From mboxrd@z Thu Jan 1 00:00:00 1970 From: tytso@mit.edu Subject: Re: unified page and buffer cache? Date: Sat, 8 May 2010 08:52:15 -0400 Message-ID: <20100508125215.GH18762@thunk.org> References: <4BCC7C05.8000803@cfl.rr.com> <4BE417CB.7000806@cfl.rr.com> <20100507135329.GA17259@parisc-linux.org> <4BE4359E.9020400@cfl.rr.com> <20100508004617.GF18762@thunk.org> <1273280081.2444.5.camel@faldara> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Matthew Wilcox , linux-fsdevel@vger.kernel.org, Linux-kernel To: Phillip Susi Return-path: Received: from thunk.org ([69.25.196.29]:46205 "EHLO thunker.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752426Ab0EHMwW (ORCPT ); Sat, 8 May 2010 08:52:22 -0400 Content-Disposition: inline In-Reply-To: <1273280081.2444.5.camel@faldara> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Fri, May 07, 2010 at 08:54:41PM -0400, Phillip Susi wrote: > Would it be possible to somehow keep the current buffer heads, but > associate them with the inode such that readahead() on the directory > would work? Maybe. Try it and send patches. :-) As I said, it's not high on my priority list and I'm *way* behind on all sorts of other, much higher priority tasks. If you were going to do the patches (which would have all sorts of complications since you'd have to make sure directory pages didn't get written back via the page write cache or via direct reclaim), I wouldn't reject them out of hand since there are a few other advantages of doing things that way.... *However*, I suspect it would be easier for you to simply use the FIEMAP ioctl had deal with directories separately from files..... - Ted