xfs
[Top] [All Lists]

XFS BUG (was Re: 3ware driver broken with 2.4.22/23 ?)

To: linux-xfs@xxxxxxxxxxx
Subject: XFS BUG (was Re: 3ware driver broken with 2.4.22/23 ?)
From: Lukas Hejtmanek <xhejtman@xxxxxxxxxxxxxxxxx>
Date: Tue, 30 Dec 2003 20:36:20 +0100
In-reply-to: <1072742359.21939.2.camel@bubbles.imr-net.com>
References: <20031221112113.GE916@mail.muni.cz> <3FE645E3.30602@rackable.com> <1072503925.27022.222.camel@menion.home> <20031229234902.GL916@mail.muni.cz> <1072742359.21939.2.camel@bubbles.imr-net.com>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mutt/1.5.4i
Hello,

that was a discusion about 3ware driver. It has resulted in XFS bug most
probably.

With kernel 2.4.22 or 2.4.23 (SGI XFS snapshot-2.4.23-2003-12-01_00:33_UTC with
no debug enabled) 

dd if=/dev/zero of=/file_on_xfs_partition bs=448k count=23405

lockup whole server. xfs partition on SCSI disc. It happen when file becomes
about 5GB of size.

First I've discovered it with iozone benchmark and 10GB file..

2.4.20 kernel with IDE disc seems to be OK as well as 2.6.0 kernel on the same
configuration as 2.4.23 kernel was.

On Mon, Dec 29, 2003 at 03:59:19PM -0800, Joshua Schmidlkofer wrote:
> On Mon, 2003-12-29 at 15:49, Lukas Hejtmanek wrote:
> > On Fri, Dec 26, 2003 at 09:45:25PM -0800, Joshua Schmidlkofer wrote:
> > > 
> > > >    Generally not with such a small rev difference.  You could try the 
> > > > latest driver, and firmware in the 7.7.  The driver source is on the 
> > > > Red 
> > > > Hat drivers disk.  You should be able to drop in the .c, and .h in 
> > > > drivers/scsi, and recompile.
> > > > http://3ware.com/support/download.asp?code=5&id=7.7.0&softtype=Driver&releasenotes=&os=Windows
> > > > 
> > > > PS- Personally I'd suspect an XFS bug.  Try reiserfs.  I've been 
> > > > running 
> > > > 2.4.23pres, and 2.4.23 on hundreds of 3ware of numerous different 
> > > > types. 
> > > >   With no issue with the prior firmware release.
> > > 
> > > There are a lot of people, running RAID5 3ware's w/ Terrabyte arrays.  I
> > > don't want to say it is not an XFS bug, but I find that highly suspect. 
> > 
> > Well, with ext3 parition iozone program finishes OK. So it looks like some 
> > XFS
> > bug.
> 
> FWIW have you sent this on to the XFS list?
> 
> thanks,
>   Joshua
> 
> 

-- 
Lukáš Hejtmánek


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