]>
Commit | Line | Data |
---|---|---|
e9be9d5e MS |
1 | /* |
2 | * | |
3 | * Copyright (C) 2011 Novell Inc. | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify it | |
6 | * under the terms of the GNU General Public License version 2 as published by | |
7 | * the Free Software Foundation. | |
8 | */ | |
9 | ||
10 | #include <linux/fs.h> | |
11 | #include <linux/slab.h> | |
5b825c3a | 12 | #include <linux/cred.h> |
e9be9d5e | 13 | #include <linux/xattr.h> |
5201dc44 | 14 | #include <linux/posix_acl.h> |
5f8415d6 | 15 | #include <linux/ratelimit.h> |
e9be9d5e MS |
16 | #include "overlayfs.h" |
17 | ||
ba1e563c CR |
18 | |
19 | static dev_t ovl_get_pseudo_dev(struct dentry *dentry) | |
20 | { | |
21 | struct ovl_entry *oe = dentry->d_fsdata; | |
22 | ||
23 | return oe->lowerstack[0].layer->pseudo_dev; | |
24 | } | |
25 | ||
e9be9d5e MS |
26 | int ovl_setattr(struct dentry *dentry, struct iattr *attr) |
27 | { | |
28 | int err; | |
29 | struct dentry *upperdentry; | |
1175b6b8 | 30 | const struct cred *old_cred; |
e9be9d5e | 31 | |
cf9a6784 MS |
32 | /* |
33 | * Check for permissions before trying to copy-up. This is redundant | |
34 | * since it will be rechecked later by ->setattr() on upper dentry. But | |
35 | * without this, copy-up can be triggered by just about anybody. | |
36 | * | |
37 | * We don't initialize inode->size, which just means that | |
38 | * inode_newsize_ok() will always check against MAX_LFS_FILESIZE and not | |
39 | * check for a swapfile (which this won't be anyway). | |
40 | */ | |
31051c85 | 41 | err = setattr_prepare(dentry, attr); |
cf9a6784 MS |
42 | if (err) |
43 | return err; | |
44 | ||
e9be9d5e MS |
45 | err = ovl_want_write(dentry); |
46 | if (err) | |
47 | goto out; | |
48 | ||
acff81ec MS |
49 | err = ovl_copy_up(dentry); |
50 | if (!err) { | |
51 | upperdentry = ovl_dentry_upper(dentry); | |
52 | ||
b99c2d91 MS |
53 | if (attr->ia_valid & (ATTR_KILL_SUID|ATTR_KILL_SGID)) |
54 | attr->ia_valid &= ~ATTR_MODE; | |
55 | ||
5955102c | 56 | inode_lock(upperdentry->d_inode); |
1175b6b8 | 57 | old_cred = ovl_override_creds(dentry->d_sb); |
e9be9d5e | 58 | err = notify_change(upperdentry, attr, NULL); |
1175b6b8 | 59 | revert_creds(old_cred); |
b81de061 KK |
60 | if (!err) |
61 | ovl_copyattr(upperdentry->d_inode, dentry->d_inode); | |
5955102c | 62 | inode_unlock(upperdentry->d_inode); |
e9be9d5e MS |
63 | } |
64 | ovl_drop_write(dentry); | |
65 | out: | |
66 | return err; | |
67 | } | |
68 | ||
5b712091 MS |
69 | int ovl_getattr(const struct path *path, struct kstat *stat, |
70 | u32 request_mask, unsigned int flags) | |
e9be9d5e | 71 | { |
a528d35e | 72 | struct dentry *dentry = path->dentry; |
72b608f0 | 73 | enum ovl_path_type type; |
e9be9d5e | 74 | struct path realpath; |
1175b6b8 | 75 | const struct cred *old_cred; |
5b712091 | 76 | bool is_dir = S_ISDIR(dentry->d_inode->i_mode); |
a0c5ad30 | 77 | bool samefs = ovl_same_sb(dentry->d_sb); |
1175b6b8 | 78 | int err; |
e9be9d5e | 79 | |
72b608f0 | 80 | type = ovl_path_real(dentry, &realpath); |
1175b6b8 | 81 | old_cred = ovl_override_creds(dentry->d_sb); |
a528d35e | 82 | err = vfs_getattr(&realpath, stat, request_mask, flags); |
72b608f0 AG |
83 | if (err) |
84 | goto out; | |
85 | ||
86 | /* | |
a0c5ad30 AG |
87 | * For non-dir or same fs, we use st_ino of the copy up origin, if we |
88 | * know it. This guaranties constant st_dev/st_ino across copy up. | |
72b608f0 AG |
89 | * |
