 | 3345 | } |
d03262c7 | 3346 | rcu_read_unlock(); |
475f6387 | 3347 | |
de37aa51 | 3348 | memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid)); |
bea7eafd OS |
3349 | fi_args->nodesize = fs_info->nodesize; |
3350 | fi_args->sectorsize = fs_info->sectorsize; | |
3351 | fi_args->clone_alignment = fs_info->sectorsize; | |
80a773fb | 3352 | |
137c5418 JT |
3353 | if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) { |
3354 | fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy); | |
3355 | fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy); | |
3356 | fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO; | |
3357 | } | |
3358 | ||
0fb408a5 JT |
3359 | if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) { |
3360 | fi_args->generation = fs_info->generation; | |
3361 | fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION; | |
3362 | } | |
3363 | ||
49bac897 JT |
3364 | if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) { |
3365 | memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid, | |
3366 | sizeof(fi_args->metadata_uuid)); | |
3367 | fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID; | |
3368 | } | |
3369 | ||
027ed2f0 LZ |
3370 | if (copy_to_user(arg, fi_args, sizeof(*fi_args))) |
3371 | ret = -EFAULT; | |
475f6387 | 3372 | |
027ed2f0 LZ |
3373 | kfree(fi_args); |
3374 | return ret; | |
475f6387 JS |
3375 | } |
3376 | ||
2ff7e61e JM |
3377 | static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, |
3378 | void __user *arg) | |
475f6387 JS |
3379 | { |
3380 | struct btrfs_ioctl_dev_info_args *di_args; | |
3381 | struct btrfs_device *dev; | |
475f6387 JS |
3382 | int ret = 0; |
3383 | char *s_uuid = NULL; | |
475f6387 | 3384 | |
475f6387 JS |
3385 | di_args = memdup_user(arg, sizeof(*di_args)); |
3386 | if (IS_ERR(di_args)) | |
3387 | return PTR_ERR(di_args); | |
3388 | ||
dd5f9615 | 3389 | if (!btrfs_is_empty_uuid(di_args->uuid)) |
475f6387 JS |
3390 | s_uuid = di_args->uuid; |
3391 | ||
c5593ca3 | 3392 | rcu_read_lock(); |
e4319cd9 | 3393 | dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid, |
b2598edf | 3394 | NULL); |
475f6387 JS |
3395 | |
3396 | if (!dev) { | |
3397 | ret = -ENODEV; | |
3398 | goto out; | |
3399 | } | |
3400 | ||
3401 | di_args->devid = dev->devid; | |
7cc8e58d MX |
3402 | di_args->bytes_used = btrfs_device_get_bytes_used(dev); |
3403 | di_args->total_bytes = btrfs_device_get_total_bytes(dev); | |
475f6387 | 3404 | memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); |
a27202fb | 3405 | if (dev->name) { |
672d5990 MT |
3406 | strncpy(di_args->path, rcu_str_deref(dev->name), |
3407 | sizeof(di_args->path) - 1); | |
a27202fb JM |
3408 | di_args->path[sizeof(di_args->path) - 1] = 0; |
3409 | } else { | |
99ba55ad | 3410 | di_args->path[0] = '\0'; |
a27202fb | 3411 | } |
475f6387 JS |
3412 | |
3413 | out: | |
c5593ca3 | 3414 | rcu_read_unlock(); |
475f6387 JS |
3415 | if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) |
3416 | ret = -EFAULT; | |
3417 | ||
3418 | kfree(di_args); | |
3419 | return ret; | |
3420 | } | |
3421 | ||
6ef5ed0d JB |
3422 | static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) |
3423 | { | |
496ad9aa | 3424 | struct inode *inode = file_inode(file); |
0b246afa | 3425 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
6ef5ed0d JB |
3426 | struct btrfs_root *root = BTRFS_I(inode)->root; |
3427 | struct btrfs_root *new_root; | |
3428 | struct btrfs_dir_item *di; | |
3429 | struct btrfs_trans_handle *trans; | |
2a2b5d62 | 3430 | struct btrfs_path *path = NULL; |
6ef5ed0d | 3431 | struct btrfs_disk_key disk_key; |
6ef5ed0d JB |
3432 | u64 objectid = 0; |
3433 | u64 dir_id; | |
3c04ce01 | 3434 | int ret; |
6ef5ed0d JB |
3435 | |
3436 | if (!capable(CAP_SYS_ADMIN)) | |
3437 | return -EPERM; | |
3438 | ||
3c04ce01 MX |
3439 | ret = mnt_want_write_file(file); |
3440 | if (ret) | |
3441 | return ret; | |
3442 | ||
3443 | if (copy_from_user(&objectid, argp, sizeof(objectid))) { | |
3444 | ret = -EFAULT; | |
3445 | goto out; | |
3446 | } | |
6ef5ed0d JB |
3447 | |
3448 | if (!objectid) | |
1cecf579 | 3449 | objectid = BTRFS_FS_TREE_OBJECTID; |
6ef5ed0d | 3450 | |
56e9357a | 3451 | new_root = btrfs_get_fs_root(fs_info, objectid, true); |
3c04ce01 MX |
3452 | if (IS_ERR(new_root)) { |
3453 | ret = PTR_ERR(new_root); | |
3454 | goto out; | |
3455 | } | |
2a2b5d62 JB |
3456 | if (!is_fstree(new_root->root_key.objectid)) { |
3457 | ret = -ENOENT; | |
3458 | goto out_free; | |
3459 | } | |
6ef5ed0d | 3460 | |
6ef5ed0d | 3461 | path = btrfs_alloc_path(); |
3c04ce01 MX |
3462 | if (!path) { |
3463 | ret = -ENOMEM; | |
2a2b5d62 | 3464 | goto out_free; |
3c04ce01 | 3465 | } |
6ef5ed0d JB |
3466 | |
3467 | trans = btrfs_start_transaction(root, 1); | |
98d5dc13 | 3468 | if (IS_ERR(trans)) { |
3c04ce01 | 3469 | ret = PTR_ERR(trans); |
2a2b5d62 | 3470 | goto out_free; |
6ef5ed0d JB |
3471 | } |
3472 | ||
0b246afa JM |
3473 | dir_id = btrfs_super_root_dir(fs_info->super_copy); |
3474 | di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, | |
6ef5ed0d | 3475 | dir_id, "default", 7, 1); |
cf1e99a4 | 3476 | if (IS_ERR_OR_NULL(di)) { |
2a2b5d62 | 3477 | btrfs_release_path(path); |
3a45bb20 | 3478 | btrfs_end_transaction(trans); |
0b246afa | 3479 | btrfs_err(fs_info, |
5d163e0e | 3480 | "Umm, you don't have the default diritem, this isn't going to work"); |
3c04ce01 | 3481 | ret = -ENOENT; |
2a2b5d62 | 3482 | goto out_free; |
6ef5ed0d JB |
3483 | } |
3484 | ||
3485 | btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); | |
3486 | btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); | |
3487 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
2a2b5d62 | 3488 | btrfs_release_path(path); |
6ef5ed0d | 3489 | |
0b246afa | 3490 | btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); |
3a45bb20 | 3491 | btrfs_end_transaction(trans); |
2a2b5d62 | 3492 | out_free: |
00246528 | 3493 | btrfs_put_root(new_root); |
2a2b5d62 | 3494 | btrfs_free_path(path); |
3c04ce01 MX |
3495 | out: |
3496 | mnt_drop_write_file(file); | |
3497 | return ret; | |
6ef5ed0d JB |
3498 | } |
3499 | ||
c065f5b1 SY |
3500 | static void get_block_group_info(struct list_head *groups_list, |
3501 | struct btrfs_ioctl_space_info *space) | |
bf5fc093 | 3502 | { |
32da5386 | 3503 | struct btrfs_block_group *block_group; |
bf5fc093 JB |
3504 | |
3505 | space->total_bytes = 0; | |
3506 | space->used_bytes = 0; | |
3507 | space->flags = 0; | |
3508 | list_for_each_entry(block_group, groups_list, list) { | |
3509 | space->flags = block_group->flags; | |
b3470b5d | 3510 | space->total_bytes += block_group->length; |
bf38be65 | 3511 | space->used_bytes += block_group->used; |
bf5fc093 JB |
3512 | } |
3513 | } | |
3514 | ||
2ff7e61e JM |
3515 | static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, |
3516 | void __user *arg) | |
1406e432 JB |
3517 | { |
3518 | struct btrfs_ioctl_space_args space_args; | |
3519 | struct btrfs_ioctl_space_info space; | |
3520 | struct btrfs_ioctl_space_info *dest; | |
7fde62bf | 3521 | struct btrfs_ioctl_space_info *dest_orig; |
13f2696f | 3522 | struct btrfs_ioctl_space_info __user *user_dest; |
1406e432 | 3523 | struct btrfs_space_info *info; |
315d8e98 CIK |
3524 | static const u64 types[] = { |
3525 | BTRFS_BLOCK_GROUP_DATA, | |
3526 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3527 | BTRFS_BLOCK_GROUP_METADATA, | |
3528 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA | |
3529 | }; | |
bf5fc093 | 3530 | int num_types = 4; |
7fde62bf | 3531 | int alloc_size; |
1406e432 | 3532 | int ret = 0; |
51788b1b | 3533 | u64 slot_count = 0; |
bf5fc093 | 3534 | int i, c; |
1406e432 JB |
3535 | |
3536 | if (copy_from_user(&space_args, | |
3537 | (struct btrfs_ioctl_space_args __user *)arg, | |
3538 | sizeof(space_args))) | |
3539 | return -EFAULT; | |
3540 | ||
bf5fc093 JB |
3541 | for (i = 0; i < num_types; i++) { |
3542 | struct btrfs_space_info *tmp; | |
3543 | ||
3544 | info = NULL; | |
72804905 | 3545 | list_for_each_entry(tmp, &fs_info->space_info, list) { |
bf5fc093 JB |
3546 | if (tmp->flags == types[i]) { |
3547 | info = tmp; | |
