[PATCH V3 2/2] drm/bridge: ti-sn65dsi83: Add TI SN65DSI83 and SN65DSI84 driver

[Dave Stevenson]

On Thu, 6 May 2021 at 21:51, Marek Vasut <marex at denx.de> wrote:
>
> On 5/6/21 6:03 PM, Frieder Schrempf wrote:
> > On 06.05.21 17:46, Marek Vasut wrote:
> >> On 5/6/21 5:38 PM, Frieder Schrempf wrote:
> >> [...]
> >>> Works on i.MX8MM with SN65DSI84 and a single link LVDS display (1024x600) and from my perspective everything else also looks good. Thanks for your work!
> >>>
> >>> I have two remarks:
> >>>
> >>> 1. In my test I couldn't get it to work with four DSI lanes enabled (only with two) but I'm quite sure that the DSIM driver is to blame as everything on the bridge level looks good (also setting the DSI EQ register didn't help as you suggested, Marek).
> >>
> >> I suspect there is indeed something with the DSIM going on, I'll keep you posted if I find something out.
> >>
> >>> 2. When I set MEDIA_BUS_FMT_RGB888_1X7X4_SPWG in the panel driver I get distorted colors. I need to use MEDIA_BUS_FMT_RGB888_1X24 to make it work, but this is not valid for LVDS. Again I don't think this driver is to blame as I can't see where it does anything wrong, but my experience here is very limited so I still want to mention it.
> >>
> >> Hmm, in that conversion supposed to happen in this bridge driver or should MXSFB handle the SPWG pixel format ? Or should the DSIM bridge do something about that ?
> >
> > As far as I understand it the conversion is already done by the DSI84 without any extra configuration necessary. The only thing that needs to be done is selecting the LVDS output format via CHx_24BPP_MODE and CHx_24BPP_FORMAT1 which the driver currently hardcodes to 24bpp aka MEDIA_BUS_FMT_RGB888_1X7X4_SPWG. I think the DSI input format is always 24bpp aka MEDIA_BUS_FMT_RGB888_1X24.
>
> The DSI is MEDIA_BUS_FMT_RGB888_1X24, yes.
>
> So maybe this bridge driver has to somehow deal with
> MEDIA_BUS_FMT_RGB888_1X7X4_SPWG ? Except I haven't seen such a thing
> implemented in other bridge drivers, so input would be welcome on this.

I'm claiming no knowledge of whether this is the correct approach or
not, but Toshiba TC358775 is also a DSI to LVDS bridge which appears
to handle both formats.
https://elixir.free-electrons.com/linux/latest/source/drivers/gpu/drm/bridge/tc358775.c#L457

> > So I wonder where the format actually is evaluated. Could it be that it is passed down to the LCDIF and changes its output format which causes the data passed by DSIM to the DSI84 to already be in the SPWG format? If that's the case we maybe need a way to specify MEDIA_BUS_FMT_RGB888_1X24 as input bus format for the DSI84 so it doesn't pass on the panel's format? Only a wild guess, no idea if it really works like that.
>
> I _think_ the bridge must somehow handle the
> MEDIA_BUS_FMT_RGB888_1X7X4_SPWG <-> MEDIA_BUS_FMT_RGB888_1X24 conversion.

I've not looked at where the interchange happens, but as you're
setting the DSI format in struct mipi_dsi_device to
MIPI_DSI_FMT_RGB888 doesn't that provide the configuration side to the
DSI transmitter?
Otherwise presumably it needs to support the atomic_get_input_bus_fmts
and/or atomic_get_output_bus_fmts functions in drm_bridge_funcs.

  Dave