90 | * If filesystem supports NFS export ops, this also guaranties | |
91 | * persistent st_ino across mount cycle. | |
92 | */ | |
a0c5ad30 | 93 | if (!is_dir || samefs) { |
72b608f0 AG |
94 | if (OVL_TYPE_ORIGIN(type)) { |
95 | struct kstat lowerstat; | |
5b712091 | 96 | u32 lowermask = STATX_INO | (!is_dir ? STATX_NLINK : 0); |
72b608f0 AG |
97 | |
98 | ovl_path_lower(dentry, &realpath); | |
99 | err = vfs_getattr(&realpath, &lowerstat, | |
5b712091 | 100 | lowermask, flags); |
72b608f0 AG |
101 | if (err) |
102 | goto out; | |
103 | ||
72b608f0 | 104 | /* |
359f392c | 105 | * Lower hardlinks may be broken on copy up to different |
72b608f0 AG |
106 | * upper files, so we cannot use the lower origin st_ino |
107 | * for those different files, even for the same fs case. | |
359f392c AG |
108 | * With inodes index enabled, it is safe to use st_ino |
109 | * of an indexed hardlinked origin. The index validates | |
110 | * that the upper hardlink is not broken. | |
72b608f0 | 111 | */ |
359f392c AG |
112 | if (is_dir || lowerstat.nlink == 1 || |
113 | ovl_test_flag(OVL_INDEX, d_inode(dentry))) | |
72b608f0 | 114 | stat->ino = lowerstat.ino; |
a0c5ad30 AG |
115 | |
116 | if (samefs) | |
117 | WARN_ON_ONCE(stat->dev != lowerstat.dev); | |
118 | else | |
119 | stat->dev = ovl_get_pseudo_dev(dentry); | |
72b608f0 | 120 | } |
a0c5ad30 AG |
121 | if (samefs) { |
122 | /* | |
123 | * When all layers are on the same fs, all real inode | |
124 | * number are unique, so we use the overlay st_dev, | |
125 | * which is friendly to du -x. | |
126 | */ | |
127 | stat->dev = dentry->d_sb->s_dev; | |
128 | } else if (!OVL_TYPE_UPPER(type)) { | |
129 | /* | |
130 | * For non-samefs setup, to make sure that st_dev/st_ino | |
131 | * pair is unique across the system, we use a unique | |
132 | * anonymous st_dev for lower layer inode. | |
133 | */ | |
134 | stat->dev = ovl_get_pseudo_dev(dentry); | |
135 | } | |
136 | } else { | |
5b712091 | 137 | /* |
5b712091 MS |
138 | * Always use the overlay st_dev for directories, so 'find |
139 | * -xdev' will scan the entire overlay mount and won't cross the | |
140 | * overlay mount boundaries. | |
a0c5ad30 AG |
141 | * |
142 | * If not all layers are on the same fs the pair {real st_ino; | |
143 | * overlay st_dev} is not unique, so use the non persistent | |
144 | * overlay st_ino for directories. | |
5b712091 MS |
145 | */ |
146 | stat->dev = dentry->d_sb->s_dev; | |
147 | stat->ino = dentry->d_inode->i_ino; | |
72b608f0 | 148 | } |
5b712091 MS |
149 | |
150 | /* | |
151 | * It's probably not worth it to count subdirs to get the | |
152 | * correct link count. nlink=1 seems to pacify 'find' and | |
153 | * other utilities. | |
154 | */ | |
155 | if (is_dir && OVL_TYPE_MERGE(type)) | |
156 | stat->nlink = 1; | |
157 | ||
5f8415d6 AG |
158 | /* |
159 | * Return the overlay inode nlinks for indexed upper inodes. | |
160 | * Overlay inode nlink counts the union of the upper hardlinks | |
161 | * and non-covered lower hardlinks. It does not include the upper | |
162 | * index hardlink. | |
163 | */ | |
164 | if (!is_dir && ovl_test_flag(OVL_INDEX, d_inode(dentry))) | |
165 | stat->nlink = dentry->d_inode->i_nlink; | |
166 | ||
72b608f0 | 167 | out: |
1175b6b8 | 168 | revert_creds(old_cred); |
72b608f0 | 169 | |
1175b6b8 | 170 | return err; |
e9be9d5e MS |
171 | } |
172 | ||
173 | int ovl_permission(struct inode *inode, int mask) | |
174 | { | |
09d8b586 MS |
175 | struct inode *upperinode = ovl_inode_upper(inode); |
