[vfio-users] nVidia Code 43

rndbit rndbit at sysret.net
Mon Dec 14 13:04:41 UTC 2015


I updated my Win8.1 VM xml with changes i did not have from Hristo's
win10.xml. This is what i have discovered:

* 359.06 driver installer will not install driver claiming it can not
find nvidia GPU
* Installing driver from C:\NVIDIA via device manager's "update driver"
function works
* On boot nvidia card has code 43
* Disabling and reenabling card reports no error code, says device is
working!

However GPUZ does not display half of information and 3d applications
still run on emulated GPU. That is probably because i tried this on
laptop with 960m.

Need emulated GPU so i can enter pre-boot password. Besides this
configuration worked fine with AMD card, windows would switch from
emulated one to passed-through during boot. Not with nvidia card. Oh well.

On 2015.12.11 19:22, Alan Jenkins wrote:
> Hey Philip,
>
> Yeah my 2 halfs of the card each have their own IOMMU groups:
>
> /sys/kernel/iommu_groups/33/devices
> /sys/kernel/iommu_groups/33/devices/0000:06:00.0
> /sys/kernel/iommu_groups/33/devices/0000:06:00.1
> /sys/kernel/iommu_groups/34
> /sys/kernel/iommu_groups/34/devices
> /sys/kernel/iommu_groups/34/devices/0000:07:00.0
> /sys/kernel/iommu_groups/34/devices/0000:07:00.1
>
> However if you are struggling to get a 660 working I am worried about
> my chances with my weirder dual GPU card. Have you tried both
> Tianocore and Seabios or just Tianocore as could be that your 660 is
> not compatible with UEFI booting (I had to reflash my 690's with the
> firmware of another manufacturer's 690 just to obtain that ability). 
>
> I just finished some testing with remapping to different PCI slots
> inside the VM. Turns out that I cannot fully recreate the mappings the
> physical machine has inside the virtual machine as there is a
> restriction on the slot you can assign that states that it has to be
> >= 1 whereas on my physical machine they are in slot 0 so rather than
> being:
>
> 0000:06:00.0
> 0000:06:00.1
> 0000:07:00.0
> 0000:07:00.1
>
> the closest I can make it is:
>
> 0000:06:01.0
> 0000:06:01.1
> 0000:07:01.0
> 0000:07:01.1
>
> Not sure if this is what could be upsetting it or not but it is
> annoying that I cannot replicate the real world inside the VM. That
> being said the Arch VM I setup was able to run Unigine Heaven fine on
> the 690 but it doesn't seem to want to enable MultiGPU mode meaning it
> only uses one of the 2 GPU's resulting in less than 50% of the
> performance of running it on the host OS.
>
> Copying those mappings over to the Windows VM resulted in Windows
> stating initially that the cards were working fine but needed to
> reinstall their drivers. After the reinstall though I ended up with
> the same Code 43 error.
>
> Would anyone mind sharing a working nVidia based XML?
>
> Thanks,
>
> Alan Jenkins
>
> On Fri, 11 Dec 2015 at 16:30 Philip Abernethy <chais.z3r0 at gmail.com
> <mailto:chais.z3r0 at gmail.com>> wrote:
>
>     On Fri, 11 Dec 2015 13:26 Alan Jenkins
>     <alan.james.jenkins at gmail.com
>     <mailto:alan.james.jenkins at gmail.com>> wrote:
>
>     Has anyone else had success with one?
>
>     Sadly all I can say currently is, that I have the exact same
>     problem with my GTX 660. Despite everything I've tried and what
>     was suggested by the people here, I can't get that thing to work,
>     while older and newer cards work just fine.
>
>     The only thing I haven't tried yet is using a different slot for
>     the card. Does your card have a IOMMU group on its own? Because
>     mine shares one with the PCI controller.
>
>     Philip Abernethy
>
>
>
> _______________________________________________
> vfio-users mailing list
> vfio-users at redhat.com
> https://www.redhat.com/mailman/listinfo/vfio-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/vfio-users/attachments/20151214/c11111c4/attachment.htm>


More information about the vfio-users mailing list