From owner-freebsd-bugs  Mon Jun 26 11:31:34 1995
Return-Path: bugs-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id LAA27109
          for bugs-outgoing; Mon, 26 Jun 1995 11:31:34 -0700
Received: from gvr.win.tue.nl (root@gvr.win.tue.nl [131.155.210.19])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id LAA27099
          for <bugs@freebsd.org>; Mon, 26 Jun 1995 11:31:28 -0700
Received: by gvr.win.tue.nl (8.6.10/1.53)
    id UAA07287; Mon, 26 Jun 1995 20:31:07 +0200
From: guido@gvr.win.tue.nl (Guido van Rooij)
Message-Id: <199506261831.UAA07287@gvr.win.tue.nl>
Subject: Re: (fwd) "kernel: mb_map full" crash (fwd)
To: davidg@Root.COM
Date: Mon, 26 Jun 1995 20:31:06 +0200 (MET DST)
Cc: Andrew@dca.net, bugs@freebsd.org
In-Reply-To: <199506260641.XAA00657@corbin.Root.COM> from "David Greenman" at Jun 25, 95 11:41:55 pm
X-Mailer: ELM [version 2.4 PL24]
Content-Type: text
Content-Length: 781       
Sender: bugs-owner@freebsd.org
Precedence: bulk

David Greenman wrote:
> 
>    It's caused by an insufficient number of mbuf clusters being available. You
> can increase the mbuf cluster pool size with options "NMBCLUSTERS=1024" (or
> 2048 or whatever if necessary; each cluster takes 2K of memory). Newer
> versions of FreeBSD (including 2.0.5R), have a more dynamic calculation for
> this which is based on maxusers.

Is it known why this crashes the machine? The code itsself seems to dynamically
adjust the number of mbufs when it's out of mbuf space. I recently
spotted a SunOS 4.1.4 (or what is the latest 4.1.y version) also behaves
badly when it printf's the message that it's had to drop a packet because
of an mbuf shortage. Btw: untill it does the printf, it seems the number
of mbufs is succesfully increased.

-Guido