]>
Commit | Line | Data |
---|---|---|
e9be9d5e MS |
1 | /* |
2 | * | |
3 | * Copyright (C) 2011 Novell Inc. | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify it | |
6 | * under the terms of the GNU General Public License version 2 as published by | |
7 | * the Free Software Foundation. | |
8 | */ | |
9 | ||
5b825c3a | 10 | #include <uapi/linux/magic.h> |
e9be9d5e MS |
11 | #include <linux/fs.h> |
12 | #include <linux/namei.h> | |
13 | #include <linux/xattr.h> | |
e9be9d5e | 14 | #include <linux/mount.h> |
e9be9d5e MS |
15 | #include <linux/parser.h> |
16 | #include <linux/module.h> | |
cc259639 | 17 | #include <linux/statfs.h> |
f45827e8 | 18 | #include <linux/seq_file.h> |
d837a49b | 19 | #include <linux/posix_acl_xattr.h> |
e9be9d5e MS |
20 | #include "overlayfs.h" |
21 | ||
22 | MODULE_AUTHOR("Miklos Szeredi <miklos@szeredi.hu>"); | |
23 | MODULE_DESCRIPTION("Overlay filesystem"); | |
24 | MODULE_LICENSE("GPL"); | |
25 | ||
e9be9d5e MS |
26 | |
27 | struct ovl_dir_cache; | |
28 | ||
a78d9f0d MS |
29 | #define OVL_MAX_STACK 500 |
30 | ||
688ea0e5 MS |
31 | static bool ovl_redirect_dir_def = IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_DIR); |
32 | module_param_named(redirect_dir, ovl_redirect_dir_def, bool, 0644); | |
33 | MODULE_PARM_DESC(ovl_redirect_dir_def, | |
34 | "Default to on or off for the redirect_dir feature"); | |
e9be9d5e | 35 | |
02bcd157 AG |
36 | static bool ovl_index_def = IS_ENABLED(CONFIG_OVERLAY_FS_INDEX); |
37 | module_param_named(index, ovl_index_def, bool, 0644); | |
38 | MODULE_PARM_DESC(ovl_index_def, | |
39 | "Default to on or off for the inodes index feature"); | |
40 | ||
4155c10a MS |
41 | static void ovl_entry_stack_free(struct ovl_entry *oe) |
42 | { | |
43 | unsigned int i; | |
44 | ||
45 | for (i = 0; i < oe->numlower; i++) | |
46 | dput(oe->lowerstack[i].dentry); | |
47 | } | |
48 | ||
e9be9d5e MS |
49 | static void ovl_dentry_release(struct dentry *dentry) |
50 | { | |
51 | struct ovl_entry *oe = dentry->d_fsdata; | |
52 | ||
53 | if (oe) { | |
4155c10a | 54 | ovl_entry_stack_free(oe); |
e9be9d5e MS |
55 | kfree_rcu(oe, rcu); |
56 | } | |
57 | } | |
58 | ||
b0990fbb AG |
59 | static int ovl_check_append_only(struct inode *inode, int flag) |
60 | { | |
61 | /* | |
62 | * This test was moot in vfs may_open() because overlay inode does | |
63 | * not have the S_APPEND flag, so re-check on real upper inode | |
64 | */ | |
65 | if (IS_APPEND(inode)) { | |
66 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) | |
67 | return -EPERM; | |
68 | if (flag & O_TRUNC) | |
69 | return -EPERM; | |
70 | } | |
71 | ||
72 | return 0; | |
73 | } | |
74 | ||
2d902671 MS |
75 | static struct dentry *ovl_d_real(struct dentry *dentry, |
76 | const struct inode *inode, | |
495e6429 | 77 | unsigned int open_flags, unsigned int flags) |
d101a125 MS |
78 | { |
79 | struct dentry *real; | |
b0990fbb | 80 | int err; |
d101a125 | 81 | |
cd91304e MS |
82 | if (flags & D_REAL_UPPER) |
83 | return ovl_dentry_upper(dentry); | |
84 | ||
ca4c8a3a | 85 | if (!d_is_reg(dentry)) { |
d101a125 MS |
86 | if (!inode || inode == d_inode(dentry)) |
87 | return dentry; | |
88 | goto bug; | |
89 | } | |
90 | ||
2d902671 | 91 | if (open_flags) { |
b0990fbb | 92 | err = ovl_open_maybe_copy_up(dentry, open_flags); |
2d902671 MS |
93 | if (err) |
94 | return ERR_PTR(err); | |
95 | } | |
96 | ||
d101a125 | 97 | real = ovl_dentry_upper(dentry); |
b0990fbb AG |
98 | if (real && (!inode || inode == d_inode(real))) { |
99 | if (!inode) { | |
100 | err = ovl_check_append_only(d_inode(real), open_flags); | |
101 | if (err) | |
102 | return ERR_PTR(err); | |
103 | } | |
d101a125 | 104 | return real; |
b0990fbb | 105 | } |
d101a125 MS |
106 | |
107 | real = ovl_dentry_lower(dentry); | |
108 | if (!real) | |
109 | goto bug; | |
110 | ||
c4fcfc16 | 111 | /* Handle recursion */ |
495e6429 | 112 | real = d_real(real, inode, open_flags, 0); |
c4fcfc16 | 113 | |
d101a125 MS |
114 | if (!inode || inode == d_inode(real)) |
115 | return real; | |
d101a125 | 116 | bug: |
656189d2 | 117 | WARN(1, "ovl_d_real(%pd4, %s:%lu): real dentry not found\n", dentry, |
d101a125 MS |
118 | inode ? inode->i_sb->s_id : "NULL", inode ? inode->i_ino : 0); |
119 | return dentry; | |
120 | } | |
121 | ||
7c03b5d4 MS |
122 | static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags) |
123 | { | |
124 | struct ovl_entry *oe = dentry->d_fsdata; | |
125 | unsigned int i; | |
126 | int ret = 1; | |
127 | ||
128 | for (i = 0; i < oe->numlower; i++) { | |
129 | struct dentry *d = oe->lowerstack[i].dentry; | |
130 | ||
131 | if (d->d_flags & DCACHE_OP_REVALIDATE) { | |
132 | ret = d->d_op->d_revalidate(d, flags); | |
133 | if (ret < 0) | |
134 | return ret; | |
135 | if (!ret) { | |
136 | if (!(flags & LOOKUP_RCU)) | |
137 | d_invalidate(d); | |
138 | return -ESTALE; | |
139 | } | |
140 | } | |
141 | } | |
142 | return 1; | |
143 | } | |
144 | ||
145 | static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags) | |
146 | { | |
147 | struct ovl_entry *oe = dentry->d_fsdata; | |
148 | unsigned int i; | |
149 | int ret = 1; | |
150 | ||
151 | for (i = 0; i < oe->numlower; i++) { | |
152 | struct dentry *d = oe->lowerstack[i].dentry; | |
153 | ||
154 | if (d->d_flags & DCACHE_OP_WEAK_REVALIDATE) { | |
155 | ret = d->d_op->d_weak_revalidate(d, flags); | |
156 | if (ret <= 0) | |
157 | break; | |
158 | } | |
159 | } | |
160 | return ret; | |
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, |
e9be9d5e MS |
166 | }; |
167 | ||
7c03b5d4 MS |
168 | static const struct dentry_operations ovl_reval_dentry_operations = { |
169 | .d_release = ovl_dentry_release, | |
d101a125 | 170 | .d_real = ovl_d_real, |
7c03b5d4 MS |
171 | .d_revalidate = ovl_dentry_revalidate, |
172 | .d_weak_revalidate = ovl_dentry_weak_revalidate, | |
173 | }; | |
174 | ||
