From owner-freebsd-bugs  Fri Apr 18 04:20:47 1997
Return-Path: <owner-bugs>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id EAA25680
          for bugs-outgoing; Fri, 18 Apr 1997 04:20:47 -0700 (PDT)
Received: from dg-rtp.dg.com (dg-rtp.rtp.dg.com [128.222.1.2])
          by freefall.freebsd.org (8.8.5/8.8.5) with SMTP id EAA25661
          for <freebsd-bugs@freefall.freebsd.org>; Fri, 18 Apr 1997 04:20:43 -0700 (PDT)
Received: by dg-rtp.dg.com (5.4R3.10/dg-rtp-v02)
	id AA05162; Fri, 18 Apr 1997 07:20:09 -0400
Received: from ponds by dg-rtp.dg.com.rtp.dg.com; Fri, 18 Apr 1997 07:20 EDT
Received: from lakes.water.net (lakes [10.0.0.3]) by ponds.water.net (8.8.3/8.7.3) with ESMTP id GAA10496; Fri, 18 Apr 1997 06:41:37 -0400 (EDT)
Received: (from rivers@localhost) by lakes.water.net (8.8.3/8.6.9) id GAA07390; Fri, 18 Apr 1997 06:48:17 -0400 (EDT)
Date: Fri, 18 Apr 1997 06:48:17 -0400 (EDT)
From: Thomas David Rivers <ponds!rivers@dg-rtp.dg.com>
Message-Id: <199704181048.GAA07390@lakes.water.net>
To: ponds!labinfo.iet.unipi.it!luigi, ponds!lakes.water.net!rivers
Subject: Re: kern/3304: NFS V2 readdir hangs
Cc: ponds!freefall.freebsd.org!freebsd-bugs
Content-Type: text
Sender: owner-bugs@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

Luigi Rizzo writes:
> 
> Hi,
> 
> I followed the interesting (and very detailed) analysis of the NFS
> problem (PR kern/3304), and I have a possible patch.
> 
> looking at nfs_rcvlock() is implemented, I notice that the only way to
> return EALREADY is to have rep->r_mrep != NULL
> 
> The variable comes (via nfs_reply() ) from nfs_request() where the code
> is something like the following (I have reported only a few relevant
> lines, from a 2.2.1-R system:
> 
>     $Id: nfs_socket.c,v 1.18 1996/10/11 10:15:33 dfr Exp $
> 
> 
>  923:	tryagain:
>  ...
>  933:		rep->r_mrep = NULL;
>  ...
>  979:			error = nfs_reply(rep);
>  ...
> 1064:                                goto tryagain;
> 
> the problem is, r_mrep seems only to be set to something != NULL in
> nfs_reply(). While I don't understand the code very well, it seems
> to me that the initialization in line 933 should be moved outside
> the retry loop, before line 923. Otherwise the locking cannot possibly
> work because repeated calls to nfs_reply() use the same parameters.

 Well - what good eyes you have :-)  I didn't notice that...


> 
> Can someone (Dave ?) try the above fix and see if it works ?

 I'll give it a try later this morning (my time, EST ) and let you know.

	- Dave Rivers -