Kernel 3.6.2 Power Regressions (Sandy Bridge)

shadyabhi · 2012-11-16 13:43:55

I have noticed that if I boot/resume my laptop without power cord plugged in, my laptop generally boots up without those cpu scaling issues.

Last edited by shadyabhi (2012-11-16 13:44:13)

donniezazen · 2012-11-16 17:13:08

I am also booting up to high temperatures.

Phobophobia · 2012-11-17 08:29:28

LordChaos73 wrote:

Phobophobia wrote:
How high is "very high"? There are at least a few 13" ultrabooks with 1080p screens, and I don't know of any currently produced laptops of any size that have a higher resolution than that.
Have you been sleeping under a rock lately ?
http://www.apple.com/macbook-pro/

Derp.

Rolinh · 2012-11-17 10:19:00

Question: do you also get kernel panics (usually at boot-up) once in a while with kernel 3.6 ?

LordChaos73 · 2012-11-17 11:27:19

Rolinh wrote:

Question: do you also get kernel panics (usually at boot-up) once in a while with kernel 3.6 ?

Nope, never had any. Sorry.

LordChaos73 · 2012-11-17 13:33:23

After resume from suspend powertop is always reporting the following, which seems very strange to me (I'm on a wireless network by the way):

The battery reports a discharge rate of 16.2 W
The estimated remaining time is 2 hours, 46 minutes

Summary: 378.9 wakeups/second, 15.9 GPU ops/seconds, 0.0 VFS ops/sec and 7.6% CPU use

Power est. Usage Events/s Category Description
10.3 W 0.0 pkts/s Device Network interface: eth0 (r8169)

Before suspending the laptop, it is usually consuming about 8-9W.

And after a reboot, idling:

The battery reports a discharge rate of 8.80 W
The estimated remaining time is 4 hours, 1 minutes

Summary: 194.4 wakeups/second, 0.1 GPU ops/seconds, 0.0 VFS ops/sec and 40.3% CPU use

Power est. Usage Events/s Category Description
10.3 W 0.0 pkts/s Device Network interface: eth0 (r8169)

This doesn't seem right, now does it?

Last edited by LordChaos73 (2012-11-17 14:10:15)

altercation · 2012-11-17 22:25:03

Rolinh wrote:

Question: do you also get kernel panics (usually at boot-up) once in a while with kernel 3.6 ?

Yep. Regular enough under 3.6 (majority of the time boot is fine, but then boom, panic). Rolled back to 3.5 for this and the power regression.

Schrottfresse · 2012-11-18 14:22:27

altercation wrote:

Rolinh wrote:
Question: do you also get kernel panics (usually at boot-up) once in a while with kernel 3.6 ?
Yep. Regular enough under 3.6 (majority of the time boot is fine, but then boom, panic). Rolled back to 3.5 for this and the power regression.

Me too.

r0b0t · 2012-11-19 10:44:10

coretemp-isa-0000
Adapter: ISA adapter
Physical id 0: +45.0 C (high = +86.0 C, crit = +100.0 C)
Core 0: +45.0 C (high = +86.0 C, crit = +100.0 C)
Core 1: +45.0 C (high = +86.0 C, crit = +100.0 C)
After update, I just installed intel-dri update and ati-dri and as you see the temperature it's OK now, you may try this.

shadyabhi · 2012-11-20 04:07:27

This has happened to me like 5th or 6th already. I never shutdown, I suspend my computer. So, at the time of resuming, the cpu scaling starts to work on every 3rd resume.

So, generally when I resume my laptop, I just check by powertop if everything is fine & if it isn't I suspend it again and resume. This almost always works the third time. This could be the wierdest thing ever but can anyone also replicate this? (I generally have my laptop on battery when I'm doing this)

Also, whenever CPU scaling works, GPU always goes into RC6 state.

Last edited by shadyabhi (2012-11-20 04:11:08)

mabl · 2012-11-20 08:48:40

Hi,

I'm on an Lenovo X220 with i7-2620M, Linux 3.6.6-1-ARCH (unpatched) and see the same issue.

shadyabhi wrote:

This has happened to me like 5th or 6th already. I never shutdown, I suspend my computer. So, at the time of resuming, the cpu scaling starts to work on every 3rd resume.

Once I see the error, resuspend never helps...

shadyabhi wrote:

Also, whenever CPU scaling works, GPU always goes into RC6 state.

