[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openbsd-bugs
Subject:    Re: VM guest restarting
From:       Mike Larkin <mlarkin () azathoth ! net>
Date:       2017-06-20 5:34:01
Message-ID: 20170620053401.GD63267 () azathoth ! net
[Download RAW message or body]

On Thu, Jun 08, 2017 at 06:33:35AM -0700, Mike Larkin wrote:
> On Fri, Jun 02, 2017 at 08:38:14AM -0400, Bryan Steele wrote:
> > On Fri, Jun 02, 2017 at 05:56:32AM -0500, Michael Graves wrote:
> > > On 2017-06-01 23:37, Mike Larkin wrote:
> > > > 
> > > > This is the second report I've seen of the FX4100 crashing. I'll have to
> > > > add
> > > > some diagnostic code to see why it fails since I don't have access to
> > > > this
> > > > hardware.
> > > > 
> > > > One thing I see that's odd is you are using the old style of booting.
> > > > Can you
> > > > make sure you got the firmware via fw_update after the new snap?
> > > > 
> > > > -ml
> > > 
> > > I ran fw_update without any packages being updated.  Here is the output of
> > > 'fw_update -i'  and a md5 sum of the vmm* files if that helps.
> > > 
> > > # fw_update -i
> > > Installed: radeondrm-firmware-20150927 vmm-firmware-1.10.2p3
> > > 
> > > # md5 /etc/firmware/vmm-bios*
> > > MD5 (/etc/firmware/vmm-bios) = 2946eaf040aa4fd31c13edd442b36665
> > > MD5 (/etc/firmware/vmm-bios-license) = 4bb44edd8d967f805ae4df6386c59aaa
> > > 
> > >
> > 
> > I'm the other FX-4100 owner, I think Mike wanted you to test if BIOS
> > boots work for you, for me I get SeaBIOS and boot(8) out, but crashes
> > and restarts the VM as soon as it loads the kernel.
> > 
> > For BIOS boots, you simply don't need to pass a kernel image via
> > -b or boot, just a disk image with a valid boot sector.
> > 
> > -Bryan.
> > 
> 
> So Bryan and I have made a little progress here, we've made it past this error
> by disabling NX in the guest (which makes no sense since the CPU supports it
> and the feature is enabled properly). However, it hangs later with an interrupt
> issue.
> 
> Still looking at it. Looks like this cpu is the problem child model of AMD.
> 
> -ml
>

I just committed part of the fix. While this does not yet make this family of
CPUs usable for vmm, we are getting closer.

-ml 

[prev in list] [next in list] [prev in thread] [next in thread]