[Intel-gfx] DisplayPort support

[Chris Wilson]

On Wed, 27 Apr 2011 18:27:51 +1000, Paul McGarry <paul at paulmcgarry.com> wrote:
> On Wed, Apr 27, 2011 at 4:49 PM, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> > DP has worked for most people the last 18 months or so... There is always
> > to unhappy exception. Can you please append drm.debug=0xe to you kernel
> > boot parameters and attach the dmesg from booting?
> > -Chris
> 
> Hi Chris,
> 
> Please find that attached.
> It is from a boot with only the DP monitor attached.

The unusual sequence appears to be:

[   30.081906] [drm:intel_dp_i2c_init], i2c_init DPDDC-B
[   30.082413] [drm:intel_dp_aux_ch], dp_aux_ch timeout status 0x5143003e
[   30.082415] [drm:intel_dp_i2c_aux_ch], aux_ch failed -110
[   30.082921] [drm:intel_dp_aux_ch], dp_aux_ch timeout status 0x5143003e
[   30.082923] [drm:intel_dp_i2c_aux_ch], aux_ch failed -110
[   30.082986] [drm:intel_dp_i2c_init], i2c_init DPDDC-C
[   30.083193] [drm:intel_dp_i2c_aux_ch], aux_i2c nack
[   30.083401] [drm:intel_dp_i2c_aux_ch], aux_i2c nack
[   30.083427] [drm:intel_dp_i2c_init], i2c_init DPDDC-D
[   30.083934] [drm:intel_dp_aux_ch], dp_aux_ch timeout status 0x5143003e
[   30.083935] [drm:intel_dp_i2c_aux_ch], aux_ch failed -110
[   30.084442] [drm:intel_dp_aux_ch], dp_aux_ch timeout status 0x5143003e
[   30.084443] [drm:intel_dp_i2c_aux_ch], aux_ch failed -110

I wonder what side-effects that explicit NAK from the monitor does during
initialisation of the DP i2c channel.

During detection we also see a discrepancy for DPDDC-D:

[   30.172250] [drm:drm_helper_probe_single_connector_modes],
[CONNECTOR:19:DP-2]
[   30.172757] [drm:intel_dp_aux_ch], dp_aux_ch timeout status 0x5013003e
[   30.172758] [drm:ironlake_dp_detect], DPCD: 0000

So instead of sending the expected 20 bytes to query the connection, we
just sent the 1 byte header. It seems our problems start with that nack.

Lets start with a WARN for the AUX_I2C_REPLY_NACK so that I can see the
callpath for that initial probe and so work out the command being sent
that the monitor doesn't like.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre