YAMA et al rely on on i_uid/i_gid to be populated in order to perform their checks. While these really cannot be guarenteed as the underlying filesystem may not even have the concept, they are expected to be filled when possible. To quote Al Viro:
"Ideally, yes, we'd want to have ->i_uid used only by fs-specific code and helpers used by that fs (including those that are implicit defaults). [...] In practice we have enough places where uid/gid is used directly to make setting them practically a requirement - places like /proc/<pid>/ can get away with not doing that, but only because shitloads of syscalls are not allowed on those anyway, permissions or no permissions. In anything general-purpose you really need to set it." Copy up the underlying filesystem information into the overlayfs inode when we create it. BugLink: http://bugs.launchpad.net/bugs/944386 Signed-off-by: Andy Whitcroft <a...@canonical.com> --- fs/overlayfs/dir.c | 2 ++ fs/overlayfs/overlayfs.h | 6 ++++++ fs/overlayfs/super.c | 1 + 3 files changed, 9 insertions(+) After a long hiatus I have had time to look into the issues highlighted by the i_uid/i_gid requirements from the VFS. I have identified a number of places which definatly did need the ids copying up and those are reflected in the patch below. I am not 100% convinced I have hit all of the places this might be needed but it cirtainly helps with the issues I was seeing with link and YAMA (which given YAMA is now gaining the link constraints in mainline in v3.6 we will see more issues here). were seeing and identify the places where Please consider for overlayfs. -apw diff --git a/fs/overlayfs/dir.c b/fs/overlayfs/dir.c index c914c97..084e527 100644 --- a/fs/overlayfs/dir.c +++ b/fs/overlayfs/dir.c @@ -304,6 +304,7 @@ static int ovl_create_object(struct dentry *dentry, int mode, dev_t rdev, } } ovl_dentry_update(dentry, newdentry); + ovl_copyattr(newdentry->d_inode, inode); d_instantiate(dentry, inode); inode = NULL; newdentry = NULL; @@ -446,6 +447,7 @@ static int ovl_link(struct dentry *old, struct inode *newdir, new->d_fsdata); if (!newinode) goto link_fail; + ovl_copyattr(upperdir->d_inode, newinode); ovl_dentry_version_inc(new->d_parent); ovl_dentry_update(new, newdentry); diff --git a/fs/overlayfs/overlayfs.h b/fs/overlayfs/overlayfs.h index 1dd05f7..3495a55 100644 --- a/fs/overlayfs/overlayfs.h +++ b/fs/overlayfs/overlayfs.h @@ -56,6 +56,12 @@ int ovl_removexattr(struct dentry *dentry, const char *name); struct inode *ovl_new_inode(struct super_block *sb, umode_t mode, struct ovl_entry *oe); +static inline void ovl_copyattr(struct inode *from, struct inode *to) +{ + to->i_uid = from->i_uid; + to->i_gid = from->i_gid; +} + /* dir.c */ extern const struct inode_operations ovl_dir_inode_operations; diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c index 1d2d1e2..23cac54 100644 --- a/fs/overlayfs/super.c +++ b/fs/overlayfs/super.c @@ -343,6 +343,7 @@ static int ovl_do_lookup(struct dentry *dentry) oe); if (!inode) goto out_dput; + ovl_copyattr(realdentry->d_inode, inode); } if (upperdentry) -- 1.7.9.5 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/