From owner-freebsd-hackers  Mon Jul 14 17:03:56 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id RAA23135
          for hackers-outgoing; Mon, 14 Jul 1997 17:03:56 -0700 (PDT)
Received: from zephyr.isi.edu (zephyr.isi.edu [128.9.160.160])
          by hub.freebsd.org (8.8.5/8.8.5) with SMTP id RAA23130;
          Mon, 14 Jul 1997 17:03:49 -0700 (PDT)
Received: from roo.isi.edu by zephyr.isi.edu (5.65c/5.61+local-26)
	id <AA14675>; Mon, 14 Jul 1997 17:03:04 -0700
Message-Id: <199707150003.AA14675@zephyr.isi.edu>
X-Mailer: exmh version 1.6.4 10/10/95
To: dyson@FreeBSD.ORG
Cc: hackers@FreeBSD.ORG
Reply-To: hutton@ISI.EDU
Subject: Re: vm_page_alloc_contig again 
In-Reply-To: Your message of Mon, 14 Jul 1997 18:45:04 -0500.
             <199707142345.SAA00363@dyson.iquest.net> 
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Mon, 14 Jul 97 17:02:03 PDT
From: Anne Hutton <hutton@ISI.EDU>
Sender: owner-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk


> vm_page_alloc_contig works best before the system is fully up.  It is used
> to allocate medium sized contiguous chunks at bootup time.
> 

we use vm_page_alloc_contig in a device driver...so it is at boot time and it 
is a small chunk - 1/2 meg.

It allocates ok and the device driver works but we suspect that the pages 
aren't (really) wired down. We see a page fault and the machine panics when we 
are forwarding packets through the NIC from another NIC on the same host. Is 
there a known problem with the pages not being wired down once allocated by 
vm_page_alloc_contig at boot time?


Anne
 

> We don't have a very reliable method for allocating large amounts of
> contiguous memory in the kernel.  If you don't need contiguous memory,
> then there are other ways to allocate virtually contiguous memory.  The
> simplest is kernel malloc.  The memory will be virtually contiguous,
> but will not likely be physically contiguous.
>