FW V4.90(ABPL.1)C0 seems to break IPV6 traffic on GS1350-26HP

I have applied FW V4.90(ABPL.1)C0 to GS1350-26HP as I needed to have port information exposed in SNMP. This FW does indeed brings this feature, but I am encountering a serious issue which forced me to rollback to previous FW version.

As soon as I got the FW updated to V4.90(ABPL.1)C0 users complained about connection issues on Nebula APs connected downstream of the switch.

Upon further investigation I have the following conclusions:

1/ The switch is Nebula managed, but because I suspected the switch was the source of my connectivity issues, I have defined interfaces for all VLANs exposed by the switch using WebGUI. Note that the VLAN interfaces were already defined and working in Nebula, but were not defined in the WebGUI of the new FW when I first connected.

2/ I have activated and setup IPV6 for every such interface.

3/ I can ping the gateway in IPV6 from the switch itself reliably and without issue which makes me believe that I have proper IPV6 settings.

4/ Mobile devices connected to AP downstream of the switch do get IPV6 addresses from the router with RA mechanism (managed mode SLAAC and DHCPV6 addresses) which means that broadcast requests pass through the switch.

5/ Mobile devices connected to AP downstream of the switch cannot reliably ping the gateway in IPV6. When doing specific ping queries in IPV6 to the gateway, about one request out of 20 gets a reply, with random TTL sometimes 10ms sometimes 2000+ms. I then get 3 to 5 replies then no replies for another 20 request and so on. As by default Android and iOS choose to use IPV6 connectivty by default when available, the mobiles devices get disconnected when trying to access IPV6 services (streaming services for instance).

6/ If I disable the RA mechanism forcing mobile device to only use IPV4, connectivity is working fine to outside services and all IPV4 services are accessible.

7/ Mobile devices connected to AP which is NOT downstream of the switch get IPV6 address with the same mechanism. They have no issue pinging the gateway in IPV6.

8/ Wired devices such as PC and servers connected downstream of the switch cannot ping the gateway in IPV6 neither, but can reliably connect in IPV4.

9/ I have verified IPV6 configuration of the switch itself and everything appears good and as stated above. The switch itself does not suffer connectivity issues when pinging the gateway in IPV6.

10/ I have rolled back to V4.70(ABPL.5) | 10/25/2022 and I immediately get all the above issues resolved.

So the overall symptoms are that IPV6 unicast packets do not get pass the switch itself.

My conclusion is that either V4.90(ABPL.1)C0 breaks connectivity in IPV6 completely, or it encounters an issue with previous settings from the previous FW version. I have not tried to reset switch settings completely as it would require a complete reconfiguration of all ports and VLAN mappings.

Let me know if I can provide you with configuration or any other data.