PSOD on the ESXi host when child port is created on ENS enabled host

Products

• VMware NSX Networking

Issue/Introduction

• ESXi host is prepared as a Host Transport Node in NSX.
• ESXi host is configured in ENS (enhanced data path) mode,
  or ESXi host is using a smart network card (SmartNIC).
• A child VLAN port is created for a VM connected to overlay segment.
• ESXi host PSODs once such VM is powered on.
• Backtrace (in vmkernel-zdump coredump file) may look similar to the example below:
  0x4538b301ba00:[0x420019fc0802][e-mail].l2#1.1.7.0.22667792+0x12 stack: 0x4538b301ba88, 0x420018a73067, 0x955650000000007a, 0x43038d002b80, 0x4000089
  0x4538b301bb10:[0x420019ff01ac][e-mail].l2#1.1.7.0.22667792+0x6e1 stack: 0x41ffd8c214a0, 0x20000000f, 0x400007c0400007b, 0x400007e0400007d, 0x40000800400007f
  0x4538b301bbf0:[0x420018a44243]NetEvent_PostEvent@vmkernel#nover+0x198 stack: 0x0, 0x0, 0x3, 0x4301ac60a630, 0x5a
  0x4538b301bc80:[0x420018a7bc66]Portset_PostEvent@vmkernel#nover+0x33 stack: 0x420018ebc87e, 0x4538b301bce4, 0x430052a98c20, 0x420018aa2410, 0x400007a
  0x4538b301bcb0:[0x420018aa240f]vmk_NetEventPortsetEventPost@vmkernel#nover+0x5c stack: 0x0, 0x420018b283e2, 0x0, 0x43038d0649c0, 0x500000000
  0x4538b301bcd0:[0x420018b283e1]EnsPostPortEvent@vmkernel#nover+0x3a stack: 0x500000000, 0x400007a00000004, 0x0, 0x0, 0x0
  0x4538b301bd10:[0x420018b2d1b0]Ens_PortEnable@vmkernel#nover+0x1b1 stack: 0x1b301005e, 0x42001a422aac, 0x420018b2d177, 0x0, 0x500000000
  0x4538b301bd70:[0x420018ae9b17]Vmxnet3VMKDev_Enable@vmkernel#nover+0x33bc stack: 0x4303e8002200, 0x0, 0x3f00000001, 0x42000400007a, 0x8000000002
  0x4538b301bee0:[0x420018cb4f6e]User_UWVMK64SyscallHandler@vmkernel#nover+0x183 stack: 0x1, 0x1, 0x420018954b64, 0xe5b93bf4c0, 0xe5ffd05360
  0x4538b301bf40:[0x42001894b518]SyscallUWVMK64@vmkernel#nover+0x90 stack: 0x0, 0x0, 0x1, 0x1, 0xe5b93bf4c0

Environment

VMware NSX 4.1.2.3 or lower. 
VMware NSX-T Data Center 3.2.3.2 or lower.

Cause

This issue is triggered when host is configured in ENS mode, or the host has a SmartNIC in it's hardware configuration, and a VLAN child ports are created on a VM connected to overlay segment.
Once a VM with VLAN child port(s) is powered on, ENS does the callback of a specific vdl2 function to apply the child port filter with parent port ID. Due to this issue, VLAN child port is not present in vdl2, and this will cause the host to panic and PSOD.

Resolution

This issue is fixed in:
NSX-T Data Center 3.2.4 and higher.
NSX 4.1.2.4 and higher.