xfs
[Top] [All Lists]

RE: XFS SB Validate error

To: "Eric Sandeen" <sandeen@xxxxxxx>
Subject: RE: XFS SB Validate error
From: "Yom, Francis" <fyom@xxxxxxxxxxx>
Date: Thu, 31 Jan 2002 22:31:53 -0500
Cc: <linux-xfs@xxxxxxxxxxx>, <lord@xxxxxxx>, <spstarr@xxxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
Thread-index: AcGqzV3bug2SBzyvTgiLZqZzTO5fVgAAswyw
Thread-topic: XFS SB Validate error
Hi Eric,

In /usr/src/linux/fs/pagebuf/page_buf_io.c, for pagebuf_write_full_page,
it says:

        _unmark_delalloc(page)

So it appears to be the forced shutdown, but I don't know what had
caused the forced shutdown.  It truly is a mystery to me.  I run
VMWorkstation 3.0 on xfs, could this be a problem.  Although the VMWare
goons keep telling me that xfs is unsupported, I have never had a
problem up to this time.  I don't want to make any conlusions just yet
either.  I am keeping a very careful eye to see if it happens again so I
can document it better.  This is the very first time I ever encountered
SB Validate errors since using xfs from 2.4.14.  My kernel also uses the
preemptive and RMAP12a patches.  Could the Rmap12a patch be a source?
This is the last major modification I've made to my 2.4.17 kernel since
Monday (4 days ago).  I implemented this patch because I was have
stability problems when I would push my system to it's memory limit by
running multiple VMWare sessions - the VM wouldn't swap properly.  The
Rmap12a patch completely fixed my VM problems.  My system runs very
steady now at full RAM utilization.  

I am not using any neat stuff like xfs_freeze, realtime, etc. etc.
although I would love to play with it later.  ;-)

-Francis

-----Original Message-----
From: Eric Sandeen [mailto:sandeen@xxxxxxx] 
Sent: Thursday, January 31, 2002 10:11 PM
To: Yom, Francis
Cc: linux-xfs@xxxxxxxxxxx; lord@xxxxxxx; spstarr@xxxxxxxx
Subject: RE: XFS SB Validate error


Ok, a couple things... There have been several changes in 2.4.17 - can
you look at the bottom of pagebuf_write_full_page() in page_buf_io.c?
About 8 lines from the bottom of the function, does it say
block_flush_page(page,0), or _unmark_delalloc(page)?  If it's the
latter, then it was probably the forced shutdown that caused the
corruption, although I'm not sure what's causing the shutdown.

Are you doing anything fun with this filesystem?  Using LVM snapshots,
playing with xfs_freeze, realtime subvolumes?  :)

-Eric


On Thu, 31 Jan 2002, Yom, Francis wrote:

> Jan 31 17:21:09 sera kernel: EFSCORRUPTED returned from file 
> xfs_ialloc.c line 1265


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