13cf199d AG |
175 | static struct kmem_cache *ovl_inode_cachep; |
176 | ||
177 | static struct inode *ovl_alloc_inode(struct super_block *sb) | |
178 | { | |
179 | struct ovl_inode *oi = kmem_cache_alloc(ovl_inode_cachep, GFP_KERNEL); | |
180 | ||
b3885bd6 HN |
181 | if (!oi) |
182 | return NULL; | |
183 | ||
04a01ac7 | 184 | oi->cache = NULL; |
cf31c463 | 185 | oi->redirect = NULL; |
04a01ac7 | 186 | oi->version = 0; |
13c72075 | 187 | oi->flags = 0; |
09d8b586 | 188 | oi->__upperdentry = NULL; |
25b7713a | 189 | oi->lower = NULL; |
a015dafc | 190 | mutex_init(&oi->lock); |
25b7713a | 191 | |
13cf199d AG |
192 | return &oi->vfs_inode; |
193 | } | |
194 | ||
195 | static void ovl_i_callback(struct rcu_head *head) | |
196 | { | |
197 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
198 | ||
199 | kmem_cache_free(ovl_inode_cachep, OVL_I(inode)); | |
200 | } | |
201 | ||
202 | static void ovl_destroy_inode(struct inode *inode) | |
203 | { | |
09d8b586 MS |
204 | struct ovl_inode *oi = OVL_I(inode); |
205 | ||
206 | dput(oi->__upperdentry); | |
cf31c463 | 207 | kfree(oi->redirect); |
4edb83bb | 208 | ovl_dir_cache_free(inode); |
a015dafc | 209 | mutex_destroy(&oi->lock); |
09d8b586 | 210 | |
13cf199d AG |
211 | call_rcu(&inode->i_rcu, ovl_i_callback); |
212 | } | |
213 | ||
ad204488 | 214 | static void ovl_free_fs(struct ovl_fs *ofs) |
e9be9d5e | 215 | { |
dd662667 | 216 | unsigned i; |
e9be9d5e | 217 | |
ad204488 MS |
218 | dput(ofs->indexdir); |
219 | dput(ofs->workdir); | |
220 | if (ofs->workdir_locked) | |
221 | ovl_inuse_unlock(ofs->workbasedir); | |
222 | dput(ofs->workbasedir); | |
223 | if (ofs->upperdir_locked) | |
224 | ovl_inuse_unlock(ofs->upper_mnt->mnt_root); | |
225 | mntput(ofs->upper_mnt); | |
226 | for (i = 0; i < ofs->numlower; i++) { | |
227 | mntput(ofs->lower_layers[i].mnt); | |
228 | free_anon_bdev(ofs->lower_layers[i].pseudo_dev); | |
2a9c6d06 | 229 | } |
ad204488 MS |
230 | kfree(ofs->lower_layers); |
231 | ||
232 | kfree(ofs->config.lowerdir); | |
233 | kfree(ofs->config.upperdir); | |
234 | kfree(ofs->config.workdir); | |
235 | if (ofs->creator_cred) | |
236 | put_cred(ofs->creator_cred); | |
237 | kfree(ofs); | |
e9be9d5e MS |
238 | } |
239 | ||
a9075cdb MS |
240 | static void ovl_put_super(struct super_block *sb) |
241 | { | |
242 | struct ovl_fs *ofs = sb->s_fs_info; | |
243 | ||
244 | ovl_free_fs(ofs); | |
245 | } | |
246 | ||
e593b2bf AG |
247 | static int ovl_sync_fs(struct super_block *sb, int wait) |
248 | { | |
ad204488 | 249 | struct ovl_fs *ofs = sb->s_fs_info; |
e593b2bf AG |
250 | struct super_block *upper_sb; |
251 | int ret; | |
252 | ||
ad204488 | 253 | if (!ofs->upper_mnt) |
e593b2bf | 254 | return 0; |
ad204488 | 255 | upper_sb = ofs->upper_mnt->mnt_sb; |
e593b2bf AG |
256 | if (!upper_sb->s_op->sync_fs) |
257 | return 0; | |
258 | ||
259 | /* real inodes have already been synced by sync_filesystem(ovl_sb) */ | |
260 | down_read(&upper_sb->s_umount); | |
261 | ret = upper_sb->s_op->sync_fs(upper_sb, wait); | |
262 | up_read(&upper_sb->s_umount); | |
263 | return ret; | |
264 | } | |
265 | ||
cc259639 AW |
266 | /** |
267 | * ovl_statfs | |
268 | * @sb: The overlayfs super block | |
269 | * @buf: The struct kstatfs to fill in with stats | |
270 | * | |
271 | * Get the filesystem statistics. As writes always target the upper layer | |
4ebc5818 | 272 | * filesystem pass the statfs to the upper filesystem (if it exists) |
cc259639 AW |
273 | */ |
274 | static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf) | |
275 | { | |
276 | struct ovl_fs *ofs = dentry->d_sb->s_fs_info; | |
277 | struct dentry *root_dentry = dentry->d_sb->s_root; | |
278 | struct path path; | |
279 | int err; | |
280 | ||
4ebc5818 | 281 | ovl_path_real(root_dentry, &path); |
cc259639 AW |
282 | |
283 | err = vfs_statfs(&path, buf); | |
284 | if (!err) { | |
6b2d5fe4 | 285 | buf->f_namelen = ofs->namelen; |
cc259639 AW |
286 | buf->f_type = OVERLAYFS_SUPER_MAGIC; |
287 | } | |
288 | ||
289 | return err; | |
290 | } | |
291 | ||
02bcd157 | 292 | /* Will this overlay be forced to mount/remount ro? */ |
ad204488 | 293 | static bool ovl_force_readonly(struct ovl_fs *ofs) |
02bcd157 | 294 | { |
ad204488 | 295 | return (!ofs->upper_mnt || !ofs->workdir); |
02bcd157 AG |
296 | } |
297 | ||
f45827e8 EZ |
298 | /** |
299 | * ovl_show_options | |
300 | * | |
301 | * Prints the mount options for a given superblock. | |
302 | * Returns zero; does not fail. | |
303 | */ | |
304 | static int ovl_show_options(struct seq_file *m, struct dentry *dentry) | |
305 | { | |
306 | struct super_block *sb = dentry->d_sb; | |
ad204488 | 307 | struct ovl_fs *ofs = sb->s_fs_info; |
f45827e8 | 308 | |
ad204488 MS |
309 | seq_show_option(m, "lowerdir", ofs->config.lowerdir); |
310 | if (ofs->config.upperdir) { | |
311 | seq_show_option(m, "upperdir", ofs->config.upperdir); | |
312 | seq_show_option(m, "workdir", ofs->config.workdir); | |
53a08cb9 | 313 | } |
ad204488 | 314 | if (ofs->config.default_permissions) |
8d3095f4 | 315 | seq_puts(m, ",default_permissions"); |
ad204488 | 316 | if (ofs->config.redirect_dir != ovl_redirect_dir_def) |
c5bef3a7 | 317 | seq_printf(m, ",redirect_dir=%s", |
ad204488 MS |
318 | ofs->config.redirect_dir ? "on" : "off"); |
319 | if (ofs->config.index != ovl_index_def) | |
02bcd157 | 320 | seq_printf(m, ",index=%s", |
ad204488 | 321 | ofs->config.index ? "on" : "off"); |
f45827e8 EZ |
322 | return 0; |
323 | } | |
324 | ||
3cdf6fe9 SL |
325 | static int ovl_remount(struct super_block *sb, int *flags, char *data) |
326 | { | |
ad204488 | 327 | struct ovl_fs *ofs = sb->s_fs_info; |
3cdf6fe9 | 328 | |
ad204488 | 329 | if (!(*flags & MS_RDONLY) && ovl_force_readonly(ofs)) |
3cdf6fe9 SL |
330 | return -EROFS; |
331 | ||
332 | return 0; | |
333 | } | |
334 | ||
