Bug 55161

Using the fn+brightness keys allows it to be set in acpi above 10 maxing at 15, but values over 10 do nothing more or less than 10

This may be useful info:

~$ uname -r
3.6.11-030611-generic
~$ xbacklight -set 0;acpi -c | grep LCD;xbacklight -get;acpi -c | grep LCD
Cooling 0: LCD 10 of 10
0.000000
Cooling 0: LCD 10 of 10
~$ xbacklight -set 100;acpi -c | grep LCD;xbacklight -get;acpi -c | grep LCD
Cooling 0: LCD 0 of 10
100.000000
Cooling 0: LCD 0 of 10
~$ xbacklight -set 50;acpi -c | grep LCD;xbacklight -get;acpi -c | grep LCD
Cooling 0: LCD 5 of 10
50.000000
Cooling 0: LCD 0 of 10
~$ ls /sys/class/backlight/*
/sys/class/backlight/acpi_video0:
actual_brightness  brightness  max_brightness  subsystem  uevent
bl_power           device      power           type

/sys/class/backlight/intel_backlight:
actual_brightness  brightness  max_brightness  subsystem  uevent
bl_power           device      power           type



~$ uname -r
3.9.0-030900rc2-generic
~$ xbacklight -set 0;acpi -c | grep LCD;xbacklight -get;acpi -c | grep LCD
Cooling 0: LCD 0 of 100
100.000000
Cooling 0: LCD 10 of 100
~$ xbacklight -set 100;acpi -c | grep LCD;xbacklight -get;acpi -c | grep LCD
Cooling 0: LCD 10 of 100
90.000000
Cooling 0: LCD 9 of 100
~$ xbacklight -set 50;acpi -c | grep LCD;xbacklight -get;acpi -c | grep LCD
Cooling 0: LCD 5 of 100
95.000000
Cooling 0: LCD 10 of 100
~$ ls /sys/class/backlight/*
/sys/class/backlight/acpi_video0:
actual_brightness  brightness  max_brightness  subsystem  uevent
bl_power           device      power           type

/sys/class/backlight/intel_backlight:
actual_brightness  brightness  max_brightness  subsystem  uevent
bl_power           device      power           type

I am not sure if this is a ACPI issue for a issue with the way the driver interacts with the ACPI, so the Product field may be wrong

Backlight still seems to be a horrible hit&miss issue :(

So atm I can only throw a few patches at you, hoping that some of them stick. Obviously retesting with latest upstream (3.9-rc or drm-intel-nightly) would be good, ubuntu should have ppas for those.

If that doesn't fix things, please try the following two patches (on top of -nightly):

https://patchwork.kernel.org/patch/2254211/
https://patchwork.kernel.org/patch/2254221/

would the libdrm packages in this ppa be what i want?
https://launchpad.net/~xorg-edgers/+archive/ppa?field.series_filter=raring
i have tried 3.9 RC3 on quantal 12.10

will those patches be in 3.9 RC4 (ubuntu mainline)
i have never had much luck with compiling
as i said i can setup a ssh connection for you and you can try anything you think will work (email me directly if you want to take me up on that offer)

in the mean time i am going to put my testing hdd in this system and try the nightly ppa on it with 3.9 rc3 and report back here

raring with kernel 3.9 rc3 and xorg edgers ppa

test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness 
91
976
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD;xbacklight -get;
91
976
Cooling 0: LCD 9 of 100
91.000000
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD;xbacklight -get;
91
976
Cooling 0: LCD 10 of 100
90.000000
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD;xbacklight -get;cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD;
90
976
Cooling 0: LCD 9 of 100
91.000000
91
976
Cooling 0: LCD 10 of 100
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD;xbacklight -set 5;cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD;
91
976
Cooling 0: LCD 10 of 100
5
976
Cooling 0: LCD 1 of 100
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD;xbacklight -set 0;cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD;
5
976
Cooling 0: LCD 1 of 100
0
976
Cooling 0: LCD 0 of 100
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/max_brightness
100
4882
test@A54C-NB91:~$ uname -r
3.9.0-030900rc3-generic

