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 | ||
3bb7df92 | 116 | static int ovl_revalidate_real(struct dentry *d, unsigned int flags, bool weak) |
7c03b5d4 | 117 | { |
7c03b5d4 MS |
118 | int ret = 1; |
119 | ||
3bb7df92 MS |
120 | if (weak) { |
121 | if (d->d_flags & DCACHE_OP_WEAK_REVALIDATE) | |
122 | ret = d->d_op->d_weak_revalidate(d, flags); | |
123 | } else if (d->d_flags & DCACHE_OP_REVALIDATE) { | |
124 | ret = d->d_op->d_revalidate(d, flags); | |
125 | if (!ret) { | |
126 | if (!(flags & LOOKUP_RCU)) | |
127 | d_invalidate(d); | |
128 | ret = -ESTALE; | |
7c03b5d4 MS |
129 | } |
130 | } | |
3bb7df92 | 131 | return ret; |
7c03b5d4 MS |
132 | } |
133 | ||
3bb7df92 MS |
134 | static int ovl_dentry_revalidate_common(struct dentry *dentry, |
135 | unsigned int flags, bool weak) | |
7c03b5d4 MS |
136 | { |
137 | struct ovl_entry *oe = dentry->d_fsdata; | |
bccece1e | 138 | struct dentry *upper; |
7c03b5d4 MS |
139 | unsigned int i; |
140 | int ret = 1; | |
141 | ||
bccece1e MS |
142 | upper = ovl_dentry_upper(dentry); |
143 | if (upper) | |
144 | ret = ovl_revalidate_real(upper, flags, weak); | |
145 | ||
3bb7df92 MS |
146 | for (i = 0; ret > 0 && i < oe->numlower; i++) { |
147 | ret = ovl_revalidate_real(oe->lowerstack[i].dentry, flags, | |
148 | weak); | |
7c03b5d4 MS |
149 | } |
150 | return ret; | |
151 | } | |
152 | ||
3bb7df92 MS |
153 | static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags) |
154 | { | |
155 | return ovl_dentry_revalidate_common(dentry, flags, false); | |
156 | } | |
157 | ||
158 | static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags) | |
159 | { | |
160 | return ovl_dentry_revalidate_common(dentry, flags, true); | |
161 | } | |
162 | ||
e9be9d5e MS |
163 | static const struct dentry_operations ovl_dentry_operations = { |
164 | .d_release = ovl_dentry_release, | |
d101a125 | 165 | .d_real = ovl_d_real, |
7c03b5d4 MS |
166 | .d_revalidate = ovl_dentry_revalidate, |
167 | .d_weak_revalidate = ovl_dentry_weak_revalidate, | |
168 | }; | |
169 | ||
13cf199d AG |
170 | static struct kmem_cache *ovl_inode_cachep; |
171 | ||
172 | static struct inode *ovl_alloc_inode(struct super_block *sb) | |
173 | { | |
174 | struct ovl_inode *oi = kmem_cache_alloc(ovl_inode_cachep, GFP_KERNEL); | |
175 | ||
b3885bd6 HN |
176 | if (!oi) |
177 | return NULL; | |
178 | ||
04a01ac7 | 179 | oi->cache = NULL; |
cf31c463 | 180 | oi->redirect = NULL; |
04a01ac7 | 181 | oi->version = 0; |
13c72075 | 182 | oi->flags = 0; |
09d8b586 | 183 | oi->__upperdentry = NULL; |
25b7713a | 184 | oi->lower = NULL; |
2664bd08 | 185 | oi->lowerdata = NULL; |
a015dafc | 186 | mutex_init(&oi->lock); |
25b7713a | 187 | |
13cf199d AG |
188 | return &oi->vfs_inode; |
189 | } | |
190 | ||
0b269ded | 191 | static void ovl_free_inode(struct inode *inode) |
13cf199d | 192 | { |
0b269ded | 193 | struct ovl_inode *oi = OVL_I(inode); |
13cf199d | 194 | |
0b269ded AV |
195 | kfree(oi->redirect); |
196 | mutex_destroy(&oi->lock); | |
197 | kmem_cache_free(ovl_inode_cachep, oi); | |
13cf199d AG |
198 | } |
199 | ||
200 | static void ovl_destroy_inode(struct inode *inode) | |
201 | { | |
09d8b586 MS |
202 | struct ovl_inode *oi = OVL_I(inode); |
203 | ||
204 | dput(oi->__upperdentry); | |
31747eda | 205 | iput(oi->lower); |
2664bd08 VG |
206 | if (S_ISDIR(inode->i_mode)) |
207 | ovl_dir_cache_free(inode); | |
208 | else | |
209 | iput(oi->lowerdata); | |
13cf199d AG |
210 | } |
211 | ||
ad204488 | 212 | static void ovl_free_fs(struct ovl_fs *ofs) |
e9be9d5e | 213 | { |
dd662667 | 214 | unsigned i; |
e9be9d5e | 215 | |
0be0bfd2 | 216 | iput(ofs->workbasedir_trap); |
146d62e5 AG |
217 | iput(ofs->indexdir_trap); |
218 | iput(ofs->workdir_trap); | |
219 | iput(ofs->upperdir_trap); | |
ad204488 MS |
220 | dput(ofs->indexdir); |
221 | dput(ofs->workdir); | |
222 | if (ofs->workdir_locked) | |
223 | ovl_inuse_unlock(ofs->workbasedir); | |
224 | dput(ofs->workbasedir); | |
225 | if (ofs->upperdir_locked) | |
226 | ovl_inuse_unlock(ofs->upper_mnt->mnt_root); | |
227 | mntput(ofs->upper_mnt); | |
94375f9d AG |
228 | for (i = 1; i < ofs->numlayer; i++) { |
229 | iput(ofs->layers[i].trap); | |
230 | mntput(ofs->layers[i].mnt); | |
146d62e5 | 231 | } |
94375f9d | 232 | kfree(ofs->layers); |
b7bf9908 | 233 | for (i = 0; i < ofs->numfs; i++) |
07f1e596 AG |
234 | free_anon_bdev(ofs->fs[i].pseudo_dev); |
235 | kfree(ofs->fs); | |
ad204488 MS |
236 | |
237 | kfree(ofs->config.lowerdir); | |
238 | kfree(ofs->config.upperdir); | |
239 | kfree(ofs->config.workdir); | |
438c84c2 | 240 | kfree(ofs->config.redirect_mode); |
ad204488 MS |
241 | if (ofs->creator_cred) |
242 | put_cred(ofs->creator_cred); | |
243 | kfree(ofs); | |
e9be9d5e MS |
244 | } |
245 | ||
a9075cdb MS |
246 | static void ovl_put_super(struct super_block *sb) |
247 | { | |
248 | struct ovl_fs *ofs = sb->s_fs_info; | |
249 | ||
250 | ovl_free_fs(ofs); | |
251 | } | |
252 | ||
e8d4bfe3 | 253 | /* Sync real dirty inodes in upper filesystem (if it exists) */ |
e593b2bf AG |
254 | static int ovl_sync_fs(struct super_block *sb, int wait) |
255 | { | |
ad204488 | 256 | struct ovl_fs *ofs = sb->s_fs_info; |
e593b2bf AG |
257 | struct super_block *upper_sb; |
258 | int ret; | |
259 | ||
ad204488 | 260 | if (!ofs->upper_mnt) |
e593b2bf | 261 | return 0; |
e8d4bfe3 CX |
262 | |
263 | /* | |
264 | * If this is a sync(2) call or an emergency sync, all the super blocks | |
265 | * will be iterated, including upper_sb, so no need to do anything. | |
266 | * | |
267 | * If this is a syncfs(2) call, then we do need to call | |
268 | * sync_filesystem() on upper_sb, but enough if we do it when being | |
269 | * called with wait == 1. | |
270 | */ | |
271 | if (!wait) | |
e593b2bf AG |
272 | return 0; |
273 | ||
e8d4bfe3 CX |
274 | upper_sb = ofs->upper_mnt->mnt_sb; |
275 | ||
e593b2bf | 276 | down_read(&upper_sb->s_umount); |
e8d4bfe3 | 277 | ret = sync_filesystem(upper_sb); |
e593b2bf | 278 | up_read(&upper_sb->s_umount); |
e8d4bfe3 | 279 | |
e593b2bf AG |
280 | return ret; |
281 | } | |
282 | ||
cc259639 AW |
283 | /** |
284 | * ovl_statfs | |
285 | * @sb: The overlayfs super block | |
286 | * @buf: The struct kstatfs to fill in with stats | |
287 | * | |
288 | * Get the filesystem statistics. As writes always target the upper layer | |
4ebc5818 | 289 | * filesystem pass the statfs to the upper filesystem (if it exists) |
cc259639 AW |
290 | */ |
291 | static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf) | |
292 | { | |
293 | struct ovl_fs *ofs = dentry->d_sb->s_fs_info; | |
294 | struct dentry *root_dentry = dentry->d_sb->s_root; | |
295 | struct path path; | |
296 | int err; | |
297 | ||
4ebc5818 | 298 | ovl_path_real(root_dentry, &path); |
cc259639 AW |
299 | |
300 | err = vfs_statfs(&path, buf); | |
301 | if (!err) { | |
6b2d5fe4 | 302 | buf->f_namelen = ofs->namelen; |
cc259639 AW |
303 | buf->f_type = OVERLAYFS_SUPER_MAGIC; |
304 | } | |
305 | ||
306 | return err; | |
307 | } | |
308 | ||
02bcd157 | 309 | /* Will this overlay be forced to mount/remount ro? */ |
ad204488 | 310 | static bool ovl_force_readonly(struct ovl_fs *ofs) |
02bcd157 | 311 | { |
ad204488 | 312 | return (!ofs->upper_mnt || !ofs->workdir); |
02bcd157 AG |
313 | } |
314 | ||
438c84c2 MS |
315 | static const char *ovl_redirect_mode_def(void) |
316 | { | |
317 | return ovl_redirect_dir_def ? "on" : "off"; | |
318 | } | |
319 | ||
795939a9 AG |
320 | enum { |
321 | OVL_XINO_OFF, | |
322 | OVL_XINO_AUTO, | |
323 | OVL_XINO_ON, | |
324 | }; | |
325 | ||
326 | static const char * const ovl_xino_str[] = { | |
327 | "off", | |
328 | "auto", | |
329 | "on", | |
330 | }; | |
331 | ||
332 | static inline int ovl_xino_def(void) | |
333 | { | |
334 | return ovl_xino_auto_def ? OVL_XINO_AUTO : OVL_XINO_OFF; | |
335 | } | |
336 | ||
f45827e8 EZ |
337 | /** |
338 | * ovl_show_options | |
339 | * | |
340 | * Prints the mount options for a given superblock. | |
341 | * Returns zero; does not fail. | |
342 | */ | |
343 | static int ovl_show_options(struct seq_file *m, struct dentry *dentry) | |
344 | { | |
345 | struct super_block *sb = dentry->d_sb; | |
ad204488 | 346 | struct ovl_fs *ofs = sb->s_fs_info; |
f45827e8 | 347 | |
ad204488 MS |
348 | seq_show_option(m, "lowerdir", ofs->config.lowerdir); |
349 | if (ofs->config.upperdir) { | |
350 | seq_show_option(m, "upperdir", ofs->config.upperdir); | |
351 | seq_show_option(m, "workdir", ofs->config.workdir); | |
53a08cb9 | 352 | } |
ad204488 | 353 | if (ofs->config.default_permissions) |
8d3095f4 | 354 | seq_puts(m, ",default_permissions"); |
438c84c2 MS |
355 | if (strcmp(ofs->config.redirect_mode, ovl_redirect_mode_def()) != 0) |
356 | seq_printf(m, ",redirect_dir=%s", ofs->config.redirect_mode); | |
ad204488 | 357 | if (ofs->config.index != ovl_index_def) |
438c84c2 | 358 | seq_printf(m, ",index=%s", ofs->config.index ? "on" : "off"); |
f168f109 AG |
359 | if (ofs->config.nfs_export != ovl_nfs_export_def) |
360 | seq_printf(m, ",nfs_export=%s", ofs->config.nfs_export ? | |
361 | "on" : "off"); | |
0f831ec8 | 362 | if (ofs->config.xino != ovl_xino_def() && !ovl_same_fs(sb)) |
795939a9 | 363 | seq_printf(m, ",xino=%s", ovl_xino_str[ofs->config.xino]); |
d5791044 VG |
364 | if (ofs->config.metacopy != ovl_metacopy_def) |
365 | seq_printf(m, ",metacopy=%s", | |
366 | ofs->config.metacopy ? "on" : "off"); | |
f45827e8 EZ |
367 | return 0; |
368 | } | |
369 | ||
3cdf6fe9 SL |
370 | static int ovl_remount(struct super_block *sb, int *flags, char *data) |
371 | { | |
ad204488 | 372 | struct ovl_fs *ofs = sb->s_fs_info; |
3cdf6fe9 | 373 | |
1751e8a6 | 374 | if (!(*flags & SB_RDONLY) && ovl_force_readonly(ofs)) |
3cdf6fe9 SL |
375 | return -EROFS; |
376 | ||
377 | return 0; | |
378 | } | |
379 | ||
e9be9d5e | 380 | static const struct super_operations ovl_super_operations = { |
13cf199d | 381 | .alloc_inode = ovl_alloc_inode, |
0b269ded | 382 | .free_inode = ovl_free_inode, |
13cf199d AG |
383 | .destroy_inode = ovl_destroy_inode, |
384 | .drop_inode = generic_delete_inode, | |
e9be9d5e | 385 | .put_super = ovl_put_super, |
e593b2bf | 386 | .sync_fs = ovl_sync_fs, |
cc259639 | 387 | .statfs = ovl_statfs, |
f45827e8 | 388 | .show_options = ovl_show_options, |
3cdf6fe9 | 389 | .remount_fs = ovl_remount, |
e9be9d5e MS |
390 | }; |
391 | ||
392 | enum { | |
393 | OPT_LOWERDIR, | |
394 | OPT_UPPERDIR, | |
395 | OPT_WORKDIR, | |
8d3095f4 | 396 | OPT_DEFAULT_PERMISSIONS, |
438c84c2 | 397 | OPT_REDIRECT_DIR, |
02bcd157 AG |
398 | OPT_INDEX_ON, |
399 | OPT_INDEX_OFF, | |
f168f109 AG |
400 | OPT_NFS_EXPORT_ON, |
401 | OPT_NFS_EXPORT_OFF, | |
795939a9 AG |
402 | OPT_XINO_ON, |
403 | OPT_XINO_OFF, | |
404 | OPT_XINO_AUTO, | |
d5791044 VG |
405 | OPT_METACOPY_ON, |
406 | OPT_METACOPY_OFF, | |
e9be9d5e MS |
407 | OPT_ERR, |
408 | }; | |
409 | ||
410 | static const match_table_t ovl_tokens = { | |
411 | {OPT_LOWERDIR, "lowerdir=%s"}, | |
412 | {OPT_UPPERDIR, "upperdir=%s"}, | |
413 | {OPT_WORKDIR, "workdir=%s"}, | |
8d3095f4 | 414 | {OPT_DEFAULT_PERMISSIONS, "default_permissions"}, |
438c84c2 | 415 | {OPT_REDIRECT_DIR, "redirect_dir=%s"}, |
02bcd157 AG |
416 | {OPT_INDEX_ON, "index=on"}, |
417 | {OPT_INDEX_OFF, "index=off"}, | |
f168f109 AG |
418 | {OPT_NFS_EXPORT_ON, "nfs_export=on"}, |
419 | {OPT_NFS_EXPORT_OFF, "nfs_export=off"}, | |
795939a9 AG |
420 | {OPT_XINO_ON, "xino=on"}, |
421 | {OPT_XINO_OFF, "xino=off"}, | |
422 | {OPT_XINO_AUTO, "xino=auto"}, | |
d5791044 VG |
423 | {OPT_METACOPY_ON, "metacopy=on"}, |
424 | {OPT_METACOPY_OFF, "metacopy=off"}, | |
e9be9d5e MS |
425 | {OPT_ERR, NULL} |
426 | }; | |
427 | ||
91c77947 MS |
428 | static char *ovl_next_opt(char **s) |
429 | { | |
430 | char *sbegin = *s; | |
431 | char *p; | |
432 | ||
433 | if (sbegin == NULL) | |
434 | return NULL; | |
435 | ||
436 | for (p = sbegin; *p; p++) { | |
437 | if (*p == '\\') { | |
438 | p++; | |
439 | if (!*p) | |
440 | break; | |
441 | } else if (*p == ',') { | |
442 | *p = '\0'; | |
443 | *s = p + 1; | |
444 | return sbegin; | |
445 | } | |
446 | } | |
447 | *s = NULL; | |
448 | return sbegin; | |
449 | } | |
450 | ||
438c84c2 MS |
451 | static int ovl_parse_redirect_mode(struct ovl_config *config, const char *mode) |
452 | { | |
453 | if (strcmp(mode, "on") == 0) { | |
454 | config->redirect_dir = true; | |
455 | /* | |
456 | * Does not make sense to have redirect creation without | |
457 | * redirect following. | |
458 | */ | |
459 | config->redirect_follow = true; | |
460 | } else if (strcmp(mode, "follow") == 0) { | |
461 | config->redirect_follow = true; | |
462 | } else if (strcmp(mode, "off") == 0) { | |
463 | if (ovl_redirect_always_follow) | |
464 | config->redirect_follow = true; | |
465 | } else if (strcmp(mode, "nofollow") != 0) { | |
1bd0a3ae | 466 | pr_err("bad mount option \"redirect_dir=%s\"\n", |
438c84c2 MS |
467 | mode); |
468 | return -EINVAL; | |
469 | } | |
470 | ||
471 | return 0; | |
472 | } | |
473 | ||
e9be9d5e MS |
474 | static int ovl_parse_opt(char *opt, struct ovl_config *config) |
475 | { | |
476 | char *p; | |
d5791044 | 477 | int err; |
d47748e5 | 478 | bool metacopy_opt = false, redirect_opt = false; |
e9be9d5e | 479 | |
438c84c2 MS |
480 | config->redirect_mode = kstrdup(ovl_redirect_mode_def(), GFP_KERNEL); |
481 | if (!config->redirect_mode) | |
482 | return -ENOMEM; | |
483 | ||
91c77947 | 484 | while ((p = ovl_next_opt(&opt)) != NULL) { |
e9be9d5e MS |
485 | int token; |
486 | substring_t args[MAX_OPT_ARGS]; | |
487 | ||
488 | if (!*p) | |
489 | continue; | |
490 | ||
491 | token = match_token(p, ovl_tokens, args); | |
492 | switch (token) { | |
493 | case OPT_UPPERDIR: | |
494 | kfree(config->upperdir); | |
495 | config->upperdir = match_strdup(&args[0]); | |
496 | if (!config->upperdir) | |
497 | return -ENOMEM; | |
498 | break; | |
499 | ||
500 | case OPT_LOWERDIR: | |
501 | kfree(config->lowerdir); | |
502 | config->lowerdir = match_strdup(&args[0]); | |
503 | if (!config->lowerdir) | |
504 | return -ENOMEM; | |
505 | break; | |
506 | ||
507 | case OPT_WORKDIR: | |
508 | kfree(config->workdir); | |
509 | config->workdir = match_strdup(&args[0]); | |
510 | if (!config->workdir) | |
511 | return -ENOMEM; | |
512 | break; | |
513 | ||
8d3095f4 MS |
514 | case OPT_DEFAULT_PERMISSIONS: |
515 | config->default_permissions = true; | |
516 | break; | |
517 | ||
438c84c2 MS |
518 | case OPT_REDIRECT_DIR: |
519 | kfree(config->redirect_mode); | |
520 | config->redirect_mode = match_strdup(&args[0]); | |
521 | if (!config->redirect_mode) | |
522 | return -ENOMEM; | |
d47748e5 | 523 | redirect_opt = true; |
a6c60655 MS |
524 | break; |
525 | ||
02bcd157 AG |
526 | case OPT_INDEX_ON: |
527 | config->index = true; | |
528 | break; | |
529 | ||
530 | case OPT_INDEX_OFF: | |
531 | config->index = false; | |
532 | break; | |
533 | ||
f168f109 AG |
534 | case OPT_NFS_EXPORT_ON: |
535 | config->nfs_export = true; | |
536 | break; | |
537 | ||
538 | case OPT_NFS_EXPORT_OFF: | |
539 | config->nfs_export = false; | |
540 | break; | |
541 | ||
795939a9 AG |
542 | case OPT_XINO_ON: |
543 | config->xino = OVL_XINO_ON; | |
544 | break; | |
545 | ||
546 | case OPT_XINO_OFF: | |
547 | config->xino = OVL_XINO_OFF; | |
548 | break; | |
549 | ||
550 | case OPT_XINO_AUTO: | |
551 | config->xino = OVL_XINO_AUTO; | |
552 | break; | |
553 | ||
d5791044 VG |
554 | case OPT_METACOPY_ON: |
555 | config->metacopy = true; | |
d47748e5 | 556 | metacopy_opt = true; |
d5791044 VG |
557 | break; |
558 | ||
559 | case OPT_METACOPY_OFF: | |
560 | config->metacopy = false; | |
561 | break; | |
562 | ||
e9be9d5e | 563 | default: |
1bd0a3ae | 564 | pr_err("unrecognized mount option \"%s\" or missing value\n", |
565 | p); | |
e9be9d5e MS |
566 | return -EINVAL; |
567 | } | |
568 | } | |
71cbad7e | 569 | |
570 | /* Workdir is useless in non-upper mount */ | |
571 | if (!config->upperdir && config->workdir) { | |
