funny sockstat behavior in 4.5-R

[John Fox]

Hello,

I've got a rather strange thing occurring on my RADIUS server.
For all I know it could have been happening for some time,
because I have only recently begun the behavior that brought
this problem into the light.

In short, sockstat is behaving rather strangely.  Most of the 
time, there is no problem, but it on occassion exits, printing
only 'close(fstat):'.

I have more specific details below:

Uname data:
===========
> FreeBSD ego.xxxxx.net 4.5-STABLE FreeBSD 4.5-STABLE #0


Usual behavior:
===============
> [ego:/home/jjf/src/nagios/nsca/radiusd]# sockstat |grep radiusd
> root     radiusd  72559   10 udp4   *:1646                *:*                  
> root     radiusd  72559   16 udp4   *:1648                *:*                  
> root     radiusd  72558    9 udp4   *:1645                *:*                  
> root     radiusd  72558   10 udp4   *:1646                *:*                  
> root     radiusd  72558   13 udp4   *:1647                *:*                  
> root     radiusd  72558   16 udp4   *:1648                *:*                  
> root     radiusd  72555    9 udp4   *:1645                *:*                  
> root     radiusd  72555   13 udp4   *:1647                *:*                  

Strange behavior:
=================
> [ego:/home/jjf/src/nagios/nsca/radiusd]# sockstat |grep radiusd
> close(fstat): 

dmesg output:
=============
> pid 47410 (fstat), uid 0: exited on signal 10
> < SNIP >
> pid 12421 (fstat), uid 0: exited on signal 10


snapshot from 'top':
====================
> last pid: 25314;  load averages:  3.57,  4.30,  5.35 up 57+14:06:15  12:30:49
> 208 processes: 9 running, 198 sleeping, 1 zombie
> CPU states: 48.6% user,  0.0% nice, 48.1% system,  1.6% interrupt,  1.8% idle
> Mem: 303M Active, 70M Inact, 88M Wired, 22M Cache, 61M Buf, 18M Free
> Swap: 1024M Total, 30M Used, 993M Free, 2% Inuse


I've verified to myself that the fstat errors are related 
to the strange sockstat behavior; I ran the 'sockstat'
grep shown above again and again (about five times) until
I got the error condition.  I then checked dmesg, and
confirmed that new fstat errors had been appeneded to it.

I just ran sockstat 100 times in a row -- throwing away
its STDOUT, and got the following results:

> # for num in `jot 100 1 100`;do sockstat >/dev/null;done
> fstat: can't read mount table at 0x1800945
> close(fstat): 
> close(fstat): 
> close(fstat): 
> close(fstat): 
> close(fstat): 
> close(fstat): 
> Use of uninitialized value at /usr/bin/sockstat line 78, <PIPE> chunk 486.
> Use of uninitialized value at /usr/bin/sockstat line 79, <PIPE> chunk 486.
> Use of uninitialized value at /usr/bin/sockstat line 79, <PIPE> chunk 486.
> close(fstat): 
> fstat: can't read mount table at 0x1800945

Those 'unitialized value' lines clued me in that sockstat is a Perl script.
Perhaps my problem is related to Perl?

I have searched the listing of open bug reports for 'sockstat', but to 
no avail.  Any advice or suggestions would be most welcome.


Thank you,

John
-- 
+---------------------------------------------------------------------------+
| John Fox <jjf at mind.net>     |    System Administrator   | InfoStructure   |
+---------------------------------------------------------------------------+
|  "The people and friends that we have lost, the dreams that have faded... |
|  never forget them."  -- Yuna, Final Fantasy X                            |
+---------------------------------------------------------------------------+