[ovs-git] [ovn-org/ovn] 24ac0a: Add VXLAN support for non-VTEP datapath bindings

Ihar Hrachyshka noreply at github.com
Thu Sep 17 07:12:17 UTC 2020


  Branch: refs/heads/branch-20.09
  Home:   https://github.com/ovn-org/ovn
  Commit: 24ac0a496ace531065d6b7c410962caaa0984d6b
      https://github.com/ovn-org/ovn/commit/24ac0a496ace531065d6b7c410962caaa0984d6b
  Author: Ihar Hrachyshka <ihrachys at redhat.com>
  Date:   2020-09-17 (Thu, 17 Sep 2020)

  Changed paths:
    M NEWS
    M controller/physical.c
    M include/ovn/logical-fields.h
    M lib/ovn-util.h
    M northd/ovn-northd.c
    M ovn-architecture.7.xml
    M ovn-nb.xml
    M tests/ovn-macros.at
    M tests/ovn.at

  Log Message:
  -----------
  Add VXLAN support for non-VTEP datapath bindings

Because of limited space in VXLAN VNI to pass over all three of -
datapath id, ingress port, egress port - the implementation ignores
ingress; and splits the remaining 24 bits of VNI into two chunks, 12
bits each - one for datapath and one for egress port.

This also bumps priority for ramp switch flows to 110 to prioritize
them over regular non-ramp VXLAN flows.

Limitations: because ingress port is not passed, ACLs that rely on it
won't work with VXLAN; reduced number of networks and ports per
network (max 4096 for both).

NB consumers may use NB_Global options:max_tunid to determine maximum
capacity for logical switches supported by the setup.

Renamed MLF_RCV_FROM_VXLAN_BIT into MLF_RCV_FROM_RAMP_BIT to reflect
the new use case.

Added test scenarios that ping through VXLAN tunnel between two
hypervisors added. Also max_tunid is validated.

Signed-off-by: Ihar Hrachyshka <ihrachys at redhat.com>
Signed-off-by: Numan Siddique <numans at ovn.org>




More information about the git mailing list