Kernel 3.6.2 Power Regressions (Sandy Bridge)

firewalker · 2013-06-17 16:01:09

Yes, I did recompile the kernel. I guess you could only compile the i915 module.

Chais · 2013-06-24 06:40:10

With 3.9.7 even my i5 goes up to 93°C. Any improvements on the bug filed?

hrkfdn · 2013-06-24 08:55:15

This reply came in on a different (but related) bugreport: https://bugzilla.kernel.org/show_bug.cgi?id=58971#c56

Recompiling my kernel and trying this now..

Last edited by hrkfdn (2013-06-24 16:20:14)

jsravn · 2013-06-26 00:02:05

Latest workaround is to set /sys/power/pm_async to 0. So far this seems to have fixed it for me.

cris9288 · 2013-06-27 03:20:47

jsravn wrote:

Latest workaround is to set /sys/power/pm_async to 0. So far this seems to have fixed it for me.

tried this and still got hot wakeups. I set up a file in /etc/tmpfiles.d to echo 0 to pm_async on boot. After a suspend and resume, my temp was up to 50 C and climbing. Usually I boot/resume at 28 C.

alexyakushev · 2013-07-02 18:43:22

I confirm that setting /sys/power/pm_async to 0 seems to fix the issue for me as well. Once after I did that, the laptop failed to wake up after suspension, but since then it was all nice and quiet.

I'm running 3.9.6-1-ARCH x86_64

wlmeng11 · 2013-07-02 22:41:46

If anyone is looking for a stable release to fall back to, 3.5.6-1-ARCH x86_64 is perfectly fine for me on a Thinkpad X220 with a i5-2520m.

cris9288 · 2013-07-06 20:45:41

wlmeng11 wrote:

If anyone is looking for a stable release to fall back to, 3.5.6-1-ARCH x86_64 is perfectly fine for me on a Thinkpad X220 with a i5-2520m.

It's interesting that this seems to vary. The last stable release on my machine was the last 3.7 release.

Thaodan · 2013-07-07 00:59:02

my laptop (thinkpad edge e530, core3rd) works fine with 3.9.5-pf. my pc with 3.10 and core2nd too

Last edited by Thaodan (2013-07-07 01:35:13)

WonderWoofy · 2013-07-07 01:12:47

Thaodan, this thread is about power regressions with Sandy Bridge processors specifically. So I am not sure why you are reporting on a ThinkPad E535 (which should be AMD BTW) and a core2duo. It seems kind of cruel to give false hope to some poor sufferer of this bug who fails to read your post in detail.

Thaodan · 2013-07-07 01:22:43

WonderWoofy wrote:

Thaodan, this thread is about power regressions with Sandy Bridge processors specifically. So I am not sure why you are reporting on a ThinkPad E535 (which should be AMD BTW) and a core2duo. It seems kind of cruel to give false hope to some poor sufferer of this bug who fails to read your post in detail.

My laptop has core3nd I meant E530 and I forgot the n at the 2nd. I wrote about the heat issues.

Last edited by Thaodan (2013-07-07 01:34:51)

WonderWoofy · 2013-07-07 01:56:41

Ah, okay that makes more sense then...

jsravn · 2013-07-14 18:56:23

alexyakushev wrote:

I confirm that setting /sys/power/pm_async to 0 seems to fix the issue for me as well. Once after I did that, the laptop failed to wake up after suspension, but since then it was all nice and quiet.
I'm running 3.9.6-1-ARCH x86_64

Interesting... what machine are you on? I'm on an x220 i5. And since I made that pm_async change, I haven't had the issue a single time. Before it was happening like every 3rd suspend/resume.

Note I was on 3.9.7 when I tried this workaround. Maybe try a later kernel and see if it improves.

donniezazen · 2013-07-14 19:42:33

jsravn wrote:

Latest workaround is to set /sys/power/pm_async to 0. So far this seems to have fixed it for me.

How exactly have to you set it up?

Thaodan · 2013-07-15 00:02:44

donniezazen wrote:

jsravn wrote:
Latest workaround is to set /sys/power/pm_async to 0. So far this seems to have fixed it for me.
How exactly have to you set it up?

It dont need this workaround but for such stuff I have a /etc/systemd/system/rc.local.service:

[Unit]
Description=/etc/rc.local Compatibility

[Service]
Type=forking
ExecStart=/etc/rc.local 
TimeoutSec=0
StandardOutput=tty
RemainAfterExit=yes
SysVStartPriority=99

[Install]
WantedBy=multi-user.target

Then just write a /etc/rc.local with your stuff and make it executable.
I just it to write the target file of tuxonice_file to /sys/power/tuxonice/file/target

The Compiler · 2013-07-16 06:56:54

donniezazen wrote:

jsravn wrote:
Latest workaround is to set /sys/power/pm_async to 0. So far this seems to have fixed it for me.
How exactly have to you set it up?

IMHO the nicest way to do this, is to do a tmpfiles.d entry, e.g. I have this in /etc/tmpfiles.d/pm_async.conf:

w /sys/power/pm_async - - - - 0

See "man tmpfiles.d" for details.

