Re: [Intel-gfx] [PATCH v2 1/2] drm/i915/display: ignore long HPDs based on a flag

["Govindapillai, Vinod" <vinod.govindapillai@intel.com>]

On Mon, 2022-10-10 at 14:08 +0200, Andrzej Hajda wrote:
> On 10.10.2022 10:34, Vinod Govindapillai wrote:
> > Some panels generate long HPD events even while connected to
> > the port. This cause some unexpected CI execution issues. A
> > new flag is added to track if such spurious long HPDs can be
> > ignored and are not processed further if the flag is set.
> > 
> > v2: Address patch styling comments (Jani Nikula)
> > 
> > Signed-off-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
> 
> Yeah, especially bat-rpls-2 is quite buggy[1] (500 long HPDs per log).
> Do you plan to integrate it to CI? I guess i915 should be immune to such 
> misbehavior anyway.
> 
> [1]: 
> https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_12228/bat-rpls-2/dmesg0.txt
> 
> Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
> 
> Regards
> Andrzej

Thanks. Yes. IGT patch is floated to take this into use in CI
https://patchwork.freedesktop.org/series/109531/

> 
> 
> > ---
> >   drivers/gpu/drm/i915/display/intel_display_core.h | 11 +++++++++++
> >   drivers/gpu/drm/i915/display/intel_dp.c           |  7 +++++++
> >   2 files changed, 18 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/display/intel_display_core.h
> > b/drivers/gpu/drm/i915/display/intel_display_core.h
> > index 96cf994b0ad1..f84a2e760ef4 100644
> > --- a/drivers/gpu/drm/i915/display/intel_display_core.h
> > +++ b/drivers/gpu/drm/i915/display/intel_display_core.h
> > @@ -170,6 +170,17 @@ struct intel_hotplug {
> >          * blocked behind the non-DP one.
> >          */
> >         struct workqueue_struct *dp_wq;
> > +
> > +       /*
> > +        * Flag to track if long HPDs need not to be processed
> > +        *
> > +        * Some panels generate long HPDs while keep connected to the port.
> > +        * This can cause issues with CI tests results. In CI systems we
> > +        * don't expect to disconnect the panels and could ignore the long
> > +        * HPDs generated from the faulty panels. This flag can be used as
> > +        * cue to ignore the long HPDs and can be set / unset using debugfs.
> > +        */
> > +       bool ignore_long_hpd;
> >   };
> >   
> >   struct intel_vbt_data {
> > diff --git a/drivers/gpu/drm/i915/display/intel_dp.c b/drivers/gpu/drm/i915/display/intel_dp.c
> > index 70b06806ec0d..87a0d01d8003 100644
> > --- a/drivers/gpu/drm/i915/display/intel_dp.c
> > +++ b/drivers/gpu/drm/i915/display/intel_dp.c
> > @@ -5092,6 +5092,13 @@ intel_dp_hpd_pulse(struct intel_digital_port *dig_port, bool long_hpd)
> >                 return IRQ_HANDLED;
> >         }
> >   
> > +       if (i915->display.hotplug.ignore_long_hpd && long_hpd) {
> > +               drm_dbg_kms(&i915->drm, "[ENCODER:%d:%s] ignoring long hpd\n",
> > +                           dig_port->base.base.base.id,
> > +                           dig_port->base.base.name);
> > +               return IRQ_HANDLED;
> > +       }
> > +
> >         drm_dbg_kms(&i915->drm, "got hpd irq on [ENCODER:%d:%s] - %s\n",
> >                     dig_port->base.base.base.id,
> >                     dig_port->base.base.name,
>