xfrm: Refuse to insert 32 bit userspace socket policies on 64 bit systems
authorSteffen Klassert <steffen.klassert@secunet.com>
Thu, 1 Feb 2018 07:49:23 +0000 (08:49 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 8 Apr 2018 12:26:29 +0000 (14:26 +0200)
commit 19d7df69fdb2636856dc8919de72fc1bf8f79598 upstream.

We don't have a compat layer for xfrm, so userspace and kernel
structures have different sizes in this case. This results in
a broken configuration, so refuse to configure socket policies
when trying to insert from 32 bit userspace as we do it already
with policies inserted via netlink.

Reported-and-tested-by: syzbot+e1a1577ca8bcb47b769a@syzkaller.appspotmail.com
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/xfrm/xfrm_state.c

index bab20c6..8f13fb5 100644 (file)
@@ -2050,6 +2050,11 @@ int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen
        struct xfrm_mgr *km;
        struct xfrm_policy *pol = NULL;
 
+#ifdef CONFIG_COMPAT
+       if (in_compat_syscall())
+               return -EOPNOTSUPP;
+#endif
+
        if (!optval && !optlen) {
                xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
                xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);