Oh I see what you meant, still reproducible, how to I apply the patches?
http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-intel-nightly/current/

test@A54C-NB91:~$ uname -r
3.9.0-994-generic
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/max_brightness;acpi -c | grep LCD
100
4882
Cooling 0: LCD 10 of 100
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness
91
976
test@A54C-NB91:~$ xbacklight -set 10
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness
10
976
test@A54C-NB91:~$ acpi -c | grep LCD
Cooling 0: LCD 1 of 100
test@A54C-NB91:~$ xbacklight -get
99.000000
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD
99
976
Cooling 0: LCD 10 of 100
test@A54C-NB91:~$ echo "brightness keys to max"
brightness keys to max
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD
99
976
Cooling 0: LCD 15 of 100
test@A54C-NB91:~$ echo "brightness keys to min"
brightness keys to min
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD
99
976
Cooling 0: LCD 0 of 100
test@A54C-NB91:~$ xbacklight -set 100
test@A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/brightness;acpi -c | grep LCD
100
976
Cooling 0: LCD 10 of 100
test@A54C-NB91:~$

the patch had no effect :(

Ok, i figure out how to compile it (and got a lot of dependences for compiling)

here are the command i ran:

wget https://www.kernel.org/pub/linux/kernel/v3.x/testing/linux-3.9-rc3.tar.xz
tar -xvf ./linux-3.9-rc3.tar.xz
cd linux-3.9-rc3
patch -p1 < ../kernel-patch1 
patch -p1 < ../kernel-patch2
cp /boot/config-3.9.0-030900rc3-generic .config
make oldconfig
make clean
make -j `getconf _NPROCESSORS_ONLN` deb-pkg LOCALVERSION=-custom
sudo dpkg -i ../linux-*.deb

bug still present in 
http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-intel-nightly/2013-03-23-raring/
also present in 3.9.0 rc4

(In reply to comment #7)
> the patch had no effect :(

FWIW the patches shouldn't have much effect on how your backlight actually works. They should merely report more consistent results from the sysfs files.

Note that you're cating the cached "brightness" value. You might get more interesting information by including the "actual_brightness" sysfs file too, which is supposed to be the value queried from the hardware. (Which it currently isn't in intel_backlight, but the second patch suggested by Daniel fixes that.)

this helpful?
chad@ASUS-A54C-NB91:~$ uname -r 
3.9.0-030900rc4-generic
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/max_brightness;acpi -c |grep LCD
100
4882
Cooling 0: LCD 10 of 100
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c |grep LCD
90
100
976
976
Cooling 0: LCD 10 of 100
chad@ASUS-A54C-NB91:~$ xbacklight -set 10
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c |grep LCD
99
10
976
976
Cooling 0: LCD 1 of 100
chad@ASUS-A54C-NB91:~$ acpi -c | grep LCD
Cooling 0: LCD 1 of 100
chad@ASUS-A54C-NB91:~$ echo "brightness keys to max"
brightness keys to max
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c |grep LCD
85
10
976
976
Cooling 0: LCD 15 of 100
chad@ASUS-A54C-NB91:~$ echo "brightness keys to min"
brightness keys to min
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c |grep LCD
100
10
976
976
Cooling 0: LCD 0 of 100
chad@ASUS-A54C-NB91:~$ xbacklight -set 100
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c |grep LCD
90
100
976
976
Cooling 0: LCD 10 of 100
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c |grep LCD;xbacklight -get;cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c |grep LCD
90
100
976
976
Cooling 0: LCD 10 of 100
90.000000
91
90
976
976
Cooling 0: LCD 9 of 100

(In reply to comment #10)
> this helpful?

Is this with or without the patches from comment #3?

did those patches make it into the 3.9 rc 4 mainline kernel?
*goes to compile mainline kernel from kernel.org with patches*

(In reply to comment #12)
> did those patches make it into the 3.9 rc 4 mainline kernel?

No, they won't make it to 3.9. They're not in any of the intel drm trees either.