176 | struct inode *realinode = upperinode ?: ovl_inode_lower(inode); | |
c0ca3d70 | 177 | const struct cred *old_cred; |
e9be9d5e MS |
178 | int err; |
179 | ||
e9be9d5e | 180 | /* Careful in RCU walk mode */ |
e9be9d5e MS |
181 | if (!realinode) { |
182 | WARN_ON(!(mask & MAY_NOT_BLOCK)); | |
a999d7e1 | 183 | return -ECHILD; |
e9be9d5e MS |
184 | } |
185 | ||
c0ca3d70 VG |
186 | /* |
187 | * Check overlay inode with the creds of task and underlying inode | |
188 | * with creds of mounter | |
189 | */ | |
190 | err = generic_permission(inode, mask); | |
191 | if (err) | |
192 | return err; | |
193 | ||
194 | old_cred = ovl_override_creds(inode->i_sb); | |
09d8b586 MS |
195 | if (!upperinode && |
196 | !special_file(realinode->i_mode) && mask & MAY_WRITE) { | |
754f8cb7 | 197 | mask &= ~(MAY_WRITE | MAY_APPEND); |
500cac3c VG |
198 | /* Make sure mounter can read file for copy up later */ |
199 | mask |= MAY_READ; | |
200 | } | |
9c630ebe | 201 | err = inode_permission(realinode, mask); |
c0ca3d70 VG |
202 | revert_creds(old_cred); |
203 | ||
204 | return err; | |
e9be9d5e MS |
205 | } |
206 | ||
6b255391 | 207 | static const char *ovl_get_link(struct dentry *dentry, |
fceef393 AV |
208 | struct inode *inode, |
209 | struct delayed_call *done) | |
e9be9d5e | 210 | { |
1175b6b8 VG |
211 | const struct cred *old_cred; |
212 | const char *p; | |
e9be9d5e | 213 | |
6b255391 AV |
214 | if (!dentry) |
215 | return ERR_PTR(-ECHILD); | |
216 | ||
1175b6b8 | 217 | old_cred = ovl_override_creds(dentry->d_sb); |
7764235b | 218 | p = vfs_get_link(ovl_dentry_real(dentry), done); |
1175b6b8 VG |
219 | revert_creds(old_cred); |
220 | return p; | |
e9be9d5e MS |
221 | } |
222 | ||
0956254a | 223 | bool ovl_is_private_xattr(const char *name) |
e9be9d5e | 224 | { |
fe2b7595 AG |
225 | return strncmp(name, OVL_XATTR_PREFIX, |
226 | sizeof(OVL_XATTR_PREFIX) - 1) == 0; | |
e9be9d5e MS |
227 | } |
228 | ||
1d88f183 MS |
229 | int ovl_xattr_set(struct dentry *dentry, struct inode *inode, const char *name, |
230 | const void *value, size_t size, int flags) | |
e9be9d5e MS |
231 | { |
232 | int err; | |
1d88f183 MS |
233 | struct dentry *upperdentry = ovl_i_dentry_upper(inode); |
234 | struct dentry *realdentry = upperdentry ?: ovl_dentry_lower(dentry); | |
1175b6b8 | 235 | const struct cred *old_cred; |
e9be9d5e MS |
236 | |
237 | err = ovl_want_write(dentry); | |
238 | if (err) | |
239 | goto out; | |
240 | ||
1d88f183 MS |
241 | if (!value && !upperdentry) { |
242 | err = vfs_getxattr(realdentry, name, NULL, 0); | |
0e585ccc AG |
243 | if (err < 0) |
244 | goto out_drop_write; | |
245 | } | |
246 | ||
1d88f183 MS |
247 | if (!upperdentry) { |
248 | err = ovl_copy_up(dentry); | |
249 | if (err) | |
250 | goto out_drop_write; | |
e9be9d5e | 251 | |
1d88f183 MS |
252 | realdentry = ovl_dentry_upper(dentry); |
253 | } | |
0e585ccc | 254 | |
1175b6b8 | 255 | old_cred = ovl_override_creds(dentry->d_sb); |
0e585ccc | 256 | if (value) |
1d88f183 | 257 | err = vfs_setxattr(realdentry, name, value, size, flags); |
0e585ccc AG |
258 | else { |
259 | WARN_ON(flags != XATTR_REPLACE); | |
1d88f183 | 260 | err = vfs_removexattr(realdentry, name); |
0e585ccc | 261 | } |
1175b6b8 | 262 | revert_creds(old_cred); |
e9be9d5e MS |
263 | |
264 | out_drop_write: | |
265 | ovl_drop_write(dentry); | |
266 | out: | |
267 | return err; | |
268 | } | |
269 | ||
1d88f183 | 270 | int ovl_xattr_get(struct dentry *dentry, struct inode *inode, const char *name, |
