How FLR has been implemented inside the Corvin's code for the nVidia GPU.

Hello Corvin and everyone.

I'm continuing the development of the initial project of Abhinav Chavali,
the VMM Accelerator support for QEMU :

https://summerofcode.withgoogle.com/archive/2025/projects/lRkVElCJ
It seems that it works fine,but I would like to know.

How was FLR implemented in the code Corvin wrote to pass the NVIDIA GPU to
a Linux-based VM? I ask because when I use the Corvin's patched bhyve, or
perhaps his vmm, I don't see any system freezes when I load and unload the
vmm module from memory. It used to happen at the beginning,I remember well,
but later not anymore. However, when I do it with qemu accelerated with
bhyve, the problem still occurs. I asked Claude to look into Corvin's code
to understand how he handled it, but even after doing so, the problem
persists.

-- 
Mario.