[PATCH] drm/armada: Fix compile fail

[Daniel Vetter]

On Mon, Jan 2, 2017 at 3:20 PM, Russell King - ARM Linux
<linux at armlinux.org.uk> wrote:
> On Fri, Dec 30, 2016 at 05:33:38PM +0100, Daniel Vetter wrote:
>> I reported the include issue for tracepoints a while ago, but nothing
>> seems to have happened. Now it bit us, since the drm_mm_print
>> conversion was broken for armada. Fix them both.
>
> Nothing's happened because I haven't had an opportunity to investigate
> yet.  It's not as simple as you think...

Yeah, I don't get it either tbh.

> This has been tested over several kernel versions before submission,
> and it's also been sitting in my build tree for a while before sending
> upstream.  At no point have I seen the failure you are reporting.
>
> The tracepoints have been in use on 4.7 and 4.8 kernels, and it builds
> fine there without any changes.  It's been in my nightly builder, which
> has found no problems with it, eg, 4.9 allmodconfig:
>
>   CC [M]  drivers/gpu/drm/armada/armada_crtc.o
>   CC [M]  drivers/gpu/drm/armada/armada_drv.o
>   CC [M]  drivers/gpu/drm/armada/armada_fb.o
>   CC [M]  drivers/gpu/drm/armada/armada_fbdev.o
>   CC [M]  drivers/gpu/drm/armada/armada_gem.o
>   CC [M]  drivers/gpu/drm/armada/armada_overlay.o
>   CC [M]  drivers/gpu/drm/armada/armada_trace.o
>   CC [M]  drivers/gpu/drm/armada/armada_510.o
>   CC [M]  drivers/gpu/drm/armada/armada_debugfs.o
>   LD [M]  drivers/gpu/drm/armada/armada.o
>
> So the question remains: why do you see it, and I don't - there's
> something different in our build environments that's triggering it.
> I always build with a separated object tree - maybe you build in the
> same tree.  Does the problem disappear if you build using a separated
> object tree (O=/path/to/object/tree) ?

Yup. That also seems to be the setup 0day uses. I've also tried
disabling tracepoints, that didn't help either. I haven't checked
whether built-in or module changes anything, but my config builds
armada as a module.

> However, given the contents of your patch, it seems only relevant for
> separated object tree builds.
>
> So... confused.
>
> I'll investigate further once I've updated my cubox tree to 4.9.

We've had the same fun in i915 with tracepoints, and looking around
we're not the only ones opting for this solution. I'm not sure whether
tracepoints in modules is a bit the odd thing out, or whether
something in kbuild is amiss. I don't really grok it enough to figure
out what's really going on ... The hack gets things back on track at
least.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch