Re: [PATCH v6 0/8] drm/tidss: Use new connector model for tidss

[Tomi Valkeinen]

On 09/05/2023 12:30, Aradhya Bhatia wrote:
> Hi all,
>
> I have picked up this long standing series from Nikhil Devshatwar[1].
>
> This series moves the tidss to using new connectoe model, where the SoC
> driver (tidss) creates the connector and all the bridges are attached
> with the flag DRM_BRIDGE_ATTACH_NO_CONNECTOR. It also now creates bridge
> to support format negotiation and and 'simple' encoder to expose it to
> the userspace.
>
> Since the bridges do not create the connector, the bus_format and
> bus_flag is set via atomic hooks.
>
> Support format negotiations in the tfp410, sii902x and mhdp-8546 bridge
> drivers as a first step before moving the connector model.
>
> These patches were tested on AM625-SK EVM, AM625 SoC based BeaglePlay,
> and J721E-SK. Display support for AM625 SoC has not been added upstream
> and is a WIP. To test this series on AM625 based platforms, basic
> display support patches, (for driver + devicetree), can be found in
> the "next_AttachNoConn" branch on my github fork[2].

Not exactly related to this series, but I was testing this with J7 EVM 
and a Dell DP monitor.

I'm seeing DPCD errors quite often, which also sometimes cause link 
training errors. E.g.:

cdns-mhdp8546 a000000.dp-bridge: Failed to read DPCD addr 0

cdns-mhdp8546 a000000.dp-bridge: Failed to adjust Link Training.

I tested retrying the read/write when it fails, and that seemed to help 
a bit, but not always. Are you using the firmware that's on the upstream 
linux-firmware repo?

 Tomi