On Sun, Jun 11, 2023 at 08:17:56PM +0000, Parav Pandit wrote:
>
> > From: Michael S. Tsirkin <[email protected]>
> > Sent: Sunday, June 11, 2023 4:09 PM
>
> > > > It's not conformant to that statement then :( It's a SHOULD which
> > > > means if you know exactly what you are doing, there could be exceptions.
> > > > In this case it's a SHOULD because it was added after 1.0 and we did
> > > > not find a way to negotiate this requirement since it's the feature
> > > > negotiation
> > itself.
> > >
> > > True. May be in future such device can update the revision id.
> >
> > You can already change it but that doesn't imply anything at all, in 1.0
> > revisions
> > are vendor specific.
>
> I was suggesting that when such a fundamental change is done which cannot be
> communicated via features bit or before the reset,
> a new revision and attaching more advance/different behavior.
> This way driver knows what difference to expect for new revision.
But you don't know which revision do existing devices use:
Drivers MUST match any PCI Revision ID value.
we can come up with a mechanism, several ideas come to mind.
Before we bother, the actual need has to be understood.
--
MST
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
