route-map encapsulation gretap #19786

pguibert6WIND · 2025-10-17T15:26:27Z

Add ability to encapsulation L3VPN traffic directly into GRETAP interface, by using route-map.

pguibert6WIND · 2025-10-20T07:12:27Z

ci:rerun

donaldsharp · 2025-10-20T13:33:55Z

zebra/interface.c

 	} else if (IS_ZEBRA_IF_VXLAN(ifp)) {
 		zebra_vxlan_if_dump_vty(vty, zebra_if);
-	} else if (IS_ZEBRA_IF_GRE(ifp)) {
+	} else if (IS_ZEBRA_IF_GRE(ifp) || IS_ZEBRA_IF_GRETAP(ifp)) {


sure looks like you are missing this code in the json portion ( and this is EXACTLY why I didn't want a json specific function.

you are right. I completely forgot the json part. Adding the support for it.

donaldsharp · 2025-10-20T13:34:35Z

zebra/zebra_nb_state.c

 	case ZEBRA_IF_GRE:
 		type = "frr-zebra:zif-gre";
 		break;
+	case ZEBRA_IF_GRETAP:


Why no changes in zebra/zapi_msg.c. Shouldn't we just change the IS_ZEBRA_IF_GRE to encompass a GRETAP interface?

There is a change in zebra/zapi_msg.c... or I did not understand the question.

--- a/zebra/zapi_msg.c +++ b/zebra/zapi_msg.c @@ -77,6 +77,7 @@ static void zserv_encode_interface(struct stream *s, struct interface *ifp) stream_putl(s, ifp->mtu6); stream_putl(s, ifp->bandwidth); stream_putl(s, zif->link_ifindex); + stream_putl(s, ifp->zif_type);

I agree that GRE and GRETAP interfaces are similar.
But what about users that already have GRETAP interfaces used today for an other purpose?

there are tests of if (IS_ZEBRA_GRE) in zapi_msg.c that you did not include the new test for to test if it is IS_ZEBRA_GRETAP`

ha ok. yes, I need to clarify. today, only NHRP uses it, so NHRP is only GRE. But we have to take care of remaining code. Later GRE_SOURCE_SET and GRE_GET zapi messages should cover other kind of tunnels too.
So I added a comment.

donaldsharp · 2025-10-20T13:36:37Z

bgpd/bgp_routemap.c


 /* Set nexthop to object */
 struct rmap_l3vpn_nexthop_encapsulation_set {
-	uint8_t protocol;


Why are we completely dropping the protocol?

About protocol, I need to think..

Link layer protocol is not precise enough to know I am over a GRETAP (LLT ETHERNET) or a GRE interface (LLT_IPGRE).

Since it was storing protocol and allowing matches against it, who have you broken with this change?

The IP protocol of the 'rmap_l3vpn_nexthop_encapsulation_set'
structure is replaced by the zif_type, as interfaces used as backbone
for mpls ca either be ptp L3 GRE interfaces or ptp L2 GRE (GRETAP)
interfaces.

I was considering IP6GRE and GRE equal for BGP.
I will use the same mechanism for IP6GRETAP and GRETAP interfaces.
So one new keywords: gretap

donaldsharp · 2025-10-20T13:38:24Z

a) Documentation update
b) topotest to show this works.

pguibert6WIND · 2025-10-20T19:13:36Z

ci:rerun

yang/frr-bgp-route-map.yang

pguibert6WIND · 2025-10-21T12:15:21Z

ci:rerun

pguibert6WIND · 2025-10-21T14:32:55Z

ci:rerun

pguibert6WIND · 2025-10-22T08:13:06Z

@vayetze , there is a topotest added.

Add gretap interface information. > Interface l2gre0 is up, line protocol is up > Link ups: 0 last: (never) > Link downs: 0 last: (never) > vrf: default > index 8 metric 0 mtu 1462 speed 0 txqlen 1000 > flags: <UP,LOWER_UP,BROADCAST,RUNNING,MULTICAST> > MPLS Not specified by CLI > Multicast config is Not specified by CLI > Shutdown config is Not specified by CLI > Type: Ethernet > HWaddr: 2e:dd:45:c2:12:44 > inet6 fe80::2cdd:45ff:fec2:1244/64 > Interface Type GRETAP > Interface Slave Type None > VTEP IP: 192.168.0.10 , remote 192.168.0.103 > protodown: off Some ZAPI code is GRE-specific for now, as only NHRP uses GRE information. Signed-off-by: Philippe Guibert <[email protected]>

The interface kind is needed by some daemons in order to apply some filtering rules. The struct interface is shared between daemons via the ZAPI socket. Signed-off-by: Philippe Guibert <[email protected]>

Add new defines to detect ip6gre and ip6gretap interfaces. Signed-off-by: Philippe Guibert <[email protected]>

Use the zebra interface type for the bgp route-map encapsulation gre command. The IP protocol of the 'rmap_l3vpn_nexthop_encapsulation_set' structure is replaced by the zif_type, as interfaces used as backbone for mpls ca either be ptp L3 GRE interfaces or ptp L2 GRE (GRETAP) interfaces. Signed-off-by: Philippe Guibert <[email protected]>

Add support for gretap encapsulation type. Signed-off-by: Philippe Guibert <[email protected]>

The bgp_vpnv4_gre test validates imported routes as ok when route-map is detached. Whereas, The nexthop of MPLS paths must either be a labelled nexthop or point to a GRE interface (by using route-map). The case where this check does not matter is when local importation is performed by using the 'import vrf' command. Fix this by applying the automatic validation only when this command is used to import VPN paths. Enforce the current topotest by really checking the invalidity of the path. Apply black style to modified topotest. Fixes: 93fd9cb ("bgpd: Validate imported routes next-hop that is in a default VRF") Signed-off-by: Philippe Guibert <[email protected]>

The previous commit broke the bgp_vpn_import_nexthop_default_vrf test. This test reproduced issue FRRouting#12502. By default, BGP imported MPLS paths can not be selected, while the nexthop is not located under an MPLS backbone, except for some exceptions. One of them is that the path is learned by an eBGP peer located in a connected network segment. This is what is described in [0], section b). For that, the per-interface 'mpls bgp forwarding' command aims at helping validating the incoming BGP path. Add config command for test 'vpnv4_import_allowas_in' and for 'vpnv4_import_nexthop_default_vrf' that have been implemented on the same model. [0] https://datatracker.ietf.org/doc/html/rfc4364#section-10 Link: FRRouting@34f7a57 Link: FRRouting#12502 Signed-off-by: Philippe Guibert <[email protected]>

Add a test that demonstrates that an MPLS path can be installed over a GRETAP interface, if a route-map autorises it. Signed-off-by: Philippe Guibert <[email protected]>

The previous commit broke some IGP tests that were not using MPLS paths. Move the test to be an eBGP peering, and use the interAS option that allows bgp paths to be imported. Signed-off-by: Philippe Guibert <[email protected]>

Like it has been done on previous commits, use the inter-AS option B mechanism to autorise ebgp connection to convey directly MPLS paths without any MPLS backbone support. Signed-off-by: Philippe Guibert <[email protected]>

Signed-off-by: Philippe Guibert <[email protected]>

pguibert6WIND · 2025-10-29T17:07:29Z

@ton31337 @donaldsharp , any updates?

frrbot bot added bgp yang zebra labels Oct 17, 2025

github-actions bot added master size/M labels Oct 17, 2025

donaldsharp reviewed Oct 20, 2025

View reviewed changes

pguibert6WIND force-pushed the rmap_encapsulation_gretap branch from 1e3f3a9 to 1b0243e Compare October 20, 2025 14:00

frrbot bot added bugfix documentation labels Oct 20, 2025

github-actions bot added size/L and removed size/M labels Oct 20, 2025

ton31337 reviewed Oct 21, 2025

View reviewed changes

yang/frr-bgp-route-map.yang Outdated Show resolved Hide resolved

pguibert6WIND force-pushed the rmap_encapsulation_gretap branch 2 times, most recently from 0505ea9 to a326503 Compare October 21, 2025 09:02

pguibert6WIND force-pushed the rmap_encapsulation_gretap branch from a326503 to 2962f58 Compare October 21, 2025 12:44

vayetze added the Need TopoTest label Oct 21, 2025

pguibert6WIND force-pushed the rmap_encapsulation_gretap branch 4 times, most recently from ba6605d to 6b82fe6 Compare October 24, 2025 15:25

frrbot bot added the libfrr label Oct 24, 2025

pguibert6WIND removed the Need TopoTest label Oct 27, 2025

pguibert6WIND force-pushed the rmap_encapsulation_gretap branch 3 times, most recently from 82def60 to 22bb12c Compare October 27, 2025 16:44

pguibert6WIND added 11 commits October 28, 2025 11:04

zebra, link: zif_type attribute to struct interface

2acc407

The interface kind is needed by some daemons in order to apply some filtering rules. The struct interface is shared between daemons via the ZAPI socket. Signed-off-by: Philippe Guibert <[email protected]>

zebra, lib: add support for ip6gre and ip6gretap interface

2444744

Add new defines to detect ip6gre and ip6gretap interfaces. Signed-off-by: Philippe Guibert <[email protected]>

bgpd, yang, doc: add gretap encapsulation type in route-map

8b85936

Add support for gretap encapsulation type. Signed-off-by: Philippe Guibert <[email protected]>

topotests: bgp_vpnv4_gre, add l3vpn route-map test for gretap

c75050f

Add a test that demonstrates that an MPLS path can be installed over a GRETAP interface, if a route-map autorises it. Signed-off-by: Philippe Guibert <[email protected]>

topotests: bgp_srv6l3vpn_sid, fix add 'mpls bgp forwarding'

1c4829e

Like it has been done on previous commits, use the inter-AS option B mechanism to autorise ebgp connection to convey directly MPLS paths without any MPLS backbone support. Signed-off-by: Philippe Guibert <[email protected]>

topotests: bgp_vrf_route_leak_basic, add r2 peering

5dc35cb

Signed-off-by: Philippe Guibert <[email protected]>

pguibert6WIND force-pushed the rmap_encapsulation_gretap branch from 22bb12c to 5dc35cb Compare October 28, 2025 10:05

ton31337 approved these changes Nov 5, 2025

View reviewed changes

route-map encapsulation gretap #19786

Are you sure you want to change the base?

route-map encapsulation gretap #19786

Conversation

pguibert6WIND commented Oct 17, 2025

Uh oh!

pguibert6WIND commented Oct 20, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

donaldsharp commented Oct 20, 2025

Uh oh!

pguibert6WIND commented Oct 20, 2025

Uh oh!

Uh oh!

pguibert6WIND commented Oct 21, 2025

Uh oh!

pguibert6WIND commented Oct 21, 2025

Uh oh!

pguibert6WIND commented Oct 22, 2025

Uh oh!

pguibert6WIND commented Oct 29, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants