xfs
[Top] [All Lists]

Re: Linux w/XFS on Athlon systems

To: "Gabe E. Nydick" <gnydick@xxxxxxxxxxxxx>, "Drew 'Cheetah!' Maxwell" <cheetah@xxxxxxxxxxxxx>
Subject: Re: Linux w/XFS on Athlon systems
From: Seth Mos <knuffie@xxxxxxxxx>
Date: Thu, 06 Dec 2001 23:57:21 +0100
Cc: Jesse Hall <jdhall@xxxxxxxx>, <linux-xfs@xxxxxxxxxxx>
In-reply-to: <B835232B.9B6%gnydick@clubphoto.com>
References: <Pine.BSI.4.10.10112062034020.8606-100000@xs4.xs4all.nl>
Sender: owner-linux-xfs@xxxxxxxxxxx
At 13:29 6-12-2001 -0800, Gabe E. Nydick wrote:
Seth,

    The system we have works fine with 1000 simultaneous bonnie++ processes
running under a PIII kernel.  Do you think it could still be the memory?

I had one system with 64 MB of ram of which about 1 was faulty. Sometimes it booted. It died at various places. corrupted some files crashed some processes, some stuck in D state, hanging kupdate, oopses, panics. I had seen about every error known to man in about 2 hours untill I removed one of the dims and everything was well.


That was 4 months ago. The machine is still running to this date and has never crashed _once_ since that time or even give any form of trouble.

Bonnie excercises disk. The Bonnie proces stays in RAM and your disk cache might wander right over the bad piece of ram without crashing the kernel. You enable the Athlon option, it reorganizes some ways of reading and writing to the ram to speed it up and suddenly the kernel code is trying to use the faulty ram. Seems plausible to me.

Did you enable mtrr by any chance? I could make a Eicon Diva server card stop working by enabling or disabling mtrr. Because of remapping some buffers and PCI ports the driver was not the same spot where the driver thought it should have been.


-- Seth Every program has two purposes one for which it was written and another for which it wasn't I use the last kind.


<Prev in Thread] Current Thread [Next in Thread>