FS#41580 - [linux] Backlight control failure on linux 3.16

Attached to Project: Arch Linux
Opened by Martin Roth (Captain_Rage) - Saturday, 16 August 2014, 22:39 GMT
Last edited by Doug Newgard (Scimmia) - Sunday, 01 October 2017, 15:48 GMT
Task Type Bug Report
Category Kernel
Status Closed
Assigned To Tobias Powalowski (tpowa)
Architecture x86_64
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:
Upon booting the follwing message appears:

[FAILED] Failed to start Load/Save Screen Backlight Brightness of backlight:acpi_video0
See 'systemctl status systemd-backlight@backlight:acpi_video0.service' for details.

Controlling the brightness via hotkeys doesn't work anymore. Related:
https://bugzilla.redhat.com/show_bug.cgi?id=1062638

Additional info:
* package version(s)
* config and/or log files etc.


Steps to reproduce:
Boot the system using systemd.
This task depends upon

Closed by  Doug Newgard (Scimmia)
Sunday, 01 October 2017, 15:48 GMT
Reason for closing:  Fixed
Comment by Dave Reisner (falconindy) - Sunday, 17 August 2014, 00:39 GMT
Please run as root:

strace /usr/lib/systemd/systemd-backlight load backlight:acpi_video0 &>strace.log

And attach the log file.

Doesn't seem at all related to the linked bug since the included work from Zbigniew is included as of v214.
Comment by Martin Roth (Captain_Rage) - Sunday, 17 August 2014, 19:38 GMT
strace run as root. See attached log.
Comment by Dave Reisner (falconindy) - Sunday, 17 August 2014, 20:39 GMT
You didn't run the command I asked for. You misspelled the name of the backlight.
Comment by Martin Roth (Captain_Rage) - Sunday, 17 August 2014, 20:57 GMT
My apologies for the typo. Now it should be correct. Please see the attached log.
Comment by Martin Roth (Captain_Rage) - Monday, 18 August 2014, 09:56 GMT
Interestingly, pressing Fn+F6 still works (the button that makes the screen go almost completely dark).
Comment by Dave Reisner (falconindy) - Wednesday, 20 August 2014, 01:12 GMT
Can you post the output from 'systemctl status systemd-backlight@backlight:acpi_video.service' ? It's not clear what's starting this service if the device doesn't exist...
Comment by Martin Roth (Captain_Rage) - Wednesday, 20 August 2014, 08:39 GMT
[martin@Wisp ~]$ systemctl status systemd-backlight@backlight:acpi_video.service
● systemd-backlight@backlight:acpi_video.service - Load/Save Screen Backlight Brightness of backlight:acpi_video
Loaded: loaded (/usr/lib/systemd/system/systemd-backlight@.service; static)
Active: inactive (dead)
Docs: man:systemd-backlight@.service(8)


The only thing I'm sure about is that it was working before some update and that I haven't fiddled with anything. I'm not sure what conclusion to make from this command, but hopefully it is helpful.
Comment by Martin Roth (Captain_Rage) - Wednesday, 20 August 2014, 14:11 GMT
Comment by brent saner (sanerb) - Friday, 22 August 2014, 21:20 GMT
I get incredibly similar behaviour- on my laptop, when it is plugged in, the backlight is up all the way (and cannot be dimmed; the display can only be turned off by turning the dimmer completely down as noted in M. Roth's 8.18.2014 comment).

When on battery, however, the backlight seems to not be on at all as it is dimmed extremely dark.

However, I did not see any failures from the systemd unit:

[root@workhorse ~]# systemctl status systemd-backlight@*
● systemd-backlight@backlight:intel_backlight.service - Load/Save Screen Backlight Brightness of backlight:intel_backlight
Loaded: loaded (/usr/lib/systemd/system/systemd-backlight@.service; static)
Active: active (exited) since Fri 2014-08-22 16:58:04 EDT; 10min ago
Docs: man:systemd-backlight@.service(8)
Process: 325 ExecStart=/usr/lib/systemd/systemd-backlight load %i (code=exited, status=0/SUCCESS)
Main PID: 325 (code=exited, status=0/SUCCESS)

Aug 22 16:58:04 workhorse systemd[1]: Started Load/Save Screen Backlight Brightness of backlight:intel_backlight.



Hardware:
[root@workhorse ~]# lspci | egrep -i 'vga|display'
00:02.0 VGA compatible controller: Intel Corporation 3rd Gen Core processor Graphics Controller (rev 09)
01:00.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Mars [Radeon HD 8670A/8670M/8750M] (rev ff)

No failed systemd services at all, actually..

