Date: Thu, 17 Jan 2002 21:35:03 +0100
From: Wessel Dankers <wsl@fruit.eu.org>
To: Steve Lord <lord@sgi.com>
Subject: Re: Large file I/O error
Message-ID: <20020117213503.D25522@fruit.eu.org>
References: <20020117200844.A25522@fruit.eu.org> <1011294784.13534.695.camel@jen.americas.sgi.com> <20020117203146.C25522@fruit.eu.org> <1011296110.13545.725.camel@jen.americas.sgi.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.2.5i
In-Reply-To: <1011296110.13545.725.camel@jen.americas.sgi.com>; from lord@sgi.com on Thu, Jan 17, 2002 at 01:35:10PM -0600
Sender: owner-linux-xfs@oss.sgi.com
Precedence: bulk
Status: O
Content-Length: 1122
Lines: 31

On 2002-01-17 13:35:10-0600, Steve Lord wrote:
> On Thu, 2002-01-17 at 13:31, Wessel Dankers wrote:
> > On 2002-01-17 13:13:04-0600, Steve Lord wrote:
> > > Someone else just pointed out that forced shutdown is overwriting the
> > > super block - which is not good. There appear to be a bunch of dirty
> > > buffers with a zero disk address in them left behind. It is being
> > > worked on.
> > 
> > Ok, great!
> > 
> > Any idea what's causing the forced shutdown in the first place, though?
> 
> Buffered I/O beyond 2^44 bytes into the sparse file probably. There is
> a hole in the code here. The linux VM uses a 32 bit page number
> index to index cache data. So 2^32 * 4096 is as bit as you can go.
> XFS is letting file offsets bigger than this in and not handling the
> results.

Ack. If I create the file to be exactly 2^44 bytes the problem goes away.
(mkfs fails though, ENOSPC ;)

The downside is that any user can cause a forced shutdown which is a bit
worrisome from a security perspective. =(

Kind regards,

--
Wessel Dankers <wsl@fruit.eu.org>

We had to turn off that service to comply with the CDA Bill.