Subject: kern/30077: anything accesses filesystem sleeps forever on vnlock, on nfs server
To: None <kern-bug-people@netbsd.org, gnats-admin@netbsd.org,>
From: None <carton@Ivy.NET>
List: netbsd-bugs
Date: 04/28/2005 00:29:01
>Number:         30077
>Category:       kern
>Synopsis:       anything accesses filesystem sleeps forever on vnlock, on nfs server
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Apr 28 00:29:00 +0000 2005
>Originator:     Miles Nordin
>Release:        NetBSD 2.0_RC5
>Organization:
Levin had long ago observed that when people made you feel
uncomfortable by their excessive tractability and submissiveness, they
would soon make your life unbearable by being excessively demanding
and carping.	-- Tolstoy, _Anna Karenina_
>Environment:
System: NetBSD castrovalva 2.0_RC5 NetBSD 2.0_RC5 (CASTROVALVA-$Revision: 1.8 $) #2: Tue Feb 15 19:00:10 EST 2005 carton@castrovalva:/scratch/src/sys/arch/alpha/compile/CASTROVALVA alpha
Architecture: alpha
Machine: alpha
>Description:
a filesystem, /export, is ``frozen'' meaning anything that accesses it 
sleeps forever in vnlock and can't be killed.  I can run 'disklabel' on 
the disk that holds /export with no problems, and there are no errors in 
dmesg about the disk itself.  other filesystems are fine.

you can see three 'find's left behind by the daily script.

castrovalva:/usr/home/carton$ mount
/dev/sd0a on / type ffs (soft dependencies, local)
/dev/sd0d on /usr type ffs (soft dependencies, local)
mfs:435 on /tmp type mfs (synchronous, nodev, local)
/dev/sd1d on /scratch type ffs (nosuid, nodev, soft dependencies, NFS exported, local)
/dev/sd2d on /export type ffs (nosuid, nodev, soft dependencies, NFS exported, local)
castrovalva:/usr/home/carton$ sudo kill -9 465
castrovalva:/usr/home/carton$ ps alxw | grep nfsd
   0   465     1   0 -18  0    88   448 uvn_fp2  DL   ??   8:16.61 nfsd: server
 405 22495  8059   1   2  0   232   888 netio    S+   p7   0:00.01 grep nfsd
castrovalva:/usr/home/carton$ ps alxw | grep vnlock
   0   437     1   0  -2  0   456     0 vnlock   DWs  ??   0:09.62 /usr/sbin/mountd
   0  7498 18581   2  -2  0   256     0 vnlock   DW   ??   0:44.08 find / ( ! -fstype local -o -fstype rdonly -o -fstype fdesc -o -
   0 24462  6339   4  -2  0   248     0 vnlock   DW   ??   0:42.62 find / ( ! -fstype local -o -fstype rdonly -o -fstype fdesc -o -
   0 29164 10187   2  -2  0   256     0 vnlock   DW   ??   0:44.14 find / ( ! -fstype local -o -fstype rdonly -o -fstype fdesc -o -
 405 20344  8059   3   2  0  1104     0 netio    R+   p7   0:00.00 grep vnlock (ksh93)
   0 26367  7705   0  -2  0    32     0 vnlock   DW+  q0   0:00.05 umount -f /export
 405 27368  7705   0  -2  4    88     0 vnlock   DWN  q0   0:00.01 ls -l /export 
castrovalva:/usr/home/carton$ dmesg | tail
nfsd send error 65
nfsd send error 65
nfsd send error 65
nfsd send error 65
nfsd send error 65
nfsd send error 65
nfsd send error 65
nfsd send error 65
nfsd send error 65
tlp0: transmit underrun; new threshold: 96/256 bytes
castrovalva:/usr/home/carton$
>How-To-Repeat:
box has been up for 67 days.  it's used as an NFS server.  the error 65 
``host unreachable'' is probably caused by quagga removing routes to an 
NFS client.  i'll try to repeat it after rebooting but am not hopeful.
>Fix:
unknown