Bug:when dhcp relay downlink interface is configured with vrf dhcp relay fails

[lucas-li-micas]

Is it platform specific

generic

Importance or Severity

Medium

Description of the bug

server-------(eth49)relay(eth50)-------client

When isc-dhcp downlink interface(eth50) is configured with vrf1 and uplink interface(eth49) is in default vrf, dhcp v4 relay fails, see topoloy shown above.

Steps to Reproduce

1.add vlan100 and vlan member
2.bind vlan100 to vrf1
3.configure vlan100 ip and dhcp relay helper
4.start dhcp v4 request from client

Actual Behavior and Expected Behavior

Actual Behaviour:

The dhcp relay request packet is sent to server but offer packet from server is sent but dropped by uplink interface.
If we bind uplink interface to vrf1 as downlink interface does, dhcp request packet is proposed by relay device but not sent to dhcp server, resulting dhcp relay failure.

Expected Behaviour:

If uplink and downlink interfaces are in the same vrf1, dhcp relay will succeed.

Relevant log output

2025 Jun 20 03:38:01.614413 sonic ALERT dhcp_relay#dhcpmon[116]: dhcpmon detected disparity in DHCP Relay behavior. Duration: 63648 (sec) for vlan: 'Agg-Vlan100'
2025 Jun 20 03:38:01.614597 sonic NOTICE dhcp_relay#dhcpmon[116]: :- publish: EVENT_PUBLISHED: {"sonic-events-dhcp-relay:dhcp-relay-disparity":{"duration":"63648","timestamp":"2025-06-20T03:38:01.614413Z","vlan":"Agg-Vlan100"}}
2025 Jun 20 03:38:01.614881 sonic NOTICE dhcp_relay#dhcpmon[116]: [ Agg-Vlan100-Snapshot rx/tx] Discover: 358/ 208, Offer: 0/ 0, Request: 0/ 0, ACK: 0/ 0
2025 Jun 20 03:38:01.614967 sonic NOTICE dhcp_relay#dhcpmon[116]: [ Agg-Vlan100- Current rx/tx] Discover: 363/ 208, Offer: 0/ 0, Request: 0/ 0, ACK: 0/ 0

Output of show version, show techsupport

SONiC Software Version: SONiC.202505.0-dirty-20250605.054827
SONiC OS Version: 12
Distribution: Debian 12.11
Kernel: 6.1.0-29-2-amd64
Build commit: 0a0488097
Build date: Thu Jun  5 10:51:24 UTC 2025
Built by: micasrd@micasrd-PowerEdge-R730

Platform: x86_64-micas_m2-w6510-48gt4v-r0
HwSKU: M2-W6510-48GT4V
ASIC: broadcom
ASIC Count: 1
Serial Number: 1234567890123
Model Number: 01018913
Hardware Revision: 100
Uptime: 08:16:21 up  2:25,  3 users,  load average: 2.51, 1.48, 1.11
Date: Thu 19 Jun 2025 08:16:21

Attach files (if any)

No response

[bingwang-ms]

The support of vrf in dhcp-relay is not verified. So probably it's a limitation in dhcp-relay.
@yaqiangz Can you help take a look and confirm?

[yaqiangz]

I don't think current dhcp_relay support vrf

[AnantKishorSharma]

Hi @lucas-li-micas @bingwang-ms @yaqiangz — can this be closed as resolved by the native v4 relay agent (opt-in via DEVICE_METADATA['localhost']['has_sonic_dhcpv4_relay'] = 'True')? It's VRF-aware: it binds the server-facing socket per-VRF via SO_BINDTODEVICE and supports a server VRF distinct from the client VLAN's VRF. There was a bug in that path, but PR sonic-net/sonic-dhcp-relay#104 fixed it — moving a VLAN to a non-default VRF used to leave the upstream socket bound to the old VRF (OFFER/ACK never returned). This capability is absent from the legacy ISC dhcrelay path this issue was filed against, and I'm not aware of any plan to add it there.