1bd0a3ae | 572 | pr_info("option \"workdir=%s\" is useless in a non-upper mount, ignore\n", |
71cbad7e | 573 | config->workdir); |
574 | kfree(config->workdir); | |
575 | config->workdir = NULL; | |
576 | } | |
577 | ||
d5791044 VG |
578 | err = ovl_parse_redirect_mode(config, config->redirect_mode); |
579 | if (err) | |
580 | return err; | |
581 | ||
d47748e5 MS |
582 | /* |
583 | * This is to make the logic below simpler. It doesn't make any other | |
584 | * difference, since config->redirect_dir is only used for upper. | |
585 | */ | |
586 | if (!config->upperdir && config->redirect_follow) | |
587 | config->redirect_dir = true; | |
588 | ||
589 | /* Resolve metacopy -> redirect_dir dependency */ | |
590 | if (config->metacopy && !config->redirect_dir) { | |
591 | if (metacopy_opt && redirect_opt) { | |
1bd0a3ae | 592 | pr_err("conflicting options: metacopy=on,redirect_dir=%s\n", |
d47748e5 MS |
593 | config->redirect_mode); |
594 | return -EINVAL; | |
595 | } | |
596 | if (redirect_opt) { | |
597 | /* | |
598 | * There was an explicit redirect_dir=... that resulted | |
599 | * in this conflict. | |
600 | */ | |
1bd0a3ae | 601 | pr_info("disabling metacopy due to redirect_dir=%s\n", |
d47748e5 MS |
602 | config->redirect_mode); |
603 | config->metacopy = false; | |
604 | } else { | |
605 | /* Automatically enable redirect otherwise. */ | |
606 | config->redirect_follow = config->redirect_dir = true; | |
607 | } | |
d5791044 VG |
608 | } |
609 | ||
610 | return 0; | |
e9be9d5e MS |
611 | } |
612 | ||
613 | #define OVL_WORKDIR_NAME "work" | |
02bcd157 | 614 | #define OVL_INDEXDIR_NAME "index" |
e9be9d5e | 615 | |
ad204488 | 616 | static struct dentry *ovl_workdir_create(struct ovl_fs *ofs, |
6b8aa129 | 617 | const char *name, bool persist) |
e9be9d5e | 618 | { |
ad204488 MS |
619 | struct inode *dir = ofs->workbasedir->d_inode; |
620 | struct vfsmount *mnt = ofs->upper_mnt; | |
e9be9d5e MS |
621 | struct dentry *work; |
622 | int err; | |
623 | bool retried = false; | |
6b8aa129 | 624 | bool locked = false; |
e9be9d5e | 625 | |
5955102c | 626 | inode_lock_nested(dir, I_MUTEX_PARENT); |
6b8aa129 AG |
627 | locked = true; |
628 | ||
e9be9d5e | 629 | retry: |
ad204488 | 630 | work = lookup_one_len(name, ofs->workbasedir, strlen(name)); |
e9be9d5e MS |
631 | |
632 | if (!IS_ERR(work)) { | |
c11b9fdd MS |
633 | struct iattr attr = { |
634 | .ia_valid = ATTR_MODE, | |
32a3d848 | 635 | .ia_mode = S_IFDIR | 0, |
c11b9fdd | 636 | }; |
e9be9d5e MS |
637 | |
638 | if (work->d_inode) { | |
639 | err = -EEXIST; | |
640 | if (retried) | |
641 | goto out_dput; | |
642 | ||
6b8aa129 AG |
643 | if (persist) |
644 | goto out_unlock; | |
645 | ||
e9be9d5e | 646 | retried = true; |
eea2fb48 | 647 | ovl_workdir_cleanup(dir, mnt, work, 0); |
e9be9d5e MS |
648 | dput(work); |
649 | goto retry; | |
650 | } | |
651 | ||
95a1c815 MS |
652 | work = ovl_create_real(dir, work, OVL_CATTR(attr.ia_mode)); |
653 | err = PTR_ERR(work); | |
654 | if (IS_ERR(work)) | |
655 | goto out_err; | |
c11b9fdd | 656 | |
cb348edb MS |
657 | /* |
658 | * Try to remove POSIX ACL xattrs from workdir. We are good if: | |
659 | * | |
660 | * a) success (there was a POSIX ACL xattr and was removed) | |
661 | * b) -ENODATA (there was no POSIX ACL xattr) | |
662 | * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported) | |
663 | * | |
664 | * There are various other error values that could effectively | |
665 | * mean that the xattr doesn't exist (e.g. -ERANGE is returned | |
666 | * if the xattr name is too long), but the set of filesystems | |
667 | * allowed as upper are limited to "normal" ones, where checking | |
668 | * for the above two errors is sufficient. | |
669 | */ | |
c11b9fdd | 670 | err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_DEFAULT); |
e1ff3dd1 | 671 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
672 | goto out_dput; |
673 | ||
674 | err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_ACCESS); | |
e1ff3dd1 | 675 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
676 | goto out_dput; |
677 | ||
678 | /* Clear any inherited mode bits */ | |
679 | inode_lock(work->d_inode); | |
680 | err = notify_change(work, &attr, NULL); | |
681 | inode_unlock(work->d_inode); | |
682 | if (err) | |
683 | goto out_dput; | |
6b8aa129 AG |
684 | } else { |
685 | err = PTR_ERR(work); | |
686 | goto out_err; | |
e9be9d5e MS |
687 | } |
688 | out_unlock: | |
6b8aa129 AG |
689 | if (locked) |
690 | inode_unlock(dir); | |
e9be9d5e MS |
691 | |
692 | return work; | |
693 | ||
694 | out_dput: | |
695 | dput(work); | |
6b8aa129 | 696 | out_err: |
1bd0a3ae | 697 | pr_warn("failed to create directory %s/%s (errno: %i); mounting read-only\n", |
ad204488 | 698 | ofs->config.workdir, name, -err); |
6b8aa129 | 699 | work = NULL; |
e9be9d5e MS |
700 | goto out_unlock; |
701 | } | |
702 | ||
91c77947 MS |
703 | static void ovl_unescape(char *s) |
704 | { | |
705 | char *d = s; | |
706 | ||
707 | for (;; s++, d++) { | |
708 | if (*s == '\\') | |
709 | s++; | |
710 | *d = *s; | |
711 | if (!*s) | |
712 | break; | |
713 | } | |
714 | } | |
715 | ||
ab508822 MS |
716 | static int ovl_mount_dir_noesc(const char *name, struct path *path) |
717 | { | |
a78d9f0d | 718 | int err = -EINVAL; |
ab508822 | 719 | |
a78d9f0d | 720 | if (!*name) { |
1bd0a3ae | 721 | pr_err("empty lowerdir\n"); |
a78d9f0d MS |
722 | goto out; |
723 | } | |
ab508822 MS |
724 | err = kern_path(name, LOOKUP_FOLLOW, path); |
725 | if (err) { | |
1bd0a3ae | 726 | pr_err("failed to resolve '%s': %i\n", name, err); |
ab508822 MS |
727 | goto out; |
728 | } | |
729 | err = -EINVAL; | |
7c03b5d4 | 730 | if (ovl_dentry_weird(path->dentry)) { |
1bd0a3ae | 731 | pr_err("filesystem on '%s' not supported\n", name); |
ab508822 MS |
732 | goto out_put; |
733 | } | |
2b8c30e9 | 734 | if (!d_is_dir(path->dentry)) { |
1bd0a3ae | 735 | pr_err("'%s' not a directory\n", name); |
ab508822 MS |
736 | goto out_put; |
737 | } | |
738 | return 0; | |
739 | ||
740 | out_put: | |
8aafcb59 | 741 | path_put_init(path); |
ab508822 MS |
742 | out: |
743 | return err; | |
744 | } | |
745 | ||
746 | static int ovl_mount_dir(const char *name, struct path *path) | |
747 | { | |
748 | int err = -ENOMEM; | |
749 | char *tmp = kstrdup(name, GFP_KERNEL); | |
750 | ||
751 | if (tmp) { | |
752 | ovl_unescape(tmp); | |
753 | err = ovl_mount_dir_noesc(tmp, path); | |
7c03b5d4 | 754 | |
bccece1e | 755 | if (!err && path->dentry->d_flags & DCACHE_OP_REAL) { |
7925dad8 MS |
756 | pr_err("filesystem on '%s' not supported as upperdir\n", |
757 | tmp); | |
758 | path_put_init(path); | |
759 | err = -EINVAL; | |
760 | } | |
ab508822 MS |
761 | kfree(tmp); |
762 | } | |
763 | return err; | |
764 | } | |
765 | ||
6b2d5fe4 MS |
766 | static int ovl_check_namelen(struct path *path, struct ovl_fs *ofs, |
767 | const char *name) | |
ab508822 | 768 | { |
ab508822 | 769 | struct kstatfs statfs; |
6b2d5fe4 MS |
770 | int err = vfs_statfs(path, &statfs); |
771 | ||
772 | if (err) | |
1bd0a3ae | 773 | pr_err("statfs failed on '%s'\n", name); |
6b2d5fe4 MS |
774 | else |
775 | ofs->namelen = max(ofs->namelen, statfs.f_namelen); | |
776 | ||
777 | return err; | |
778 | } | |
779 | ||
780 | static int ovl_lower_dir(const char *name, struct path *path, | |
f4288844 | 781 | struct ovl_fs *ofs, int *stack_depth) |
6b2d5fe4 | 782 | { |
e487d889 | 783 | int fh_type; |
6b2d5fe4 | 784 | int err; |
ab508822 | 785 | |
a78d9f0d | 786 | err = ovl_mount_dir_noesc(name, path); |
ab508822 MS |
787 | if (err) |
788 | goto out; | |
789 | ||
6b2d5fe4 MS |
790 | err = ovl_check_namelen(path, ofs, name); |
791 | if (err) | |
ab508822 | 792 | goto out_put; |
6b2d5fe4 | 793 | |
ab508822 MS |
794 | *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth); |
795 | ||
02bcd157 | 796 | /* |
f168f109 AG |
797 | * The inodes index feature and NFS export need to encode and decode |
798 | * file handles, so they require that all layers support them. | |
