Ben pointed out off list that controllers will expect the flow metadata to be
cleared when traversing a patch port.
---
ofproto/ofproto-dpif.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c
index d5bd0b4..d75a63c 100644
--- a/ofproto/ofproto-dpif.c
+++ b/ofproto/ofproto-dpif.c
@@ -5508,6 +5508,10 @@ compose_output_action__(struct action_xlate_ctx *ctx,
uint16_t ofp_port,
struct priority_to_dscp *pdscp;
uint32_t out_port, odp_port;
+ /* If 'struct flow' gets additional metadata, we'll need to zero it out
+ * before traversing a patch port. */
+ BUILD_ASSERT_DECL(FLOW_WC_SEQ == 18);
+
if (!ofport) {
xlate_report(ctx, "Nonexistent output port");
return;
@@ -5537,6 +5541,9 @@ compose_output_action__(struct action_xlate_ctx *ctx,
uint16_t ofp_port,
ctx->ofproto = ofproto_dpif_cast(peer->up.ofproto);
ctx->flow.in_port = peer->up.ofp_port;
+ ctx->flow.metadata = htonll(0);
+ memset(&ctx->flow.tunnel, 0, sizeof ctx->flow.tunnel);
+ memset(ctx->flow.regs, 0, sizeof ctx->flow.regs);
xlate_table_action(ctx, ctx->flow.in_port, 0, true);
ctx->flow = old_flow;
ctx->ofproto = ofproto_dpif_cast(ofport->up.ofproto);
--
1.7.9.5
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
