[Intel-gfx] Valid DP connection without EDID?

[Takashi Iwai]

At Tue, 18 Sep 2012 10:04:33 -0400,
Adam Jackson wrote:
> 
> On Tue, 2012-09-18 at 13:01 +0200, Takashi Iwai wrote:
> 
> > > I started a patch series for this a bit ago, I'll send it on
> > > momentarily.
> > 
> > Thanks!  I evaluated it now (with a typo fix suggested by Jani).
> > Unfortunately, it doesn't improve the situation.
> > 
> > The fetch of downstream ports succeeds, and it gets 0x09.  So, this
> > indicates again it's a VGA downstream port.  But that's all, so far.
> > The 0x09 is reported no matter whether the VGA cable is plugged or
> > not, so this can't be used as the detection of the downstream port
> > plug state.
> 
> Sorry, there's a bug in the patch.  link_configuration[0] is not
> DP_SINK_COUNT, I have no idea why I thought it was.  Try this on top of
> the series:
> 
> ===
> diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
> index 9809c53..b6b9a18 100644
> --- a/drivers/gpu/drm/i915/intel_dp.c
> +++ b/drivers/gpu/drm/i915/intel_dp.c
> @@ -2098,15 +2098,22 @@ intel_dp_detect_dpcd(struct intel_dp *intel_dp)
>  
>         if (!intel_dp_get_dpcd(intel_dp))
>                 return connector_status_disconnected;
> -       
> +
>         /* if there's no downstream port, we're done */
>         if (!(dpcd[DP_DOWNSTREAMPORT_PRESENT] & DP_DWN_STRM_PORT_PRESENT))
>                 return connector_status_connected;
>  
>         /* If we're HPD-aware, SINK_COUNT changes dynamically */
>         hpd = !!(intel_dp->downstream_ports[0] & DP_DS_PORT_HPD);
> -       if (hpd && (intel_dp->link_configuration[0] & DP_SINK_COUNT_MASK))
> -               return connector_status_connected;
> +       if (hpd) {
> +               uint8_t sink_count;
> +               if (!intel_dp_aux_native_read_retry(intel_dp, DP_SINK_COUNT,
> +                                                   &sink_count, 1))
> +                       return connector_status_unknown;
> +               sink_count &= DP_SINK_COUNT_MASK;
> +               return sink_count ? connector_status_connected
> +                                 : connector_status_disconnected;
> +       }
>  
>         /* If no HPD, poke DDC gently */
>         if (drm_probe_ddc(&intel_dp->adapter))
> ===

Woohoo, the patch works!

This also avoids the driver spewing tons of error messages 
  [drm:intel_dp_i2c_aux_ch] *ERROR* too many retries, giving up
and
  [drm:intel_dp_complete_link_train] *ERROR* failed to train DP, aborting

I guess it's because now the driver detects the disconnection
properly.  The hotplug / -unplug also seems working.

Could you brush up and resend the patches for merging to 3.7 kernel?
Or, would you like to fix the multiple branch devices first?

When you resubmit patches, feel free to add:
  Tested-by: Takashi Iwai <tiwai at suse.de>


Thanks!

Takashi


> If that doesn't work then the HPD-capable bit is useless - or if we're
> lucky just needs quirking by branch OUI - and we should just fall
> through to the drm_probe_ddc() path.  What is the branch OUI, anyway?
> 
> There's a third possibility, which is that HPD does work but that we're
> not doing enough to enable it.  The DP 1.1a spec has a non-normative
> appendix describing one way a device could go about doing that as an
> optional feature, but the method described does not match how we're
> currently handling sink-specific IRQs.  I have no idea what the 1.2 spec
> says on this point though.
> 
> - ajax