Re: [tlug] apparent hardware problem(s) -- but where?

[Kalin KOZHUHAROV <me.kalin@example.com>]

Hello Chris,

I think you nailed the problem, but somehow you missed the solution ;-)

On Tue, Nov 17, 2009 at 13:37, BALUTA Chris <baluta@example.com> wrote:
[snip]
>  - memtest86 has been run throughout this process. On 2 or 3 occasions,
> there were no errors. On the others, there were a plethora of errors.
I think it is obvious that you have a  faulty memory, very often the
case I have seen. I'd bet 99% this will be the issue reading your
description. (Additionally you may have other issues, but let's fix
the obvious first)

You have a few options:

prep: Get later SysRescueCD (http://www.sysresccd.org/Download) it
includes new version of memtes86+ and redo the testing.

1. If you have multiple modules of memory remove all, then
all-but-one, all-but-two, etc. and run the memtest to check if there
is a single faulty module. It has to pass at least a few times before
you claim it is working. If it is a laptop, it may contain RAM
on-board which may be faulty; if in warranty send for repair, if not -
use linux and live with it (see next). If it is a faulty removable
module - replace it or live without it.

2. Even if it is the on-board memory, use boot parameters to the
kernel to exclude the faulty region (see
Documentation/kernel-parameters.txt) like memmap=100M$0x186900 (to
exclude 100MB, starting at address 0x186900). You can generate the
exact positions from the memetest output. You can add several such
exclusions, make them a bit bigger than the faulty regions just in
case. (This is linux-only solution, no idea how to go this in other
OSes)


Please report "success" here when you are done ;-)
Cheers,
Kalin.