Bug: [vpp] VLAN router interface (SVI) attribute update fails with SAI_STATUS_FAILURE, wedging orchagent on dual-ToR

[lolyu]

Is it platform specific

generic (VPP platform / docker-syncd-vpp, libsaivs)

Importance or Severity

High

Description of the bug

On the VPP platform, updating a SAI_ROUTER_INTERFACE of type VLAN (i.e. the Vlan<x> SVI, realized in VPP as a BVI) always fails with SAI_STATUS_FAILURE.

SwitchVpp::vpp_update_router_interface() (vslib vslib/vpp/SwitchVppRif.cpp) reads SAI_ROUTER_INTERFACE_ATTR_PORT_ID before it branches on the RIF type:

attr.id = SAI_ROUTER_INTERFACE_ATTR_PORT_ID;
status = get(SAI_OBJECT_TYPE_ROUTER_INTERFACE, object_id, 1, &attr);
if (status != SAI_STATUS_SUCCESS) {
    SWSS_LOG_ERROR("attr SAI_ROUTER_INTERFACE_ATTR_PORT_ID was not passed");
    return SAI_STATUS_FAILURE;   // <-- VLAN RIF has no PORT_ID => always here
}

A VLAN router interface has no SAI_ROUTER_INTERFACE_ATTR_PORT_ID (it has SAI_ROUTER_INTERFACE_ATTR_VLAN_ID instead), so the get() fails and the whole set() returns SAI_STATUS_FAILURE. The create and remove paths already branch on TYPE == VLAN first; only the update path does not.

The failure is reliably triggered on dual-ToR, where IntfsOrch issues a SAI_ROUTER_INTERFACE_ATTR_SRC_MAC_ADDRESS update on the Vlan1000 RIF (shared gateway MAC reconcile). orchagent turns the sticky SAI failure into a hard error, re-logs it continuously, and asks syncd to dump:

ERR    swss#orchagent: :- doTask: Failed to set router interface mac 22:ad:a2:34:08:76 for port Vlan1000, rv:-1
NOTICE swss#orchagent: :- notifySyncd: sending syncd: SYNCD_INVOKE_DUMP

orchagent then stays unhealthy (the RIF SET remains in m_toSync and retries forever), which stalls further programming on that Executor.

Steps to Reproduce

1. Deploy a dual-ToR VPP KVM testbed (vms-kvm-vpp-dual-tor / t0-dualtor), platform vpp, docker-syncd-vpp.
2. Bring up the Vlan1000 SVI with a router interface (VLAN-type RIF) — normal dual-ToR minigraph does this.
3. Let IntfsOrch issue a SAI_ROUTER_INTERFACE_ATTR_SRC_MAC_ADDRESS (or MTU) update on the VLAN RIF (happens automatically during mux/gateway-MAC reconcile; can also be forced with config interface ... that re-derives the SVI MAC).
4. Observe orchagent logging Failed to set router interface mac ... for port Vlan1000, rv:-1 and issuing SYNCD_INVOKE_DUMP.

Actual Behavior and Expected Behavior

Actual: Any attribute update to a VLAN (SVI/BVI) router interface returns SAI_STATUS_FAILURE because vpp_update_router_interface() requires PORT_ID. On dual-ToR this wedges orchagent with a sticky, endlessly-retried SAI failure and repeated SYNCD_INVOKE_DUMP.

Expected: Updating a VLAN router interface should succeed. The update should be dispatched by RIF type (as create/remove already are) and applied to the BVI (bvi<vlan_id>), mirroring vpp_create_bvi_interface(). Supported attributes (SRC_MAC, MTU) should be applied to the BVI and persisted to the SAI object store; unsupported attributes should not fail the call.

Relevant log output

# orchagent (host syslog) — repeats until the RIF SET is satisfied
ERR    swss#orchagent: :- doTask: Failed to set router interface mac 22:ad:a2:34:08:76 for port Vlan1000, rv:-1
NOTICE swss#orchagent: :- notifySyncd: sending syncd: SYNCD_INVOKE_DUMP

# /var/log/swss/sairedis.rec — VLAN RIF is created with TYPE_VLAN + VLAN_ID (no PORT_ID) ...
c|SAI_OBJECT_TYPE_ROUTER_INTERFACE:oid:0x60000000005d7|SAI_ROUTER_INTERFACE_ATTR_VIRTUAL_ROUTER_ID=oid:0x3000000000002|SAI_ROUTER_INTERFACE_ATTR_SRC_MAC_ADDRESS=00:AA:BB:CC:DD:EE|SAI_ROUTER_INTERFACE_ATTR_TYPE=SAI_ROUTER_INTERFACE_TYPE_VLAN|SAI_ROUTER_INTERFACE_ATTR_VLAN_ID=oid:0x260000000005bc|SAI_ROUTER_INTERFACE_ATTR_MTU=9100|SAI_ROUTER_INTERFACE_ATTR_NAT_ZONE_ID=0

# ... then a later SRC_MAC 'set' on that same VLAN RIF is what fails in vpp_update_router_interface():
s|SAI_OBJECT_TYPE_ROUTER_INTERFACE:oid:0x60000000005d7|SAI_ROUTER_INTERFACE_ATTR_SRC_MAC_ADDRESS=22:AD:A2:34:08:76

Output of show version, show techsupport

SONiC Software Version: SONiC.master.1182394-fc9ea17bb
Distribution: Debian 13.5
Kernel: 6.12.41+deb13-sonic-amd64
Build commit: fc9ea17bb
Platform: x86_64-kvm_x86_64-r0
HwSKU: Force10-S6000
ASIC: vpp
ASIC Count: 1

Fix

Dispatch the update by RIF type before the PORT_ID lookup and handle the VLAN case on the BVI:

• sonic-sairedis: [vpp] Handle VLAN router interface attribute updates on the BVI sonic-sairedis#2022 — [vpp] Handle VLAN router interface attribute updates on the BVI

Related (dual-ToR trigger / root cause of the spurious SRC_MAC reconcile):

• sonic-swss: [dualtor][intfsorch] Preserve port MAC when updating router interface MAC sonic-swss#4816 — [dualtor][intfsorch] Preserve port MAC when updating router interface MAC