From: Eric Sandeen Subject: Re: [PATCH 3/3] ext4: Use readahead when reading an inode from the inode table Date: Fri, 26 Sep 2008 08:47:24 -0500 Message-ID: <48DCE7EC.5000801@redhat.com> References: <1222275213-16268-1-git-send-email-tytso@mit.edu> <1222275213-16268-2-git-send-email-tytso@mit.edu> <1222275213-16268-3-git-send-email-tytso@mit.edu> <20080926082107.GT10950@webber.adilger.int> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: "Theodore Ts'o" , Ext4 Developers List To: Andreas Dilger Return-path: Received: from mx2.redhat.com ([66.187.237.31]:40888 "EHLO mx2.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751475AbYIZNrs (ORCPT ); Fri, 26 Sep 2008 09:47:48 -0400 In-Reply-To: <20080926082107.GT10950@webber.adilger.int> Sender: linux-ext4-owner@vger.kernel.org List-ID: Andreas Dilger wrote: > On Sep 24, 2008 12:53 -0400, Theodore Ts'o wrote: >> @@ -514,8 +514,10 @@ static void ext4_put_super(struct super_block *sb) >> + if (sbi->s_proc) { >> + remove_proc_entry("inode_readahead_bits", sbi->s_proc); >> remove_proc_entry(sb->s_id, ext4_proc_root); > > As a general UI interface, specifying the "bits to shift the filesystem > block number" seems like an easy-to-implement but is fairly bad from > a usability point of view. I'd much prefer to specify this as a > number of kB to readahead, and it can be converted internally to the > number of blocks to readahead. It isn't fatal if we do a bit of rounding > on the input value to match a full blocksize. Or specify that it must be a power of two in some range, and reject other values - either way. I had the same thought about the nr of bits as a user input value... -Eric