Bug 14820 - high power consumption before suspend to ram
Summary: high power consumption before suspend to ram
Status: CLOSED INSUFFICIENT_DATA
Alias: None
Product: Power Management
Classification: Unclassified
Component: Hibernation/Suspend (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: Zhang Rui
URL: www.mygnu.de
Keywords:
Depends on:
Blocks:
 
Reported: 2009-12-15 23:34 UTC by mithrandir
Modified: 2013-03-27 07:32 UTC (History)
5 users (show)

See Also:
Kernel Version: 3.2
Tree: Mainline
Regression: Yes


Attachments
Hight Power State (before suspend) (3.92 KB, text/plain)
2009-12-24 01:00 UTC, mithrandir
Details
Low Power State (after suspend) (3.80 KB, text/plain)
2009-12-24 01:01 UTC, mithrandir
Details
dmesg output (boot, then suspend, then resume) (58.14 KB, text/plain)
2009-12-24 01:02 UTC, mithrandir
Details

Description mithrandir 2009-12-15 23:34:21 UTC
On my Dell Precision M65 the power consumption after a clean boot is much higher than after a suspend to ram cycle. After a clean boot it uses aproximately 1900mA and after a suspend to ram cycle it uses aproximately 1400mA. I cannot find any hints in dmesg.

The problem seems to be the same as described in http://bugzilla.kernel.org/show_bug.cgi?id=11122 for HP Laptops.

I tried this with kernel versions from 2.6.29 to 2.6.32. The missbehaviour was always reproducible. Thus it was reintroduced in 2.6.28 or 2.6.29.

Jürgen
Comment 1 Andrew Morton 2009-12-18 23:00:35 UTC
I'll reassign this to ACPI.
Comment 2 Len Brown 2009-12-22 02:34:30 UTC
please attach the output from powertop -d for the high-power
and low-power cases.
Comment 3 mithrandir 2009-12-24 01:00:21 UTC
Created attachment 24275 [details]
Hight Power State (before suspend)
Comment 4 mithrandir 2009-12-24 01:01:00 UTC
Created attachment 24276 [details]
Low Power State (after suspend)
Comment 5 mithrandir 2009-12-24 01:02:28 UTC
Created attachment 24277 [details]
dmesg output (boot, then suspend, then resume)
Comment 6 mithrandir 2009-12-24 01:06:30 UTC
The outputs I just posted are from latest zen-kernel, but the same problem is also present with vanilla sources (tried 2.6.32 without any patches).
Comment 7 Zhang Rui 2009-12-30 08:07:07 UTC
please set CONFIG_PM_DEBUG, and run this test:

"echo core > /sys/power/pm_test"
"echo mem > /sys/power/state"
wait about 10s

does the power consumption drop after the system coming back?
Comment 8 mithrandir 2010-01-07 18:07:25 UTC
I tried to do so, but the system does not come back when I echoed core > pm_test before. The screen stays black. When not echoing core > pm_test the supend cycle works and the system of course consumes less power afterwards. Afterwards the test cycle with echo core > pm_test works and the system also consumes less power. 

I don´t think, that this information is of any use...

sorry

Jürgen
Comment 9 Zhang Rui 2010-01-13 08:18:03 UTC
that's weird.

what do you get after running "echo core > /sys/power/pm_test" and "echo mem > /sys/power/state"?
you said the screen is black, is the screen still black after 20 seconds?
does the system hangs? say, ping this laptop and see if there is response.
do you have to reset to get the system back to work?
Comment 10 mithrandir 2010-01-14 23:42:38 UTC
The screen is still black after 20 seconds. The Laptop is not pingable. Sometimes I noticed blinking keyboard LEDs. I have to reset the laptop to bring is back to work.
Comment 11 Zhang Rui 2010-01-21 08:02:54 UTC
please try to echo processors/platform/devices/freezer instead of "core" to /sys/power/pm_test, and see with which value the S3 test can resume successfully.
Comment 12 Zhang Rui 2010-02-21 07:06:52 UTC
ping...
Comment 13 mithrandir 2010-02-25 19:54:27 UTC
Sorry, I was busy the last weeks. I tried to echo processors/platform/devices/freezer into pm_test. I used a fresh boot every time and was on battery power. Platform, devices and freezer worked.  Core and Processors crashed as reported before. The power usage went down after platform and devices. The power usage stayed high after core and processors. One strange thing I noticed was that running one cycle with devices after the freezer cycle (this time without reboot) the system also crashed.
Comment 14 Zhang Rui 2010-02-26 02:03:28 UTC
Hah, an interesting bug.

Rafael,
have you ever seen a system that works well in S3, but fails in the pm_test?
Comment 15 Rafael J. Wysocki 2010-02-26 20:11:15 UTC
No, I haven't, but I bet this the bug discussed here:

http://patchwork.kernel.org/patch/81830/
Comment 16 Zhang Rui 2010-03-01 06:02:29 UTC
(In reply to comment #15)
> No, I haven't, but I bet this the bug discussed here:
> 
> http://patchwork.kernel.org/patch/81830/

I'm afraid this won't help because the patch above is a fix for hibernation.
Comment 17 Zhang Rui 2011-04-19 07:54:03 UTC
does the problem still exist in the latest upstream kernel?
Comment 18 mithrandir 2011-05-05 22:48:34 UTC
The problem still persists in 2.6.37.5-zen, which was the latest one I tried.
Comment 19 mithrandir 2011-05-30 11:01:24 UTC
The problem is still there in 2.6.38.4.
Comment 20 Zhang Rui 2012-01-18 01:55:56 UTC
It's great that kernel bugzilla is back.

mithrandir,
can you please verify if the problem still exists in the latest upstream
kernel?
Comment 21 mithrandir 2012-01-22 18:52:39 UTC
The problem is still there with 3.0 kernel, which is the last one I tried. I will write again when I have built a newer kernel.

mithrandir
Comment 22 mithrandir 2012-03-28 17:22:26 UTC
The problem is still there with 3.2 kernel.

mithrandir
Comment 23 Zhang Rui 2012-12-11 01:22:03 UTC
(In reply to comment #13)
> Sorry, I was busy the last weeks. I tried to echo
> processors/platform/devices/freezer into pm_test. I used a fresh boot every
> time and was on battery power. Platform, devices and freezer worked.  Core
> and
> Processors crashed as reported before. The power usage went down after
> platform
> and devices.

You mean the system consumes less power after resuming from "echo device > /sys/power/pm_test && echo mem > /sys/power/state", but not for "freezer", right?

can you attach the output of the latest powertop tool both before and after pm_test?
Comment 24 Zhang Rui 2013-03-08 05:56:44 UTC
ping...
Comment 25 Zhang Rui 2013-03-27 07:32:34 UTC
Hmm,
I think it is okay to close this bug report as there is no response from the bug reporter for one year.
But please feel free to reopen it if the problem still exists in the latest upstream kernel.

Note You need to log in before you can comment on or make changes to this bug.