bug: PXE issue - Machine are stuck in boot loop if booted from onboard nic

[vinodchitraliNVIDIA]

Version

main

Describe the bug.

In GB200, if the host is booted from onboard nic, the secondary ipxe chain loader timeout and goes into boot loop forever.

see the logs

Minimum reproducible example

ZERO dpu

Relevant log output

>>Start HTTP Boot over IPv4.....         version 02.04.14
  Station IP address is 10.85.111.10      date 20251204

  URI: http://10.86.160.86:8080/public/blobs/internal/aarch64/ipxe.efi
  File Size: 379392 Bytes
  Downloading...100%TcpOnAppAbort: connection reset issued by application for TCB 7851D62198
Loading driver at 0x078536BC000 EntryPoint=0x078536C58B0
 Image - snponly.efi
EFI handle 0x785cd77298 systab 0x78629c0018
EFI does not provide protocol ConsoleControl
EFI protocol ef9fc172-a1b2-4693-b327-6d32fc416042 is at 0x786088e200
EFI protocol AcpiTable is at 0x785dab80a0
EFI protocol 8b843e20-8132-4852-90cc-551a4e4a7f1c is at 0x78607f3200
EFI protocol 3152bca5-eade-433d-862e-c01cdc291f44 is at 0x78607d81a8
EFI does not provide configuration table eb9d2d30-2d88-11d3-9a16-0090273fc14d
EFI configuration table f2fd1544-9794-4a2c-992e-e5bbcf20e394 is at 0x78622c0000
EFI configuration table eb9d2d31-2d88-11d3-9a16-0090273fc14d is at 0xffff0000
EFI image base address 0x78536bc000
EFI image device path VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x1412
1FFF)/PciRoot(0x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1)/IPv4(0.
0.0.0,TCP,DHCP,10.85.111.10,10.85.111.1,255.255.255.128)/Dns(10.86.160.88)/Uri(http://10.86.160.86:8
080/public/blobs/internal/aarch64/ipxe.efi)
carbide initialising devices...
EFISVC VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) created ManagedNetwork
Sb child UNKNOWN<0x7851d67098,ManagedNetwork>
autoexec.ipxe... Not found (https://ipxe.org/2d0c618e)
EFISVC VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) removing ManagedNetwor
kSb child UNKNOWN<0x7851d67098,ManagedNetwork>
remove-symbol-file /out/nvidia/bootloader/uefi/Server_RELEASE/Build/Server/RELEASE_GCC5/AARCH64/NetworkPkg/Dhcp6Dxe/Dhcp6Dxe/DEBUG/Dhcp6Dxe.dll 0x5FAF8000
EFIDRV connecting our drivers
EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) has driver "MNPONLY"
EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) disconnecting existing
 drivers

snp->undi.shutdown()  8000h:6h

snp->undi.stop()  8000h:6h
EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) connecting new drivers

EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) has driver "NIIONLY"
EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) DRIVER_START
EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) is not supported by dr
iver "MNPONLY": Error 0x7f5a4083 (https://ipxe.org/7f5a4083)
EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) using driver "NIIONLY"


snp->undi.start()  3h:8000h


Carbide
carbide v2025.12.05-rc1-0-22-gdebug -- Carbide Bare Metal provisioning -- http://nvidia.com
Features: DNS HTTP HTTPS iSCSI TFTP VLAN AoE EFI Menu
Forge - v2025.12.05-rc1-0-22-gdebug
DHCP 0x785371a608 entering discovery state
Configuring [dhcp] (net0 d0:f4:05:6c:c6:38)...DHCP 0x785371a608 DHCPDISCOVER
DHCP 0x785371a608 DHCPOFFER from 10.85.111.3:67 (10.42.0.237/10.42.0.237) for 10.85.111.10 proxy
DHCP 0x785371a608 entering request state
DHCP 0x785371a608 DHCPREQUEST to 10.42.0.237:67 for 10.85.111.10
DHCP 0x785371a608 DHCPREQUEST to 10.42.0.237:67 for 10.85.111.10
DHCP 0x785371a608 DHCPACK from 10.85.111.2:67 (10.42.0.237/10.42.0.237) for 10.85.111.10
DHCP 0x785371a608 entering ProxyDHCP state
DHCP 0x785371a608 ProxyDHCP REQUEST to 10.42.0.237
DHCP 0x785371a608 DHCPACK from 10.85.111.3:67 (10.42.0.237/10.42.0.237)
DHCP 0x785371a608 DHCPACK from 10.85.111.2:67 (10.42.0.237/10.42.0.237)
DHCP 0x785371a608 ProxyDHCP REQUEST to 10.42.0.237
.DHCP 0x785371a608 ProxyDHCP REQUEST to 10.42.0.237
.DHCP 0x785371a608 ProxyDHCP REQUEST to 10.42.0.237
.. ok
EFIDRV disconnecting our drivers                
EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) DRIVER_STOP child VenH
w(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0x5)/Pci(0x0
,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1)/MAC(D0F4056CC638,0x1)
EFIDRV VenHw(1E5A432C-0466-4D31-B009-D4D9239271D3)/MemoryMapped(0xB,0x14120000,0x14121FFF)/PciRoot(0
x5)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x6,0x0)/Pci(0x0,0x0)/MAC(D0F4056CC638,0x1) DRIVER_STOP
EFI image unloading
EFIDRV disconnecting our drivers
EFI image unloaded
EFIDRV reconnecting old drivers

Other/Misc.

No response

Code of Conduct

• I agree to follow NCX Infra Controller's Code of Conduct
• I have searched the open bugs and have found no duplicates for this bug report