bruce@think.ARPA (Bruce J. Nemnich) (05/04/86)
I remember my DEQNAs coming (about 18 months ago) with a DEQNA manual addendum that says two DEQNAs are not supported in a backplane yet. Presumably this is becuase of a hardware bug. This wouldn't surprise me, since there are rather serious other hardware bugs in the DEQNA. The most serious bug is that the device hangs every once in a while. How often seems to be proportional to the general amount of traffic on the net (not necessarily addressed to the card) and to the speed of the processor (only 2 data points here, but it hangs with a KDJ-11 much more frequently than with a KDF-11), and quite possibly with the length of the ether (much more frequent wedgings after doubling the length of our ether from about 500-1000 feet). The symptom I check for in my driver is that a transmit command never completes--I check it every 5 seconds, and do a device initialization if it appears hung. The machine I am on now has hung and been reset 6 times in 310 hours and 1.3 million packets. There are other bugs with respect to handshaking flags in the Buffer Descriptor Lists. The bugs are either in the documentation or the card; but they aren't too bad since one can program around them. -- --Bruce Nemnich, Thinking Machines Corporation, Cambridge, MA --bruce@think.com, ihnp4!think!bruce; +1 617 876 1111
hurf@batcomputer.TN.CORNELL.EDU (Hurf Sheldon) (05/06/86)
We found that under ultrix 1.1 the uvaxen will crash if a packet larger than 1kbyte is received. Ultrix 1.2 is fixed in this respect. The qe0 restart does seem to be ubiquitous, however the frequency is down by a factor of 4 to 6 in 1.2 vs 1.1. We have a vaxstationII and a uvax with different configurations & these symptoms were/are the same on both. hurf sheldon Plasma Studies