e9be9d5e | 335 | static const struct super_operations ovl_super_operations = { |
13cf199d AG |
336 | .alloc_inode = ovl_alloc_inode, |
337 | .destroy_inode = ovl_destroy_inode, | |
338 | .drop_inode = generic_delete_inode, | |
e9be9d5e | 339 | .put_super = ovl_put_super, |
e593b2bf | 340 | .sync_fs = ovl_sync_fs, |
cc259639 | 341 | .statfs = ovl_statfs, |
f45827e8 | 342 | .show_options = ovl_show_options, |
3cdf6fe9 | 343 | .remount_fs = ovl_remount, |
e9be9d5e MS |
344 | }; |
345 | ||
346 | enum { | |
347 | OPT_LOWERDIR, | |
348 | OPT_UPPERDIR, | |
349 | OPT_WORKDIR, | |
8d3095f4 | 350 | OPT_DEFAULT_PERMISSIONS, |
a6c60655 MS |
351 | OPT_REDIRECT_DIR_ON, |
352 | OPT_REDIRECT_DIR_OFF, | |
02bcd157 AG |
353 | OPT_INDEX_ON, |
354 | OPT_INDEX_OFF, | |
e9be9d5e MS |
355 | OPT_ERR, |
356 | }; | |
357 | ||
358 | static const match_table_t ovl_tokens = { | |
359 | {OPT_LOWERDIR, "lowerdir=%s"}, | |
360 | {OPT_UPPERDIR, "upperdir=%s"}, | |
361 | {OPT_WORKDIR, "workdir=%s"}, | |
8d3095f4 | 362 | {OPT_DEFAULT_PERMISSIONS, "default_permissions"}, |
a6c60655 MS |
363 | {OPT_REDIRECT_DIR_ON, "redirect_dir=on"}, |
364 | {OPT_REDIRECT_DIR_OFF, "redirect_dir=off"}, | |
02bcd157 AG |
365 | {OPT_INDEX_ON, "index=on"}, |
366 | {OPT_INDEX_OFF, "index=off"}, | |
e9be9d5e MS |
367 | {OPT_ERR, NULL} |
368 | }; | |
369 | ||
91c77947 MS |
370 | static char *ovl_next_opt(char **s) |
371 | { | |
372 | char *sbegin = *s; | |
373 | char *p; | |
374 | ||
375 | if (sbegin == NULL) | |
376 | return NULL; | |
377 | ||
378 | for (p = sbegin; *p; p++) { | |
379 | if (*p == '\\') { | |
380 | p++; | |
381 | if (!*p) | |
382 | break; | |
383 | } else if (*p == ',') { | |
384 | *p = '\0'; | |
385 | *s = p + 1; | |
386 | return sbegin; | |
387 | } | |
388 | } | |
389 | *s = NULL; | |
390 | return sbegin; | |
391 | } | |
392 | ||
e9be9d5e MS |
393 | static int ovl_parse_opt(char *opt, struct ovl_config *config) |
394 | { | |
395 | char *p; | |
396 | ||
91c77947 | 397 | while ((p = ovl_next_opt(&opt)) != NULL) { |
e9be9d5e MS |
398 | int token; |
399 | substring_t args[MAX_OPT_ARGS]; | |
400 | ||
401 | if (!*p) | |
402 | continue; | |
403 | ||
404 | token = match_token(p, ovl_tokens, args); | |
405 | switch (token) { | |
406 | case OPT_UPPERDIR: | |
407 | kfree(config->upperdir); | |
408 | config->upperdir = match_strdup(&args[0]); | |
409 | if (!config->upperdir) | |
410 | return -ENOMEM; | |
411 | break; | |
412 | ||
413 | case OPT_LOWERDIR: | |
414 | kfree(config->lowerdir); | |
415 | config->lowerdir = match_strdup(&args[0]); | |
416 | if (!config->lowerdir) | |
417 | return -ENOMEM; | |
418 | break; | |
419 | ||
420 | case OPT_WORKDIR: | |
421 | kfree(config->workdir); | |
422 | config->workdir = match_strdup(&args[0]); | |
423 | if (!config->workdir) | |
424 | return -ENOMEM; | |
425 | break; | |
426 | ||
8d3095f4 MS |
427 | case OPT_DEFAULT_PERMISSIONS: |
428 | config->default_permissions = true; | |
429 | break; | |
430 | ||
a6c60655 MS |
431 | case OPT_REDIRECT_DIR_ON: |
432 | config->redirect_dir = true; | |
433 | break; | |
434 | ||
435 | case OPT_REDIRECT_DIR_OFF: | |
436 | config->redirect_dir = false; | |
437 | break; | |
438 | ||
02bcd157 AG |
439 | case OPT_INDEX_ON: |
440 | config->index = true; | |
441 | break; | |
442 | ||
443 | case OPT_INDEX_OFF: | |
444 | config->index = false; | |
445 | break; | |
446 | ||
e9be9d5e | 447 | default: |
bead55ef | 448 | pr_err("overlayfs: unrecognized mount option \"%s\" or missing value\n", p); |
e9be9d5e MS |
449 | return -EINVAL; |
450 | } | |
451 | } | |
71cbad7e | 452 | |
453 | /* Workdir is useless in non-upper mount */ | |
454 | if (!config->upperdir && config->workdir) { | |
455 | pr_info("overlayfs: option \"workdir=%s\" is useless in a non-upper mount, ignore\n", | |
456 | config->workdir); | |
457 | kfree(config->workdir); | |
458 | config->workdir = NULL; | |
459 | } | |
460 | ||
e9be9d5e MS |
461 | return 0; |
462 | } | |
463 | ||
464 | #define OVL_WORKDIR_NAME "work" | |
02bcd157 | 465 | #define OVL_INDEXDIR_NAME "index" |
e9be9d5e | 466 | |
ad204488 | 467 | static struct dentry *ovl_workdir_create(struct ovl_fs *ofs, |
6b8aa129 | 468 | const char *name, bool persist) |
e9be9d5e | 469 | { |
ad204488 MS |
470 | struct inode *dir = ofs->workbasedir->d_inode; |
471 | struct vfsmount *mnt = ofs->upper_mnt; | |
e9be9d5e MS |
472 | struct dentry *work; |
473 | int err; | |
474 | bool retried = false; | |
6b8aa129 | 475 | bool locked = false; |
e9be9d5e MS |
476 | |
477 | err = mnt_want_write(mnt); | |
478 | if (err) | |
6b8aa129 | 479 | goto out_err; |
e9be9d5e | 480 | |
5955102c | 481 | inode_lock_nested(dir, I_MUTEX_PARENT); |
6b8aa129 AG |
482 | locked = true; |
483 | ||
e9be9d5e | 484 | retry: |
ad204488 | 485 | work = lookup_one_len(name, ofs->workbasedir, strlen(name)); |
e9be9d5e MS |
486 | |
487 | if (!IS_ERR(work)) { | |
c11b9fdd MS |
488 | struct iattr attr = { |
489 | .ia_valid = ATTR_MODE, | |
32a3d848 | 490 | .ia_mode = S_IFDIR | 0, |
c11b9fdd | 491 | }; |
e9be9d5e MS |
492 | |
493 | if (work->d_inode) { | |
494 | err = -EEXIST; | |
495 | if (retried) | |
496 | goto out_dput; | |
497 | ||
6b8aa129 AG |
498 | if (persist) |
499 | goto out_unlock; | |
500 | ||
e9be9d5e | 501 | retried = true; |
eea2fb48 | 502 | ovl_workdir_cleanup(dir, mnt, work, 0); |
e9be9d5e MS |
503 | dput(work); |
504 | goto retry; | |
505 | } | |
506 | ||
32a3d848 AV |
507 | err = ovl_create_real(dir, work, |
508 | &(struct cattr){.mode = S_IFDIR | 0}, | |
509 | NULL, true); | |
e9be9d5e MS |
510 | if (err) |
511 | goto out_dput; | |
c11b9fdd | 512 | |
cb348edb MS |
513 | /* |
514 | * Try to remove POSIX ACL xattrs from workdir. We are good if: | |
515 | * | |
516 | * a) success (there was a POSIX ACL xattr and was removed) | |
517 | * b) -ENODATA (there was no POSIX ACL xattr) | |
518 | * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported) | |
519 | * | |
520 | * There are various other error values that could effectively | |
521 | * mean that the xattr doesn't exist (e.g. -ERANGE is returned | |
522 | * if the xattr name is too long), but the set of filesystems | |
523 | * allowed as upper are limited to "normal" ones, where checking | |
524 | * for the above two errors is sufficient. | |
525 | */ | |
c11b9fdd | 526 | err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_DEFAULT); |
e1ff3dd1 | 527 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
528 | goto out_dput; |
529 | ||
530 | err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_ACCESS); | |
e1ff3dd1 | 531 | if (err && err != -ENODATA && err != -EOPNOTSUPP) |
c11b9fdd MS |
532 | goto out_dput; |
533 | ||
534 | /* Clear any inherited mode bits */ | |
535 | inode_lock(work->d_inode); | |
536 | err = notify_change(work, &attr, NULL); | |
537 | inode_unlock(work->d_inode); | |
538 | if (err) | |
539 | goto out_dput; | |
6b8aa129 AG |
540 | } else { |
541 | err = PTR_ERR(work); | |
542 | goto out_err; | |
e9be9d5e MS |
543 | } |
544 | out_unlock: | |
e9be9d5e | 545 | mnt_drop_write(mnt); |
6b8aa129 AG |
546 | if (locked) |
547 | inode_unlock(dir); | |
e9be9d5e MS |
548 | |
549 | return work; | |
550 | ||
551 | out_dput: | |
552 | dput(work); | |
6b8aa129 AG |
553 | out_err: |
554 | pr_warn("overlayfs: failed to create directory %s/%s (errno: %i); mounting read-only\n", | |
ad204488 | 555 | ofs->config.workdir, name, -err); |
6b8aa129 | 556 | work = NULL; |
e9be9d5e MS |
557 | goto out_unlock; |
558 | } | |
559 | ||
91c77947 MS |
560 | static void ovl_unescape(char *s) |
561 | { | |
562 | char *d = s; | |
563 | ||
564 | for (;; s++, d++) { | |
565 | if (*s == '\\') | |
566 | s++; | |
567 | *d = *s; | |
568 | if (!*s) | |
569 | break; | |
570 | } | |
571 | } | |
572 | ||
ab508822 MS |
573 | static int ovl_mount_dir_noesc(const char *name, struct path *path) |
574 | { | |
a78d9f0d | 575 | int err = -EINVAL; |
ab508822 | 576 | |
a78d9f0d MS |
577 | if (!*name) { |
578 | pr_err("overlayfs: empty lowerdir\n"); | |
579 | goto out; | |
580 | } | |
ab508822 MS |
581 | err = kern_path(name, LOOKUP_FOLLOW, path); |
582 | if (err) { | |
583 | pr_err("overlayfs: failed to resolve '%s': %i\n", name, err); | |
584 | goto out; | |
585 | } | |
586 | err = -EINVAL; | |
7c03b5d4 | 587 | if (ovl_dentry_weird(path->dentry)) { |
ab508822 MS |
588 | pr_err("overlayfs: filesystem on '%s' not supported\n", name); |
589 | goto out_put; | |
590 | } | |
2b8c30e9 | 591 | if (!d_is_dir(path->dentry)) { |
ab508822 MS |
592 | pr_err("overlayfs: '%s' not a directory\n", name); |
593 | goto out_put; | |
594 | } | |
595 | return 0; | |
596 | ||
597 | out_put: | |
8aafcb59 | 598 | path_put_init(path); |
ab508822 MS |
599 | out: |
600 | return err; | |
601 | } | |
602 | ||
603 | static int ovl_mount_dir(const char *name, struct path *path) | |
604 | { | |
605 | int err = -ENOMEM; | |
606 | char *tmp = kstrdup(name, GFP_KERNEL); | |
607 | ||
608 | if (tmp) { | |
609 | ovl_unescape(tmp); | |
610 | err = ovl_mount_dir_noesc(tmp, path); | |
7c03b5d4 MS |
611 | |
612 | if (!err) | |
613 | if (ovl_dentry_remote(path->dentry)) { | |
614 | pr_err("overlayfs: filesystem on '%s' not supported as upperdir\n", | |
615 | tmp); | |
8aafcb59 | 616 | path_put_init(path); |
7c03b5d4 MS |
617 | err = -EINVAL; |
618 | } | |
ab508822 MS |
619 | kfree(tmp); |
620 | } | |
621 | return err; | |
622 | } | |
623 | ||
6b2d5fe4 MS |
624 | static int ovl_check_namelen(struct path *path, struct ovl_fs *ofs, |
625 | const char *name) | |
ab508822 | 626 | { |
ab508822 | 627 | struct kstatfs statfs; |
6b2d5fe4 MS |
628 | int err = vfs_statfs(path, &statfs); |
629 | ||
630 | if (err) | |
631 | pr_err("overlayfs: statfs failed on '%s'\n", name); | |
632 | else | |
633 | ofs->namelen = max(ofs->namelen, statfs.f_namelen); | |
634 | ||
635 | return err; | |
636 | } | |
637 | ||
638 | static int ovl_lower_dir(const char *name, struct path *path, | |
639 | struct ovl_fs *ofs, int *stack_depth, bool *remote) | |
640 | { | |
641 | int err; | |
ab508822 | 642 | |
a78d9f0d | 643 | err = ovl_mount_dir_noesc(name, path); |
ab508822 MS |
644 | if (err) |
645 | goto out; | |
646 | ||
6b2d5fe4 MS |
647 | err = ovl_check_namelen(path, ofs, name); |
648 | if (err) | |
ab508822 | 649 | goto out_put; |
6b2d5fe4 | 650 | |
ab508822 MS |
651 | *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth); |
652 | ||
7c03b5d4 MS |
653 | if (ovl_dentry_remote(path->dentry)) |
654 | *remote = true; | |
655 | ||
02bcd157 AG |
656 | /* |
657 | * The inodes index feature needs to encode and decode file | |
658 | * handles, so it requires that all layers support them. | |
659 | */ | |
660 | if (ofs->config.index && !ovl_can_decode_fh(path->dentry->d_sb)) { | |
661 | ofs->config.index = false; | |
662 | pr_warn("overlayfs: fs on '%s' does not support file handles, falling back to index=off.\n", name); | |
663 | } | |
664 | ||
ab508822 MS |
665 | return 0; |
666 | ||
667 | out_put: | |
8aafcb59 | 668 | path_put_init(path); |
ab508822 MS |
669 | out: |
670 | return err; | |
671 | } | |
672 | ||
e9be9d5e MS |
673 | /* Workdir should not be subdir of upperdir and vice versa */ |
674 | static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir) | |
675 | { | |
676 | bool ok = false; | |
677 | ||
678 | if (workdir != upperdir) { | |
679 | ok = (lock_rename(workdir, upperdir) == NULL); | |
680 | unlock_rename(workdir, upperdir); | |
681 | } | |
682 | return ok; | |
683 | } | |
684 | ||
a78d9f0d MS |
685 | static unsigned int ovl_split_lowerdirs(char *str) |
686 | { | |
687 | unsigned int ctr = 1; | |
688 | char *s, *d; | |
689 | ||
690 | for (s = d = str;; s++, d++) { | |
691 | if (*s == '\\') { | |
692 | s++; | |
693 | } else if (*s == ':') { | |
694 | *d = '\0'; | |