3548 | break; | |
3549 | } | |
3550 | } | |
bf5fc093 JB |
3551 | |
3552 | if (!info) | |
3553 | continue; | |
3554 | ||
3555 | down_read(&info->groups_sem); | |
3556 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3557 | if (!list_empty(&info->block_groups[c])) | |
3558 | slot_count++; | |
3559 | } | |
3560 | up_read(&info->groups_sem); | |
3561 | } | |
7fde62bf | 3562 | |
36523e95 DS |
3563 | /* |
3564 | * Global block reserve, exported as a space_info | |
3565 | */ | |
3566 | slot_count++; | |
3567 | ||
7fde62bf CM |
3568 | /* space_slots == 0 means they are asking for a count */ |
3569 | if (space_args.space_slots == 0) { | |
3570 | space_args.total_spaces = slot_count; | |
3571 | goto out; | |
3572 | } | |
bf5fc093 | 3573 | |
51788b1b | 3574 | slot_count = min_t(u64, space_args.space_slots, slot_count); |
bf5fc093 | 3575 | |
7fde62bf | 3576 | alloc_size = sizeof(*dest) * slot_count; |
bf5fc093 | 3577 | |
7fde62bf CM |
3578 | /* we generally have at most 6 or so space infos, one for each raid |
3579 | * level. So, a whole page should be more than enough for everyone | |
3580 | */ | |
09cbfeaf | 3581 | if (alloc_size > PAGE_SIZE) |
7fde62bf CM |
3582 | return -ENOMEM; |
3583 | ||
1406e432 | 3584 | space_args.total_spaces = 0; |
8d2db785 | 3585 | dest = kmalloc(alloc_size, GFP_KERNEL); |
7fde62bf CM |
3586 | if (!dest) |
3587 | return -ENOMEM; | |
3588 | dest_orig = dest; | |
1406e432 | 3589 | |
7fde62bf | 3590 | /* now we have a buffer to copy into */ |
bf5fc093 JB |
3591 | for (i = 0; i < num_types; i++) { |
3592 | struct btrfs_space_info *tmp; | |
3593 | ||
51788b1b DR |
3594 | if (!slot_count) |
3595 | break; | |
3596 | ||
bf5fc093 | 3597 | info = NULL; |
72804905 | 3598 | list_for_each_entry(tmp, &fs_info->space_info, list) { |
bf5fc093 JB |
3599 | if (tmp->flags == types[i]) { |
3600 | info = tmp; | |
3601 | break; | |
3602 | } | |
3603 | } | |
7fde62bf | 3604 | |
bf5fc093 JB |
3605 | if (!info) |
3606 | continue; | |
3607 | down_read(&info->groups_sem); | |
3608 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3609 | if (!list_empty(&info->block_groups[c])) { | |
c065f5b1 SY |
3610 | get_block_group_info(&info->block_groups[c], |
3611 | &space); | |
bf5fc093 JB |
3612 | memcpy(dest, &space, sizeof(space)); |
3613 | dest++; | |
3614 | space_args.total_spaces++; | |
51788b1b | 3615 | slot_count--; |
bf5fc093 | 3616 | } |
51788b1b DR |
3617 | if (!slot_count) |
3618 | break; | |
bf5fc093 JB |
3619 | } |
3620 | up_read(&info->groups_sem); | |
1406e432 | 3621 | } |
1406e432 | 3622 | |
36523e95 DS |
3623 | /* |
3624 | * Add global block reserve | |
3625 | */ | |
3626 | if (slot_count) { | |
0b246afa | 3627 | struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; |
36523e95 DS |
3628 | |
3629 | spin_lock(&block_rsv->lock); | |
3630 | space.total_bytes = block_rsv->size; | |
3631 | space.used_bytes = block_rsv->size - block_rsv->reserved; | |
3632 | spin_unlock(&block_rsv->lock); | |
3633 | space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; | |
3634 | memcpy(dest, &space, sizeof(space)); | |
3635 | space_args.total_spaces++; | |
3636 | } | |
3637 | ||
2eec6c81 | 3638 | user_dest = (struct btrfs_ioctl_space_info __user *) |
7fde62bf CM |
3639 | (arg + sizeof(struct btrfs_ioctl_space_args)); |
3640 | ||
3641 | if (copy_to_user(user_dest, dest_orig, alloc_size)) | |
3642 | ret = -EFAULT; | |
3643 | ||
3644 | kfree(dest_orig); | |
3645 | out: | |
3646 | if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) | |
1406e432 JB |
3647 | ret = -EFAULT; |
3648 | ||
3649 | return ret; | |
3650 | } | |
3651 | ||
9a8c28be MX |
3652 | static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, |
3653 | void __user *argp) | |
46204592 | 3654 | { |
46204592 SW |
3655 | struct btrfs_trans_handle *trans; |
3656 | u64 transid; | |
db5b493a | 3657 | int ret; |
46204592 | 3658 | |
d4edf39b | 3659 | trans = btrfs_attach_transaction_barrier(root); |
ff7c1d33 MX |
3660 | if (IS_ERR(trans)) { |
3661 | if (PTR_ERR(trans) != -ENOENT) | |
3662 | return PTR_ERR(trans); | |
3663 | ||
3664 | /* No running transaction, don't bother */ | |
3665 | transid = root->fs_info->last_trans_committed; | |
3666 | goto out; | |
3667 | } | |
46204592 | 3668 | transid = trans->transid; |
32cc4f87 | 3669 | ret = btrfs_commit_transaction_async(trans); |
8b2b2d3c | 3670 | if (ret) { |
3a45bb20 | 3671 | btrfs_end_transaction(trans); |
db5b493a | 3672 | return ret; |
8b2b2d3c | 3673 | } |
ff7c1d33 | 3674 | out: |
46204592 SW |
3675 | if (argp) |
3676 | if (copy_to_user(argp, &transid, sizeof(transid))) | |
3677 | return -EFAULT; | |
3678 | return 0; | |
3679 | } | |
3680 | ||
2ff7e61e | 3681 | static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, |
9a8c28be | 3682 | void __user *argp) |
46204592 | 3683 | { |
46204592 SW |
3684 | u64 transid; |
3685 | ||
3686 | if (argp) { | |
3687 | if (copy_from_user(&transid, argp, sizeof(transid))) | |
3688 | return -EFAULT; | |
3689 | } else { | |
3690 | transid = 0; /* current trans */ | |
3691 | } | |
2ff7e61e | 3692 | return btrfs_wait_for_commit(fs_info, transid); |
46204592 SW |
3693 | } |
3694 | ||
b8e95489 | 3695 | static long btrfs_ioctl_scrub(struct file *file, void __user *arg) |
475f6387 | 3696 | { |
0b246afa | 3697 | struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); |
475f6387 | 3698 | struct btrfs_ioctl_scrub_args *sa; |
b8e95489 | 3699 | int ret; |
475f6387 JS |
3700 | |
3701 | if (!capable(CAP_SYS_ADMIN)) | |
3702 | return -EPERM; | |
3703 | ||
3704 | sa = memdup_user(arg, sizeof(*sa)); | |
3705 | if (IS_ERR(sa)) | |
3706 | return PTR_ERR(sa); | |
3707 | ||
b8e95489 MX |
3708 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) { |
3709 | ret = mnt_want_write_file(file); | |
3710 | if (ret) | |
3711 | goto out; | |
3712 | } | |
3713 | ||
0b246afa | 3714 | ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, |
63a212ab SB |
3715 | &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, |
3716 | 0); | |
475f6387 | 3717 | |
5afe6ce7 FM |
3718 | /* |
3719 | * Copy scrub args to user space even if btrfs_scrub_dev() returned an | |
3720 | * error. This is important as it allows user space to know how much | |
3721 | * progress scrub has done. For example, if scrub is canceled we get | |
3722 | * -ECANCELED from btrfs_scrub_dev() and return that error back to user | |
3723 | * space. Later user space can inspect the progress from the structure | |
3724 | * btrfs_ioctl_scrub_args and resume scrub from where it left off | |
3725 | * previously (btrfs-progs does this). | |
3726 | * If we fail to copy the btrfs_ioctl_scrub_args structure to user space | |
3727 | * then return -EFAULT to signal the structure was not copied or it may | |
3728 | * be corrupt and unreliable due to a partial copy. | |
3729 | */ | |
3730 | if (copy_to_user(arg, sa, sizeof(*sa))) | |
475f6387 JS |
3731 | ret = -EFAULT; |
3732 | ||
b8e95489 MX |
3733 | if (!(sa->flags & BTRFS_SCRUB_READONLY)) |
3734 | mnt_drop_write_file(file); | |
3735 | out: | |
475f6387 JS |
3736 | kfree(sa); |
3737 | return ret; | |
3738 | } | |
3739 | ||
2ff7e61e | 3740 | static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) |
475f6387 JS |
3741 | { |
3742 | if (!capable(CAP_SYS_ADMIN)) | |
3743 | return -EPERM; | |
3744 | ||
2ff7e61e | 3745 | return btrfs_scrub_cancel(fs_info); |
475f6387 JS |
3746 | } |
3747 | ||
2ff7e61e | 3748 | static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, |
475f6387 JS |
3749 | void __user *arg) |
3750 | { | |
3751 | struct btrfs_ioctl_scrub_args *sa; | |
3752 | int ret; | |
3753 | ||
3754 | if (!capable(CAP_SYS_ADMIN)) | |
3755 | return -EPERM; | |
3756 | ||
3757 | sa = memdup_user(arg, sizeof(*sa)); | |
3758 | if (IS_ERR(sa)) | |
3759 | return PTR_ERR(sa); | |
3760 | ||
2ff7e61e | 3761 | ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); |
475f6387 | 3762 | |
4fa99b00 | 3763 | if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) |
475f6387 JS |
3764 | ret = -EFAULT; |
3765 | ||
3766 | kfree(sa); | |
3767 | return ret; | |
3768 | } | |
3769 | ||
2ff7e61e | 3770 | static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, |