0eb45fc3 | 271 | void *value, size_t size) |
e9be9d5e | 272 | { |
1175b6b8 VG |
273 | ssize_t res; |
274 | const struct cred *old_cred; | |
1d88f183 MS |
275 | struct dentry *realdentry = |
276 | ovl_i_dentry_upper(inode) ?: ovl_dentry_lower(dentry); | |
52148463 | 277 | |
1175b6b8 VG |
278 | old_cred = ovl_override_creds(dentry->d_sb); |
279 | res = vfs_getxattr(realdentry, name, value, size); | |
280 | revert_creds(old_cred); | |
281 | return res; | |
e9be9d5e MS |
282 | } |
283 | ||
a082c6f6 MS |
284 | static bool ovl_can_list(const char *s) |
285 | { | |
286 | /* List all non-trusted xatts */ | |
287 | if (strncmp(s, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) != 0) | |
288 | return true; | |
289 | ||
290 | /* Never list trusted.overlay, list other trusted for superuser only */ | |
291 | return !ovl_is_private_xattr(s) && capable(CAP_SYS_ADMIN); | |
292 | } | |
293 | ||
e9be9d5e MS |
294 | ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size) |
295 | { | |
b581755b | 296 | struct dentry *realdentry = ovl_dentry_real(dentry); |
e9be9d5e | 297 | ssize_t res; |
7cb35119 MS |
298 | size_t len; |
299 | char *s; | |
1175b6b8 | 300 | const struct cred *old_cred; |
e9be9d5e | 301 | |
1175b6b8 | 302 | old_cred = ovl_override_creds(dentry->d_sb); |
b581755b | 303 | res = vfs_listxattr(realdentry, list, size); |
1175b6b8 | 304 | revert_creds(old_cred); |
e9be9d5e MS |
305 | if (res <= 0 || size == 0) |
306 | return res; | |
307 | ||
e9be9d5e | 308 | /* filter out private xattrs */ |
7cb35119 MS |
309 | for (s = list, len = res; len;) { |
310 | size_t slen = strnlen(s, len) + 1; | |
e9be9d5e | 311 | |
7cb35119 MS |
312 | /* underlying fs providing us with an broken xattr list? */ |
313 | if (WARN_ON(slen > len)) | |
314 | return -EIO; | |
e9be9d5e | 315 | |
7cb35119 | 316 | len -= slen; |
a082c6f6 | 317 | if (!ovl_can_list(s)) { |
e9be9d5e | 318 | res -= slen; |
7cb35119 | 319 | memmove(s, s + slen, len); |
e9be9d5e | 320 | } else { |
7cb35119 | 321 | s += slen; |
e9be9d5e MS |
322 | } |
323 | } | |
324 | ||
325 | return res; | |
326 | } | |
327 | ||
39a25b2b VG |
328 | struct posix_acl *ovl_get_acl(struct inode *inode, int type) |
329 | { | |
09d8b586 | 330 | struct inode *realinode = ovl_inode_real(inode); |
1175b6b8 VG |
331 | const struct cred *old_cred; |
332 | struct posix_acl *acl; | |
39a25b2b | 333 | |
5201dc44 | 334 | if (!IS_ENABLED(CONFIG_FS_POSIX_ACL) || !IS_POSIXACL(realinode)) |
39a25b2b VG |
335 | return NULL; |
336 | ||
1175b6b8 | 337 | old_cred = ovl_override_creds(inode->i_sb); |
5201dc44 | 338 | acl = get_acl(realinode, type); |
1175b6b8 VG |
339 | revert_creds(old_cred); |
340 | ||
341 | return acl; | |
39a25b2b VG |
342 | } |
343 | ||
59be0971 | 344 | static bool ovl_open_need_copy_up(struct dentry *dentry, int flags) |
e9be9d5e | 345 | { |
59be0971 AG |
346 | if (ovl_dentry_upper(dentry) && |
347 | ovl_dentry_has_upper_alias(dentry)) | |
e9be9d5e MS |
348 | return false; |
349 | ||
59be0971 | 350 | if (special_file(d_inode(dentry)->i_mode)) |
e9be9d5e MS |
351 | return false; |
352 | ||
353 | if (!(OPEN_FMODE(flags) & FMODE_WRITE) && !(flags & O_TRUNC)) | |
354 | return false; | |
355 | ||
356 | return true; | |
357 | } | |
358 | ||
2d902671 | 359 | int ovl_open_maybe_copy_up(struct dentry *dentry, unsigned int file_flags) |
e9be9d5e | 360 | { |
2d902671 | 361 | int err = 0; |
e9be9d5e | 362 | |
59be0971 | 363 | if (ovl_open_need_copy_up(dentry, file_flags)) { |
