From owner-freebsd-current  Thu Apr 10 14:26:24 1997
Return-Path: <owner-current>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id OAA01435
          for current-outgoing; Thu, 10 Apr 1997 14:26:24 -0700 (PDT)
Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.50])
          by freefall.freebsd.org (8.8.5/8.8.5) with SMTP id OAA01426;
          Thu, 10 Apr 1997 14:26:18 -0700 (PDT)
Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id OAA09726; Thu, 10 Apr 1997 14:06:08 -0700
From: Terry Lambert <terry@lambert.org>
Message-Id: <199704102106.OAA09726@phaeton.artisoft.com>
Subject: Re: NFS/mmap freeze in 2.2R
To: kachun@zippo.com (Kachun Lee)
Date: Thu, 10 Apr 1997 14:06:08 -0700 (MST)
Cc: dyson@freebsd.org, freebsd-current@freebsd.org
In-Reply-To: <199704101748.KAA25163@dopey.pathlink.com> from "Kachun Lee" at Apr 10, 97 10:38:21 am
X-Mailer: ELM [version 2.4 PL24]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-current@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

> While you are at it, maybe, you want to check this out also. We noticed we have 
> been getting these messages:
> 
> -------------
> Apr  8 23:21:12 lois /kernel: vnode_pager_getpages: I/O read error
> Apr  8 23:21:12 lois /kernel: vm_fault: pager input (probably hardware) error, 
> P
> ID 11987 failure
> Apr  8 23:21:12 lois /kernel: pid 11987 (in.nnrpd), uid 7003: exited on signal 
> 11
> --------------
> 
> since 2.2R or later with our NFS mounted news servers. They appeared at
> the NFS clients during news was expiring at the NFS server. That leaded
> me to believe that the message was triggered when someone was reading
> an article while the NFS server removed it - both systems kept running
> though.

Are you running the INN msync() patch?  There is an error in the INN
use of mapped files when the file is extended.  The problem occurs when:

1)	The file is not a multiple of a page size in length
2)	The file is mmap'ed, and the last page is zero filled
3)	The file is extended
4)	The mmap'ed area is accessed between the previous end of the
	file and the page boundry following the end of the file, without
	an intervening msync().

This is a bug in INN's use of the mmap() interface, not a FreeBSD bug.


(If you are running the msync() patch, then never mind).


					Regards,
					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.