02bcd157 | 799 | */ |
e487d889 | 800 | fh_type = ovl_can_decode_fh(path->dentry->d_sb); |
f168f109 | 801 | if ((ofs->config.nfs_export || |
e487d889 | 802 | (ofs->config.index && ofs->config.upperdir)) && !fh_type) { |
02bcd157 | 803 | ofs->config.index = false; |
f168f109 | 804 | ofs->config.nfs_export = false; |
1bd0a3ae | 805 | pr_warn("fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n", |
f168f109 | 806 | name); |
02bcd157 AG |
807 | } |
808 | ||
e487d889 AG |
809 | /* Check if lower fs has 32bit inode numbers */ |
810 | if (fh_type != FILEID_INO32_GEN) | |
0f831ec8 | 811 | ofs->xino_mode = -1; |
e487d889 | 812 | |
ab508822 MS |
813 | return 0; |
814 | ||
815 | out_put: | |
8aafcb59 | 816 | path_put_init(path); |
ab508822 MS |
817 | out: |
818 | return err; | |
819 | } | |
820 | ||
e9be9d5e MS |
821 | /* Workdir should not be subdir of upperdir and vice versa */ |
822 | static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir) | |
823 | { | |
824 | bool ok = false; | |
825 | ||
826 | if (workdir != upperdir) { | |
827 | ok = (lock_rename(workdir, upperdir) == NULL); | |
828 | unlock_rename(workdir, upperdir); | |
829 | } | |
830 | return ok; | |
831 | } | |
832 | ||
a78d9f0d MS |
833 | static unsigned int ovl_split_lowerdirs(char *str) |
834 | { | |
835 | unsigned int ctr = 1; | |
836 | char *s, *d; | |
837 | ||
838 | for (s = d = str;; s++, d++) { | |
839 | if (*s == '\\') { | |
840 | s++; | |
841 | } else if (*s == ':') { | |
842 | *d = '\0'; | |
843 | ctr++; | |
844 | continue; | |
845 | } | |
846 | *d = *s; | |
847 | if (!*s) | |
848 | break; | |
849 | } | |
850 | return ctr; | |
851 | } | |
852 | ||
0eb45fc3 AG |
853 | static int __maybe_unused |
854 | ovl_posix_acl_xattr_get(const struct xattr_handler *handler, | |
855 | struct dentry *dentry, struct inode *inode, | |
856 | const char *name, void *buffer, size_t size) | |
857 | { | |
1d88f183 | 858 | return ovl_xattr_get(dentry, inode, handler->name, buffer, size); |
0eb45fc3 AG |
859 | } |
860 | ||
0c97be22 AG |
861 | static int __maybe_unused |
862 | ovl_posix_acl_xattr_set(const struct xattr_handler *handler, | |
863 | struct dentry *dentry, struct inode *inode, | |
864 | const char *name, const void *value, | |
865 | size_t size, int flags) | |
d837a49b MS |
866 | { |
867 | struct dentry *workdir = ovl_workdir(dentry); | |
09d8b586 | 868 | struct inode *realinode = ovl_inode_real(inode); |
d837a49b MS |
869 | struct posix_acl *acl = NULL; |
870 | int err; | |
871 | ||
872 | /* Check that everything is OK before copy-up */ | |
873 | if (value) { | |
874 | acl = posix_acl_from_xattr(&init_user_ns, value, size); | |
875 | if (IS_ERR(acl)) | |
876 | return PTR_ERR(acl); | |
877 | } | |
878 | err = -EOPNOTSUPP; | |
879 | if (!IS_POSIXACL(d_inode(workdir))) | |
880 | goto out_acl_release; | |
881 | if (!realinode->i_op->set_acl) | |
882 | goto out_acl_release; | |
883 | if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) { | |
884 | err = acl ? -EACCES : 0; | |
885 | goto out_acl_release; | |
886 | } | |
887 | err = -EPERM; | |
888 | if (!inode_owner_or_capable(inode)) | |
889 | goto out_acl_release; | |
890 | ||
891 | posix_acl_release(acl); | |
892 | ||
fd3220d3 MS |
893 | /* |
894 | * Check if sgid bit needs to be cleared (actual setacl operation will | |
895 | * be done with mounter's capabilities and so that won't do it for us). | |
896 | */ | |
897 | if (unlikely(inode->i_mode & S_ISGID) && | |
898 | handler->flags == ACL_TYPE_ACCESS && | |
899 | !in_group_p(inode->i_gid) && | |
900 | !capable_wrt_inode_uidgid(inode, CAP_FSETID)) { | |
901 | struct iattr iattr = { .ia_valid = ATTR_KILL_SGID }; | |
902 | ||
903 | err = ovl_setattr(dentry, &iattr); | |
904 | if (err) | |
905 | return err; | |
906 | } | |
907 | ||
1d88f183 | 908 | err = ovl_xattr_set(dentry, inode, handler->name, value, size, flags); |
ce31513a | 909 | if (!err) |
09d8b586 | 910 | ovl_copyattr(ovl_inode_real(inode), inode); |
ce31513a MS |
911 | |
912 | return err; | |
d837a49b MS |
913 | |
914 | out_acl_release: | |
915 | posix_acl_release(acl); | |
916 | return err; | |
917 | } | |
918 | ||
0eb45fc3 AG |
919 | static int ovl_own_xattr_get(const struct xattr_handler *handler, |
920 | struct dentry *dentry, struct inode *inode, | |
921 | const char *name, void *buffer, size_t size) | |
922 | { | |
48fab5d7 | 923 | return -EOPNOTSUPP; |
0eb45fc3 AG |
924 | } |
925 | ||
d837a49b MS |
926 | static int ovl_own_xattr_set(const struct xattr_handler *handler, |
927 | struct dentry *dentry, struct inode *inode, | |
928 | const char *name, const void *value, | |
929 | size_t size, int flags) | |
930 | { | |
48fab5d7 | 931 | return -EOPNOTSUPP; |
d837a49b MS |
932 | } |
933 | ||
0eb45fc3 AG |
934 | static int ovl_other_xattr_get(const struct xattr_handler *handler, |
935 | struct dentry *dentry, struct inode *inode, | |
936 | const char *name, void *buffer, size_t size) | |
937 | { | |
1d88f183 | 938 | return ovl_xattr_get(dentry, inode, name, buffer, size); |
0eb45fc3 AG |
939 | } |
940 | ||
0e585ccc AG |
941 | static int ovl_other_xattr_set(const struct xattr_handler *handler, |
942 | struct dentry *dentry, struct inode *inode, | |
943 | const char *name, const void *value, | |
944 | size_t size, int flags) | |
945 | { | |
1d88f183 | 946 | return ovl_xattr_set(dentry, inode, name, value, size, flags); |
0e585ccc AG |
947 | } |
948 | ||
0c97be22 AG |
949 | static const struct xattr_handler __maybe_unused |
950 | ovl_posix_acl_access_xattr_handler = { | |
d837a49b MS |
951 | .name = XATTR_NAME_POSIX_ACL_ACCESS, |
952 | .flags = ACL_TYPE_ACCESS, | |
0eb45fc3 | 953 | .get = ovl_posix_acl_xattr_get, |
d837a49b MS |
954 | .set = ovl_posix_acl_xattr_set, |
955 | }; | |
956 | ||
0c97be22 AG |
957 | static const struct xattr_handler __maybe_unused |
958 | ovl_posix_acl_default_xattr_handler = { | |
d837a49b MS |
959 | .name = XATTR_NAME_POSIX_ACL_DEFAULT, |
960 | .flags = ACL_TYPE_DEFAULT, | |
0eb45fc3 | 961 | .get = ovl_posix_acl_xattr_get, |
d837a49b MS |
962 | .set = ovl_posix_acl_xattr_set, |
963 | }; | |
964 | ||
965 | static const struct xattr_handler ovl_own_xattr_handler = { | |
966 | .prefix = OVL_XATTR_PREFIX, | |
0eb45fc3 | 967 | .get = ovl_own_xattr_get, |
d837a49b MS |
968 | .set = ovl_own_xattr_set, |
969 | }; | |
970 | ||
971 | static const struct xattr_handler ovl_other_xattr_handler = { | |
972 | .prefix = "", /* catch all */ | |
0eb45fc3 | 973 | .get = ovl_other_xattr_get, |
d837a49b MS |
974 | .set = ovl_other_xattr_set, |
975 | }; | |
976 | ||
977 | static const struct xattr_handler *ovl_xattr_handlers[] = { | |
0c97be22 | 978 | #ifdef CONFIG_FS_POSIX_ACL |
d837a49b MS |
979 | &ovl_posix_acl_access_xattr_handler, |
980 | &ovl_posix_acl_default_xattr_handler, | |
0c97be22 | 981 | #endif |
d837a49b MS |
982 | &ovl_own_xattr_handler, |
983 | &ovl_other_xattr_handler, | |
984 | NULL | |
985 | }; | |
986 | ||
146d62e5 AG |
987 | static int ovl_setup_trap(struct super_block *sb, struct dentry *dir, |
988 | struct inode **ptrap, const char *name) | |
989 | { | |
990 | struct inode *trap; | |
991 | int err; | |
992 | ||
993 | trap = ovl_get_trap_inode(sb, dir); | |
1dac6f5b AB |
994 | err = PTR_ERR_OR_ZERO(trap); |
995 | if (err) { | |
146d62e5 | 996 | if (err == -ELOOP) |
1bd0a3ae | 997 | pr_err("conflicting %s path\n", name); |
146d62e5 AG |
998 | return err; |
999 | } | |
1000 | ||
1001 | *ptrap = trap; | |
1002 | return 0; | |
1003 | } | |
1004 | ||
0be0bfd2 AG |
1005 | /* |
1006 | * Determine how we treat concurrent use of upperdir/workdir based on the | |
1007 | * index feature. This is papering over mount leaks of container runtimes, | |
1008 | * for example, an old overlay mount is leaked and now its upperdir is | |
1009 | * attempted to be used as a lower layer in a new overlay mount. | |
1010 | */ | |
1011 | static int ovl_report_in_use(struct ovl_fs *ofs, const char *name) | |
