[Intel-gfx] [PATCH 0/3] BYT DSI Dual Link Support

[Jani Nikula]

On Mon, 24 Nov 2014, "Singh, Gaurav K" <gaurav.k.singh at intel.com> wrote:
> Hi Jani,
>
> Thanks for the review comments.
>
> Regarding the first 2 patches, I was doing almost the same thing in my 
> 3rd and 4th patch. But your patches are more generic.
>
> Regarding the 3rd patch, I have a comment:
>
> Since in case of dual link panels, few panels may require sequence to be 
> sent only on Port A or Port C or both. In that case, 
> for_each_dsi_port(port, intel_dsi->ports) will cause it to be sent to 
> both ports.
> To resolve this, in the earlier patches, intel_dsi->port was used which 
> gets calculated to either 0 or 1 in mipi_exec_send_packet(). This value 
> of 0 or 1 is dependent on sequence block#53.
>
>  From now on as we will be using the _PORT3() macro for using proper 
> MIPI regs, then for this scenario, we may need to have some 
> workaround/hardcode type of code again. May I know your suggestion on this?

Perhaps patch 3/3 was a bad place for the for_each_dsi_port example - it
should have been put into intel_dsi.c. Maybe you'll need to add the port
as parameter to the functions in intel_dsi_cmd.c, and let the caller
decide which port should be used? I want to avoid any platform/panel
specific special casing in intel_dsi.c and intel_dsi_cmd.c.

I think the general case for for_each_dsi_port is in intel_dsi.c anyway.

BR,
Jani.


>
> With regards,
> Gaurav
>
> On 11/14/2014 8:24 PM, Jani Nikula wrote:
>> Hi Shobhit and Gaurav -
>>
>> I've been pondering this whole MIPI DSI pipes vs. ports thing and
>> discussing with Ville and others. Rather than try and fail in explaining
>> the ideas, here are some concrete patches to describe what I'd like to
>> be done first.
>>
>> The most important thing is that we don't confuse the pipes and
>> ports. Getting confused was easy with the pipe B mapping to port C, and
>> the register defines being very confused/confusing about it. These
>> patches attempt to fix that. Before adding dual link support, there's a
>> simple function mapping the pipe to port.
>>
>> Next up is expanding that to handle multiple ports driven from one
>> pipe. That's handled by adding intel_dsi->ports bitmap that has the bit
>> set for each port that is to be driven. I've added the bitmap and some
>> helpers to iterate over the configured ports, but there's no actual
>> support for doing the configuration. I'm hoping you could take over from
>> here. There's a sample patch about the usage.
>>
>> I'm sorry it's taken me so long to reply. With the new stuff coming in,
>> I really think it's important to get the foundation right
>> first. Especially because I'm to blame for getting some of the port/pipe
>> stuff confused in the first place...
>>
>> BR,
>> Jani.
>>
>>
>>
>> Jani Nikula (3):
>>    drm/i915/dsi: clean up MIPI DSI pipe vs. port usage
>>    drm/i915/dsi: add ports to intel_dsi to describe the ports being
>>      driven
>>    drm/i915/dsi: an example how to handle dual link for each port
>>
>>   drivers/gpu/drm/i915/i915_reg.h      | 303 ++++++++++++++++++-----------------
>>   drivers/gpu/drm/i915/intel_dsi.c     | 151 ++++++++---------
>>   drivers/gpu/drm/i915/intel_dsi.h     |  19 +++
>>   drivers/gpu/drm/i915/intel_dsi_cmd.c |  76 ++++-----
>>   4 files changed, 290 insertions(+), 259 deletions(-)
>>
>

-- 
Jani Nikula, Intel Open Source Technology Center