Re: Re: Re: [PATCH v3 1/2] mailbox: pcc: support polling mode when there is no platform IRQ

[Andrea Tomassetti]

On 25/12/04 01:14PM, Sudeep Holla wrote:
> On Thu, Dec 04, 2025 at 01:59:38PM +0100, Andrea Tomassetti wrote:
> > On 25/12/03 10:28AM, Sudeep Holla wrote:
> > > On Tue, Dec 02, 2025 at 11:12:14AM +0100, Andrea Tomassetti wrote:
> > > > The goal is to allow clients to submit a message in both irq and polling
> > > > mode of the pcc mailbox. The ACPI specification does not require a
> > > > platform irq for pcc channels. Let's implement the case where it is not
> > > > available.
> > > >
> > > 
> > > Just curious if you have a real use case for this polling mode on your
> > > platforms or ...
> > > 
> > > > Tested-by: Thibault Cantori <thibault.cantori@xxxxxxxxxxx>
> > > > Co-developed-by: Olivier Dautricourt <olivier.dautricourt@xxxxxxxxxxx>
> > > > Signed-off-by: Olivier Dautricourt <olivier.dautricourt@xxxxxxxxxxx>
> > > > Signed-off-by: Andrea Tomassetti <andrea.tomassetti@xxxxxxxxxxx>
> > > > Reported-by: kernel test robot <lkp@xxxxxxxxx>
> > > > Closes: https://lore.kernel.org/oe-kbuild-all/202511120243.soxAFpqQ-lkp@xxxxxxxxx/
> > > > Closes: https://lore.kernel.org/oe-kbuild-all/202511120558.Cln7LF6M-lkp@xxxxxxxxx/
> > > > Closes: https://lore.kernel.org/oe-kbuild-all/202511120729.R3XQNSnx-lkp@xxxxxxxxx/
> > > 
> > > You are just trying to fix these warnings. If it is latter, we don't have to
> > > add support for polling mode especially if it can't be tested on real
> > > platforms.
> > >
> > In our target product, we're still investigating if PCC-based SCMI communication will
> > rely on interrupts or polling. When we started looking into it we realized that polling
> > wasn't supported and that's why we decided to work on and send this patch. We thought it
> > could have been beneficial to other members of the community and it brings the driver a
> > bit closer to the ACPI specifications.
> > 
> > We're using ARM Fast Models for prototyping and that's how we validated and tested this patch.
> > 
> 
> I wouldn't consider that as real platform especially if it is not std. AEM
> models that are well maintained. Many Fast models are short lived and never
> maintained long term, so I don't want to push any feature based on that alone
> unless you have a real platform with missing or broken interrupt that needs
> this polling feature.
> 
> It is burden for long term maintenance if there is no regular way to test this
> polling mode feature.
> 
Fair, I totally get your point. Thank you very much for your time.

Regards,
Andrea
> -- 
> Regards,
> Sudeep