Re: [Xen-devel] Nested Virtualisation and UEFI

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 04.09.15 at 10:47, <james-xen@xxxxxxxxxxxxxx> wrote:
> I am trying to get a nested virtualisation environment working where it is 
> started with bios='ovmf'.  When my 
> hvm configuration is bios='seabios' and I use the xen.gz everything starts 
> correctly.  Using the xen.efi from 
> the same build when bios='ovmf' causesthe guest Linux kernel to block when 
> accessing the emulated disk 
> controller (scsi or ide) as below.
> 
> Is this nested configuration expected to work?

I don't think there's any expectation either way. You may well be the
first one to every try. I wouldn't even dare to guess whether the
issue is in the BIOS, Xen, qemu, or any combination thereof.

>  If so, are there other 
> things that need to change apart from 
> switching the bios configuration value?  I can provide more information from 
> the guest XEN / kernel logs or the 
> guest configuration file if that would be useful.
> 
> [   31.848214] ata1: lost interrupt (Status 0x50)

This might hint at a problem with interrupts. A full, maximally verbose
log (hypervisor and kernel) may tell.

> [   31.855823] ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 
> frozen
> [   31.866516] ata1.00: failed command: READ DMA
> [   31.873472] ata1.00: cmd c8/00:08:00:00:00/00:00:00:00:00/e0 tag 0 dma 
> 4096 in
> [   31.873472]          res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 
> (timeout)
> [   31.896759] ata1.00: status: { DRDY }
> [   31.902526] ata1: soft resetting link
> [   32.076309] ata1.00: configured for MWDMA2
> [   32.083559] ata1.00: device reported invalid CHS sector 0
> [   32.094014] ata1: EH complete

Whereas this might also hint at the BIOS reporting an IOMMU when
there is none, or some other issue with setting up DMA. Again, full
logs may provide some insight.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel