Re: [PATCH v3 1/2] dt-bindings: drm/bridge: Add no-hpd property

[Tomi Valkeinen]

Hi,

On 12/03/2025 14:52, Dmitry Baryshkov wrote:
> On Wed, Mar 12, 2025 at 11:56:41AM +0530, Harikrishna Shenoy wrote:
> > On 05/02/25 19:03, Dmitry Baryshkov wrote:
> > > On Wed, Feb 05, 2025 at 12:52:52PM +0100, Krzysztof Kozlowski wrote:
> > > > On 05/02/2025 12:50, Harikrishna Shenoy wrote:
> > > > > From: Rahul T R <r-ravikumar@xxxxxx>
> > > > >
> > > > > The mhdp bridge can work without its HPD pin hooked up to the connector,
> > > > > but the current bridge driver throws an error when hpd line is not
> > > > > connected to the connector. For such cases, we need an indication for
> > > > > no-hpd, using which we can bypass the hpd detection and instead use the
> > > > > auxiliary channels connected to the DP connector to confirm the
> > > > > connection.
> > > > > So add no-hpd property to the bindings, to disable hpd when not
> > > > > connected or unusable due to DP0-HPD not connected to correct HPD
> > > > > pin on SOC like in case of J721S2.
> > > > >
> > > > > Signed-off-by: Rahul T R <r-ravikumar@xxxxxx>
> > > >
> > > > Why are you sending over and over the same? You already got feedback.
> > > > Then you send v2. You got the same feedback.
> > > >
> > > > Now you send v3?
> > > >
> > > > So the same feedback, but this time: NAK
> > >
> > > Krzysztof's email forced me to take a look at the actual boards that you
> > > are trying to enable. I couldn't stop by notice that the HPD signal
> > > _is_ connected to a GPIO pin. Please stop hacking the bridge driver and
> > > use the tools that are already provided to you: add the HPD pin to the
> > > dp-controller device node. And then fix any possible issues coming from
> > > the bridge driver not being able to handle HPD signals being delivered
> > > by the DRM framework via the .hpd_notify() callback.
> > >
> > > TL;DR: also a NAK from my side, add HPD gpio to dp-controller.
> > We tried implementing a interrupt based HPD functionality as HPD signal is
> > connected to GPIO0_18 pin, we were able to get interrupt based HPD working
> > however to route this signal to SoC we are loosing audio capability due to
> > MUX conflict. Due to board level limitations to
> > route the signal to SoC, we will not be able to support interrupt
> > based HPD and polling seems a possible way without loosing on audio
> > capability.
>
> Still NAK for the no-hpd property. HPD pin is a requirement for
> DisplayPort to work, as it is used e.g. for the 'attention' IRQs being
> sent by the DP sink. I'm not sure what kind of idea you HW engineers had
> in mind.

It's true that for normal DP functionality the HPD is required, but 
afaik DP works "fine" without HPD too. This is not the first board that 
has DP connector, but doesn't have HPD, that I have seen or worked on. 
Polling can be used for the IRQs too.

For eDP HPD is optional, and some of the cases I've worked with involved 
a chip intended for eDP, but used with a full DP connector, and no HPD. 
However, in this particular case the DP chip supports full DP, so it's 
just a board design error.

My question is, is J721s2 EVM something that's used widely? Or is it a 
rare board? If it's a rare one, maybe there's no point in solving this 
in upstream? But if it's widely used, I don't see why we wouldn't 
support it in upstream. The HW is broken, but we need to live with it.

Another question is, if eDP support is added to the cdns-mhdp driver, 
and used with a panel that doesn't have an HPD, how would that code look 
like? If that would be solved with a "no-hpd" property, identical to the 
one proposed in this series, then... There's even less reason to not 
support this.

Disclaimer: I didn't study the schematics, and I haven't thought or 
looked at how eDP is implemented in other drm drivers.

 Tomi