Re: [PATCH 22/35] drm/panel: drop the orientation workaround from get_modes()

[Romain TISSERAND <romain.tisserand@gmail.com>]

On Mon, Oct 26, 2026, Dmitry Baryshkov wrote:
> The exceptions are sun6i_mipi_dsi, sti_dvo and the Tegra DSI, which get
> the panel only after the DRM device has been registered.
> None of the in-tree boards using them has a rotated panel.

Two are on their way for sun6i_mipi_dsi, both A33 handhelds with the
panel mounted for landscape use, rotation = <270>: the Miyoo PocketGo S30
(ILI9488, panel series at [1]) and the Miyoo A30 (ST7701, not posted yet).

I sent a fix for this gap on 1 November [2]: sun6i_dsi_bind() reads the
rotation from the panel node, a child of the DSI node, and sets the
orientation on the connector before the DRM device is registered. It
does not rely on the panel driver, so it keeps working once the
get_modes() workaround is gone, and it skips port/ports nodes, so it
does not mind whether the board also links the panel through the graph.
Tested on the S30; it still applies to drm-misc-next as of 12 November.

Could your repost of 15+ take it, or point at it for sun6i_mipi_dsi? It
reads the DT itself rather than panel->orientation from your 23/35,
because the panel is not there yet at bind time. I am happy to rebase it
on top of your series if you prefer.

[1] https://lore.kernel.org/all/20261109160706.2272332-1-romain.tisserand@gmail.com/
[2] https://lore.kernel.org/all/20261101203406.1360656-1-romain.tisserand@gmail.com/

Romain