Commit | Line | Data |
---|---|---|
d2912cb1 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
e9be9d5e MS |
2 | /* |
3 | * | |
4 | * Copyright (C) 2011 Novell Inc. | |
e9be9d5e MS |
5 | */ |
6 | ||
5b825c3a | 7 | #include <uapi/linux/magic.h> |
e9be9d5e MS |
8 | #include <linux/fs.h> |
9 | #include <linux/namei.h> | |
10 | #include <linux/xattr.h> | |
e9be9d5e | 11 | #include <linux/mount.h> |
e9be9d5e MS |
12 | #include <linux/parser.h> |
13 | #include <linux/module.h> | |
cc259639 | 14 | #include <linux/statfs.h> |
f45827e8 | 15 | #include <linux/seq_file.h> |
d837a49b | 16 | #include <linux/posix_acl_xattr.h> |
e487d889 | 17 | #include <linux/exportfs.h> |
e9be9d5e MS |
18 | #include "overlayfs.h" |
19 | ||
20 | MODULE_AUTHOR("Miklos Szeredi <miklos@szeredi.hu>"); | |
21 | MODULE_DESCRIPTION("Overlay filesystem"); | |
22 | MODULE_LICENSE("GPL"); | |
23 | ||
e9be9d5e MS |
24 | |
25 | struct ovl_dir_cache; | |
26 | ||
a78d9f0d MS |
27 | #define OVL_MAX_STACK 500 |
28 | ||
688ea0e5 MS |
29 | static bool ovl_redirect_dir_def = IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_DIR); |
30 | module_param_named(redirect_dir, ovl_redirect_dir_def, bool, 0644); | |
253e7483 | 31 | MODULE_PARM_DESC(redirect_dir, |
688ea0e5 | 32 | "Default to on or off for the redirect_dir feature"); |
e9be9d5e | 33 | |
438c84c2 MS |
34 | static bool ovl_redirect_always_follow = |
35 | IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW); | |
36 | module_param_named(redirect_always_follow, ovl_redirect_always_follow, | |
37 | bool, 0644); | |
253e7483 | 38 | MODULE_PARM_DESC(redirect_always_follow, |
438c84c2 MS |
39 | "Follow redirects even if redirect_dir feature is turned off"); |
40 | ||
02bcd157 AG |
41 | static bool ovl_index_def = IS_ENABLED(CONFIG_OVERLAY_FS_INDEX); |
42 | module_param_named(index, ovl_index_def, bool, 0644); | |
253e7483 | 43 | MODULE_PARM_DESC(index, |
02bcd157 AG |
44 | "Default to on or off for the inodes index feature"); |
45 | ||
f168f109 AG |
46 | static bool ovl_nfs_export_def = IS_ENABLED(CONFIG_OVERLAY_FS_NFS_EXPORT); |
47 | module_param_named(nfs_export, ovl_nfs_export_def, bool, 0644); | |
253e7483 | 48 | MODULE_PARM_DESC(nfs_export, |
f168f109 AG |
49 | "Default to on or off for the NFS export feature"); |
50 | ||
795939a9 AG |
51 | static bool ovl_xino_auto_def = IS_ENABLED(CONFIG_OVERLAY_FS_XINO_AUTO); |
52 | module_param_named(xino_auto, ovl_xino_auto_def, bool, 0644); | |
253e7483 | 53 | MODULE_PARM_DESC(xino_auto, |
795939a9 AG |
54 | "Auto enable xino feature"); |
55 | ||
4155c10a MS |
56 | static void ovl_entry_stack_free(struct ovl_entry *oe) |
57 | { | |
58 | unsigned int i; | |
59 | ||
60 | for (i = 0; i < oe->numlower; i++) | |
61 | dput(oe->lowerstack[i].dentry); | |
62 | } | |
63 | ||
d5791044 VG |
64 | static bool ovl_metacopy_def = IS_ENABLED(CONFIG_OVERLAY_FS_METACOPY); |
65 | module_param_named(metacopy, ovl_metacopy_def, bool, 0644); | |
253e7483 | 66 | MODULE_PARM_DESC(metacopy, |
d5791044 VG |
67 | "Default to on or off for the metadata only copy up feature"); |
68 | ||
e9be9d5e MS |
69 | static void ovl_dentry_release(struct dentry *dentry) |
70 | { | |
71 | struct ovl_entry *oe = dentry->d_fsdata; | |
72 | ||
73 | if (oe) { | |
4155c10a | 74 | ovl_entry_stack_free(oe); |
e9be9d5e MS |
75 | kfree_rcu(oe, rcu); |
76 | } | |
77 | } | |
78 | ||
2d902671 | 79 | static struct dentry *ovl_d_real(struct dentry *dentry, |
fb16043b | 80 | const struct inode *inode) |
d101a125 MS |
81 | { |
82 | struct dentry *real; | |
83 | ||
e8c985ba MS |
84 | /* It's an overlay file */ |
85 | if (inode && d_inode(dentry) == inode) | |
86 | return dentry; | |
87 | ||
ca4c8a3a | 88 | if (!d_is_reg(dentry)) { |
d101a125 MS |
89 | if (!inode || inode == d_inode(dentry)) |
90 | return dentry; | |
91 | goto bug; | |
92 | } | |
93 | ||
94 | real = ovl_dentry_upper(dentry); | |
2c3d7358 | 95 | if (real && (inode == d_inode(real))) |
d101a125 MS |
96 | return real; |
97 | ||
2c3d7358 VG |
98 | if (real && !inode && ovl_has_upperdata(d_inode(dentry))) |
99 | return real; | |
100 | ||
101 | real = ovl_dentry_lowerdata(dentry); | |
d101a125 MS |
102 | if (!real) |
103 | goto bug; | |
104 | ||
c4fcfc16 | 105 | /* Handle recursion */ |
fb16043b | 106 | real = d_real(real, inode); |
c4fcfc16 | 107 | |
d101a125 MS |
108 | if (!inode || inode == d_inode(real)) |
109 | return real; | |
d101a125 | 110 | bug: |
656189d2 | 111 | WARN(1, "ovl_d_real(%pd4, %s:%lu): real dentry not found\n", dentry, |
d101a125 MS |
112 | inode ? inode->i_sb->s_id : "NULL", inode ? inode->i_ino : 0); |
113 | return dentry; | |
114 | } | |
115 | ||
7c03b5d4 MS |
116 | static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags) |
117 | { | |
118 | struct ovl_entry *oe = dentry->d_fsdata; | |
119 | unsigned int i; | |
120 | int ret = 1; | |
121 | ||
122 | for (i = 0; i < oe->numlower; i++) { | |
123 | struct dentry *d = oe->lowerstack[i].dentry; | |
124 | ||
125 | if (d->d_flags & DCACHE_OP_REVALIDATE) { | |
126 | ret = d->d_op->d_revalidate(d, flags); | |
127 | if (ret < 0) | |
128 | return ret; | |
129 | if (!ret) { | |
130 | if (!(flags & LOOKUP_RCU)) | |
131 | d_invalidate(d); | |
132 | return -ESTALE; | |
133 | } | |
134 | } | |
135 | } | |
136 | return 1; | |
137 | } | |
138 | ||
139 | static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags) | |
140 | { | |
141 | struct ovl_entry *oe = dentry->d_fsdata; | |
142 | unsigned int i; | |
143 | int ret = 1; | |
144 | ||
145 | for (i = 0; i < oe->numlower; i++) { | |
146 | struct dentry *d = oe->lowerstack[i].dentry; | |
147 | ||
148 | if (d->d_flags & DCACHE_OP_WEAK_REVALIDATE) { | |
149 | ret = d->d_op->d_weak_revalidate(d, flags); | |
150 | if (ret <= 0) | |
151 | break; | |
152 | } | |
153 | } | |
154 | return ret; | |
155 | } | |
156 | ||
e9be9d5e MS |
157 | static const struct dentry_operations ovl_dentry_operations = { |
158 | .d_release = ovl_dentry_release, | |
d101a125 | 159 | .d_real = ovl_d_real, |
e9be9d5e MS |
160 | }; |
161 | ||
7c03b5d4 MS |
162 | static const struct dentry_operations ovl_reval_dentry_operations = { |
163 | .d_release = ovl_dentry_release, | |
d101a125 | 164 | .d_real = ovl_d_real, |
7c03b5d4 MS |
165 | .d_revalidate = ovl_dentry_revalidate, |
166 | .d_weak_revalidate = ovl_dentry_weak_revalidate, | |
167 | }; | |
168 | ||
13cf199d AG |
169 | static struct kmem_cache *ovl_inode_cachep; |
170 | ||
171 | static struct inode *ovl_alloc_inode(struct super_block *sb) | |
172 | { | |
173 | struct ovl_inode *oi = kmem_cache_alloc(ovl_inode_cachep, GFP_KERNEL); | |
174 | ||
b3885bd6 HN |
175 | if (!oi) |
176 | return NULL; | |
177 | ||
04a01ac7 | 178 | oi->cache = NULL; |
cf31c463 | 179 | oi->redirect = NULL; |
04a01ac7 | 180 | oi->version = 0; |
13c72075 | 181 | oi->flags = 0; |
09d8b586 | 182 | oi->__upperdentry = NULL; |
25b7713a | 183 | oi->lower = NULL; |
2664bd08 | 184 | oi->lowerdata = NULL; |
a015dafc | 185 | mutex_init(&oi->lock); |
25b7713a | 186 | |
13cf199d AG |
187 | return &oi->vfs_inode; |
188 | } | |
189 | ||
0b269ded | 190 | static void ovl_free_inode(struct inode *inode) |
13cf199d | 191 | { |
0b269ded | 192 | struct ovl_inode *oi = OVL_I(inode); |
13cf199d | 193 | |
0b269ded AV |
194 | kfree(oi->redirect); |
195 | mutex_destroy(&oi->lock); | |
196 | kmem_cache_free(ovl_inode_cachep, oi); | |
13cf199d AG |
197 | } |
198 | ||
199 | static void ovl_destroy_inode(struct inode *inode) | |
200 | { | |
09d8b586 MS |
201 | struct ovl_inode *oi = OVL_I(inode); |
202 | ||
203 | dput(oi->__upperdentry); | |
31747eda | 204 | iput(oi->lower); |
2664bd08 VG |
205 | if (S_ISDIR(inode->i_mode)) |
206 | ovl_dir_cache_free(inode); | |
207 | else | |
208 | iput(oi->lowerdata); | |
13cf199d AG |
209 | } |
210 | ||
ad204488 | 211 | static void ovl_free_fs(struct ovl_fs *ofs) |
e9be9d5e | 212 | { |
dd662667 | 213 | unsigned i; |
e9be9d5e | 214 | |
0be0bfd2 | 215 | iput(ofs->workbasedir_trap); |
146d62e5 AG |
216 | iput(ofs->indexdir_trap); |
217 | iput(ofs->workdir_trap); | |
218 | iput(ofs->upperdir_trap); | |
ad204488 MS |
219 | dput(ofs->indexdir); |
220 | dput(ofs->workdir); | |
