[Simh] VAX 3900: Possible Networking Bug triggered by Clustering in VMS 7.3

[Mark Benson]

Hi,

Had a really interesting night playing  with my new SimH VAX cluster. I have tripped over what may be a bug in the networking in SimH... or may not. I don't know enough to make out exactly what it is that broke.

Host environment: 

NetBSD 5.1.2 amd64
libpcap-1.1.1 compiled from pkgsrc

The system is running dual NICs and an ipnat/ipfilter gateway, if that matters at all.

The symptoms:

Initially, enabling the disk shadowing server (added 'SHADOWING=2' to MODPARAMS.DAT and AUTOGEN'd) caused the following:

 - Cluster communications between nodes stopped working almost completely.
 - starting DECnet via @STARTNET crashed the whole VAX and caused it to crashdump and reboot (see http://dectec.info/DECnet_crash_dump.txt)

I initially thought the trigger was the Shadowing and turned it off. I have subsequently discovered that the issue persists even after the shadow server is disabled (commneted out line in MODPARAMS.DAT and AUTOGEN'd). Cluster communications remain spotty and (work on some boots and not others) and DECnet still crashes.

I have another node (RaspberryPi) running a Debian Linux derivative. It is running a similarly hand compiled SimH vax binary albeit with a different version of libpcap (I can't tell what exact version but I think it's 1.3.0 ish). It exhibits non of the described issues.

So something specific to my NetBSD machine is causing the crash/network issue. 

That's all I got. I'll gladly test/verify any other details if required.

-- 

Mark Benson

http://DECtec.info
Twitter: @DECtecInfo
HECnet: STAR69::MARK

Online Resource & Mailing List for DEC Enthusiasts.