[nSLUG] Re: hung install - any ideas?

Dop Ganger nslug at fop.ns.ca
Fri Feb 1 19:09:56 AST 2002


On Fri, 1 Feb 2002, Mike Spencer wrote:

> Dop Ganger <nslug at fop.ns.ca> quoth:
> 
> > So I'd tend to say it's either buggered cache (try disabling level 1
> > and/or level 2 cache in the bios) or a dicky ram stick.
> 
> Would what happens in calibrate_delay() (or the  __delay() asm func it
> calls) get trashed by a buggered cache?

Yes; if there's a cell that's marginal it might be able to get to the
point of doing the bogomips loops, but note the inner loop is very short
and runs the same code over and over again, which may be enough to push
the cell over the edge and return random data which gets executed, and
then leads to lockup. Disabling the cache will help show whether this is
the case or not; note, however, that it's not impossible that this may
mask a different hardware problem, since the machine will be running
slower and unknownbadpieceofhardware might be able to limp along at the
slower speed.

Another scenario is a bad memory cell in just the spot where the bogomips
test is read from, which would again lead to lockup. memtest86 should show
if this is the problem, although cache corruption may well cause memtest86
to lock up too.

A couple of things to note for Ron that I should have said in my previous
mail is that if it *is* a dicky ram stick, there's a patch to exclude bad
cells available at http://rick.vanrein.org/linux/badram/, and also
underclocking the machine might help get it down to the point where the
marginal hardware is no longer marginal.

Cheers... Dop.




More information about the nSLUG mailing list