Linux Mint - Configured for one GPU with passthrough, libvirtd.service malfunctioning
Linux Mint - Configured for one GPU with passthrough, libvirtd.service malfunctioning
Hello, I'm attempting to configure a single GPU passthrough on my AMD PC using an Nvidia RTX 3070. I've followed the instructions from the YT video thoroughly, double-checking each step. Despite careful review, after setting up the passthrough, I still can't boot into my VM through Virtual Machine Manager. Once I restart, it shows "QEMU/KVM - Connecting..." but when I try to launch a new machine, I receive an error saying "no active connection to install on." After some research, I discovered that "libvirtd.service" seems to be behaving oddly—even with sudo/root access, I can't restart or stop it. By examining the logs, I noticed "dnsmasq" and "libvirtd" are preventing a proper reboot, but they appear to work without issues when I manually reboot my PC. However, after rebooting, I can't restart the VM either. Manually booting from terminal with "sudo virsh start win11" just displays a terminal window with "thinking." It's frustrating since I'm a Windows admin and feel quite lost here. Could you help me figure out where I went wrong? Thanks in advance.
Log entries for libvirtd show a failure in loading the virtualization daemon. The service was enabled but crashed shortly after startup, with repeated restart attempts and system alerts about unit processes not stopping properly. System logs indicate ongoing issues with dnsmasq and dnsmasq units remaining active despite attempts to terminate them.
I don't have direct access to system details or GPU information. Please run a compatibility check or use your device's settings to identify the active driver.
Sun Jun 23 05:57:57 2024 +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.183.01 | Driver Version: 535.183.01 | CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3070 Off | 00000000:09:00.0 On | N/A | | 0% 47C P8 24W / 220W | 1605MiB / 8192MiB | 30% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+
You should first remove the NVIDIA drivers, then attempt a passthrough again. The LSPCI command displays the currently installed drivers. It lists three: one open-source default, the NVIDIA driver you’re using, and a third passthrough driver whose name you can’t recall. Using LSPCI, expand it with `lspci --help` to identify the active driver. Once identified, blacklist its name in the drivers file so it won’t be bound to your system anymore.
This seems to raise some concerns. Removing the video driver kernel module could leave users without hardware acceleration even when the virtual machine isn't active. This undermines the goal, as the host Linux would become ineffective without it, making running Linux impractical. The overall approach doesn't align with the instructions, which emphasize enabling drivers before setting up the VM with GPU passthrough. The intended outcome should involve installing Linux with hardware acceleration, using a VM with GPU passthrough, and restoring full acceleration when the VM is shut down. I haven't tested it myself, but it appears to be an interesting solution. I joined the discussion to check if others have found success with this method.
Only if his CPU includes graphics drivers. That wasn't the request OP made. I managed to set up GPU passthrough using the Intel CPU as the main processor and the GPU as a pass-through. The setup involved a screen with two HDMI inputs. But it seems that wasn't what OP intended, so I'd suggest he consider giving it up or owning two GPUs for a smoother experience. GPU passthrough can be tricky, but my approach should work. If it does, you'll definitely see the results in the guide. Fun fact—my system runs both operating systems, one with GPU passthrough and another focused on gaming. Ez switcheroo!
Bruh, the CPU can indeed support graphics like Intel CPUs do in mini PCs. I’ve got an Intel CPU on a tower and am pretty sure some AMD models have graphics too. What exactly are you referring to? Linux can run headless with GPU forwarding (I haven’t tried it before), and you can also use the CPU for Linux while enabling GPU pass-through for virtual machines. That’s my go-to approach. Both methods technically involve single GPU forwarding since the CPU isn’t labeled as a GPU. I made an earlier mistake by confusing the terms, thinking about headless operation but focusing on GPU forwarding instead. It’s still experimental, and I’m just trying to assist.