Rolinh · 2013-07-16 13:16:12

cris9288 wrote:

wlmeng11 wrote:
If anyone is looking for a stable release to fall back to, 3.5.6-1-ARCH x86_64 is perfectly fine for me on a Thinkpad X220 with a i5-2520m.
It's interesting that this seems to vary. The last stable release on my machine was the last 3.7 release.

For my laptop (Thinkpad T420, Core i7-2620M), this is also the latest stable release. Things started to go bad with 3.6 release.

trsk · 2013-07-16 15:57:38

Rolinh wrote:

cris9288 wrote:
wlmeng11 wrote:
If anyone is looking for a stable release to fall back to, 3.5.6-1-ARCH x86_64 is perfectly fine for me on a Thinkpad X220 with a i5-2520m.
It's interesting that this seems to vary. The last stable release on my machine was the last 3.7 release.
For my laptop (Thinkpad T420, Core i7-2620M), this is also the latest stable release. Things started to go bad with 3.6 release.

It is 3.5.6-1-ARCH for me as well.

WonderWoofy · 2013-07-16 17:12:08

trsk wrote:

Rolinh wrote:
cris9288 wrote:
It's interesting that this seems to vary. The last stable release on my machine was the last 3.7 release.
For my laptop (Thinkpad T420, Core i7-2620M), this is also the latest stable release. Things started to go bad with 3.6 release.
It is 3.5.6-1-ARCH for me as well.

Ooo, no bueno! Going back to 3.5 means that you are now using an end-of-life kernel. There are no updates or security fixes happening to that. So you might want to humor the idea of using the linux-lts34 (currently 3.4.53-1-ARCH) package from the AUR if you are willing to compile your own kernel. The config included is sane and all inclusive like the normal ARCH kernel, so if you are okay with a semi-long compile time you don't even have to change anything. OTOH, you have a Core i7 quad with hyper-threading, so you could probably compile with -j10 pretty sanely.

donniezazen · 2013-07-17 11:17:05

The Compiler wrote:

IMHO the nicest way to do this, is to do a tmpfiles.d entry, e.g. I have this in /etc/tmpfiles.d/pm_async.conf:
w /sys/power/pm_async - - - - 0
See "man tmpfiles.d" for details.

Thanks. Do I have to enable any service file or are tempfiles enabled by default?

PS:- There some relevant info on wiki on modifying /sys parameters.

Note: This method may not work to set options in /sys since the systemd-tmpfiles-setup service may run before the appropriate device modules
is loaded. In this case you could check whether the module has a parameter for the option you want to set with modinfo <module> and set this option 
with a config file in /etc/modprobe.d. Otherwise you will have to write a udev rule to set the appropriate attribute as soon as the device appears.

firewalker · 2013-07-17 11:49:35

It seem to be resolved.

https://bugzilla.kernel.org/show_bug.cgi?id=58971

commit 7dcd2677ea912573d9ed4bcd629b0023b2d11505
Author: Konstantin Khlebnikov <khlebnikov@openvz.org>
Date:   Wed Jul 17 10:22:58 2013 +0400

    drm/i915: fix long-standing SNB regression in power consumption after resume

Thaodan · 2013-07-17 13:45:13

Anyone found a workaround for the texture corruption after suspend to disk?

doug piston · 2013-07-17 14:21:57

firewalker wrote:

It seem to be resolved.

https://bugzilla.kernel.org/show_bug.cgi?id=58971

commit 7dcd2677ea912573d9ed4bcd629b0023b2d11505
Author: Konstantin Khlebnikov <khlebnikov@openvz.org>
Date:   Wed Jul 17 10:22:58 2013 +0400

    drm/i915: fix long-standing SNB regression in power consumption after resume

Good news first thing in the morning.

trsk · 2013-07-17 16:51:44

Can anyone confirm that the problem is fixed on the X220? I will only be able to test it in a few days when I get the laptop back.

dr_undecided · 2013-07-17 19:50:25

Sweet!
What kernel version will it be included in? When do you estimate it will reach the core repository?

Arch Linux

#526 2013-06-17 16:01:09

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#527 2013-06-24 06:40:10

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#528 2013-06-24 08:55:15

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#529 2013-06-26 00:02:05

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#530 2013-06-27 03:20:47

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#531 2013-07-02 18:43:22

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#532 2013-07-02 22:41:46

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#533 2013-07-06 20:45:41

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#534 2013-07-07 00:59:02

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#535 2013-07-07 01:12:47

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#536 2013-07-07 01:22:43

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#537 2013-07-07 01:56:41

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#538 2013-07-14 18:56:23

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#539 2013-07-14 19:42:33

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#540 2013-07-15 00:02:44

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#541 2013-07-16 06:56:54

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#542 2013-07-16 13:16:12

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#543 2013-07-16 15:57:38

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#544 2013-07-16 17:12:08

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#545 2013-07-17 11:17:05

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#546 2013-07-17 11:49:35

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#547 2013-07-17 13:45:13

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#548 2013-07-17 14:21:57

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#549 2013-07-17 16:51:44

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#550 2013-07-17 19:50:25

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Board footer