[Top] [All Lists]

Re: Oops on recovery with Linux-2.4.11-xfs on SPARC

To: Sean Neakums <sneakums@xxxxxxxx>, Linux XFS <linux-xfs@xxxxxxxxxxx>
Subject: Re: Oops on recovery with Linux-2.4.11-xfs on SPARC
From: Seth Mos <knuffie@xxxxxxxxx>
Date: Thu, 11 Oct 2001 12:11:30 +0200
In-reply-to: <6ulmii5wxp.fsf@zork.zork.net>
Sender: owner-linux-xfs@xxxxxxxxxxx
At 10:48 11-10-2001 +0100, Sean Neakums wrote:
I was feeling kinda adventurous and had been having excellent results
with XFS on an IA-32 box running Debian unstable, so I moved a SPARC
Debian unstable install over to 2.4.11 from the CVS tree.  I built the
kernel with the sparc64 cross-compiler from Debian's gcc-3.0-sparc64
package, version 3.0.2.

Sparc is unknown teritory at the moment. You are probably one of the first.

Everything worked OK initially.  Yesterday I moved /usr, /var and /tmp
over to XFS , but this morning I started having a few problems.  The
biggest one was kernel oopsing on mount during recovery (see below).
(I'm not sure if I decoded this properly.)

xfs_repair got to phase three before dying with a Bus Error, but
whatever it managed to do was enough the let the filesystems be mounted
again, until the next reboot.

Lucky you :-)

I also encountered some strange failures with apt and dpkg, giving
strange errors about files not being found, even though they really
were there (e.g. apt was trying to move a file from the
archives/partial into archives and failed, claiming the file was not
there, but it was).

This sounds like the behaviour that people spotted when kernels were compiled with early gcc 2.96 versions. There might be a few problems in the XFS code.

Another problem I had was in creating symlinks: doing

ln -s lib state

would give me a symbolic link called state that pointed not to lib,
but to state.  This symptom *only* cropped up after I had mounted an
XFS filesystem.  Before mounting, symlink creation worked fine on my
ext2 root; aftermounting an XFS filesystem it failed as described, on
both ext2 and XFS filesystems.

This could be related to the symlink problem that has just been discussed on lkml. 2.4.12 should fix this but this one is also reported to be broken (just don't know what exactly yet).


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>