]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - fs/overlayfs/super.c
ovl: multi-layer lookup
[mirror_ubuntu-hirsute-kernel.git] / fs / overlayfs / super.c
CommitLineData
e9be9d5e
MS
1/*
2 *
3 * Copyright (C) 2011 Novell Inc.
4 *
5 * This program is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 as published by
7 * the Free Software Foundation.
8 */
9
10#include <linux/fs.h>
11#include <linux/namei.h>
12#include <linux/xattr.h>
13#include <linux/security.h>
14#include <linux/mount.h>
15#include <linux/slab.h>
16#include <linux/parser.h>
17#include <linux/module.h>
18#include <linux/sched.h>
cc259639 19#include <linux/statfs.h>
f45827e8 20#include <linux/seq_file.h>
e9be9d5e
MS
21#include "overlayfs.h"
22
23MODULE_AUTHOR("Miklos Szeredi <miklos@szeredi.hu>");
24MODULE_DESCRIPTION("Overlay filesystem");
25MODULE_LICENSE("GPL");
26
ef94b186 27#define OVERLAYFS_SUPER_MAGIC 0x794c7630
cc259639 28
f45827e8
EZ
29struct ovl_config {
30 char *lowerdir;
31 char *upperdir;
32 char *workdir;
33};
34
e9be9d5e
MS
35/* private information held for overlayfs's superblock */
36struct ovl_fs {
37 struct vfsmount *upper_mnt;
dd662667
MS
38 unsigned numlower;
39 struct vfsmount **lower_mnt;
e9be9d5e 40 struct dentry *workdir;
cc259639 41 long lower_namelen;
f45827e8
EZ
42 /* pathnames of lower and upper dirs, for show_options */
43 struct ovl_config config;
e9be9d5e
MS
44};
45
46struct ovl_dir_cache;
47
48/* private information held for every overlayfs dentry */
49struct ovl_entry {
50 struct dentry *__upperdentry;
e9be9d5e
MS
51 struct ovl_dir_cache *cache;
52 union {
53 struct {
54 u64 version;
55 bool opaque;
56 };
57 struct rcu_head rcu;
58 };
dd662667
MS
59 unsigned numlower;
60 struct path lowerstack[];
e9be9d5e
MS
61};
62
63const char *ovl_opaque_xattr = "trusted.overlay.opaque";
64
dd662667
MS
65static struct dentry *__ovl_dentry_lower(struct ovl_entry *oe)
66{
67 return oe->numlower ? oe->lowerstack[0].dentry : NULL;
68}
e9be9d5e
MS
69
70enum ovl_path_type ovl_path_type(struct dentry *dentry)
71{
72 struct ovl_entry *oe = dentry->d_fsdata;
1afaba1e 73 enum ovl_path_type type = 0;
e9be9d5e
MS
74
75 if (oe->__upperdentry) {
1afaba1e
MS
76 type = __OVL_PATH_UPPER;
77
dd662667 78 if (oe->numlower) {
e9be9d5e 79 if (S_ISDIR(dentry->d_inode->i_mode))
1afaba1e
MS
80 type |= __OVL_PATH_MERGE;
81 } else if (!oe->opaque) {
82 type |= __OVL_PATH_PURE;
e9be9d5e 83 }
9d7459d8
MS
84 } else {
85 if (oe->numlower > 1)
86 type |= __OVL_PATH_MERGE;
e9be9d5e 87 }
1afaba1e 88 return type;
e9be9d5e
MS
89}
90
91static struct dentry *ovl_upperdentry_dereference(struct ovl_entry *oe)
92{
71d50928 93 return lockless_dereference(oe->__upperdentry);
e9be9d5e
MS
94}
95
96void ovl_path_upper(struct dentry *dentry, struct path *path)
97{
98 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
99 struct ovl_entry *oe = dentry->d_fsdata;
100
101 path->mnt = ofs->upper_mnt;
102 path->dentry = ovl_upperdentry_dereference(oe);
103}
104
105enum ovl_path_type ovl_path_real(struct dentry *dentry, struct path *path)
106{
107
108 enum ovl_path_type type = ovl_path_type(dentry);
109
1afaba1e 110 if (!OVL_TYPE_UPPER(type))
e9be9d5e
MS
111 ovl_path_lower(dentry, path);
112 else
113 ovl_path_upper(dentry, path);
114
115 return type;
116}
117
118struct dentry *ovl_dentry_upper(struct dentry *dentry)
119{
120 struct ovl_entry *oe = dentry->d_fsdata;
121
122 return ovl_upperdentry_dereference(oe);
123}
124
125struct dentry *ovl_dentry_lower(struct dentry *dentry)
126{
127 struct ovl_entry *oe = dentry->d_fsdata;
128
dd662667 129 return __ovl_dentry_lower(oe);
e9be9d5e
MS
130}
131
132struct dentry *ovl_dentry_real(struct dentry *dentry)
133{
134 struct ovl_entry *oe = dentry->d_fsdata;
135 struct dentry *realdentry;
136
137 realdentry = ovl_upperdentry_dereference(oe);
138 if (!realdentry)
dd662667 139 realdentry = __ovl_dentry_lower(oe);
e9be9d5e
MS
140
141 return realdentry;
142}
143
144struct dentry *ovl_entry_real(struct ovl_entry *oe, bool *is_upper)
145{
146 struct dentry *realdentry;
147
148 realdentry = ovl_upperdentry_dereference(oe);
149 if (realdentry) {
150 *is_upper = true;
151 } else {
dd662667 152 realdentry = __ovl_dentry_lower(oe);
e9be9d5e
MS
153 *is_upper = false;
154 }
155 return realdentry;
156}
157
158struct ovl_dir_cache *ovl_dir_cache(struct dentry *dentry)
159{
160 struct ovl_entry *oe = dentry->d_fsdata;
161
162 return oe->cache;
163}
164
165void ovl_set_dir_cache(struct dentry *dentry, struct ovl_dir_cache *cache)
166{
167 struct ovl_entry *oe = dentry->d_fsdata;
168
169 oe->cache = cache;
170}
171
172void ovl_path_lower(struct dentry *dentry, struct path *path)
173{
e9be9d5e
MS
174 struct ovl_entry *oe = dentry->d_fsdata;
175
dd662667 176 *path = oe->numlower ? oe->lowerstack[0] : (struct path) { NULL, NULL };
e9be9d5e
MS
177}
178
179int ovl_want_write(struct dentry *dentry)
180{
181 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
182 return mnt_want_write(ofs->upper_mnt);
183}
184
185void ovl_drop_write(struct dentry *dentry)
186{
187 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
188 mnt_drop_write(ofs->upper_mnt);
189}
190
191struct dentry *ovl_workdir(struct dentry *dentry)
192{
193 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
194 return ofs->workdir;
195}
196
197bool ovl_dentry_is_opaque(struct dentry *dentry)
198{
199 struct ovl_entry *oe = dentry->d_fsdata;
200 return oe->opaque;
201}
202
203void ovl_dentry_set_opaque(struct dentry *dentry, bool opaque)
204{
205 struct ovl_entry *oe = dentry->d_fsdata;
206 oe->opaque = opaque;
207}
208
209void ovl_dentry_update(struct dentry *dentry, struct dentry *upperdentry)
210{
211 struct ovl_entry *oe = dentry->d_fsdata;
212
213 WARN_ON(!mutex_is_locked(&upperdentry->d_parent->d_inode->i_mutex));
214 WARN_ON(oe->__upperdentry);
215 BUG_ON(!upperdentry->d_inode);
216 /*
217 * Make sure upperdentry is consistent before making it visible to
218 * ovl_upperdentry_dereference().
219 */
220 smp_wmb();
221 oe->__upperdentry = upperdentry;
222}
223
224void ovl_dentry_version_inc(struct dentry *dentry)
225{
226 struct ovl_entry *oe = dentry->d_fsdata;
227
228 WARN_ON(!mutex_is_locked(&dentry->d_inode->i_mutex));
229 oe->version++;
230}
231
232u64 ovl_dentry_version_get(struct dentry *dentry)
233{
234 struct ovl_entry *oe = dentry->d_fsdata;
235
236 WARN_ON(!mutex_is_locked(&dentry->d_inode->i_mutex));
237 return oe->version;
238}
239
240bool ovl_is_whiteout(struct dentry *dentry)
241{
242 struct inode *inode = dentry->d_inode;
243
244 return inode && IS_WHITEOUT(inode);
245}
246
247static bool ovl_is_opaquedir(struct dentry *dentry)
248{
249 int res;
250 char val;
251 struct inode *inode = dentry->d_inode;
252
253 if (!S_ISDIR(inode->i_mode) || !inode->i_op->getxattr)
254 return false;
255
256 res = inode->i_op->getxattr(dentry, ovl_opaque_xattr, &val, 1);
257 if (res == 1 && val == 'y')
258 return true;
259
260 return false;
261}
262
263static void ovl_dentry_release(struct dentry *dentry)
264{
265 struct ovl_entry *oe = dentry->d_fsdata;
266
267 if (oe) {
dd662667
MS
268 unsigned int i;
269
e9be9d5e 270 dput(oe->__upperdentry);
dd662667
MS
271 for (i = 0; i < oe->numlower; i++)
272 dput(oe->lowerstack[i].dentry);
e9be9d5e
MS
273 kfree_rcu(oe, rcu);
274 }
275}
276
277static const struct dentry_operations ovl_dentry_operations = {
278 .d_release = ovl_dentry_release,
279};
280
dd662667 281static struct ovl_entry *ovl_alloc_entry(unsigned int numlower)
e9be9d5e 282{
dd662667
MS
283 size_t size = offsetof(struct ovl_entry, lowerstack[numlower]);
284 struct ovl_entry *oe = kzalloc(size, GFP_KERNEL);
285
286 if (oe)
287 oe->numlower = numlower;
288
289 return oe;
e9be9d5e
MS
290}
291
292static inline struct dentry *ovl_lookup_real(struct dentry *dir,
293 struct qstr *name)
294{
295 struct dentry *dentry;
296
297 mutex_lock(&dir->d_inode->i_mutex);
298 dentry = lookup_one_len(name->name, dir, name->len);
299 mutex_unlock(&dir->d_inode->i_mutex);
300
301 if (IS_ERR(dentry)) {
302 if (PTR_ERR(dentry) == -ENOENT)
303 dentry = NULL;
304 } else if (!dentry->d_inode) {
305 dput(dentry);
306 dentry = NULL;
307 }
308 return dentry;
309}
310
5ef88da5
MS
311/*
312 * Returns next layer in stack starting from top.
313 * Returns -1 if this is the last layer.
314 */
315int ovl_path_next(int idx, struct dentry *dentry, struct path *path)
316{
317 struct ovl_entry *oe = dentry->d_fsdata;
318
319 BUG_ON(idx < 0);
320 if (idx == 0) {
321 ovl_path_upper(dentry, path);
322 if (path->dentry)
323 return oe->numlower ? 1 : -1;
324 idx++;
325 }
326 BUG_ON(idx > oe->numlower);
327 *path = oe->lowerstack[idx - 1];
328
329 return (idx < oe->numlower) ? idx + 1 : -1;
330}
331
e9be9d5e
MS
332struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
333 unsigned int flags)
334{
335 struct ovl_entry *oe;
3d3c6b89
MS
336 struct ovl_entry *poe = dentry->d_parent->d_fsdata;
337 struct path *stack = NULL;
338 struct dentry *upperdir, *upperdentry = NULL;
339 unsigned int ctr = 0;
e9be9d5e 340 struct inode *inode = NULL;
3d3c6b89
MS
341 bool upperopaque = false;
342 struct dentry *this, *prev = NULL;
343 unsigned int i;
e9be9d5e
MS
344 int err;
345
3d3c6b89 346 upperdir = ovl_upperdentry_dereference(poe);
e9be9d5e 347 if (upperdir) {
3d3c6b89
MS
348 this = ovl_lookup_real(upperdir, &dentry->d_name);
349 err = PTR_ERR(this);
350 if (IS_ERR(this))
351 goto out;
352
353 /*
354 * If this is not the lowermost layer, check whiteout and opaque
355 * directory.
356 */
357 if (poe->numlower && this) {
358 if (ovl_is_whiteout(this)) {
359 dput(this);
360 this = NULL;
361 upperopaque = true;
362 } else if (ovl_is_opaquedir(this)) {
363 upperopaque = true;
e9be9d5e
MS
364 }
365 }
3d3c6b89 366 upperdentry = prev = this;
e9be9d5e 367 }
3d3c6b89
MS
368
369 if (!upperopaque && poe->numlower) {
370 err = -ENOMEM;
371 stack = kcalloc(poe->numlower, sizeof(struct path), GFP_KERNEL);
372 if (!stack)
373 goto out_put_upper;
e9be9d5e
MS
374 }
375
3d3c6b89
MS
376 for (i = 0; !upperopaque && i < poe->numlower; i++) {
377 bool opaque = false;
378 struct path lowerpath = poe->lowerstack[i];
379
380 opaque = false;
381 this = ovl_lookup_real(lowerpath.dentry, &dentry->d_name);
382 err = PTR_ERR(this);
383 if (IS_ERR(this))
384 goto out_put;
385 if (!this)
386 continue;
387
388 /*
389 * If this is not the lowermost layer, check whiteout and opaque
390 * directory.
391 */
392 if (i < poe->numlower - 1) {
393 if (ovl_is_whiteout(this)) {
394 dput(this);
395 break;
396 } else if (ovl_is_opaquedir(this)) {
397 opaque = true;
398 }
399 }
400 /*
401 * If this is a non-directory then stop here.
402 *
403 * FIXME: check for opaqueness maybe better done in remove code.
404 */
405 if (!S_ISDIR(this->d_inode->i_mode)) {
406 opaque = true;
407 } else if (prev && (!S_ISDIR(prev->d_inode->i_mode) ||
408 !S_ISDIR(this->d_inode->i_mode))) {
409 if (prev == upperdentry)
410 upperopaque = true;
411 dput(this);
412 break;
413 }
414 stack[ctr].dentry = this;
415 stack[ctr].mnt = lowerpath.mnt;
416 ctr++;
417 prev = this;
418 if (opaque)
419 break;
e9be9d5e
MS
420 }
421
3d3c6b89
MS
422 oe = ovl_alloc_entry(ctr);
423 err = -ENOMEM;
424 if (!oe)
425 goto out_put;
426
427 if (upperdentry || ctr) {
e9be9d5e
MS
428 struct dentry *realdentry;
429
3d3c6b89
MS
430 realdentry = upperdentry ? upperdentry : stack[0].dentry;
431
e9be9d5e
MS
432 err = -ENOMEM;
433 inode = ovl_new_inode(dentry->d_sb, realdentry->d_inode->i_mode,
434 oe);
435 if (!inode)
3d3c6b89 436 goto out_free_oe;
e9be9d5e
MS
437 ovl_copyattr(realdentry->d_inode, inode);
438 }
439
3d3c6b89 440 oe->opaque = upperopaque;
e9be9d5e 441 oe->__upperdentry = upperdentry;
3d3c6b89
MS
442 memcpy(oe->lowerstack, stack, sizeof(struct path) * ctr);
443 kfree(stack);
e9be9d5e
MS
444 dentry->d_fsdata = oe;
445 d_add(dentry, inode);
446
447 return NULL;
448
3d3c6b89 449out_free_oe:
e9be9d5e 450 kfree(oe);
3d3c6b89
MS
451out_put:
452 for (i = 0; i < ctr; i++)
453 dput(stack[i].dentry);
454 kfree(stack);
455out_put_upper:
456 dput(upperdentry);
e9be9d5e
MS
457out:
458 return ERR_PTR(err);
459}
460
461struct file *ovl_path_open(struct path *path, int flags)
462{
463 return dentry_open(path, flags, current_cred());
464}
465
466static void ovl_put_super(struct super_block *sb)
467{
468 struct ovl_fs *ufs = sb->s_fs_info;
dd662667 469 unsigned i;
e9be9d5e
MS
470
471 dput(ufs->workdir);
472 mntput(ufs->upper_mnt);
dd662667
MS
473 for (i = 0; i < ufs->numlower; i++)
474 mntput(ufs->lower_mnt[i]);
e9be9d5e 475
f45827e8
EZ
476 kfree(ufs->config.lowerdir);
477 kfree(ufs->config.upperdir);
478 kfree(ufs->config.workdir);
e9be9d5e
MS
479 kfree(ufs);
480}
481
cc259639
AW
482/**
483 * ovl_statfs
484 * @sb: The overlayfs super block
485 * @buf: The struct kstatfs to fill in with stats
486 *
487 * Get the filesystem statistics. As writes always target the upper layer
488 * filesystem pass the statfs to the same filesystem.
489 */
490static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf)
491{
492 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
493 struct dentry *root_dentry = dentry->d_sb->s_root;
494 struct path path;
495 int err;
496
497 ovl_path_upper(root_dentry, &path);
498
499 err = vfs_statfs(&path, buf);
500 if (!err) {
501 buf->f_namelen = max(buf->f_namelen, ofs->lower_namelen);
502 buf->f_type = OVERLAYFS_SUPER_MAGIC;
503 }
504
505 return err;
506}
507
f45827e8
EZ
508/**
509 * ovl_show_options
510 *
511 * Prints the mount options for a given superblock.
512 * Returns zero; does not fail.
513 */
514static int ovl_show_options(struct seq_file *m, struct dentry *dentry)
515{
516 struct super_block *sb = dentry->d_sb;
517 struct ovl_fs *ufs = sb->s_fs_info;
518
519 seq_printf(m, ",lowerdir=%s", ufs->config.lowerdir);
520 seq_printf(m, ",upperdir=%s", ufs->config.upperdir);
521 seq_printf(m, ",workdir=%s", ufs->config.workdir);
522 return 0;
523}
524
e9be9d5e
MS
525static const struct super_operations ovl_super_operations = {
526 .put_super = ovl_put_super,
cc259639 527 .statfs = ovl_statfs,
f45827e8 528 .show_options = ovl_show_options,
e9be9d5e
MS
529};
530
531enum {
532 OPT_LOWERDIR,
533 OPT_UPPERDIR,
534 OPT_WORKDIR,
535 OPT_ERR,
536};
537
538static const match_table_t ovl_tokens = {
539 {OPT_LOWERDIR, "lowerdir=%s"},
540 {OPT_UPPERDIR, "upperdir=%s"},
541 {OPT_WORKDIR, "workdir=%s"},
542 {OPT_ERR, NULL}
543};
544
91c77947
MS
545static char *ovl_next_opt(char **s)
546{
547 char *sbegin = *s;
548 char *p;
549
550 if (sbegin == NULL)
551 return NULL;
552
553 for (p = sbegin; *p; p++) {
554 if (*p == '\\') {
555 p++;
556 if (!*p)
557 break;
558 } else if (*p == ',') {
559 *p = '\0';
560 *s = p + 1;
561 return sbegin;
562 }
563 }
564 *s = NULL;
565 return sbegin;
566}
567
e9be9d5e
MS
568static int ovl_parse_opt(char *opt, struct ovl_config *config)
569{
570 char *p;
571
91c77947 572 while ((p = ovl_next_opt(&opt)) != NULL) {
e9be9d5e
MS
573 int token;
574 substring_t args[MAX_OPT_ARGS];
575
576 if (!*p)
577 continue;
578
579 token = match_token(p, ovl_tokens, args);
580 switch (token) {
581 case OPT_UPPERDIR:
582 kfree(config->upperdir);
583 config->upperdir = match_strdup(&args[0]);
584 if (!config->upperdir)
585 return -ENOMEM;
586 break;
587
588 case OPT_LOWERDIR:
589 kfree(config->lowerdir);
590 config->lowerdir = match_strdup(&args[0]);
591 if (!config->lowerdir)
592 return -ENOMEM;
593 break;
594
595 case OPT_WORKDIR:
596 kfree(config->workdir);
597 config->workdir = match_strdup(&args[0]);
598 if (!config->workdir)
599 return -ENOMEM;
600 break;
601
602 default:
603 return -EINVAL;
604 }
605 }
606 return 0;
607}
608
609#define OVL_WORKDIR_NAME "work"
610
611static struct dentry *ovl_workdir_create(struct vfsmount *mnt,
612 struct dentry *dentry)
613{
614 struct inode *dir = dentry->d_inode;
615 struct dentry *work;
616 int err;
617 bool retried = false;
618
619 err = mnt_want_write(mnt);
620 if (err)
621 return ERR_PTR(err);
622
623 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
624retry:
625 work = lookup_one_len(OVL_WORKDIR_NAME, dentry,
626 strlen(OVL_WORKDIR_NAME));
627
628 if (!IS_ERR(work)) {
629 struct kstat stat = {
630 .mode = S_IFDIR | 0,
631 };
632
633 if (work->d_inode) {
634 err = -EEXIST;
635 if (retried)
636 goto out_dput;
637
638 retried = true;
639 ovl_cleanup(dir, work);
640 dput(work);
641 goto retry;
642 }
643
644 err = ovl_create_real(dir, work, &stat, NULL, NULL, true);
645 if (err)
646 goto out_dput;
647 }
648out_unlock:
649 mutex_unlock(&dir->i_mutex);
650 mnt_drop_write(mnt);
651
652 return work;
653
654out_dput:
655 dput(work);
656 work = ERR_PTR(err);
657 goto out_unlock;
658}
659
91c77947
MS
660static void ovl_unescape(char *s)
661{
662 char *d = s;
663
664 for (;; s++, d++) {
665 if (*s == '\\')
666 s++;
667 *d = *s;
668 if (!*s)
669 break;
670 }
671}
672
e9be9d5e
MS
673static int ovl_mount_dir(const char *name, struct path *path)
674{
675 int err;
91c77947
MS
676 char *tmp = kstrdup(name, GFP_KERNEL);
677
678 if (!tmp)
679 return -ENOMEM;
e9be9d5e 680
91c77947
MS
681 ovl_unescape(tmp);
682 err = kern_path(tmp, LOOKUP_FOLLOW, path);
e9be9d5e 683 if (err) {
91c77947 684 pr_err("overlayfs: failed to resolve '%s': %i\n", tmp, err);
e9be9d5e
MS
685 err = -EINVAL;
686 }
91c77947 687 kfree(tmp);
e9be9d5e
MS
688 return err;
689}
690
691static bool ovl_is_allowed_fs_type(struct dentry *root)
692{
693 const struct dentry_operations *dop = root->d_op;
694
695 /*
696 * We don't support:
697 * - automount filesystems
698 * - filesystems with revalidate (FIXME for lower layer)
699 * - filesystems with case insensitive names
700 */
701 if (dop &&
702 (dop->d_manage || dop->d_automount ||
703 dop->d_revalidate || dop->d_weak_revalidate ||
704 dop->d_compare || dop->d_hash)) {
705 return false;
706 }
707 return true;
708}
709
710/* Workdir should not be subdir of upperdir and vice versa */
711static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir)
712{
713 bool ok = false;
714
715 if (workdir != upperdir) {
716 ok = (lock_rename(workdir, upperdir) == NULL);
717 unlock_rename(workdir, upperdir);
718 }
719 return ok;
720}
721
722static int ovl_fill_super(struct super_block *sb, void *data, int silent)
723{
724 struct path lowerpath;
725 struct path upperpath;
726 struct path workpath;
727 struct inode *root_inode;
728 struct dentry *root_dentry;
729 struct ovl_entry *oe;
730 struct ovl_fs *ufs;
cc259639 731 struct kstatfs statfs;
dd662667
MS
732 struct vfsmount *mnt;
733 unsigned int i;
e9be9d5e
MS
734 int err;
735
f45827e8
EZ
736 err = -ENOMEM;
737 ufs = kzalloc(sizeof(struct ovl_fs), GFP_KERNEL);
738 if (!ufs)
e9be9d5e
MS
739 goto out;
740
f45827e8
EZ
741 err = ovl_parse_opt((char *) data, &ufs->config);
742 if (err)
743 goto out_free_config;
744
e9be9d5e
MS
745 /* FIXME: workdir is not needed for a R/O mount */
