Hello all.

got a few questions regarding win10, gpu passthru and the config of the vm, but first thing first.

my first attempts at passing thru my old gtx570 only worked once and second boot always resulted in a hardlock of host.

now i have replaced it with a gtx970 and things work much better.
my host is running fedora 22 so it only have qemu 2.3.1.
this means i have to use the old style libvirt settings of hiding kvm and so on to get the nvidia drivers working.

this works just fine, so my next step was to see if i can get a newer version of qemu and use the hv_vendor_id workaround instead. checked out the stable-2.5 branch from git and compiled it and setup the wrapper scrips as documented in previous mails on this list and then edited the libvirt xml file to point to the right wrapper script.

vm booted just fine but at this stage i still have the old hyperv and kvm settings turned off. since hv_vendor_id is supposed to make that not needed i tried to turn them back on again and as soon as i did win10 crashes at boot with system thread exception. it doesnt matter which one i turn back on again, just one is enough to cause a crash. i tried each one in order to see if i could find the one causing it, but any single option of those is enough to cause the crash.

i suspect this crash only happens when hv_vendor_id is specified.

any advice on how to get win10 to boot without system thread exception with hv_vendor_id specified and the hyperv and kvm options turned on?
i read something about x2apic but i don't fully understand what to do.

i also tried to change cpu model to core2duo with no success.
do we know what driver is causing the system thread exception?
anyone analyzed the crashdump via a debugger?

_______________________________________________
vfio-users mailing list
vfio-users@redhat.com
https://www.redhat.com/mailman/listinfo/vfio-users

Reply via email to