Date: Fri, 29 Jun 2001 22:49:57 -0400 (EDT)
From: J Landman <landman@mediaone.net>
To: Simon Matter <Simon.Matter@ch.sauter-bc.com>
cc: <linux-xfs@oss.sgi.com>
Subject: Re: Problem fixed (I think)
In-Reply-To: <3B3CEF07.479A0857@ch.sauter-bc.com>
Message-ID: <Pine.LNX.4.32.0106292242400.32586-100000@dsl-squash.corp.sgi.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-linux-xfs@oss.sgi.com
Precedence: bulk

On Fri, 29 Jun 2001, Simon Matter wrote:

> J Landman schrieb:
>
> > Folks:
> >
> >   I had some problems with xfs that I think I fixed.  Basically I had
> > cp/mv and others core dump rather than work.  What I did is I backed up my
>
> Can you let me know on how many disks your filesystem was. And
> was it IDE or SCSI. If IDE, what kind of chip and what kind of settings?

2 disks, Seagate 181 GB 7200 RPM Ultra 160's, in LVD connection.  Using
Adaptec 29160.  Using the aic7xxx module.  Booting off of a Reiserfs
partition (will change that when XFS gets into the kernel).

> > xfs based data, rebuilt the raido, rebuilt the filesystem, and it now
>
> Ah, RAID 0, so you have more than one disk? IDE? with DMA (enabled
> for most chipsets by default in the RH7.1-XFS kernel?

Not IDE.  I dont have enough IDE channels to make a good RAID0.

> >
> >   I had built the original file system with the RedHat 7.1 iso based
> > images, and tools.  I changed distros to Mandrake.  Thats when it broke.
>
> What kernel did you use when it broke? Are you shure you know exactly
> when it broke?

I used the 2.4.3 mandrake source (included in 8.0). I made a copy of the
original tree, patched my cloned tree, and built from there.  This was
using the mandrake specific patches of about 1 month ago.

> >   During this time, I was unable to build the development tree (same
> > crashes that everyone else was getting, using wither the 2.91.66 or the
> > 2.96.3 compilers).  I was also unable to apply the patches to a 2.4.5-5mdk
> > Mandrake source tree.  Got lots of rej's.  I am hoping that the work to
>
> Did you check filesystem consistency somehow? For ex. with rpm -Va
> which will show all changes in the distribution file?

??

I used xfs_check for file system consistency checking, and xfs_repair -n.
No errors were flagged.  The RPM database shouldn't have been corrupted.
I didn't check it, but I can do that.

>
> >
> > get it into the kernel accelerates, so we wont have the problems in the
> > not too distant future.
> >
> >   Regardless of the minor annoyances, good work folks!
> >
> >  --
> >  Joe Landman,
> >  landman@mediaone.net
>
> I was suffering filesystem corruption for awhile now and it was very
> difficult
> to reproduce it. In my case it is not XFS or Raid related but has something
> to
> do with the IDE subsystem.

This was scsi based.  The odd thing is that tar, bru, and others worked
just fine.  It was the packages in fileutils-4.x that died.  Considering
my build environment is using makefiles with lots of cp/mv stuff, I had to
copy it to my IDE to make it work for the last few weeks.  I emulated copy
with

	tar -cf - ./ | (cd /target && tar -xvf - )

which worked fine.  No data corruption that I could detect.

-- 
 Joe Landman,
 landman@mediaone.net