[PATCH frr 0/2] Fix leaked EVPN routes having wrong nexthop on IPv4 via IPv6 routes

[Gabriel Goller <g.goller@proxmox.com>]

When we have a EVPN fabric with IPv6 VTEPs and leak them into the default
routing table (or any other VRF) (via exit-nodes in our case) the nexthop is
wrong. The nexthop is correct in the vrf (e.g. `ip route show vrf vrf_evpn`) but
the import-vrf does not correctly check the BGP MP properties for the IPv6
nexthop and inserts a invalid (0.0.0.0) IPv4 nexthop. The BGP MP properties are
not checked because a legacy bgp flag is set (BGP_ATTR_NEXT_HOP), which
circumvents the BGP MP check.

Fix this by not setting this flag when a IPv6 nexthop is used. Backport an
upstream commit and make another small change on top of it so that this also
works with EVPN routes and not only general BGP MP.

Thanks @HannesL for the quick testing.

@Thomas you can also cherry pick the patches from the other frr series and do a
single bump, or we can apply this patch as-is and think over other series.


frr:

Gabriel Goller (2):
  frr: backport #21166 and #21958, fixing EVPN IPv4 routes with IPv6
    nexhtop
  bump to version 10.6.1-1+pve2

 debian/changelog                              |   6 +
 debian/patches/series                         |   2 +
 ...R_NEXT_HOP-flag-handling-in-bgp_attr.patch | 149 ++++++++++++++++++
 ...v6-nexthops-when-importing-EVPN-IPv4.patch | 107 +++++++++++++
 4 files changed, 264 insertions(+)
 create mode 100644 debian/patches/upstream/0005-bgpd-fix-BGP_ATTR_NEXT_HOP-flag-handling-in-bgp_attr.patch
 create mode 100644 debian/patches/upstream/0006-bgpd-preserve-IPv6-nexthops-when-importing-EVPN-IPv4.patch


Summary over all repositories:
  4 files changed, 264 insertions(+), 0 deletions(-)

-- 
Generated by murpp 0.11.0