xfs
[Top] [All Lists]

Re: AIO and XFS?

To: James Braid <james.braid@xxxxxxxxx>
Subject: Re: AIO and XFS?
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Thu, 3 Oct 2002 14:26:39 +0100
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <018601c26a92$ceac8f20$0902000a@corp.peace.com>; from james.braid@peace.com on Thu, Oct 03, 2002 at 04:10:28PM +1200
References: <20021002132651.A2141@infradead.org> <018601c26a92$ceac8f20$0902000a@corp.peace.com>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mutt/1.2.5.1i
On Thu, Oct 03, 2002 at 04:10:28PM +1200, James Braid wrote:
> > > Or even better, has anyone got AIO and XFS working already?
> > 
> > Of course it's possible.  But it's not trivial.  I'd suggest you wait
> > for aio getting fully implemented in 2.5, you can be sure 
> > we'll adds supporti
> > for XFS there as soon as possible.
> > 
> 
> I think I kinda have XFS and AIO working now on 2.4.20pre8aa2, it is
> failing on one test case, 5.t, which says that it does "- Write from a
> mmap() of the same file. (5.t)".
> 
> All I have done so far is add the following 4 lines into
> fs/xfs/linux/xfs_file.c:
> 
>         .kvec_read      = generic_file_kvec_read,
>         .kvec_write     = generic_file_kvec_write,
>         .aio_read       = generic_file_aio_read,
>         .aio_write      = generic_file_aio_write,
> 
> In the file_operations struct after the fsync entry in there.
> Anybody got any ideas why it would be failing on that specific test
> case?

Because you don't have any XFS-intetrnal locks or checks.
It's suprising that you got at least so far..


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