rjg@nis.mn.org (Robert J. Granvin) (03/08/89)
Now, I seem to have a strange problem, and I'm once again begging the
net for suggestions or solutions ... :-)
Basic information: We have tty0 set up as a unidirectional inbound
serial line. It's hardwired to a minicomputer. The gettydef that
is being used on the receiving (Microport) side is one that is
identical to one that is in use in several other locations. Anyways,
the gettydef isn't the issue, because the configuration has worked
flawlessly for a number of months, and nothing has changed.
Until now.
This situation is guaranteed to occur every time upon boot or any
kind of power fail recovery. It will also occur at random times
following the completion of a UUCP communication.
What happens is that when getty spawns, the init that brings it up
sticks around. After a little bit (30-40 seconds), what appears to
happen is that init will die, but will take the getty with it. Of
course, getty will respawn and bring along a new "stuck" init with
it. The situation gets worse, because sometimes multiple inits for
the one port will start collecting, each one wiping out the getty
and respawning it again. This gets to be real ugly, and anything
that's going to drop the port on a rapid basis is going to obviously
cause havoc with any connection attempt.
The situation can be resolved by turning the port from respawn to
off, wait for the inits to go away, and then turning it back on and
re-initing it. It'll run flawlessly for days (usually) before
starting itself up again independently (or on the next reboot).
The frustrating part is that there is no apparent rationale for
what's happening, and it never occurred before. It's beginning
to gain a life of it's own. :-)
[Running Uport '286 v2.4, if it matters]
Any ideas? (email, please).
--
Robert J. Granvin "BUGS: ps is unavailable (and is most
National Information Services needed) when the system is hung."
rjg@nis.mn.org --MKS Toolkit manual
{amdahl,hpda}!bungia!nis!rjg