[Intel-gfx] [REGRESSION BISECTED] backlight control stops workin with 3.14 and later

[Jani Nikula]

On Sat, 23 Aug 2014, Hans de Goede <hdegoede at redhat.com> wrote:
> Hi,
>
> On 08/22/2014 08:48 AM, Jani Nikula wrote:
>> On Fri, 22 Aug 2014, Bertrik Sikken <bertrik at sikken.nl> wrote:
>>> On 19-8-2014 3:29, Jani Nikula wrote:
>>>> On Tue, 19 Aug 2014, Bertrik Sikken <bertrik at sikken.nl> wrote:
>>>>>> On Sun, 17 Aug 2014, Bertrik Sikken <bertrik at sikken.nl> wrote:
>>>>>>> On 15-8-2014 3:43, Jani Nikula wrote:
>>>>>>>> On Thu, 14 Aug 2014, Bertrik Sikken <bertrik at sikken.nl> wrote:
>>>>>>>
>>>>>>>>> Attached is dmesg output from booting kernel 3.14-2 (debian unstable)
>>>>>>>>> with drm.debug=0xe and the samsung_laptop module enabled, from my
>>>>>>>>> Samsung N150plus netbook.
>>>>>>>>
>>>>>>>> Have you tried 3.15?
>>>>>>>
>>>>>>> I've built the v3.15 kernel (using the .config file from debian
>>>>>>> unstable and doing make oldconfig).
>>>>>>>
>>>>>>> The backlight is at maximum brightness after boot and I can't control
>>>>>>> it using the backlight buttons, nor by writing to
>>>>>>> /sys/class/backlight/samsung/brightness
>>>>>>> (say half the value or 1/10th of max_brightness)
>>>>>>>
>>>>>>> Backlight does work when writing
>>>>>>> /sys/class/backlight/intel_backlight/brightness
>>>>>>
>>>>>> How about disabling samsung backlight module with 3.15?
>>>>>
>>>>> I'm not sure what you mean by that.
>>>>>
>>>>> As I understand it, there are three ways to control the backlight on this
>>>>> netbook: using intel_backlight, samsung_laptop (using a "sabi" interface)
>>>>> and acpi_video.
>>>>> Backlight control using the samsung_laptop driver no longer seems to work
>>>>> after the change. If I disable it (e.g. by blacklisting it), I expect it
>>>>> to no longer work at all obviously.
>>>>>
>>>>> What do you want me to test exactly?
>>>>
>>>> If the intel_backlight interface works in 3.15, I presume the problem is
>>>> that you have a non-functional samsung backlight interface that is
>>>> preferred over intel_backlight by your userspace.
>>>
>>> I tested the intel_backlight interface in linux v3.15 with the samsung
>>> backlight module blacklisted. The intel_backlight interface still works.
>> 
>> I read that as, "I no longer have problems with backlight".
>
> No, just NO. This thread has "REGRESSION" in the subject for a reason,
> userspace will prefer a firmware backlight interface over a raw backlight
> interface, so the only way to get userspace to use the intel backlight
> driver is to blacklist the samsung_laptop and acpi_video (Bertrik can you
> confirm that acpi_video also is broken?) so that the intel one is the only
> remaining.

Fair enough.

I'm going to need a bunch of attached files that will be awkward on the
mailing list, so please file a new bug against DRM/Intel at [1]. Refer
this mail thread, and attach the output of intel_reg_dumper (from
intel-gpu-tools [2]) on a working and a non-working kernel. Also attach
/sys/kernel/debug/dri/0/i915_opregion

On the broken kernel, does the samsung backlight interface work *before*
you've loaded i915.ko?

Thanks,
Jani.


[1] https://bugs.freedesktop.org/enter_bug.cgi?product=DRI
[2] http://cgit.freedesktop.org/xorg/app/intel-gpu-tools/

-- 
Jani Nikula, Intel Open Source Technology Center