tunnel: Correctly check for internal device.

With header caching we check to see if the next device in the stack
is an OVS device and, if so, cache that flow as well.  However, the
test for this called internal_dev_get_vport() assuming that it would
return NULL if the device is not an internal device.  It doesn't,
however, it just returns the offset from the device where the vport
data structure would be if it were an internal device.  This changes
it to explicitly check for an internal device first to avoid a panic.

Bug #3470

Reported-by: Ram Jothikumar <rjothikumar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Reviewed-by: Justin Pettit <jpettit@nicira.com>

diff --git a/datapath/tunnel.c b/datapath/tunnel.c
index 77f976f..ad45228 100644 (file)
--- a/datapath/tunnel.c
+++ b/datapath/tunnel.c
@@ -1266,17 +1266,15 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
 
                if (likely(cache)) {
                        int orig_len = skb->len - cache->len;
-                       struct vport *cache_vport = internal_dev_get_vport(rt_dst(rt).dev);
 
                        skb->protocol = htons(ETH_P_IP);
-
                        iph->tot_len = htons(skb->len - skb_network_offset(skb));
                        ip_send_check(iph);
 
-                       if (likely(cache_vport)) {
+                       if (is_internal_dev(rt_dst(rt).dev)) {
                                OVS_CB(skb)->flow = cache->flow;
                                compute_ip_summed(skb, true);
-                               vport_receive(cache_vport, skb);
+                               vport_receive(internal_dev_get_vport(rt_dst(rt).dev), skb);
                                sent_len += orig_len;
                        } else {
                                int err;