b27f7c0c | 3771 | void __user *arg) |
c11d2c23 SB |
3772 | { |
3773 | struct btrfs_ioctl_get_dev_stats *sa; | |
3774 | int ret; | |
3775 | ||
c11d2c23 SB |
3776 | sa = memdup_user(arg, sizeof(*sa)); |
3777 | if (IS_ERR(sa)) | |
3778 | return PTR_ERR(sa); | |
3779 | ||
b27f7c0c DS |
3780 | if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { |
3781 | kfree(sa); | |
3782 | return -EPERM; | |
3783 | } | |
3784 | ||
2ff7e61e | 3785 | ret = btrfs_get_dev_stats(fs_info, sa); |
c11d2c23 | 3786 | |
eee99577 | 3787 | if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa))) |
c11d2c23 SB |
3788 | ret = -EFAULT; |
3789 | ||
3790 | kfree(sa); | |
3791 | return ret; | |
3792 | } | |
3793 | ||
2ff7e61e JM |
3794 | static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, |
3795 | void __user *arg) | |
3f6bcfbd SB |
3796 | { |
3797 | struct btrfs_ioctl_dev_replace_args *p; | |
3798 | int ret; | |
3799 | ||
3800 | if (!capable(CAP_SYS_ADMIN)) | |
3801 | return -EPERM; | |
3802 | ||
3803 | p = memdup_user(arg, sizeof(*p)); | |
3804 | if (IS_ERR(p)) | |
3805 | return PTR_ERR(p); | |
3806 | ||
3807 | switch (p->cmd) { | |
3808 | case BTRFS_IOCTL_DEV_REPLACE_CMD_START: | |
bc98a42c | 3809 | if (sb_rdonly(fs_info->sb)) { |
adfa97cb ID |
3810 | ret = -EROFS; |
3811 | goto out; | |
3812 | } | |
c3e1f96c | 3813 | if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) { |
e57138b3 | 3814 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
3f6bcfbd | 3815 | } else { |
2ff7e61e | 3816 | ret = btrfs_dev_replace_by_ioctl(fs_info, p); |
c3e1f96c | 3817 | btrfs_exclop_finish(fs_info); |
3f6bcfbd SB |
3818 | } |
3819 | break; | |
3820 | case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: | |
0b246afa | 3821 | btrfs_dev_replace_status(fs_info, p); |
3f6bcfbd SB |
3822 | ret = 0; |
3823 | break; | |
3824 | case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: | |
17d202b9 | 3825 | p->result = btrfs_dev_replace_cancel(fs_info); |
97282031 | 3826 | ret = 0; |
3f6bcfbd SB |
3827 | break; |
3828 | default: | |
3829 | ret = -EINVAL; | |
3830 | break; | |
3831 | } | |
3832 | ||
d3a53286 | 3833 | if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p))) |
3f6bcfbd | 3834 | ret = -EFAULT; |
adfa97cb | 3835 | out: |
3f6bcfbd SB |
3836 | kfree(p); |
3837 | return ret; | |
3838 | } | |
3839 | ||
d7728c96 JS |
3840 | static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) |
3841 | { | |
3842 | int ret = 0; | |
3843 | int i; | |
740c3d22 | 3844 | u64 rel_ptr; |
d7728c96 | 3845 | int size; |
806468f8 | 3846 | struct btrfs_ioctl_ino_path_args *ipa = NULL; |
d7728c96 JS |
3847 | struct inode_fs_paths *ipath = NULL; |
3848 | struct btrfs_path *path; | |
3849 | ||
82b22ac8 | 3850 | if (!capable(CAP_DAC_READ_SEARCH)) |
d7728c96 JS |
3851 | return -EPERM; |
3852 | ||
3853 | path = btrfs_alloc_path(); | |
3854 | if (!path) { | |
3855 | ret = -ENOMEM; | |
3856 | goto out; | |
3857 | } | |
3858 | ||
3859 | ipa = memdup_user(arg, sizeof(*ipa)); | |
3860 | if (IS_ERR(ipa)) { | |
3861 | ret = PTR_ERR(ipa); | |
3862 | ipa = NULL; | |
3863 | goto out; | |
3864 | } | |
3865 | ||
3866 | size = min_t(u32, ipa->size, 4096); | |
3867 | ipath = init_ipath(size, root, path); | |
3868 | if (IS_ERR(ipath)) { | |
3869 | ret = PTR_ERR(ipath); | |
3870 | ipath = NULL; | |
3871 | goto out; | |
3872 | } | |
3873 | ||
3874 | ret = paths_from_inode(ipa->inum, ipath); | |
3875 | if (ret < 0) | |
3876 | goto out; | |
3877 | ||
3878 | for (i = 0; i < ipath->fspath->elem_cnt; ++i) { | |
745c4d8e JM |
3879 | rel_ptr = ipath->fspath->val[i] - |
3880 | (u64)(unsigned long)ipath->fspath->val; | |
740c3d22 | 3881 | ipath->fspath->val[i] = rel_ptr; |
d7728c96 JS |
3882 | } |
3883 | ||
718dc5fa OS |
3884 | ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, |
3885 | ipath->fspath, size); | |
d7728c96 JS |
3886 | if (ret) { |
3887 | ret = -EFAULT; | |
3888 | goto out; | |
3889 | } | |
3890 | ||
3891 | out: | |
3892 | btrfs_free_path(path); | |
3893 | free_ipath(ipath); | |
3894 | kfree(ipa); | |
3895 | ||
3896 | return ret; | |
3897 | } | |
3898 | ||
3899 | static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) | |
3900 | { | |
3901 | struct btrfs_data_container *inodes = ctx; | |
3902 | const size_t c = 3 * sizeof(u64); | |
3903 | ||
3904 | if (inodes->bytes_left >= c) { | |
3905 | inodes->bytes_left -= c; | |
3906 | inodes->val[inodes->elem_cnt] = inum; | |
3907 | inodes->val[inodes->elem_cnt + 1] = offset; | |
3908 | inodes->val[inodes->elem_cnt + 2] = root; | |
3909 | inodes->elem_cnt += 3; | |
3910 | } else { | |
3911 | inodes->bytes_missing += c - inodes->bytes_left; | |
3912 | inodes->bytes_left = 0; | |
3913 | inodes->elem_missed += 3; | |
3914 | } | |
3915 | ||
3916 | return 0; | |
3917 | } | |
3918 | ||
2ff7e61e | 3919 | static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, |
d24a67b2 | 3920 | void __user *arg, int version) |
d7728c96 JS |
3921 | { |
3922 | int ret = 0; | |
3923 | int size; | |
d7728c96 JS |
3924 | struct btrfs_ioctl_logical_ino_args *loi; |
3925 | struct btrfs_data_container *inodes = NULL; | |
3926 | struct btrfs_path *path = NULL; | |
d24a67b2 | 3927 | bool ignore_offset; |
d7728c96 JS |
3928 | |
3929 | if (!capable(CAP_SYS_ADMIN)) | |
3930 | return -EPERM; | |
3931 | ||
3932 | loi = memdup_user(arg, sizeof(*loi)); | |
7b9ea627 SV |
3933 | if (IS_ERR(loi)) |
3934 | return PTR_ERR(loi); | |
d7728c96 | 3935 | |
d24a67b2 ZB |
3936 | if (version == 1) { |
3937 | ignore_offset = false; | |
b115e3bc | 3938 | size = min_t(u32, loi->size, SZ_64K); |
d24a67b2 ZB |
3939 | } else { |
3940 | /* All reserved bits must be 0 for now */ | |
3941 | if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) { | |
3942 | ret = -EINVAL; | |
3943 | goto out_loi; | |
3944 | } | |
3945 | /* Only accept flags we have defined so far */ | |
3946 | if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) { | |
3947 | ret = -EINVAL; | |
3948 | goto out_loi; | |
3949 | } | |
3950 | ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET; | |
b115e3bc | 3951 | size = min_t(u32, loi->size, SZ_16M); |
d24a67b2 ZB |
3952 | } |
3953 | ||
d7728c96 JS |
3954 | path = btrfs_alloc_path(); |
3955 | if (!path) { | |
3956 | ret = -ENOMEM; | |
3957 | goto out; | |
3958 | } | |
3959 | ||
d7728c96 JS |
3960 | inodes = init_data_container(size); |
3961 | if (IS_ERR(inodes)) { | |
3962 | ret = PTR_ERR(inodes); | |
3963 | inodes = NULL; | |
3964 | goto out; | |
3965 | } | |
3966 | ||
2ff7e61e | 3967 | ret = iterate_inodes_from_logical(loi->logical, fs_info, path, |
d24a67b2 | 3968 | build_ino_list, inodes, ignore_offset); |
df031f07 | 3969 | if (ret == -EINVAL) |
d7728c96 JS |
3970 | ret = -ENOENT; |
3971 | if (ret < 0) | |
3972 | goto out; | |
3973 | ||
718dc5fa OS |
3974 | ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, |
3975 | size); | |
d7728c96 JS |
3976 | if (ret) |
3977 | ret = -EFAULT; | |
3978 | ||
3979 | out: | |
3980 | btrfs_free_path(path); | |
f54de068 | 3981 | kvfree(inodes); |
d24a67b2 | 3982 | out_loi: |
d7728c96 JS |
3983 | kfree(loi); |
3984 | ||
3985 | return ret; | |
3986 | } | |
3987 | ||
008ef096 | 3988 | void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info, |
c9e9f97b ID |
3989 | struct btrfs_ioctl_balance_args *bargs) |
3990 | { | |
3991 | struct btrfs_balance_control *bctl = fs_info->balance_ctl; | |
3992 | ||
3993 | bargs->flags = bctl->flags; | |
3994 | ||
3009a62f | 3995 | if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) |
837d5b6e ID |
3996 | bargs->state |= BTRFS_BALANCE_STATE_RUNNING; |
3997 | if (atomic_read(&fs_info->balance_pause_req)) | |
3998 | bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; | |
a7e99c69 ID |
3999 | if (atomic_read(&fs_info->balance_cancel_req)) |
4000 | bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; | |
837d5b6e | 4001 | |
c9e9f97b ID |
4002 | memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); |
4003 | memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); | |
4004 | memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); | |
19a39dce | 4005 | |
008ef096 DS |
