Re: [Intel-gfx] [Regression] WARNING: drivers/gpu/drm/i915/i915_gem.c:4525 i915_gem_free_object

[Daniel Vetter <daniel@ffwll.ch>]

On Mon, Mar 16, 2015 at 01:21:58PM -0400, Steven Rostedt wrote:
> On Mon, 16 Mar 2015 17:43:33 +0100
> Daniel Vetter <daniel@ffwll.ch> wrote:
> 
> 
> > Is linux-next also affected or is this just a case of i915 maintainers
> > having failed to push the right patch to -fixes?
> 
> I just confirmed, linux-next boots without the warning.
> 
> > 
> > I'll try to figure out what fell apart here meanwhile. To help out, can
> > you please boot with drm.debug=0xe and grab the full dmesg for booting?
> > Looks like it falls over in the initial fbcon modeset, which is indeed
> > strange ...
> 
> Here they are:
> 
>  http://rostedt.homelinux.com/private/dmesg-next
> 
>  http://rostedt.homelinux.com/private/dmesg-4.0

The unnerving part is that I didn't find any patch not anything else that
might fix this frontbuffer_bits leak. I also couldn't spot the bug.

The real reason why -next works and 4.0 doesn't seems to be that 4.0
reconstructs a bogus initial framebuffer and then falls over when cleaning
that up. -next just doesn't reconstruct that framebuffer and hence also
can't fall over on cleanup.

Can you please cherry pick 42a7b088127f (\"drm/i915: Make sure the primary
plane is enabled before reading out the fb state\") from -next to 4.0-rc
to test whether this is indeed the difference in ducttape?

Thanks, Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch