On Fri, 26 Apr 1996, Jason Thorpe wrote:
> I see two major things that have changed between my working sources and 
> broken sources:
> 
> 	- SBus DMA burst size is now checked

This is a change that I put in, but it should compile out on a non-Sun4m
kernel and in a generic kernel should not run on a SS1. The only change
that would affect a SS1 or 1+ is that sbus.c now reads the PROM burst-size
property. I don't think that reading a node property could cause the system
to crash...

If people see this effect on Sun4m's, please let me know, since this might
then be the culprit...

> 	- Calibrated delay was introduced.  This changed delay()
> 	  significantly.

This is what I first thought of (it might be confusing some timing-sensitive
piece of the SCSI code), but that doesn't explain a crash in malloc. If
I get a chance tomorrow I'll try to build a -current kernel on my SS20
and see if it works or fails.

-Aaron