e9be9d5e | 364 | err = ovl_want_write(dentry); |
2d902671 | 365 | if (!err) { |
9aba6521 | 366 | err = ovl_copy_up_flags(dentry, file_flags); |
2d902671 MS |
367 | ovl_drop_write(dentry); |
368 | } | |
e9be9d5e MS |
369 | } |
370 | ||
2d902671 | 371 | return err; |
e9be9d5e MS |
372 | } |
373 | ||
d719e8f2 MS |
374 | int ovl_update_time(struct inode *inode, struct timespec *ts, int flags) |
375 | { | |
376 | struct dentry *alias; | |
377 | struct path upperpath; | |
378 | ||
379 | if (!(flags & S_ATIME)) | |
380 | return 0; | |
381 | ||
382 | alias = d_find_any_alias(inode); | |
383 | if (!alias) | |
384 | return 0; | |
385 | ||
386 | ovl_path_upper(alias, &upperpath); | |
387 | if (upperpath.dentry) { | |
388 | touch_atime(&upperpath); | |
389 | inode->i_atime = d_inode(upperpath.dentry)->i_atime; | |
390 | } | |
391 | ||
392 | dput(alias); | |
393 | ||
394 | return 0; | |
395 | } | |
396 | ||
e9be9d5e MS |
397 | static const struct inode_operations ovl_file_inode_operations = { |
398 | .setattr = ovl_setattr, | |
399 | .permission = ovl_permission, | |
400 | .getattr = ovl_getattr, | |
e9be9d5e | 401 | .listxattr = ovl_listxattr, |
39a25b2b | 402 | .get_acl = ovl_get_acl, |
d719e8f2 | 403 | .update_time = ovl_update_time, |
e9be9d5e MS |
404 | }; |
405 | ||
406 | static const struct inode_operations ovl_symlink_inode_operations = { | |
407 | .setattr = ovl_setattr, | |
6b255391 | 408 | .get_link = ovl_get_link, |
e9be9d5e | 409 | .getattr = ovl_getattr, |
e9be9d5e | 410 | .listxattr = ovl_listxattr, |
d719e8f2 | 411 | .update_time = ovl_update_time, |
e9be9d5e MS |
412 | }; |
413 | ||
b1eaa950 AG |
414 | /* |
415 | * It is possible to stack overlayfs instance on top of another | |
416 | * overlayfs instance as lower layer. We need to annonate the | |
417 | * stackable i_mutex locks according to stack level of the super | |
418 | * block instance. An overlayfs instance can never be in stack | |
419 | * depth 0 (there is always a real fs below it). An overlayfs | |
420 | * inode lock will use the lockdep annotaion ovl_i_mutex_key[depth]. | |
421 | * | |
422 | * For example, here is a snip from /proc/lockdep_chains after | |
423 | * dir_iterate of nested overlayfs: | |
424 | * | |
425 | * [...] &ovl_i_mutex_dir_key[depth] (stack_depth=2) | |
426 | * [...] &ovl_i_mutex_dir_key[depth]#2 (stack_depth=1) | |
427 | * [...] &type->i_mutex_dir_key (stack_depth=0) | |
428 | */ | |
429 | #define OVL_MAX_NESTING FILESYSTEM_MAX_STACK_DEPTH | |
430 | ||
431 | static inline void ovl_lockdep_annotate_inode_mutex_key(struct inode *inode) | |
432 | { | |
433 | #ifdef CONFIG_LOCKDEP | |
434 | static struct lock_class_key ovl_i_mutex_key[OVL_MAX_NESTING]; | |
435 | static struct lock_class_key ovl_i_mutex_dir_key[OVL_MAX_NESTING]; | |
4eae06de | 436 | static struct lock_class_key ovl_i_lock_key[OVL_MAX_NESTING]; |
b1eaa950 AG |
437 | |
438 | int depth = inode->i_sb->s_stack_depth - 1; | |
439 | ||
440 | if (WARN_ON_ONCE(depth < 0 || depth >= OVL_MAX_NESTING)) | |
441 | depth = 0; | |
442 | ||
443 | if (S_ISDIR(inode->i_mode)) | |
444 | lockdep_set_class(&inode->i_rwsem, &ovl_i_mutex_dir_key[depth]); | |
445 | else | |
446 | lockdep_set_class(&inode->i_rwsem, &ovl_i_mutex_key[depth]); | |
4eae06de AG |
447 | |
448 | lockdep_set_class(&OVL_I(inode)->lock, &ovl_i_lock_key[depth]); | |
b1eaa950 AG |
449 | #endif |
450 | } | |
451 | ||
ca4c8a3a | 452 | static void ovl_fill_inode(struct inode *inode, umode_t mode, dev_t rdev) |
