sventek@LBL.ARPA@sri-unix.UUCP (10/11/84)
We just installed a second UBA on our 11/780 which used to run 4.2. After
completing the installation, complete with checks that the new hardware
was working correctly, we modified our configuration file to reflect our
new addition, and built a new kernel. During boot, the system successfully
finds uba0 and all of the devices specified there, and announces that it
has found uba1. unfortunately, as soon as it tries to access the devices
on uba1 (via their probe routines), the system crashes.
Another twist to this situation is that the generic kernel boots successfully,
finds both uba0 and uba1, and even autoconfigures devices found on the second
uba. We found that if we did not specify the csr and vector addresses for
devices on uba1, we could not even successfully build the new kernel.
Needless to say, we are somewhat confounded by this situation, and need to
bring the system up ASAP.
I scanned the archives of unix-wizards from the last couple of years, and
the only entries which I could find with regards to a second UBA concerned
special kernel mods necessary to make it work on 750's. If someone has
seen problems similar to ours, could you please let me know how you
resolved it? please respond directly to me, as the machine which is
down is our exploder for unix-wizards. I will also summarize to the
mailing list after we have determined what is going on.
Joe Sventek <sventek@lbl.arpa>
ucbvax!sventek%ucbernie
P.S. Sorry that I cannot give you the exact error message when the
kernel crashes, but I am sending this from home, with the console
listings still sitting on the decwriter.sventek@LBL.ARPA (11/02/84)
A hearty thanks to Rick Adams <rick@seismo> for pointing out my problem. It
seems that our 4.2 system was attempting to access a variable using NUBA+1.
The fix is to declare that we have 4 uba's in the config file (as the
generic system does), knowing that it will only find 2 nexi associated with
an UBA. System came up like a champ, and has given us no problems.
A minor gotcha for those considering embarking on the second UBA path and
which have hosts with more than one network interface. If you plan on
spreading your network interfaces over the two UBA's, you MUST place the
interface whose address you want used for originated IP packets on UBA0.
This is independent of the order in which the devices are listed in the
configuration file. As it stands now, most of our r* protocol use is
broken until I reboot with our Imp interface bound to UBA0.
Joe Sventek <JSSventek@lbl.arpa>