I am attaching an strace.log in the event it helps, but being that the service unit starts successfully but the *behaviour* is b0rk3n, I'm not certain it will help.
It's also a possibility that it's failing (or just not enabled) for the Radeon "Display controller" but the VGA controller is working okay, and it never attempts to load the Radeon backlighting.

However, this is a clear regression- this was working just fine before an upgrade. List of package activity that could have caused this (as they were all installed/removed/upgraded between known states of working vs. non-working) is also attached.


Comment by Martin Roth (Captain_Rage) - Monday, 25 August 2014, 23:07 GMT
You might be right that this is not caused by systemd. The issue could have been caused by a kernal upgrade. My machine is also running the open source drivers for an AMD APU.
Comment by brent saner (sanerb) - Tuesday, 26 August 2014, 02:18 GMT
i suspect strongly that it might be something with ATI drivers or the kernel upgrade.

Martin, can you do the following for me and paste it?

for i in $(pacman -Qqs '(^|-)ati') linux ; do pacman -Qi ${i} | egrep '^Name|^Version|^Install Date' ; echo ; done

(if you're wary about pasting one-liners from strangers on the intarwebz, mine looks like this:

[bts@workhorse ~]$ for i in $(pacman -Qqs '(^|-)ati') linux ; do pacman -Qi ${i} | egrep '^Name|^Version|^Install Date' ; echo ; done
Name : ati-dri
Version : 10.2.6-1
Install Date : Thu 21 Aug 2014 09:15:38 PM EDT

Name : xf86-video-ati
Version : 1:7.4.0-3
Install Date : Thu 21 Aug 2014 09:21:22 PM EDT

Name : linux
Version : 3.16.1-1
Install Date : Thu 21 Aug 2014 09:19:25 PM EDT

)

All three packages were upgraded (for me) during the same upgrade run, and definitely falls into the range of time between reboots (as noted in the package list at https://bugs.archlinux.org/task/41580?getfile=12125 ) in which the issue came up. I'd like to see if you would have better luck noticing which package was the closest to about when you first noticed this issue.

What's stranger, to me, is that I don't see ANY backlight device for ATI/AMD devices (acpi_video0), only Intel (intel_backlight) despite me apparently having both (per my lspci output above).

If you can give the same lspci output I did (lspci | egrep -i 'vga|display'), I'd be curious to see that as well.


Dave- can you provide any input or perspective on what is/might be happening here? Should an upstream bug be filed?
Comment by Dave Reisner (falconindy) - Tuesday, 26 August 2014, 02:25 GMT
pacman. You wanted pacman. Never yaourt (especially not on our bug tracker).

> Dave- can you provide any input or perspective on what is/might be happening here? Should an upstream bug be filed?
Seems like something changed in the kernel. Not much for Arch to do. If you want to file a bug report, I'd suggest bisecting it to the broken commit first.
Comment by brent saner (sanerb) - Tuesday, 26 August 2014, 02:51 GMT
"pacman. You wanted pacman. Never yaourt (especially not on our bug tracker)."

CRIMINY, you're right. Sorry; that one must have slipped in there. Fixed.

"Seems like something changed in the kernel. Not much for Arch to do. If you want to file a bug report, I'd suggest bisecting it to the broken commit first."

Yeah, I'm going through the commits now between 15.5 and 16.1. I'll pop back in if I find anything.
Comment by Martin Roth (Captain_Rage) - Tuesday, 26 August 2014, 10:42 GMT
[martin@Wisp ~]$ for i in $(pacman -Qqs '(^|-)ati') linux ; do pacman -Qi ${i} | egrep '^Name|^Version|^Install Date' ; echo ; done
Name : ati-dri
Version : 10.2.6-1
Install Date : Fri 22 Aug 2014 02:34:28 AM CEST

Name : lib32-ati-dri
Version : 10.2.6-1
Install Date : Fri 22 Aug 2014 02:34:31 AM CEST

Name : xf86-video-ati
Version : 1:7.4.0-3
Install Date : Wed 06 Aug 2014 11:33:12 AM CEST

Name : linux
Version : 3.16.1-1
Install Date : Thu 14 Aug 2014 06:08:33 PM CEST


I haven't tried downgrading the kernel. Doing so might be an easy way rule out that it's a problem with another package.

