From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from iad1-shared-relay2.dreamhost.com ([208.113.157.41]:45155 "EHLO iad1-shared-relay2.dreamhost.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752403AbdF3KVU (ORCPT ); Fri, 30 Jun 2017 06:21:20 -0400 Received: from iad1-shared-relay1.dreamhost.com (iad1-shared-relay1.dreamhost.com [208.113.157.50]) by iad1-shared-relay2.dreamhost.com (Postfix) with ESMTP id C22053E567F for ; Thu, 29 Jun 2017 15:21:08 -0700 (PDT) Received: from ware.dreamhost.com (ware.dreamhost.com [64.111.127.160]) by iad1-shared-relay1.dreamhost.com (Postfix) with ESMTP id 0946AB40084 for ; Thu, 29 Jun 2017 15:21:08 -0700 (PDT) In-Reply-To: <20170629134510.GA32385@infradead.org> References: <20170629134510.GA32385@infradead.org> From: Tang Junhui Date: Tue, 9 May 2017 12:05:08 -0700 Subject: [PATCH 02/19] bcache: fix sequential large write IO bypass MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 To: linux-block@vger.kernel.org Message-Id: <20170629222107.D41A5100128@ware.dreamhost.com> Sender: linux-block-owner@vger.kernel.org List-Id: linux-block@vger.kernel.org Sequential write IOs were tested with bs=1M by FIO in writeback cache mode, these IOs were expected to be bypassed, but actually they did not. We debug the code, and find in check_should_bypass(): if (!congested && mode == CACHE_MODE_WRITEBACK && op_is_write(bio_op(bio)) && (bio-＞bi_opf & REQ_SYNC)) goto rescale that means, If in writeback mode, a write IO with REQ_SYNC flag will not be bypassed though it is a sequential large IO, It's not a correct thing to do actually, so this patch remove these codes. Signed-off-by: tang.junhui Reviewed-by: Kent Overstreet Reviewed-by: Eric Wheeler Cc: stable@vger.kernel.org --- drivers/md/bcache/request.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c index 019b3df..958072a 100644 --- a/drivers/md/bcache/request.c +++ b/drivers/md/bcache/request.c @@ -400,12 +400,6 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio) if (!congested && !dc->sequential_cutoff) goto rescale; - if (!congested && - mode == CACHE_MODE_WRITEBACK && - op_is_write(bio->bi_opf) && - op_is_sync(bio->bi_opf)) - goto rescale; - spin_lock(&dc->io_lock); hlist_for_each_entry(i, iohash(dc, bio->bi_iter.bi_sector), hash) -- 1.8.3.1