Subject: Problems with 1.5.2, KFQSA & KA640
To: None <port-vax@netbsd.org>
From: Michael Schneider <ms@silke.rt.schwaben.de>
List: port-vax
Date: 11/03/2001 22:28:41
Hello, and thank you all for your comments so far.
Still, my KFQSA is giving me headaches..
To summarize what had happened until now:
I have a KFQSA in a MV3400 (ka640 cpu). I have a spare ka650 cpu. I have
3 RF71 drives connected. The whole shebang is running 1.5.2.
DSSI Termination shoudl be correct, and the KFQSA has been configured
correctly, i would say.
The MV3400 runns for weeks without problems as long as i disable the
cache-
enabling code in ka650.c.
If i use a kernel with enables the CPU cache, i get errors like
/netbsd: uda1: controller error, sa=0101510 (unknown error code)
within minutes of operation.
Sometimes, the error code is 0101511. Never somethin else
Sometimes, a stray interrupt (vector 0x318, ipl 23) hits just befor the
controller error. But this rarely happens at all and has probably no
significance.
The error occurs only when i run the ka640 or ka650 with enabled
CPU-caches.
Theoretically, this should not matter, but it does here.
Additionaly, it seems like the error comes only when the controller has
to
switch between drives, as in a file copy from one partition to another
and
the partitions live on different drives. Or, during fsck at startup,
when
two or more drives are affected.
A simple "cp" is enough...
I've put the KFQSA into another, "known good" VAX (a MV3500), and get
the same
problems with NetBSD 1.5.2. So it might be the controller.
I then tried OpenVMS, and alas! everything worked! Even with 2 disks.
I do not actually know whether VMS enables the CPU cache, but i would
think so.
So, is there someone using a KFQSA with more than one drive, with a
ka640
or ka650 cpu?
Does this work with 1.5 or higher for you?
I am prepared to dig into the mscp-driver (well, mentaly prepared. My
wife
calls this a "hyperventilating ego"...), but only if i am at least a
little
bit convinced that it is a driver bug. After all, my KFQSA could still
be broken...
Thank you, and sorry for the far too long post.
ms
--
Michael Schneider email: ms@silke.rt.schwaben.de
Germany http://www.vaxcluster.de
People disagree with me. I just ignore them.
(Linus Torvalds)