2 * Copyright (C) 2011 Novell Inc.
3 * Copyright (C) 2016 Red Hat, Inc.
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.
11 #include <linux/cred.h>
12 #include <linux/namei.h>
13 #include <linux/xattr.h>
14 #include <linux/ratelimit.h>
15 #include <linux/mount.h>
16 #include <linux/exportfs.h>
17 #include "overlayfs.h"
18 #include "ovl_entry.h"
20 struct ovl_lookup_data {
29 static int ovl_check_redirect(struct dentry *dentry, struct ovl_lookup_data *d,
30 size_t prelen, const char *post)
33 char *s, *next, *buf = NULL;
35 res = vfs_getxattr(dentry, OVL_XATTR_REDIRECT, NULL, 0);
37 if (res == -ENODATA || res == -EOPNOTSUPP)
41 buf = kzalloc(prelen + res + strlen(post) + 1, GFP_KERNEL);
48 res = vfs_getxattr(dentry, OVL_XATTR_REDIRECT, buf, res);
54 for (s = buf; *s++ == '/'; s = next) {
55 next = strchrnul(s, '/');
60 if (strchr(buf, '/') != NULL)
63 memmove(buf + prelen, buf, res);
64 memcpy(buf, d->name.name, prelen);
70 d->name.name = d->redirect;
71 d->name.len = strlen(d->redirect);
79 pr_warn_ratelimited("overlayfs: failed to get redirect (%i)\n", res);
82 pr_warn_ratelimited("overlayfs: invalid redirect (%s)\n", buf);
86 static int ovl_acceptable(void *ctx, struct dentry *dentry)
91 static struct ovl_fh *ovl_get_origin_fh(struct dentry *dentry)
94 struct ovl_fh *fh = NULL;
96 res = vfs_getxattr(dentry, OVL_XATTR_ORIGIN, NULL, 0);
98 if (res == -ENODATA || res == -EOPNOTSUPP)
102 /* Zero size value means "copied up but origin unknown" */
106 fh = kzalloc(res, GFP_KERNEL);
108 return ERR_PTR(-ENOMEM);
110 res = vfs_getxattr(dentry, OVL_XATTR_ORIGIN, fh, res);
114 if (res < sizeof(struct ovl_fh) || res < fh->len)
117 if (fh->magic != OVL_FH_MAGIC)
120 /* Treat larger version and unknown flags as "origin unknown" */
121 if (fh->version > OVL_FH_VERSION || fh->flags & ~OVL_FH_FLAG_ALL)
124 /* Treat endianness mismatch as "origin unknown" */
125 if (!(fh->flags & OVL_FH_FLAG_ANY_ENDIAN) &&
126 (fh->flags & OVL_FH_FLAG_BIG_ENDIAN) != OVL_FH_FLAG_CPU_ENDIAN)
136 pr_warn_ratelimited("overlayfs: failed to get origin (%i)\n", res);
139 pr_warn_ratelimited("overlayfs: invalid origin (%*phN)\n", res, fh);
143 static struct dentry *ovl_get_origin(struct dentry *dentry,
144 struct vfsmount *mnt)
146 struct dentry *origin = NULL;
147 struct ovl_fh *fh = ovl_get_origin_fh(dentry);
150 if (IS_ERR_OR_NULL(fh))
151 return (struct dentry *)fh;
154 * Make sure that the stored uuid matches the uuid of the lower
155 * layer where file handle will be decoded.
157 if (!uuid_equal(&fh->uuid, &mnt->mnt_sb->s_uuid))
160 bytes = (fh->len - offsetof(struct ovl_fh, fid));
161 origin = exportfs_decode_fh(mnt, (struct fid *)fh->fid,
162 bytes >> 2, (int)fh->type,
163 ovl_acceptable, NULL);
164 if (IS_ERR(origin)) {
165 /* Treat stale file handle as "origin unknown" */
166 if (origin == ERR_PTR(-ESTALE))
171 if (ovl_dentry_weird(origin) ||
172 ((d_inode(origin)->i_mode ^ d_inode(dentry)->i_mode) & S_IFMT))
180 pr_warn_ratelimited("overlayfs: invalid origin (%pd2)\n", origin);
186 static bool ovl_is_opaquedir(struct dentry *dentry)
188 return ovl_check_dir_xattr(dentry, OVL_XATTR_OPAQUE);
191 static int ovl_lookup_single(struct dentry *base, struct ovl_lookup_data *d,
192 const char *name, unsigned int namelen,
193 size_t prelen, const char *post,
199 this = lookup_one_len_unlocked(name, base, namelen);
203 if (err == -ENOENT || err == -ENAMETOOLONG)
210 if (ovl_dentry_weird(this)) {
211 /* Don't support traversing automounts and other weirdness */
215 if (ovl_is_whiteout(this)) {
216 d->stop = d->opaque = true;
219 if (!d_can_lookup(this)) {
226 if (!d->last && ovl_is_opaquedir(this)) {
227 d->stop = d->opaque = true;
230 err = ovl_check_redirect(this, d, prelen, post);
247 static int ovl_lookup_layer(struct dentry *base, struct ovl_lookup_data *d,
250 /* Counting down from the end, since the prefix can change */
251 size_t rem = d->name.len - 1;
252 struct dentry *dentry = NULL;
255 if (d->name.name[0] != '/')
256 return ovl_lookup_single(base, d, d->name.name, d->name.len,
259 while (!IS_ERR_OR_NULL(base) && d_can_lookup(base)) {
260 const char *s = d->name.name + d->name.len - rem;
261 const char *next = strchrnul(s, '/');
262 size_t thislen = next - s;
265 /* Verify we did not go off the rails */
266 if (WARN_ON(s[-1] != '/'))
269 err = ovl_lookup_single(base, d, s, thislen,
270 d->name.len - rem, next, &base);
280 if (WARN_ON(rem >= d->name.len))
288 static int ovl_check_origin(struct dentry *upperdentry,
289 struct path *lowerstack, unsigned int numlower,
290 struct path **stackp, unsigned int *ctrp)
292 struct vfsmount *mnt;
293 struct dentry *origin = NULL;
297 for (i = 0; i < numlower; i++) {
298 mnt = lowerstack[i].mnt;
299 origin = ovl_get_origin(upperdentry, mnt);
301 return PTR_ERR(origin);
312 *stackp = kmalloc(sizeof(struct path), GFP_KERNEL);
317 **stackp = (struct path) { .dentry = origin, .mnt = mnt };
324 * Verify that @fh matches the origin file handle stored in OVL_XATTR_ORIGIN.
325 * Return 0 on match, -ESTALE on mismatch, < 0 on error.
327 static int ovl_verify_origin_fh(struct dentry *dentry, const struct ovl_fh *fh)
329 struct ovl_fh *ofh = ovl_get_origin_fh(dentry);
338 if (fh->len != ofh->len || memcmp(fh, ofh, fh->len))
346 * Verify that an inode matches the origin file handle stored in upper inode.
348 * If @set is true and there is no stored file handle, encode and store origin
349 * file handle in OVL_XATTR_ORIGIN.
351 * Return 0 on match, -ESTALE on mismatch, < 0 on error.
353 int ovl_verify_origin(struct dentry *dentry, struct vfsmount *mnt,
354 struct dentry *origin, bool is_upper, bool set)
360 fh = ovl_encode_fh(origin, is_upper);
365 err = ovl_verify_origin_fh(dentry, fh);
366 if (set && err == -ENODATA)
367 err = ovl_do_setxattr(dentry, OVL_XATTR_ORIGIN, fh, fh->len, 0);
376 inode = d_inode(origin);
377 pr_warn_ratelimited("overlayfs: failed to verify origin (%pd2, ino=%lu, err=%i)\n",
378 origin, inode ? inode->i_ino : 0, err);
383 * Verify that an index entry name matches the origin file handle stored in
384 * OVL_XATTR_ORIGIN and that origin file handle can be decoded to lower path.
385 * Return 0 on match, -ESTALE on mismatch or stale origin, < 0 on error.
387 int ovl_verify_index(struct dentry *index, struct path *lowerstack,
388 unsigned int numlower)
390 struct ovl_fh *fh = NULL;
392 struct path origin = { };
393 struct path *stack = &origin;
394 unsigned int ctr = 0;
401 * Directory index entries are going to be used for looking up
402 * redirected upper dirs by lower dir fh when decoding an overlay
403 * file handle of a merge dir. Whiteout index entries are going to be
404 * used as an indication that an exported overlay file handle should
405 * be treated as stale (i.e. after unlink of the overlay inode).
406 * We don't know the verification rules for directory and whiteout
407 * index entries, because they have not been implemented yet, so return
408 * EROFS if those entries are found to avoid corrupting an index that
409 * was created by a newer kernel.
412 if (d_is_dir(index) || ovl_is_whiteout(index))
416 if (index->d_name.len < sizeof(struct ovl_fh)*2)
420 len = index->d_name.len / 2;
421 fh = kzalloc(len, GFP_KERNEL);
426 if (hex2bin((u8 *)fh, index->d_name.name, len) || len != fh->len)
429 err = ovl_verify_origin_fh(index, fh);
433 err = ovl_check_origin(index, lowerstack, numlower, &stack, &ctr);
439 /* Check if index is orphan and don't warn before cleaning it */
440 if (d_inode(index)->i_nlink == 1 &&
441 ovl_get_nlink(index, origin.dentry, 0) == 0)
450 pr_warn_ratelimited("overlayfs: failed to verify index (%pd2, ftype=%x, err=%i)\n",
451 index, d_inode(index)->i_mode & S_IFMT, err);
456 * Lookup in indexdir for the index entry of a lower real inode or a copy up
457 * origin inode. The index entry name is the hex representation of the lower
460 * If the index dentry in negative, then either no lower aliases have been
461 * copied up yet, or aliases have been copied up in older kernels and are
464 * If the index dentry for a copy up origin inode is positive, but points
465 * to an inode different than the upper inode, then either the upper inode
466 * has been copied up and not indexed or it was indexed, but since then
467 * index dir was cleared. Either way, that index cannot be used to indentify
470 int ovl_get_index_name(struct dentry *origin, struct qstr *name)
476 fh = ovl_encode_fh(origin, false);
481 n = kzalloc(fh->len * 2, GFP_KERNEL);
483 s = bin2hex(n, fh, fh->len);
484 *name = (struct qstr) QSTR_INIT(n, s - n);
493 static struct dentry *ovl_lookup_index(struct dentry *dentry,
494 struct dentry *upper,
495 struct dentry *origin)
497 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
498 struct dentry *index;
503 err = ovl_get_index_name(origin, &name);
507 index = lookup_one_len_unlocked(name.name, ofs->indexdir, name.len);
509 err = PTR_ERR(index);
510 pr_warn_ratelimited("overlayfs: failed inode index lookup (ino=%lu, key=%*s, err=%i);\n"
511 "overlayfs: mount with '-o index=off' to disable inodes index.\n",
512 d_inode(origin)->i_ino, name.len, name.name,
517 inode = d_inode(index);
518 if (d_is_negative(index)) {
519 if (upper && d_inode(origin)->i_nlink > 1) {
520 pr_warn_ratelimited("overlayfs: hard link with origin but no index (ino=%lu).\n",
521 d_inode(origin)->i_ino);
527 } else if (upper && d_inode(upper) != inode) {
528 pr_warn_ratelimited("overlayfs: wrong index found (index=%pd2, ino=%lu, upper ino=%lu).\n",
529 index, inode->i_ino, d_inode(upper)->i_ino);
531 } else if (ovl_dentry_weird(index) || ovl_is_whiteout(index) ||
532 ((inode->i_mode ^ d_inode(origin)->i_mode) & S_IFMT)) {
534 * Index should always be of the same file type as origin
535 * except for the case of a whiteout index. A whiteout
536 * index should only exist if all lower aliases have been
537 * unlinked, which means that finding a lower origin on lookup
538 * whose index is a whiteout should be treated as an error.
540 pr_warn_ratelimited("overlayfs: bad index found (index=%pd2, ftype=%x, origin ftype=%x).\n",
541 index, d_inode(index)->i_mode & S_IFMT,
542 d_inode(origin)->i_mode & S_IFMT);
552 index = ERR_PTR(-EIO);
557 * Returns next layer in stack starting from top.
558 * Returns -1 if this is the last layer.