1012 | { | |
1013 | if (ofs->config.index) { | |
1bd0a3ae | 1014 | pr_err("%s is in-use as upperdir/workdir of another mount, mount with '-o index=off' to override exclusive upperdir protection.\n", |
0be0bfd2 AG |
1015 | name); |
1016 | return -EBUSY; | |
1017 | } else { | |
1bd0a3ae | 1018 | pr_warn("%s is in-use as upperdir/workdir of another mount, accessing files from both mounts will result in undefined behavior.\n", |
0be0bfd2 AG |
1019 | name); |
1020 | return 0; | |
1021 | } | |
1022 | } | |
1023 | ||
146d62e5 AG |
1024 | static int ovl_get_upper(struct super_block *sb, struct ovl_fs *ofs, |
1025 | struct path *upperpath) | |
6ee8acf0 | 1026 | { |
5064975e | 1027 | struct vfsmount *upper_mnt; |
6ee8acf0 MS |
1028 | int err; |
1029 | ||
ad204488 | 1030 | err = ovl_mount_dir(ofs->config.upperdir, upperpath); |
6ee8acf0 MS |
1031 | if (err) |
1032 | goto out; | |
1033 | ||
1034 | /* Upper fs should not be r/o */ | |
1035 | if (sb_rdonly(upperpath->mnt->mnt_sb)) { | |
1bd0a3ae | 1036 | pr_err("upper fs is r/o, try multi-lower layers mount\n"); |
6ee8acf0 MS |
1037 | err = -EINVAL; |
1038 | goto out; | |
1039 | } | |
1040 | ||
ad204488 | 1041 | err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir); |
6ee8acf0 MS |
1042 | if (err) |
1043 | goto out; | |
1044 | ||
146d62e5 AG |
1045 | err = ovl_setup_trap(sb, upperpath->dentry, &ofs->upperdir_trap, |
1046 | "upperdir"); | |
1047 | if (err) | |
1048 | goto out; | |
1049 | ||
5064975e MS |
1050 | upper_mnt = clone_private_mount(upperpath); |
1051 | err = PTR_ERR(upper_mnt); | |
1052 | if (IS_ERR(upper_mnt)) { | |
1bd0a3ae | 1053 | pr_err("failed to clone upperpath\n"); |
5064975e MS |
1054 | goto out; |
1055 | } | |
1056 | ||
1057 | /* Don't inherit atime flags */ | |
1058 | upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME); | |
ad204488 | 1059 | ofs->upper_mnt = upper_mnt; |
8c25741a | 1060 | |
8c25741a MS |
1061 | if (ovl_inuse_trylock(ofs->upper_mnt->mnt_root)) { |
1062 | ofs->upperdir_locked = true; | |
8c25741a | 1063 | } else { |
0be0bfd2 AG |
1064 | err = ovl_report_in_use(ofs, "upperdir"); |
1065 | if (err) | |
1066 | goto out; | |
8c25741a MS |
1067 | } |
1068 | ||
6ee8acf0 MS |
1069 | err = 0; |
1070 | out: | |
1071 | return err; | |
1072 | } | |
1073 | ||
146d62e5 AG |
1074 | static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs, |
1075 | struct path *workpath) | |
8ed61dc3 | 1076 | { |
2ba9d57e | 1077 | struct vfsmount *mnt = ofs->upper_mnt; |
8ed61dc3 | 1078 | struct dentry *temp; |
e487d889 | 1079 | int fh_type; |
8ed61dc3 MS |
1080 | int err; |
1081 | ||
2ba9d57e AG |
1082 | err = mnt_want_write(mnt); |
1083 | if (err) | |
1084 | return err; | |
1085 | ||
ad204488 MS |
1086 | ofs->workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false); |
1087 | if (!ofs->workdir) | |
2ba9d57e | 1088 | goto out; |
8ed61dc3 | 1089 | |
146d62e5 AG |
1090 | err = ovl_setup_trap(sb, ofs->workdir, &ofs->workdir_trap, "workdir"); |
1091 | if (err) | |
1092 | goto out; | |
1093 | ||
8ed61dc3 MS |
1094 | /* |
1095 | * Upper should support d_type, else whiteouts are visible. Given | |
1096 | * workdir and upper are on same fs, we can do iterate_dir() on | |
1097 | * workdir. This check requires successful creation of workdir in | |
1098 | * previous step. | |
1099 | */ | |
1100 | err = ovl_check_d_type_supported(workpath); | |
1101 | if (err < 0) | |
2ba9d57e | 1102 | goto out; |
8ed61dc3 MS |
1103 | |
1104 | /* | |
1105 | * We allowed this configuration and don't want to break users over | |
1106 | * kernel upgrade. So warn instead of erroring out. | |
1107 | */ | |
1108 | if (!err) | |
1bd0a3ae | 1109 | pr_warn("upper fs needs to support d_type.\n"); |
8ed61dc3 MS |
1110 | |
1111 | /* Check if upper/work fs supports O_TMPFILE */ | |
ad204488 MS |
1112 | temp = ovl_do_tmpfile(ofs->workdir, S_IFREG | 0); |
1113 | ofs->tmpfile = !IS_ERR(temp); | |
1114 | if (ofs->tmpfile) | |
8ed61dc3 MS |
1115 | dput(temp); |
1116 | else | |
1bd0a3ae | 1117 | pr_warn("upper fs does not support tmpfile.\n"); |
8ed61dc3 MS |
1118 | |
1119 | /* | |
1120 | * Check if upper/work fs supports trusted.overlay.* xattr | |
1121 | */ | |
ad204488 | 1122 | err = ovl_do_setxattr(ofs->workdir, OVL_XATTR_OPAQUE, "0", 1, 0); |
8ed61dc3 | 1123 | if (err) { |
ad204488 | 1124 | ofs->noxattr = true; |
a683737b | 1125 | ofs->config.index = false; |
d5791044 | 1126 | ofs->config.metacopy = false; |
1bd0a3ae | 1127 | pr_warn("upper fs does not support xattr, falling back to index=off and metacopy=off.\n"); |
2ba9d57e | 1128 | err = 0; |
8ed61dc3 | 1129 | } else { |
ad204488 | 1130 | vfs_removexattr(ofs->workdir, OVL_XATTR_OPAQUE); |
8ed61dc3 MS |
1131 | } |
1132 | ||
1133 | /* Check if upper/work fs supports file handles */ | |
e487d889 AG |
1134 | fh_type = ovl_can_decode_fh(ofs->workdir->d_sb); |
1135 | if (ofs->config.index && !fh_type) { | |
ad204488 | 1136 | ofs->config.index = false; |
1bd0a3ae | 1137 | pr_warn("upper fs does not support file handles, falling back to index=off.\n"); |
8ed61dc3 MS |
1138 | } |
1139 | ||
e487d889 AG |
1140 | /* Check if upper fs has 32bit inode numbers */ |
1141 | if (fh_type != FILEID_INO32_GEN) | |
0f831ec8 | 1142 | ofs->xino_mode = -1; |
e487d889 | 1143 | |
f168f109 AG |
1144 | /* NFS export of r/w mount depends on index */ |
1145 | if (ofs->config.nfs_export && !ofs->config.index) { | |
1bd0a3ae | 1146 | pr_warn("NFS export requires \"index=on\", falling back to nfs_export=off.\n"); |
f168f109 AG |
1147 | ofs->config.nfs_export = false; |
1148 | } | |
2ba9d57e AG |
1149 | out: |
1150 | mnt_drop_write(mnt); | |
1151 | return err; | |
8ed61dc3 MS |
1152 | } |
1153 | ||
146d62e5 AG |
1154 | static int ovl_get_workdir(struct super_block *sb, struct ovl_fs *ofs, |
1155 | struct path *upperpath) | |
520d7c86 MS |
1156 | { |
1157 | int err; | |
bca44b52 | 1158 | struct path workpath = { }; |
520d7c86 | 1159 | |
ad204488 | 1160 | err = ovl_mount_dir(ofs->config.workdir, &workpath); |
520d7c86 MS |
1161 | if (err) |
1162 | goto out; | |
1163 | ||
1164 | err = -EINVAL; | |
bca44b52 | 1165 | if (upperpath->mnt != workpath.mnt) { |
1bd0a3ae | 1166 | pr_err("workdir and upperdir must reside under the same mount\n"); |
520d7c86 MS |
1167 | goto out; |
1168 | } | |
bca44b52 | 1169 | if (!ovl_workdir_ok(workpath.dentry, upperpath->dentry)) { |
1bd0a3ae | 1170 | pr_err("workdir and upperdir must be separate subtrees\n"); |
520d7c86 MS |
1171 | goto out; |
1172 | } | |
1173 | ||
8c25741a MS |
1174 | ofs->workbasedir = dget(workpath.dentry); |
1175 | ||
8c25741a | 1176 | if (ovl_inuse_trylock(ofs->workbasedir)) { |
ad204488 | 1177 | ofs->workdir_locked = true; |
520d7c86 | 1178 | } else { |
0be0bfd2 AG |
1179 | err = ovl_report_in_use(ofs, "workdir"); |
1180 | if (err) | |
1181 | goto out; | |
520d7c86 MS |
1182 | } |
1183 | ||
0be0bfd2 AG |
1184 | err = ovl_setup_trap(sb, ofs->workbasedir, &ofs->workbasedir_trap, |
1185 | "workdir"); | |
1186 | if (err) | |
1187 | goto out; | |
1188 | ||
146d62e5 | 1189 | err = ovl_make_workdir(sb, ofs, &workpath); |
bca44b52 | 1190 | |
520d7c86 | 1191 | out: |
bca44b52 MS |
1192 | path_put(&workpath); |
1193 | ||
520d7c86 MS |
1194 | return err; |
1195 | } | |
1196 | ||
146d62e5 AG |
1197 | static int ovl_get_indexdir(struct super_block *sb, struct ovl_fs *ofs, |
1198 | struct ovl_entry *oe, struct path *upperpath) | |
f7e3a7d9 | 1199 | { |
2ba9d57e | 1200 | struct vfsmount *mnt = ofs->upper_mnt; |
f7e3a7d9 MS |
1201 | int err; |
1202 | ||
2ba9d57e AG |
1203 | err = mnt_want_write(mnt); |
1204 | if (err) | |
1205 | return err; | |
1206 | ||
f7e3a7d9 | 1207 | /* Verify lower root is upper root origin */ |
d9768076 | 1208 | err = ovl_verify_origin(upperpath->dentry, oe->lowerstack[0].dentry, |
05122443 | 1209 | true); |
f7e3a7d9 | 1210 | if (err) { |
1bd0a3ae | 1211 | pr_err("failed to verify upper root origin\n"); |
f7e3a7d9 MS |
1212 | goto out; |
1213 | } | |
1214 | ||
ad204488 MS |
1215 | ofs->indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true); |
1216 | if (ofs->indexdir) { | |
146d62e5 AG |
1217 | err = ovl_setup_trap(sb, ofs->indexdir, &ofs->indexdir_trap, |
1218 | "indexdir"); | |
1219 | if (err) | |
1220 | goto out; | |
1221 | ||
ad1d615c AG |
1222 | /* |
1223 | * Verify upper root is exclusively associated with index dir. | |
1224 | * Older kernels stored upper fh in "trusted.overlay.origin" | |
1225 | * xattr. If that xattr exists, verify that it is a match to | |
1226 | * upper dir file handle. In any case, verify or set xattr | |
1227 | * "trusted.overlay.upper" to indicate that index may have | |
1228 | * directory entries. | |
1229 | */ | |
1230 | if (ovl_check_origin_xattr(ofs->indexdir)) { | |
1231 | err = ovl_verify_set_fh(ofs->indexdir, OVL_XATTR_ORIGIN, | |
1232 | upperpath->dentry, true, false); | |
1233 | if (err) | |
1bd0a3ae | 1234 | pr_err("failed to verify index dir 'origin' xattr\n"); |
ad1d615c AG |
1235 | } |
1236 | err = ovl_verify_upper(ofs->indexdir, upperpath->dentry, true); | |
f7e3a7d9 | 1237 | if (err) |
1bd0a3ae | 1238 | pr_err("failed to verify index dir 'upper' xattr\n"); |
f7e3a7d9 MS |
1239 | |
1240 | /* Cleanup bad/stale/orphan index entries */ | |
1241 | if (!err) | |
1eff1a1d | 1242 | err = ovl_indexdir_cleanup(ofs); |
f7e3a7d9 | 1243 | } |
ad204488 | 1244 | if (err || !ofs->indexdir) |
1bd0a3ae | 1245 | pr_warn("try deleting index dir or mounting with '-o index=off' to disable inodes index.\n"); |
f7e3a7d9 MS |
1246 | |
1247 | out: | |
2ba9d57e | 1248 | mnt_drop_write(mnt); |
f7e3a7d9 MS |
1249 | return err; |
1250 | } | |
1251 | ||
9df085f3 AG |
1252 | static bool ovl_lower_uuid_ok(struct ovl_fs *ofs, const uuid_t *uuid) |
1253 | { | |
1254 | unsigned int i; | |
1255 | ||
7e63c87f | 1256 | if (!ofs->config.nfs_export && !ofs->upper_mnt) |
9df085f3 AG |
1257 | return true; |
1258 | ||
1b81dddd | 1259 | for (i = 0; i < ofs->numfs; i++) { |
9df085f3 AG |
1260 | /* |
1261 | * We use uuid to associate an overlay lower file handle with a | |
1262 | * lower layer, so we can accept lower fs with null uuid as long | |
1263 | * as all lower layers with null uuid are on the same fs. | |
7e63c87f AG |
1264 | * if we detect multiple lower fs with the same uuid, we |
1265 | * disable lower file handle decoding on all of them. | |
9df085f3 | 1266 | */ |
1b81dddd AG |
1267 | if (ofs->fs[i].is_lower && |
1268 | uuid_equal(&ofs->fs[i].sb->s_uuid, uuid)) { | |
07f1e596 | 1269 | ofs->fs[i].bad_uuid = true; |
9df085f3 | 1270 | return false; |
7e63c87f | 1271 | } |
9df085f3 AG |
1272 | } |
1273 | return true; | |
1274 | } | |
1275 | ||
5148626b | 1276 | /* Get a unique fsid for the layer */ |
9df085f3 | 1277 | static int ovl_get_fsid(struct ovl_fs *ofs, const struct path *path) |
5148626b | 1278 | { |
9df085f3 | 1279 | struct super_block *sb = path->mnt->mnt_sb; |
5148626b AG |
1280 | unsigned int i; |
1281 | dev_t dev; | |
1282 | int err; | |
7e63c87f | 1283 | bool bad_uuid = false; |
5148626b | 1284 | |
07f1e596 AG |
1285 | for (i = 0; i < ofs->numfs; i++) { |
1286 | if (ofs->fs[i].sb == sb) | |
1287 | return i; | |
5148626b AG |
1288 | } |
1289 | ||
9df085f3 | 1290 | if (!ovl_lower_uuid_ok(ofs, &sb->s_uuid)) { |
7e63c87f AG |
1291 | bad_uuid = true; |
1292 | if (ofs->config.index || ofs->config.nfs_export) { | |
1293 | ofs->config.index = false; | |
1294 | ofs->config.nfs_export = false; | |
1bd0a3ae | 1295 | pr_warn("%s uuid detected in lower fs '%pd2', falling back to index=off,nfs_export=off.\n", |
7e63c87f AG |
1296 | uuid_is_null(&sb->s_uuid) ? "null" : |
1297 | "conflicting", | |
1298 | path->dentry); | |
1299 | } | |
9df085f3 AG |
1300 | } |
1301 | ||
5148626b AG |
1302 | err = get_anon_bdev(&dev); |
1303 | if (err) { | |
1bd0a3ae | 1304 | pr_err("failed to get anonymous bdev for lowerpath\n"); |
5148626b AG |
1305 | return err; |
1306 | } | |
1307 | ||
07f1e596 AG |
1308 | ofs->fs[ofs->numfs].sb = sb; |
1309 | ofs->fs[ofs->numfs].pseudo_dev = dev; | |
1310 | ofs->fs[ofs->numfs].bad_uuid = bad_uuid; | |
5148626b | 1311 | |
07f1e596 | 1312 | return ofs->numfs++; |
5148626b AG |
1313 | } |
1314 | ||
94375f9d AG |
1315 | static int ovl_get_layers(struct super_block *sb, struct ovl_fs *ofs, |
1316 | struct path *stack, unsigned int numlower) | |
520d7c86 MS |
1317 | { |
1318 | int err; | |
1319 | unsigned int i; | |
13464165 | 1320 | struct ovl_layer *layers; |
520d7c86 MS |
1321 | |
1322 | err = -ENOMEM; | |
13464165 MS |
1323 | layers = kcalloc(numlower + 1, sizeof(struct ovl_layer), GFP_KERNEL); |
1324 | if (!layers) | |
520d7c86 | 1325 | goto out; |
13464165 | 1326 | ofs->layers = layers; |
5148626b | 1327 | |
07f1e596 AG |
1328 | ofs->fs = kcalloc(numlower + 1, sizeof(struct ovl_sb), GFP_KERNEL); |
1329 | if (ofs->fs == NULL) | |
5148626b AG |
1330 | goto out; |
1331 | ||
07f1e596 AG |
1332 | /* idx/fsid 0 are reserved for upper fs even with lower only overlay */ |
1333 | ofs->numfs++; | |
1334 | ||
13464165 MS |
1335 | layers[0].mnt = ofs->upper_mnt; |
1336 | layers[0].idx = 0; | |
1337 | layers[0].fsid = 0; | |
94375f9d AG |
1338 | ofs->numlayer = 1; |
1339 | ||
07f1e596 | 1340 | /* |
b7bf9908 AG |
1341 | * All lower layers that share the same fs as upper layer, use the same |
1342 | * pseudo_dev as upper layer. Allocate fs[0].pseudo_dev even for lower | |
1343 | * only overlay to simplify ovl_fs_free(). | |
1b81dddd | 1344 | * is_lower will be set if upper fs is shared with a lower layer. |
07f1e596 | 1345 | */ |
b7bf9908 AG |
1346 | err = get_anon_bdev(&ofs->fs[0].pseudo_dev); |
1347 | if (err) { | |
1348 | pr_err("failed to get anonymous bdev for upper fs\n"); | |
1349 | goto out; | |
1350 | } | |
1351 | ||
07f1e596 AG |
1352 | if (ofs->upper_mnt) { |
1353 | ofs->fs[0].sb = ofs->upper_mnt->mnt_sb; | |
1b81dddd | 1354 | ofs->fs[0].is_lower = false; |
07f1e596 AG |
1355 | } |
1356 | ||
520d7c86 MS |
1357 | for (i = 0; i < numlower; i++) { |
1358 | struct vfsmount *mnt; | |
146d62e5 | 1359 | struct inode *trap; |
5148626b | 1360 | int fsid; |
520d7c86 | 1361 | |
9df085f3 | 1362 | err = fsid = ovl_get_fsid(ofs, &stack[i]); |
5148626b | 1363 | if (err < 0) |
520d7c86 | 1364 | goto out; |
520d7c86 | 1365 | |
146d62e5 AG |
1366 | err = ovl_setup_trap(sb, stack[i].dentry, &trap, "lowerdir"); |
1367 | if (err) | |
1368 | goto out; | |
1369 | ||
0be0bfd2 AG |
1370 | if (ovl_is_inuse(stack[i].dentry)) { |
1371 | err = ovl_report_in_use(ofs, "lowerdir"); | |
1372 | if (err) | |
1373 | goto out; | |
1374 | } | |
1375 | ||
520d7c86 MS |
1376 | mnt = clone_private_mount(&stack[i]); |
1377 | err = PTR_ERR(mnt); | |
1378 | if (IS_ERR(mnt)) { | |
1bd0a3ae | 1379 | pr_err("failed to clone lowerpath\n"); |
146d62e5 | 1380 | iput(trap); |
520d7c86 MS |
1381 | goto out; |
1382 | } | |
5148626b | 1383 | |
520d7c86 MS |
1384 | /* |
1385 | * Make lower layers R/O. That way fchmod/fchown on lower file | |
1386 | * will fail instead of modifying lower fs. | |
1387 | */ | |
1388 | mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME; | |
1389 | ||
13464165 MS |
1390 | layers[ofs->numlayer].trap = trap; |
1391 | layers[ofs->numlayer].mnt = mnt; | |
1392 | layers[ofs->numlayer].idx = ofs->numlayer; | |
1393 | layers[ofs->numlayer].fsid = fsid; | |
1394 | layers[ofs->numlayer].fs = &ofs->fs[fsid]; | |
94375f9d | 1395 | ofs->numlayer++; |
1b81dddd | 1396 | ofs->fs[fsid].is_lower = true; |
520d7c86 | 1397 | } |
e487d889 | 1398 | |
795939a9 AG |
1399 | /* |
1400 | * When all layers on same fs, overlay can use real inode numbers. | |
1401 | * With mount option "xino=on", mounter declares that there are enough | |
1402 | * free high bits in underlying fs to hold the unique fsid. | |
1403 | * If overlayfs does encounter underlying inodes using the high xino | |
1404 | * bits reserved for fsid, it emits a warning and uses the original | |
1405 | * inode number. | |
1406 | */ | |
07f1e596 | 1407 | if (ofs->numfs - !ofs->upper_mnt == 1) { |
0f831ec8 AG |
1408 | if (ofs->config.xino == OVL_XINO_ON) |
1409 | pr_info("\"xino=on\" is useless with all layers on same fs, ignore.\n"); | |
1410 | ofs->xino_mode = 0; | |
53afcd31 AG |
1411 | } else if (ofs->config.xino == OVL_XINO_OFF) { |
1412 | ofs->xino_mode = -1; | |
0f831ec8 | 1413 | } else if (ofs->config.xino == OVL_XINO_ON && ofs->xino_mode < 0) { |
795939a9 | 1414 | /* |
07f1e596 AG |
1415 | * This is a roundup of number of bits needed for encoding |
1416 | * fsid, where fsid 0 is reserved for upper fs even with | |
1417 | * lower only overlay. | |
795939a9 AG |
1418 | */ |
1419 | BUILD_BUG_ON(ilog2(OVL_MAX_STACK) > 31); | |
07f1e596 | 1420 | ofs->xino_mode = ilog2(ofs->numfs - 1) + 1; |
795939a9 AG |
1421 | } |
1422 | ||
0f831ec8 | 1423 | if (ofs->xino_mode > 0) { |
1bd0a3ae | 1424 | pr_info("\"xino\" feature enabled using %d upper inode bits.\n", |
0f831ec8 | 1425 | ofs->xino_mode); |
795939a9 | 1426 | } |
e487d889 | 1427 | |
520d7c86 MS |
1428 | err = 0; |
1429 | out: | |
1430 | return err; | |
1431 | } | |
1432 | ||
4155c10a | 1433 | static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb, |
ad204488 | 1434 | struct ovl_fs *ofs) |
53dbb0b4 MS |
1435 | { |
1436 | int err; | |
1437 | char *lowertmp, *lower; | |
4155c10a MS |
1438 | struct path *stack = NULL; |
1439 | unsigned int stacklen, numlower = 0, i; | |
4155c10a | 1440 | struct ovl_entry *oe; |
53dbb0b4 MS |
1441 | |
1442 | err = -ENOMEM; | |
ad204488 | 1443 | lowertmp = kstrdup(ofs->config.lowerdir, GFP_KERNEL); |
53dbb0b4 | 1444 | if (!lowertmp) |
4155c10a | 1445 | goto out_err; |
53dbb0b4 MS |
1446 | |
1447 | err = -EINVAL; | |
1448 | stacklen = ovl_split_lowerdirs(lowertmp); | |
1449 | if (stacklen > OVL_MAX_STACK) { | |
1bd0a3ae | 1450 | pr_err("too many lower directories, limit is %d\n", |
53dbb0b4 | 1451 | OVL_MAX_STACK); |
4155c10a | 1452 | goto out_err; |
ad204488 | 1453 | } else if (!ofs->config.upperdir && stacklen == 1) { |
1bd0a3ae | 1454 | pr_err("at least 2 lowerdir are needed while upperdir nonexistent\n"); |
4155c10a | 1455 | goto out_err; |
f168f109 AG |
1456 | } else if (!ofs->config.upperdir && ofs->config.nfs_export && |
1457 | ofs->config.redirect_follow) { | |
1bd0a3ae | 1458 | pr_warn("NFS export requires \"redirect_dir=nofollow\" on non-upper mount, falling back to nfs_export=off.\n"); |
f168f109 | 1459 | ofs->config.nfs_export = false; |
53dbb0b4 MS |
1460 | } |
1461 | ||
1462 | err = -ENOMEM; | |
1463 | stack = kcalloc(stacklen, sizeof(struct path), GFP_KERNEL); | |
1464 | if (!stack) | |
4155c10a | 1465 | goto out_err; |
53dbb0b4 MS |
1466 | |
1467 | err = -EINVAL; | |
1468 | lower = lowertmp; | |
1469 | for (numlower = 0; numlower < stacklen; numlower++) { | |
ad204488 | 1470 | err = ovl_lower_dir(lower, &stack[numlower], ofs, |
f4288844 | 1471 | &sb->s_stack_depth); |
53dbb0b4 | 1472 | if (err) |
4155c10a | 1473 | goto out_err; |
53dbb0b4 MS |
1474 | |
1475 | lower = strchr(lower, '\0') + 1; | |
1476 | } | |
1477 | ||
1478 | err = -EINVAL; | |
1479 | sb->s_stack_depth++; | |
1480 | if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) { | |
1bd0a3ae | 1481 | pr_err("maximum fs stacking depth exceeded\n"); |
4155c10a | 1482 | goto out_err; |
53dbb0b4 MS |
1483 | } |
1484 | ||
94375f9d | 1485 | err = ovl_get_layers(sb, ofs, stack, numlower); |
4155c10a MS |
1486 | if (err) |
1487 | goto out_err; | |
1488 | ||
1489 | err = -ENOMEM; | |
1490 | oe = ovl_alloc_entry(numlower); | |
1491 | if (!oe) | |
1492 | goto out_err; | |
1493 | ||
1494 | for (i = 0; i < numlower; i++) { | |
1495 | oe->lowerstack[i].dentry = dget(stack[i].dentry); | |
94375f9d | 1496 | oe->lowerstack[i].layer = &ofs->layers[i+1]; |
4155c10a | 1497 | } |
53dbb0b4 | 1498 | |
53dbb0b4 | 1499 | out: |
53dbb0b4 MS |
1500 | for (i = 0; i < numlower; i++) |
1501 | path_put(&stack[i]); | |
1502 | kfree(stack); | |
4155c10a MS |
1503 | kfree(lowertmp); |
1504 | ||
1505 | return oe; | |
1506 | ||
1507 | out_err: | |
1508 | oe = ERR_PTR(err); | |
53dbb0b4 MS |
1509 | goto out; |
1510 | } | |
1511 | ||
146d62e5 AG |
1512 | /* |
1513 | * Check if this layer root is a descendant of: | |
1514 | * - another layer of this overlayfs instance | |
1515 | * - upper/work dir of any overlayfs instance | |
146d62e5 | 1516 | */ |
0be0bfd2 AG |
1517 | static int ovl_check_layer(struct super_block *sb, struct ovl_fs *ofs, |
1518 | struct dentry *dentry, const char *name) | |
146d62e5 | 1519 | { |
9179c21d | 1520 | struct dentry *next = dentry, *parent; |
146d62e5 AG |
1521 | int err = 0; |
1522 | ||
9179c21d | 1523 | if (!dentry) |
146d62e5 AG |
1524 | return 0; |
1525 | ||
9179c21d MS |
1526 | parent = dget_parent(next); |
1527 | ||
1528 | /* Walk back ancestors to root (inclusive) looking for traps */ | |
1529 | while (!err && parent != next) { | |
0be0bfd2 | 1530 | if (ovl_lookup_trap_inode(sb, parent)) { |
146d62e5 | 1531 | err = -ELOOP; |
1bd0a3ae | 1532 | pr_err("overlapping %s path\n", name); |
0be0bfd2 AG |
1533 | } else if (ovl_is_inuse(parent)) { |
1534 | err = ovl_report_in_use(ofs, name); | |
146d62e5 | 1535 | } |
146d62e5 | 1536 | next = parent; |
9179c21d MS |
1537 | parent = dget_parent(next); |
1538 | dput(next); | |
146d62e5 AG |
1539 | } |
1540 | ||
9179c21d | 1541 | dput(parent); |
146d62e5 AG |
1542 | |
1543 | return err; | |
1544 | } | |
1545 | ||
1546 | /* | |
1547 | * Check if any of the layers or work dirs overlap. | |
1548 | */ | |
1549 | static int ovl_check_overlapping_layers(struct super_block *sb, | |
1550 | struct ovl_fs *ofs) | |
1551 | { | |
1552 | int i, err; | |
1553 | ||
1554 | if (ofs->upper_mnt) { | |
0be0bfd2 AG |
1555 | err = ovl_check_layer(sb, ofs, ofs->upper_mnt->mnt_root, |
1556 | "upperdir"); | |
146d62e5 AG |
1557 | if (err) |
1558 | return err; | |
1559 | ||
1560 | /* | |
1561 | * Checking workbasedir avoids hitting ovl_is_inuse(parent) of | |
1562 | * this instance and covers overlapping work and index dirs, | |
1563 | * unless work or index dir have been moved since created inside | |
1564 | * workbasedir. In that case, we already have their traps in | |
1565 | * inode cache and we will catch that case on lookup. | |
1566 | */ | |
0be0bfd2 | 1567 | err = ovl_check_layer(sb, ofs, ofs->workbasedir, "workdir"); |
146d62e5 AG |
1568 | if (err) |
1569 | return err; | |
1570 | } | |
1571 | ||
94375f9d | 1572 | for (i = 1; i < ofs->numlayer; i++) { |
0be0bfd2 | 1573 | err = ovl_check_layer(sb, ofs, |
94375f9d | 1574 | ofs->layers[i].mnt->mnt_root, |
146d62e5 AG |
1575 | "lowerdir"); |
1576 | if (err) | |
1577 | return err; | |
1578 | } | |
1579 | ||
1580 | return 0; | |
1581 | } | |
1582 | ||
2effc5c2 AG |
1583 | static struct dentry *ovl_get_root(struct super_block *sb, |
1584 | struct dentry *upperdentry, | |
1585 | struct ovl_entry *oe) | |
1586 | { | |
1587 | struct dentry *root; | |
62c832ed AG |
1588 | struct ovl_path *lowerpath = &oe->lowerstack[0]; |
1589 | unsigned long ino = d_inode(lowerpath->dentry)->i_ino; | |
1590 | int fsid = lowerpath->layer->fsid; | |
1591 | struct ovl_inode_params oip = { | |
1592 | .upperdentry = upperdentry, | |
1593 | .lowerpath = lowerpath, | |
1594 | }; | |
2effc5c2 AG |
1595 | |
1596 | root = d_make_root(ovl_new_inode(sb, S_IFDIR, 0)); | |
1597 | if (!root) | |
1598 | return NULL; | |
1599 | ||
1600 | root->d_fsdata = oe; | |
1601 | ||
1602 | if (upperdentry) { | |
62c832ed AG |
1603 | /* Root inode uses upper st_ino/i_ino */ |
1604 | ino = d_inode(upperdentry)->i_ino; | |
1605 | fsid = 0; | |
2effc5c2 AG |
1606 | ovl_dentry_set_upper_alias(root); |
1607 | if (ovl_is_impuredir(upperdentry)) | |
1608 | ovl_set_flag(OVL_IMPURE, d_inode(root)); | |
1609 | } | |
1610 | ||
1611 | /* Root is always merge -> can have whiteouts */ | |
1612 | ovl_set_flag(OVL_WHITEOUTS, d_inode(root)); | |
1613 | ovl_dentry_set_flag(OVL_E_CONNECTED, root); | |
1614 | ovl_set_upperdata(d_inode(root)); | |
