Re: [PATCH 11/20] drm/i915/dp: Reprobe connector if getting/acking device IRQs fails

[Imre Deak <imre.deak@intel.com>]

On Thu, Jul 03, 2025 at 02:28:01PM +0300, Luca Coelho wrote:
> On Thu, 2025-06-26 at 13:56 +0300, Imre Deak wrote:
> > On Thu, Jun 26, 2025 at 01:46:27PM +0300, Jani Nikula wrote:
> > > On Thu, 26 Jun 2025, Imre Deak <imre.deak@intel.com> wrote:
> > > > On Thu, Jun 26, 2025 at 01:23:12PM +0300, Jani Nikula wrote:
> > > > > On Thu, 26 Jun 2025, Imre Deak <imre.deak@intel.com> wrote:
> > > > > > On Thu, Jun 26, 2025 at 12:12:11PM +0300, Jani Nikula wrote:
> > > > > > > On Thu, 26 Jun 2025, Imre Deak <imre.deak@intel.com> wrote:
> > > > > > > > From: Imre Deak <imre.deak@gmail.com>
> > > > > > > > 
> > > > > > > > An AUX access failure during HPD IRQ handling should be handled by
> > > > > > > > falling back to a full connector detection, ensure that if the failure
> > > > > > > > happens while reading/acking a device service IRQ.
> > > > > > > > 
> > > > > > > > Signed-off-by: Imre Deak <imre.deak@gmail.com>
> > > > > > > > ---
> > > > > > > >  drivers/gpu/drm/i915/display/intel_dp.c | 21 +++++++++++++++------
> > > > > > > >  1 file changed, 15 insertions(+), 6 deletions(-)
> > > > > > > > 
> > > > > > > > diff --git a/drivers/gpu/drm/i915/display/intel_dp.c b/drivers/gpu/drm/i915/display/intel_dp.c
> > > > > > > > index 7793a72983abd..7eb208d2c321b 100644
> > > > > > > > --- a/drivers/gpu/drm/i915/display/intel_dp.c
> > > > > > > > +++ b/drivers/gpu/drm/i915/display/intel_dp.c
> > > > > > > > @@ -5393,16 +5393,20 @@ void intel_dp_check_link_state(struct intel_dp *intel_dp)
> > > > > > > >  	intel_encoder_link_check_queue_work(encoder, 0);
> > > > > > > >  }
> > > > > > > >  
> > > > > > > > -static void intel_dp_check_device_service_irq(struct intel_dp *intel_dp)
> > > > > > > > +static bool intel_dp_check_device_service_irq(struct intel_dp *intel_dp)
> > > > > > > 
> > > > > > > I don't think "check" is very intuitive in function names. Check
> > > > > > > something, but then what? Is it like an assert or does it do something
> > > > > > > active or what?
> > > > > > > 
> > > > > > > What does a boolean return from a check function mean?
> > > > > > > 
> > > > > > > It's not obvious to the reader at all.
> > > > > > 
> > > > > > I agree, but in this patch I didn't want to change the function name.
> > > > > 
> > > > > Arguably adding a return value changes the meaning already...
> > > > > 
> > > > > > 
> > > > > > > 
> > > > > > > >  {
> > > > > > > >  	struct intel_display *display = to_intel_display(intel_dp);
> > > > > > > >  	u8 val;
> > > > > > > >  
> > > > > > > >  	if (drm_dp_dpcd_readb(&intel_dp->aux,
> > > > > > > > -			      DP_DEVICE_SERVICE_IRQ_VECTOR, &val) != 1 || !val)
> > > > > > > > -		return;
> > > > > > > > +			      DP_DEVICE_SERVICE_IRQ_VECTOR, &val) != 1)
> > > > > > > > +		return true;
> > > > > > > 
> > > > > > > Looks like true means the check failed... while usually true for boolean
> > > > > > > functions means success.
> > > > > > 
> > > > > > The function returns true as before if a full connector detection is needed.
> > > > > 
> > > > > But it didn't return anything before! And that meaning is not conveyed
> > > > > to the reader in *any* reasonable way!
> > > > 
> > > > This function is the counterpart of intel_dp_check_link_service_irq()
> > > > both functions having the same purpose, reading and handling HPD IRQs.
> > > > The latter one's return value is true if a reprobe is needed and this
> > > > patch doesn't change that, it keeps the two functions behave the same
> > > > way.
> > > > 
> > > > > The absolute minimum is to add a comment (mind you, kernel-doc is
> > > > > overkill) stating what the return value means.
> > > > 
> > > > The function name will change in a follow-up patch and I think that
> > > > doesn't require a comment on the return value.
> > > > 
> > > > > > > 
> > > > > > > >  
> > > > > > > > -	drm_dp_dpcd_writeb(&intel_dp->aux, DP_DEVICE_SERVICE_IRQ_VECTOR, val);
> > > > > > > > +	if (!val)
> > > > > > > > +		return false;
> > > > > > > > +
> > > > > > > > +	if (drm_dp_dpcd_writeb(&intel_dp->aux, DP_DEVICE_SERVICE_IRQ_VECTOR, val) != 1)
> > > > > > > > +		return true;
> > > > > > > >  
> > > > > > > >  	if (val & DP_AUTOMATED_TEST_REQUEST)
> > > > > > > >  		intel_dp_test_request(intel_dp);
> > > > > > > 
> > > > > > > Whoa, it's not a *check* function at all?! It actually *handles* the
> > > > > > > service irqs.
> > > > > > > 
> > > > > > > Can we rephrase the function name?
> > > > > > 
> > > > > > I want to keep the function name in this patch. In the following patches
> > > > > > I will separate this part and rename it to
> > > > > > intel_dp_get_and_ack_device_service_irq(). 
> > > > > 
> > > > > Right, saw that now. But even for that function name the meaning of the
> > > > > return value is ambiguous.
> > > > 
> > > > All the get/ack IRQ functions in intel_dp.c return true for success.
> > > 
> > > Argh. You just said it doesn't mean success/failure, it means if full
> > > connector detection is needed?!
> > 
> > intel_dp_check_device_service_irq(),
> > intel_dp_check_link_service_irq() -> return value indicates if a
> > connector detection is needed.
> > 
> > intel_dp_get_and_ack_device_service_irq(),
> > intel_dp_get_and_ack_link_service_irq() -> return value indicates if
> > getting/acking the IRQ succeeded,s imilarly to
> > intel_dp_get_sink_irq_esi(), intel_dp_ack_sink_irq_esi().
> 
> Do we need to distinguish between when getting or acking failed?

No, in either case the IRQ shouldn't be handled and a full detection
should be scheduled for the connector.

> You may have handled the irq but failed to ack (theoretically).  Do
> you just abort the whole thing in either case?

If reading or acking the IRQs fail, which would be due to the relevant
AUX read/write failing, then the IRQ will not be handled and a full
connector detection will be scheduled.

> I still tend to agree with Jani that the idea of actually handling the
> interrupt is not clear in the function name. 

Yes, I agree it's not clear, but that is the current name, which I don't
want to change in this patch. In patch 13 I separate the function into
intel_dp_get_and_ack_device_service_irq() and
intel_dp_handle_device_service_irq().

> _get_and_ack doesn't imply that either.

intel_dp_get_and_ack_device_irq(), added in patch 13, will only get and
ack the IRQ, the handling happening in
intel_dp_handle_device_service_irq().

> But this is getting too nitpicky at this point, so I'll leave it up to
> you and Jani. :)