On Fri, Oct 18, 2024 at 04:01:17PM +0000, Arnd Bergmann wrote:
> On Fri, Oct 18, 2024, at 15:42, Dmitry Baryshkov wrote:
> > On Fri, Oct 18, 2024 at 03:08:45PM +0000, Arnd Bergmann wrote:
> >> From: Arnd Bergmann <a...@arndb.de>
> >>
> >> The imx display drivers use the new bridge connector helpers but don't
> >> explicitly select CONFIG_DRM_DISPLAY_HELPER, leading to a link failure
> >> in rare configurations:
> >>
> >> ld.lld-20: error: undefined symbol: drm_bridge_connector_init
> >> >>> referenced by imx-ldb.c
> >> >>> drivers/gpu/drm/imx/ipuv3/imx-ldb.o:(imx_ldb_bind) in
> >> >>> archive vmlinux.a
> >> >>> referenced by parallel-display.c
> >> >>>
> >> >>> drivers/gpu/drm/imx/ipuv3/parallel-display.o:(imx_pd_bind) in archive
> >> >>> vmlinux.a
> >>
> >> Move DRM_BRIDGE_CONNECTOR out of the DRM_DISPLAY_HELPER block and
> >> instead select that symbol indirectly.
> >>
> >> Fixes: f94b9707a1c9 ("drm/imx: parallel-display: switch to
> >> imx_legacy_bridge / drm_bridge_connector")
> >> Signed-off-by: Arnd Bergmann <a...@arndb.de>
> >
> > I think a correct fix is to select DRM_DISPLAY_HELPER instead.
>
> I think I just didn't express that clearly enough, selecting
> DRM_DISPLAY_HELPER is what I did and what I tried to explain:
>
> >> config DRM_BRIDGE_CONNECTOR
> >> bool
> >> select DRM_DISPLAY_HDMI_STATE_HELPER
> >> + select DRM_DISPLAY_HELPER
> >> help
> >> DRM connector implementation terminating DRM bridge chains.
>
> I had to move the 'if' block just to avoid a circular
> dependency.
>
> Or did you mean the imx driver is the one that should
> 'select DRM_DISPLAY_HELPER'? That would also work, but
> adds a bit more complexity.
Yes, I proposed this solution. I think this is what other drivers do.
--
With best wishes
Dmitry
