Re: [PATCH v2 13/27] drm/tegra: Move subdevice infrastructure to host1x

[Thierry Reding <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

[-- Attachment #1: Type: text/plain, Size: 1622 bytes --]

On Mon, Nov 04, 2013 at 11:20:55AM +0100, Daniel Vetter wrote:
> On Mon, Oct 07, 2013 at 10:34:30AM +0200, Thierry Reding wrote:
> > +static struct drm_bus drm_host1x_bus = {
> > +	.bus_type = DRIVER_BUS_HOST1X,
> > +	.set_busid = drm_host1x_set_busid,
> > +};
> 
> Imo this needs to die asap, i.e. before it lands in drm-next. I've just
> spent a bit of time trying to rid ourselves of this midlayer
> brain-damange, and now new stuff crops up. See
> 
> http://cgit.freedesktop.org/~danvet/drm/log/?h=drm-init-cleanup
> 
> drm_bus is a terribly midlayer madness disaster and needs to just burn
> down. So what exactly is the reason here to not go with the drm_platform.c
> stuff until I've gotten around to completely rip it all out?

Dave asked me whether it was possible to move the Tegra DRM driver back
into drivers/gpu/drm (from drivers/gpu/host1x/drm) and introducing this
new bus was the only solution I saw (besides perhaps using cross-
subsystem global variables) to do that.

Now, I'm not a big fan at all of the whole drm_bus shebang myself, so
whatever help you need in getting rid of it, please let me know. The
drm_bus implementation that this adds is about 34 lines, so it shouldn't
be difficult to get rid of. If you throw out drm_driver.bus completely,
then it should be a matter of just deleting that code and the rest
should be able to continue working as is. The only reason I added it is
because the kernel crashes if its not there, depending on what userspace
is run.

So I find myself in the middle of controversy again... I seem to have a
knack for it.

Thierry

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]