home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.unix.questions:15103 comp.sys.next.sysadmin:7274
- Newsgroups: comp.unix.questions,comp.sys.next.sysadmin
- Path: sparky!uunet!spool.mu.edu!sol.ctr.columbia.edu!usenet.ucs.indiana.edu!teg
- From: teg@usenet.ucs.indiana.edu (Tim Gurbick)
- Subject: Re: Help: syslogd crash
- Message-ID: <C07A7x.Dz5@usenet.ucs.indiana.edu>
- Followup-To: comp.sys.next.sysadmin
- Summary: same here, older OS
- Keywords: syslogd,core dump,syslog.conf,broken
- Sender: teg@indiana.edu
- Organization: Basil Fawlty Knocking Shops, Ltd.
- References: <1992Dec23.140758.13105@puug.pt>
- Date: Sat, 2 Jan 1993 00:19:08 GMT
- Lines: 47
-
- In article <1992Dec23.140758.13105@puug.pt> artur@morgaine.puug.pt (Artur Romao) writes:
- >I have the new (?) syslog, with the levels and facilities scheme. I added
- >local7 level to /etc/syslogd.conf some days ago, then sent kill -HUP to
- >syslogd and all went well. Today I added local6, and syslogd answered with
- >a beautiful core dump :-( to kill -HUP.
- >I'm on a NeXT, if that's important...
-
- I've just recently noticed similar behaviour on my NeXTcube running 2.1.
- Rebooting fails in /etc/rc because it freezes right after the "starting
- early daemons" line. Cutting the relevant code out of rc and feeding it
- to sh manually works fine;
-
- if [ -f /usr/etc/syslogd -a -f /etc/syslog.conf ]; then
- (echo 'works')
- fi
-
- prints "works" as expected.
-
- The REALLY weird part is that running syslogd by hand (as root) works just
- fine, and the system goes-on merrily. I can ctrl-C /etc/rc at the time it
- freezes, and it indicates syslog went down on a signal 2 Bad File Number,
- and manually executing syslogd -d generates no complaints at all. /dev/log
- is created at run time, and /dev/klog never goes-away. I can also sighup
- my running syslogd and it continues to work.
-
- The posibly relevant bit of source I looked at was:
-
- /*
- * Check for errors on TTY's due to loss of tty
- */
- if ((e == EIO || e == EBADF) && f->f_type != F_FILE) {
- f->f_file = open(f->f_un.f_fname,
- O_WRONLY|O_APPEND, 0);
- if (f->f_file < 0) {
- f->f_type = F_UNUSED;
- logerror(f->f_un.f_fname);
-
- but I can think of no reason why it would work for almost a year without
- complaints and then start dying suddenly. The running syslogd (launched
- by hand) has no controlling tty according to /bin/ps, but considerably
- more environment attached (via ps -ew). Anybody know if there's anything
- funny going-on with netinfo here? I can't see anything wrong from a unix
- standpoint, which is why followups should probably go to c.s.n.s. (that
- and the fact that I don't read comp.unix.questions).
-
- Thanks in advance
-
-