Re: PROBLEM: oom killer and swap weirdness on 2.6.3* kernels

From: dave b
Date: Mon Jul 26 2010 - 10:06:20 EST


On 2 June 2010 06:52, David Rientjes <rientjes@xxxxxxxxxx> wrote:
> On Thu, 27 May 2010, dave b wrote:
>
>> That was just a simple test case with dd. That test case might be
>> invalid - but it is trying to trigger out of memory - doing this any
>> other way still causes the problem. I note that playing with some bios
>> settings I was actually able to trigger what appeared to be graphics
>> corruption issues when I launched kde applications ... nothing shows
>> up in dmesg so this might just be a conflict between xorg and the
>> kernel with those bios settings...
>>
>> Anyway, This is no longer a 'problem' for me since I disabled
>> overcommit and altered the values for dirty_ratio and
>> dirty_background_ratio - and I cannot trigger it.
>>
>
> Disabling overcommit should always do it, but I'd be interested to know if
> restoring dirty_ratio to 40 would help your usecase.
>
Actually it turns out on 2.6.34.1 I can trigger this issue. What it
really is, is that linux doesn't invoke the oom killer when it should
and kill something off. This is *really* annoying.

I used the follow script - (on 2.6.34.1)
cat ./scripts/disable_over_commit
#!/bin/bash
echo 2 > /proc/sys/vm/overcommit_memory
echo 40 > /proc/sys/vm/dirty_ratio
echo 5 > /proc/sys/vm/dirty_background_ratio

And I was still able to reproduce this bug.
Here is some c code to trigger the condition I am talking about.


#include <stdlib.h>
#include <stdio.h>

int main(void)
{
while(1)
{
malloc(1000);
}

return 0;
}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/