221 | if (ofs->workdir_locked) | |
222 | ovl_inuse_unlock(ofs->workbasedir); | |
223 | dput(ofs->workbasedir); | |
224 | if (ofs->upperdir_locked) | |
225 | ovl_inuse_unlock(ofs->upper_mnt->mnt_root); | |
226 | mntput(ofs->upper_mnt); | |
146d62e5 AG |
227 | for (i = 0; i < ofs->numlower; i++) { |
228 | iput(ofs->lower_layers[i].trap); | |
ad204488 | 229 | mntput(ofs->lower_layers[i].mnt); |
146d62e5 | 230 | } |
5148626b AG |
231 | for (i = 0; i < ofs->numlowerfs; i++) |
232 | free_anon_bdev(ofs->lower_fs[i].pseudo_dev); | |
ad204488 | 233 | kfree(ofs->lower_layers); |
5148626b | 234 | kfree(ofs->lower_fs); |
ad204488 MS |
235 | |
236 | kfree(ofs->config.lowerdir); | |
237 | kfree(ofs->config.upperdir); | |
238 | kfree(ofs->config.workdir); | |
438c84c2 | 239 | kfree(ofs->config.redirect_mode); |
ad204488 MS |
240 | if (ofs->creator_cred) |
241 | put_cred(ofs->creator_cred); | |
242 | kfree(ofs); | |
e9be9d5e MS |
243 | } |
244 | ||
a9075cdb MS |
245 | static void ovl_put_super(struct super_block *sb) |
246 | { | |
247 | struct ovl_fs *ofs = sb->s_fs_info; | |
248 | ||
249 | ovl_free_fs(ofs); | |
250 | } | |
251 | ||
e8d4bfe3 | 252 | /* Sync real dirty inodes in upper filesystem (if it exists) */ |
e593b2bf AG |
253 | static int ovl_sync_fs(struct super_block *sb, int wait) |
254 | { | |
ad204488 | 255 | struct ovl_fs *ofs = sb->s_fs_info; |
e593b2bf AG |
256 | struct super_block *upper_sb; |
257 | int ret; | |
258 | ||
ad204488 | 259 | if (!ofs->upper_mnt) |
e593b2bf | 260 | return 0; |
e8d4bfe3 CX |
261 | |
262 | /* | |
263 | * If this is a sync(2) call or an emergency sync, all the super blocks | |
264 | * will be iterated, including upper_sb, so no need to do anything. | |
265 | * | |
266 | * If this is a syncfs(2) call, then we do need to call | |
267 | * sync_filesystem() on upper_sb, but enough if we do it when being | |
268 | * called with wait == 1. | |
269 | */ | |
270 | if (!wait) | |
e593b2bf AG |
271 | return 0; |
272 | ||
e8d4bfe3 CX |
273 | upper_sb = ofs->upper_mnt->mnt_sb; |
274 | ||
e593b2bf | 275 | down_read(&upper_sb->s_umount); |
e8d4bfe3 | 276 | ret = sync_filesystem(upper_sb); |
e593b2bf | 277 | up_read(&upper_sb->s_umount); |
e8d4bfe3 | 278 | |
e593b2bf AG |
279 | return ret; |
280 | } | |
281 | ||
cc259639 AW |
282 | /** |
283 | * ovl_statfs | |
284 | * @sb: The overlayfs super block | |
285 | * @buf: The struct kstatfs to fill in with stats | |
286 | * | |
287 | * Get the filesystem statistics. As writes always target the upper layer | |
4ebc5818 | 288 | * filesystem pass the statfs to the upper filesystem (if it exists) |
cc259639 AW |
289 | */ |
290 | static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf) | |
291 | { | |
292 | struct ovl_fs *ofs = dentry->d_sb->s_fs_info; | |
293 | struct dentry *root_dentry = dentry->d_sb->s_root; | |
294 | struct path path; | |
295 | int err; | |
296 | ||
4ebc5818 | 297 | ovl_path_real(root_dentry, &path); |
cc259639 AW |
298 | |
299 | err = vfs_statfs(&path, buf); | |
300 | if (!err) { | |
6b2d5fe4 | 301 | buf->f_namelen = ofs->namelen; |
cc259639 AW |
302 | buf->f_type = OVERLAYFS_SUPER_MAGIC; |
303 | } | |
304 | ||
305 | return err; | |
306 | } | |
307 | ||
02bcd157 | 308 | /* Will this overlay be forced to mount/remount ro? */ |
ad204488 | 309 | static bool ovl_force_readonly(struct ovl_fs *ofs) |
02bcd157 | 310 | { |
ad204488 | 311 | return (!ofs->upper_mnt || !ofs->workdir); |
02bcd157 AG |
312 | } |
313 | ||
438c84c2 MS |
314 | static const char *ovl_redirect_mode_def(void) |
315 | { | |
316 | return ovl_redirect_dir_def ? "on" : "off"; | |
317 | } | |
318 | ||
795939a9 AG |
319 | enum { |
320 | OVL_XINO_OFF, | |
321 | OVL_XINO_AUTO, | |
322 | OVL_XINO_ON, | |
323 | }; | |
324 | ||
325 | static const char * const ovl_xino_str[] = { | |
326 | "off", | |
327 | "auto", | |
328 | "on", | |
329 | }; | |
330 | ||
331 | static inline int ovl_xino_def(void) | |
332 | { | |
333 | return ovl_xino_auto_def ? OVL_XINO_AUTO : OVL_XINO_OFF; | |
334 | } | |
335 | ||
f45827e8 EZ |
336 | /** |
337 | * ovl_show_options | |
338 | * | |
339 | * Prints the mount options for a given superblock. | |
340 | * Returns zero; does not fail. | |
341 | */ | |
342 | static int ovl_show_options(struct seq_file *m, struct dentry *dentry) | |
343 | { | |
344 | struct super_block *sb = dentry->d_sb; | |
ad204488 | 345 | struct ovl_fs *ofs = sb->s_fs_info; |
f45827e8 | 346 | |
ad204488 MS |
347 | seq_show_option(m, "lowerdir", ofs->config.lowerdir); |
348 | if (ofs->config.upperdir) { | |
349 | seq_show_option(m, "upperdir", ofs->config.upperdir); | |
350 | seq_show_option(m, "workdir", ofs->config.workdir); | |
53a08cb9 | 351 | } |
ad204488 | 352 | if (ofs->config.default_permissions) |
8d3095f4 | 353 | seq_puts(m, ",default_permissions"); |
438c84c2 MS |
354 | if (strcmp(ofs->config.redirect_mode, ovl_redirect_mode_def()) != 0) |
355 | seq_printf(m, ",redirect_dir=%s", ofs->config.redirect_mode); | |
ad204488 | 356 | if (ofs->config.index != ovl_index_def) |
438c84c2 | 357 | seq_printf(m, ",index=%s", ofs->config.index ? "on" : "off"); |
f168f109 AG |
358 | if (ofs->config.nfs_export != ovl_nfs_export_def) |
359 | seq_printf(m, ",nfs_export=%s", ofs->config.nfs_export ? | |
360 | "on" : "off"); | |
795939a9 AG |
361 | if (ofs->config.xino != ovl_xino_def()) |
362 | seq_printf(m, ",xino=%s", ovl_xino_str[ofs->config.xino]); | |
d5791044 VG |
363 | if (ofs->config.metacopy != ovl_metacopy_def) |
364 | seq_printf(m, ",metacopy=%s", | |
365 | ofs->config.metacopy ? "on" : "off"); | |
f45827e8 EZ |
366 | return 0; |
367 | } | |
368 | ||
3cdf6fe9 SL |
369 | static int ovl_remount(struct super_block *sb, int *flags, char *data) |
370 | { | |
ad204488 | 371 | struct ovl_fs *ofs = sb->s_fs_info; |
3cdf6fe9 | 372 | |
1751e8a6 | 373 | if (!(*flags & SB_RDONLY) && ovl_force_readonly(ofs)) |
3cdf6fe9 SL |
374 | return -EROFS; |
375 | ||
376 | return 0; | |
377 | } | |
378 | ||
e9be9d5e | 379 | static const struct super_operations ovl_super_operations = { |
13cf199d | 380 | .alloc_inode = ovl_alloc_inode, |
0b269ded | 381 | .free_inode = ovl_free_inode, |
13cf199d AG |
382 | .destroy_inode = ovl_destroy_inode, |
383 | .drop_inode = generic_delete_inode, | |
e9be9d5e | 384 | .put_super = ovl_put_super, |
e593b2bf | 385 | .sync_fs = ovl_sync_fs, |
cc259639 | 386 | .statfs = ovl_statfs, |
f45827e8 | 387 | .show_options = ovl_show_options, |
3cdf6fe9 | 388 | .remount_fs = ovl_remount, |
e9be9d5e MS |
389 | }; |
390 | ||
391 | enum { | |
392 | OPT_LOWERDIR, | |
393 | OPT_UPPERDIR, | |
394 | OPT_WORKDIR, | |
8d3095f4 | 395 | OPT_DEFAULT_PERMISSIONS, |
438c84c2 | 396 | OPT_REDIRECT_DIR, |
02bcd157 AG |
397 | OPT_INDEX_ON, |
398 | OPT_INDEX_OFF, | |
f168f109 AG |
399 | OPT_NFS_EXPORT_ON, |
400 | OPT_NFS_EXPORT_OFF, | |
795939a9 AG |
401 | OPT_XINO_ON, |
402 | OPT_XINO_OFF, | |
403 | OPT_XINO_AUTO, | |
d5791044 VG |
404 | OPT_METACOPY_ON, |
405 | OPT_METACOPY_OFF, | |
e9be9d5e MS |
406 | OPT_ERR, |
407 | }; | |
408 | ||
409 | static const match_table_t ovl_tokens = { | |
410 | {OPT_LOWERDIR, "lowerdir=%s"}, | |
411 | {OPT_UPPERDIR, "upperdir=%s"}, | |
412 | {OPT_WORKDIR, "workdir=%s"}, | |
8d3095f4 | 413 | {OPT_DEFAULT_PERMISSIONS, "default_permissions"}, |
438c84c2 | 414 | {OPT_REDIRECT_DIR, "redirect_dir=%s"}, |
02bcd157 AG |
415 | {OPT_INDEX_ON, "index=on"}, |
416 | {OPT_INDEX_OFF, "index=off"}, | |
f168f109 AG |
417 | {OPT_NFS_EXPORT_ON, "nfs_export=on"}, |
418 | {OPT_NFS_EXPORT_OFF, "nfs_export=off"}, | |
795939a9 AG |
419 | {OPT_XINO_ON, "xino=on"}, |
420 | {OPT_XINO_OFF, "xino=off"}, | |
421 | {OPT_XINO_AUTO, "xino=auto"}, | |
d5791044 VG |
422 | {OPT_METACOPY_ON, "metacopy=on"}, |
423 | {OPT_METACOPY_OFF, "metacopy=off"}, | |
e9be9d5e MS |
424 | {OPT_ERR, NULL} |
425 | }; | |
426 | ||
91c77947 MS |
427 | static char *ovl_next_opt(char **s) |
428 | { | |
429 | char *sbegin = *s; | |
430 | char *p; | |
431 | ||
432 | if (sbegin == NULL) | |
433 | return NULL; | |
434 | ||
435 | for (p = sbegin; *p; p++) { | |
436 | if (*p == '\\') { | |
437 | p++; | |
438 | if (!*p) | |
439 | break; | |
440 | } else if (*p == ',') { | |
441 | *p = '\0'; | |
