Re: [PATCH 00/17] [RFC] OvmfPkg: Add support for microvm machine type

["Yao, Jiewen" <jiewen.yao@intel.com>]

Thank you. That confirms my guess.

If you can add all those info in readme, that would be great, especially the URL https://qemu.readthedocs.io/en/latest/system/i386/microvm.html, as well as the scope.

I have not reviewed all detail, and I think it is good idea.
Acked-by: Jiewen Yao <Jiewen.yao@intel.com>


> -----Original Message-----
> From: Gerd Hoffmann <kraxel@redhat.com>
> Sent: Tuesday, August 31, 2021 7:27 PM
> To: Yao, Jiewen <jiewen.yao@intel.com>
> Cc: devel@edk2.groups.io; Marc-André Lureau
> <marcandre.lureau@redhat.com>; Anthony Perard
> <anthony.perard@citrix.com>; Ard Biesheuvel <ardb+tianocore@kernel.org>;
> Justen, Jordan L <jordan.l.justen@intel.com>; Julien Grall <julien@xen.org>;
> Stefan Berger <stefanb@linux.ibm.com>
> Subject: Re: [PATCH 00/17] [RFC] OvmfPkg: Add support for microvm machine
> type
> 
> On Tue, Aug 31, 2021 at 10:31:31AM +0000, Yao, Jiewen wrote:
> > Hi Gerd
> > I would like to confirm a couple of things:
> >
> > -1) Since this is a big new feature, please follow process to submit Bugzilla.
> >
> > 0) I hope we can include a readme file to introduce more background and the
> scope of the new project.
> 
> Sure, will do both.
> 
> > 1) What is the device list of microvm?
> 
> It is a minimal virtual machine designed for modern (aka virtio-aware) guests.
> Most legacy isa devices (pic, pit, rtc) are optional and can be turned
> off.  storage/network/etc. uses virtio-mmio.
> 
> Optional pcie support is available and any pcie device supported by qemu
> cane plugged in (including virtio-pci if you prefer that over
> virtio-mmio).
> 
> > Or do you mean
> https://qemu.readthedocs.io/en/latest/system/i386/microvm.html?
> 
> Yes, it's this.
> 
> > 2) How many QEMU configurations do we want to support?
> 
> This series adds just the basics.  I want add support for virtio-mmio
> and pcie later on, once the basics are sorted and merged.
> 
> > 3) What is the reason to add a new dsc/fdf instead of add a MACRO/PCD in
> OVMF?
> 
> microvm borrowed a bunch of ideas from arm virt.  So it doesn't work
> like a classic pc chipset in many areas.  There is no pci config space
> access via ioport 0xcf8 for example.  Also power management works
> completely different (see patch #14).
> 
> Adapting at runtime would need glue code basically switching between
> between different *Lib implementations, depending on machine type.
> Similar to DxePciLibI440FxQ35 for pci / pcie, but for a bunch of more
> places too.
> 
> Xen moved to a separate dsc/fdf too recently, I suspect for simliar
> reasons.
> 
> take care,
>   Gerd