Success stories with GVT-g or primary GPU passtrough?

[tilllt]

Ok, i got my Gigabyte Aero 15 to replace my ancient Macbook(s) Pro. Since i still need to finish up some old FCP7 Projects, i need OSX running with Graphics acceleration. Unfortunately the Aero 15 - although it features a nice GTX1060, is still using Optimus grahics switching. So i am waging the options, as mentioned in my other thread here: http://www.insanelymac.com/forum/topic/329903-osx-in-qemu-kvm-kholia-tutorial-no-network/

 

So i followed the Kholia Guide and i got Sierra running. I have the virtio-net virtual Network Card running, next thing is graphics.

 

I am waging my options between:

1. GVT-g shared Intel Graphics with the internal Intel 630 GPU.

2. Passthrough of the primary Intel GPU, running on a headless Lubuntu Hypervisor.

 

As you can imagine, both options have Pros and Cons. I'd love to hear a success story for either one, since both options are not really well documented and i am not sure if i am knowledgeable enough to function as a guinea pig for this setup process. The main doubts about Option 1 is that i am not sure if it is usable already or rather a proof-of-concept still. In any case it would mean to use VNC as a protocol for viewing the Guest output, since not even SPiCE is implemented yet...

 

Some relevant Infos:

https://github.com/kholia/OSX-KVM

https://01.org/igvt-g/blogs/wangbo85/2017/gvt-g-upstream-status-update-were-transition-phase

https://github.com/intel/gvt-linux/wiki/GVTg_Setup_Guide

https://www.reddit.com/r/archlinux/comments/5wy8x4/has_anyone_passed_through_intel_hd_to_guest_vms/

https://www.kraxel.org/blog/2017/01/virtual-gpu-support-landing-upstream/

[nvflash]

I'm not sure I can help with your problem, I did something simpler with my setup. My motherboard will only output video at post to the first PCI-E slot that has a graphics card in it, and that creates trouble if I want to use that card for PCI Passthough.

 

I had to add:

video=efifb:off

 To the /etc/default/grub file, looks like this:

GRUB_CMDLINE_LINUX_DEFAULT="quiet amd_iommu=on vfio_iommu_type1.allow_unsafe_interrupts=1 video=efifb:off"

Shuts down video output to the post video device when grub loads the initfs.

 

Works for me, but I'm not sure that's what you are looking for.

 

I've never really looked into Optimus, I mean, I know basically what it is, but I'm not sure if each GPU can be used to drive it's own display, or if anyone ever tried it.

 

If you could, I would try and drive an exteral dispaly with the GTX 1060. I assume, at post the Intel 630 is driving the LVDS dispaly. If you blacklist the kernel modules for the nVidia card, you maybe able it pass it via PCI Passthough, tho I'm not sure how to rout it to an external display.

 

Anyway, could you tell me how you setup virtio networking?

 

How did you created the bridge, what does your qemu-ifup script look like, and what command are you using for the network device for qemu?

 

I tried:

-netdev type=tap,script=/etc/qemu-ifup,id=net0 -device virtio-net-pci,netdev=net0

 But I get errors, please, if you have time, give me a step by step of how you got it working on the host side.