442 | *s = p + 1; | |
443 | return sbegin; | |
444 | } | |
445 | } | |
446 | *s = NULL; | |
447 | return sbegin; | |
448 | } | |
449 | ||
438c84c2 MS |
450 | static int ovl_parse_redirect_mode(struct ovl_config *config, const char *mode) |
451 | { | |
452 | if (strcmp(mode, "on") == 0) { | |
453 | config->redirect_dir = true; | |
454 | /* | |
455 | * Does not make sense to have redirect creation without | |
456 | * redirect following. | |
457 | */ | |
458 | config->redirect_follow = true; | |
459 | } else if (strcmp(mode, "follow") == 0) { | |
460 | config->redirect_follow = true; | |
461 | } else if (strcmp(mode, "off") == 0) { | |
462 | if (ovl_redirect_always_follow) | |
463 | config->redirect_follow = true; | |
464 | } else if (strcmp(mode, "nofollow") != 0) { | |
465 | pr_err("overlayfs: bad mount option \"redirect_dir=%s\"\n", | |
466 | mode); | |
467 | return -EINVAL; | |
468 | } | |
469 | ||
470 | return 0; | |
471 | } | |
472 | ||
e9be9d5e MS |
473 | static int ovl_parse_opt(char *opt, struct ovl_config *config) |
474 | { | |
475 | char *p; | |
d5791044 | 476 | int err; |
d47748e5 | 477 | bool metacopy_opt = false, redirect_opt = false; |
e9be9d5e | 478 | |
438c84c2 MS |
479 | config->redirect_mode = kstrdup(ovl_redirect_mode_def(), GFP_KERNEL); |
480 | if (!config->redirect_mode) | |
481 | return -ENOMEM; | |
482 | ||
91c77947 | 483 | while ((p = ovl_next_opt(&opt)) != NULL) { |
e9be9d5e MS |
484 | int token; |
485 | substring_t args[MAX_OPT_ARGS]; | |
486 | ||
487 | if (!*p) | |
488 | continue; | |
489 | ||
490 | token = match_token(p, ovl_tokens, args); | |
491 | switch (token) { | |
492 | case OPT_UPPERDIR: | |
493 | kfree(config->upperdir); | |
494 | config->upperdir = match_strdup(&args[0]); | |
495 | if (!config->upperdir) | |
496 | return -ENOMEM; | |
497 | break; | |
498 | ||
499 | case OPT_LOWERDIR: | |
500 | kfree(config->lowerdir); | |
501 | config->lowerdir = match_strdup(&args[0]); | |
502 | if (!config->lowerdir) | |
503 | return -ENOMEM; | |
504 | break; | |
505 | ||
506 | case OPT_WORKDIR: | |
507 | kfree(config->workdir); | |
508 | config->workdir = match_strdup(&args[0]); | |
509 | if (!config->workdir) | |
510 | return -ENOMEM; | |
511 | break; | |
512 | ||
8d3095f4 MS |
513 | case OPT_DEFAULT_PERMISSIONS: |
514 | config->default_permissions = true; | |
515 | break; | |
516 | ||
438c84c2 MS |
517 | case OPT_REDIRECT_DIR: |
518 | kfree(config->redirect_mode); | |
519 | config->redirect_mode = match_strdup(&args[0]); | |
520 | if (!config->redirect_mode) | |
521 | return -ENOMEM; | |
d47748e5 | 522 | redirect_opt = true; |
a6c60655 MS |
523 | break; |
524 | ||
02bcd157 AG |
525 | case OPT_INDEX_ON: |
526 | config->index = true; | |
527 | break; | |
528 | ||
529 | case OPT_INDEX_OFF: | |
530 | config->index = false; | |
531 | break; | |
532 | ||
f168f109 AG |
533 | case OPT_NFS_EXPORT_ON: |
534 | config->nfs_export = true; | |
535 | break; | |
536 | ||
537 | case OPT_NFS_EXPORT_OFF: | |
538 | config->nfs_export = false; | |
539 | break; | |
540 | ||
795939a9 AG |
541 | case OPT_XINO_ON: |
542 | config->xino = OVL_XINO_ON; | |
543 | break; | |
544 | ||
545 | case OPT_XINO_OFF: | |
546 | config->xino = OVL_XINO_OFF; | |
547 | break; | |
548 | ||
549 | case OPT_XINO_AUTO: | |
550 | config->xino = OVL_XINO_AUTO; | |
551 | break; | |
552 | ||
d5791044 VG |
553 | case OPT_METACOPY_ON: |
554 | config->metacopy = true; | |
d47748e5 | 555 | metacopy_opt = true; |
d5791044 VG |
556 | break; |
557 | ||
558 | case OPT_METACOPY_OFF: | |
559 | config->metacopy = false; | |
560 | break; | |
561 | ||
e9be9d5e | 562 | default: |
bead55ef | 563 | pr_err("overlayfs: unrecognized mount option \"%s\" or missing value\n", p); |
e9be9d5e MS |
564 | return -EINVAL; |
565 | } | |
566 | } | |
71cbad7e | 567 | |
568 | /* Workdir is useless in non-upper mount */ | |
569 | if (!config->upperdir && config->workdir) { | |
570 | pr_info("overlayfs: option \"workdir=%s\" is useless in a non-upper mount, ignore\n", | |
571 | config->workdir); | |
572 | kfree(config->workdir); | |
573 | config->workdir = NULL; | |
574 | } | |
575 | ||
d5791044 VG |
576 | err = ovl_parse_redirect_mode(config, config->redirect_mode); |
577 | if (err) | |
578 | return err; | |
579 | ||
d47748e5 MS |
580 | /* |
581 | * This is to make the logic below simpler. It doesn't make any other | |
582 | * difference, since config->redirect_dir is only used for upper. | |
583 | */ | |
584 | if (!config->upperdir && config->redirect_follow) | |
585 | config->redirect_dir = true; | |
586 | ||
587 | /* Resolve metacopy -> redirect_dir dependency */ | |
588 | if (config->metacopy && !config->redirect_dir) { | |
589 | if (metacopy_opt && redirect_opt) { | |
590 | pr_err("overlayfs: conflicting options: metacopy=on,redirect_dir=%s\n", | |
591 | config->redirect_mode); | |
592 | return -EINVAL; | |
593 | } | |
594 | if (redirect_opt) { | |
595 | /* | |
596 | * There was an explicit redirect_dir=... that resulted | |
597 | * in this conflict. | |
598 | */ | |
599 | pr_info("overlayfs: disabling metacopy due to redirect_dir=%s\n", | |
600 | config->redirect_mode); | |
601 | config->metacopy = false; | |
602 | } else { | |
603 | /* Automatically enable redirect otherwise. */ | |
604 | config->redirect_follow = config->redirect_dir = true; | |
605 | } | |
d5791044 VG |
606 | } |
607 | ||
608 | return 0; | |
e9be9d5e MS |
609 | } |
610 | ||
611 | #define OVL_WORKDIR_NAME "work" | |
02bcd157 | 612 | #define OVL_INDEXDIR_NAME "index" |
e9be9d5e | 613 | |
ad204488 | 614 | static struct dentry *ovl_workdir_create(struct ovl_fs *ofs, |
6b8aa129 | 615 | const char *name, bool persist) |
e9be9d5e | 616 | { |
ad204488 MS |
617 | struct inode *dir = ofs->workbasedir->d_inode; |
618 | struct vfsmount *mnt = ofs->upper_mnt; | |
e9be9d5e MS |
619 | struct dentry *work; |
620 | int err; | |
621 | bool retried = false; | |
6b8aa129 | 622 | bool locked = false; |
e9be9d5e | 623 | |
5955102c | 624 | inode_lock_nested(dir, I_MUTEX_PARENT); |
6b8aa129 AG |
625 | locked = true; |
626 | ||
e9be9d5e | 627 | retry: |
ad204488 | 628 | work = lookup_one_len(name, ofs->workbasedir, strlen(name)); |
e9be9d5e MS |
629 | |
630 | if (!IS_ERR(work)) { | |
c11b9fdd MS |
631 | struct iattr attr = { |
632 | .ia_valid = ATTR_MODE, | |
32a3d848 | 633 | .ia_mode = S_IFDIR | 0, |
c11b9fdd | 634 | }; |
e9be9d5e MS |
635 | |
636 | if (work->d_inode) { | |
637 | err = -EEXIST; | |
638 | if (retried) | |
639 | goto out_dput; | |
640 | ||
6b8aa129 AG |
641 | if (persist) |
642 | goto out_unlock; | |
643 | ||
e9be9d5e | 644 | retried = true; |
eea2fb48 | 645 | ovl_workdir_cleanup(dir, mnt, work, 0); |
e9be9d5e MS |
646 | dput(work); |
647 | goto retry; | |
648 | } | |
649 | ||
95a1c815 MS |
650 | work = ovl_create_real(dir, work, OVL_CATTR(attr.ia_mode)); |
651 | err = PTR_ERR(work); | |
652 | if (IS_ERR(work)) | |
653 | goto out_err; | |
c11b9fdd | 654 | |
cb348edb MS |
655 | /* |
656 | * Try to remove POSIX ACL xattrs from workdir. We are good if: | |
657 | * | |
658 | * a) success (there was a POSIX ACL xattr and was removed) | |
659 | * b) -ENODATA (there was no POSIX ACL xattr) | |
660 | * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported) | |
661 | * | |
662 | * There are various other error values that could effectively | |
663 | * mean that the xattr doesn't exist (e.g. -ERANGE is returned | |
664 | * if the xattr name is too long), but the set of filesystems | |
665 | * allowed as upper are limited to "normal" ones, where checking | |
666 | * for the above two errors is sufficient. | |
667 | */ | |
c11b9fdd | 668 | err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_DEFAULT); |
e1ff3dd1 | 669 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
670 | goto out_dput; |
671 | ||
672 | err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_ACCESS); | |
e1ff3dd1 | 673 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
674 | goto out_dput; |
675 | ||
676 | /* Clear any inherited mode bits */ | |
677 | inode_lock(work->d_inode); | |
678 | err = notify_change(work, &attr, NULL); | |
679 | inode_unlock(work->d_inode); | |
