xfs
[Top] [All Lists]

Re: Irreparable 'corrupt dinode ... error 990' Revisited

To: Linux-XFS Mailing List <linux-xfs@xxxxxxxxxxx>
Subject: Re: Irreparable 'corrupt dinode ... error 990' Revisited
From: Chris Wedgwood <cw@xxxxxxxx>
Date: Thu, 5 Aug 2004 03:15:18 -0700
In-reply-to: <20040805024808.GF10358@xxxxxxxxxxxxxxxxxxxx>
References: <20040804153356.GD26826@xxxxxxxxxxxxxxxxxxxx> <20040804164037.GC18867@xxxxxxxxxxxxxxxxxxxx> <20040804173651.GA1107@xxxxxxxxxxxxxxxxxxxxx> <20040805024808.GF10358@xxxxxxxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Thu, Aug 05, 2004 at 10:48:08AM +0800, Federico Sevilla III wrote:

> Ack. No, I don't have ECC memory on this system, as it is a clone
> workstation.  I already ran MemTest86 on this before, and remember
> it passing the basic tests consistently.

that just means it probably doesn't have bad ram, but it doesn't mean
you didn't get an ecc error (memory errors are statistically possible
even with the best ram, cheap ram is just more likely than expensive
ram usually)

> How sensitive is XFS to RAM quality?

tree structures don't tollerate memory corruption well

> Would anyone know how sensitive other Linux filesystems (ext3,
> ReiserFS and JFS) are to RAM quality?

reiserfs is also sensitive to marginal ram --- this is relative to
ext2 which is much less sensitive it seems

don't know about ext3 or jfs

> Does the kernel necessarily have to panic when RAM has very slight
> problems

no, usually it doesn't even notice, things just silently go bad

if the kernel can detect it it usually panics as it doesn't presently
have the infrustructure to deal with things.  fwiw there are no ECC
drivers in mainline anyhow for the most part if we are talking ia32,
for other platforms (ie. ia64) you get events and the kernel will
usually just panic


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