On Wed, Apr 16, 2014 at 11:49 PM, Andy Zhou <az...@nicira.com> wrote:
> diff --git a/datapath/actions.c b/datapath/actions.c
> index 820075f..9307ee1 100644
> --- a/datapath/actions.c
> +++ b/datapath/actions.c
> +static int execute_recirc(struct datapath *dp, struct sk_buff *skb,
> + const struct nlattr *a)
> +{
> + struct sw_flow_key recirc_key;
> + const struct vport *p = OVS_CB(skb)->input_vport;
> + uint32_t hash = OVS_CB(skb)->pkt_key->ovs_flow_hash;
> + int err;
> +
> + err = ovs_flow_extract(skb, p->port_no, &recirc_key);
I think there's a possibility of OVS_CB(skb)->input_vport being NULL
if the packet comes from userspace through ovs_packet_cmd_execute().
> diff --git a/datapath/flow_netlink.c b/datapath/flow_netlink.c
> index f464a5b..6edf54b 100644
> --- a/datapath/flow_netlink.c
> +++ b/datapath/flow_netlink.c
> @@ -902,6 +911,10 @@ int ovs_nla_put_flow(const struct sw_flow_key *swkey,
> output->ovs_flow_hash)))
> goto nla_put_failure;
>
> + if ((swkey->recirc_id) &&
> + (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id)))
> + goto nla_put_failure;
I think the extra sets of parentheses around each clause is
unnecessary - and actually I see this is in the hash patch as well.
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
