removing "link down"

[Michael Johnston]

In article <1990Aug16.000058.8119 at rice.edu> you write:
>
>We have the following problem on our Sun.
>
>System   :   Sun/3 180  Sun OS 4.0.1
>Problem  :  /dev/cua* occasionally goes into "link down" mode.
>Diagnosis:
>
> o There is no LCK_* file in /usr/spool/uucp. "tip" | uucp is not running.
> o Modem is working/ready. Power cycling it does nothing.
> o Likely to happen when tip is killed externally by a user (i.e. kill -9)
> o Subsequent tip execution results in "link down" messages from the system.
> o Only way to remove this condition we know of is to reboot the system.
>
>If anybody knows of ways to fix the problem without rebooting the system,
>I am very interested in knowing how.  Thank you very much in advance.
>

We have had this problem here exactly as you describe it. Generally, we
wind up with a process that is <defunct> and unkillable. Look for a
<defunct> or process owned by 'uucp' and then get its pid. Next, try:

    gcore <pid> 

Don't ask me why this works. It shouldn't. As I take it once a process is
locked into core with a proiority lower than 28 (I think) it is
unkillable.  We discovered this accidentally while trying to get and debug
a core dump of the offending tip process. Sun is apparently aware of the
problem and acknowledges that it is difficult to trace. I'd be interested
in seeing if this works for others as it has for us. It has worked for us
every single time and this happens at least once a week around here (on
our NFS server to boot!). Try convincing users that you have to reboot the
server so you can use the modem. Sometimes it's not so easy.

Michael R. Johnston                          / mikej at marob.masa.com
System Administrator         Temporary Email: 
Lilink Public Xenix                          \ mikej at marob