Re: 3.13.?: Strange / dangerous fan policy...

[Manuel Krause <manuelkrause@netscape.net>]

On 2014-04-07 13:45, Rafael J. Wysocki wrote:
> On Monday, April 07, 2014 01:17:51 AM Manuel Krause wrote:
>> On 2014-04-06 04:43, Guenter Roeck wrote:
>>> On 04/05/2014 07:37 PM, Manuel Krause wrote:
>>>> On 2014-04-01 01:47, Guenter Roeck wrote:
>>>>> On 03/31/2014 04:37 PM, Manuel Krause wrote:
>>>>>> On 2014-03-20 21:21, Manuel Krause wrote:
>>>>>>> On 2014-03-11 22:59, Manuel Krause wrote:
>>>>>>>> On 2014-03-10 02:49, Manuel Krause wrote:
>>>>>>>>> On 2014-03-09 18:58, Rafael J. Wysocki wrote:
>>>>>>>>>> On Sunday, March 09, 2014 01:10:25 AM Manuel Krause wrote:
>>>>>>>>>>> On 2014-03-08 16:59, Guenter Roeck wrote:
>>>>>>>>>>>> On 03/08/2014 03:08 AM, Jean Delvare wrote:
>>>>>>>>>>>>> On Fri, 7 Mar 2014 14:52:30 -0800, Guenter Roeck wrote:
>>>>>>>>>>>>>> On Fri, Mar 07, 2014 at 11:04:29PM +0100, Manuel Krause
>>>>>>>>>>>>>> wrote:
>>>>>>> [SNIP]
>>>>>>>
>>>>>>> Long time no reply from you... Have I overseen a unwritten
>>>>>>> convention? Or were my charts that unusable for your
>>>>>>> analysis/work?
>>>>>>>
>>>>>>> Two days ago, I tried the 3.14.0-rc7-vanilla. And the problem
>>>>>>> persists. "Strange / dangerous fan policy..."
>>>>>>>
>>>>>>> Since kernel 3.13.6 I've managed to 'fix' the potential
>>>>>>> overheating problem by manually issuing a:
>>>>>>> "echo 1 > /sys/class/thermal/cooling_device3/cur_state" *)
>>>>>>> _before_ obviously critical temperatures occur. Remind: This
>>>>>>> particular setting may only work for my system! ...and keeps
>>>>>>> working for 3.14-rc.
>>>>>>>
>>>>>>> In the following I'd like to present you a modified output
>>>>>>> of my
>>>>>>> /sys/class/thermal, that I've written a script for (for my
>>>>>>> system), that shows the results in the way of
>>>>>>> linux/Documentation/thermal/sysfs-api.txt, point 3:
>>>>>>> {I've uploded the files to pastebin, to not swamp you and the
>>>>>>> lists with so many lines of logs.}
>>>>>>>
>>>>>>> For the last good kernel -- 3.12.14 -- in-use:
>>>>>>>    http://pastebin.com/HL1PNcda
>>>>>>> For my first bad kernel revision 3.13 -- at critical temp:
>>>>>>>    http://pastebin.com/98hgf1a9
>>>>>>> For the last bad kernel -- 3.14.0-rc7 -- at critical temp:
>>>>>>>    http://pastebin.com/MuTwTnjD
>>>>>>> For the last bad kernel -- 3.14.0-rc7 -- after issuing the
>>>>>>>    *) command:
>>>>>>>    http://pastebin.com/2peda54z
>>>>>>>
>>>>>>> Please, have a look at them! And maybe, give me hints on how I
>>>>>>> can help you to further debug this issue, as my manual method
>>>>>>> works but it's annoying.
>>>>>>>
>>>>>>> And, PLEASE CC: ME, as I'm not on the lists. Or lead this
>>>>>>> Email-thread to someone in charge.
>>>>>>>
>>>>>>> Thank you for your work && best regards,
>>>>>>> Manuel Krause
>>>>>>>
>>>>>>
>>>>>> This is still BUG 71711
>>>>>> https://bugzilla.kernel.org/show_bug.cgi?id=71711
>>>>>>
>>>>>> 3.12.15 works very well
>>>>>> 3.13.7 fails
>>>>>> 3.14.0-rc8 fails
>>>>>>
>>>>>
>>>>> Best you can do would really be to bisect the problem.
>>>>> Unfortunately only you (or someone else with an affected system)
>>>>> can do that. Once the culprit is known it would be much easier
>>>>> to get it fixed.
>>>>>
>>>>> To answer your earlier question: I don't think you did anything