746 err = -EINVAL;
f45827e8
EZ
747 if (!ufs->config.upperdir || !ufs->config.lowerdir ||
748 !ufs->config.workdir) {
e9be9d5e
MS
749 pr_err("overlayfs: missing upperdir or lowerdir or workdir\n");
750 goto out_free_config;
751 }
752
753 err = -ENOMEM;
dd662667 754 oe = ovl_alloc_entry(1);
e9be9d5e 755 if (oe == NULL)
f45827e8 756 goto out_free_config;
e9be9d5e 757
f45827e8 758 err = ovl_mount_dir(ufs->config.upperdir, &upperpath);
e9be9d5e
MS
759 if (err)
760 goto out_free_oe;
761
f45827e8 762 err = ovl_mount_dir(ufs->config.lowerdir, &lowerpath);
e9be9d5e
MS
763 if (err)
764 goto out_put_upperpath;
765
f45827e8 766 err = ovl_mount_dir(ufs->config.workdir, &workpath);
e9be9d5e
MS
767 if (err)
768 goto out_put_lowerpath;
769
770 err = -EINVAL;
771 if (!S_ISDIR(upperpath.dentry->d_inode->i_mode) ||
772 !S_ISDIR(lowerpath.dentry->d_inode->i_mode) ||
773 !S_ISDIR(workpath.dentry->d_inode->i_mode)) {
774 pr_err("overlayfs: upperdir or lowerdir or workdir not a directory\n");
775 goto out_put_workpath;
776 }
777
778 if (upperpath.mnt != workpath.mnt) {
779 pr_err("overlayfs: workdir and upperdir must reside under the same mount\n");
780 goto out_put_workpath;
781 }
782 if (!ovl_workdir_ok(workpath.dentry, upperpath.dentry)) {
783 pr_err("overlayfs: workdir and upperdir must be separate subtrees\n");
784 goto out_put_workpath;
785 }
786
787 if (!ovl_is_allowed_fs_type(upperpath.dentry)) {
788 pr_err("overlayfs: filesystem of upperdir is not supported\n");
789 goto out_put_workpath;
790 }
791
792 if (!ovl_is_allowed_fs_type(lowerpath.dentry)) {
793 pr_err("overlayfs: filesystem of lowerdir is not supported\n");
794 goto out_put_workpath;
795 }
796
cc259639
AW
797 err = vfs_statfs(&lowerpath, &statfs);
798 if (err) {
799 pr_err("overlayfs: statfs failed on lowerpath\n");
800 goto out_put_workpath;
801 }
802 ufs->lower_namelen = statfs.f_namelen;
803
69c433ed
MS
804 sb->s_stack_depth = max(upperpath.mnt->mnt_sb->s_stack_depth,
805 lowerpath.mnt->mnt_sb->s_stack_depth) + 1;
806
807 err = -EINVAL;
808 if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) {
809 pr_err("overlayfs: maximum fs stacking depth exceeded\n");
810 goto out_put_workpath;
811 }
812
e9be9d5e
MS
813 ufs->upper_mnt = clone_private_mount(&upperpath);
814 err = PTR_ERR(ufs->upper_mnt);
815 if (IS_ERR(ufs->upper_mnt)) {
816 pr_err("overlayfs: failed to clone upperpath\n");
817 goto out_put_workpath;
818 }
819
dd662667
MS
820 ufs->lower_mnt = kcalloc(1, sizeof(struct vfsmount *), GFP_KERNEL);
821 if (ufs->lower_mnt == NULL)
e9be9d5e 822 goto out_put_upper_mnt;
dd662667
MS
823
824 mnt = clone_private_mount(&lowerpath);
825 err = PTR_ERR(mnt);
826 if (IS_ERR(mnt)) {
827 pr_err("overlayfs: failed to clone lowerpath\n");
828 goto out_put_lower_mnt;
e9be9d5e 829 }
dd662667
MS
830 /*
831 * Make lower_mnt R/O. That way fchmod/fchown on lower file
832 * will fail instead of modifying lower fs.
