From owner-freebsd-bugs  Thu Dec 30 11:38:52 1999
Delivered-To: freebsd-bugs@freebsd.org
Received: from apollo.backplane.com (apollo.backplane.com [216.240.41.2])
	by hub.freebsd.org (Postfix) with ESMTP
	id 15619151A2; Thu, 30 Dec 1999 11:38:42 -0800 (PST)
	(envelope-from dillon@apollo.backplane.com)
Received: (from dillon@localhost)
	by apollo.backplane.com (8.9.3/8.9.1) id LAA77035;
	Thu, 30 Dec 1999 11:38:41 -0800 (PST)
	(envelope-from dillon)
Date: Thu, 30 Dec 1999 11:38:41 -0800 (PST)
From: Matthew Dillon <dillon@apollo.backplane.com>
Message-Id: <199912301938.LAA77035@apollo.backplane.com>
To: peter@FreeBSD.ORG, assar@sics.se, freebsd-bugs@FreeBSD.ORG
Subject: Re: kern/15707: bad trap in mprotect
References:  <199912301906.LAA50571@freefall.freebsd.org>
Sender: owner-freebsd-bugs@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.org


:
:Synopsis: bad trap in mprotect
:
:Responsible-Changed-From-To: freebsd-bugs->dillon
:Responsible-Changed-By: peter
:Responsible-Changed-When: Thu Dec 30 11:04:49 PST 1999
:Responsible-Changed-Why: 
:Matt, can you take a look at this please?

    Sure.  I may not be able to get to it until early january.  I can
    make a few suggestions to Assar in the mean time:  First thing to
    do is to turn off DIAGNOSTIC in your kernel config.  It appears
    to be there in two places so be sure you comment out both.

    I doubt that that DIAGNOSTIC is the problem, but it has been known to
    cause problems.  ElectricFence makes very heavy use of the VM system by
    allocating two pages around each malloc().  Due to the number of 
    vm_map_entry's that will be created, it is possible that the kernel 
    is running out of KVM, so please also check:

	apollo:/usr/src/sys/vm# vmstat -z
	...

    and

	apollo:/usr/src/sys/vm# vmstat -m
	...

    While you are running your test, run the above command once a second
    or so.  The panic is definitely a bug - at the very least it should
    tell us if its running out resources.

    However, since the fault address is non-NULL, there is a good chance
    that the huge size of the map_entry list is causing a previously 
    unknown race condition to rear its ugly head.  Most likely you have
    found a real bug somewhere in the code.

    I should be able to reproduce the crash when I have time in January
    to take up the problem.

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>


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