Indeed, and not only this, the GPU is then clocked to its maximum frequency:

# cat /sys/kernel/debug/dri/0/i915_cur_delayinfo
GT_PERF_STATUS: 0x00000d29
RPSTAT1: 0x00048d00
Render p-state ratio: 13
Render p-state VID: 41
Render p-state limit: 255
CAGF: 650MHz
RP CUR UP EI: 61329us
RP CUR UP: 4us
RP PREV UP: 0us
RP CUR DOWN EI: 0us
RP CUR DOWN: 0us
RP PREV DOWN: 0us
Lowest (RPN) frequency: 650MHz
Nominal (RP1) frequency: 650MHz
Max non-overclocked (RP0) frequency: 1300MHz

# cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq
800000

############ After standby #################################

# cat /sys/kernel/debug/dri/0/i915_cur_delayinfo
GT_PERF_STATUS: 0x00001ac6
RPSTAT1: 0x00049a19
Render p-state ratio: 26
Render p-state VID: 198
Render p-state limit: 255
CAGF: 1300MHz
RP CUR UP EI: 9142us
RP CUR UP: 9142us
RP PREV UP: 66000us
RP CUR DOWN EI: 191074us
RP CUR DOWN: 0us
RP PREV DOWN: 0us
Lowest (RPN) frequency: 650MHz
Nominal (RP1) frequency: 650MHz
Max non-overclocked (RP0) frequency: 1300MHz

# cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq
2700000

Notice the CAGF which is now stuck at 1300MHz. Also, the GT_PERF_STATUS and RPSTAT1 don't change when doing another standby - I just don't know what it means, though...

LordChaos73 · 2012-11-20 22:52:33

mabl wrote:

Notice the CAGF which is now stuck at 1300MHz. Also, the GT_PERF_STATUS and RPSTAT1 don't change when doing another standby - I just don't know what it means, though...

Unfortunately, I can't reproduce this. The only thing I'm seeing is a ~5W increase in battery discharge rate after suspend. But I can't find the root cause; the cpufreq governor seems to work fine and my GPU is still on 650Mhz when I have resumed my session. Since I can reproduce this issue using a Ubuntu live cd as well (both 12.04 & 12.10), I'm thinking I've hit another bug. Or could this be caused by a buggy BIOS as well?

KaiSforza · 2012-11-20 23:55:24

LordChaos73 wrote:

mabl wrote:
Notice the CAGF which is now stuck at 1300MHz. Also, the GT_PERF_STATUS and RPSTAT1 don't change when doing another standby - I just don't know what it means, though...
Unfortunately, I can't reproduce this. The only thing I'm seeing is a ~5W increase in battery discharge rate after suspend. But I can't find the root cause; the cpufreq governor seems to work fine and my GPU is still on 650Mhz when I have resumed my session. Since I can reproduce this issue using a Ubuntu live cd as well (both 12.04 & 12.10), I'm thinking I've hit another bug. Or could this be caused by a buggy BIOS as well?

If you're reproducing this with an Ubuntu live cd, then I'd say it's not just related to 3.6.x. 3.6 wasn't released till the end of September of this year, and I highly doubt that Ubuntu was using it on those live cd's.

chx · 2012-11-21 04:24:44

mabl wrote:

Hi,
I'm on an Lenovo X220 with i7-2620M, Linux 3.6.6-1-ARCH (unpatched) and see the same issue.
shadyabhi wrote:
Also, whenever CPU scaling works, GPU always goes into RC6 state.
Indeed, and not only this, the GPU is then clocked to its maximum frequency:

Lenovo T420, I can confirm this report.

hendry · 2012-11-21 10:50:47

donniezazen wrote:

I am also booting up to high temperatures.

Me too. I basically stopped suspending thinking that might work around the problem.

x220:~$ grep HandleLid /etc/systemd/logind.conf
HandleLidSwitch=ignore

And it still goes up to high temperatures. I think after playing a Flash video seems to trigger the problem. Really hard to reproduce.

I'm also getting occasional kernel panics when I share my internet connection. Again I can't reproduce it.

Anyway this is definitely a low point for me, thinkpads and Linux. I'm basically unproductive since the machine is too hot to use a lot of the time. ~80C.

I can't use an older kernel readily since Virtualbox doesn't work. Nightmare.

martinc2 · 2012-11-21 11:09:45

hendry wrote:

I can't use an older kernel readily since Virtualbox doesn't work. Nightmare.

