oom - out of memory

oom - out of memory

[Kestutis Kupciunas]

hello, linux memory managers,

thing i am eager to clarify is oom, out of memory problem,
which doesn't work as it is supposed to (at least i think it
doesn't do the trick). Having the system fully utilizing all the
memory available on box and requesting more simply "hangs"
the box. 
Going into more details: i have noticed this behavior
with all 2.[23].x kernels i have used (not sure about the previous series).
usually problem arises when manipulating LARGE sets of large images
under X (with gimp, imagemagick tools). as i open more images, naturally,
memory/swap usage grows, and when it grows to the bounds, keyboard stops
responding, screen stops repainting, hdd led's going crazy. all box
services stop responding - i'm unable to connect from remote box. *RESET* :(
this behavior isnt my box specific - i've vitnessed it happening on
a bunch of different intels as well. The only chracteristics that apply
to all those boxes are that all of them are x86.
but according to the oom() function, the pid which is requesting
memory when it's out, is beeing killed with a message.
i didnt find any message in logs later...
im not a 'kernel hacker', so maybe somebody could analyze the lifecycle
of linux-mm memory allocating up to the bounds and over?
or is there something i don't get right?
sorry for the messy english


regards,
ydum
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://humbolt.geo.uu.nl/Linux-MM/

oom - out of memory

[David Kulp]

I have just the same problem with the same kernels: system hangs when
one process requires more than total RAM.  I can't kill processes or
otherwise get any response.  No syslog messages.  I don't know where
to start to try to track down this problem -- but I thought monitoring
this list would be a start.  Ironically -- considering the recent
'ammo' thread, I had no trouble with this in FreeBSD.    )-:

If someone needs specific reproducible test cases or other details,
I'd love to try and help.

-d

ps. This isn't an X problem as could possibly be hypothesized from the 
original post: this occurs when running non-X apps, too.

Kestutis Kupciunas writes:
 > hello, linux memory managers,
 > 
 > thing i am eager to clarify is oom, out of memory problem,
 > which doesn't work as it is supposed to (at least i think it
 > doesn't do the trick). Having the system fully utilizing all the
 > memory available on box and requesting more simply "hangs"
 > the box. 
 > Going into more details: i have noticed this behavior
 > with all 2.[23].x kernels i have used (not sure about the previous series).
 > usually problem arises when manipulating LARGE sets of large images
 > under X (with gimp, imagemagick tools). as i open more images, naturally,
 > memory/swap usage grows, and when it grows to the bounds, keyboard stops
 > responding, screen stops repainting, hdd led's going crazy. all box
 > services stop responding - i'm unable to connect from remote box. *RESET* :(
 > this behavior isnt my box specific - i've vitnessed it happening on
 > a bunch of different intels as well. The only chracteristics that apply
 > to all those boxes are that all of them are x86.
 > but according to the oom() function, the pid which is requesting
 > memory when it's out, is beeing killed with a message.
 > i didnt find any message in logs later...
 > im not a 'kernel hacker', so maybe somebody could analyze the lifecycle
 > of linux-mm memory allocating up to the bounds and over?
 > or is there something i don't get right?
 > sorry for the messy english
 > 
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://humbolt.geo.uu.nl/Linux-MM/

Re: oom - out of memory

[Andrea Arcangeli]

On Thu, 23 Sep 1999, David Kulp wrote:

>I have just the same problem with the same kernels: system hangs when
>one process requires more than total RAM.  I can't kill processes or
>otherwise get any response.  No syslog messages.  I don't know where
>to start to try to track down this problem -- but I thought monitoring
>this list would be a start.  Ironically -- considering the recent
>'ammo' thread, I had no trouble with this in FreeBSD.    )-:

I fixed a potential oom deadlock and some omm related bug in the VM at
2.2.10 time. Unfortunately it's not been merged in the stock kernel yet
(it also breaks all architectures and I fixed only i386 and Alpha as I
don't have access to other hardware).

Please apply this my fix and give us back any positive/negative feedback.

	ftp://ftp.suse.com/pub/people/andrea/kernel-patches/2.2.12/oom-2.2.12-I

Thanks!

Andrea

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://humbolt.geo.uu.nl/Linux-MM/