Re: 2.2.12 Memory Leak?

Robert Dinse (nanook@eskimo.com)
Wed, 1 Sep 1999 06:57:15 -0700 (PDT)


On Wed, 1 Sep 1999, Andrea Arcangeli wrote:

> Date: Wed, 1 Sep 1999 14:43:46 +0200 (CEST)
> From: Andrea Arcangeli <andrea@suse.de>
> To: Robert Dinse <nanook@eskimo.com>
> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>, ericj@eskimo.com, evol@eskimo.com,
> linux-kernel@vger.rutgers.edu
> Subject: Re: 2.2.12 Memory Leak?
>
> On Wed, 1 Sep 1999, Robert Dinse wrote:
>
> > But with 2.2.12, free mem just keeps decreasing, then swapd starts to
> >increase, eventually it gets to a point where anything you try to do says
> >"killed" and all you can do is reboot and start over.
>
> So basically you meant that you have a buffer-leakage. Could you press
> SYSRQ+M and show me the state of the buffers?
>
> Maybe you may want to give a try also to this my two other patches:
>
> ftp://ftp.suse.com/pub/people/andrea/kernel-patches/my-2.2.12/no-swapout-2.2.10-B
> ftp://ftp.suse.com/pub/people/andrea/kernel-patches/my-2.2.12/shrink_all_cache-2.2.10-A
>
> they should to be applyed at the same time.
>
> If the above patches won't make difference it means somebody forget to
> release some buffer or something similar.
>
> If you can reproduce on x86 then you can debug the thing with the ikd
> patch.
>
> ftp://ftp.suse.com/pub/people/andrea/kernel-patches/ikd/2.2.12-ikd1.gz
>
> Andrea

I can't right now because I didn't enable the key when I built the kernel.
At a later time I will build a kernel with that enabled and try those patches,
at least if 2.2.13pre3 with fixes to address same isn't already out.

At this point I've gone back to 2.2.10, but applied the Sparc assembly
patch that Dave sent me for 2.2.12 since the assembly was same, and see what
happens with that. I've got to get some sleep, been up all night. Crash on the
news server is a real painful experience owing to the hour-long fsck that
follows each time.

Even with your schedualer patch, 2.2.12 also seems noticably slower than
2.2.10 was with same, even before it manages to run itself out of memory.

It would be my obvious preference to work as stay as close to the
distributed tree and current and get the bugs worked out in that so that the
fixes are propogated to future releases, but right now I just have to get A
stable 2.2.x kernel so that the machines will at least stay up and running when
I'm not actively babysitting them.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/