According to D. J. Vanecek:
>
>	f) Some iconoclasts say that over-aggressive Linux code
>	is responsible for part of the problem. Apparently relying
>	on the published spex for DMA is over-aggressive. But, Linux
>	does rupture the bleeding edge from time to time. (Why I
>	run 'BSD reason #12312)
>

Hmmmm I know I will get a spew of "works fine for me"'s but when I was
running an AHA1542B in a 486DX/25 motherboard (Micronics one) I could
guarantee a memory parity error just by thrashing the scsi disk for a
while - unpacking part of the X11R6 distribution was all it took.  I
was able to _somewhat_ reduce the error by changing the bus on/off
times in the scsi driver.  The problem was with me from about NetBSD
0.9 times (or there abouts) and only really went away when I changed
the motherboard for a 486DX/33 (no brand one).

-- 
Brett Lymn, Computer Systems Administrator, British Aerospace Australia
===============================================================================
  What do you get when you cross a cantaloup with a dog?        Melancholy :-P