>>>>> wrong.
>>>>> I guess everyone else is just as clueless as I am (if not,
>>>>> speak up
>>>>> and help ;-).
>>>>>
>>>>> Guenter
>>>>>
>>>>
>>>> I've now bisected two times. From two different kernel origins,
>>>> just to be sure, as I'm new to this stupid-and-lengthy method,
>>>> and, to be sure, I haven't given a false positive inbetween due
>>>> to boredom.
>>>>
>>>
>>> Not really. Keep in mint that you were able to track down the bad
>>> commit
>>> among more than 10,000 commits in a reasonably short period of time.
>>>
>>>> In the end it says each time:
>>>> # git bisect bad | tee -a /var/log/bisect.log
>>>> cc8ef52707341e67a12067d6ead991d56ea017ca is the first bad commit
>>>> commit cc8ef52707341e67a12067d6ead991d56ea017ca
>>>> Author: Zhang Rui <rui.zhang@intel.com>
>>>> Date:   Wed Sep 25 20:39:45 2013 +0800
>>>>
>>>>       ACPI / AC: convert ACPI ac driver to platform bus
>>>>
>>>>       Signed-off-by: Zhang Rui <rui.zhang@intel.com>
>>>>       Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
>>>>
>>> Off to the two of you...
>>>
>>> Guenter
>>>
>>>> :040000 040000 5a0d397cfcbf53c03390f2805b83754cb7837d84
>>>> 4a2af1454f65d67f1d1a507c08e3b9ef3ffe57e7 M      drivers
>>>>
>>>>
>>>> Please help me, on how I can help debug this more, and please
>>>> also read the newest from
>>>> https://bugzilla.kernel.org/show_bug.cgi?id=71711
>>>>
>>>> Manuel Krause
>>>>
>>>>
>>>>
>>>
>>
>> Sorry, that I've forgotton to add the following last night: After
>> the first bisection round, I was so glad about a result that
>> time, that I reverted this mentioned patch from the 3.13.8
>> kernel, but this didn't fix it.
>
> This means that the commit in question didn't introduce the problem
> you're seeing.
>
> Please check out commit 7f2dc5c4bcbf (Merge tag 'dm-3.13-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm),
> build a kernel from that and see if you can reprocude the problem with it.
> If so, it can be used as your new "first known bad" kernel for bisection.
> Otherwise, you can use it as the "first good" one and commit cc8ef52707341
> as "first known bad".
>
> Thanks!
>

Sorry, for any inconvenience, but you should forget about what 
I've written, that reverting the patch in question from 3.13.x 
didn't fix it. Of course it didn't fix it, as the patch doesn't 
cleanly revert from release-kernels at all. My mistake!

I' ve been guided by Guenter Roeck through two more bisecting 
sessions/ways on this, that always pointed to the commit in question.

Some citation:
Me:
>>> O.k. I've now followed your latest directions:
>>> git checkout -b testing cc8ef52707341e67a12067d6ead991d56ea017ca
>>> => result after rebuild was BAD =>
>>> git revert cc8ef52707341e67a12067d6ead991d56ea017ca
>>> => result after rebuild was GOOD
>>>
[ ...]
>>> Reverting that commit in question from this very git tree makes the
>>> kernel work as expected.
[ ... ]
Guenter:
>> Report the results you have above. That should show without question
>> that cc8ef52707341e67a12067d6ead991d56ea017ca is the bad commit,
>> and it should be easy to reproduce.

That seems to be all I can do for you for now. Please let me know 
of any preliminary patches to test!
And I want to add special thanks to Guenter Roeck for his 
always-just-in-time assistance over so many days,

Manuel Krause