RE: Booting Issue

[Connor Davis <davisc@xxxxxxxxxxxx>]

> -----Original Message-----
> From: Paul Durrant <xadimgnik@xxxxxxxxx>
> Sent: Thursday, December 17, 2020 12:15 PM
> To: Connor Davis <davisc@xxxxxxxxxxxx>; win-pv-devel@xxxxxxxxxxxxxxxxxxxx
> Subject: RE: Booting Issue
> 
> Notice: This message originated outside of ainfosec.com
> 
> 
> 
> > -----Original Message-----
> > From: win-pv-devel <win-pv-devel-bounces@xxxxxxxxxxxxxxxxxxxx> On Behalf Of 
> > Connor Davis
> > Sent: 17 December 2020 18:51
> > To: win-pv-devel@xxxxxxxxxxxxxxxxxxxx
> > Subject: Booting Issue
> >
> > Hi All,
> >
> > What is the expected behavior wrt Windows booting up if the Xen platform 
> > PCI device isn't present
> > after a previous run where the PV drivers loaded and ran correctly with the 
> > platform device present?
> > On my system it is currently preventing Windows from booting.
> 
> Hi Connor,
> 
>   What is *supposed* to happen is that the XENFILT driver detects the absence 
> of the PCI to which XENBUS is bound and prevents
> emulated devices from being unplugged, which should allow the boot to 
> continue.
> 
> >
> > Another possibly related issue is occasionally an update will cause Windows 
> > to complain that
> > a boot critical file is missing, usually pointing to xenbus.sys. Is there a 
> > way to tell Windows that
> > the PV drivers are not boot critical so that Windows will continue booting 
> > even if it doesn't load
> > the PV drivers properly?
> >
> 
>   Not seen that before... perhaps it's a new 'feature' of Windows. XENBUS 
> would be considered boot critical as it is the parent of
> XENVBD which, on the previous boot, was presumably hosting the system disk. 
> This will need some investigation unfortunately. What
> version of Windows are you running? 10, Server 2019?

I'm currently testing Windows 10 1909. One major caveat here: we are running 
Windows on top of MicroV, which implements the
subset of the Xen interface to boot and run the PV drivers (sidenote: we hope 
to start open sourcing and upstreaming this work next year).
The only devices that are not given to Windows are network devices; those are 
passed through to Linux service VMs. So Windows has
complete control over the disk. In fact, xenvbd.sys is never installed - we are 
only using xenbus, xeniface, xenvif, and xennet.

> 
>   Cheers,
> 
>     Paul

Thanks,
Connor