evpn vxlan fixes batch1 #20046

soumyar-roy · 2025-11-13T23:31:27Z

This is batch commit MR for multiple evpn vxlan related bugs found internally and fixed.

github-actions · 2025-11-13T23:31:39Z

This pull request has conflicts, please resolve those before we can evaluate the pull request.

ton31337 · 2025-11-14T07:22:31Z

zebra/zebra_rib.c

 					vtep_ip.ipaddr_v6 = tmp_nh->gate.ipv6;
 				}
-				zebra_rib_queue_evpn_route_del(tmp_nh->vrf_id, &vtep_ip, &ere->p);
+				/* zebra_rib_queue_evpn_route_del(re->vrf_id, &vtep_ip, &ere->p); */


Can we drop this call at all then?

ton31337 · 2025-11-14T07:23:47Z

zebra/zebra_vxlan.c

+ * given vxlan interface. For deletion of vni from the vni table please use
+ * zebra_vxlan_if_vni_del()
+ */
+static int zebra_vxlan_if_vni_remove(struct zebra_if *zif, vni_t vni)


If we don't care about the return code, let's use void().

ton31337 · 2025-11-14T07:44:35Z

zebra/zebra_evpn_mh.h

 	/* SVI associated with the VLAN */
 	struct zebra_if *vlan_zif;
+	/* VNI count */
+	uint8_t vni_count;


Maybe let's name it vni_refcnt? Because now I'm reading it as a VNI mapping, but it's a reference count actually?

soumyar-roy · 2025-11-17T17:25:30Z

ci:rerun

Set the vlan_id/bridge_if when the l3vni is configured. We were neglecting to set this info if the l3vni is configured after we read in the bridge/vlan info. This happens even at startup. Before: ``` alfred# show evpn vni 10 VNI: 10 Type: L3 Tenant VRF: vrf-red Vlan: 0 <============================ Bridge: - <=========================== Local Vtep Ip: 2.2.2.1 Vxlan-Intf: vxlan0 SVI-If: vlan10 State: Up VNI Filter: none System MAC: aa:bb:cc:00:12:11 Router MAC: aa:bb:cc:00:12:11 L2 VNIs: 20 ``` After: ``` alfred# show evpn vni 10 VNI: 10 Type: L3 Tenant VRF: vrf-red Vlan: 10 Bridge: bridge Local Vtep Ip: 2.2.2.1 Vxlan-Intf: vxlan0 SVI-If: vlan10 State: Up VNI Filter: none System MAC: aa:bb:cc:00:12:11 Router MAC: aa:bb:cc:00:12:11 L2 VNIs: 20 ``` Ticket: #3283598 Testing Done: Manual, evpn-min Signed-off-by: Stephen Worley <[email protected]>

Issue is: when new VNI is added, the access vlan - '1' acts as a placeholder till the sequence of updating the info for the new VNI is complete. In the update sequence the placeholder is dereferenced when the actual access-vlan for the VNI is updated. Hence when an actual vlan1 bridge exists this current code path assumes this is a placedolder and removes the vlan1-vni mapping. fix is to keep a ref count of the vni’s associated with the VLAN. So, when we reference and deference the VLAN->VNI mapping based on this ref count we could take the appropriate action. Ticket:#3102128 Testing Done: UT Signed-off-by: Ashwini Reddy <[email protected]>

Signed-off-by: Soumya Roy <[email protected]>

Add VNI's associated bridge and vlan info in json output format. torm-11# show evpn vni detail VNI: 1008 Type: L2 Vlan: 1008 Bridge: bridge ... Ticket:#3208813 Reviewed By: Testing Done: torm-11# show evpn vni detail json [ { "vni":1008, "type":"L2", "vlan":1008, <<< New field "bridge":"bridge", <<< New field "vrf":"vrf3", "vxlanInterface":"vxlan0", "ifindex":15, "vtepIp":"27.0.0.15", "mcastGroup":"239.1.1.108", "advertiseGatewayMacip":"No", "advertiseSviMacip":"No", "numMacs":18, "numArpNd":42, "numRemoteVteps":[ "27.0.0.18", "27.0.0.17", "27.0.0.16" ] }, ] Signed-off-by: Chirag Shah <[email protected]>

Add test to verify that L2 VNI JSON output includes "vlan" and "bridge" fields that were missing before. The test checks both PE1 and PE2 for VNI 101 to ensure the JSON output from 'show evpn vni <l2vni> json' contains: - "vlan" field with the VLAN ID - "bridge" field with the bridge interface name This validates the fix from commit 1a5ad2356b ("zebra: add vni bridge vlan info in json output"). Signed-off-by: Soumya Roy <[email protected]>

Add a debug for deleting the NH for a evpn-vrf route. Signed-off-by: Stephen Worley <[email protected]>

Below is output of the change: Ticket: #4074735 tor-12# show evpn vni VNI Type VxLAN IF # MACs # ARPs # Remote VTEPs Tenant VRF VLAN BRIDGE 10400 L2 vxlan48 5 11 3 default 400 br_default 10300 L2 vxlan48 1 3 3 default 300 br_default 10100 L2 vxlan48 13 27 3 default 100 br_default 10200 L2 vxlan48 1 3 3 default 200 br_default tor-12# show evpn vni json { "10400":{ "vni":10400, "type":"L2", "vxlanIf":"vxlan48", "numMacs":5, "numArpNd":11, "numRemoteVteps":3, "tenantVrf":"default", "vlan":400, "bridge":"br_default", "remoteVteps":[ "10.10.3.14", "10.10.3.13", "10.10.3.11" ] }, "10300":{ "vni":10300, "type":"L2", "vxlanIf":"vxlan48", "numMacs":1, "numArpNd":3, "numRemoteVteps":3, "tenantVrf":"default", "vlan":300, "bridge":"br_default", "remoteVteps":[ "10.10.3.14", "10.10.3.13", "10.10.3.11" ] }, tor-12# show evpn arp-cache vni 10400 ip 2011:11:44::124 json { "uptime":"00:00:06", "ip":"2011:11:44::124", "type":"local", "state":"active", "mac":"00:02:00:00:00:08", "vlan":400, "localSequence":0, "remoteSequence":0, "detectionCount":0, "isDuplicate":false } tor-12# show evpn arp-cache vni 10400 ip 2011:11:44::124 IP: 2011:11:44::124 Type: local State: active Uptime: 00:00:11 MAC: 00:02:00:00:00:08 VLAN: 400 Sync-info: - Flags: Router Local Seq: 0 Remote Seq: 0 Signed-off-by: Ashwini Reddy <[email protected]>

Verify 'show evpn vni' includes VLAN and BRIDGE columns. Tests commit b99ffd9. Signed-off-by: Soumya Roy <[email protected]>

Add test case to verify that L3 VNI displays correct VLAN and bridge information instead of showing "Vlan: 0" and "Bridge: -". The test adds L3 VNI 999 with VLAN 999 on bridge br999 to the existing bgp_evpn_vxlan_topo1 topology and verifies that "show evpn vni 999" displays: - "Vlan: 999" (not 0) - "Bridge: br999" (not "-") This validates the fix from commit 3ad2a78 ("zebra: set vlan_id/bridge_if on l3vni configure") which addresses the issue where L3 VNI configuration was not populating these fields when the VNI is configured after bridge/VLAN interfaces are read. Signed-off-by: Soumya Roy <[email protected]>

We are missing some fpm hooks for rmac installation. Add them. Signed-off-by: Stephen Worley <[email protected]>

Let's assume the DUT receives sticky mac of a remote host, but it already learnt the same mac as local sticky mac. Currently, Zebra reprograms the local sticky to remote sticky MAC. When this happens, we don't log any message about the mac movement. So adding a log. UT after the fix: Leaf1 ------- Leaf2(DUT) To recreate this scenario, we need to configure route-map filter on Leaf1 to deny MACIP route coming from Leaf2. So that Leaf1 can generate MACIP route for the same sticky MAC and the DUT(already has local sticky MAC) can receive and process the remote sticky MAC. 2024-10-16T07:45:13.107733+00:00 leaf2 zebra[450183]: [V6SBB-9YZQG] Received remote mac add for MAC 00:0b:11:bb:00:01 VNI 10 VTEP 1.1.1.1, but it is already learnt as local sticky MAC on intf swp3 Ticket: #3803354 Signed-off-by: Krishnasamy R <[email protected]>

Rename vni_count to vni_refcnt to reflect what it is being used for Signed-off-by: Soumya Roy <[email protected]>

ton31337 · 2025-11-18T07:05:55Z

Could you fix Verify Source check before merging?

soumyar-roy · 2025-11-18T07:12:05Z

Could you fix Verify Source check before merging?

"
18-Nov-2025 05:05:19 | < WARNING: quoted string split across lines
18-Nov-2025 05:05:19 | < #2185: FILE: /tmp/f1-3206634/zebra_evpn_mac.c:2185:
"
It has a fail for this, for multiline string split, but isn't splitting here better than having them together ?>>if (IS_ZEBRA_DEBUG_VXLAN)
zlog_debug(" Add/Update %sMAC %pEA intf %s(%u) VID %u -> VNI %u%s, "
"entry exists and has not changed ",
sticky ? "sticky " : "",
macaddr, ifp->name,
ifp->ifindex, vid, zevpn->vni,
local_inactive
? " local_inactive"
: "");

Signed-off-by: Soumya Roy <[email protected]>

soumyar-roy · 2025-11-18T22:02:15Z

Could you fix Verify Source check before merging?

Fixed it

frrbot bot added bugfix tests Topotests, make check, etc zebra labels Nov 13, 2025

github-actions bot added master size/L conflicts labels Nov 13, 2025

soumyar-roy marked this pull request as draft November 13, 2025 23:31

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch from 9d4818f to 1bd1f0b Compare November 13, 2025 23:52

github-actions bot removed the conflicts label Nov 13, 2025

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch from 1bd1f0b to fc3ac12 Compare November 14, 2025 00:01

github-actions bot added size/S size/M and removed size/L size/S labels Nov 14, 2025

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch from fc9bb2d to 5afcfe7 Compare November 14, 2025 00:15

github-actions bot added size/L and removed size/M labels Nov 14, 2025

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch 9 times, most recently from b72e8f3 to 059e501 Compare November 14, 2025 07:14

ton31337 reviewed Nov 14, 2025

View reviewed changes

soumyar-roy marked this pull request as ready for review November 14, 2025 18:42

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch from 3861f2d to 34940b0 Compare November 17, 2025 21:11

sworleys and others added 8 commits November 17, 2025 15:09

tests: add test for access VLAN VNI count feature

fafb244

Signed-off-by: Soumya Roy <[email protected]>

zebra: add debug for evpn vrf NH deletion

efd774f

Add a debug for deleting the NH for a evpn-vrf route. Signed-off-by: Stephen Worley <[email protected]>

tests: add test for VNI summary VLAN/BRIDGE fields

9b3398b

Verify 'show evpn vni' includes VLAN and BRIDGE columns. Tests commit b99ffd9. Signed-off-by: Soumya Roy <[email protected]>

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch 2 times, most recently from 969609a to 3d16e6d Compare November 18, 2025 05:02

soumyar-roy and others added 4 commits November 17, 2025 21:03

zebra: add missing rmac FPM hooks

6d402f7

We are missing some fpm hooks for rmac installation. Add them. Signed-off-by: Stephen Worley <[email protected]>

zebra: Rename vni_count to vni_refcnt

5f23716

Rename vni_count to vni_refcnt to reflect what it is being used for Signed-off-by: Soumya Roy <[email protected]>

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch from 3d16e6d to ff4308a Compare November 18, 2025 05:04

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch from ff4308a to 913c753 Compare November 18, 2025 19:55

zebra: Fix build warning zlog_warn and string split

c336a6b

Signed-off-by: Soumya Roy <[email protected]>

soumyar-roy force-pushed the soumya/cl_to_frr_batch1 branch from 913c753 to c336a6b Compare November 18, 2025 19:58

ton31337 approved these changes Nov 20, 2025

View reviewed changes

ton31337 merged commit e46d1fb into FRRouting:master Nov 20, 2025
17 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

evpn vxlan fixes batch1 #20046

evpn vxlan fixes batch1 #20046

Uh oh!

soumyar-roy commented Nov 13, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Nov 13, 2025

Uh oh!

ton31337 Nov 14, 2025

Uh oh!

soumyar-roy Nov 14, 2025

Uh oh!

ton31337 Nov 14, 2025

Uh oh!

soumyar-roy Nov 14, 2025

Uh oh!

ton31337 Nov 14, 2025

Uh oh!

soumyar-roy Nov 14, 2025

Uh oh!

soumyar-roy commented Nov 17, 2025

Uh oh!

ton31337 commented Nov 18, 2025

Uh oh!

soumyar-roy commented Nov 18, 2025 •

edited

Loading

Uh oh!

soumyar-roy commented Nov 18, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

evpn vxlan fixes batch1 #20046

evpn vxlan fixes batch1 #20046

Uh oh!

Conversation

soumyar-roy commented Nov 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Nov 13, 2025

Uh oh!

ton31337 Nov 14, 2025

Choose a reason for hiding this comment

Uh oh!

soumyar-roy Nov 14, 2025

Choose a reason for hiding this comment

Uh oh!

ton31337 Nov 14, 2025

Choose a reason for hiding this comment

Uh oh!

soumyar-roy Nov 14, 2025

Choose a reason for hiding this comment

Uh oh!

ton31337 Nov 14, 2025

Choose a reason for hiding this comment

Uh oh!

soumyar-roy Nov 14, 2025

Choose a reason for hiding this comment

Uh oh!

soumyar-roy commented Nov 17, 2025

Uh oh!

ton31337 commented Nov 18, 2025

Uh oh!

soumyar-roy commented Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

soumyar-roy commented Nov 18, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

soumyar-roy commented Nov 13, 2025 •

edited

Loading

soumyar-roy commented Nov 18, 2025 •

edited

Loading