From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 27 Nov 2001 02:16:32 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 27 Nov 2001 02:16:13 -0500 Received: from ns.suse.de ([213.95.15.193]:63759 "HELO Cantor.suse.de") by vger.kernel.org with SMTP id ; Tue, 27 Nov 2001 02:16:04 -0500 To: "David S. Miller" Cc: linux-kernel@vger.kernel.org, neilb@cse.unsw.edu.au Subject: Re: Fix knfsd readahead cache in 2.4.15 In-Reply-To: <15362.18626.303009.379772@charged.uio.no.suse.lists.linux.kernel> <15362.53694.192797.275363@esther.cse.unsw.edu.au.suse.lists.linux.kernel> <20011126.155347.45872112.davem@redhat.com.suse.lists.linux.kernel> From: Andi Kleen Date: 27 Nov 2001 08:16:03 +0100 In-Reply-To: "David S. Miller"'s message of "27 Nov 2001 00:58:40 +0100" Message-ID: X-Mailer: Gnus v5.7/Emacs 20.7 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org "David S. Miller" writes: > > There are other problems remaining, this function is a logical > mess. It is also broken by design. It forces file systems that need more state to maintain read/write ahead to maintain their own cache (e.g. XFS which needs to cache extent allocations) Better would be to just fix nfsd to cache open file handles, then all this mess would be done by the VFS. In addition it would be faster because it wouldn't need to execute an "open()" for every rpc operation. -Andi