4006 | spin_lock(&fs_info->balance_lock); |
4007 | memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); | |
4008 | spin_unlock(&fs_info->balance_lock); | |
c9e9f97b ID |
4009 | } |
4010 | ||
9ba1f6e4 | 4011 | static long btrfs_ioctl_balance(struct file *file, void __user *arg) |
c9e9f97b | 4012 | { |
496ad9aa | 4013 | struct btrfs_root *root = BTRFS_I(file_inode(file))->root; |
c9e9f97b ID |
4014 | struct btrfs_fs_info *fs_info = root->fs_info; |
4015 | struct btrfs_ioctl_balance_args *bargs; | |
4016 | struct btrfs_balance_control *bctl; | |
ed0fb78f | 4017 | bool need_unlock; /* for mut. excl. ops lock */ |
c9e9f97b ID |
4018 | int ret; |
4019 | ||
4020 | if (!capable(CAP_SYS_ADMIN)) | |
4021 | return -EPERM; | |
4022 | ||
e54bfa31 | 4023 | ret = mnt_want_write_file(file); |
9ba1f6e4 LB |
4024 | if (ret) |
4025 | return ret; | |
4026 | ||
ed0fb78f | 4027 | again: |
c3e1f96c | 4028 | if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) { |
ed0fb78f ID |
4029 | mutex_lock(&fs_info->balance_mutex); |
4030 | need_unlock = true; | |
4031 | goto locked; | |
4032 | } | |
4033 | ||
4034 | /* | |
01327610 | 4035 | * mut. excl. ops lock is locked. Three possibilities: |
ed0fb78f ID |
4036 | * (1) some other op is running |
4037 | * (2) balance is running | |
4038 | * (3) balance is paused -- special case (think resume) | |
4039 | */ | |
c9e9f97b | 4040 | mutex_lock(&fs_info->balance_mutex); |
ed0fb78f ID |
4041 | if (fs_info->balance_ctl) { |
4042 | /* this is either (2) or (3) */ | |
3009a62f | 4043 | if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { |
ed0fb78f | 4044 | mutex_unlock(&fs_info->balance_mutex); |
dccdb07b DS |
4045 | /* |
4046 | * Lock released to allow other waiters to continue, | |
4047 | * we'll reexamine the status again. | |
4048 | */ | |
ed0fb78f ID |
4049 | mutex_lock(&fs_info->balance_mutex); |
4050 | ||
4051 | if (fs_info->balance_ctl && | |
3009a62f | 4052 | !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) { |
ed0fb78f ID |
4053 | /* this is (3) */ |
4054 | need_unlock = false; | |
4055 | goto locked; | |
4056 | } | |
4057 | ||
4058 | mutex_unlock(&fs_info->balance_mutex); | |
ed0fb78f ID |
4059 | goto again; |
4060 | } else { | |
4061 | /* this is (2) */ | |
4062 | mutex_unlock(&fs_info->balance_mutex); | |
4063 | ret = -EINPROGRESS; | |
4064 | goto out; | |
4065 | } | |
4066 | } else { | |
4067 | /* this is (1) */ | |
4068 | mutex_unlock(&fs_info->balance_mutex); | |
e57138b3 | 4069 | ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; |
ed0fb78f ID |
4070 | goto out; |
4071 | } | |
4072 | ||
4073 | locked: | |
c9e9f97b ID |
4074 | |
4075 | if (arg) { | |
4076 | bargs = memdup_user(arg, sizeof(*bargs)); | |
4077 | if (IS_ERR(bargs)) { | |
4078 | ret = PTR_ERR(bargs); | |
ed0fb78f | 4079 | goto out_unlock; |
c9e9f97b | 4080 | } |
de322263 ID |
4081 | |
4082 | if (bargs->flags & BTRFS_BALANCE_RESUME) { | |
4083 | if (!fs_info->balance_ctl) { | |
4084 | ret = -ENOTCONN; | |
4085 | goto out_bargs; | |
4086 | } | |
4087 | ||
4088 | bctl = fs_info->balance_ctl; | |
4089 | spin_lock(&fs_info->balance_lock); | |
4090 | bctl->flags |= BTRFS_BALANCE_RESUME; | |
4091 | spin_unlock(&fs_info->balance_lock); | |
4092 | ||
4093 | goto do_balance; | |
4094 | } | |
c9e9f97b ID |
4095 | } else { |
4096 | bargs = NULL; | |
4097 | } | |
4098 | ||
ed0fb78f | 4099 | if (fs_info->balance_ctl) { |
837d5b6e ID |
4100 | ret = -EINPROGRESS; |
4101 | goto out_bargs; | |
4102 | } | |
4103 | ||
8d2db785 | 4104 | bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); |
c9e9f97b ID |
4105 | if (!bctl) { |
4106 | ret = -ENOMEM; | |
4107 | goto out_bargs; | |
4108 | } | |
4109 | ||
c9e9f97b ID |
4110 | if (arg) { |
4111 | memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); | |
4112 | memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); | |
4113 | memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); | |
4114 | ||
4115 | bctl->flags = bargs->flags; | |
f43ffb60 ID |
4116 | } else { |
4117 | /* balance everything - no filters */ | |
4118 | bctl->flags |= BTRFS_BALANCE_TYPE_MASK; | |
c9e9f97b ID |
4119 | } |
4120 | ||
8eb93459 DS |
4121 | if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { |
4122 | ret = -EINVAL; | |
0f89abf5 | 4123 | goto out_bctl; |
8eb93459 DS |
4124 | } |
4125 | ||
de322263 | 4126 | do_balance: |
c9e9f97b | 4127 | /* |
c3e1f96c GR |
4128 | * Ownership of bctl and exclusive operation goes to btrfs_balance. |
4129 | * bctl is freed in reset_balance_state, or, if restriper was paused | |
4130 | * all the way until unmount, in free_fs_info. The flag should be | |
4131 | * cleared after reset_balance_state. | |
c9e9f97b | 4132 | */ |
ed0fb78f ID |
4133 | need_unlock = false; |
4134 | ||
6fcf6e2b | 4135 | ret = btrfs_balance(fs_info, bctl, bargs); |
0f89abf5 | 4136 | bctl = NULL; |
ed0fb78f | 4137 | |
d00c2d9c | 4138 | if ((ret == 0 || ret == -ECANCELED) && arg) { |
c9e9f97b ID |
4139 | if (copy_to_user(arg, bargs, sizeof(*bargs))) |
4140 | ret = -EFAULT; | |
4141 | } | |
4142 | ||
0f89abf5 CE |
4143 | out_bctl: |
4144 | kfree(bctl); | |
c9e9f97b ID |
4145 | out_bargs: |
4146 | kfree(bargs); | |
ed0fb78f | 4147 | out_unlock: |
c9e9f97b | 4148 | mutex_unlock(&fs_info->balance_mutex); |
ed0fb78f | 4149 | if (need_unlock) |
c3e1f96c | 4150 | btrfs_exclop_finish(fs_info); |
ed0fb78f | 4151 | out: |
e54bfa31 | 4152 | mnt_drop_write_file(file); |
c9e9f97b ID |
4153 | return ret; |
4154 | } | |
4155 | ||
2ff7e61e | 4156 | static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) |
837d5b6e ID |
4157 | { |
4158 | if (!capable(CAP_SYS_ADMIN)) | |
4159 | return -EPERM; | |
4160 | ||
4161 | switch (cmd) { | |
4162 | case BTRFS_BALANCE_CTL_PAUSE: | |
0b246afa | 4163 | return btrfs_pause_balance(fs_info); |
a7e99c69 | 4164 | case BTRFS_BALANCE_CTL_CANCEL: |
0b246afa | 4165 | return btrfs_cancel_balance(fs_info); |
837d5b6e ID |
4166 | } |
4167 | ||
4168 | return -EINVAL; | |
4169 | } | |
4170 | ||
2ff7e61e | 4171 | static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, |
19a39dce ID |
4172 | void __user *arg) |
4173 | { | |
19a39dce ID |
4174 | struct btrfs_ioctl_balance_args *bargs; |
4175 | int ret = 0; | |
4176 | ||
4177 | if (!capable(CAP_SYS_ADMIN)) | |
4178 | return -EPERM; | |
4179 | ||
4180 | mutex_lock(&fs_info->balance_mutex); | |
4181 | if (!fs_info->balance_ctl) { | |
4182 | ret = -ENOTCONN; | |
4183 | goto out; | |
4184 | } | |
4185 | ||
8d2db785 | 4186 | bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); |
19a39dce ID |
4187 | if (!bargs) { |
4188 | ret = -ENOMEM; | |
4189 | goto out; | |
4190 | } | |
4191 | ||
008ef096 | 4192 | btrfs_update_ioctl_balance_args(fs_info, bargs); |
19a39dce ID |
4193 | |
4194 | if (copy_to_user(arg, bargs, sizeof(*bargs))) | |
4195 | ret = -EFAULT; | |
4196 | ||
4197 | kfree(bargs); | |
4198 | out: | |
4199 | mutex_unlock(&fs_info->balance_mutex); | |
4200 | return ret; | |
4201 | } | |
4202 | ||
905b0dda | 4203 | static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) |
5d13a37b | 4204 | { |
0b246afa JM |
4205 | struct inode *inode = file_inode(file); |
4206 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
5d13a37b | 4207 | struct btrfs_ioctl_quota_ctl_args *sa; |
5d13a37b | 4208 | int ret; |
5d13a37b AJ |
4209 | |
4210 | if (!capable(CAP_SYS_ADMIN)) | |
4211 | return -EPERM; | |
4212 | ||
905b0dda MX |
4213 | ret = mnt_want_write_file(file); |
4214 | if (ret) | |
4215 | return ret; | |
5d13a37b AJ |
4216 | |
4217 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4218 | if (IS_ERR(sa)) { |
4219 | ret = PTR_ERR(sa); | |
4220 | goto drop_write; | |
4221 | } | |
5d13a37b | 4222 | |
0b246afa | 4223 | down_write(&fs_info->subvol_sem); |
5d13a37b AJ |
4224 | |
4225 | switch (sa->cmd) { | |
4226 | case BTRFS_QUOTA_CTL_ENABLE: | |
340f1aa2 | 4227 | ret = btrfs_quota_enable(fs_info); |
5d13a37b AJ |
4228 | break; |
4229 | case BTRFS_QUOTA_CTL_DISABLE: | |
340f1aa2 | 4230 | ret = btrfs_quota_disable(fs_info); |
5d13a37b | 4231 | break; |
5d13a37b AJ |
4232 | default: |
4233 | ret = -EINVAL; | |
4234 | break; | |
4235 | } | |
4236 | ||