e9be9d5e | 453 | { |
e9be9d5e MS |
454 | inode->i_ino = get_next_ino(); |
455 | inode->i_mode = mode; | |
d719e8f2 | 456 | inode->i_flags |= S_NOCMTIME; |
2a3a2a3f MS |
457 | #ifdef CONFIG_FS_POSIX_ACL |
458 | inode->i_acl = inode->i_default_acl = ACL_DONT_CACHE; | |
459 | #endif | |
e9be9d5e | 460 | |
b1eaa950 AG |
461 | ovl_lockdep_annotate_inode_mutex_key(inode); |
462 | ||
ca4c8a3a MS |
463 | switch (mode & S_IFMT) { |
464 | case S_IFREG: | |
465 | inode->i_op = &ovl_file_inode_operations; | |
466 | break; | |
467 | ||
e9be9d5e | 468 | case S_IFDIR: |
e9be9d5e MS |
469 | inode->i_op = &ovl_dir_inode_operations; |
470 | inode->i_fop = &ovl_dir_operations; | |
471 | break; | |
472 | ||
473 | case S_IFLNK: | |
474 | inode->i_op = &ovl_symlink_inode_operations; | |
475 | break; | |
476 | ||
51f7e52d | 477 | default: |
e9be9d5e | 478 | inode->i_op = &ovl_file_inode_operations; |
ca4c8a3a | 479 | init_special_inode(inode, mode, rdev); |
e9be9d5e | 480 | break; |
51f7e52d MS |
481 | } |
482 | } | |
e9be9d5e | 483 | |
5f8415d6 AG |
484 | /* |
485 | * With inodes index enabled, an overlay inode nlink counts the union of upper | |
486 | * hardlinks and non-covered lower hardlinks. During the lifetime of a non-pure | |
487 | * upper inode, the following nlink modifying operations can happen: | |
488 | * | |
489 | * 1. Lower hardlink copy up | |
490 | * 2. Upper hardlink created, unlinked or renamed over | |
491 | * 3. Lower hardlink whiteout or renamed over | |
492 | * | |
493 | * For the first, copy up case, the union nlink does not change, whether the | |
494 | * operation succeeds or fails, but the upper inode nlink may change. | |
495 | * Therefore, before copy up, we store the union nlink value relative to the | |
496 | * lower inode nlink in the index inode xattr trusted.overlay.nlink. | |
497 | * | |
498 | * For the second, upper hardlink case, the union nlink should be incremented | |
499 | * or decremented IFF the operation succeeds, aligned with nlink change of the | |
500 | * upper inode. Therefore, before link/unlink/rename, we store the union nlink | |
501 | * value relative to the upper inode nlink in the index inode. | |
502 | * | |
503 | * For the last, lower cover up case, we simplify things by preceding the | |
504 | * whiteout or cover up with copy up. This makes sure that there is an index | |
505 | * upper inode where the nlink xattr can be stored before the copied up upper | |
506 | * entry is unlink. | |
507 | */ | |
508 | #define OVL_NLINK_ADD_UPPER (1 << 0) | |
509 | ||
510 | /* | |
511 | * On-disk format for indexed nlink: | |
512 | * | |
513 | * nlink relative to the upper inode - "U[+-]NUM" | |
514 | * nlink relative to the lower inode - "L[+-]NUM" | |
515 | */ | |
516 | ||
517 | static int ovl_set_nlink_common(struct dentry *dentry, | |
518 | struct dentry *realdentry, const char *format) | |
519 | { | |
520 | struct inode *inode = d_inode(dentry); | |
521 | struct inode *realinode = d_inode(realdentry); | |
522 | char buf[13]; | |
523 | int len; | |
524 | ||
525 | len = snprintf(buf, sizeof(buf), format, | |
526 | (int) (inode->i_nlink - realinode->i_nlink)); | |
527 | ||
6787341a MS |
528 | if (WARN_ON(len >= sizeof(buf))) |
529 | return -EIO; | |
530 | ||
5f8415d6 AG |
531 | return ovl_do_setxattr(ovl_dentry_upper(dentry), |
532 | OVL_XATTR_NLINK, buf, len, 0); | |
533 | } | |
534 | ||
535 | int ovl_set_nlink_upper(struct dentry *dentry) | |
