gumkowsk@nadc.arpa (D. Gumkowski) (11/10/89)
Hopefully someone who has experienced a like problem or someone who is
wizardly in the networking arena can shed some light on a situation that
is occurring at our site. Virtually each time our PSN is reloaded our
VAX 11/780 running an almost vanilla Berkeley UNIX 4.3 (definitely vanilla
in the network area) crashes with a "panic: Segmentation fault".
Some basic history:
The problem started occurring in the timeframe that our IMP was changed
to a PSN. Near the same time we also incorporated the bug fixes that
were available from Berkeley. The "netimp" directory in the bug fixes
directory was applied to our system.
Some observations:
(1) We no longer get messages telling us that our IMP (PSN) is going
down in 30 seconds or messages telling us that our IMP (PSN) has
been marked down. We do get the "interface reset" message.
(2) If we take the PSN down with "/etc/config" before the reload occurs
and bring it back up after the reload occurs, the system does not
crash. This is not feasible since we are usually not notified when
the network control center reloads our PSN.
(3) The crash appears to occur (though not absolutely positive) in the
module _impstart which is in "if_imp.c". We are currently running
Berkeley version 7.5 dated 5/6/88. The previous version before
upgrading "netimp" was Berkeley version 7.1 dated 6/4/86.
(4) All other network operations seem to be working fine.
Questions:
(1) Are we running software that is known to be bad?
(2) Is the network control center we are tied to not appropriately
telling our system that the node is going to go down?
Thank-you in advance for any help anyone can give.
David A. Gumkowski, Naval Air Development Center
<gumkowsk@nadc.arpa>