[919] in NetBSD-Development
hanging problem solved
daemon@ATHENA.MIT.EDU (ghudson@MIT.EDU)
Sat Jul 29 20:34:30 1995
From: ghudson@MIT.EDU
Date: Sat, 29 Jul 1995 20:31:13 -0400
To: pc-dialup@MIT.EDU
So, the hanging problem we were having is that, when any device was
initiating bus accesses for DMA (the ethernet or SCSI card), and you
were writing to the console, the pccons fillw() routine would hang
while writing spaces across the bottom line of the screen after
scrolling.
This is either a Pentium bug or a PCI bug. Since the actual
instructions in fillw() are relatively innocuous (and Charles tried
rewriting it to avoid string instructions, and it still hung), we
can't tell much more without a logic analyzer. However, the problem
goes away if you turn off video memory caching. No, we don't know
why, but it works.
So that was the remaining hardware problem (except that the machine
hangs on reboot). Since John Kohl has (apparently) fixed the software
problem with released read-only volumes, I don't see any technical
problems in the way of putting the machine into production.
jaws is running the libafs1_0A.o from granola at the moment; there may
be some instability due to this.
