Re: [REGRESSION] [2.6.37-rc1] Fan noise after suspend to ram/disk

Previous thread: by Looney on Thursday, November 4, 2010 - 9:06 am. (1 message)

Next thread: [PATCH 1/1] crypto: api.c: adding doxygen comments to api by Mark Allyn on Thursday, November 4, 2010 - 9:32 am. (3 messages)
From: Maciej Rutecki
Date: Thursday, November 4, 2010 - 9:06 am

1. 
Last known good kernel: 2.6.36
Bad kernel: 2.6.37-rc1
===============================

2. Problem:
When I resume computer from s2ram/s2disk and current CPU temp is below lowest 
trip point, and -- from unknown reason -- fan start in maximum speed it 
doesn't turn off (it should).

It doesn't happen always but only when fan starts work and any of trip points 
aren't exceed. It seems that system only check temperature when exceed trip 
point.
===============================

3. /sys/cass/thermal/ info:

I have:
maciek@gumis:/sys/class/thermal$ ls
cooling_device0  cooling_device2  cooling_device4  cooling_device6  
thermal_zone1  thermal_zone3
cooling_device1  cooling_device3  cooling_device5  thermal_zone0    
thermal_zone2  thermal_zone4

maciek@gumis:/sys/class/thermal$ cat cooling_device*/type 
LCD
Processor
Processor
Fan
Fan
Fan
Fan

===============================

Colling device from 3 to 6 are inform what is current fan speed, for example:
cooling_device3 = 0
cooling_device4 = 0
cooling_device5 = 0
cooling_device6 = 0 
Means that fan is off.

cooling_device3 = 0
cooling_device4 = 0
cooling_device5 = 0
cooling_device6 = 1
Means fan works on 40% of full speed (1st trip point is exceed).

cooling_device3 = 1
cooling_device4 = 1
cooling_device5 = 1
cooling_device6 = 1
Means fan works on full speed (all (4) trip points is exceed).
===============================

5. State when fan works on full speed after resume:
maciek@gumis:/sys/class/thermal$ cat cooling_device*/cur_state
0
0
0
1
1
1
1
(Runs on full speed)

maciek@gumis:/sys/class/thermal$ cat thermal_zone0/trip_point_*
256000
critical
78000
active
70000
active
60000
active
50000
active

(It's OK for current CPU temp)

"sensors" command shows:
maciek@gumis:/sys/class/thermal$ sensors
acpitz-virtual-0
Adapter: Virtual device
temp1:       +37.0°C  (crit = +256.0°C)                  
temp2:       +42.0°C  (crit = +90.0°C)                  
temp3:       +33.0°C  (crit = ...
From: Zhang Rui
Date: Thursday, November 4, 2010 - 5:27 pm

could you please attach the output of "grep . /sys/class/thermal/*/*"
when this happens?

thanks,


--

From: Maciej Rutecki
Date: Friday, November 5, 2010 - 12:49 pm

Thanks for the answer.

maciek@gumis:~$ grep . ...
From: Zhang Rui
Date: Sunday, November 7, 2010 - 11:58 pm

IMO, the only one that may cause this regression is commit
c57b62f5b1e6dd69ff8c96f6db7f86ea31c0e21f
some app may enable polling via procfs and update the thermal zone
status.

could you please attach the output of "grep . /proc/acpi/thermal/*/*" in
2.6.36 kernel?

thanks,
rui



--

From: Maciej Rutecki
Date: Monday, November 8, 2010 - 10:35 am

I revert this commit from 2.6.37-rc1. Still the same:
maciek@gumis:~$ grep . ...
From: Zhang Rui
Date: Sunday, November 14, 2010 - 8:16 pm

could you please apply the patch attached, on top of 2.6.37-rc1, and
then attach the dmesg output if the fan is noise after suspend/resume.

thanks,

From: Zhang Rui
Date: Monday, November 15, 2010 - 6:26 pm

it seems that it's not the thermal driver that changes the Fan speed.
Please apply the debug patch attached, rebuild with CONFIG_ACPI_DEBUG
set and reboot with "acpi.debug_level=0x004
acpi.debug_layer=0x04A10000", and then attach the dmesg output after the
same test.
Please attach the acpidump output of this laptop as well.

thanks,

From: Maciej Rutecki
Date: Tuesday, November 16, 2010 - 8:46 am

Regards
-- 
Maciej Rutecki
http://www.maciek.unixy.pl
--

From: Maciej Rutecki
Date: Tuesday, November 16, 2010 - 10:31 am

2.6.37-rc1 + your latest patch, suspend to ram&resume:
http://unixy.pl/maciek/download/kernel/2.6.37-rc1/gumis/dmesg-debug-2.6.37-
rc1.txt.zip

Thanks
-- 
Maciej Rutecki
http://www.maciek.unixy.pl
--

From: Maciej Rutecki
Date: Tuesday, November 16, 2010 - 12:28 pm

Revert 3e384ee6c687cb397581ee8f9440fc8220cfac80 seems solve problem:
https://bugzilla.kernel.org/show_bug.cgi?id=22932#c3

Thanks
-- 
Maciej Rutecki
http://www.maciek.unixy.pl
--

From: Rafael J. Wysocki
Date: Tuesday, November 16, 2010 - 3:35 pm

Please do the test I asked for in the bug entry.

Thanks,
Rafael
--

Previous thread: by Looney on Thursday, November 4, 2010 - 9:06 am. (1 message)

Next thread: [PATCH 1/1] crypto: api.c: adding doxygen comments to api by Mark Allyn on Thursday, November 4, 2010 - 9:32 am. (3 messages)