Re: [PATCH 2/3] drm/vc4: Don't try to initialize FBDEV if we're only bound to V3D.

[Daniel Vetter]

On Mon, Apr 24, 2017 at 10:26:45AM -0400, Alex Deucher wrote:
> On Fri, Apr 21, 2017 at 6:53 PM, Eric Anholt <eric@xxxxxxxxxx> wrote:
> > Daniel Vetter <daniel@xxxxxxxx> writes:
> >
> >> On Wed, Apr 19, 2017 at 7:55 PM, Eric Anholt <eric@xxxxxxxxxx> wrote:
> >>> Daniel Vetter <daniel@xxxxxxxx> writes:
> >>>> On Tue, Apr 18, 2017 at 9:11 PM, Eric Anholt <eric@xxxxxxxxxx> wrote:
> >>>>> The FBDEV initialization would throw an error in dmesg, when we just
> >>>>> want to silently not initialize fbdev on a V3D-only VC4 instance.
> >>>>>
> >>>>> Signed-off-by: Eric Anholt <eric@xxxxxxxxxx>
> >>>>
> >>>> Hm, this shouldn't be an error really, you might want to hotplug more
> >>>> connectors later on. What exactly complains?
> >>>
> >>> drm_fb_helper_init() throws an error if the passed in connector count is
> >>> 0, so drm_fb_cma_helper() printks an error.
> >>
> >> Oh, _that_ thing. The error in there is correct, but (almost) everyone
> >> gets this parameter wrong. This isn't the max number of connectors the
> >> fb helper will light up, but just the max number of connectors _per_
> >> crtc when driving in hw clone mode. There's two problems with that:
> >> - fb helpers don't support hw clone mode, we select 1:1 crtcs for each
> >> active connector
> >> - I mentioned that everyone gets this wrong?
> >>
> >> If you're moderately bored it'd be great to nuke the max_connector
> >> argument from drm_fb_helper_init, and hard-code it to 1 (with a big
> >> comment explaining that this needs to be changed, probably with
> >> dynamic reallocation, once someone gets around to implementing hw
> >> clone mode).
> >>
> >> If you're less bored, just hardcode this to 1 in vc4 and done. Plus a
> >> TODO.rst entry would be great in that case.
> >
> > If I'm driving a GPU with no display subsystem at all, it seems like I
> > shouldn't initialize fbdev for it, right?
> 
> That's what we do for radeon/amdgpu on hw without display blocks.

Yeah I got slightly distracted with the error :-)

Still might make sense to put this logic into the helpers, if there's no
crtc then don't bother initializing the fbcon. But imo the better cleanup
is getting rid of the connectors parameter, since most drivers get it
wrong, then respin the vc4 patch to explain what it's really doing (it's
not about the error really, it's about not registering and fbdev that's
not doing anything).
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch