linux-kernel - Re: CPU scheduler weirdness?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <alpine.LNX.2.00.0908171758340.2764@cinke.fazekas.hu>
Date:	Tue, 18 Aug 2009 21:49:31 +0200 (CEST)
From:	Marton Balint <cus@...ekas.hu>
To:	Andreas Mohr <andi@...as.de>
cc:	linux-kernel@...r.kernel.org, mingo@...e.hu, peterz@...radead.org
Subject: Re: CPU scheduler weirdness?

Hi,

On Thu, 13 Aug 2009, Andreas Mohr wrote:
> On Thu, Aug 13, 2009 at 05:39:10PM +0200, Marton Balint wrote:
>>>> Does anybody have any idea what can cause this?
>>>
>>> /sys/devices/system/cpu/sched_smt_power_savings , perhaps?
>>
>> Thanks for the tip, tuning the sched_mc_power_savings setting helped! The
>> original value of it was 0, but after setting it to 1, the two
>> cpu-intensive processes got scheduled to different CPU cores, as
>> expected.
>
> Heh, I did expect it to not help, and indeed that thing helping in this
> way points to a... BUG, plain and simple.
>
> http://lwn.net/Articles/297306/
> lists possible settings as
>
> "
> The power savings and performance of the given workload in an under
> utilised system can be controlled by setting values of 0, 1 or 2 to
> /sys/devices/system/cpu/sched_mc_power_savings with 0 being highest
> performance and least power savings and level 2 indicating maximum
> power savings even at the cost of slight performance degradation.
> "
>
> which is exactly opposite to what I'd have expected to be normal,
> unconfigured behaviour in your case.
>
>> Setting it back to 0 casused the two cpu-intensive processes to run on
>> the same CPU again. So I guess I will just set it to 1 after booting the
>> system.
>
> ...which would indicate a level=1 or level=2 (maximum powersaving)
> behaviour. Something either seems reversed or really weird.
> But it could just be opaque if correct behaviour due to a much more complex
> load balancing algo in the scheduler or so.
>
> Comments, anyone?

In the meantime, I was able to create a tiny C program which always 
succesfully reproduces the bug. It's basically an endless loop which does 
not stop while the process is running on the last CPU core. The program 
creates multiple instances of itself, to be able to keep all of the CPU 
cores busy. After 1 second, the processes running on other than the last 
CPU core die, the processes running on the last CPU core remain stuck 
there...

I tested it on my dual core system, if someone could test it on a quad 
core and report back that would probably be useful.

Usage: ./schedtest <number of CPU cores>

And don't forget to kill the stuck processes after using the program! :)

Regards,
   Marton
View attachment "schedtest.c" of type "TEXT/x-c++src" (593 bytes)