5d13a37b | 4237 | kfree(sa); |
0b246afa | 4238 | up_write(&fs_info->subvol_sem); |
905b0dda MX |
4239 | drop_write: |
4240 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4241 | return ret; |
4242 | } | |
4243 | ||
905b0dda | 4244 | static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) |
5d13a37b | 4245 | { |
0b246afa JM |
4246 | struct inode *inode = file_inode(file); |
4247 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4248 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
5d13a37b AJ |
4249 | struct btrfs_ioctl_qgroup_assign_args *sa; |
4250 | struct btrfs_trans_handle *trans; | |
4251 | int ret; | |
4252 | int err; | |
4253 | ||
4254 | if (!capable(CAP_SYS_ADMIN)) | |
4255 | return -EPERM; | |
4256 | ||
905b0dda MX |
4257 | ret = mnt_want_write_file(file); |
4258 | if (ret) | |
4259 | return ret; | |
5d13a37b AJ |
4260 | |
4261 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4262 | if (IS_ERR(sa)) { |
4263 | ret = PTR_ERR(sa); | |
4264 | goto drop_write; | |
4265 | } | |
5d13a37b AJ |
4266 | |
4267 | trans = btrfs_join_transaction(root); | |
4268 | if (IS_ERR(trans)) { | |
4269 | ret = PTR_ERR(trans); | |
4270 | goto out; | |
4271 | } | |
4272 | ||
5d13a37b | 4273 | if (sa->assign) { |
9f8a6ce6 | 4274 | ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst); |
5d13a37b | 4275 | } else { |
39616c27 | 4276 | ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst); |
5d13a37b AJ |
4277 | } |
4278 | ||
e082f563 | 4279 | /* update qgroup status and info */ |
280f8bd2 | 4280 | err = btrfs_run_qgroups(trans); |
e082f563 | 4281 | if (err < 0) |
0b246afa JM |
4282 | btrfs_handle_fs_error(fs_info, err, |
4283 | "failed to update qgroup status and info"); | |
3a45bb20 | 4284 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4285 | if (err && !ret) |
4286 | ret = err; | |
4287 | ||
4288 | out: | |
4289 | kfree(sa); | |
905b0dda MX |
4290 | drop_write: |
4291 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4292 | return ret; |
4293 | } | |
4294 | ||
905b0dda | 4295 | static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) |
5d13a37b | 4296 | { |
0b246afa | 4297 | struct inode *inode = file_inode(file); |
0b246afa | 4298 | struct btrfs_root *root = BTRFS_I(inode)->root; |
5d13a37b AJ |
4299 | struct btrfs_ioctl_qgroup_create_args *sa; |
4300 | struct btrfs_trans_handle *trans; | |
4301 | int ret; | |
4302 | int err; | |
4303 | ||
4304 | if (!capable(CAP_SYS_ADMIN)) | |
4305 | return -EPERM; | |
4306 | ||
905b0dda MX |
4307 | ret = mnt_want_write_file(file); |
4308 | if (ret) | |
4309 | return ret; | |
5d13a37b AJ |
4310 | |
4311 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4312 | if (IS_ERR(sa)) { |
4313 | ret = PTR_ERR(sa); | |
4314 | goto drop_write; | |
4315 | } | |
5d13a37b | 4316 | |
d86e56cf MX |
4317 | if (!sa->qgroupid) { |
4318 | ret = -EINVAL; | |
4319 | goto out; | |
4320 | } | |
4321 | ||
5d13a37b AJ |
4322 | trans = btrfs_join_transaction(root); |
4323 | if (IS_ERR(trans)) { | |
4324 | ret = PTR_ERR(trans); | |
4325 | goto out; | |
4326 | } | |
4327 | ||
5d13a37b | 4328 | if (sa->create) { |
49a05ecd | 4329 | ret = btrfs_create_qgroup(trans, sa->qgroupid); |
5d13a37b | 4330 | } else { |
3efbee1d | 4331 | ret = btrfs_remove_qgroup(trans, sa->qgroupid); |
5d13a37b AJ |
4332 | } |
4333 | ||
3a45bb20 | 4334 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4335 | if (err && !ret) |
4336 | ret = err; | |
4337 | ||
4338 | out: | |
4339 | kfree(sa); | |
905b0dda MX |
4340 | drop_write: |
4341 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4342 | return ret; |
4343 | } | |
4344 | ||
905b0dda | 4345 | static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) |
5d13a37b | 4346 | { |
0b246afa | 4347 | struct inode *inode = file_inode(file); |
0b246afa | 4348 | struct btrfs_root *root = BTRFS_I(inode)->root; |
5d13a37b AJ |
4349 | struct btrfs_ioctl_qgroup_limit_args *sa; |
4350 | struct btrfs_trans_handle *trans; | |
4351 | int ret; | |
4352 | int err; | |
4353 | u64 qgroupid; | |
4354 | ||
4355 | if (!capable(CAP_SYS_ADMIN)) | |
4356 | return -EPERM; | |
4357 | ||
905b0dda MX |
4358 | ret = mnt_want_write_file(file); |
4359 | if (ret) | |
4360 | return ret; | |
5d13a37b AJ |
4361 | |
4362 | sa = memdup_user(arg, sizeof(*sa)); | |
905b0dda MX |
4363 | if (IS_ERR(sa)) { |
4364 | ret = PTR_ERR(sa); | |
4365 | goto drop_write; | |
4366 | } | |
5d13a37b AJ |
4367 | |
4368 | trans = btrfs_join_transaction(root); | |
4369 | if (IS_ERR(trans)) { | |
4370 | ret = PTR_ERR(trans); | |
4371 | goto out; | |
4372 | } | |
4373 | ||
4374 | qgroupid = sa->qgroupid; | |
4375 | if (!qgroupid) { | |
4376 | /* take the current subvol as qgroup */ | |
4377 | qgroupid = root->root_key.objectid; | |
4378 | } | |
4379 | ||
f0042d5e | 4380 | ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim); |
5d13a37b | 4381 | |
3a45bb20 | 4382 | err = btrfs_end_transaction(trans); |
5d13a37b AJ |
4383 | if (err && !ret) |
4384 | ret = err; | |
4385 | ||
4386 | out: | |
4387 | kfree(sa); | |
905b0dda MX |
4388 | drop_write: |
4389 | mnt_drop_write_file(file); | |
5d13a37b AJ |
4390 | return ret; |
4391 | } | |
4392 | ||
2f232036 JS |
4393 | static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) |
4394 | { | |
0b246afa JM |
4395 | struct inode *inode = file_inode(file); |
4396 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
2f232036 JS |
4397 | struct btrfs_ioctl_quota_rescan_args *qsa; |
4398 | int ret; | |
4399 | ||
4400 | if (!capable(CAP_SYS_ADMIN)) | |
4401 | return -EPERM; | |
4402 | ||
4403 | ret = mnt_want_write_file(file); | |
4404 | if (ret) | |
4405 | return ret; | |
4406 | ||
4407 | qsa = memdup_user(arg, sizeof(*qsa)); | |
4408 | if (IS_ERR(qsa)) { | |
4409 | ret = PTR_ERR(qsa); | |
4410 | goto drop_write; | |
4411 | } | |
4412 | ||
4413 | if (qsa->flags) { | |
4414 | ret = -EINVAL; | |
4415 | goto out; | |
4416 | } | |
4417 | ||
0b246afa | 4418 | ret = btrfs_qgroup_rescan(fs_info); |
2f232036 JS |
4419 | |
4420 | out: | |
4421 | kfree(qsa); | |
4422 | drop_write: | |
4423 | mnt_drop_write_file(file); | |
4424 | return ret; | |
4425 | } | |
4426 | ||
b929c1d8 MPS |
4427 | static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info, |
4428 | void __user *arg) | |
2f232036 | 4429 | { |
0afb603a | 4430 | struct btrfs_ioctl_quota_rescan_args qsa = {0}; |
2f232036 JS |
4431 | int ret = 0; |
4432 | ||
4433 | if (!capable(CAP_SYS_ADMIN)) | |
4434 | return -EPERM; | |
4435 | ||
0b246afa | 4436 | if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { |
0afb603a GR |
4437 | qsa.flags = 1; |
4438 | qsa.progress = fs_info->qgroup_rescan_progress.objectid; | |
2f232036 JS |
4439 | } |
4440 | ||
0afb603a | 4441 | if (copy_to_user(arg, &qsa, sizeof(qsa))) |
2f232036 JS |
4442 | ret = -EFAULT; |
4443 | ||
2f232036 JS |
4444 | return ret; |
4445 | } | |
4446 | ||
b929c1d8 MPS |
4447 | static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info, |
4448 | void __user *arg) | |
57254b6e | 4449 | { |
57254b6e JS |
4450 | if (!capable(CAP_SYS_ADMIN)) |
4451 | return -EPERM; | |
4452 | ||
0b246afa | 4453 | return btrfs_qgroup_wait_for_completion(fs_info, true); |
57254b6e JS |
4454 | } |
4455 | ||
abccd00f | 4456 | static long _btrfs_ioctl_set_received_subvol(struct file *file, |
e4fed17a | 4457 | struct user_namespace *mnt_userns, |
abccd00f | 4458 | struct btrfs_ioctl_received_subvol_args *sa) |
8ea05e3a | 4459 | { |
496ad9aa | 4460 | struct inode *inode = file_inode(file); |
0b246afa | 4461 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
8ea05e3a AB |
4462 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4463 | struct btrfs_root_item *root_item = &root->root_item; | |
4464 | struct btrfs_trans_handle *trans; | |
95582b00 | 4465 | struct timespec64 ct = current_time(inode); |
8ea05e3a | 4466 | int ret = 0; |
dd5f9615 | 4467 | int received_uuid_changed; |
8ea05e3a | 4468 | |
e4fed17a | 4469 | if (!inode_owner_or_capable(mnt_userns, inode)) |
bd60ea0f DS |
4470 | return -EPERM; |
4471 | ||
8ea05e3a AB |
4472 | ret = mnt_want_write_file(file); |
4473 | if (ret < 0) | |
4474 | return ret; | |
4475 | ||
