data corruption with current (maybe sis chipset related?)

[Terry Lambert]

Heiko Schaefer wrote:
> > Terry mentioned that he explained it to Bosco, and that he had a patch to
> > work around it in a different way. But that was from November, time has
> > passed since then.
> 
> a nice workaround would be great, of course. assuming terry knows what
> he's doing, i'm sure such code would be preferrable to turning on these
> options by default. but until then, turning them on sounds like the safest
> bet to me. i'd sacrifice some speed for reliability anyday. and so far i
> felt that freebsd always had that philosophy.

It's Bosko's workaround.  I have a binary patch with two .o's
(machdep.o and locore.o) that I've given a couple of people who
agreed not to disassemble it which contains a fix based on my
understanding of the problem; it worked for them.  Being binary,
it's not really appropriate to -current any more.  Bosko's is
preferred, because it doesn't disclose the nature of the bug,
so it's OK to pass around as source code.

-- Terry