536 | { | |
537 | return ovl_set_nlink_common(dentry, ovl_dentry_upper(dentry), "U%+i"); | |
538 | } | |
539 | ||
540 | int ovl_set_nlink_lower(struct dentry *dentry) | |
541 | { | |
542 | return ovl_set_nlink_common(dentry, ovl_dentry_lower(dentry), "L%+i"); | |
543 | } | |
544 | ||
caf70cb2 AG |
545 | unsigned int ovl_get_nlink(struct dentry *lowerdentry, |
546 | struct dentry *upperdentry, | |
547 | unsigned int fallback) | |
5f8415d6 AG |
548 | { |
549 | int nlink_diff; | |
550 | int nlink; | |
551 | char buf[13]; | |
552 | int err; | |
553 | ||
554 | if (!lowerdentry || !upperdentry || d_inode(lowerdentry)->i_nlink == 1) | |
555 | return fallback; | |
556 | ||
557 | err = vfs_getxattr(upperdentry, OVL_XATTR_NLINK, &buf, sizeof(buf) - 1); | |
558 | if (err < 0) | |
559 | goto fail; | |
560 | ||
561 | buf[err] = '\0'; | |
562 | if ((buf[0] != 'L' && buf[0] != 'U') || | |
563 | (buf[1] != '+' && buf[1] != '-')) | |
564 | goto fail; | |
565 | ||
566 | err = kstrtoint(buf + 1, 10, &nlink_diff); | |
567 | if (err < 0) | |
568 | goto fail; | |
569 | ||
570 | nlink = d_inode(buf[0] == 'L' ? lowerdentry : upperdentry)->i_nlink; | |
571 | nlink += nlink_diff; | |
572 | ||
573 | if (nlink <= 0) | |
574 | goto fail; | |
575 | ||
576 | return nlink; | |
577 | ||
578 | fail: | |
579 | pr_warn_ratelimited("overlayfs: failed to get index nlink (%pd2, err=%i)\n", | |
580 | upperdentry, err); | |
581 | return fallback; | |
582 | } | |
583 | ||
ca4c8a3a | 584 | struct inode *ovl_new_inode(struct super_block *sb, umode_t mode, dev_t rdev) |
51f7e52d MS |
585 | { |
586 | struct inode *inode; | |
587 | ||
588 | inode = new_inode(sb); | |
589 | if (inode) | |
ca4c8a3a | 590 | ovl_fill_inode(inode, mode, rdev); |
51f7e52d MS |
591 | |
592 | return inode; | |
593 | } | |
594 | ||
595 | static int ovl_inode_test(struct inode *inode, void *data) | |
596 | { | |
25b7713a | 597 | return inode->i_private == data; |
51f7e52d MS |
598 | } |
599 | ||
600 | static int ovl_inode_set(struct inode *inode, void *data) | |
601 | { | |
25b7713a | 602 | inode->i_private = data; |
51f7e52d MS |
603 | return 0; |
604 | } | |
605 | ||
b9ac5c27 MS |
606 | static bool ovl_verify_inode(struct inode *inode, struct dentry *lowerdentry, |
607 | struct dentry *upperdentry) | |
608 | { | |
d2c9584e AG |
609 | if (S_ISDIR(inode->i_mode)) { |
610 | /* Real lower dir moved to upper layer under us? */ | |
611 | if (!lowerdentry && ovl_inode_lower(inode)) | |
612 | return false; | |
613 | ||
614 | /* Lookup of an uncovered redirect origin? */ | |
615 | if (!upperdentry && ovl_inode_upper(inode)) | |
616 | return false; | |
617 | } | |
618 | ||
939ae4ef AG |
619 | /* |
620 | * Allow non-NULL lower inode in ovl_inode even if lowerdentry is NULL. | |
621 | * This happens when finding a copied up overlay inode for a renamed | |
622 | * or hardlinked overlay dentry and lower dentry cannot be followed | |
623 | * by origin because lower fs does not support file handles. | |
624 | */ | |
625 | if (lowerdentry && ovl_inode_lower(inode) != d_inode(lowerdentry)) | |
b9ac5c27 MS |
626 | return false; |
627 | ||
628 | /* | |
629 | * Allow non-NULL __upperdentry in inode even if upperdentry is NULL. | |
630 | * This happens when finding a lower alias for a copied up hard link. | |
631 | */ | |
632 | if (upperdentry && ovl_inode_upper(inode) != d_inode(upperdentry)) | |
633 | return false; | |
634 | ||
635 | return true; | |
636 | } | |
637 | ||
