On 24/04/2017 14:35, Simon Horman wrote:
On Tue, Apr 18, 2017 at 03:18:55PM +0300, Roi Dayan wrote:
On 14/04/2017 04:11, Joe Stringer wrote:
On 7 April 2017 at 06:12, Roi Dayan <[email protected]> wrote:
From: Paul Blakey <[email protected]>
Signed-off-by: Paul Blakey <[email protected]>
Reviewed-by: Roi Dayan <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
---
<snip>
diff --git a/lib/netdev.h b/lib/netdev.h
index d6c07c1..6d2db7d 100644
--- a/lib/netdev.h
+++ b/lib/netdev.h
@@ -156,6 +156,29 @@ int netdev_send(struct netdev *, int qid, struct
dp_packet_batch *,
bool may_steal, bool concurrent_txq);
void netdev_send_wait(struct netdev *, int qid);
+/* Flow offloading. */
+struct offload_info {
+ const void *port_hmap_obj; /* To query ports info from netdev port map */
+ ovs_be16 tp_dst_port; /* Destination port for tunnel in SET action */
Is this assuming there is only ever one tunnel destination port? What
about multiple output?
Yes. we currently only support single dst port output. if there is more than
1 we fail with EOPNOTSUPP and fallback to OVS datapath.
the check is done in dpif-netlink.c parse_flow_put()
FWIW, from my PoV this is a shortcoming of the feature-set provided by
this patch-set and is something I would like to see lifted at some point in
the future - I would be interested in doing so if it isn't on anyone else's
todo list. Perhaps it would be useful to formally list what offloads are
supported by the API?
The above notwithstanding, I would expect that output to a single tunnel
vport covers a lot of use-cases.
Yes, we do want to support multiple destinations and considered it as a
later feature which is on our TODO, though we haven't started working on
this yet.
As for a formal list of what is offloads are supported, I guess we can
list it later in the documentation.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
