Build error: MSM DRM

[Rob Clark]

On Mon, Feb 24, 2014 at 6:15 AM, Russell King - ARM Linux
<linux at arm.linux.org.uk> wrote:
> Automated build testing of allmodconfig discovered this error in the
> MSM DRM driver:
>
> drivers/gpu/drm/msm/hdmi/hdmi.o:(.rodata+0x10): multiple definition of `__mod_of_device_table'
> drivers/gpu/drm/msm/adreno/a3xx_gpu.o:(.rodata+0x4dc): first defined here
>
> It appears that this will happen whenever the MSM DRM driver is built
> as a module.  It's unclear what the correct fix is - since making it a
> "bool" will make it unusable if the rest of DRM is configured modular,
> and it's not trivial to split these files.

fwiw, I was hoping https://lkml.org/lkml/2014/1/27/340 would get
merged, as that should also fix this build error (same in tilcdc)..
but making it a bool instead of tristate is probably ok for a short
term solution.

> Since this is yet another multiple-device monstrosity pretending to look
> like a single device (though slightly saner than other stuff, but still
> ultimately unsafe - nothing guarantees for instance that hdmi_pdev
> remains valid) the right solution to this would be to convert the thing
> to use the component infrastructure included in v3.14-rc, which then
> allows the MSM driver to be multiple stand alone modules.

haven't had a chance to look at the component infrastructure stuff
yet, but this was my plan.  It is the better long term solution.

BR,
-R

> This is clearly what the driver needs if it's actually made up from
> multiple drivers.
>
> While there's no examples of how to do this in mainline yet, two examples
> of such converted drivers have been posted on LAKML and elsewhere - which
> are the imx-drm and armada-drm drivers.
>
> --
> FTTC broadband for 0.8mile line: now at 9.7Mbps down 460kbps up... slowly
> improving, and getting towards what was expected from it.
> _______________________________________________
> dri-devel mailing list
> dri-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/dri-devel