From patchwork Wed Apr 6 18:44:06 2011 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Josef Bacik X-Patchwork-Id: 690341 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by demeter1.kernel.org (8.14.4/8.14.3) with ESMTP id p36It0Xq028461 for ; Wed, 6 Apr 2011 18:55:00 GMT Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756566Ab1DFSy5 (ORCPT ); Wed, 6 Apr 2011 14:54:57 -0400 Received: from mx1.redhat.com ([209.132.183.28]:61417 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756508Ab1DFSy4 (ORCPT ); Wed, 6 Apr 2011 14:54:56 -0400 Received: from int-mx12.intmail.prod.int.phx2.redhat.com (int-mx12.intmail.prod.int.phx2.redhat.com [10.5.11.25]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id p36IsuRQ001645 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Wed, 6 Apr 2011 14:54:56 -0400 Received: from test1244.test.redhat.com (test1244.test.redhat.com [10.10.10.244]) by int-mx12.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id p36Ist0c016401 for ; Wed, 6 Apr 2011 14:54:55 -0400 From: Josef Bacik To: linux-btrfs@vger.kernel.org Subject: [PATCH] Btrfs: do not use async submit for small DIO io's Date: Wed, 6 Apr 2011 14:44:06 -0400 Message-Id: <1302115446-3046-1-git-send-email-josef@redhat.com> X-Scanned-By: MIMEDefang 2.68 on 10.5.11.25 Sender: linux-btrfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-btrfs@vger.kernel.org X-Greylist: IP, sender and recipient auto-whitelisted, not delayed by milter-greylist-4.2.6 (demeter1.kernel.org [140.211.167.41]); Wed, 06 Apr 2011 18:55:00 +0000 (UTC) When looking at our DIO performance Chris said that for small IO's doing the async submit stuff tends to be more overhead than it's worth. With this on top of my other fixes I get about a 17-20% speedup doing a sequential dd with 4k IO's. Basically if we don't have to split the bio for the map length it's small enough to be directly submitted, otherwise go back to the async submit. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 24 +++++++++++++++++++----- 1 files changed, 19 insertions(+), 5 deletions(-) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5144d5c..93ed346 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -5955,7 +5955,7 @@ static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev, static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode, int rw, u64 file_offset, int skip_sum, - u32 *csums) + u32 *csums, int async_submit) { int write = rw & REQ_WRITE; struct btrfs_root *root = BTRFS_I(inode)->root; @@ -5966,13 +5966,24 @@ static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode, if (ret) goto err; - if (write && !skip_sum) { + if (skip_sum) + goto map; + + if (write && async_submit) { ret = btrfs_wq_submit_bio(root->fs_info, inode, rw, bio, 0, 0, file_offset, __btrfs_submit_bio_start_direct_io, __btrfs_submit_bio_done); goto err; + } else if (write) { + /* + * If we aren't doing async submit, calculate the csum of the + * bio now. + */ + ret = btrfs_csum_one_bio(root, inode, bio, file_offset, 1); + if (ret) + goto err; } else if (!skip_sum) { ret = btrfs_lookup_bio_sums_dio(root, inode, bio, file_offset, csums); @@ -5980,7 +5991,8 @@ static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode, goto err; } - ret = btrfs_map_bio(root, rw, bio, 0, 1); +map: + ret = btrfs_map_bio(root, rw, bio, 0, async_submit); err: bio_put(bio); return ret; @@ -6002,6 +6014,7 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, int nr_pages = 0; u32 *csums = dip->csums; int ret = 0; + int async_submit = 0; int write = rw & REQ_WRITE; map_length = orig_bio->bi_size; @@ -6017,6 +6030,7 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, goto submit; } + async_submit = 1; bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS); if (!bio) return -ENOMEM; @@ -6037,7 +6051,7 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, atomic_inc(&dip->pending_bios); ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum, - csums); + csums, async_submit); if (ret) { bio_put(bio); atomic_dec(&dip->pending_bios); @@ -6076,7 +6090,7 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip, submit: ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum, - csums); + csums, async_submit); if (!ret) return 0;