From patchwork Wed Apr 24 05:35:17 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: NeilBrown X-Patchwork-Id: 2482311 Return-Path: X-Original-To: patchwork-linux-nfs@patchwork.kernel.org Delivered-To: patchwork-process-083081@patchwork2.kernel.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by patchwork2.kernel.org (Postfix) with ESMTP id A0D2CDF25A for ; Wed, 24 Apr 2013 05:35:30 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751243Ab3DXFf3 (ORCPT ); Wed, 24 Apr 2013 01:35:29 -0400 Received: from cantor2.suse.de ([195.135.220.15]:52461 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751179Ab3DXFf3 (ORCPT ); Wed, 24 Apr 2013 01:35:29 -0400 Received: from relay1.suse.de (unknown [195.135.220.254]) by mx2.suse.de (Postfix) with ESMTP id AB348A3DDF; Wed, 24 Apr 2013 07:35:27 +0200 (CEST) Date: Wed, 24 Apr 2013 15:35:17 +1000 From: NeilBrown To: "Myklebust, Trond" Cc: NFS Subject: [PATCH-RFC] NFS: don't use STABLE writes during writeback. Message-ID: <20130424153517.4222da09@notabene.brown> X-Mailer: Claws Mail 3.9.0 (GTK+ 2.24.14; x86_64-suse-linux-gnu) Mime-Version: 1.0 Sender: linux-nfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-nfs@vger.kernel.org Hi Trond, you might remember little over two years ago we discussed an issue with small O_SYNC writes which resulted in commit b31268ac793fd300da66b9c28bbf0a200339ab96 The Subject: of the emails was "Small O_SYNC writes are no longer NFS_DATA_SYNC" This solved the problem at the time but created a new one. Now *all* small writes are NFS_FILE_SYNC whether they are O_SYNC or not. The following seems to work for me without apparently introducing any regressions. Thoughts? Thanks, NeilBrown commit b31268ac793fd300da66b9c28bbf0a200339ab96 FS: Use stable writes when not doing a bulk flush was a bit heavy handed. The particular problem that lead to this patch was that small writes to an O_SYNC file we being written as UNSTABLE writes followed by a commit. This is appropriate for large writes (which require multiple NFS requests) but for small writes (single NFS request), using NFS_FILE_SYNC is more efficient. So that patch causes the code to select between the two methods depending on how many nfs requests get generated. Unfortunately this ends up applying to non O_SYNC writes as well. In particular if you memory-map a file and update random pages, then when they are eventually written out by writeback they will go as NFS_FILE_SYNC. This is inefficient and slows down the application. So: only set FLUSH_COND_STABLE when wbc->sync_mode is WB_SYNC_ALL. With this patch: O_SYNC writes are NFS_FILE_SYNC for single requests, and NFS_UNSTABLE followed by COMMIT for multiple requests Writing immediately before close of fsync follow the same pattern. Non-O_SYNC writes without an fsync of close eventually get flushed out as UNSTABLE and a commit follows eventually as appropriate. Signed-off-by: NeilBrown diff --git a/fs/nfs/write.c b/fs/nfs/write.c index c483cc5..05220b5 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -225,11 +225,14 @@ static void nfs_mark_uptodate(struct page *page, unsigned int base, unsigned int static int wb_priority(struct writeback_control *wbc) { + int ret = 0; if (wbc->for_reclaim) return FLUSH_HIGHPRI | FLUSH_STABLE; + if (wbc->sync_mode == WB_SYNC_ALL) + ret = FLUSH_COND_STABLE; if (wbc->for_kupdate || wbc->for_background) - return FLUSH_LOWPRI | FLUSH_COND_STABLE; - return FLUSH_COND_STABLE; + ret |= FLUSH_LOWPRI; + return ret; } /*