Sounds great, Brent! Good luck and report if you find anything!
Comment by brent saner (sanerb) - Wednesday, 27 August 2014, 01:33 GMT
Martin-

If you can downgrade (note that it's not officially supported, BUT possible. Before doing so, please read https://wiki.archlinux.org/index.php/Downgrading_packages and https://wiki.archlinux.org/index.php/Arch_Rollback_Machine) while I look over the commits to the kernel, that'd be a HUGE help (as bisecting can be a somewhat tedious process).

If you're on x86_64, you can use this package: http://seblu.net/a/arm/2014/08/12/core/os/x86_64/linux-3.15.8-1-x86_64.pkg.tar.xz
If you're on i686, you can use this package: http://seblu.net/a/arm/2014/08/12/core/os/i686/linux-3.15.8-1-i686.pkg.tar.xz
Comment by brent saner (sanerb) - Wednesday, 27 August 2014, 01:34 GMT
(you may also need the associated linux-headers package of the same version at the same url)
Comment by brent saner (sanerb) - Wednesday, 27 August 2014, 03:45 GMT
scratch that- had some free time, it's definitely kernel-related. (sigh) back to commit reading and bisecting.
Comment by brent saner (sanerb) - Sunday, 07 September 2014, 15:29 GMT
Apologies for the delays- between traveling, work, etc. I had to squeeze in a kernel bisect.

Here's the result:

751109aad5834375ca9ed0dcfcd85a00cbf872b5 is the first bad commit
commit 751109aad5834375ca9ed0dcfcd85a00cbf872b5
Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Date: Thu Jun 5 22:47:35 2014 +0200

ACPI / video: Change the default for video.use_native_backlight to 1

Now that we're hoping to have resolved all of the problems with
video.use_native_backlight=1, make that the default at last.

Link: http://marc.info/?l=linux-acpi&m=139716088401106&w=2
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

:040000 040000 d8f9ae057c052cd94f1adcfc0b298db2c728a315 933b38080034b141fc0c31f54f38f5b232f69c17 M drivers




And here's the full log:

[bts@workhorse linux-git]$ git bisect log
git bisect start
# bad: [19583ca584d6f574384e17fe7613dfaeadcdc4a6] Linux 3.16
git bisect bad 19583ca584d6f574384e17fe7613dfaeadcdc4a6
# good: [fad01e866afdbe01a1f3ec06a39c3a8b9e197014] Linux 3.15-rc8
git bisect good fad01e866afdbe01a1f3ec06a39c3a8b9e197014
# good: [7b215de3d0abbc4f6daf2efd19e8809af0564490] Merge branch 'i2c/for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux into next
git bisect good 7b215de3d0abbc4f6daf2efd19e8809af0564490
# bad: [0430e49b6e7c6b5e076be8fefdee089958c9adad] ima: introduce ima_kernel_read()
git bisect bad 0430e49b6e7c6b5e076be8fefdee089958c9adad
# good: [2937f5efa5754754daf46de745f67350f7f06ec2] Merge branch 'for_linus' of git://cavan.codon.org.uk/platform-drivers-x86
git bisect good 2937f5efa5754754daf46de745f67350f7f06ec2
# good: [bc1dfff04a5d4064ba0db1fab13f84ab4f333d2b] Merge branch 'drm-nouveau-next' of git://anongit.freedesktop.org/git/nouveau/linux-2.6 into drm-next
git bisect good bc1dfff04a5d4064ba0db1fab13f84ab4f333d2b
# good: [412dd3a6daf0cadce1b2d6a34fa3713f40255579] Merge tag 'xfs-for-linus-3.16-rc1' of git://oss.sgi.com/xfs/xfs
git bisect good 412dd3a6daf0cadce1b2d6a34fa3713f40255579
# good: [c31c24b8251fd44962a9b0bf82c770653bf07f6e] Merge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux
git bisect good c31c24b8251fd44962a9b0bf82c770653bf07f6e
# good: [16b9057804c02e2d351e9c8f606e909b43cbd9e7] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
git bisect good 16b9057804c02e2d351e9c8f606e909b43cbd9e7
# bad: [19c1940feab777bb037c665a09f495d08a6c4e6c] Merge tag 'pm+acpi-3.16-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
git bisect bad 19c1940feab777bb037c665a09f495d08a6c4e6c
# good: [af76004cf8b4f368583bda22d7e348e40a338b91] Merge tag 'backlight-for-linus-3.16' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/backlight
git bisect good af76004cf8b4f368583bda22d7e348e40a338b91
# bad: [589e18a973bd6bb8abd2c6d4d8a1dcf5ae1dff61] Merge branch 'pm-cpufreq'
git bisect bad 589e18a973bd6bb8abd2c6d4d8a1dcf5ae1dff61
# good: [830bcac4e483d347087ed51eed564d1467e1e363] cpufreq: intel_pstate: Remove duplicate CPU ID check
git bisect good 830bcac4e483d347087ed51eed564d1467e1e363
# good: [9d674f2107b7dc801db9780f5f6b645f4905372f] Merge branch 'acpi-hotplug'
git bisect good 9d674f2107b7dc801db9780f5f6b645f4905372f
# bad: [de815a6d00da0f8a59e8aebf8efe12e289552a8f] Merge branches 'acpi-general' and 'acpi-video'
git bisect bad de815a6d00da0f8a59e8aebf8efe12e289552a8f
# good: [a4714a898e85205e1118ec923cde43d88eb105f6] ACPI: Fix bug when ACPI reset register is implemented in system memory
git bisect good a4714a898e85205e1118ec923cde43d88eb105f6
# bad: [751109aad5834375ca9ed0dcfcd85a00cbf872b5] ACPI / video: Change the default for video.use_native_backlight to 1
git bisect bad 751109aad5834375ca9ed0dcfcd85a00cbf872b5
# first bad commit: [751109aad5834375ca9ed0dcfcd85a00cbf872b5] ACPI / video: Change the default for video.use_native_backlight to 1


I'll see if I can test against the 3.17 branch to see if it still persists.
Comment by brent saner (sanerb) - Monday, 08 September 2014, 14:38 GMT
I still can't get 3.17 to boot; it seems to lock up on udev stuff.

But yeah, the above should be the patch that's causing this particular issue.

Do the arch maintainers report bugs upstream?
Comment by Tobias Powalowski (tpowa) - Tuesday, 09 September 2014, 12:04 GMT
Please report this to upstream maintainers. They will probably help you and a bisction is a very good start for them.
Comment by brent saner (sanerb) - Tuesday, 09 September 2014, 21:10 GMT
Thanks, Tobias. I'll pass it up to them!
Comment by brent saner (sanerb) - Wednesday, 10 September 2014, 01:42 GMT
For those interested in following:

https://bugzilla.kernel.org/show_bug.cgi?id=84221
Comment by Hong Xu (xuhdev) - Friday, 19 December 2014, 02:46 GMT
Seems everyone thinks it's an ATI related problem, but I also have this problem on an Nvidia card. The only difference is that I do not have this problem for every boot. It happens sometimes.
Comment by brent saner (sanerb) - Friday, 19 December 2014, 04:20 GMT
Hong Xu-

I'd open a new ticket at bugzilla.kernel.org and reference my bug (linked above) as well as the one I reference in that bug (84221). It'll also help to provide a dmidecode output since you'll probably be asked for it.

As for it only happening sometimes, can you confirm that:


1.) You are using the same kernel version every boot, and

2.) It happens truly at random and can't be reproduced at will (i.e. it occurs if you boot without A/C plugged in, but does not occur if you boot with A/C plugged in, etc.)

if it indeed can't be reproduced at-will, it might be a recent software update, so check your pacman logs and see what was installed/upgraded since the issue first started happening.
Comment by brent saner (sanerb) - Friday, 19 December 2014, 04:22 GMT
Oh, forgot to mention-

Upstream found a fix and should have a patch in one of the releases for 3.18 to fix the issues I was having with this (probably 3.18.2 if I had to guess).
Comment by Hong Xu (xuhdev) - Friday, 19 December 2014, 08:49 GMT
Thanks -- let's wait for 3.18.2 and see what would happen
Comment by Márcio Sousa Rocha (marciosr) - Monday, 22 December 2014, 22:02 GMT
Seems I'm too late, but I have the same problem on a notebook Samsung Ative Book 6 (np670z5e) with intel graphics. With linux-lts everything works nice but with latest kernels 3.17.x and mainline kernel (3.18) is not possible to set the backlight.
Arch linux full updated.
Acpi is instaled and I have the package samsung-laptop instaled too, but removing it do not change anything.

journalctl -xb shows the line below:
Dez 22 19:29:00 archnote systemd-backlight[24269]: Failed to get backlight or LED device 'backlight:acpi_video0': No such file or directory

Maybe we will have some look with the 3.18.2 release.
Thanks.
Comment by mattia (nTia89) - Sunday, 01 October 2017, 15:36 GMT
is this issue still valid?
Comment by brent saner (sanerb) - Sunday, 01 October 2017, 15:38 GMT
mattia- upstream has resolved back in 3.18.something. i had a scare on 4.12.x as it ironically seemed to have regressed, but i haven't noticed it be an issue since.

Loading...