From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [patch 0/2]btrfs: add two ioctls to do metadata readahead Date: Mon, 19 Jul 2010 23:58:05 -0400 Message-ID: <20100720035805.GA18049@infradead.org> References: <1279094536.1528.93.camel@sli10-desk.sh.intel.com> <20100719080354.GA12808@infradead.org> <4C445A81.5010808@linux.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Christoph Hellwig , Shaohua Li , Chris Mason , "linux-btrfs@vger.kernel.org" , "Wu, Fengguang" To: Arjan van de Ven Return-path: In-Reply-To: <4C445A81.5010808@linux.intel.com> List-ID: On Mon, Jul 19, 2010 at 07:00:33AM -0700, Arjan van de Ven wrote: > On 7/19/2010 1:03 AM, Christoph Hellwig wrote: > >This seems like iteration 66 of the ill fated mobling readahead crap. > > I think you need to actually look at what this patch does before > rendering this judgment; something you apparently have not done. Thanks a lot for the attack, but of course I have. It's the same stupid idea of throwing almost generic code into filesystems that you previous attempt did. It's completely generic code for checking what pages are in-memory (plus constaints) and then force read-ahead on them again. The only filesystems specific bit is which address_space to select. The whole apporach is also useful to extN / xfs / etc if you allow it on any address_space.