Ran on patched copy of 3.9 rc4 from kernel.org

chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
100
976
chad@ASUS-A54C-NB91:~$ xbacklight -set 0
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
100
976
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
100
976
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
85
4882
chad@ASUS-A54C-NB91:~$ clear

chad@ASUS-A54C-NB91:~$ echo brightness to max via keys
brightness to max via keys
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
85
4882
chad@ASUS-A54C-NB91:~$ echo brightness to min via keys
brightness to min via keys
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
100
976
chad@ASUS-A54C-NB91:~$ xbacklight -set 0
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
100
976
chad@ASUS-A54C-NB91:~$ xbacklight -set 100
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
90
4882
chad@ASUS-A54C-NB91:~$ xbacklight -get
90.000000
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
91
3924
chad@ASUS-A54C-NB91:~$ xbacklight -get
91.000000
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
90
4882
chad@ASUS-A54C-NB91:~$ xbacklight -set 50
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
95
2201
chad@ASUS-A54C-NB91:~$ xbacklight -get
95.000000
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/*/actual_brightness
90
4882
-------------------------------------------------------------------
chad@ASUS-A54C-NB91:~$ uname -r
3.9.0-rc4-custom
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/max_brightness;acpi -c |grep LCD
100
4882
Cooling 0: LCD 10 of 100
chad@ASUS-A54C-NB91:~$ ls /sys/class/backlight/*
/sys/class/backlight/acpi_video0:
actual_brightness  brightness  max_brightness  subsystem  uevent
bl_power           device      power           type

/sys/class/backlight/intel_backlight:
actual_brightness  brightness  max_brightness  subsystem  uevent
bl_power           device      power           type
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c | grep LCD
90
100
4882
976
Cooling 0: LCD 10 of 100
chad@ASUS-A54C-NB91:~$ xbacklight -set 10
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c | grep LCD
99
10
1225
976
Cooling 0: LCD 1 of 100
chad@ASUS-A54C-NB91:~$ echo "brightness keys to max"
brightness keys to max
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c | grep LCD
85
10
4882
976
Cooling 0: LCD 15 of 100
chad@ASUS-A54C-NB91:~$ echo "brightness keys to min"
brightness keys to min
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c | grep LCD
100
10
976
976
Cooling 0: LCD 0 of 100
chad@ASUS-A54C-NB91:~$ xbacklight -set 100
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c | grep LCD
90
100
4882
976
Cooling 0: LCD 10 of 100
chad@ASUS-A54C-NB91:~$ xbacklight -get 
90.000000
chad@ASUS-A54C-NB91:~$ cat /sys/class/backlight/{acpi_video0,intel_backlight}/{actual_,}brightness;acpi -c | grep LCD
91
90
3924
976
Cooling 0: LCD 9 of 100

Just a hunch, please try acpi_osi="!Windows 2012" on the kernel cmdline, or try reverting:

commit a57f7f9175b8ccbc9df83ac13860488913115de4
Author: Bob Moore <robert.moore@intel.com>
Date:   Fri Aug 17 10:55:02 2012 +0800

    ACPICA: Add Windows8/Server2012 string for _OSI method.

and report back.

Also, providing a dmesg with drm.debug=0xe might prove useful when tracking down what goes on with the backlight changes.

Created attachment 97861 [details]
DMESG output

and crap i just notied i made a typo in the cmdline, now to redo all of that
at least that means there is still a chance that will work :)

Created attachment 97871 [details]
Requested DMESG output (fixed)

AWESOME IT WORKED!!!!!!
there is a minor issue, but it will not stop me from using kernels > 3.6.11

The only issue is the Fn keys allow the brightness to go 4 steps over the maximum, there are 11 brightness levels (0-10) and 16 levels
there are 5 levels that are 100%

(In reply to comment #19)
> AWESOME IT WORKED!!!!!!
> there is a minor issue, but it will not stop me from using kernels > 3.6.11

Ooookay, in a strange way that sort of was not the answer I was hoping for, because this means taming some BIOS dragons. But thanks for reporting back, and at least you have a workaround now. :)

> The only issue is the Fn keys allow the brightness to go 4 steps over the
> maximum, there are 11 brightness levels (0-10) and 16 levels
> there are 5 levels that are 100%

Now I have a hunch about that too. Please mount the debugfs and attach the output of 'hexdump < /sys/kernel/debug/dri/0/i915_opregion' (assuming that's where the debugfs is mounted, etc.).

(Note to developers: I presume both issues are related to our ACPI opregion handling, and changes in the way BIOS caters for Windows 8.)