Have you tried using virtualbox-host-source instead of virtualbox-host-modules? Works fine for me with 3.5.6 kernel

johnLate · 2012-11-22 00:51:52

I have a "X220 Tablet" running 3.7.0-rc6 and also suffer from this problem.

To "fix" it, I go through a cycle of "echo mem > /sys/power/state", powerbutton, "powertop" and repeat if "GPU active" shows 100%.

If I don't do this, I have a battery life time of only 1.5 hours instead of >6.

hongleong · 2012-11-22 03:20:01

hendry wrote:

I can't use an older kernel readily since Virtualbox doesn't work.

I reverted to kernel 3.5.6 along with the older virtualbox-host-modules-4.2.0-5 to match the kernel version. VirtualBox has been running fine so far. Here are the commands:

# cd /var/cache/pacman/pkg
# pacman -U linux-3.5.6-1-x86_64.pkg.tar.xz virtualbox-host-modules-4.2.0-5-x86_64.pkg.tar.xz

Rebooted and tested. It has been 8 days since reverting and so far so good. Hope that helps.

mabl · 2012-11-22 08:00:11

You can also automate it a bit ^.^

#!/bin/bash

COUNTER=0
while  grep --quiet 0x00001ac6 /sys/kernel/debug/dri/0/i915_cur_delayinfo ; do
  let COUNTER=COUNTER+1 

  echo Suspend resume cycle $COUNTER

  rtcwake -s 2 -m mem
  sleep 5
done

echo "Suspend problems solved after $COUNTER tries"

This relies on the GT_PERF_STATUS output - I found it to be an reliable indicator for my machine.

Last edited by mabl (2012-11-22 08:10:38)

chx · 2012-11-22 11:28:02

After downgrading the kernel, my brightness control no longer works hope this gets fixed. For now, it's 3.5.4, otherwise laptop cooks.

thesofty · 2012-11-23 19:53:50

I'm also affected by the sandy bridge - problem on my new lenovo X1.
Because it was a 1'st installation on the Laptop downgrading was not an option for me. So I did the patch on the actual arch-kernel 3.6.6-1, using abs. It seens to work for me, at least much better than before.
But it does't fix the problem completely. My expierience is that after booting the kernel, the power regression problem still occurs sometimes and sometimes not. When I close the lid my laptop switches to suspend mode, and after waking it up again, the fix seems to work perfectly.

Last edited by thesofty (2012-11-23 20:08:27)

The Compiler · 2012-11-23 20:30:07

From the freedesktop.org bug:

Daniel Vetter 2012-11-22 20:28:28 UTC
Thanks everyone for testing, I'll tentatively close this for now. Jesse is still working on a real fix instead of duct-tape, so we might ask for a bit of testing still.

Mark 2012-11-23 14:58:21 UTC
So when will we see this? Kernel update 2.??

Daniel Vetter 2012-11-23 17:18:39 UTC
Going into 3.8 atm. It requires quite some code rework, so unlikely to get backported. The real fix Jesse is working on though might (since that will be less invasive).

So there is hope this is fixed in 3.8 for everyone

spinacz · 2012-11-24 08:58:16

thesofty wrote:

I'm also affected by the sandy bridge - problem on my new lenovo X1.
Because it was a 1'st installation on the Laptop downgrading was not an option for me.

You can downgrading system with this repo . After downgrade i added this line to pacman.conf

IgnorePkg   = linux linux-headers linux-api-headers

now i am waiting for solution.

Arch Linux

#151 2012-11-16 13:43:55

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#152 2012-11-16 17:13:08

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#153 2012-11-17 08:29:28

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#154 2012-11-17 10:19:00

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#155 2012-11-17 11:27:19

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#156 2012-11-17 13:33:23

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#157 2012-11-17 22:25:03

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#158 2012-11-18 14:22:27

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#159 2012-11-19 10:44:10

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#160 2012-11-20 04:07:27

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#161 2012-11-20 08:48:40

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#162 2012-11-20 22:52:33

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#163 2012-11-20 23:55:24

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#164 2012-11-21 04:24:44

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#165 2012-11-21 10:50:47

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#166 2012-11-21 11:09:45

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#167 2012-11-22 00:51:52

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#168 2012-11-22 03:20:01

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#169 2012-11-22 08:00:11

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#170 2012-11-22 11:28:02

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#171 2012-11-23 19:53:50

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#172 2012-11-23 20:30:07

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#173 2012-11-24 08:58:16

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#174 2012-11-24 10:58:56

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

#175 2012-11-24 16:37:14

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Board footer