[3221] in testers
8.1.4 sparc: non-/-cwd system processes
daemon@ATHENA.MIT.EDU (John Hawkinson)
Thu Jun 12 21:35:06 1997
Date: Thu, 12 Jun 1997 21:34:47 -0400
To: testers@MIT.EDU
From: John Hawkinson <jhawk@MIT.EDU>
Under 8.1.4, the following system processes seem to have non-/ working
directories:
# /usr/proc/bin/pwdx `ps -fu 0,1 | awk 'NR!=1{print $2}'` 2> /dev/null | egrep -v '/$|/.*system process'
126: /etc
157: /var/spool/cron/atjobs
255: /afs/dev.mit.edu/system/sun4x_55/os/usr/lib/vold
3583: /afs/athena.mit.edu/user/j/h/jhawk
3708: /afs/athena.mit.edu/user/j/h/jhawk
3722: /afs/athena.mit.edu/user/j/h/jhawk
3721: /afs/athena.mit.edu/user/j/h/jhawk
[bart-savagewood!jhawk] ~# ps -lp126,157,255
F S UID PID PPID C PRI NI ADDR SZ WCHAN TTY TIME CMD
8 S 0 126 1 0 41 20 f5d87cc8 405 f596c1c6 ? 1:33 in.named
8 S 0 157 1 0 41 20 f5edccd0 345 f5edcd40 ? 0:02 cron
8 S 0 255 1 0 61 20 f5f07998 506 f5ecfade ? 0:03 vold
This is problematic if they core dump.
named has /etc because of the presence of "directory /etc" in /etc/named.boot.
I think that's reasonable, so a /etc/core symlink should be added,
pointing at either /core or /var/athena/core (not sure which we'd like to
enforce here, they should point to the same place, of course, once
/core is changed to point to /var/athena/core).
I'm not sure what to think about cron and vold. I assume that this is
desired behavior on their part. A bug report should be submitted to
Sun regarding the fact they do not document the fact that cron and
vold chdir() silently. It should be referenced in the man page,
otherwise users cannot know where to expect core dumps.
--jhawk
