Date: 2016-12-21, 17-18:30 CET
Participants: Jarno R, Ben P, Ben M-C, Jiri B, Simon H, Zoltan B, Jan S
Summary:
* Status/progress/coordination of work packages:
1 - L3 packet handling in kernel datapath (net-next)
Upstreaming to net-next done, depending on Eric G's VLAN fixes.
Yi Yang has backported the net-next patches and submitted to ovs-dev on
2016-12-28
AP (Jarno): coordinate review of Yi's backported patches
Firsts tests backported patches on 4.9 kernel showed instabilities
Jiri: Better to start testing with net or net-next kernel.
AP (Jiri): help Zoltan set up test environment.
Q: How to match on presence/absence of Ethernet header in kernel datapath?
AP (Jiri): Check that the presence/absence of the Ethernet attribute in
the netlink flow mask is enough for the datapath to match on "packet_type"
(Ethernet vs L3). Can the kernel datapath handle an "all-zero" Ethernet mask
attribute as indicator to match on presence of Ethernet header without
matching on MAC addresses?
Jarno: How to properly reflect this in flow printouts?
2 - Layer 3 tunnel configuration (RedHat)
RTNETLINK patches on OVS are prerequisite to start testing L3 tunnels on
kernel datapath.
This patch is on the critical line.
Redhat (Eric G) working on this. Common understanding with Pravin
achieved.
AP (Jiri): Send update on status and ETA (probably after RedHat
conference next week)
3 - L3 ports in non-PTAP bridge (Ericssson)
Packet_type field introduced in structs dp_packet and flow.
Porting of earlier user-space patches to packet_type representation
completed.
L3 tunnel handling simplified compared to earlier patch sets. Impact now
solely on generic tunnel handler in ofproto. No longer impact on native
tnl-ports.
Functional tests done on netdev datapath. Make check successful.
Waiting for WPs 1 and 2 and internal clearance before submission.
4 - PTAP bridge (Ericsson)
Implementation ongoing: OXM field PACKET_TYPE added, bridge property
"packet-type-aware" added, alternative prerequisite "packet_type=(1,Ethertype)"
for existing L3 OXM match fields done, Interfacing PTAP bridge with versatile
tunnel ports done
Functional testing ongoing.
Error handling to be done.
Packet In/Out and Table Features not started.
5 - Basic EXT-382 (Ericsson)
Not started yet. Zoltan can start soon while test of L3 tunneling on
kernel datapath is waiting for WP 1 and 2.
6 - VXLAN-GPE tunnel (Intel)
Yi Yang has started to adapt the earlier patches to the new packet_type
concept (WP3)
7 - NSH MD1 Match Fields (Intel)
Yi Yang will adapt earlier patches to new OXM point codes and make sure
the fields have the right pre-requisites so that they are only used as packet
match fields, not as metadata fields. This can run in parallel to WPs 4&5.
8 - Encap/Decap NSH MD1 (Intel)
Depends on WP 5.
9 - GEN_TLV registers (VMware)
Jarno will have time to start with this in 2-3 weeks.
10 - NSH MD2 Support (Intel)
Depend on WP 9.
* ONF specification drafts for NSH and Generic TLV Mapping
Ongoing merger of ONF with ON.Lab stresses the importance of open-source
implementations for evolving the ONF specifications. The implementation of NSH
in OVS is welcome in this context.
Jan will draft a short ONF specification for NSH using the official ONF spec
template, based on the content of the Google doc for presentation and
discussion in the Oped datapath WG in ONF.
AP (Ben M-C): Check if there is any example spec to look at.
The draft spec for handling generic TLV fields will be started when the
design in OVS is clearer.
* Any other issue
-
Next meeting: Tue 2016-01-24, 17:00 CET
Link to the Google design doc:
https://docs.google.com/document/d/1oWMYUH8sjZJzWa72o2q9kU0N6pNE-rwZcLH3-kbbDR8/edit
Regards, Jan
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