6eaf0111 AG |
638 | struct inode *ovl_get_inode(struct dentry *dentry, struct dentry *upperdentry, |
639 | struct dentry *index) | |
51f7e52d | 640 | { |
09d8b586 MS |
641 | struct dentry *lowerdentry = ovl_dentry_lower(dentry); |
642 | struct inode *realinode = upperdentry ? d_inode(upperdentry) : NULL; | |
51f7e52d | 643 | struct inode *inode; |
6eaf0111 AG |
644 | /* Already indexed or could be indexed on copy up? */ |
645 | bool indexed = (index || (ovl_indexdir(dentry->d_sb) && !upperdentry)); | |
d2c9584e AG |
646 | struct dentry *origin = indexed ? lowerdentry : NULL; |
647 | bool is_dir; | |
6eaf0111 AG |
648 | |
649 | if (WARN_ON(upperdentry && indexed && !lowerdentry)) | |
650 | return ERR_PTR(-EIO); | |
51f7e52d | 651 | |
09d8b586 MS |
652 | if (!realinode) |
653 | realinode = d_inode(lowerdentry); | |
654 | ||
6eaf0111 | 655 | /* |
d2c9584e AG |
656 | * Copy up origin (lower) may exist for non-indexed non-dir upper, but |
657 | * we must not use lower as hash key in that case. | |
658 | * Hash non-dir that is or could be indexed by origin inode. | |
659 | * Hash dir that is or could be merged by origin inode. | |
660 | * Hash pure upper and non-indexed non-dir by upper inode. | |
6eaf0111 | 661 | */ |
d2c9584e AG |
662 | is_dir = S_ISDIR(realinode->i_mode); |
663 | if (is_dir) | |
664 | origin = lowerdentry; | |
665 | ||
666 | if (upperdentry || origin) { | |
667 | struct inode *key = d_inode(origin ?: upperdentry); | |
668 | unsigned int nlink = is_dir ? 1 : realinode->i_nlink; | |
b9ac5c27 MS |
669 | |
670 | inode = iget5_locked(dentry->d_sb, (unsigned long) key, | |
671 | ovl_inode_test, ovl_inode_set, key); | |
09d8b586 | 672 | if (!inode) |
b9ac5c27 | 673 | goto out_nomem; |
09d8b586 | 674 | if (!(inode->i_state & I_NEW)) { |
b9ac5c27 MS |
675 | /* |
676 | * Verify that the underlying files stored in the inode | |
677 | * match those in the dentry. | |
678 | */ | |
679 | if (!ovl_verify_inode(inode, lowerdentry, upperdentry)) { | |
680 | iput(inode); | |
681 | inode = ERR_PTR(-ESTALE); | |
682 | goto out; | |
683 | } | |
684 | ||
09d8b586 MS |
685 | dput(upperdentry); |
686 | goto out; | |
687 | } | |
e6d2ebdd | 688 | |
d2c9584e AG |
689 | /* Recalculate nlink for non-dir due to indexing */ |
690 | if (!is_dir) | |
691 | nlink = ovl_get_nlink(lowerdentry, upperdentry, nlink); | |
5f8415d6 | 692 | set_nlink(inode, nlink); |
e6d2ebdd MS |
693 | } else { |
694 | inode = new_inode(dentry->d_sb); | |
695 | if (!inode) | |
b9ac5c27 | 696 | goto out_nomem; |
e9be9d5e | 697 | } |
e6d2ebdd | 698 | ovl_fill_inode(inode, realinode->i_mode, realinode->i_rdev); |
09d8b586 | 699 | ovl_inode_init(inode, upperdentry, lowerdentry); |
13c72075 MS |
700 | |
701 | if (upperdentry && ovl_is_impuredir(upperdentry)) | |
702 | ovl_set_flag(OVL_IMPURE, inode); | |
703 | ||
b79e05aa | 704 | /* Check for non-merge dir that may have whiteouts */ |
d2c9584e | 705 | if (is_dir) { |
b79e05aa AG |
706 | struct ovl_entry *oe = dentry->d_fsdata; |
707 | ||
708 | if (((upperdentry && lowerdentry) || oe->numlower > 1) || | |
709 | ovl_check_origin_xattr(upperdentry ?: lowerdentry)) { | |
710 | ovl_set_flag(OVL_WHITEOUTS, inode); | |
711 | } | |
712 | } | |
713 | ||
e6d2ebdd MS |
714 | if (inode->i_state & I_NEW) |
715 | unlock_new_inode(inode); | |
716 | out: | |
e9be9d5e | 717 | return inode; |
b9ac5c27 MS |
718 | |
719 | out_nomem: | |
720 | inode = ERR_PTR(-ENOMEM); | |
721 | goto out; | |
e9be9d5e | 722 | } |