[TUTORIAL] virtiofsd in PVE 8.0.x

Very green to proxmox and all of this really, but was able to get this working with this very helpful guide! One problem I'm having is that if my VM shuts down it seems the virtiofs daemon is still running. Even if I kill the process it still wont work again until I reboot the whole server. Anyone know what is going on?

Failed to start transient service unit: Unit virtiofs-daemon-200.service was already loaded or has a fragment file.
hookscript error for 200 on pre-start: command '/var/lib/vz/snippets/200.hook.sh 200 pre-start' failed: exit code 1
 
How do I get this working with iommu for GPU passthrough? I'm running PVE 8.3.2 if that matters/this guide is outdated... but if not, I get this:
Code:
 virtiofsd[12061]: Waiting for daemon failed: HandleRequest(InvalidParam)

When I research that it's apparently a bug in an older version of the virtiofsd but I tried version v1.10.1 (default and v1.13.0 (latest release I built)

I still get that log, any one have any ideas? I really need the access to it directly with GPU for serving media but this is way out of my level I think.
 
For whoever end up here, having issues. The tutorial is great, the hookscript that is suggested to change the args of the qemu before executing the VM doesn't work for me unless the following patch is applied. I'm on PVE 8.3.2.
 
How do I get this working with iommu for GPU passthrough? I'm running PVE 8.3.2 if that matters/this guide is outdated... but if not, I get this:
Code:
 virtiofsd[12061]: Waiting for daemon failed: HandleRequest(InvalidParam)

When I research that it's apparently a bug in an older version of the virtiofsd but I tried version v1.10.1 (default and v1.13.0 (latest release I built)

I still get that log, any one have any ideas? I really need the access to it directly with GPU for serving amedia but this is way out of my level I think.
Same problem here, virtiofsd's sandboxed child process exits for me as soon as the guest VM tries to initialize _something_. Unclear what triggers it precisely and I can find what the child is logging anywhere.

The good news, however, is that enabling the NUMA args in the options list in the Qemu config file seems to prevent this. Specifically, using cleptes's scripts seems to work with NUMA on for me in an Arch VM with GPU passthrough.

Hi, I created a Bash hookscript based on this forum thread, extended it with configs and what not. Hope it helps anyone :) Any suggestions and what not are always welcome on project page: https://github.com/JSubelj/vfs-pve-hook.sh
Thank you, this seems like a much cleaner option till we get built-in support.