0b246afa | 4476 | down_write(&fs_info->subvol_sem); |
8ea05e3a | 4477 | |
4a0cc7ca | 4478 | if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { |
8ea05e3a AB |
4479 | ret = -EINVAL; |
4480 | goto out; | |
4481 | } | |
4482 | ||
4483 | if (btrfs_root_readonly(root)) { | |
4484 | ret = -EROFS; | |
4485 | goto out; | |
4486 | } | |
4487 | ||
dd5f9615 SB |
4488 | /* |
4489 | * 1 - root item | |
4490 | * 2 - uuid items (received uuid + subvol uuid) | |
4491 | */ | |
4492 | trans = btrfs_start_transaction(root, 3); | |
8ea05e3a AB |
4493 | if (IS_ERR(trans)) { |
4494 | ret = PTR_ERR(trans); | |
4495 | trans = NULL; | |
4496 | goto out; | |
4497 | } | |
4498 | ||
4499 | sa->rtransid = trans->transid; | |
4500 | sa->rtime.sec = ct.tv_sec; | |
4501 | sa->rtime.nsec = ct.tv_nsec; | |
4502 | ||
dd5f9615 SB |
4503 | received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, |
4504 | BTRFS_UUID_SIZE); | |
4505 | if (received_uuid_changed && | |
d87ff758 | 4506 | !btrfs_is_empty_uuid(root_item->received_uuid)) { |
d1957791 | 4507 | ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid, |
d87ff758 NB |
4508 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, |
4509 | root->root_key.objectid); | |
4510 | if (ret && ret != -ENOENT) { | |
4511 | btrfs_abort_transaction(trans, ret); | |
4512 | btrfs_end_transaction(trans); | |
4513 | goto out; | |
4514 | } | |
4515 | } | |
8ea05e3a AB |
4516 | memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); |
4517 | btrfs_set_root_stransid(root_item, sa->stransid); | |
4518 | btrfs_set_root_rtransid(root_item, sa->rtransid); | |
3cae210f QW |
4519 | btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); |
4520 | btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); | |
4521 | btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); | |
4522 | btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); | |
8ea05e3a | 4523 | |
0b246afa | 4524 | ret = btrfs_update_root(trans, fs_info->tree_root, |
8ea05e3a AB |
4525 | &root->root_key, &root->root_item); |
4526 | if (ret < 0) { | |
3a45bb20 | 4527 | btrfs_end_transaction(trans); |
8ea05e3a | 4528 | goto out; |
dd5f9615 SB |
4529 | } |
4530 | if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { | |
cdb345a8 | 4531 | ret = btrfs_uuid_tree_add(trans, sa->uuid, |
dd5f9615 SB |
4532 | BTRFS_UUID_KEY_RECEIVED_SUBVOL, |
4533 | root->root_key.objectid); | |
4534 | if (ret < 0 && ret != -EEXIST) { | |
66642832 | 4535 | btrfs_abort_transaction(trans, ret); |
efd38150 | 4536 | btrfs_end_transaction(trans); |
8ea05e3a | 4537 | goto out; |
dd5f9615 SB |
4538 | } |
4539 | } | |
3a45bb20 | 4540 | ret = btrfs_commit_transaction(trans); |
abccd00f | 4541 | out: |
0b246afa | 4542 | up_write(&fs_info->subvol_sem); |
abccd00f HM |
4543 | mnt_drop_write_file(file); |
4544 | return ret; | |
4545 | } | |
4546 | ||
4547 | #ifdef CONFIG_64BIT | |
4548 | static long btrfs_ioctl_set_received_subvol_32(struct file *file, | |
4549 | void __user *arg) | |
4550 | { | |
4551 | struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; | |
4552 | struct btrfs_ioctl_received_subvol_args *args64 = NULL; | |
4553 | int ret = 0; | |
4554 | ||
4555 | args32 = memdup_user(arg, sizeof(*args32)); | |
7b9ea627 SV |
4556 | if (IS_ERR(args32)) |
4557 | return PTR_ERR(args32); | |
abccd00f | 4558 | |
8d2db785 | 4559 | args64 = kmalloc(sizeof(*args64), GFP_KERNEL); |
84dbeb87 DC |
4560 | if (!args64) { |
4561 | ret = -ENOMEM; | |
abccd00f HM |
4562 | goto out; |
4563 | } | |
4564 | ||
4565 | memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); | |
4566 | args64->stransid = args32->stransid; | |
4567 | args64->rtransid = args32->rtransid; | |
4568 | args64->stime.sec = args32->stime.sec; | |
4569 | args64->stime.nsec = args32->stime.nsec; | |
4570 | args64->rtime.sec = args32->rtime.sec; | |
4571 | args64->rtime.nsec = args32->rtime.nsec; | |
4572 | args64->flags = args32->flags; | |
4573 | ||
e4fed17a | 4574 | ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64); |
abccd00f HM |
4575 | if (ret) |
4576 | goto out; | |
4577 | ||
4578 | memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); | |
4579 | args32->stransid = args64->stransid; | |
4580 | args32->rtransid = args64->rtransid; | |
4581 | args32->stime.sec = args64->stime.sec; | |
4582 | args32->stime.nsec = args64->stime.nsec; | |
4583 | args32->rtime.sec = args64->rtime.sec; | |
4584 | args32->rtime.nsec = args64->rtime.nsec; | |
4585 | args32->flags = args64->flags; | |
4586 | ||
4587 | ret = copy_to_user(arg, args32, sizeof(*args32)); | |
4588 | if (ret) | |
4589 | ret = -EFAULT; | |
4590 | ||
4591 | out: | |
4592 | kfree(args32); | |
4593 | kfree(args64); | |
4594 | return ret; | |
4595 | } | |
4596 | #endif | |
4597 | ||
4598 | static long btrfs_ioctl_set_received_subvol(struct file *file, | |
4599 | void __user *arg) | |
4600 | { | |
4601 | struct btrfs_ioctl_received_subvol_args *sa = NULL; | |
4602 | int ret = 0; | |
4603 | ||
4604 | sa = memdup_user(arg, sizeof(*sa)); | |
7b9ea627 SV |
4605 | if (IS_ERR(sa)) |
4606 | return PTR_ERR(sa); | |
abccd00f | 4607 | |
e4fed17a | 4608 | ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa); |
abccd00f HM |
4609 | |
4610 | if (ret) | |
4611 | goto out; | |
4612 | ||
8ea05e3a AB |
4613 | ret = copy_to_user(arg, sa, sizeof(*sa)); |
4614 | if (ret) | |
4615 | ret = -EFAULT; | |
4616 | ||
4617 | out: | |
4618 | kfree(sa); | |
8ea05e3a AB |
4619 | return ret; |
4620 | } | |
4621 | ||
b929c1d8 MPS |
4622 | static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info, |
4623 | void __user *arg) | |
867ab667 | 4624 | { |
a1b83ac5 | 4625 | size_t len; |
867ab667 | 4626 | int ret; |
a1b83ac5 AJ |
4627 | char label[BTRFS_LABEL_SIZE]; |
4628 | ||
0b246afa JM |
4629 | spin_lock(&fs_info->super_lock); |
4630 | memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); | |
4631 | spin_unlock(&fs_info->super_lock); | |
a1b83ac5 AJ |
4632 | |
4633 | len = strnlen(label, BTRFS_LABEL_SIZE); | |
867ab667 | 4634 | |
4635 | if (len == BTRFS_LABEL_SIZE) { | |
0b246afa JM |
4636 | btrfs_warn(fs_info, |
4637 | "label is too long, return the first %zu bytes", | |
4638 | --len); | |
867ab667 | 4639 | } |
4640 | ||
867ab667 | 4641 | ret = copy_to_user(arg, label, len); |
867ab667 | 4642 | |
4643 | return ret ? -EFAULT : 0; | |
4644 | } | |
4645 | ||
a8bfd4ab | 4646 | static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) |
4647 | { | |
0b246afa JM |
4648 | struct inode *inode = file_inode(file); |
4649 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4650 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4651 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
a8bfd4ab | 4652 | struct btrfs_trans_handle *trans; |
4653 | char label[BTRFS_LABEL_SIZE]; | |
4654 | int ret; | |
4655 | ||
4656 | if (!capable(CAP_SYS_ADMIN)) | |
4657 | return -EPERM; | |
4658 | ||
4659 | if (copy_from_user(label, arg, sizeof(label))) | |
4660 | return -EFAULT; | |
4661 | ||
4662 | if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { | |
0b246afa | 4663 | btrfs_err(fs_info, |
5d163e0e JM |
4664 | "unable to set label with more than %d bytes", |
4665 | BTRFS_LABEL_SIZE - 1); | |
a8bfd4ab | 4666 | return -EINVAL; |
4667 | } | |
4668 | ||
4669 | ret = mnt_want_write_file(file); | |
4670 | if (ret) | |
4671 | return ret; | |
4672 | ||
a8bfd4ab | 4673 | trans = btrfs_start_transaction(root, 0); |
4674 | if (IS_ERR(trans)) { | |
4675 | ret = PTR_ERR(trans); | |
4676 | goto out_unlock; | |
4677 | } | |
4678 | ||
0b246afa | 4679 | spin_lock(&fs_info->super_lock); |
a8bfd4ab | 4680 | strcpy(super_block->label, label); |
0b246afa | 4681 | spin_unlock(&fs_info->super_lock); |
3a45bb20 | 4682 | ret = btrfs_commit_transaction(trans); |
a8bfd4ab | 4683 | |
4684 | out_unlock: | |
a8bfd4ab | 4685 | mnt_drop_write_file(file); |
4686 | return ret; | |
4687 | } | |
4688 | ||
2eaa055f JM |
4689 | #define INIT_FEATURE_FLAGS(suffix) \ |
4690 | { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ | |
4691 | .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ | |
4692 | .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } | |
4693 | ||
d5131b65 | 4694 | int btrfs_ioctl_get_supported_features(void __user *arg) |
2eaa055f | 4695 | { |
4d4ab6d6 | 4696 | static const struct btrfs_ioctl_feature_flags features[3] = { |
2eaa055f JM |
4697 | INIT_FEATURE_FLAGS(SUPP), |
4698 | INIT_FEATURE_FLAGS(SAFE_SET), | |
4699 | INIT_FEATURE_FLAGS(SAFE_CLEAR) | |
4700 | }; | |
4701 | ||
4702 | if (copy_to_user(arg, &features, sizeof(features))) | |
4703 | return -EFAULT; | |
4704 | ||
4705 | return 0; | |
4706 | } | |
4707 | ||
b929c1d8 MPS |
4708 | static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info, |
4709 | void __user *arg) | |
2eaa055f | 4710 | { |
0b246afa | 4711 | struct btrfs_super_block *super_block = fs_info->super_copy; |
2eaa055f JM |
4712 | struct btrfs_ioctl_feature_flags features; |
4713 | ||
4714 | features.compat_flags = btrfs_super_compat_flags(super_block); | |
4715 | features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); | |
4716 | features.incompat_flags = btrfs_super_incompat_flags(super_block); | |
4717 | ||
4718 | if (copy_to_user(arg, &features, sizeof(features))) | |
4719 | return -EFAULT; | |
4720 | ||
4721 | return 0; | |
4722 | } | |
4723 | ||
2ff7e61e | 4724 | static int check_feature_bits(struct btrfs_fs_info *fs_info, |
3b02a68a | 4725 | enum btrfs_feature_set set, |
2eaa055f JM |
4726 | u64 change_mask, u64 flags, u64 supported_flags, |
4727 | u64 safe_set, u64 safe_clear) | |
4728 | { | |
f10152bc | 4729 | const char *type = btrfs_feature_set_name(set); |
3b02a68a | 4730 | char *names; |
2eaa055f JM |
4731 | u64 disallowed, unsupported; |
4732 | u64 set_mask = flags & change_mask; | |
4733 | u64 clear_mask = ~flags & change_mask; | |
4734 | ||
4735 | unsupported = set_mask & ~supported_flags; | |
4736 | if (unsupported) { | |
3b02a68a JM |
4737 | names = btrfs_printable_features(set, unsupported); |
4738 | if (names) { | |
0b246afa JM |
4739 | btrfs_warn(fs_info, |
4740 | "this kernel does not support the %s feature bit%s", | |
4741 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
4742 | kfree(names); |
4743 | } else | |
0b246afa JM |
4744 | btrfs_warn(fs_info, |
4745 | "this kernel does not support %s bits 0x%llx", | |
4746 | type, unsupported); | |
2eaa055f JM |
4747 | return -EOPNOTSUPP; |
4748 | } | |
4749 | ||
4750 | disallowed = set_mask & ~safe_set; | |
4751 | if (disallowed) { | |
3b02a68a JM |
4752 | names = btrfs_printable_features(set, disallowed); |
4753 | if (names) { | |
0b246afa JM |
4754 | btrfs_warn(fs_info, |
4755 | "can't set the %s feature bit%s while mounted", | |
4756 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
4757 | kfree(names); |
4758 | } else | |
0b246afa JM |
4759 | btrfs_warn(fs_info, |
4760 | "can't set %s bits 0x%llx while mounted", | |
4761 | type, disallowed); | |
2eaa055f JM |
4762 | return -EPERM; |
4763 | } | |
4764 | ||
4765 | disallowed = clear_mask & ~safe_clear; | |
4766 | if (disallowed) { | |
3b02a68a JM |
4767 | names = btrfs_printable_features(set, disallowed); |
4768 | if (names) { | |
0b246afa JM |
4769 | btrfs_warn(fs_info, |
4770 | "can't clear the %s feature bit%s while mounted", | |
4771 | names, strchr(names, ',') ? "s" : ""); | |
3b02a68a JM |
4772 | kfree(names); |
4773 | } else | |
0b246afa JM |
4774 | btrfs_warn(fs_info, |
4775 | "can't clear %s bits 0x%llx while mounted", | |
4776 | type, disallowed); | |
2eaa055f JM |
4777 | return -EPERM; |
4778 | } | |
4779 | ||
4780 | return 0; | |
4781 | } | |
4782 | ||
2ff7e61e JM |
4783 | #define check_feature(fs_info, change_mask, flags, mask_base) \ |
4784 | check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ | |
2eaa055f JM |
4785 | BTRFS_FEATURE_ ## mask_base ## _SUPP, \ |
4786 | BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ | |
4787 | BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) | |
4788 | ||
4789 | static int btrfs_ioctl_set_features(struct file *file, void __user *arg) | |
4790 | { | |
0b246afa JM |
4791 | struct inode *inode = file_inode(file); |
4792 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4793 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4794 | struct btrfs_super_block *super_block = fs_info->super_copy; | |
2eaa055f JM |
4795 | struct btrfs_ioctl_feature_flags flags[2]; |
4796 | struct btrfs_trans_handle *trans; | |
4797 | u64 newflags; | |
4798 | int ret; | |
4799 | ||
4800 | if (!capable(CAP_SYS_ADMIN)) | |
4801 | return -EPERM; | |
4802 | ||
4803 | if (copy_from_user(flags, arg, sizeof(flags))) | |
4804 | return -EFAULT; | |
4805 | ||
4806 | /* Nothing to do */ | |
4807 | if (!flags[0].compat_flags && !flags[0].compat_ro_flags && | |
4808 | !flags[0].incompat_flags) | |
4809 | return 0; | |
4810 | ||
2ff7e61e | 4811 | ret = check_feature(fs_info, flags[0].compat_flags, |
2eaa055f JM |
4812 | flags[1].compat_flags, COMPAT); |
4813 | if (ret) | |
4814 | return ret; | |
4815 | ||
2ff7e61e | 4816 | ret = check_feature(fs_info, flags[0].compat_ro_flags, |
2eaa055f JM |
4817 | flags[1].compat_ro_flags, COMPAT_RO); |
4818 | if (ret) | |
4819 | return ret; | |
4820 | ||
2ff7e61e | 4821 | ret = check_feature(fs_info, flags[0].incompat_flags, |
2eaa055f JM |
4822 | flags[1].incompat_flags, INCOMPAT); |
4823 | if (ret) | |
4824 | return ret; | |
4825 | ||
7ab19625 DS |
4826 | ret = mnt_want_write_file(file); |
4827 | if (ret) | |
4828 | return ret; | |
4829 | ||
8051aa1a | 4830 | trans = btrfs_start_transaction(root, 0); |
7ab19625 DS |
4831 | if (IS_ERR(trans)) { |
4832 | ret = PTR_ERR(trans); | |
4833 | goto out_drop_write; | |
4834 | } | |
2eaa055f | 4835 | |
0b246afa | 4836 | spin_lock(&fs_info->super_lock); |
2eaa055f JM |
4837 | newflags = btrfs_super_compat_flags(super_block); |
4838 | newflags |= flags[0].compat_flags & flags[1].compat_flags; | |
4839 | newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); | |
4840 | btrfs_set_super_compat_flags(super_block, newflags); | |
4841 | ||
4842 | newflags = btrfs_super_compat_ro_flags(super_block); | |
4843 | newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; | |
4844 | newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); | |
4845 | btrfs_set_super_compat_ro_flags(super_block, newflags); | |
4846 | ||
4847 | newflags = btrfs_super_incompat_flags(super_block); | |
4848 | newflags |= flags[0].incompat_flags & flags[1].incompat_flags; | |
4849 | newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); | |
4850 | btrfs_set_super_incompat_flags(super_block, newflags); | |
0b246afa | 4851 | spin_unlock(&fs_info->super_lock); |
2eaa055f | 4852 | |
3a45bb20 | 4853 | ret = btrfs_commit_transaction(trans); |
7ab19625 DS |
4854 | out_drop_write: |
4855 | mnt_drop_write_file(file); | |
4856 | ||
4857 | return ret; | |
2eaa055f JM |
4858 | } |
4859 | ||
2351f431 JB |
4860 | static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat) |
4861 | { | |
4862 | struct btrfs_ioctl_send_args *arg; | |
4863 | int ret; | |
4864 | ||
4865 | if (compat) { | |
4866 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) | |
4867 | struct btrfs_ioctl_send_args_32 args32; | |
4868 | ||
4869 | ret = copy_from_user(&args32, argp, sizeof(args32)); | |
4870 | if (ret) | |
4871 | return -EFAULT; | |
4872 | arg = kzalloc(sizeof(*arg), GFP_KERNEL); | |
4873 | if (!arg) | |
4874 | return -ENOMEM; | |
4875 | arg->send_fd = args32.send_fd; | |
4876 | arg->clone_sources_count = args32.clone_sources_count; | |
4877 | arg->clone_sources = compat_ptr(args32.clone_sources); | |
4878 | arg->parent_root = args32.parent_root; | |
4879 | arg->flags = args32.flags; | |
4880 | memcpy(arg->reserved, args32.reserved, | |
4881 | sizeof(args32.reserved)); | |
4882 | #else | |
4883 | return -ENOTTY; | |
4884 | #endif | |
4885 | } else { | |
4886 | arg = memdup_user(argp, sizeof(*arg)); | |
4887 | if (IS_ERR(arg)) | |
4888 | return PTR_ERR(arg); | |
4889 | } | |
4890 | ret = btrfs_ioctl_send(file, arg); | |
4891 | kfree(arg); | |
4892 | return ret; | |
4893 | } | |
4894 | ||
f46b5a66 CH |
4895 | long btrfs_ioctl(struct file *file, unsigned int |
4896 | cmd, unsigned long arg) | |
4897 | { | |
0b246afa JM |
4898 | struct inode *inode = file_inode(file); |
4899 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); | |
4900 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
4bcabaa3 | 4901 | void __user *argp = (void __user *)arg; |
f46b5a66 CH |
4902 | |
4903 | switch (cmd) { | |
6cbff00f CH |
4904 | case FS_IOC_GETVERSION: |
4905 | return btrfs_ioctl_getversion(file, argp); | |
40cf931f | 4906 | case FS_IOC_GETFSLABEL: |
b929c1d8 | 4907 | return btrfs_ioctl_get_fslabel(fs_info, argp); |
40cf931f ES |
4908 | case FS_IOC_SETFSLABEL: |
4909 | return btrfs_ioctl_set_fslabel(file, argp); | |
f7039b1d | 4910 | case FITRIM: |
b929c1d8 | 4911 | return btrfs_ioctl_fitrim(fs_info, argp); |
f46b5a66 | 4912 | case BTRFS_IOC_SNAP_CREATE: |
fa0d2b9b | 4913 | return btrfs_ioctl_snap_create(file, argp, 0); |
fdfb1e4f | 4914 | case BTRFS_IOC_SNAP_CREATE_V2: |
fa0d2b9b | 4915 | return btrfs_ioctl_snap_create_v2(file, argp, 0); |
3de4586c | 4916 | case BTRFS_IOC_SUBVOL_CREATE: |
fa0d2b9b | 4917 | return btrfs_ioctl_snap_create(file, argp, 1); |
6f72c7e2 AJ |
4918 | case BTRFS_IOC_SUBVOL_CREATE_V2: |
4919 | return btrfs_ioctl_snap_create_v2(file, argp, 1); | |
76dda93c | 4920 | case BTRFS_IOC_SNAP_DESTROY: |
949964c9 MPS |
4921 | return btrfs_ioctl_snap_destroy(file, argp, false); |
4922 | case BTRFS_IOC_SNAP_DESTROY_V2: | |
4923 | return btrfs_ioctl_snap_destroy(file, argp, true); | |
0caa102d LZ |
4924 | case BTRFS_IOC_SUBVOL_GETFLAGS: |
4925 | return btrfs_ioctl_subvol_getflags(file, argp); | |
4926 | case BTRFS_IOC_SUBVOL_SETFLAGS: | |
4927 | return btrfs_ioctl_subvol_setflags(file, argp); | |
6ef5ed0d JB |
4928 | case BTRFS_IOC_DEFAULT_SUBVOL: |
4929 | return btrfs_ioctl_default_subvol(file, argp); | |
f46b5a66 | 4930 | case BTRFS_IOC_DEFRAG: |
1e701a32 CM |
4931 | return btrfs_ioctl_defrag(file, NULL); |
4932 | case BTRFS_IOC_DEFRAG_RANGE: | |
4933 | return btrfs_ioctl_defrag(file, argp); | |
f46b5a66 | 4934 | case BTRFS_IOC_RESIZE: |
198605a8 | 4935 | return btrfs_ioctl_resize(file, argp); |
f46b5a66 | 4936 | case BTRFS_IOC_ADD_DEV: |
2ff7e61e | 4937 | return btrfs_ioctl_add_dev(fs_info, argp); |
f46b5a66 | 4938 | case BTRFS_IOC_RM_DEV: |
da24927b | 4939 | return btrfs_ioctl_rm_dev(file, argp); |
6b526ed7 AJ |
4940 | case BTRFS_IOC_RM_DEV_V2: |
4941 | return btrfs_ioctl_rm_dev_v2(file, argp); | |
475f6387 | 4942 | case BTRFS_IOC_FS_INFO: |
2ff7e61e | 4943 | return btrfs_ioctl_fs_info(fs_info, argp); |
475f6387 | 4944 | case BTRFS_IOC_DEV_INFO: |
2ff7e61e | 4945 | return btrfs_ioctl_dev_info(fs_info, argp); |
f46b5a66 | 4946 | case BTRFS_IOC_BALANCE: |
9ba1f6e4 | 4947 | return btrfs_ioctl_balance(file, NULL); |
ac8e9819 CM |
4948 | case BTRFS_IOC_TREE_SEARCH: |
4949 | return btrfs_ioctl_tree_search(file, argp); | |
cc68a8a5 GH |
4950 | case BTRFS_IOC_TREE_SEARCH_V2: |
4951 | return btrfs_ioctl_tree_search_v2(file, argp); | |
ac8e9819 CM |
4952 | case BTRFS_IOC_INO_LOOKUP: |
4953 | return btrfs_ioctl_ino_lookup(file, argp); | |
d7728c96 JS |
4954 | case BTRFS_IOC_INO_PATHS: |
4955 | return btrfs_ioctl_ino_to_path(root, argp); | |
4956 | case BTRFS_IOC_LOGICAL_INO: | |
d24a67b2 ZB |
4957 | return btrfs_ioctl_logical_to_ino(fs_info, argp, 1); |
4958 | case BTRFS_IOC_LOGICAL_INO_V2: | |
4959 | return btrfs_ioctl_logical_to_ino(fs_info, argp, 2); | |
1406e432 | 4960 | case BTRFS_IOC_SPACE_INFO: |
2ff7e61e | 4961 | return btrfs_ioctl_space_info(fs_info, argp); |
9b199859 FDBM |
4962 | case BTRFS_IOC_SYNC: { |
4963 | int ret; | |
4964 | ||
9db4dc24 | 4965 | ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false); |
9b199859 FDBM |
4966 | if (ret) |
4967 | return ret; | |
0b246afa | 4968 | ret = btrfs_sync_fs(inode->i_sb, 1); |
2fad4e83 DS |
4969 | /* |
4970 | * The transaction thread may want to do more work, | |
01327610 | 4971 | * namely it pokes the cleaner kthread that will start |
2fad4e83 DS |
4972 | * processing uncleaned subvols. |
4973 | */ | |
0b246afa | 4974 | wake_up_process(fs_info->transaction_kthread); |
9b199859 FDBM |
4975 | return ret; |
4976 | } | |
46204592 | 4977 | case BTRFS_IOC_START_SYNC: |
9a8c28be | 4978 | return btrfs_ioctl_start_sync(root, argp); |
46204592 | 4979 | case BTRFS_IOC_WAIT_SYNC: |
2ff7e61e | 4980 | return btrfs_ioctl_wait_sync(fs_info, argp); |
475f6387 | 4981 | case BTRFS_IOC_SCRUB: |
b8e95489 | 4982 | return btrfs_ioctl_scrub(file, argp); |
475f6387 | 4983 | case BTRFS_IOC_SCRUB_CANCEL: |
2ff7e61e | 4984 | return btrfs_ioctl_scrub_cancel(fs_info); |
475f6387 | 4985 | case BTRFS_IOC_SCRUB_PROGRESS: |
2ff7e61e | 4986 | return btrfs_ioctl_scrub_progress(fs_info, argp); |
c9e9f97b | 4987 | case BTRFS_IOC_BALANCE_V2: |
9ba1f6e4 | 4988 | return btrfs_ioctl_balance(file, argp); |
837d5b6e | 4989 | case BTRFS_IOC_BALANCE_CTL: |
2ff7e61e | 4990 | return btrfs_ioctl_balance_ctl(fs_info, arg); |
19a39dce | 4991 | case BTRFS_IOC_BALANCE_PROGRESS: |
2ff7e61e | 4992 | return btrfs_ioctl_balance_progress(fs_info, argp); |
8ea05e3a AB |
4993 | case BTRFS_IOC_SET_RECEIVED_SUBVOL: |
4994 | return btrfs_ioctl_set_received_subvol(file, argp); | |
abccd00f HM |
4995 | #ifdef CONFIG_64BIT |
4996 | case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: | |
4997 | return btrfs_ioctl_set_received_subvol_32(file, argp); | |
4998 | #endif | |
31db9f7c | 4999 | case BTRFS_IOC_SEND: |
2351f431 JB |
5000 | return _btrfs_ioctl_send(file, argp, false); |
5001 | #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT) | |
5002 | case BTRFS_IOC_SEND_32: | |
5003 | return _btrfs_ioctl_send(file, argp, true); | |
5004 | #endif | |
c11d2c23 | 5005 | case BTRFS_IOC_GET_DEV_STATS: |
2ff7e61e | 5006 | return btrfs_ioctl_get_dev_stats(fs_info, argp); |
5d13a37b | 5007 | case BTRFS_IOC_QUOTA_CTL: |
905b0dda | 5008 | return btrfs_ioctl_quota_ctl(file, argp); |
5d13a37b | 5009 | case BTRFS_IOC_QGROUP_ASSIGN: |
905b0dda | 5010 | return btrfs_ioctl_qgroup_assign(file, argp); |
5d13a37b | 5011 | case BTRFS_IOC_QGROUP_CREATE: |
905b0dda | 5012 | return btrfs_ioctl_qgroup_create(file, argp); |
5d13a37b | 5013 | case BTRFS_IOC_QGROUP_LIMIT: |
905b0dda | 5014 | return btrfs_ioctl_qgroup_limit(file, argp); |
2f232036 JS |
5015 | case BTRFS_IOC_QUOTA_RESCAN: |
5016 | return btrfs_ioctl_quota_rescan(file, argp); | |
5017 | case BTRFS_IOC_QUOTA_RESCAN_STATUS: | |
b929c1d8 | 5018 | return btrfs_ioctl_quota_rescan_status(fs_info, argp); |
57254b6e | 5019 | case BTRFS_IOC_QUOTA_RESCAN_WAIT: |
b929c1d8 | 5020 | return btrfs_ioctl_quota_rescan_wait(fs_info, argp); |
3f6bcfbd | 5021 | case BTRFS_IOC_DEV_REPLACE: |
2ff7e61e | 5022 | return btrfs_ioctl_dev_replace(fs_info, argp); |
2eaa055f | 5023 | case BTRFS_IOC_GET_SUPPORTED_FEATURES: |
d5131b65 | 5024 | return btrfs_ioctl_get_supported_features(argp); |
2eaa055f | 5025 | case BTRFS_IOC_GET_FEATURES: |
b929c1d8 | 5026 | return btrfs_ioctl_get_features(fs_info, argp); |
2eaa055f JM |
5027 | case BTRFS_IOC_SET_FEATURES: |
5028 | return btrfs_ioctl_set_features(file, argp); | |
b64ec075 TM |
5029 | case BTRFS_IOC_GET_SUBVOL_INFO: |
5030 | return btrfs_ioctl_get_subvol_info(file, argp); | |
42e4b520 TM |
5031 | case BTRFS_IOC_GET_SUBVOL_ROOTREF: |
5032 | return btrfs_ioctl_get_subvol_rootref(file, argp); | |
23d0b79d TM |
5033 | case BTRFS_IOC_INO_LOOKUP_USER: |
5034 | return btrfs_ioctl_ino_lookup_user(file, argp); | |
14605409 BB |
5035 | case FS_IOC_ENABLE_VERITY: |
5036 | return fsverity_ioctl_enable(file, (const void __user *)argp); | |
5037 | case FS_IOC_MEASURE_VERITY: | |
5038 | return fsverity_ioctl_measure(file, argp); | |
f46b5a66 CH |
5039 | } |
5040 | ||
5041 | return -ENOTTY; | |
5042 | } | |
4c63c245 LD |
5043 | |
5044 | #ifdef CONFIG_COMPAT | |
5045 | long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) | |
5046 | { | |
2a362249 JM |
5047 | /* |
5048 | * These all access 32-bit values anyway so no further | |
5049 | * handling is necessary. | |
5050 | */ | |
4c63c245 | 5051 | switch (cmd) { |
4c63c245 LD |
5052 | case FS_IOC32_GETVERSION: |
5053 | cmd = FS_IOC_GETVERSION; | |
5054 | break; | |
4c63c245 LD |
5055 | } |
5056 | ||
5057 | return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); | |
5058 | } | |
5059 | #endif |