From owner-freebsd-hackers  Fri Feb 28 13:58:08 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id NAA20334
          for hackers-outgoing; Fri, 28 Feb 1997 13:58:08 -0800 (PST)
Received: from plains.nodak.edu (plains.NoDak.edu [134.129.111.64])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id NAA20326
          for <hackers@freebsd.org>; Fri, 28 Feb 1997 13:58:05 -0800 (PST)
Received: (from tinguely@localhost) by plains.nodak.edu (8.8.4/8.8.3) id PAA20337; Fri, 28 Feb 1997 15:57:36 -0600 (CST)
Date: Fri, 28 Feb 1997 15:57:36 -0600 (CST)
From: Mark Tinguely <tinguely@plains.nodak.edu>
Message-Id: <199702282157.PAA20337@plains.nodak.edu>
To: dg@root.com, terry@lambert.org
Subject: Re: Assumptions about kmem_malloc()...
Cc: hackers@freebsd.org, jpt@msc.edu
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

>  Why doesn't it push pages around to get a contiguous space so that
>  it won't fail?

that would work if everything is swappable. It is hard to fight those holes
when drivers that need wired contigous memory come and go through lkm
loads/unloads. 

Most driver allocations would puke if their memory got moved unless we go
to a double pointer arrangement, but (aaaaaaaaaahhhhhgg) I would not have
to be the one to have to debug VM problems).

--mark.