680 | if (err) | |
681 | goto out_dput; | |
6b8aa129 AG |
682 | } else { |
683 | err = PTR_ERR(work); | |
684 | goto out_err; | |
e9be9d5e MS |
685 | } |
686 | out_unlock: | |
6b8aa129 AG |
687 | if (locked) |
688 | inode_unlock(dir); | |
e9be9d5e MS |
689 | |
690 | return work; | |
691 | ||
692 | out_dput: | |
693 | dput(work); | |
6b8aa129 AG |
694 | out_err: |
695 | pr_warn("overlayfs: failed to create directory %s/%s (errno: %i); mounting read-only\n", | |
ad204488 | 696 | ofs->config.workdir, name, -err); |
6b8aa129 | 697 | work = NULL; |
e9be9d5e MS |
698 | goto out_unlock; |
699 | } | |
700 | ||
91c77947 MS |
701 | static void ovl_unescape(char *s) |
702 | { | |
703 | char *d = s; | |
704 | ||
705 | for (;; s++, d++) { | |
706 | if (*s == '\\') | |
707 | s++; | |
708 | *d = *s; | |
709 | if (!*s) | |
710 | break; | |
711 | } | |
712 | } | |
713 | ||
ab508822 MS |
714 | static int ovl_mount_dir_noesc(const char *name, struct path *path) |
715 | { | |
a78d9f0d | 716 | int err = -EINVAL; |
ab508822 | 717 | |
a78d9f0d MS |
718 | if (!*name) { |
719 | pr_err("overlayfs: empty lowerdir\n"); | |
720 | goto out; | |
721 | } | |
ab508822 MS |
722 | err = kern_path(name, LOOKUP_FOLLOW, path); |
723 | if (err) { | |
724 | pr_err("overlayfs: failed to resolve '%s': %i\n", name, err); | |
725 | goto out; | |
726 | } | |
727 | err = -EINVAL; | |
7c03b5d4 | 728 | if (ovl_dentry_weird(path->dentry)) { |
ab508822 MS |
729 | pr_err("overlayfs: filesystem on '%s' not supported\n", name); |
730 | goto out_put; | |
731 | } | |
2b8c30e9 | 732 | if (!d_is_dir(path->dentry)) { |
ab508822 MS |
733 | pr_err("overlayfs: '%s' not a directory\n", name); |
734 | goto out_put; | |
735 | } | |
736 | return 0; | |
737 | ||
738 | out_put: | |
8aafcb59 | 739 | path_put_init(path); |
ab508822 MS |
740 | out: |
741 | return err; | |
742 | } | |
743 | ||
744 | static int ovl_mount_dir(const char *name, struct path *path) | |
745 | { | |
746 | int err = -ENOMEM; | |
747 | char *tmp = kstrdup(name, GFP_KERNEL); | |
748 | ||
749 | if (tmp) { | |
750 | ovl_unescape(tmp); | |
751 | err = ovl_mount_dir_noesc(tmp, path); | |
7c03b5d4 MS |
752 | |
753 | if (!err) | |
754 | if (ovl_dentry_remote(path->dentry)) { | |
755 | pr_err("overlayfs: filesystem on '%s' not supported as upperdir\n", | |
756 | tmp); | |
8aafcb59 | 757 | path_put_init(path); |
7c03b5d4 MS |
758 | err = -EINVAL; |
759 | } | |
ab508822 MS |
760 | kfree(tmp); |
761 | } | |
762 | return err; | |
763 | } | |
764 | ||
6b2d5fe4 MS |
765 | static int ovl_check_namelen(struct path *path, struct ovl_fs *ofs, |
766 | const char *name) | |
ab508822 | 767 | { |
ab508822 | 768 | struct kstatfs statfs; |
6b2d5fe4 MS |
769 | int err = vfs_statfs(path, &statfs); |
770 | ||
771 | if (err) | |
772 | pr_err("overlayfs: statfs failed on '%s'\n", name); | |
773 | else | |
774 | ofs->namelen = max(ofs->namelen, statfs.f_namelen); | |
775 | ||
776 | return err; | |
777 | } | |
778 | ||
779 | static int ovl_lower_dir(const char *name, struct path *path, | |
780 | struct ovl_fs *ofs, int *stack_depth, bool *remote) | |
781 | { | |
e487d889 | 782 | int fh_type; |
6b2d5fe4 | 783 | int err; |
ab508822 | 784 | |
a78d9f0d | 785 | err = ovl_mount_dir_noesc(name, path); |
ab508822 MS |
786 | if (err) |
787 | goto out; | |
788 | ||
6b2d5fe4 MS |
789 | err = ovl_check_namelen(path, ofs, name); |
790 | if (err) | |
ab508822 | 791 | goto out_put; |
6b2d5fe4 | 792 | |
ab508822 MS |
793 | *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth); |
794 | ||
7c03b5d4 MS |
795 | if (ovl_dentry_remote(path->dentry)) |
796 | *remote = true; | |
797 | ||
02bcd157 | 798 | /* |
f168f109 AG |
799 | * The inodes index feature and NFS export need to encode and decode |
800 | * file handles, so they require that all layers support them. | |
02bcd157 | 801 | */ |
e487d889 | 802 | fh_type = ovl_can_decode_fh(path->dentry->d_sb); |
f168f109 | 803 | if ((ofs->config.nfs_export || |
e487d889 | 804 | (ofs->config.index && ofs->config.upperdir)) && !fh_type) { |
02bcd157 | 805 | ofs->config.index = false; |
f168f109 AG |
806 | ofs->config.nfs_export = false; |
807 | pr_warn("overlayfs: fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n", | |
808 | name); | |
02bcd157 AG |
809 | } |
810 | ||
e487d889 AG |
811 | /* Check if lower fs has 32bit inode numbers */ |
812 | if (fh_type != FILEID_INO32_GEN) | |
813 | ofs->xino_bits = 0; | |
814 | ||
ab508822 MS |
815 | return 0; |
816 | ||
817 | out_put: | |
8aafcb59 | 818 | path_put_init(path); |
ab508822 MS |
819 | out: |
820 | return err; | |
821 | } | |
822 | ||
e9be9d5e MS |
823 | /* Workdir should not be subdir of upperdir and vice versa */ |
824 | static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir) | |
825 | { | |
826 | bool ok = false; | |
827 | ||
828 | if (workdir != upperdir) { | |
829 | ok = (lock_rename(workdir, upperdir) == NULL); | |
830 | unlock_rename(workdir, upperdir); | |
831 | } | |
832 | return ok; | |
833 | } | |
834 | ||
a78d9f0d MS |
835 | static unsigned int ovl_split_lowerdirs(char *str) |
836 | { | |
837 | unsigned int ctr = 1; | |
838 | char *s, *d; | |
839 | ||
840 | for (s = d = str;; s++, d++) { | |
841 | if (*s == '\\') { | |
842 | s++; | |
843 | } else if (*s == ':') { | |
844 | *d = '\0'; | |
845 | ctr++; | |
846 | continue; | |
847 | } | |
848 | *d = *s; | |
849 | if (!*s) | |
850 | break; | |
851 | } | |
852 | return ctr; | |
853 | } | |
854 | ||
0eb45fc3 AG |
855 | static int __maybe_unused |
856 | ovl_posix_acl_xattr_get(const struct xattr_handler *handler, | |
857 | struct dentry *dentry, struct inode *inode, | |
858 | const char *name, void *buffer, size_t size) | |
859 | { | |
1d88f183 | 860 | return ovl_xattr_get(dentry, inode, handler->name, buffer, size); |
0eb45fc3 AG |
861 | } |
862 | ||
0c97be22 AG |
863 | static int __maybe_unused |
864 | ovl_posix_acl_xattr_set(const struct xattr_handler *handler, | |
865 | struct dentry *dentry, struct inode *inode, | |
866 | const char *name, const void *value, | |
867 | size_t size, int flags) | |
d837a49b MS |
868 | { |
869 | struct dentry *workdir = ovl_workdir(dentry); | |
09d8b586 | 870 | struct inode *realinode = ovl_inode_real(inode); |
d837a49b MS |
871 | struct posix_acl *acl = NULL; |
872 | int err; | |
873 | ||
874 | /* Check that everything is OK before copy-up */ | |
875 | if (value) { | |
876 | acl = posix_acl_from_xattr(&init_user_ns, value, size); | |
877 | if (IS_ERR(acl)) | |
878 | return PTR_ERR(acl); | |
879 | } | |
880 | err = -EOPNOTSUPP; | |
881 | if (!IS_POSIXACL(d_inode(workdir))) | |
882 | goto out_acl_release; | |
883 | if (!realinode->i_op->set_acl) | |
884 | goto out_acl_release; | |
885 | if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) { | |
886 | err = acl ? -EACCES : 0; | |
887 | goto out_acl_release; | |
888 | } | |
889 | err = -EPERM; | |
890 | if (!inode_owner_or_capable(inode)) | |
891 | goto out_acl_release; | |
892 | ||
893 | posix_acl_release(acl); | |
894 | ||
fd3220d3 MS |
895 | /* |
896 | * Check if sgid bit needs to be cleared (actual setacl operation will | |
897 | * be done with mounter's capabilities and so that won't do it for us). | |
898 | */ | |
899 | if (unlikely(inode->i_mode & S_ISGID) && | |
900 | handler->flags == ACL_TYPE_ACCESS && | |
901 | !in_group_p(inode->i_gid) && | |
902 | !capable_wrt_inode_uidgid(inode, CAP_FSETID)) { | |
903 | struct iattr iattr = { .ia_valid = ATTR_KILL_SGID }; | |
904 | ||
905 | err = ovl_setattr(dentry, &iattr); | |
906 | if (err) | |
907 | return err; | |
908 | } | |
909 | ||
1d88f183 | 910 | err = ovl_xattr_set(dentry, inode, handler->name, value, size, flags); |
ce31513a | 911 | if (!err) |
09d8b586 | 912 | ovl_copyattr(ovl_inode_real(inode), inode); |
ce31513a MS |
913 | |
914 | return err; | |
d837a49b MS |
915 | |
916 | out_acl_release: | |
917 | posix_acl_release(acl); | |
918 | return err; | |
919 | } | |
920 | ||
0eb45fc3 AG |
921 | static int ovl_own_xattr_get(const struct xattr_handler *handler, |
922 | struct dentry *dentry, struct inode *inode, | |
923 | const char *name, void *buffer, size_t size) | |
924 | { | |
48fab5d7 | 925 | return -EOPNOTSUPP; |
0eb45fc3 AG |
926 | } |
927 | ||
d837a49b MS |
928 | static int ovl_own_xattr_set(const struct xattr_handler *handler, |
929 | struct dentry *dentry, struct inode *inode, | |
930 | const char *name, const void *value, | |
931 | size_t size, int flags) | |
932 | { | |
48fab5d7 | 933 | return -EOPNOTSUPP; |
d837a49b MS |
934 | } |
935 | ||
0eb45fc3 AG |
936 | static int ovl_other_xattr_get(const struct xattr_handler *handler, |
937 | struct dentry *dentry, struct inode *inode, | |
938 | const char *name, void *buffer, size_t size) | |
939 | { | |
1d88f183 | 940 | return ovl_xattr_get(dentry, inode, name, buffer, size); |
0eb45fc3 AG |
941 | } |
942 | ||
0e585ccc AG |
943 | static int ovl_other_xattr_set(const struct xattr_handler *handler, |
944 | struct dentry *dentry, struct inode *inode, | |
945 | const char *name, const void *value, | |
946 | size_t size, int flags) | |
947 | { | |
1d88f183 | 948 | return ovl_xattr_set(dentry, inode, name, value, size, flags); |
0e585ccc AG |
949 | } |
950 | ||
0c97be22 AG |
951 | static const struct xattr_handler __maybe_unused |
952 | ovl_posix_acl_access_xattr_handler = { | |
d837a49b MS |
953 | .name = XATTR_NAME_POSIX_ACL_ACCESS, |
954 | .flags = ACL_TYPE_ACCESS, | |
0eb45fc3 | 955 | .get = ovl_posix_acl_xattr_get, |
d837a49b MS |
956 | .set = ovl_posix_acl_xattr_set, |
957 | }; | |
958 | ||
0c97be22 AG |
959 | static const struct xattr_handler __maybe_unused |
960 | ovl_posix_acl_default_xattr_handler = { | |
d837a49b MS |
961 | .name = XATTR_NAME_POSIX_ACL_DEFAULT, |
962 | .flags = ACL_TYPE_DEFAULT, | |
0eb45fc3 | 963 | .get = ovl_posix_acl_xattr_get, |
d837a49b MS |
964 | .set = ovl_posix_acl_xattr_set, |
965 | }; | |
966 | ||
967 | static const struct xattr_handler ovl_own_xattr_handler = { | |
968 | .prefix = OVL_XATTR_PREFIX, | |
0eb45fc3 | 969 | .get = ovl_own_xattr_get, |
d837a49b MS |
970 | .set = ovl_own_xattr_set, |
971 | }; | |
972 | ||
973 | static const struct xattr_handler ovl_other_xattr_handler = { | |
974 | .prefix = "", /* catch all */ | |
0eb45fc3 | 975 | .get = ovl_other_xattr_get, |
d837a49b MS |
976 | .set = ovl_other_xattr_set, |
977 | }; | |
978 | ||
979 | static const struct xattr_handler *ovl_xattr_handlers[] = { | |
0c97be22 | 980 | #ifdef CONFIG_FS_POSIX_ACL |
d837a49b MS |
981 | &ovl_posix_acl_access_xattr_handler, |
982 | &ovl_posix_acl_default_xattr_handler, | |
0c97be22 | 983 | #endif |
d837a49b MS |
984 | &ovl_own_xattr_handler, |
985 | &ovl_other_xattr_handler, | |
986 | NULL | |
987 | }; | |
988 | ||
146d62e5 AG |
989 | static int ovl_setup_trap(struct super_block *sb, struct dentry *dir, |
990 | struct inode **ptrap, const char *name) | |
991 | { | |
992 | struct inode *trap; | |
993 | int err; | |
994 | ||
995 | trap = ovl_get_trap_inode(sb, dir); | |
1dac6f5b AB |
996 | err = PTR_ERR_OR_ZERO(trap); |
997 | if (err) { | |
146d62e5 AG |
998 | if (err == -ELOOP) |
999 | pr_err("overlayfs: conflicting %s path\n", name); | |
1000 | return err; | |
1001 | } | |
1002 | ||
1003 | *ptrap = trap; | |
1004 | return 0; | |
1005 | } | |
1006 | ||
0be0bfd2 AG |
1007 | /* |
1008 | * Determine how we treat concurrent use of upperdir/workdir based on the | |
1009 | * index feature. This is papering over mount leaks of container runtimes, | |
1010 | * for example, an old overlay mount is leaked and now its upperdir is | |
1011 | * attempted to be used as a lower layer in a new overlay mount. | |
1012 | */ | |
1013 | static int ovl_report_in_use(struct ovl_fs *ofs, const char *name) | |
1014 | { | |
1015 | if (ofs->config.index) { | |
1016 | pr_err("overlayfs: %s is in-use as upperdir/workdir of another mount, mount with '-o index=off' to override exclusive upperdir protection.\n", | |
1017 | name); | |
1018 | return -EBUSY; | |
1019 | } else { | |
1020 | pr_warn("overlayfs: %s is in-use as upperdir/workdir of another mount, accessing files from both mounts will result in undefined behavior.\n", | |
1021 | name); | |
1022 | return 0; | |
1023 | } | |
1024 | } | |
1025 | ||
146d62e5 AG |
1026 | static int ovl_get_upper(struct super_block *sb, struct ovl_fs *ofs, |
1027 | struct path *upperpath) | |
6ee8acf0 | 1028 | { |
5064975e | 1029 | struct vfsmount *upper_mnt; |
6ee8acf0 MS |
1030 | int err; |
1031 | ||
ad204488 | 1032 | err = ovl_mount_dir(ofs->config.upperdir, upperpath); |
6ee8acf0 MS |
1033 | if (err) |
1034 | goto out; | |
1035 | ||
1036 | /* Upper fs should not be r/o */ | |
1037 | if (sb_rdonly(upperpath->mnt->mnt_sb)) { | |
1038 | pr_err("overlayfs: upper fs is r/o, try multi-lower layers mount\n"); | |
1039 | err = -EINVAL; | |
1040 | goto out; | |
1041 | } | |
1042 | ||
ad204488 | 1043 | err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir); |
6ee8acf0 MS |
1044 | if (err) |
1045 | goto out; | |
1046 | ||
146d62e5 AG |
1047 | err = ovl_setup_trap(sb, upperpath->dentry, &ofs->upperdir_trap, |
1048 | "upperdir"); | |
1049 | if (err) | |
1050 | goto out; | |
1051 | ||
5064975e MS |
1052 | upper_mnt = clone_private_mount(upperpath); |
1053 | err = PTR_ERR(upper_mnt); | |
1054 | if (IS_ERR(upper_mnt)) { | |
1055 | pr_err("overlayfs: failed to clone upperpath\n"); | |
1056 | goto out; | |
1057 | } | |
1058 | ||
1059 | /* Don't inherit atime flags */ | |
1060 | upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME); | |
ad204488 | 1061 | ofs->upper_mnt = upper_mnt; |
8c25741a | 1062 | |
8c25741a MS |
1063 | if (ovl_inuse_trylock(ofs->upper_mnt->mnt_root)) { |
1064 | ofs->upperdir_locked = true; | |
8c25741a | 1065 | } else { |
0be0bfd2 AG |
1066 | err = ovl_report_in_use(ofs, "upperdir"); |
1067 | if (err) | |
1068 | goto out; | |
8c25741a MS |
1069 | } |
1070 | ||
6ee8acf0 MS |
1071 | err = 0; |
1072 | out: | |
1073 | return err; | |
1074 | } | |
1075 | ||
146d62e5 AG |
1076 | static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs, |
1077 | struct path *workpath) | |
8ed61dc3 | 1078 | { |
2ba9d57e | 1079 | struct vfsmount *mnt = ofs->upper_mnt; |
8ed61dc3 | 1080 | struct dentry *temp; |
e487d889 | 1081 | int fh_type; |
8ed61dc3 MS |
1082 | int err; |
1083 | ||
2ba9d57e AG |
1084 | err = mnt_want_write(mnt); |
1085 | if (err) | |
1086 | return err; | |
1087 | ||
ad204488 MS |
1088 | ofs->workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false); |
1089 | if (!ofs->workdir) | |
2ba9d57e | 1090 | goto out; |
8ed61dc3 | 1091 | |
146d62e5 AG |
1092 | err = ovl_setup_trap(sb, ofs->workdir, &ofs->workdir_trap, "workdir"); |
1093 | if (err) | |
1094 | goto out; | |
1095 | ||
8ed61dc3 MS |
1096 | /* |
1097 | * Upper should support d_type, else whiteouts are visible. Given | |
1098 | * workdir and upper are on same fs, we can do iterate_dir() on | |
1099 | * workdir. This check requires successful creation of workdir in | |
1100 | * previous step. | |
1101 | */ | |
1102 | err = ovl_check_d_type_supported(workpath); | |
1103 | if (err < 0) | |
2ba9d57e | 1104 | goto out; |
8ed61dc3 MS |
1105 | |
1106 | /* | |
1107 | * We allowed this configuration and don't want to break users over | |
1108 | * kernel upgrade. So warn instead of erroring out. | |
1109 | */ | |
1110 | if (!err) | |
1111 | pr_warn("overlayfs: upper fs needs to support d_type.\n"); | |
1112 | ||
1113 | /* Check if upper/work fs supports O_TMPFILE */ | |
ad204488 MS |
1114 | temp = ovl_do_tmpfile(ofs->workdir, S_IFREG | 0); |
1115 | ofs->tmpfile = !IS_ERR(temp); | |
1116 | if (ofs->tmpfile) | |
8ed61dc3 MS |
1117 | dput(temp); |
1118 | else | |
1119 | pr_warn("overlayfs: upper fs does not support tmpfile.\n"); | |
1120 | ||
1121 | /* | |
1122 | * Check if upper/work fs supports trusted.overlay.* xattr | |
1123 | */ | |
ad204488 | 1124 | err = ovl_do_setxattr(ofs->workdir, OVL_XATTR_OPAQUE, "0", 1, 0); |
8ed61dc3 | 1125 | if (err) { |
ad204488 | 1126 | ofs->noxattr = true; |
a683737b | 1127 | ofs->config.index = false; |
d5791044 VG |
1128 | ofs->config.metacopy = false; |
1129 | pr_warn("overlayfs: upper fs does not support xattr, falling back to index=off and metacopy=off.\n"); | |
2ba9d57e | 1130 | err = 0; |
8ed61dc3 | 1131 | } else { |
ad204488 | 1132 | vfs_removexattr(ofs->workdir, OVL_XATTR_OPAQUE); |
8ed61dc3 MS |
1133 | } |
1134 | ||
1135 | /* Check if upper/work fs supports file handles */ | |
e487d889 AG |
1136 | fh_type = ovl_can_decode_fh(ofs->workdir->d_sb); |
1137 | if (ofs->config.index && !fh_type) { | |
ad204488 | 1138 | ofs->config.index = false; |
8ed61dc3 MS |
1139 | pr_warn("overlayfs: upper fs does not support file handles, falling back to index=off.\n"); |
1140 | } | |
1141 | ||
e487d889 AG |
1142 | /* Check if upper fs has 32bit inode numbers */ |
1143 | if (fh_type != FILEID_INO32_GEN) | |
1144 | ofs->xino_bits = 0; | |
1145 | ||
f168f109 AG |
1146 | /* NFS export of r/w mount depends on index */ |
1147 | if (ofs->config.nfs_export && !ofs->config.index) { | |
1148 | pr_warn("overlayfs: NFS export requires \"index=on\", falling back to nfs_export=off.\n"); | |
1149 | ofs->config.nfs_export = false; | |
1150 | } | |
2ba9d57e AG |
1151 | out: |
1152 | mnt_drop_write(mnt); | |
1153 | return err; | |
8ed61dc3 MS |
1154 | } |
1155 | ||
146d62e5 AG |
1156 | static int ovl_get_workdir(struct super_block *sb, struct ovl_fs *ofs, |
1157 | struct path *upperpath) | |
520d7c86 MS |
1158 | { |
1159 | int err; | |
bca44b52 | 1160 | struct path workpath = { }; |
520d7c86 | 1161 | |
ad204488 | 1162 | err = ovl_mount_dir(ofs->config.workdir, &workpath); |
520d7c86 MS |
1163 | if (err) |
1164 | goto out; | |
1165 | ||
1166 | err = -EINVAL; | |
bca44b52 | 1167 | if (upperpath->mnt != workpath.mnt) { |
520d7c86 MS |
1168 | pr_err("overlayfs: workdir and upperdir must reside under the same mount\n"); |
1169 | goto out; | |
1170 | } | |
bca44b52 | 1171 | if (!ovl_workdir_ok(workpath.dentry, upperpath->dentry)) { |
520d7c86 MS |
1172 | pr_err("overlayfs: workdir and upperdir must be separate subtrees\n"); |
1173 | goto out; | |
1174 | } | |
1175 | ||
8c25741a MS |
1176 | ofs->workbasedir = dget(workpath.dentry); |
1177 | ||
8c25741a | 1178 | if (ovl_inuse_trylock(ofs->workbasedir)) { |
ad204488 | 1179 | ofs->workdir_locked = true; |
520d7c86 | 1180 | } else { |
0be0bfd2 AG |
1181 | err = ovl_report_in_use(ofs, "workdir"); |
1182 | if (err) | |
1183 | goto out; | |
520d7c86 MS |
1184 | } |
1185 | ||
0be0bfd2 AG |
1186 | err = ovl_setup_trap(sb, ofs->workbasedir, &ofs->workbasedir_trap, |
1187 | "workdir"); | |
1188 | if (err) | |
1189 | goto out; | |
1190 | ||
146d62e5 | 1191 | err = ovl_make_workdir(sb, ofs, &workpath); |
bca44b52 | 1192 | |
520d7c86 | 1193 | out: |
bca44b52 MS |
1194 | path_put(&workpath); |
1195 | ||
520d7c86 MS |
1196 | return err; |
1197 | } | |
1198 | ||
146d62e5 AG |
1199 | static int ovl_get_indexdir(struct super_block *sb, struct ovl_fs *ofs, |
1200 | struct ovl_entry *oe, struct path *upperpath) | |
f7e3a7d9 | 1201 | { |
2ba9d57e | 1202 | struct vfsmount *mnt = ofs->upper_mnt; |
f7e3a7d9 MS |
1203 | int err; |
1204 | ||
2ba9d57e AG |
1205 | err = mnt_want_write(mnt); |
1206 | if (err) | |
1207 | return err; | |
1208 | ||
f7e3a7d9 | 1209 | /* Verify lower root is upper root origin */ |
d9768076 | 1210 | err = ovl_verify_origin(upperpath->dentry, oe->lowerstack[0].dentry, |
05122443 | 1211 | true); |
f7e3a7d9 MS |
1212 | if (err) { |
1213 | pr_err("overlayfs: failed to verify upper root origin\n"); | |
1214 | goto out; | |
1215 | } | |
1216 | ||
ad204488 MS |
1217 | ofs->indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true); |
1218 | if (ofs->indexdir) { | |
146d62e5 AG |
1219 | err = ovl_setup_trap(sb, ofs->indexdir, &ofs->indexdir_trap, |
1220 | "indexdir"); | |
1221 | if (err) | |
1222 | goto out; | |
1223 | ||
ad1d615c AG |
1224 | /* |
1225 | * Verify upper root is exclusively associated with index dir. | |
1226 | * Older kernels stored upper fh in "trusted.overlay.origin" | |
1227 | * xattr. If that xattr exists, verify that it is a match to | |
1228 | * upper dir file handle. In any case, verify or set xattr | |
1229 | * "trusted.overlay.upper" to indicate that index may have | |
1230 | * directory entries. | |
1231 | */ | |
1232 | if (ovl_check_origin_xattr(ofs->indexdir)) { | |
1233 | err = ovl_verify_set_fh(ofs->indexdir, OVL_XATTR_ORIGIN, | |
1234 | upperpath->dentry, true, false); | |
1235 | if (err) | |
1236 | pr_err("overlayfs: failed to verify index dir 'origin' xattr\n"); | |
1237 | } | |
1238 | err = ovl_verify_upper(ofs->indexdir, upperpath->dentry, true); | |
f7e3a7d9 | 1239 | if (err) |
ad1d615c | 1240 | pr_err("overlayfs: failed to verify index dir 'upper' xattr\n"); |
f7e3a7d9 MS |
1241 | |
1242 | /* Cleanup bad/stale/orphan index entries */ | |
1243 | if (!err) | |
1eff1a1d | 1244 | err = ovl_indexdir_cleanup(ofs); |
f7e3a7d9 | 1245 | } |
ad204488 | 1246 | if (err || !ofs->indexdir) |
f7e3a7d9 MS |
1247 | pr_warn("overlayfs: try deleting index dir or mounting with '-o index=off' to disable inodes index.\n"); |
1248 | ||
1249 | out: | |
2ba9d57e | 1250 | mnt_drop_write(mnt); |
f7e3a7d9 MS |
1251 | return err; |
1252 | } | |
1253 | ||
9df085f3 AG |
1254 | static bool ovl_lower_uuid_ok(struct ovl_fs *ofs, const uuid_t *uuid) |
1255 | { | |
1256 | unsigned int i; | |
1257 | ||
1258 | if (!ofs->config.nfs_export && !(ofs->config.index && ofs->upper_mnt)) | |
1259 | return true; | |
1260 | ||
1261 | for (i = 0; i < ofs->numlowerfs; i++) { | |
1262 | /* | |
1263 | * We use uuid to associate an overlay lower file handle with a | |
1264 | * lower layer, so we can accept lower fs with null uuid as long | |
1265 | * as all lower layers with null uuid are on the same fs. | |
1266 | */ | |
1267 | if (uuid_equal(&ofs->lower_fs[i].sb->s_uuid, uuid)) | |
1268 | return false; | |
1269 | } | |
1270 | return true; | |
1271 | } | |
1272 | ||
5148626b | 1273 | /* Get a unique fsid for the layer */ |
9df085f3 | 1274 | static int ovl_get_fsid(struct ovl_fs *ofs, const struct path *path) |
5148626b | 1275 | { |
9df085f3 | 1276 | struct super_block *sb = path->mnt->mnt_sb; |
5148626b AG |
1277 | unsigned int i; |
1278 | dev_t dev; | |
1279 | int err; | |
1280 | ||
1281 | /* fsid 0 is reserved for upper fs even with non upper overlay */ | |
1282 | if (ofs->upper_mnt && ofs->upper_mnt->mnt_sb == sb) | |
1283 | return 0; | |
1284 | ||
1285 | for (i = 0; i < ofs->numlowerfs; i++) { | |
1286 | if (ofs->lower_fs[i].sb == sb) | |
1287 | return i + 1; | |
1288 | } | |
1289 | ||
9df085f3 AG |
1290 | if (!ovl_lower_uuid_ok(ofs, &sb->s_uuid)) { |
1291 | ofs->config.index = false; | |
1292 | ofs->config.nfs_export = false; | |
1293 | pr_warn("overlayfs: %s uuid detected in lower fs '%pd2', falling back to index=off,nfs_export=off.\n", | |
1294 | uuid_is_null(&sb->s_uuid) ? "null" : "conflicting", | |
1295 | path->dentry); | |
1296 | } | |
1297 | ||
5148626b AG |
1298 | err = get_anon_bdev(&dev); |
1299 | if (err) { | |
1300 | pr_err("overlayfs: failed to get anonymous bdev for lowerpath\n"); | |
1301 | return err; | |
1302 | } | |
1303 | ||
1304 | ofs->lower_fs[ofs->numlowerfs].sb = sb; | |
1305 | ofs->lower_fs[ofs->numlowerfs].pseudo_dev = dev; | |
1306 | ofs->numlowerfs++; | |
1307 | ||
1308 | return ofs->numlowerfs; | |
1309 | } | |
1310 | ||
146d62e5 AG |
1311 | static int ovl_get_lower_layers(struct super_block *sb, struct ovl_fs *ofs, |
1312 | struct path *stack, unsigned int numlower) | |
520d7c86 MS |
1313 | { |
1314 | int err; | |
1315 | unsigned int i; | |
1316 | ||
1317 | err = -ENOMEM; | |
ad204488 | 1318 | ofs->lower_layers = kcalloc(numlower, sizeof(struct ovl_layer), |
520d7c86 | 1319 | GFP_KERNEL); |
ad204488 | 1320 | if (ofs->lower_layers == NULL) |
520d7c86 | 1321 | goto out; |
5148626b AG |
1322 | |
1323 | ofs->lower_fs = kcalloc(numlower, sizeof(struct ovl_sb), | |
1324 | GFP_KERNEL); | |
1325 | if (ofs->lower_fs == NULL) | |
1326 | goto out; | |
1327 | ||
520d7c86 MS |
1328 | for (i = 0; i < numlower; i++) { |
1329 | struct vfsmount *mnt; | |
146d62e5 | 1330 | struct inode *trap; |
5148626b | 1331 | int fsid; |
520d7c86 | 1332 | |
9df085f3 | 1333 | err = fsid = ovl_get_fsid(ofs, &stack[i]); |
5148626b | 1334 | if (err < 0) |
520d7c86 | 1335 | goto out; |
520d7c86 | 1336 | |
146d62e5 AG |
1337 | err = ovl_setup_trap(sb, stack[i].dentry, &trap, "lowerdir"); |
1338 | if (err) | |
1339 | goto out; | |
1340 | ||
0be0bfd2 AG |
1341 | if (ovl_is_inuse(stack[i].dentry)) { |
1342 | err = ovl_report_in_use(ofs, "lowerdir"); | |
1343 | if (err) | |
1344 | goto out; | |
1345 | } | |
1346 | ||
520d7c86 MS |
1347 | mnt = clone_private_mount(&stack[i]); |
1348 | err = PTR_ERR(mnt); | |
1349 | if (IS_ERR(mnt)) { | |
1350 | pr_err("overlayfs: failed to clone lowerpath\n"); | |
146d62e5 | 1351 | iput(trap); |
520d7c86 MS |
1352 | goto out; |
1353 | } | |
5148626b | 1354 | |
520d7c86 MS |
1355 | /* |
1356 | * Make lower layers R/O. That way fchmod/fchown on lower file | |
1357 | * will fail instead of modifying lower fs. | |
1358 | */ | |
1359 | mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME; | |
1360 | ||
146d62e5 | 1361 | ofs->lower_layers[ofs->numlower].trap = trap; |
ad204488 | 1362 | ofs->lower_layers[ofs->numlower].mnt = mnt; |
d583ed7d | 1363 | ofs->lower_layers[ofs->numlower].idx = i + 1; |
5148626b AG |
1364 | ofs->lower_layers[ofs->numlower].fsid = fsid; |
1365 | if (fsid) { | |
1366 | ofs->lower_layers[ofs->numlower].fs = | |
1367 | &ofs->lower_fs[fsid - 1]; | |
1368 | } | |
ad204488 | 1369 | ofs->numlower++; |
520d7c86 | 1370 | } |
e487d889 | 1371 | |
795939a9 AG |
1372 | /* |
1373 | * When all layers on same fs, overlay can use real inode numbers. | |
1374 | * With mount option "xino=on", mounter declares that there are enough | |
1375 | * free high bits in underlying fs to hold the unique fsid. | |
1376 | * If overlayfs does encounter underlying inodes using the high xino | |
1377 | * bits reserved for fsid, it emits a warning and uses the original | |
1378 | * inode number. | |
1379 | */ | |
1380 | if (!ofs->numlowerfs || (ofs->numlowerfs == 1 && !ofs->upper_mnt)) { | |
e487d889 | 1381 | ofs->xino_bits = 0; |
795939a9 AG |
1382 | ofs->config.xino = OVL_XINO_OFF; |
1383 | } else if (ofs->config.xino == OVL_XINO_ON && !ofs->xino_bits) { | |
1384 | /* | |
1385 | * This is a roundup of number of bits needed for numlowerfs+1 | |
1386 | * (i.e. ilog2(numlowerfs+1 - 1) + 1). fsid 0 is reserved for | |
1387 | * upper fs even with non upper overlay. | |
1388 | */ | |
1389 | BUILD_BUG_ON(ilog2(OVL_MAX_STACK) > 31); | |
1390 | ofs->xino_bits = ilog2(ofs->numlowerfs) + 1; | |
1391 | } | |
1392 | ||
1393 | if (ofs->xino_bits) { | |
1394 | pr_info("overlayfs: \"xino\" feature enabled using %d upper inode bits.\n", | |
1395 | ofs->xino_bits); | |
1396 | } | |
e487d889 | 1397 | |
520d7c86 MS |
1398 | err = 0; |
1399 | out: | |
1400 | return err; | |
1401 | } | |
1402 | ||
4155c10a | 1403 | static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb, |
ad204488 | 1404 | struct ovl_fs *ofs) |
53dbb0b4 MS |
1405 | { |
1406 | int err; | |
1407 | char *lowertmp, *lower; | |
4155c10a MS |
1408 | struct path *stack = NULL; |
1409 | unsigned int stacklen, numlower = 0, i; | |
53dbb0b4 | 1410 | bool remote = false; |
4155c10a | 1411 | struct ovl_entry *oe; |
53dbb0b4 MS |
1412 | |
1413 | err = -ENOMEM; | |
ad204488 | 1414 | lowertmp = kstrdup(ofs->config.lowerdir, GFP_KERNEL); |
53dbb0b4 | 1415 | if (!lowertmp) |
4155c10a | 1416 | goto out_err; |
53dbb0b4 MS |
1417 | |
1418 | err = -EINVAL; | |
1419 | stacklen = ovl_split_lowerdirs(lowertmp); | |
1420 | if (stacklen > OVL_MAX_STACK) { | |
1421 | pr_err("overlayfs: too many lower directories, limit is %d\n", | |
1422 | OVL_MAX_STACK); | |
4155c10a | 1423 | goto out_err; |
ad204488 | 1424 | } else if (!ofs->config.upperdir && stacklen == 1) { |
53dbb0b4 | 1425 | pr_err("overlayfs: at least 2 lowerdir are needed while upperdir nonexistent\n"); |
4155c10a | 1426 | goto out_err; |
f168f109 AG |
1427 | } else if (!ofs->config.upperdir && ofs->config.nfs_export && |
1428 | ofs->config.redirect_follow) { | |
1429 | pr_warn("overlayfs: NFS export requires \"redirect_dir=nofollow\" on non-upper mount, falling back to nfs_export=off.\n"); | |
1430 | ofs->config.nfs_export = false; | |
53dbb0b4 MS |
1431 | } |
1432 | ||
1433 | err = -ENOMEM; | |
1434 | stack = kcalloc(stacklen, sizeof(struct path), GFP_KERNEL); | |
1435 | if (!stack) | |
4155c10a | 1436 | goto out_err; |
53dbb0b4 MS |
1437 | |
1438 | err = -EINVAL; | |
1439 | lower = lowertmp; | |
1440 | for (numlower = 0; numlower < stacklen; numlower++) { | |
ad204488 | 1441 | err = ovl_lower_dir(lower, &stack[numlower], ofs, |
53dbb0b4 MS |
1442 | &sb->s_stack_depth, &remote); |
1443 | if (err) | |
4155c10a | 1444 | goto out_err; |
53dbb0b4 MS |
1445 | |
1446 | lower = strchr(lower, '\0') + 1; | |
1447 | } | |
1448 | ||
1449 | err = -EINVAL; | |
1450 | sb->s_stack_depth++; | |
1451 | if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) { | |
1452 | pr_err("overlayfs: maximum fs stacking depth exceeded\n"); | |
4155c10a | 1453 | goto out_err; |
53dbb0b4 MS |
1454 | } |
1455 | ||
146d62e5 | 1456 | err = ovl_get_lower_layers(sb, ofs, stack, numlower); |
4155c10a MS |
1457 | if (err) |
1458 | goto out_err; | |
1459 | ||
1460 | err = -ENOMEM; | |
1461 | oe = ovl_alloc_entry(numlower); | |
1462 | if (!oe) | |
1463 | goto out_err; | |
1464 | ||
1465 | for (i = 0; i < numlower; i++) { | |
1466 | oe->lowerstack[i].dentry = dget(stack[i].dentry); | |
ad204488 | 1467 | oe->lowerstack[i].layer = &ofs->lower_layers[i]; |
4155c10a | 1468 | } |
53dbb0b4 MS |
1469 | |
1470 | if (remote) | |
1471 | sb->s_d_op = &ovl_reval_dentry_operations; | |
1472 | else | |
1473 | sb->s_d_op = &ovl_dentry_operations; | |
1474 | ||
53dbb0b4 | 1475 | out: |
53dbb0b4 MS |
1476 | for (i = 0; i < numlower; i++) |
1477 | path_put(&stack[i]); | |
1478 | kfree(stack); | |
4155c10a MS |
1479 | kfree(lowertmp); |
1480 | ||
1481 | return oe; | |
1482 | ||
1483 | out_err: | |
1484 | oe = ERR_PTR(err); | |
53dbb0b4 MS |
1485 | goto out; |
1486 | } | |
1487 | ||
146d62e5 AG |
1488 | /* |
1489 | * Check if this layer root is a descendant of: | |
1490 | * - another layer of this overlayfs instance | |
1491 | * - upper/work dir of any overlayfs instance | |
146d62e5 | 1492 | */ |
0be0bfd2 AG |
1493 | static int ovl_check_layer(struct super_block *sb, struct ovl_fs *ofs, |
1494 | struct dentry *dentry, const char *name) | |
146d62e5 | 1495 | { |
9179c21d | 1496 | struct dentry *next = dentry, *parent; |
146d62e5 AG |
1497 | int err = 0; |
1498 | ||
9179c21d | 1499 | if (!dentry) |
146d62e5 AG |
1500 | return 0; |
1501 | ||
9179c21d MS |
1502 | parent = dget_parent(next); |
1503 | ||
1504 | /* Walk back ancestors to root (inclusive) looking for traps */ | |
1505 | while (!err && parent != next) { | |
0be0bfd2 | 1506 | if (ovl_lookup_trap_inode(sb, parent)) { |
146d62e5 AG |
1507 | err = -ELOOP; |
1508 | pr_err("overlayfs: overlapping %s path\n", name); | |
0be0bfd2 AG |
1509 | } else if (ovl_is_inuse(parent)) { |
1510 | err = ovl_report_in_use(ofs, name); | |
146d62e5 | 1511 | } |
146d62e5 | 1512 | next = parent; |
9179c21d MS |
1513 | parent = dget_parent(next); |
1514 | dput(next); | |
146d62e5 AG |
1515 | } |
1516 | ||
9179c21d | 1517 | dput(parent); |
146d62e5 AG |
1518 | |
1519 | return err; | |
1520 | } | |
1521 | ||
1522 | /* | |
1523 | * Check if any of the layers or work dirs overlap. | |
1524 | */ | |
1525 | static int ovl_check_overlapping_layers(struct super_block *sb, | |
1526 | struct ovl_fs *ofs) | |
1527 | { | |
1528 | int i, err; | |
1529 | ||
1530 | if (ofs->upper_mnt) { | |
0be0bfd2 AG |
1531 | err = ovl_check_layer(sb, ofs, ofs->upper_mnt->mnt_root, |
1532 | "upperdir"); | |
146d62e5 AG |
1533 | if (err) |
1534 | return err; | |
1535 | ||
1536 | /* | |
1537 | * Checking workbasedir avoids hitting ovl_is_inuse(parent) of | |
1538 | * this instance and covers overlapping work and index dirs, | |
1539 | * unless work or index dir have been moved since created inside | |
1540 | * workbasedir. In that case, we already have their traps in | |
1541 | * inode cache and we will catch that case on lookup. | |
1542 | */ | |
0be0bfd2 | 1543 | err = ovl_check_layer(sb, ofs, ofs->workbasedir, "workdir"); |
146d62e5 AG |
1544 | if (err) |
1545 | return err; | |
1546 | } | |
1547 | ||
1548 | for (i = 0; i < ofs->numlower; i++) { | |
0be0bfd2 AG |
1549 | err = ovl_check_layer(sb, ofs, |
1550 | ofs->lower_layers[i].mnt->mnt_root, | |
146d62e5 AG |
1551 | "lowerdir"); |
1552 | if (err) | |
1553 | return err; | |
1554 | } | |
1555 | ||
1556 | return 0; | |
1557 | } | |
1558 | ||
e9be9d5e MS |
1559 | static int ovl_fill_super(struct super_block *sb, void *data, int silent) |
1560 | { | |
33006cdf | 1561 | struct path upperpath = { }; |
e9be9d5e | 1562 | struct dentry *root_dentry; |
4155c10a | 1563 | struct ovl_entry *oe; |
ad204488 | 1564 | struct ovl_fs *ofs; |
51f8f3c4 | 1565 | struct cred *cred; |
e9be9d5e MS |
1566 | int err; |
1567 | ||
f45827e8 | 1568 | err = -ENOMEM; |
ad204488 MS |
1569 | ofs = kzalloc(sizeof(struct ovl_fs), GFP_KERNEL); |
1570 | if (!ofs) | |
e9be9d5e MS |
1571 | goto out; |
1572 | ||
ad204488 | 1573 | ofs->creator_cred = cred = prepare_creds(); |
c6fe6254 MS |
1574 | if (!cred) |
1575 | goto out_err; | |
1576 | ||
ad204488 | 1577 | ofs->config.index = ovl_index_def; |
f168f109 | 1578 | ofs->config.nfs_export = ovl_nfs_export_def; |
795939a9 | 1579 | ofs->config.xino = ovl_xino_def(); |
d5791044 | 1580 | ofs->config.metacopy = ovl_metacopy_def; |
ad204488 | 1581 | err = ovl_parse_opt((char *) data, &ofs->config); |
f45827e8 | 1582 | if (err) |
a9075cdb | 1583 | goto out_err; |
f45827e8 | 1584 | |
e9be9d5e | 1585 | err = -EINVAL; |
ad204488 | 1586 | if (!ofs->config.lowerdir) { |
07f2af7b KK |
1587 | if (!silent) |
1588 | pr_err("overlayfs: missing 'lowerdir'\n"); | |
a9075cdb | 1589 | goto out_err; |
e9be9d5e MS |
1590 | } |
1591 | ||
53a08cb9 | 1592 | sb->s_stack_depth = 0; |
cf9a6784 | 1593 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
e487d889 | 1594 | /* Assume underlaying fs uses 32bit inodes unless proven otherwise */ |
795939a9 AG |
1595 | if (ofs->config.xino != OVL_XINO_OFF) |
1596 | ofs->xino_bits = BITS_PER_LONG - 32; | |
1597 | ||
146d62e5 AG |
1598 | /* alloc/destroy_inode needed for setting up traps in inode cache */ |
1599 | sb->s_op = &ovl_super_operations; | |
1600 | ||
ad204488 MS |
1601 | if (ofs->config.upperdir) { |
1602 | if (!ofs->config.workdir) { | |
53a08cb9 | 1603 | pr_err("overlayfs: missing 'workdir'\n"); |
a9075cdb | 1604 | goto out_err; |
53a08cb9 | 1605 | } |
e9be9d5e | 1606 | |
146d62e5 | 1607 | err = ovl_get_upper(sb, ofs, &upperpath); |
53a08cb9 | 1608 | if (err) |
a9075cdb | 1609 | goto out_err; |
2cac0c00 | 1610 | |
146d62e5 | 1611 | err = ovl_get_workdir(sb, ofs, &upperpath); |
8ed61dc3 | 1612 | if (err) |
a9075cdb | 1613 | goto out_err; |
c6fe6254 | 1614 | |
ad204488 | 1615 | if (!ofs->workdir) |
1751e8a6 | 1616 | sb->s_flags |= SB_RDONLY; |
6e88256e | 1617 | |
ad204488 MS |
1618 | sb->s_stack_depth = ofs->upper_mnt->mnt_sb->s_stack_depth; |
1619 | sb->s_time_gran = ofs->upper_mnt->mnt_sb->s_time_gran; | |
c6fe6254 | 1620 | |
e9be9d5e | 1621 | } |
ad204488 | 1622 | oe = ovl_get_lowerstack(sb, ofs); |
4155c10a MS |
1623 | err = PTR_ERR(oe); |
1624 | if (IS_ERR(oe)) | |
a9075cdb | 1625 | goto out_err; |
e9be9d5e | 1626 | |
71cbad7e | 1627 | /* If the upper fs is nonexistent, we mark overlayfs r/o too */ |
ad204488 | 1628 | if (!ofs->upper_mnt) |
1751e8a6 | 1629 | sb->s_flags |= SB_RDONLY; |
e9be9d5e | 1630 | |
ad204488 | 1631 | if (!(ovl_force_readonly(ofs)) && ofs->config.index) { |
146d62e5 | 1632 | err = ovl_get_indexdir(sb, ofs, oe, &upperpath); |
54fb347e | 1633 | if (err) |
4155c10a | 1634 | goto out_free_oe; |
6e88256e | 1635 | |
972d0093 AG |
1636 | /* Force r/o mount with no index dir */ |
1637 | if (!ofs->indexdir) { | |
1638 | dput(ofs->workdir); | |
1639 | ofs->workdir = NULL; | |
1751e8a6 | 1640 | sb->s_flags |= SB_RDONLY; |
972d0093 AG |
1641 | } |
1642 | ||
02bcd157 AG |
1643 | } |
1644 | ||
146d62e5 AG |
1645 | err = ovl_check_overlapping_layers(sb, ofs); |
1646 | if (err) | |
1647 | goto out_free_oe; | |
1648 | ||
972d0093 | 1649 | /* Show index=off in /proc/mounts for forced r/o mount */ |
f168f109 | 1650 | if (!ofs->indexdir) { |
ad204488 | 1651 | ofs->config.index = false; |
f168f109 AG |
1652 | if (ofs->upper_mnt && ofs->config.nfs_export) { |
1653 | pr_warn("overlayfs: NFS export requires an index dir, falling back to nfs_export=off.\n"); | |
1654 | ofs->config.nfs_export = false; | |
1655 | } | |
1656 | } | |
02bcd157 | 1657 | |
d5791044 VG |
1658 | if (ofs->config.metacopy && ofs->config.nfs_export) { |
1659 | pr_warn("overlayfs: NFS export is not supported with metadata only copy up, falling back to nfs_export=off.\n"); | |
1660 | ofs->config.nfs_export = false; | |
1661 | } | |
1662 | ||
8383f174 AG |
1663 | if (ofs->config.nfs_export) |
1664 | sb->s_export_op = &ovl_export_operations; | |
1665 | ||
51f8f3c4 KK |
1666 | /* Never override disk quota limits or use reserved space */ |
1667 | cap_lower(cred->cap_effective, CAP_SYS_RESOURCE); | |
1668 | ||
655042cc | 1669 | sb->s_magic = OVERLAYFS_SUPER_MAGIC; |
655042cc | 1670 | sb->s_xattr = ovl_xattr_handlers; |
ad204488 | 1671 | sb->s_fs_info = ofs; |
de2a4a50 | 1672 | sb->s_flags |= SB_POSIXACL; |
655042cc | 1673 | |
c6fe6254 | 1674 | err = -ENOMEM; |
ca4c8a3a | 1675 | root_dentry = d_make_root(ovl_new_inode(sb, S_IFDIR, 0)); |
e9be9d5e | 1676 | if (!root_dentry) |
4155c10a | 1677 | goto out_free_oe; |
e9be9d5e | 1678 | |
c62520a8 AG |
1679 | root_dentry->d_fsdata = oe; |
1680 | ||
e9be9d5e | 1681 | mntput(upperpath.mnt); |
f3a15685 | 1682 | if (upperpath.dentry) { |
c62520a8 | 1683 | ovl_dentry_set_upper_alias(root_dentry); |
13c72075 MS |
1684 | if (ovl_is_impuredir(upperpath.dentry)) |
1685 | ovl_set_flag(OVL_IMPURE, d_inode(root_dentry)); | |
f3a15685 | 1686 | } |
e9be9d5e | 1687 | |
b79e05aa AG |
1688 | /* Root is always merge -> can have whiteouts */ |
1689 | ovl_set_flag(OVL_WHITEOUTS, d_inode(root_dentry)); | |
2ca3c148 | 1690 | ovl_dentry_set_flag(OVL_E_CONNECTED, root_dentry); |
0c288874 | 1691 | ovl_set_upperdata(d_inode(root_dentry)); |
09d8b586 | 1692 | ovl_inode_init(d_inode(root_dentry), upperpath.dentry, |
2664bd08 | 1693 | ovl_dentry_lower(root_dentry), NULL); |
ed06e069 | 1694 | |
e9be9d5e | 1695 | sb->s_root = root_dentry; |
e9be9d5e MS |
1696 | |
1697 | return 0; | |
1698 | ||
4155c10a MS |
1699 | out_free_oe: |
1700 | ovl_entry_stack_free(oe); | |
b9343632 | 1701 | kfree(oe); |
4155c10a | 1702 | out_err: |
e9be9d5e | 1703 | path_put(&upperpath); |
ad204488 | 1704 | ovl_free_fs(ofs); |
e9be9d5e MS |
1705 | out: |
1706 | return err; | |
1707 | } | |
1708 | ||
1709 | static struct dentry *ovl_mount(struct file_system_type *fs_type, int flags, | |
1710 | const char *dev_name, void *raw_data) | |
1711 | { | |
1712 | return mount_nodev(fs_type, flags, raw_data, ovl_fill_super); | |
1713 | } | |
1714 | ||
1715 | static struct file_system_type ovl_fs_type = { | |
1716 | .owner = THIS_MODULE, | |
ef94b186 | 1717 | .name = "overlay", |
e9be9d5e MS |
1718 | .mount = ovl_mount, |
1719 | .kill_sb = kill_anon_super, | |
1720 | }; | |
ef94b186 | 1721 | MODULE_ALIAS_FS("overlay"); |
e9be9d5e | 1722 | |
13cf199d AG |
1723 | static void ovl_inode_init_once(void *foo) |
1724 | { | |
1725 | struct ovl_inode *oi = foo; | |
1726 | ||
1727 | inode_init_once(&oi->vfs_inode); | |
1728 | } | |
1729 | ||
e9be9d5e MS |
1730 | static int __init ovl_init(void) |
1731 | { | |
13cf199d AG |
1732 | int err; |
1733 | ||
1734 | ovl_inode_cachep = kmem_cache_create("ovl_inode", | |
1735 | sizeof(struct ovl_inode), 0, | |
1736 | (SLAB_RECLAIM_ACCOUNT| | |
1737 | SLAB_MEM_SPREAD|SLAB_ACCOUNT), | |
1738 | ovl_inode_init_once); | |
1739 | if (ovl_inode_cachep == NULL) | |
1740 | return -ENOMEM; | |
1741 | ||
1742 | err = register_filesystem(&ovl_fs_type); | |
1743 | if (err) | |
1744 | kmem_cache_destroy(ovl_inode_cachep); | |
1745 | ||
1746 | return err; | |
e9be9d5e MS |
1747 | } |
1748 | ||
1749 | static void __exit ovl_exit(void) | |
1750 | { | |
1751 | unregister_filesystem(&ovl_fs_type); | |
13cf199d AG |
1752 | |
1753 | /* | |
1754 | * Make sure all delayed rcu free inodes are flushed before we | |
1755 | * destroy cache. | |
1756 | */ | |
1757 | rcu_barrier(); | |
1758 | kmem_cache_destroy(ovl_inode_cachep); | |
1759 | ||
e9be9d5e MS |
1760 | } |
1761 | ||
1762 | module_init(ovl_init); | |
1763 | module_exit(ovl_exit); |