Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 15 Feb 2006 05:23:32 +1100
From:      Peter Jeremy <peterjeremy@optushome.com.au>
To:        Vladimir Konrad <v.konrad@lse.ac.uk>
Cc:        FreeBSD Current <freebsd-current@freebsd.org>
Subject:   Re: memtest - an idea
Message-ID:  <20060214182331.GE900@turion.vk2pj.dyndns.org>
In-Reply-To: <1139938898.2050.70.camel@localhost.localdomain>
References:  <1139938898.2050.70.camel@localhost.localdomain>

next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 2006-Feb-14 17:41:38 +0000, Vladimir Konrad wrote:
>is it viable to have a memtest like process running full-time (low
>priority) or scheduled, on running FreeBSD system to discover bad memory
>sooner than a later (without taking the system off-line)?

If you're using ECC RAM then it's fairly simple to have a background
task that reads all memory to detect and scrub single bit errors.

If you're looking for something to check non-ECC memory then it's far
more difficult.  You could write a kernel module that grabbed a free
page and did a few tests on it but this won't be able to test allocated
memory (keep in mind that the kernel is permanently wired), will miss
some pattern sensitive errors and, in the case of pattern sensitive
errors, has an equal probablility of causing a failure in a page not
being tested.

-- 
Peter Jeremy



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060214182331.GE900>