62c832ed | 1615 | ovl_inode_init(d_inode(root), &oip, ino, fsid); |
f4288844 | 1616 | ovl_dentry_update_reval(root, upperdentry, DCACHE_OP_WEAK_REVALIDATE); |
2effc5c2 AG |
1617 | |
1618 | return root; | |
1619 | } | |
1620 | ||
e9be9d5e MS |
1621 | static int ovl_fill_super(struct super_block *sb, void *data, int silent) |
1622 | { | |
33006cdf | 1623 | struct path upperpath = { }; |
e9be9d5e | 1624 | struct dentry *root_dentry; |
4155c10a | 1625 | struct ovl_entry *oe; |
ad204488 | 1626 | struct ovl_fs *ofs; |
51f8f3c4 | 1627 | struct cred *cred; |
e9be9d5e MS |
1628 | int err; |
1629 | ||
f4288844 MS |
1630 | sb->s_d_op = &ovl_dentry_operations; |
1631 | ||
f45827e8 | 1632 | err = -ENOMEM; |
ad204488 MS |
1633 | ofs = kzalloc(sizeof(struct ovl_fs), GFP_KERNEL); |
1634 | if (!ofs) | |
e9be9d5e MS |
1635 | goto out; |
1636 | ||
ad204488 | 1637 | ofs->creator_cred = cred = prepare_creds(); |
c6fe6254 MS |
1638 | if (!cred) |
1639 | goto out_err; | |
1640 | ||
ad204488 | 1641 | ofs->config.index = ovl_index_def; |
f168f109 | 1642 | ofs->config.nfs_export = ovl_nfs_export_def; |
795939a9 | 1643 | ofs->config.xino = ovl_xino_def(); |
d5791044 | 1644 | ofs->config.metacopy = ovl_metacopy_def; |
ad204488 | 1645 | err = ovl_parse_opt((char *) data, &ofs->config); |
f45827e8 | 1646 | if (err) |
a9075cdb | 1647 | goto out_err; |
f45827e8 | 1648 | |
e9be9d5e | 1649 | err = -EINVAL; |
ad204488 | 1650 | if (!ofs->config.lowerdir) { |
07f2af7b | 1651 | if (!silent) |
1bd0a3ae | 1652 | pr_err("missing 'lowerdir'\n"); |
a9075cdb | 1653 | goto out_err; |
e9be9d5e MS |
1654 | } |
1655 | ||
53a08cb9 | 1656 | sb->s_stack_depth = 0; |
cf9a6784 | 1657 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
e487d889 | 1658 | /* Assume underlaying fs uses 32bit inodes unless proven otherwise */ |
53afcd31 | 1659 | if (ofs->config.xino != OVL_XINO_OFF) { |
0f831ec8 | 1660 | ofs->xino_mode = BITS_PER_LONG - 32; |
53afcd31 AG |
1661 | if (!ofs->xino_mode) { |
1662 | pr_warn("xino not supported on 32bit kernel, falling back to xino=off.\n"); | |
1663 | ofs->config.xino = OVL_XINO_OFF; | |
1664 | } | |
1665 | } | |
795939a9 | 1666 | |
146d62e5 AG |
1667 | /* alloc/destroy_inode needed for setting up traps in inode cache */ |
1668 | sb->s_op = &ovl_super_operations; | |
1669 | ||
ad204488 MS |
1670 | if (ofs->config.upperdir) { |
1671 | if (!ofs->config.workdir) { | |
1bd0a3ae | 1672 | pr_err("missing 'workdir'\n"); |
a9075cdb | 1673 | goto out_err; |
53a08cb9 | 1674 | } |
e9be9d5e | 1675 | |
146d62e5 | 1676 | err = ovl_get_upper(sb, ofs, &upperpath); |
53a08cb9 | 1677 | if (err) |
a9075cdb | 1678 | goto out_err; |
2cac0c00 | 1679 | |
146d62e5 | 1680 | err = ovl_get_workdir(sb, ofs, &upperpath); |
8ed61dc3 | 1681 | if (err) |
a9075cdb | 1682 | goto out_err; |
c6fe6254 | 1683 | |
ad204488 | 1684 | if (!ofs->workdir) |
1751e8a6 | 1685 | sb->s_flags |= SB_RDONLY; |
6e88256e | 1686 | |
ad204488 MS |
1687 | sb->s_stack_depth = ofs->upper_mnt->mnt_sb->s_stack_depth; |
1688 | sb->s_time_gran = ofs->upper_mnt->mnt_sb->s_time_gran; | |
c6fe6254 | 1689 | |
e9be9d5e | 1690 | } |
ad204488 | 1691 | oe = ovl_get_lowerstack(sb, ofs); |
4155c10a MS |
1692 | err = PTR_ERR(oe); |
1693 | if (IS_ERR(oe)) | |
a9075cdb | 1694 | goto out_err; |
e9be9d5e | 1695 | |
71cbad7e | 1696 | /* If the upper fs is nonexistent, we mark overlayfs r/o too */ |
ad204488 | 1697 | if (!ofs->upper_mnt) |
1751e8a6 | 1698 | sb->s_flags |= SB_RDONLY; |
e9be9d5e | 1699 | |
ad204488 | 1700 | if (!(ovl_force_readonly(ofs)) && ofs->config.index) { |
146d62e5 | 1701 | err = ovl_get_indexdir(sb, ofs, oe, &upperpath); |
54fb347e | 1702 | if (err) |
4155c10a | 1703 | goto out_free_oe; |
6e88256e | 1704 | |
972d0093 AG |
1705 | /* Force r/o mount with no index dir */ |
1706 | if (!ofs->indexdir) { | |
1707 | dput(ofs->workdir); | |
1708 | ofs->workdir = NULL; | |
1751e8a6 | 1709 | sb->s_flags |= SB_RDONLY; |
972d0093 AG |
1710 | } |
1711 | ||
02bcd157 AG |
1712 | } |
1713 | ||
146d62e5 AG |
1714 | err = ovl_check_overlapping_layers(sb, ofs); |
1715 | if (err) | |
1716 | goto out_free_oe; | |
1717 | ||
972d0093 | 1718 | /* Show index=off in /proc/mounts for forced r/o mount */ |
f168f109 | 1719 | if (!ofs->indexdir) { |
ad204488 | 1720 | ofs->config.index = false; |
f168f109 | 1721 | if (ofs->upper_mnt && ofs->config.nfs_export) { |
1bd0a3ae | 1722 | pr_warn("NFS export requires an index dir, falling back to nfs_export=off.\n"); |
f168f109 AG |
1723 | ofs->config.nfs_export = false; |
1724 | } | |
1725 | } | |
02bcd157 | 1726 | |
d5791044 | 1727 | if (ofs->config.metacopy && ofs->config.nfs_export) { |
1bd0a3ae | 1728 | pr_warn("NFS export is not supported with metadata only copy up, falling back to nfs_export=off.\n"); |
d5791044 VG |
1729 | ofs->config.nfs_export = false; |
1730 | } | |
1731 | ||
8383f174 AG |
1732 | if (ofs->config.nfs_export) |
1733 | sb->s_export_op = &ovl_export_operations; | |
1734 | ||
51f8f3c4 KK |
1735 | /* Never override disk quota limits or use reserved space */ |
1736 | cap_lower(cred->cap_effective, CAP_SYS_RESOURCE); | |
1737 | ||
655042cc | 1738 | sb->s_magic = OVERLAYFS_SUPER_MAGIC; |
655042cc | 1739 | sb->s_xattr = ovl_xattr_handlers; |
ad204488 | 1740 | sb->s_fs_info = ofs; |
de2a4a50 | 1741 | sb->s_flags |= SB_POSIXACL; |
655042cc | 1742 | |
c6fe6254 | 1743 | err = -ENOMEM; |
2effc5c2 | 1744 | root_dentry = ovl_get_root(sb, upperpath.dentry, oe); |
e9be9d5e | 1745 | if (!root_dentry) |
4155c10a | 1746 | goto out_free_oe; |
e9be9d5e MS |
1747 | |
1748 | mntput(upperpath.mnt); | |
ed06e069 | 1749 | |
e9be9d5e | 1750 | sb->s_root = root_dentry; |
e9be9d5e MS |
1751 | |
1752 | return 0; | |
1753 | ||
4155c10a MS |
1754 | out_free_oe: |
1755 | ovl_entry_stack_free(oe); | |
b9343632 | 1756 | kfree(oe); |
4155c10a | 1757 | out_err: |
e9be9d5e | 1758 | path_put(&upperpath); |
ad204488 | 1759 | ovl_free_fs(ofs); |
e9be9d5e MS |
1760 | out: |
1761 | return err; | |
1762 | } | |
1763 | ||
1764 | static struct dentry *ovl_mount(struct file_system_type *fs_type, int flags, | |
1765 | const char *dev_name, void *raw_data) | |
1766 | { | |
1767 | return mount_nodev(fs_type, flags, raw_data, ovl_fill_super); | |
1768 | } | |
1769 | ||
1770 | static struct file_system_type ovl_fs_type = { | |
1771 | .owner = THIS_MODULE, | |
ef94b186 | 1772 | .name = "overlay", |
e9be9d5e MS |
1773 | .mount = ovl_mount, |
1774 | .kill_sb = kill_anon_super, | |
1775 | }; | |
ef94b186 | 1776 | MODULE_ALIAS_FS("overlay"); |
e9be9d5e | 1777 | |
13cf199d AG |
1778 | static void ovl_inode_init_once(void *foo) |
1779 | { | |
1780 | struct ovl_inode *oi = foo; | |
1781 | ||
1782 | inode_init_once(&oi->vfs_inode); | |
1783 | } | |
1784 | ||
e9be9d5e MS |
1785 | static int __init ovl_init(void) |
1786 | { | |
13cf199d AG |
1787 | int err; |
1788 | ||
1789 | ovl_inode_cachep = kmem_cache_create("ovl_inode", | |
1790 | sizeof(struct ovl_inode), 0, | |
1791 | (SLAB_RECLAIM_ACCOUNT| | |
1792 | SLAB_MEM_SPREAD|SLAB_ACCOUNT), | |
1793 | ovl_inode_init_once); | |
1794 | if (ovl_inode_cachep == NULL) | |
1795 | return -ENOMEM; | |
1796 | ||
2406a307 JX |
1797 | err = ovl_aio_request_cache_init(); |
1798 | if (!err) { | |
1799 | err = register_filesystem(&ovl_fs_type); | |
1800 | if (!err) | |
1801 | return 0; | |
1802 | ||
1803 | ovl_aio_request_cache_destroy(); | |
1804 | } | |
1805 | kmem_cache_destroy(ovl_inode_cachep); | |
13cf199d AG |
1806 | |
1807 | return err; | |
e9be9d5e MS |
1808 | } |
1809 | ||
1810 | static void __exit ovl_exit(void) | |
1811 | { | |
1812 | unregister_filesystem(&ovl_fs_type); | |
13cf199d AG |
1813 | |
1814 | /* | |
1815 | * Make sure all delayed rcu free inodes are flushed before we | |
1816 | * destroy cache. | |
1817 | */ | |
1818 | rcu_barrier(); | |
1819 | kmem_cache_destroy(ovl_inode_cachep); | |
2406a307 | 1820 | ovl_aio_request_cache_destroy(); |
e9be9d5e MS |
1821 | } |
1822 | ||
1823 | module_init(ovl_init); | |
1824 | module_exit(ovl_exit); |