MIPI DSI, DBI, and tinydrm drivers

Noralf Trønnes noralf at tronnes.org
Sun May 24 21:24:11 UTC 2020

Den 24.05.2020 22.42, skrev Paul Cercueil:
> Le dim. 24 mai 2020 à 22:14, Noralf Trønnes <noralf at tronnes.org> a écrit :
>> Den 24.05.2020 21.54, skrev Paul Cercueil:
>>>  Hi Noralf,
>>>  Le dim. 24 mai 2020 à 19:46, Noralf Trønnes <noralf at tronnes.org> a
>>> écrit :
>>>>  Den 24.05.2020 18.13, skrev Paul Cercueil:
>>>>>   Hi list,
>>>>>   I'd like to open a discussion about the current support of MIPI
>>>>> DSI and
>>>>>   DBI panels.
>>>>>   Both are standards from the MIPI alliance, both are communication
>>>>>   protocols between a LCD controller and a LCD panel, they
>>>>> generally both
>>>>>   use the same commands (DCS), the main difference is that DSI is
>>>>> serial
>>>>>   and DBI is generally parallel.
>>>>>   In the kernel right now, DSI is pretty well implemented. All the
>>>>>   infrastucture to register a DSI host, DSI device etc. is there. DSI
>>>>>   panels are implemented as regular drm_panel instances, and their
>>>>>  drivers
>>>>>   go through the DSI API to communicate with the panel, which makes
>>>>> them
>>>>>   independent of the DSI host driver.
>>>>>   DBI, on the other hand, does not have any of this. All (?) DBI
>>>>> panels
>>>>>   are implemented as tinydrm drivers, which make them impossible to
>>>>> use
>>>>>   with regular DRM drivers. Writing a standard drm_panel driver is
>>>>>   impossible, as there is no concept of host and device. All these
>>>>>  tinydrm
>>>>>   drivers register their own DBI host as they all do DBI over SPI.
>>>>>   I think this needs a good cleanup. Given that DSI and DBI are so
>>>>>   similar, it would probably make sense to fuse DBI support into the
>>>>>   current DSI code, as trying to update DBI would result in a lot
>>>>> of code
>>>>>   being duplicated. With the proper host/device registration mechanism
>>>>>   from DSI code, it would be possible to turn most of the tinydrm
>>>>> drivers
>>>>>   into regular drm_panel drivers.
>>>>>   The problem then is that these should still be available as tinydrm
>>>>>   drivers. If the DSI/DBI panels can somehow register a .update_fb()
>>>>>   callback, it would make it possible to have a panel-agnostic tinydrm
>>>>>   driver, which would then probably open a lot of doors, and help a
>>>>>  lot to
>>>>>   clean the mess.
>>>>>   I think I can help with that, I just need some guidance - I am
>>>>> fishing
>>>>>   in exotic seas here.
>>>>>   Thoughts, comments, are very welcome.
>>>>  I did look at this a few months back:
>>>>  drm/mipi-dbi: Support panel drivers
>>>> https://lists.freedesktop.org/archives/dri-devel/2019-August/228966.html
>>>>  The problem with DBI is that it has reused other busses which means we
>>>>  don't have DBI drivers, we have SPI drivers instead (6800/8080 is not
>>>>  avail. as busses in Linux yet). DSI and DPI on the other hand has
>>>>  dedicated hw controller drivers not shared with other subsystems.
>>>  I don't think that should be much of a problem. You could have a
>>>  bridge, that wraps a SPI device into a DBI host, for instance. The
>>> panel
>>>  drivers would just use the DBI API without having to know what's done
>>>  behind the scene.
>> This will be a bridge implemented in software, are we allowed to have
>> software devices in the Device Tree? I though it was just allowed to
>> describe hardware.
> It wouldn't appear in devicetree. If the panel is connected over SPI,
> then DBI is just the protocol it uses.

How do you attach a panel to the DBI device if it doesn't appear in DT?
Another problem is that the DBI panel uses SPI both for framebuffer
upload and controller initialization. How shall this be handled when the
panel driver needs SPI for init and the DBI bridge needs SPI for frame


> If probed as a SPI device driver, the panel's spi_driver would register
> an instance of the DBI/SPI host driver, then register itself as a
> dbi_driver. If probed from a DBI host it would just register itself as a
> dbi_driver.
> -Paul
>>>>  My initial tinydrm work used drm_panel, but I was not allowed to
>>>> use it
>>>>  (at least not the way I had done it).
>>>>  Noralf.
>>>>>   Cheers,
>>>>>   -Paul

More information about the dri-devel mailing list