Bug 81438

Created attachment 102938 [details]
dmesg

Please try reverting:
commit c675949ec58ca50d5a3ae3c757892f1560f6e896
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Wed Apr 9 11:31:37 2014 +0300

    drm/i915: do not setup backlight if not available according to VBT

and see if it helps.

Please attach /sys/kernel/debug/dri/0/i915_opregion

Please attach dmesg with drm.debug=0xe module parameter.

Side note, the attached dmesg has a gpu hang in it. You could file a separate bug on that.

Hi,

I don't know if it can help, but running the same laptop (MacBook 2,1), I'm having the same issue with kernel 3.16-1-686-pae (Debian version 3.16.2-3).

Can it be the cause of my problems ? :
  - "xbacklight -dec 10" returns "No outputs have backlight property"
  - backlight does not power on after a resume

I also have the "GPU HANG" in my kernel.log : I just filled a bug report at https://bugs.freedesktop.org/show_bug.cgi?id=83851

Thanks,
YvanM

Hi,

After reverting the commit c675949ec58ca50d5a3ae3c757892f1560f6e896 on kernel 3.15.5 from kernel.org, boot fails on kernel panic with this message (I don't know how to catch the kernel panic log, so I juste write the lines I think interesting here) :

Code: .......
EIP: [<f878d93b>] __intel_set_cpu_fifo_underrun_reporting+0x2b/0x2d0 [i915] SS:ESP 0068:f5407ef0
CR2: 000000000000088c
---[ end trace ... ]---
Kernel panic - not syncing: Fatal exception in interrupt
Kernel Offset: 0x0 from 0xc1000000 (relocation range: 0xc0000000-0xf81fdfff)
...

However, it also happens rarely an other kernel panic, but with no "intelligible" message.

It makes me think of a similar message I have at boot with kernels 3.15.5 from kernel.org and 3.16.1 from Debian :
[drm:i915_irq_handler] *ERROR* pipe A underrun
[drm:i915_irq_handler] *ERROR* pipe A underrun
[drm] GPU HANG: ecode -1:0x00000000, reason: Command parse error, iir 0x00008000, action: continue

I attach my kern.log : it includes boots with kernels 3.15.5 and 3.16.1

Created attachment 106500 [details]
kern.log using 3.15.5 and 3.16.1

reverting c675949ec58 on a checkout of the v3.15 tag of the kernel.org source from git fixes the issue. intel_backlight is back in /sys/class/backlight and xbacklight works again.
/sys/kernel/debug/dri/0/i915_opregion is empty. i don't know if that is expected or if i'm supposed to do something to have something logged there. debugfs is mounted on /sys/kernel/debug.
i'm attaching a dmesg with the drm.debug=0xe parameter.

Created attachment 108477 [details]
dmesg (drm.debug=0xe)

Presumed fixed in v3.16 by
commit 56c4b63aaf4c2cd91966b2a5e69e5367bea60bbe
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Tue Jun 17 15:47:05 2014 +0300

    drm/i915: default to having backlight if VBT not available

Please try v3.16, as v3.15 is no longer supported, and the fix won't be backported.

i was actually already aware of that commit. unfortunately it doesn't seem to work on the 2,1 macbook. it also seems to have been backported to 3.15 in the fedora kernel: https://bugzilla.redhat.com/show_bug.cgi?id=1118969
i haven't tested 3.16 specifically but 3.17 still has the problem.
i looked around and some other devices with the same problem seem to have quirks applied. i'm attaching a patch which applies the 'quirk_backlight_present' for the 2,1 macbook. i hope this is the proper way to do it.

Created attachment 108528 [details]
patch to apply 'quirk_backlight_present' quirk

(In reply to jens stein from comment #10)
> Created attachment 108528 [details]
> patch to apply 'quirk_backlight_present' quirk

Please send this to intel-gfx mailing list with proper Signed-off-by. Thanks.

Fix pushed to drm-intel-fixes as
commit b2a9601c587dbc5536546aa54009d1130adedd72
Author: jens stein <jens.s.stein@gmail.com>
Date:   Tue Oct 28 20:25:53 2014 +0100

    drm/i915: Ignore VBT backlight check on Macbook 2, 1

Thanks for the report, and the fix! :)