test(machine-a-tron): guard the empty-NDF-MAC Dell boot NIC through HostInit

Cargo.toml

@@ -26,7 +26,7 @@ authors = ["NVIDIA Carbide Engineering <carbide-dev@exchange.nvidia.com>"]
 
 [workspace.dependencies]
 clap = { version = "4", features = ["derive", "env"] }
-libredfish = { git = "https://github.com/NVIDIA/libredfish.git", tag = "v0.44.15" }
+libredfish = { git = "https://github.com/NVIDIA/libredfish.git", tag = "v0.44.16" }
 librms = { git = "https://github.com/NVIDIA/nv-rms-client.git", tag = "v0.9.0-rc1" }
 ansi-to-html = "0.2.2"
 

crates/api-integration-tests/tests/lib.rs

@@ -193,6 +193,13 @@ async fn test_integration() -> eyre::Result<()> {
             Ipv4Addr::new(172, 20, 0, 2),
         )
         .boxed(),
+        test_machine_a_tron_empty_ndf_mac(
+            &test_env,
+            &bmc_address_registry,
+            // Relay IP in host-inband net (zero-DPU host)
+            Ipv4Addr::new(10, 10, 11, 2),
+        )
+        .boxed(),
     ]);
 
     tokio::select! {
@@ -334,6 +341,7 @@ async fn test_metrics_integration() -> eyre::Result<()> {
         1,
         1,
         false,
+        false, // empty_ndf_mac
         &test_env,
         &bmc_address_registry,
         Ipv4Addr::new(172, 20, 0, 1),
@@ -481,6 +489,7 @@ async fn test_machine_a_tron_multidpu(
         1,
         2,
         false,
+        false, // empty_ndf_mac
         test_env,
         bmc_mock_registry,
         admin_dhcp_relay_address,
@@ -563,6 +572,7 @@ async fn test_machine_a_tron_zerodpu(
         1,
         0,
         false,
+        false, // empty_ndf_mac
         test_env,
         bmc_mock_registry,
         admin_dhcp_relay_address,
@@ -612,6 +622,36 @@ async fn test_machine_a_tron_zerodpu(
     .await
 }
 
+/// A Dell host whose boot NIC reports an empty `NetworkDeviceFunction` MAC while
+/// its `EthernetInterface` MAC stays populated -- the NicMode-stripped / iDRAC
+/// partition case. The boot interface still resolves as a `Pair` (MAC + the
+/// interface id captured at exploration), so `PollingBiosSetup`'s `is_bios_setup`
+/// MAC lookup fails and falls back to the interface id. Reaching `Ready` proves
+/// the host no longer hangs in `HostInit` / `PollingBiosSetup`.
+async fn test_machine_a_tron_empty_ndf_mac(
+    test_env: &IntegrationTestEnvironment,
+    bmc_mock_registry: &BmcMockRegistry,
+    admin_dhcp_relay_address: Ipv4Addr,
+) -> eyre::Result<()> {
+    run_machine_a_tron_test(
+        HostHardwareType::DellPowerEdgeR750,
+        1,
+        0,
+        false,
+        true, // empty_ndf_mac
+        test_env,
+        bmc_mock_registry,
+        admin_dhcp_relay_address,
+        |machine_handle| async move {
+            machine_handle
+                .wait_until_machine_up_with_api_state("Ready", Duration::from_secs(90))
+                .await?;
+            Ok::<(), eyre::Report>(())
+        },
+    )
+    .await
+}
+
 async fn test_machine_a_tron_singledpu_nic_mode(
     hw_type: HostHardwareType,
     test_env: &IntegrationTestEnvironment,
@@ -623,6 +663,7 @@ async fn test_machine_a_tron_singledpu_nic_mode(
         1,
         1,
         true,
+        false, // empty_ndf_mac
         test_env,
         bmc_mock_registry,
         admin_dhcp_relay_address,
@@ -686,6 +727,7 @@ async fn test_machine_a_tron_dual_stack(
         1,
         1,
         false,
+        false, // empty_ndf_mac
         test_env,
         bmc_mock_registry,
         admin_dhcp_relay_address,
@@ -793,6 +835,7 @@ async fn test_machine_a_tron_dual_stack_l2(
         1,
         1,
         false,
+        false, // empty_ndf_mac
         test_env,
         bmc_mock_registry,
         admin_dhcp_relay_address,
@@ -853,6 +896,7 @@ async fn run_machine_a_tron_test<F, O>(
     host_count: u32,
     dpu_per_host_count: u32,
     dpus_in_nic_mode: bool,
+    empty_ndf_mac: bool,
     test_env: &IntegrationTestEnvironment,
     bmc_mock_registry: &BmcMockRegistry,
     admin_dhcp_relay_address: Ipv4Addr,
@@ -896,6 +940,7 @@ where
                 scout_run_interval: Duration::from_secs(1),
                 network_virtualization_type: None,
                 dpus_in_nic_mode,
+                empty_ndf_mac,
                 dpu_firmware_versions: None,
                 dpu_agent_version: None,
             }),

