[vfio-users] GPU passthrough to Windows problem: BOOT_DEVICE_INACCESSIBLE & general questions on passing physical disks

[Francesco Dicarlo]

I meant passing physical drive in that manner... Passing a partition or a lvm logical volume has the same result but it's cleaner and encouraged...

I can't seem to find the references right now. After dinner will post my xml :-) currently passing an lvm logical volume built on a 240 gb ssd without problem...

Sent from my iPhone

> On 30 Oct 2015, at 21:43, Benjamin Slade <slade at jnanam.net> wrote:
> 
> 
> Francesco Dicarlo <evilsephiroth at gmail.com> writes:
> 
>> Being on arch,I'll answer too...
>> Please use this guide on arch because it is really well made.
>> https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF
>> 
>> And for the sake of simplicity, ditch qemu scripts and use xml definitions for libvirt/virt-manager.
>> 
>> You won't have problem. My setup...
>> 
>> B85m-hd3 i5 4570 16gb ram
>> Gtx 970
>> 
>> Kernel Linux-Vfio-lts installed from aur. Edited pkgbuild to not patch i915. For ovmf it isn't needed...
>> Windows 8.1/osx Yosemite with gpu passthrough working flawlessly...
>> 
>> Passing through a physical drive isn't a good solution...
>> Better create a partition or better a logical volume on an lvm for the drive and feed it to kvm...
>> Remember to set type of storage raw and not qcow2...
>> 
>> If you want I can share my win8 xml and my qemu Yosemite qemu command line...
> 
> Why is passing a physical drive not a good solution?  Part of the reason
> I wanted to do it like that was to allow for the option of dual booting.
> 
> If you could share your win8 xml and qemu command line that would be
> helpful, so that I have a good model to work from (since the guide on
> reddit is based on the qemu script rather than an xml definition).