Regression with PXE boot on OvmfPkg/VirtioNetDxe driver on aarch64 system

[aaron.young@oracle.com]

   We have encountered what we believe to be a regression in the 
OvmfPkg/VirtioNetDxe driver.

   This regression causes PXE boot to fail with the following ASSERT:

ASSERT 
[VirtioNetDxe]/builddir/build/BUILD/ovmf-92d07e48907f/OvmfPkg/VirtioNetDxe/SnpSharedHelpers.c(184): 
((BOOLEAN)(0==1))

   The failure was encountered on a aarch64/Armv8 system (Ampere server) 
using the latest mainline (edk2.git/master) code as of 8-28-2018.

   On a hunch, I reverted the following patchset which resulted in PXE 
boot being successful - which suggests this patchset is likely the culprit:

https://lists.01.org/pipermail/edk2-devel/2017-September/014679.html

   The command used to initiate the VM PXE boot/install is:

   virt-install --name guest1  --vcpus 16 --ram 32000 --arch aarch64 
--disk guest1.img,size=100 --network 
type=direct,source=enP4p1s0,model=virtio --pxe

   NOTE that I also tried simply removing the offending ASSERT() and PXE 
boot still fails with multiple errors (exceptions of some sort that I 
didn't log).


   What I'd like to know:

   1. Is this a known failure?
   2. Was aarch64 given much testing with this new patchset?
   3. Is reverting this patchset a safe workaround for now until a fix 
is found? Or is there another simple workaround perhaps?

   4. Any hunches what the issue could be?


   Thanks for any info/help/suggestions,

   -Aaron Young