crates/bmc-mock/src/hw/dell_poweredge_r750.rs

@@ -31,6 +31,10 @@ pub struct DellPowerEdgeR750<'a> {
     pub product_serial_number: Cow<'a, str>,
     pub nics: Vec<(hw::nic::SlotNumber, hw::nic::Nic<'a>)>,
     pub embedded_nic: EmbeddedNic,
+    /// Serve an empty `NetworkDeviceFunction` MAC for the boot NIC (its
+    /// `EthernetInterface` MAC stays populated), reproducing a NicMode-stripped
+    /// iDRAC partition where the boot NIC must be resolved by interface id.
+    pub empty_ndf_mac: bool,
 }
 
 pub struct EmbeddedNic {
@@ -194,8 +198,17 @@ impl DellPowerEdgeR750<'_> {
                 &network_adapter_id,
                 &function_id,
             );
+            // The NetworkDeviceFunction MAC normally mirrors the NIC's MAC, but a
+            // NicMode-stripped boot NIC (or an iDRAC partition) can leave it empty
+            // while the EthernetInterface still carries the MAC -- serve that
+            // asymmetry on request so the boot NIC resolves by interface id.
+            let ndf_mac_address = if self.empty_ndf_mac {
+                String::new()
+            } else {
+                nic.mac_address.to_string()
+            };
             let function = redfish::network_device_function::builder(func_resource)
-                .ethernet(json!({"MACAddress": &nic.mac_address}))
+                .ethernet(json!({"MACAddress": ndf_mac_address}))
                 .oem(redfish::oem::dell::network_device_function::dell_nic_info(
                     &function_id,
                     *slot,

crates/bmc-mock/src/machine_info.rs

@@ -46,6 +46,9 @@ pub struct HostMachineInfo {
     pub nvos_mac_addresses: Vec<MacAddress>,
     pub switch_serial_number: Option<String>,
     pub hw_mac_addr_pool: MacAddressPoolConfig,
+    /// Serve an empty boot-NIC `NetworkDeviceFunction` MAC (Dell hosts only),
+    /// exercising boot-NIC resolution by interface id. Defaults to false.
+    pub empty_ndf_mac: bool,
 }
 
 #[derive(Debug, Clone)]
@@ -177,6 +180,7 @@ impl HostMachineInfo {
             switch_serial_number,
             dpus,
             hw_mac_addr_pool,
+            empty_ndf_mac: false,
         }
     }
 
@@ -393,6 +397,7 @@ impl HostMachineInfo {
                 port_1: next_mac(),
                 port_2: next_mac(),
             },
+            empty_ndf_mac: self.empty_ndf_mac,
         }
     }
 

crates/machine-a-tron/src/config.rs

@@ -123,6 +123,13 @@ pub struct MachineConfig {
 
     #[serde(default)]
     pub dpu_agent_version: Option<String>,
+
+    /// If true, this host's boot-NIC `NetworkDeviceFunction` reports an empty MAC
+    /// while its `EthernetInterface` MAC stays populated -- a Dell-only quirk for
+    /// exercising boot-NIC resolution by interface id. Honored only by hosts whose
+    /// hardware models a `NetworkDeviceFunction` (currently `DellPowerEdgeR750`).
+    #[serde(default)]
+    pub empty_ndf_mac: bool,
 }
 
 #[derive(Debug, Default, Serialize, Deserialize, Clone, Eq, PartialEq)]

crates/machine-a-tron/src/host_machine.rs

@@ -106,6 +106,7 @@ impl HostMachine {
             nvos_mac_addresses: persisted_host_machine.nvos_mac_addresses.clone(),
             switch_serial_number: persisted_host_machine.switch_serial_number.clone(),
             hw_mac_addr_pool,
+            empty_ndf_mac: config.empty_ndf_mac,
         };
         let dpus = dpu_machines
             .into_iter()
@@ -182,12 +183,13 @@ impl HostMachine {
                 )
             })
             .collect::<Vec<_>>();
-        let host_info = HostMachineInfo::new(
+        let mut host_info = HostMachineInfo::new(
             config.hw_type,
             dpu_machines.iter().map(|d| d.dpu_info().clone()).collect(),
             mac_pool,
             hw_pool_config,
         );
+        host_info.empty_ndf_mac = config.empty_ndf_mac;
         let dpus = dpu_machines
             .into_iter()
             .map(|d| d.start(true))