Date: Fri, 15 Jun 2001 01:20:12 -0400
From: Joseph Fannin <jhf@rivenstone.net>
To: linux-xfs@oss.sgi.com
Subject: Re: Does XFS require DevFs?
Message-ID: <20010615012012.A24835@zion.rivenstone.net>
References: <20010614183114.A24007@zion.rivenstone.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.2.5i
In-Reply-To: <20010614183114.A24007@zion.rivenstone.net>; from jhf@rivenstone.net on Thu, Jun 14, 2001 at 06:31:14PM -0400
Sender: owner-linux-xfs@oss.sgi.com
Precedence: bulk

On Thu, Jun 14, 2001 at 06:31:14PM -0400, Joseph Fannin wrote:
>      I've searched the archives but found no answer to this question:
> Does XFS require DevFS?  Or can I build a kernel for my system
> (installed from the XFS RH7.1 .iso) that does not include devfs
> without having problems?
> 
>     I've been completely unable to build a working kernel using either
> the RH-2.96 or egcs -- the build either fails or gives me a kernel
> that produces random segfaults, oopsen, and fs corruption.  I've tried
> the .src.rpm on the .iso, the 2.4.3 patches and the CVS devel tree.
> 

    Thanks you everyone for your replies, and for putting up with my 
frustration.  Now that I know that my problem wasn't a misconfiguration
I have built two kernels now with xfs support -- v2.4.3 with the 1.0
release patches, built with egcs.  Both machines appear to be working fine.

    The problem was definitely compiler related -- even the latest release
of 2.96 didn't work for me, though I read somewhere that it did.  The
redhat-patched kernel in the src.rpm didn't build with egcs (kgcc) but
a vanilla 2.4.3 with xfs patches applied built fine.  RedHat builds and
tests their kernels with 2.96, so I suspect one of their patches doesn't
like egcs.  (The failure was an "internal compiler error" that happened
twice, the second again in the same place, after a `make clean`)

    The other build failures I got were likely due to a combination of fs
corruption (there were some complaints about missing files) and the wrong
compiler.  The hardware on this machine is solid, though I have another with
flaky memory; the problem was 100% gcc-2.96.  So don't use it for XFS
kernels.  :-)

--
Joseph Fannin
jhf@rivenstone.net

"Bull in pure form is rare; there is usually some contamination by data."
    -- William Graves Perry Jr.