NetBSD Problem Report #469
From gnats Thu Sep 8 14:51:45 1994
Received: from python.CS.ORST.EDU (root@python.CS.ORST.EDU [128.193.164.60]) by sun-lamp.cs.berkeley.edu (8.6.9/8.6.9) with ESMTP id OAA01185 for <gnats-bugs@sun-lamp.cs.berkeley.edu>; Thu, 8 Sep 1994 14:51:44 -0700
Message-Id: <199409081838.LAA15515@helix.CS.ORST.EDU>
Date: Thu, 8 Sep 1994 11:38:15 -0700
From: "Jason R. Thorpe" <thorpej@mail.cs.orst.edu>
Reply-To: thorpej@mail.cs.orst.edu
To: gnats-bugs@sun-lamp.cs.berkeley.edu
Subject: mountd cannot grok duplicate hosts in exports list
X-Send-Pr-Version: 3.2
>Number: 469
>Category: bin
>Synopsis: mountd cannot grok duplicate hosts in exports list
>Confidential: no
>Severity: serious
>Priority: high
>Responsible: gnats-admin
>State: closed
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Thu Sep 08 15:05:08 +0000 1994
>Closed-Date: Sun Jul 20 23:33:16 +0000 1997
>Last-Modified: Sun Jul 20 23:44:50 +0000 1997
>Originator: Jason R. Thorpe
>Release: NetBSD 1.0_BETA (sup'd ~9.4.94)
>Organization:
Department of Computer Science
Oregon State University
>Environment:
System: NetBSD helix 1.0_BETA NetBSD 1.0_BETA (HELIX) #15: Tue Sep 6 14:54:16 PDT 1994 thorpej@helix:/usr/src/sys/arch/hp300/compile/HELIX hp300
>Description:
If, in /etc/exports, there is a duplicate hostname (or duplicate ip
resolved from a hostname), the the entire export of that filesystem
fails. This will probably bite most people when exporting to
netgroups, especially if each netgroup has two entries for each host
in the netgroup (to work around a bug in a *really* old NFS
implementation). However, it will also bite people if a filesystem
is exported to a netgroup that is a partial subset of another
netgroup that the filesystem is also exported to.
>How-To-Repeat:
In my case:
(output from mountd compiled with -DDEBUG)
helix# obj.hp300/mountd
Getting export list.
Got line /usr/src -maproot=root INST-HPS
Making new ep fs=0x42f,0x6bb
doing opt -maproot=root INST-HPS
got host lespaul.CS.ORST.EDU
got host gibson.CS.ORST.EDU
got host strat.CS.ORST.EDU
got host fender.CS.ORST.EDU
got host typhoon.CS.ORST.EDU
got host santiam.CS.ORST.EDU
got host prism.CS.ORST.EDU
got host helix.CS.ORST.EDU
got host xanth.CS.ORST.EDU
Getting mount list.
Here we go.
...all is ok here...but...
helix# obj.hp300/mountd
Getting export list.
Got line /usr/src -maproot=root INST-HPS NETBSD
Making new ep fs=0x42f,0x6bb
doing opt -maproot=root INST-HPS NETBSD
got host lespaul.CS.ORST.EDU
got host gibson.CS.ORST.EDU
got host strat.CS.ORST.EDU
got host fender.CS.ORST.EDU
got host typhoon.CS.ORST.EDU
got host santiam.CS.ORST.EDU
got host prism.CS.ORST.EDU
got host helix.CS.ORST.EDU
got host xanth.CS.ORST.EDU
got host helix.CS.ORST.EDU
got host santiam.CS.ORST.EDU
got host facelift.CS.ORST.EDU
Can't change attributes for /usr/src.
Bad exports list line /usr/src -maprootGetting mount list.
Here we go.
...notice that helix.CS.ORST.EDU and santiam.CS.ORST.EDU are
duplicates. Also notice that whis makes the export of the
entire filesystem fail.
The same also happens if a machine as two names (that both appear in
a netgroup) that resolve to the same IP. In my case, a machine called
viper (which has two network interfaces) has three names:
viper.cs.orst.edu
viper32.cs.orst.edu
viper164.cs.orst.edu
viper.cs.orst.edu resolves to the same ip as viper32.cs.orst.edu.
If both viper and viper32 are in the netgroup (INST-SUNS), then
the export will fail.
>Fix:
Perhaps the easiest (and a fairly fast way) to fix the problem would
be to build a hash table (in memory or on disk - tmp file) of
IPs that the filesystem is being exported to. So, everytime
a host was to be added to the list, a hash lookup could be done
on the ip that the name resolves to. If you get a hit, simply
ignore the ip/name and continue.
Another (and perhaps more expensive but maybe more correct) way
would be to look for the ip in the kernel's mount list. Again, skip
the ip if it's found.
Once again, I haven't had enough coffee to come up with any code
at this time. But I've got it brewing...
The problem here is a little strange. Basically, mountd(8) causes
the kernel to execute vfs_hang_addrlist(). This calls rn_addroute()
to add radix nodes for each IP address. If an existing entry is
found, rn_addroute() returns a NULL pointer, indicating that. When
this happens, vfs_hang_addrlist() returns EPERM. It would make
more sense for it to check the existing node, and if it's the same,
return 0. I have a potential fix that does this, but I'm awaiting
comments from other parties. - mycroft
>Release-Note:
>Audit-Trail:
State-Changed-From-To: open->analyzed
State-Changed-By: mycroft
State-Changed-When: Sun Sep 18 15:04:03 PDT 1994
State-Changed-Why:
I've investigated the problem.
State-Changed-From-To: analyzed->closed
State-Changed-By: fvdl
State-Changed-When: Sun Jul 20 16:33:16 PDT 1997
State-Changed-Why:
Fixed by allowing duplicate exports in the kernel iff they're the same.
After 3 years of hard debugging. Really!
>Unformatted:
(Contact us)
$NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.