Re: [virtio-dev] Re: [PATCH v3 0/3] transport-pci: Introduce legacy registers access using AQ

["Michael S. Tsirkin" <mst@redhat.com>]

On Wed, Jun 07, 2023 at 10:27:12AM +0800, Jason Wang wrote:
> On Tue, Jun 6, 2023 at 7:56âPM Michael S. Tsirkin <mst@redhat.com> wrote:
> >
> > On Mon, Jun 05, 2023 at 10:12:59PM +0000, Parav Pandit wrote:
> > > Rolling v4 now.
> >
> > Great thanks! I think the result will be in a good shape from
> > the ABI point of view. Good job!
> >
> > I think so far Jason was the only one with significant comments
> > on the series so let's see what he says.
> 
> I can't say I like this. I prefer to do meditation on top of a modern
> device with some lightweight features like _F_LEAGCY_HEADER. I don't
> see any advantages of a new legacy ABI over _F_LEGACY_HEADER. I don't
> want to repeat but I don't get any response here
> 
> https://lore.kernel.org/virtio-dev/CACGkMEuixsBibSq_pEQvcKOH0WBQWnta4Q2c28SV8qTS8gaVcA@mail.gmail.com/#t


After trying to write this I came away with a conviction that
if hardware vendors are prepared to implement legacy
in hardware we should let them do exactly that.


> >
> > From my personal POV what will be left to do to get this
> > feature ready for merging would be
> >
> > - tightening the language:
> > - thinking of and addressing all kind of corner cases.
> 
> We need first to check whether it's possible (not too late) to find
> all the corner cases.
> 
> Thanks

For sure, but of course we need to stop arguing about whether
to have the feature in the 1st place.

> >
> >
> > Since by now from our email discussions I have I think
> > a decent understanding of how the feature is supposed to work
> > I understand how these things work but I think there are still
> > things that would be a bit unclear for an unprepared reader.
> >
> > Just to give a random example, there's no explicit
> > time where the member kind of switches to the legacy/back to modern
> > mode. I am guessing setting DRIVER bit through legacy interface
> > does this. We'll want to list that.
> >
> > I'll try to help find these issues and maybe even suggest wording.
> >
> > I've delayed commenting on these until the high level design is agreed
> > on, didn't want to make you waste time on polishing as the patches
> > undergo significant changes, and I suggest you do the same in v4: let's
> > make sure there's a wide agreement on the ABI then focus on the spec
> > language.
> >
> >
> >
> > --
> > MST
> >