From: Stefan Fritsch <sf@openbsd.org>
Subject: Re: virtio_pci change for SEV
To: Mark Kettenis <mark.kettenis@xs4all.nl>
Cc: tech@openbsd.org
Date: Tue, 29 Oct 2024 09:47:46 +0100

On Mon, 28 Oct 2024, Mark Kettenis wrote:
> Sorry.  But the comments still make no sense to me.  The different
> bus_dma(9) implementations do wildly different things on different
> architectures.

The VIRTIO_F_ACCESS_PLATFORM bit means wildly different things on 
different architectures, too. I totally agree that a more well defined 
wording in the standard would would have been nice.

> 
> Maybe instead of making a statement about our bus_dma(9)
> implementation here that I'm not sure is true.  Just write something
> like:
> 
> /* XXX Without this SEV doesn't work with a KVM/qemu hypervisor on amd64. */
> 
> for the VIRTIO_F_ACCESS_PLATFORM bit?
> 
> And can we defer the VIRTIO_F_ORDER_PLATFORM bit for now?  I think
> that should be done in a separate diff and mixing the two in a single
> discussion isn't helping in understanding what's going on.

OK, done that now.