Subject: /etc/daily gets stuck
To: None <port-alpha@netbsd.org>
From: Paul Mather <paul@gromit.dlib.vt.edu>
List: port-alpha
Date: 01/11/2001 09:06:31
For about a month or so, an annoying intermittent problem has plagued my
NetBSD/alpha system. I'm running 1.5-current (1.5.1_ALPHA, built 1st
Jan. 2001 as of now) on a DEC 3000/300. The problem is that /etc/daily
never finishes. The find command to hunt out possible core files never
terminates. It is flagged as "D" by ps. Here is a typical output:
# ps alx
UID PID PPID CPU PRI NI VSZ RSS WCHAN STAT TT TIME COMMAND
0 214 1 0 10 0 320 960 nanoslee Is ?? 0:00.94 /usr/sbin/cron
[...]
0 1054 214 0 2 0 320 1000 netio I ?? 0:00.01 /USR/SBIN/CRON (
0 1055 1054 0 10 0 1352 376 wait Is ?? 0:00.05 /bin/sh -c /bin/
0 1056 1055 3 10 0 1424 456 wait I ?? 0:00.14 /bin/sh /etc/dai
0 1057 1055 0 2 0 56 752 netio I ?? 0:00.07 tee /var/log/dai
0 1058 1055 0 2 0 1416 1976 netio I ?? 0:00.36 sendmail -t
0 1063 1056 9 -2 0 264 1136 vnlock D ?? 2:15.55 find / (! -fsty
[...]
# fstat -v -p 1063
root find 1063 wd / 24398 drwxr-xr-x 512 r
root find 1063 0* unix stream fffffe00000f9080
root find 1063 1 /tmp 3969 -rw------- 0 w
root find 1063 2* unix stream fffffe00000e3d80 <-> fffffe00000f9600
root find 1063 3 /tmp 3968 drwx------ 512 r
root find 1063 4 /tmp 3968 drwx------ 512 r
root find 1063 5 / 24398 drwxr-xr-x 512 r
# fstat -nv -p 1063
root find 1063 wd 8,0 24398 40755 512 r
root find 1063 0* unix stream fffffe00000f9080
root find 1063 1 255,0 3969 100600 0 w
root find 1063 2* unix stream fffffe00000e3d80 <-> fffffe00000f9600
root find 1063 3 255,0 3968 40700 512 r
root find 1063 4 255,0 3968 40700 512 r
root find 1063 5 8,0 24398 40755 512 r
If I read the above correctly, find has inode 24398 open, which, it
says, is the root directory (/). However, I'm told that / is inode 2 by
ls:
# ls -lid /
2 drwxr-xr-x 20 root wheel 1024 Dec 29 16:15 /
I know ls can tell me the inode corresponding to a pathname. Is there a
command to do the reverse: to tell me the pathname corresponding to an
inode number?
Here are my currently mounted devices:
# mount
/dev/sd0a on / type ffs (local)
mfs:29 on /tmp type mfs (asynchronous, local)
/dev/sd0d on /usr type ffs (NFS exported, local, soft dependencies)
/dev/sd0e on /home type ffs (local, soft dependencies)
/dev/sd1c on /scratch type ffs (local)
Although /usr is flagged as NFS exported, no hosts are currently NFS
mounting it. (I sometimes mount /usr/src and /usr/pkgsrc on my
DECstation 3100.)
Lastly, what is state "vnlock"? I'm assuming it is an attempt to lock a
pathname component. FWIW, before I rebuilt userland recently, I got a
few occasional lockup/panic reboots after going to the 1.5.1_ALPHA
kernel. When I was able to catch the message, it said: "panic: lockmgr:
sleep/spin mismatch".
Any suggestions?? I'd like to fix this, because it's an annoyance not
to get /etc/daily reports (particularly regarding possible core dumps
lying around).
Cheers,
Paul.
e-mail: paul@gromit.dlib.vt.edu
"Without music to decorate it, time is just a bunch of boring production
deadlines or dates by which bills must be paid."
--- Frank Vincent Zappa