[Intel-gfx] [Regression] BUG: unable to handle kernel NULL pointer dereference

[Chris Wilson]

On Sun, Mar 01, 2015 at 12:30:01PM +0300, Andrey Skvortsov wrote:
> On 01 Mar, Chris Wilson wrote:
> > On Sun, Mar 01, 2015 at 12:57:01AM +0300, Andrey Skvortsov wrote:
> > > On 28 Feb, Chris Wilson wrote:
> > > > On Sat, Feb 28, 2015 at 03:20:37PM +0300, Andrey Skvortsov wrote:
> > > > > Unfortunately this is not the last bug, that breaks i915/drm working
> > > > > on my laptop. Sometimes system successfully loads with couple warnings mentioned in
> > > > > previous mail:
> > > > > 
> > > > > [   26.922953] WARNING: CPU: 1 PID: 767 at drivers/gpu/drm/i915/i915_gem.c:4525 i915_gem_free_object+0x13f/0x288 [i915]()
> > > > > [   26.922954] WARN_ON(obj->frontbuffer_bits)
> > > > 
> > > > That's inocuous, but for the serious hang, you may want to try
> > > > video=SVIDEO-1:d on the kernel commandline to workaround the hang. (Check
> > > > /sys/class/drm/card/ for the actual name of the connector for TV). I
> > > > think Ville mentioned he was looking/looked at the atomic-vs-load_detect
> > > > changes that is at the heart of the issue here. (Admittedly he did say
> > > > it was worthy of a drink or two.)
> > > > -Chris
> > > 
> > > Thank you for the help. I've tried to add to the kernel command line
> > > 'video=card0-SVIDEO-1:d video=card0-VGA-1:d'. But despite of that
> > > kernel crashes with 'BUG: unable to handle kernel NULL pointer
> > > dereference at' occasionally.
> > 
> > Remove the card0- and combine both into a single video= stanza:
> > video=SVIDEO-1:d,VGA-1:d
> 
> I've just tried that, but the crash with NULL pointer dereference still happens.

Drat. In hindsight it should have been obvious since the crash is from
the driver probing the hardware during init and not the runtime output
detection. :|
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre