ipv4: keep skb->dst around in presence of IP options

Andrey Konovalov got crashes in __ip_options_echo() when a NULL skb->dst
is accessed.

ipv4_pktinfo_prepare() should not drop the dst if (evil) IP options
are present.

We could refine the test to the presence of ts_needtime or srr,
but IP options are not often used, so let's be conservative.

Thanks to syzkaller team for finding this bug.

Fixes: d826eb1 ("ipv4: PKTINFO doesnt need dst reference")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/ipv4/ip_sockglue.c

@@ -1238,7 +1238,14 @@ void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
 		pktinfo->ipi_ifindex = 0;
 		pktinfo->ipi_spec_dst.s_addr = 0;
 	}
-	skb_dst_drop(skb);
+	/* We need to keep the dst for __ip_options_echo()
+	 * We could restrict the test to opt.ts_needtime || opt.srr,
+	 * but the following is good enough as IP options are not often used.
+	 */
+	if (unlikely(IPCB(skb)->opt.optlen))
+		skb_dst_force(skb);
+	else
+		skb_dst_drop(skb);
 }
 
 int ip_setsockopt(struct sock *sk, int level,