833 */
834 mnt->mnt_flags |= MNT_READONLY;
835
836 ufs->lower_mnt[0] = mnt;
837 ufs->numlower = 1;
e9be9d5e
MS
838
839 ufs->workdir = ovl_workdir_create(ufs->upper_mnt, workpath.dentry);
840 err = PTR_ERR(ufs->workdir);
841 if (IS_ERR(ufs->workdir)) {
842 pr_err("overlayfs: failed to create directory %s/%s\n",
f45827e8 843 ufs->config.workdir, OVL_WORKDIR_NAME);
e9be9d5e
MS
844 goto out_put_lower_mnt;
845 }
846
e9be9d5e
MS
847 /* If the upper fs is r/o, we mark overlayfs r/o too */
848 if (ufs->upper_mnt->mnt_sb->s_flags & MS_RDONLY)
849 sb->s_flags |= MS_RDONLY;
850
851 sb->s_d_op = &ovl_dentry_operations;
852
853 err = -ENOMEM;
854 root_inode = ovl_new_inode(sb, S_IFDIR, oe);
855 if (!root_inode)
856 goto out_put_workdir;
857
858 root_dentry = d_make_root(root_inode);
859 if (!root_dentry)
860 goto out_put_workdir;
861
862 mntput(upperpath.mnt);
863 mntput(lowerpath.mnt);
864 path_put(&workpath);
865
866 oe->__upperdentry = upperpath.dentry;
dd662667
MS
867 oe->lowerstack[0].dentry = lowerpath.dentry;
868 oe->lowerstack[0].mnt = ufs->lower_mnt[0];
e9be9d5e
MS
869
870 root_dentry->d_fsdata = oe;
871
cc259639 872 sb->s_magic = OVERLAYFS_SUPER_MAGIC;
e9be9d5e
MS
873 sb->s_op = &ovl_super_operations;
874 sb->s_root = root_dentry;
875 sb->s_fs_info = ufs;
876
877 return 0;
878
879out_put_workdir:
880 dput(ufs->workdir);
881out_put_lower_mnt:
dd662667
MS
882 for (i = 0; i < ufs->numlower; i++)
883 mntput(ufs->lower_mnt[i]);
884 kfree(ufs->lower_mnt);
e9be9d5e
MS
885out_put_upper_mnt:
886 mntput(ufs->upper_mnt);
887out_put_workpath:
888 path_put(&workpath);
889out_put_lowerpath:
890 path_put(&lowerpath);
891out_put_upperpath:
892 path_put(&upperpath);
893out_free_oe:
894 kfree(oe);
e9be9d5e 895out_free_config:
f45827e8
EZ
896 kfree(ufs->config.lowerdir);
897 kfree(ufs->config.upperdir);
898 kfree(ufs->config.workdir);
899 kfree(ufs);
e9be9d5e
MS
900out:
901 return err;
902}
903
904static struct dentry *ovl_mount(struct file_system_type *fs_type, int flags,
905 const char *dev_name, void *raw_data)
906{
907 return mount_nodev(fs_type, flags, raw_data, ovl_fill_super);
908}
909
910static struct file_system_type ovl_fs_type = {
911 .owner = THIS_MODULE,
ef94b186 912 .name = "overlay",
e9be9d5e
MS
913 .mount = ovl_mount,
914 .kill_sb = kill_anon_super,
915};
ef94b186 916MODULE_ALIAS_FS("overlay");
e9be9d5e
MS
917
918static int __init ovl_init(void)
919{
920 return register_filesystem(&ovl_fs_type);
921}
922
923static void __exit ovl_exit(void)
924{
925 unregister_filesystem(&ovl_fs_type);
926}
927
928module_init(ovl_init);
929module_exit(ovl_exit);