695 | ctr++; | |
696 | continue; | |
697 | } | |
698 | *d = *s; | |
699 | if (!*s) | |
700 | break; | |
701 | } | |
702 | return ctr; | |
703 | } | |
704 | ||
0eb45fc3 AG |
705 | static int __maybe_unused |
706 | ovl_posix_acl_xattr_get(const struct xattr_handler *handler, | |
707 | struct dentry *dentry, struct inode *inode, | |
708 | const char *name, void *buffer, size_t size) | |
709 | { | |
1d88f183 | 710 | return ovl_xattr_get(dentry, inode, handler->name, buffer, size); |
0eb45fc3 AG |
711 | } |
712 | ||
0c97be22 AG |
713 | static int __maybe_unused |
714 | ovl_posix_acl_xattr_set(const struct xattr_handler *handler, | |
715 | struct dentry *dentry, struct inode *inode, | |
716 | const char *name, const void *value, | |
717 | size_t size, int flags) | |
d837a49b MS |
718 | { |
719 | struct dentry *workdir = ovl_workdir(dentry); | |
09d8b586 | 720 | struct inode *realinode = ovl_inode_real(inode); |
d837a49b MS |
721 | struct posix_acl *acl = NULL; |
722 | int err; | |
723 | ||
724 | /* Check that everything is OK before copy-up */ | |
725 | if (value) { | |
726 | acl = posix_acl_from_xattr(&init_user_ns, value, size); | |
727 | if (IS_ERR(acl)) | |
728 | return PTR_ERR(acl); | |
729 | } | |
730 | err = -EOPNOTSUPP; | |
731 | if (!IS_POSIXACL(d_inode(workdir))) | |
732 | goto out_acl_release; | |
733 | if (!realinode->i_op->set_acl) | |
734 | goto out_acl_release; | |
735 | if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) { | |
736 | err = acl ? -EACCES : 0; | |
737 | goto out_acl_release; | |
738 | } | |
739 | err = -EPERM; | |
740 | if (!inode_owner_or_capable(inode)) | |
741 | goto out_acl_release; | |
742 | ||
743 | posix_acl_release(acl); | |
744 | ||
fd3220d3 MS |
745 | /* |
746 | * Check if sgid bit needs to be cleared (actual setacl operation will | |
747 | * be done with mounter's capabilities and so that won't do it for us). | |
748 | */ | |
749 | if (unlikely(inode->i_mode & S_ISGID) && | |
750 | handler->flags == ACL_TYPE_ACCESS && | |
751 | !in_group_p(inode->i_gid) && | |
752 | !capable_wrt_inode_uidgid(inode, CAP_FSETID)) { | |
753 | struct iattr iattr = { .ia_valid = ATTR_KILL_SGID }; | |
754 | ||
755 | err = ovl_setattr(dentry, &iattr); | |
756 | if (err) | |
757 | return err; | |
758 | } | |
759 | ||
1d88f183 | 760 | err = ovl_xattr_set(dentry, inode, handler->name, value, size, flags); |
ce31513a | 761 | if (!err) |
09d8b586 | 762 | ovl_copyattr(ovl_inode_real(inode), inode); |
ce31513a MS |
763 | |
764 | return err; | |
d837a49b MS |
765 | |
766 | out_acl_release: | |
767 | posix_acl_release(acl); | |
768 | return err; | |
769 | } | |
770 | ||
0eb45fc3 AG |
771 | static int ovl_own_xattr_get(const struct xattr_handler *handler, |
772 | struct dentry *dentry, struct inode *inode, | |
773 | const char *name, void *buffer, size_t size) | |
774 | { | |
48fab5d7 | 775 | return -EOPNOTSUPP; |
0eb45fc3 AG |
776 | } |
777 | ||
d837a49b MS |
778 | static int ovl_own_xattr_set(const struct xattr_handler *handler, |
779 | struct dentry *dentry, struct inode *inode, | |
780 | const char *name, const void *value, | |
781 | size_t size, int flags) | |
782 | { | |
48fab5d7 | 783 | return -EOPNOTSUPP; |
d837a49b MS |
784 | } |
785 | ||
0eb45fc3 AG |
786 | static int ovl_other_xattr_get(const struct xattr_handler *handler, |
787 | struct dentry *dentry, struct inode *inode, | |
788 | const char *name, void *buffer, size_t size) | |
789 | { | |
1d88f183 | 790 | return ovl_xattr_get(dentry, inode, name, buffer, size); |
0eb45fc3 AG |
791 | } |
792 | ||
0e585ccc AG |
793 | static int ovl_other_xattr_set(const struct xattr_handler *handler, |
794 | struct dentry *dentry, struct inode *inode, | |
795 | const char *name, const void *value, | |
796 | size_t size, int flags) | |
797 | { | |
1d88f183 | 798 | return ovl_xattr_set(dentry, inode, name, value, size, flags); |
0e585ccc AG |
799 | } |
800 | ||
0c97be22 AG |
801 | static const struct xattr_handler __maybe_unused |
802 | ovl_posix_acl_access_xattr_handler = { | |
d837a49b MS |
803 | .name = XATTR_NAME_POSIX_ACL_ACCESS, |
804 | .flags = ACL_TYPE_ACCESS, | |
0eb45fc3 | 805 | .get = ovl_posix_acl_xattr_get, |
d837a49b MS |
806 | .set = ovl_posix_acl_xattr_set, |
807 | }; | |
808 | ||
0c97be22 AG |
809 | static const struct xattr_handler __maybe_unused |
810 | ovl_posix_acl_default_xattr_handler = { | |
d837a49b MS |
811 | .name = XATTR_NAME_POSIX_ACL_DEFAULT, |
812 | .flags = ACL_TYPE_DEFAULT, | |
0eb45fc3 | 813 | .get = ovl_posix_acl_xattr_get, |
d837a49b MS |
814 | .set = ovl_posix_acl_xattr_set, |
815 | }; | |
816 | ||
817 | static const struct xattr_handler ovl_own_xattr_handler = { | |
818 | .prefix = OVL_XATTR_PREFIX, | |
0eb45fc3 | 819 | .get = ovl_own_xattr_get, |
d837a49b MS |
820 | .set = ovl_own_xattr_set, |
821 | }; | |
822 | ||
823 | static const struct xattr_handler ovl_other_xattr_handler = { | |
824 | .prefix = "", /* catch all */ | |
0eb45fc3 | 825 | .get = ovl_other_xattr_get, |
d837a49b MS |
826 | .set = ovl_other_xattr_set, |
827 | }; | |
828 | ||
829 | static const struct xattr_handler *ovl_xattr_handlers[] = { | |
0c97be22 | 830 | #ifdef CONFIG_FS_POSIX_ACL |
d837a49b MS |
831 | &ovl_posix_acl_access_xattr_handler, |
832 | &ovl_posix_acl_default_xattr_handler, | |
0c97be22 | 833 | #endif |
d837a49b MS |
834 | &ovl_own_xattr_handler, |
835 | &ovl_other_xattr_handler, | |
836 | NULL | |
837 | }; | |
838 | ||
ad204488 | 839 | static int ovl_get_upper(struct ovl_fs *ofs, struct path *upperpath) |
6ee8acf0 | 840 | { |
5064975e | 841 | struct vfsmount *upper_mnt; |
6ee8acf0 MS |
842 | int err; |
843 | ||
ad204488 | 844 | err = ovl_mount_dir(ofs->config.upperdir, upperpath); |
6ee8acf0 MS |
845 | if (err) |
846 | goto out; | |
847 | ||
848 | /* Upper fs should not be r/o */ | |
849 | if (sb_rdonly(upperpath->mnt->mnt_sb)) { | |
