Re: [Xen-devel] SR-IOV VF Passthrough Issues

[Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>]

On Thu, Aug 18, 2011 at 04:15:02AM -0500, Kaushik Kumar Ram wrote:
> I am trying to passthrough a SR-IOV VF from an Intel 82599EB 10 GbE NIC to a 
> PV Linux guest (v3.0.0-rc7). 
> I am running a pvops dom0 Linux v2.6.32.44 (xen/stable-2.6.32.x branch). 
> 
> I followed the instructions on 
> http://wiki.xensource.com/xenwiki/XenPCIpassthrough. But it just did not 
> work. I got the following error when I tried loading the ixgbevf driver in 
> the guest:
> 
> ixgbevf 0000:02:10.0: device not available (can't reserve [mem 
> 0x00000000-0x00003fff 64bit])
> ixgbevf: probe of 0000:02:10.0 failed with error -22
> 
> I am not well versed on how PCI devices are setup. But I added some debug 
> statements and figured out 
> that the device's memory regions were not initialized. I also found that 
> passing the 'pci=resource_alignment=' 

Uh, in the dom0? So the lspci showed that region uninitialized?

> argument to dom0 during boot was causing the memory regions to be released 
> prior to starting the 
> guests. So I just tried removed this particular argument and it worked!  Now 
> I am able to successfully 
> passthrough the VF to the guest.

> 
> So my question is whether the 'pci=resource_alignment=' argument really 
> needed?

It is only needed if the memory region is not page aligned. I wish you included
the full output of your guest to get an idea of what the BARs and the E820 
looked
like.

> 
> Also, when I start the guest I get the following libxl error:
> 
> libxl: error: libxl_pci.c:749:libxl__device_pci_reset: The kernel doesn't 
> support reset from sysfs for PCI device 0000:02:10.0
> 
> It seems harmless. But could this be related to the error above?

Nope.
> 
> Please let me know if you need any other information.
> 
> Thanks.
> 
> --Kaushik

> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel