Re: [PATCH v3 0/2] drm/imx/lcdc: Implement DRM driver for imx21

[Rob Herring <robh@kernel.org>]

On Sat, Dec 17, 2022 at 07:38:06PM +0100, Uwe Kleine-König wrote:
> On Fri, Dec 16, 2022 at 05:57:58PM -0600, Rob Herring wrote:
> > On Fri, Dec 16, 2022 at 06:50:04PM +0100, Uwe Kleine-König wrote:
> > > Hello,
> > > 
> > > Changes since v2:
> > > 
> > >  - added allOf as Krzysztof requested
> > >  - reworked driver based on Philipp's comments
> > >    (improved error handling, different selects, moved driver to a subdirectory,
> > >    header sorting, drm_err instead of DRM_ERROR, inlined
> > >    imx_lcdc_check_mode_change, make use of dev_err_probe())
> > >  
> > > Krzysztof also pointed out that we're now having two compatibles for a
> > > single hardware. Admittedly this is unusual, but this is the chance that
> > > the (bad) compatible identifier imx21-fb gets deprecated. The hardware
> > > is called LCDC and only the linux (framebuffer) driver is called imxfb.
> > 
> > The problem is you can't have firmware (with the DTB) that supports 
> > both. Well, you can if you want to have some firmware setting that 
> > selects which one. Otherwise, it's really an OS problem to decide what 
> > to use. 
> 
> I don't understand what you intend to say here. The same applies if the
> compatible is the same for both binding alternatives, isn't it? 

Only if you have both nodes in the DT and both enabled. But 2 enabled 
nodes at the same address is also a dtc warning, so I was assuming you 
didn't do that.

> Do you consider a firmware problem better or an OS problem?

The OS created the problem, so they get to keep it. But a PC BIOS is 
full of OS compatibility switches, so...

In the end, it's the platforms' decision really. I just want what the 
implications of having 2 compatibles are to be understood.

Rob