850 | pr_err("overlayfs: upper fs is r/o, try multi-lower layers mount\n"); | |
851 | err = -EINVAL; | |
852 | goto out; | |
853 | } | |
854 | ||
ad204488 | 855 | err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir); |
6ee8acf0 MS |
856 | if (err) |
857 | goto out; | |
858 | ||
859 | err = -EBUSY; | |
860 | if (ovl_inuse_trylock(upperpath->dentry)) { | |
ad204488 MS |
861 | ofs->upperdir_locked = true; |
862 | } else if (ofs->config.index) { | |
6ee8acf0 MS |
863 | pr_err("overlayfs: upperdir is in-use by another mount, mount with '-o index=off' to override exclusive upperdir protection.\n"); |
864 | goto out; | |
865 | } else { | |
866 | pr_warn("overlayfs: upperdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n"); | |
867 | } | |
5064975e MS |
868 | |
869 | upper_mnt = clone_private_mount(upperpath); | |
870 | err = PTR_ERR(upper_mnt); | |
871 | if (IS_ERR(upper_mnt)) { | |
872 | pr_err("overlayfs: failed to clone upperpath\n"); | |
873 | goto out; | |
874 | } | |
875 | ||
876 | /* Don't inherit atime flags */ | |
877 | upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME); | |
ad204488 | 878 | ofs->upper_mnt = upper_mnt; |
6ee8acf0 MS |
879 | err = 0; |
880 | out: | |
881 | return err; | |
882 | } | |
883 | ||
ad204488 | 884 | static int ovl_make_workdir(struct ovl_fs *ofs, struct path *workpath) |
8ed61dc3 MS |
885 | { |
886 | struct dentry *temp; | |
887 | int err; | |
888 | ||
ad204488 MS |
889 | ofs->workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false); |
890 | if (!ofs->workdir) | |
8ed61dc3 MS |
891 | return 0; |
892 | ||
893 | /* | |
894 | * Upper should support d_type, else whiteouts are visible. Given | |
895 | * workdir and upper are on same fs, we can do iterate_dir() on | |
896 | * workdir. This check requires successful creation of workdir in | |
897 | * previous step. | |
898 | */ | |
899 | err = ovl_check_d_type_supported(workpath); | |
900 | if (err < 0) | |
901 | return err; | |
902 | ||
903 | /* | |
904 | * We allowed this configuration and don't want to break users over | |
905 | * kernel upgrade. So warn instead of erroring out. | |
906 | */ | |
907 | if (!err) | |
908 | pr_warn("overlayfs: upper fs needs to support d_type.\n"); | |
909 | ||
910 | /* Check if upper/work fs supports O_TMPFILE */ | |
ad204488 MS |
911 | temp = ovl_do_tmpfile(ofs->workdir, S_IFREG | 0); |
912 | ofs->tmpfile = !IS_ERR(temp); | |
913 | if (ofs->tmpfile) | |
8ed61dc3 MS |
914 | dput(temp); |
915 | else | |
916 | pr_warn("overlayfs: upper fs does not support tmpfile.\n"); | |
917 | ||
918 | /* | |
919 | * Check if upper/work fs supports trusted.overlay.* xattr | |
920 | */ | |
ad204488 | 921 | err = ovl_do_setxattr(ofs->workdir, OVL_XATTR_OPAQUE, "0", 1, 0); |
8ed61dc3 | 922 | if (err) { |
ad204488 | 923 | ofs->noxattr = true; |
8ed61dc3 MS |
924 | pr_warn("overlayfs: upper fs does not support xattr.\n"); |
925 | } else { | |
ad204488 | 926 | vfs_removexattr(ofs->workdir, OVL_XATTR_OPAQUE); |
8ed61dc3 MS |
927 | } |
928 | ||
929 | /* Check if upper/work fs supports file handles */ | |
ad204488 MS |
930 | if (ofs->config.index && |
931 | !ovl_can_decode_fh(ofs->workdir->d_sb)) { | |
932 | ofs->config.index = false; | |
8ed61dc3 MS |
933 | pr_warn("overlayfs: upper fs does not support file handles, falling back to index=off.\n"); |
934 | } | |
935 | ||
936 | return 0; | |
937 | } | |
938 | ||
ad204488 | 939 | static int ovl_get_workdir(struct ovl_fs *ofs, struct path *upperpath) |
520d7c86 MS |
940 | { |
941 | int err; | |
bca44b52 | 942 | struct path workpath = { }; |
520d7c86 | 943 | |
ad204488 | 944 | err = ovl_mount_dir(ofs->config.workdir, &workpath); |
520d7c86 MS |
945 | if (err) |
946 | goto out; | |
947 | ||
948 | err = -EINVAL; | |
bca44b52 | 949 | if (upperpath->mnt != workpath.mnt) { |
520d7c86 MS |
950 | pr_err("overlayfs: workdir and upperdir must reside under the same mount\n"); |
951 | goto out; | |
952 | } | |
bca44b52 | 953 | if (!ovl_workdir_ok(workpath.dentry, upperpath->dentry)) { |
520d7c86 MS |
954 | pr_err("overlayfs: workdir and upperdir must be separate subtrees\n"); |
955 | goto out; | |
956 | } | |
957 | ||
958 | err = -EBUSY; | |
bca44b52 | 959 | if (ovl_inuse_trylock(workpath.dentry)) { |
ad204488 MS |
960 | ofs->workdir_locked = true; |
961 | } else if (ofs->config.index) { | |
520d7c86 MS |
962 | pr_err("overlayfs: workdir is in-use by another mount, mount with '-o index=off' to override exclusive workdir protection.\n"); |
963 | goto out; | |
964 | } else { | |
965 | pr_warn("overlayfs: workdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n"); | |
966 | } | |
967 | ||
ad204488 MS |
968 | ofs->workbasedir = dget(workpath.dentry); |
969 | err = ovl_make_workdir(ofs, &workpath); | |
bca44b52 MS |
970 | if (err) |
971 | goto out; | |
972 | ||
520d7c86 MS |
973 | err = 0; |
974 | out: | |
bca44b52 MS |
975 | path_put(&workpath); |
976 | ||
520d7c86 MS |
977 | return err; |
978 | } | |
979 | ||
ad204488 | 980 | static int ovl_get_indexdir(struct ovl_fs *ofs, struct ovl_entry *oe, |
95e6d417 | 981 | struct path *upperpath) |
f7e3a7d9 MS |
982 | { |
983 | int err; | |
984 | ||
985 | /* Verify lower root is upper root origin */ | |
986 | err = ovl_verify_origin(upperpath->dentry, | |
987 | oe->lowerstack[0].layer->mnt, | |
988 | oe->lowerstack[0].dentry, | |
989 | false, true); | |
990 | if (err) { | |
991 | pr_err("overlayfs: failed to verify upper root origin\n"); | |
992 | goto out; | |
993 | } | |
994 | ||
ad204488 MS |
995 | ofs->indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true); |
996 | if (ofs->indexdir) { | |
f7e3a7d9 | 997 | /* Verify upper root is index dir origin */ |
ad204488 | 998 | err = ovl_verify_origin(ofs->indexdir, ofs->upper_mnt, |
f7e3a7d9 MS |
999 | upperpath->dentry, true, true); |
1000 | if (err) | |