560 int ovl_path_next(int idx, struct dentry *dentry, struct path *path)
562 struct ovl_entry *oe = dentry->d_fsdata;
566 ovl_path_upper(dentry, path);
568 return oe->numlower ? 1 : -1;
571 BUG_ON(idx > oe->numlower);
572 *path = oe->lowerstack[idx - 1];
574 return (idx < oe->numlower) ? idx + 1 : -1;
577 struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
580 struct ovl_entry *oe;
581 const struct cred *old_cred;
582 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
583 struct ovl_entry *poe = dentry->d_parent->d_fsdata;
584 struct ovl_entry *roe = dentry->d_sb->s_root->d_fsdata;
585 struct path *stack = NULL;
586 struct dentry *upperdir, *upperdentry = NULL;
587 struct dentry *index = NULL;
588 unsigned int ctr = 0;
589 struct inode *inode = NULL;
590 bool upperopaque = false;
591 char *upperredirect = NULL;
595 struct ovl_lookup_data d = {
596 .name = dentry->d_name,
600 .last = !poe->numlower,
604 if (dentry->d_name.len > ofs->namelen)
605 return ERR_PTR(-ENAMETOOLONG);
607 old_cred = ovl_override_creds(dentry->d_sb);
608 upperdir = ovl_dentry_upper(dentry->d_parent);
610 err = ovl_lookup_layer(upperdir, &d, &upperdentry);
614 if (upperdentry && unlikely(ovl_dentry_remote(upperdentry))) {
619 if (upperdentry && !d.is_dir) {
620 BUG_ON(!d.stop || d.redirect);
622 * Lookup copy up origin by decoding origin file handle.
623 * We may get a disconnected dentry, which is fine,
624 * because we only need to hold the origin inode in
625 * cache and use its inode number. We may even get a
626 * connected dentry, that is not under any of the lower
627 * layers root. That is also fine for using it's inode
628 * number - it's the same as if we held a reference
629 * to a dentry in lower layer that was moved under us.
631 err = ovl_check_origin(upperdentry, roe->lowerstack,
632 roe->numlower, &stack, &ctr);
638 upperredirect = kstrdup(d.redirect, GFP_KERNEL);
641 if (d.redirect[0] == '/')
644 upperopaque = d.opaque;
647 if (!d.stop && poe->numlower) {
649 stack = kcalloc(ofs->numlower, sizeof(struct path),
655 for (i = 0; !d.stop && i < poe->numlower; i++) {
656 struct path lowerpath = poe->lowerstack[i];
658 d.last = i == poe->numlower - 1;
659 err = ovl_lookup_layer(lowerpath.dentry, &d, &this);
666 stack[ctr].dentry = this;
667 stack[ctr].mnt = lowerpath.mnt;
673 if (d.redirect && d.redirect[0] == '/' && poe != roe) {
676 /* Find the current layer on the root dentry */
677 for (i = 0; i < poe->numlower; i++)
678 if (poe->lowerstack[i].mnt == lowerpath.mnt)
680 if (WARN_ON(i == poe->numlower))
685 /* Lookup index by lower inode and verify it matches upper inode */
686 if (ctr && !d.is_dir && ovl_indexdir(dentry->d_sb)) {
687 struct dentry *origin = stack[0].dentry;
689 index = ovl_lookup_index(dentry, upperdentry, origin);
691 err = PTR_ERR(index);
697 oe = ovl_alloc_entry(ctr);
702 oe->opaque = upperopaque;
703 memcpy(oe->lowerstack, stack, sizeof(struct path) * ctr);
704 dentry->d_fsdata = oe;
707 ovl_dentry_set_upper_alias(dentry);
709 upperdentry = dget(index);
711 if (upperdentry || ctr) {
712 inode = ovl_get_inode(dentry, upperdentry);
713 err = PTR_ERR(inode);
717 OVL_I(inode)->redirect = upperredirect;
719 ovl_set_flag(OVL_INDEX, inode);
722 revert_creds(old_cred);
726 d_add(dentry, inode);
731 dentry->d_fsdata = NULL;
735 for (i = 0; i < ctr; i++)
736 dput(stack[i].dentry);
740 kfree(upperredirect);
743 revert_creds(old_cred);
747 bool ovl_lower_positive(struct dentry *dentry)
749 struct ovl_entry *oe = dentry->d_fsdata;
750 struct ovl_entry *poe = dentry->d_parent->d_fsdata;
751 const struct qstr *name = &dentry->d_name;
753 bool positive = false;
757 * If dentry is negative, then lower is positive iff this is a
760 if (!dentry->d_inode)
763 /* Negative upper -> positive lower */
764 if (!ovl_dentry_upper(dentry))
767 /* Positive upper -> have to look up lower to see whether it exists */
768 for (i = 0; !done && !positive && i < poe->numlower; i++) {
770 struct dentry *lowerdir = poe->lowerstack[i].dentry;
772 this = lookup_one_len_unlocked(name->name, lowerdir,
775 switch (PTR_ERR(this)) {
782 * Assume something is there, we just couldn't
790 positive = !ovl_is_whiteout(this);