LKML reference to another backlight problem bisected to the commit mentioned in comment #15: http://article.gmane.org/gmane.linux.kernel/1443070

Created attachment 97981 [details]
requested file (hex data)

i could have told of it would be close to the BIOS level
the boot parameter was not needed on the 208 BIOS version
with that parameter is is no longer a regression
this laptop was sold with windows 7, we should keep that in mind when assuming windows 8 behavior
if you are interested is solving the remaining issues with the back-light i am up for it
these issues were present, in the 1st place (note sure about the fn key over limit bug)

Note: that file was generated under the 3.8.0-17-generic stock Ubuntu kernel
I've been using a testing hdd to test my quick setup script for raring

There are some new backlight ACPI related fixes in drm-intel-nightly branch of git://people.freedesktop.org/~danvet/drm-intel. Please see if those will fix the issue _without_ acpi_osi="!Windows 2012".

Do i compile this the same way i did in post 7?
can i get it pre compiled here?
http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-intel-nightly/current/

ok i compiled it (~47.6 min) and gave it a try like i did in post 7
i hope i did not have to use git checkout or something, i just cloned it.
it compile a rc5 kernel, i tested it and the 3.9 mainline kernel from the ubuntu mainline

bug is present in both, it is rather annoying switching out my disk drives to test this, so if i compiled that wrong, tell me how to compile it next time

(In reply to comment #25)
> Do i compile this the same way i did in post 7?
> can i get it pre compiled here?
> http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-intel-nightly/current/

At a glance, that *seems* to contain all the commits I was referring to. I don't know who maintains it, though, so I can't really endorse it in any way.

(In reply to comment #26)
> ok i compiled it (~47.6 min) and gave it a try like i did in post 7
> i hope i did not have to use git checkout or something, i just cloned it.
> it compile a rc5 kernel, i tested it and the 3.9 mainline kernel from the
> ubuntu mainline
> 
> bug is present in both, it is rather annoying switching out my disk drives to
> test this, so if i compiled that wrong, tell me how to compile it next time

Unfortunately, that does not give you the drm-intel-nightly branch. You'll need to go to your cloned tree, and checkout the drm-intel-nightly branch.

the copy in here today (may 5th)
http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-intel-nightly/current/
bug is present, works fine with the acpi_osi='!Windows 2012'

What commit number do i checkout? how would i even get these numbers?

I presume this will get fixed when [1] gets merged.

[1] http://mid.gmane.org/1381214401-24672-1-git-send-email-aaron.lu@intel.com

Please try kernel v3.13.

without the acpi_osi='!Windows 2012' parameter?
I have been using the 3.13 kernel since i believe rc3 was out and i have not had any issues
though i have been using the acpi_osi='!Windows 2012' parameter in grub

on a side note on my amd desktop with the 3.12.9 kernel if i set the cpu governor to userspace and select a frequency when i wake from suspend the frequency changes to 100%
should be reproducible on intel system that are not using the new intel_pstate cpu frequency manager
This is the script i use to manage the governor
https://github.com/GM-Script-Writer-62850/xfce-genmon-scripts/blob/master/cpu-freq-plugin

(In reply to pqwoerituytrueiwoq from comment #32)
> without the acpi_osi='!Windows 2012' parameter?
> I have been using the 3.13 kernel since i believe rc3 was out and i have not
> had any issues
> though i have been using the acpi_osi='!Windows 2012' parameter in grub

Would be nice to check without that acpi_osi setting too.

> on a side note...

Off-topic for this bug. :)

I am not 100% sure if it is working perfectly but even without the acpi_osi='!Windows 2012' parameter, the backlight seems to work much better than it did when i made the initial report
i can not tell a difference with or without the parameter, so that is a definite improvement
i would not consider the current state a regression, in its current state it is manageable, the fn brighness changes do not show up in software so if you change that during boot when the user logs in the brightness reset to whet ever it was set to via software or 100% if it was never set via software like xbacklight
now lets have some fun with 3.14rc1 :)

with the current 1.13.1 and 3.14rc1 i cant see a difference with or without the added parameter

when using the fn keys there are a few levels over 100% that are all actually 100%, what i mean by that is i hit fn+F6 to up it 20 times when i hit fn+f5 to bring it down the 1st few times don't do anything
as long as xbacklight or my backlightx script works that is good enough for me

Thanks, closing. Please file a new bug if you encounter any issues. (This one was originally filed against a patch set, so reopening it might be confusing.)

(In reply to Jani Nikula from comment #35)
> Please file a new bug if you encounter any issues. (This
> one was originally filed against a patch set, so reopening it might be
> confusing.)

Disregard that, it's just me mixing this up with another bug. Sorry for the noise.