info-vax@ucbvax.ARPA (09/14/84)
From: engvax!KVC@cit-vax I am running the TCP/IP code. I wrote support for the DEUNA, since Tektronix doesn't seem to have any. If you want the code for the DEUNA, send me mail and we can arrange something. I also rewrote TELNET, since the PTY (Pseudo-Terminal) driver from Tek does not work very well... By the way, in case anyone out there is running with Interlan's Ethernet controllers (for the UNIBUS), here is an interesting tidbit: Using the Tektronix TCP/IP, I found the network kept crashing every few days or so once we had let regular production users loose on it. The TCPACP died with "device timeout" errors. After it died, any process which tried to access the Interlan device (including a restarted TCPACP) would hang on the first QIO to the device. I spent many weeks trying to find the bug in the network code and then in the device driver from Interlan. No luck. I also tried swapping boards around with no luck. Oddly enough, it only hangs on the VMS system, never on UNIX. Finally, in a fit of desperation I called Interlan. They said that I did not have the very very latest rev level of the board etch and the micrcode. My board was ucode level "H" and etch "J". They thought a more current rev would fix it, although I was only the third site to ever report the problem. Apparently it is a very very subtle bug. As luck would have it, we ordered another board so we'd have an on-site spare. It arrived the day after I talked to Interlan and was ucode rev "J" and board rev "K". Well, I dropped it into my VAX and started sending many megabytes. So far, so good, it's still up and running! So, if any of you have any NI1010's from Intelan, check those rev levels! It wasn't much fun spending weeks scanning source code and running ANALYZE/SYSTEM to look at the driver structures for something in the damn microcode of the device! /Kevin Carosso engvax!kvc@CIT-VAX.ARPA Hughes Aircraft Co.