1001 | pr_err("overlayfs: failed to verify index dir origin\n"); | |
1002 | ||
1003 | /* Cleanup bad/stale/orphan index entries */ | |
1004 | if (!err) | |
ad204488 MS |
1005 | err = ovl_indexdir_cleanup(ofs->indexdir, |
1006 | ofs->upper_mnt, | |
f7e3a7d9 MS |
1007 | oe->lowerstack, |
1008 | oe->numlower); | |
1009 | } | |
ad204488 | 1010 | if (err || !ofs->indexdir) |
f7e3a7d9 MS |
1011 | pr_warn("overlayfs: try deleting index dir or mounting with '-o index=off' to disable inodes index.\n"); |
1012 | ||
1013 | out: | |
1014 | return err; | |
1015 | } | |
1016 | ||
ad204488 | 1017 | static int ovl_get_lower_layers(struct ovl_fs *ofs, struct path *stack, |
520d7c86 MS |
1018 | unsigned int numlower) |
1019 | { | |
1020 | int err; | |
1021 | unsigned int i; | |
1022 | ||
1023 | err = -ENOMEM; | |
ad204488 | 1024 | ofs->lower_layers = kcalloc(numlower, sizeof(struct ovl_layer), |
520d7c86 | 1025 | GFP_KERNEL); |
ad204488 | 1026 | if (ofs->lower_layers == NULL) |
520d7c86 MS |
1027 | goto out; |
1028 | for (i = 0; i < numlower; i++) { | |
1029 | struct vfsmount *mnt; | |
1030 | dev_t dev; | |
1031 | ||
1032 | err = get_anon_bdev(&dev); | |
1033 | if (err) { | |
1034 | pr_err("overlayfs: failed to get anonymous bdev for lowerpath\n"); | |
1035 | goto out; | |
1036 | } | |
1037 | ||
1038 | mnt = clone_private_mount(&stack[i]); | |
1039 | err = PTR_ERR(mnt); | |
1040 | if (IS_ERR(mnt)) { | |
1041 | pr_err("overlayfs: failed to clone lowerpath\n"); | |
1042 | free_anon_bdev(dev); | |
1043 | goto out; | |
1044 | } | |
1045 | /* | |
1046 | * Make lower layers R/O. That way fchmod/fchown on lower file | |
1047 | * will fail instead of modifying lower fs. | |
1048 | */ | |
1049 | mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME; | |
1050 | ||
ad204488 MS |
1051 | ofs->lower_layers[ofs->numlower].mnt = mnt; |
1052 | ofs->lower_layers[ofs->numlower].pseudo_dev = dev; | |
1053 | ofs->numlower++; | |
520d7c86 MS |
1054 | |
1055 | /* Check if all lower layers are on same sb */ | |
1056 | if (i == 0) | |
ad204488 MS |
1057 | ofs->same_sb = mnt->mnt_sb; |
1058 | else if (ofs->same_sb != mnt->mnt_sb) | |
1059 | ofs->same_sb = NULL; | |
520d7c86 MS |
1060 | } |
1061 | err = 0; | |
1062 | out: | |
1063 | return err; | |
1064 | } | |
1065 | ||
4155c10a | 1066 | static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb, |
ad204488 | 1067 | struct ovl_fs *ofs) |
53dbb0b4 MS |
1068 | { |
1069 | int err; | |
1070 | char *lowertmp, *lower; | |
4155c10a MS |
1071 | struct path *stack = NULL; |
1072 | unsigned int stacklen, numlower = 0, i; | |
53dbb0b4 | 1073 | bool remote = false; |
4155c10a | 1074 | struct ovl_entry *oe; |
53dbb0b4 MS |
1075 | |
1076 | err = -ENOMEM; | |
ad204488 | 1077 | lowertmp = kstrdup(ofs->config.lowerdir, GFP_KERNEL); |
53dbb0b4 | 1078 | if (!lowertmp) |
4155c10a | 1079 | goto out_err; |
53dbb0b4 MS |
1080 | |
1081 | err = -EINVAL; | |
1082 | stacklen = ovl_split_lowerdirs(lowertmp); | |
1083 | if (stacklen > OVL_MAX_STACK) { | |
1084 | pr_err("overlayfs: too many lower directories, limit is %d\n", | |
1085 | OVL_MAX_STACK); | |
4155c10a | 1086 | goto out_err; |
ad204488 | 1087 | } else if (!ofs->config.upperdir && stacklen == 1) { |
53dbb0b4 | 1088 | pr_err("overlayfs: at least 2 lowerdir are needed while upperdir nonexistent\n"); |
4155c10a | 1089 | goto out_err; |
53dbb0b4 MS |
1090 | } |
1091 | ||
1092 | err = -ENOMEM; | |
1093 | stack = kcalloc(stacklen, sizeof(struct path), GFP_KERNEL); | |
1094 | if (!stack) | |
4155c10a | 1095 | goto out_err; |
53dbb0b4 MS |
1096 | |
1097 | err = -EINVAL; | |
1098 | lower = lowertmp; | |
1099 | for (numlower = 0; numlower < stacklen; numlower++) { | |
ad204488 | 1100 | err = ovl_lower_dir(lower, &stack[numlower], ofs, |
53dbb0b4 MS |
1101 | &sb->s_stack_depth, &remote); |
1102 | if (err) | |
4155c10a | 1103 | goto out_err; |
53dbb0b4 MS |
1104 | |
1105 | lower = strchr(lower, '\0') + 1; | |
1106 | } | |
1107 | ||
1108 | err = -EINVAL; | |
1109 | sb->s_stack_depth++; | |
1110 | if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) { | |
1111 | pr_err("overlayfs: maximum fs stacking depth exceeded\n"); | |
4155c10a | 1112 | goto out_err; |
53dbb0b4 MS |
1113 | } |
1114 | ||
ad204488 | 1115 | err = ovl_get_lower_layers(ofs, stack, numlower); |
4155c10a MS |
1116 | if (err) |
1117 | goto out_err; | |
1118 | ||
1119 | err = -ENOMEM; | |
1120 | oe = ovl_alloc_entry(numlower); | |
1121 | if (!oe) | |
1122 | goto out_err; | |
1123 | ||
1124 | for (i = 0; i < numlower; i++) { | |
1125 | oe->lowerstack[i].dentry = dget(stack[i].dentry); | |
ad204488 | 1126 | oe->lowerstack[i].layer = &ofs->lower_layers[i]; |
4155c10a | 1127 | } |
53dbb0b4 MS |
1128 | |
1129 | if (remote) | |
1130 | sb->s_d_op = &ovl_reval_dentry_operations; | |
1131 | else | |
1132 | sb->s_d_op = &ovl_dentry_operations; | |
1133 | ||
53dbb0b4 | 1134 | out: |
53dbb0b4 MS |
1135 | for (i = 0; i < numlower; i++) |
1136 | path_put(&stack[i]); | |
1137 | kfree(stack); | |
4155c10a MS |
1138 | kfree(lowertmp); |
1139 | ||
1140 | return oe; | |
1141 | ||
1142 | out_err: | |
1143 | oe = ERR_PTR(err); | |
53dbb0b4 MS |
1144 | goto out; |
1145 | } | |
1146 | ||
e9be9d5e MS |
1147 | static int ovl_fill_super(struct super_block *sb, void *data, int silent) |
1148 | { | |
33006cdf | 1149 | struct path upperpath = { }; |
e9be9d5e | 1150 | struct dentry *root_dentry; |
4155c10a | 1151 | struct ovl_entry *oe; |
ad204488 | 1152 | struct ovl_fs *ofs; |
51f8f3c4 | 1153 | struct cred *cred; |
e9be9d5e MS |
1154 | int err; |
1155 | ||
f45827e8 | 1156 | err = -ENOMEM; |
ad204488 MS |
1157 | ofs = kzalloc(sizeof(struct ovl_fs), GFP_KERNEL); |
1158 | if (!ofs) | |
e9be9d5e MS |
1159 | goto out; |
1160 | ||
ad204488 | 1161 | ofs->creator_cred = cred = prepare_creds(); |
c6fe6254 MS |
1162 | if (!cred) |
1163 | goto out_err; | |
1164 | ||
ad204488 MS |
1165 | ofs->config.redirect_dir = ovl_redirect_dir_def; |
1166 | ofs->config.index = ovl_index_def; | |
1167 | err = ovl_parse_opt((char *) data, &ofs->config); | |
f45827e8 | 1168 | if (err) |
a9075cdb | 1169 | goto out_err; |
f45827e8 | 1170 | |
e9be9d5e | 1171 | err = -EINVAL; |
ad204488 | 1172 | if (!ofs->config.lowerdir) { |
07f2af7b KK |
1173 | if (!silent) |
1174 | pr_err("overlayfs: missing 'lowerdir'\n"); | |
a9075cdb | 1175 | goto out_err; |
e9be9d5e MS |
1176 | } |
1177 | ||
53a08cb9 | 1178 | sb->s_stack_depth = 0; |
cf9a6784 | 1179 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
ad204488 MS |
1180 | if (ofs->config.upperdir) { |
1181 | if (!ofs->config.workdir) { | |
53a08cb9 | 1182 | pr_err("overlayfs: missing 'workdir'\n"); |
a9075cdb | 1183 | goto out_err; |
53a08cb9 | 1184 | } |
e9be9d5e | 1185 | |
ad204488 | 1186 | err = ovl_get_upper(ofs, &upperpath); |
53a08cb9 | 1187 | if (err) |
a9075cdb | 1188 | goto out_err; |
2cac0c00 | 1189 | |
ad204488 | 1190 | err = ovl_get_workdir(ofs, &upperpath); |
8ed61dc3 | 1191 | if (err) |
a9075cdb | 1192 | goto out_err; |
c6fe6254 | 1193 | |
ad204488 | 1194 | if (!ofs->workdir) |
6e88256e MS |
1195 | sb->s_flags |= MS_RDONLY; |
1196 | ||
ad204488 MS |
1197 | sb->s_stack_depth = ofs->upper_mnt->mnt_sb->s_stack_depth; |
1198 | sb->s_time_gran = ofs->upper_mnt->mnt_sb->s_time_gran; | |
c6fe6254 | 1199 | |
e9be9d5e | 1200 | } |
ad204488 | 1201 | oe = ovl_get_lowerstack(sb, ofs); |
4155c10a MS |
1202 | err = PTR_ERR(oe); |
1203 | if (IS_ERR(oe)) | |
a9075cdb | 1204 | goto out_err; |
e9be9d5e | 1205 | |
71cbad7e | 1206 | /* If the upper fs is nonexistent, we mark overlayfs r/o too */ |
ad204488 | 1207 | if (!ofs->upper_mnt) |
e9be9d5e | 1208 | sb->s_flags |= MS_RDONLY; |
ad204488 MS |
1209 | else if (ofs->upper_mnt->mnt_sb != ofs->same_sb) |
1210 | ofs->same_sb = NULL; | |
e9be9d5e | 1211 | |
ad204488 MS |
1212 | if (!(ovl_force_readonly(ofs)) && ofs->config.index) { |
1213 | err = ovl_get_indexdir(ofs, oe, &upperpath); | |
54fb347e | 1214 | if (err) |
4155c10a | 1215 | goto out_free_oe; |
6e88256e | 1216 | |
ad204488 | 1217 | if (!ofs->indexdir) |
6e88256e | 1218 | sb->s_flags |= MS_RDONLY; |
02bcd157 AG |
1219 | } |
1220 | ||
1221 | /* Show index=off/on in /proc/mounts for any of the reasons above */ | |
ad204488 MS |
1222 | if (!ofs->indexdir) |
1223 | ofs->config.index = false; | |
02bcd157 | 1224 | |
51f8f3c4 KK |
1225 | /* Never override disk quota limits or use reserved space */ |
1226 | cap_lower(cred->cap_effective, CAP_SYS_RESOURCE); | |
1227 | ||
655042cc VG |
1228 | sb->s_magic = OVERLAYFS_SUPER_MAGIC; |
1229 | sb->s_op = &ovl_super_operations; | |
1230 | sb->s_xattr = ovl_xattr_handlers; | |
ad204488 | 1231 | sb->s_fs_info = ofs; |
655042cc VG |
1232 | sb->s_flags |= MS_POSIXACL | MS_NOREMOTELOCK; |
1233 | ||
c6fe6254 | 1234 | err = -ENOMEM; |
ca4c8a3a | 1235 | root_dentry = d_make_root(ovl_new_inode(sb, S_IFDIR, 0)); |
e9be9d5e | 1236 | if (!root_dentry) |
4155c10a | 1237 | goto out_free_oe; |
e9be9d5e MS |
1238 | |
1239 | mntput(upperpath.mnt); | |
f3a15685 | 1240 | if (upperpath.dentry) { |
55acc661 | 1241 | oe->has_upper = true; |
13c72075 MS |
1242 | if (ovl_is_impuredir(upperpath.dentry)) |
1243 | ovl_set_flag(OVL_IMPURE, d_inode(root_dentry)); | |
f3a15685 | 1244 | } |
e9be9d5e MS |
1245 | |
1246 | root_dentry->d_fsdata = oe; | |
1247 | ||
b79e05aa AG |
1248 | /* Root is always merge -> can have whiteouts */ |
1249 | ovl_set_flag(OVL_WHITEOUTS, d_inode(root_dentry)); | |
09d8b586 MS |
1250 | ovl_inode_init(d_inode(root_dentry), upperpath.dentry, |
1251 | ovl_dentry_lower(root_dentry)); | |
ed06e069 | 1252 | |
e9be9d5e | 1253 | sb->s_root = root_dentry; |
e9be9d5e MS |
1254 | |
1255 | return 0; | |
1256 | ||
4155c10a MS |
1257 | out_free_oe: |
1258 | ovl_entry_stack_free(oe); | |
b9343632 | 1259 | kfree(oe); |
4155c10a | 1260 | out_err: |
e9be9d5e | 1261 | path_put(&upperpath); |
ad204488 | 1262 | ovl_free_fs(ofs); |
e9be9d5e MS |
1263 | out: |
1264 | return err; | |
1265 | } | |
1266 | ||
1267 | static struct dentry *ovl_mount(struct file_system_type *fs_type, int flags, | |
1268 | const char *dev_name, void *raw_data) | |
1269 | { | |
1270 | return mount_nodev(fs_type, flags, raw_data, ovl_fill_super); | |
1271 | } | |
1272 | ||
1273 | static struct file_system_type ovl_fs_type = { | |
1274 | .owner = THIS_MODULE, | |
ef94b186 | 1275 | .name = "overlay", |
e9be9d5e MS |
1276 | .mount = ovl_mount, |
1277 | .kill_sb = kill_anon_super, | |
1278 | }; | |
ef94b186 | 1279 | MODULE_ALIAS_FS("overlay"); |
e9be9d5e | 1280 | |
13cf199d AG |
1281 | static void ovl_inode_init_once(void *foo) |
1282 | { | |
1283 | struct ovl_inode *oi = foo; | |
1284 | ||
1285 | inode_init_once(&oi->vfs_inode); | |
1286 | } | |
1287 | ||
e9be9d5e MS |
1288 | static int __init ovl_init(void) |
1289 | { | |
13cf199d AG |
1290 | int err; |
1291 | ||
1292 | ovl_inode_cachep = kmem_cache_create("ovl_inode", | |
1293 | sizeof(struct ovl_inode), 0, | |
1294 | (SLAB_RECLAIM_ACCOUNT| | |
1295 | SLAB_MEM_SPREAD|SLAB_ACCOUNT), | |
1296 | ovl_inode_init_once); | |
1297 | if (ovl_inode_cachep == NULL) | |
1298 | return -ENOMEM; | |
1299 | ||
1300 | err = register_filesystem(&ovl_fs_type); | |
1301 | if (err) | |
1302 | kmem_cache_destroy(ovl_inode_cachep); | |
1303 | ||
1304 | return err; | |
e9be9d5e MS |
1305 | } |
1306 | ||
1307 | static void __exit ovl_exit(void) | |
1308 | { | |
1309 | unregister_filesystem(&ovl_fs_type); | |
13cf199d AG |
1310 | |
1311 | /* | |
1312 | * Make sure all delayed rcu free inodes are flushed before we | |
1313 | * destroy cache. | |
1314 | */ | |
1315 | rcu_barrier(); | |
1316 | kmem_cache_destroy(ovl_inode_cachep); | |
1317 | ||
e9be9d5e MS |
1318 | } |
1319 | ||
1320 | module_init(ovl_init); | |
1321 | module_exit(ovl_exit); |