(Not convinced that this is a XFCE issue, so move this if necessary)
I recently tried to get GPU pass-through to work, and this time I was successful. Using a GTX 1070 for my main EOS system and GTX 1050 Ti for my VMs, this setting is working great …until I try to return from suspend. Here is a snippet of the errors I’m getting, the first one appears multiple times:
kernel: NVRM: Xid (PCI:0000:08:00): 31, pid=129983, name=xfwm4, Ch 00000010, intr 10000000. MMU Fault: ENGINE HOST0 HUBCLIENT_HOST faulted @ 0x0_040a0000. Fault is of type FAULT_PDE ACCESS_TYPE_READ
systemd-coredump[187203]: [🡕] Process 129983 (xfwm4) of user 1000 dumped core.
I’m sure this is due to my setup. I have the GTX 1050 Ti blacklisted, but it’s connected to the 2nd monitor via HDMI. When I’m only using EOS, I use the (2) DP connections from the 1070. When I use a VM, I switch to the HDMI connection. Yes, all (3) cables are always connected.
The PCI address listed in the NVRM error is the 1070. I would think that a 3rd monitor would solve this issue, but due to lack of desk space it’s just not a practical solution. I am also testing these (3) Nvidia services that are normally disabled (nvidia-persistenced.service was already enabled):
nvidia-hibernate.service
nvidia-resume.service
nvidia-suspend.service