ntpd(8) exits quietly after fork()

[Kevin Kinsey]

ntpd won't continue running on my "stratum 2" server (making our other 
systems a tad untrustworthy "as time goes by").

$ uname -a
FreeBSD ezekiel.daleco.biz 6.2-RELEASE FreeBSD 6.2-RELEASE #5: Sat Jan 
20 16:41:42 CST 2007 
root at ezekiel.daleco.biz:/usr/obj/usr/src/sys/GENERIC  i386

using ktrace and kdump I get this prior to exit:

  12195 ntpd     NAMI  "/var/run/logpriv"
  12195 ntpd     RET   connect 0
  12195 ntpd     CALL  sendto(0x4,0xbfbfdd50,0x3e,0,0,0)
  12195 ntpd     GIO   fd 4 wrote 62 bytes
        "<13>Mar 13 13:32:06 ntpd: logging to file /home/kadmin/ntp.log"
  12195 ntpd     RET   sendto 62/0x3e
  12195 ntpd     CALL  clock_gettime(0,0xbfbfeb08)
  12195 ntpd     RET   clock_gettime 0
  12195 ntpd     CALL  sigaction(0x1,0xbfbfeac0,0xbfbfeae0)
  12195 ntpd     RET   sigaction 0
  12195 ntpd     CALL  fork
  12195 ntpd     RET   fork 12196/0x2fa4
  12195 ntpd     CALL  exit(0)

After this, I can't `pgrep ntpd` or find a pid with "ps -aux".  I've 
found nothing in the logs; even using "-l $HOME/ntp.log".  A windows box 
complains "no response" using Softros' "NTS".  `ntpq -p" here shows

ntpq -p
      remote     refid      st t when poll reach   delay   offset  jitter
========================================================================
  ezekiel.daleco. .INIT.    16 u  148 1024    0    0.000    0.000 4000.00

I can confirm that ntpd will continue to run if called with "-n" (do not 
fork).  I've started a "buildworld" cycle.  Any other suggestions?  Can 
I do anything else "debugging" wise?  Machine is about as close to 
"production" as we get, so rebuild with a debug kernel isn't something 
we'd really want to do, I think....

Thanks,


Kevin Kinsey
-- 
Computer programmers do it byte by byte.