vfio problems

vfio problems

[Dominic Eschweiler]

Hi,

I just recently started to play with vfio, since the new Kernel 3.6
comes directly with an integrated vfio-stack. My problem currently is,
that I'm not able to bind the vfio-pci to the (unused) smbus controller
in my laptop. 

Here are the steps I did and the related results:


# modprobe vfio-pci

[ 1609.065705] VFIO - User Level meta-driver version: 0.3

# lspci

...
00:1f.3 SMBus: Intel Corporation 6 Series/C200 Series Chipset Family
SMBus Controller (rev 05)
...

# lspci -n -s 0000:00:1f.3

00:1f.3 0c05: 8086:1c22 (rev 05)

# echo 8086 1c22 > /sys/bus/pci/drivers/vfio-pci/new_id

[ 4485.759148] vfio-pci: probe of 0000:00:1f.3 failed with error -22


I omitted the unbind step, which is described in the documentation,
since the device is not claimed by any driver at all. Also the "error
-22" statement isn't really helping in this case. 

Any ideas?

-- 
Gruß
  Dominic

Frankfurt Institute for Advanced Studies (FIAS)
Ruth-Moufang-Straße 1
D-60438 Frankfurt am Main
Germany

Phone:  +49 69 79844114

Re: vfio problems

[Alex Williamson]

On Thu, 2012-10-04 at 11:40 +0200, Dominic Eschweiler wrote:
> Hi,
> 
> I just recently started to play with vfio, since the new Kernel 3.6
> comes directly with an integrated vfio-stack. My problem currently is,
> that I'm not able to bind the vfio-pci to the (unused) smbus controller
> in my laptop. 
> 
> Here are the steps I did and the related results:
> 
> 
> # modprobe vfio-pci
> 
> [ 1609.065705] VFIO - User Level meta-driver version: 0.3
> 
> # lspci
> 
> ...
> 00:1f.3 SMBus: Intel Corporation 6 Series/C200 Series Chipset Family
> SMBus Controller (rev 05)
> ...
> 
> # lspci -n -s 0000:00:1f.3
> 
> 00:1f.3 0c05: 8086:1c22 (rev 05)
> 
> # echo 8086 1c22 > /sys/bus/pci/drivers/vfio-pci/new_id
> 
> [ 4485.759148] vfio-pci: probe of 0000:00:1f.3 failed with error -22
> 
> 
> I omitted the unbind step, which is described in the documentation,
> since the device is not claimed by any driver at all. Also the "error
> -22" statement isn't really helping in this case. 
> 
> Any ideas?
> 

What does this report?

readlink /sys/bus/pci/devices/0000\:00\:1f.3/iommu_group

The likely cause of an EINVAL for an endpoint device is that it's not
part of an IOMMU group, which may mean you don't have an IOMMU enabled.
You can also look in /sys/kernel/iommu_groups to see the groups.
Thanks,

Alex