From owner-freebsd-arch  Tue Nov  7 14: 2: 9 2000
Delivered-To: freebsd-arch@freebsd.org
Received: from implode.root.com (root.com [209.102.106.178])
	by hub.freebsd.org (Postfix) with ESMTP id 306CE37B479
	for <arch@FreeBSD.ORG>; Tue,  7 Nov 2000 14:02:07 -0800 (PST)
Received: from implode.root.com (localhost [127.0.0.1])
	by implode.root.com (8.8.8/8.8.5) with ESMTP id NAA00641;
	Tue, 7 Nov 2000 13:59:34 -0800 (PST)
Message-Id: <200011072159.NAA00641@implode.root.com>
To: Poul-Henning Kamp <phk@critter.freebsd.dk>
Cc: Matt Dillon <dillon@earth.backplane.com>,
	Bruce Evans <bde@zeta.org.au>, Kirk McKusick <mckusick@mckusick.com>,
	arch@FreeBSD.ORG
Subject: Re: softdep panic due to blocked malloc (with traceback) 
In-reply-to: Your message of "Tue, 07 Nov 2000 20:49:45 +0100."
             <26624.973626585@critter> 
From: David Greenman <dg@root.com>
Reply-To: dg@root.com
Date: Tue, 07 Nov 2000 13:59:34 -0800
Sender: owner-freebsd-arch@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

>In message <200011071941.eA7JfZX81819@earth.backplane.com>, Matt Dillon writes:
>>:We are not talking garbage collection, we are talking dumping the
>>:vfs namecache for instance.
>>:
>>:--
>>:Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
>>
>>   Yes, but think a minute:  How does that prevent a low memory deadlock?
>>   The answer is not "because it frees up some pages" ... what if you've
>>   already freed up all the namecache pages you can and some process 
>>   continues to eat memory, and now you have none left to free?
>
>Matt, 
>
>I'll just rest the case here.  I can see from the swiftness of your
>replies that you are in "everything phk says is wrong by definition"
>mode right now.  Think about it and we can discuss it in a couple
>of weeks time maybe.

   Actually, I think Matt is entirely correct and his approach is the only one
that can effectively deal with low memory deadlocks. Basically, the I/O system
and the VM system need to be able to operate when memory is transiently tight
so that they are able to make more space available. Clearing out caches isn't
going to solve the problem for the reasons Matt already pointed out - you can't
be assured that enough cache pages will be available to resolve the problem.

-DG

David Greenman
Co-founder, The FreeBSD Project - http://www.freebsd.org
President, TeraSolutions, Inc. - http://www.terasolutions.com
Pave the road of life with opportunities.


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-arch" in the body of the message