[PATCH 04/24] GFS2: Make atime checks more efficient

From: Steven Whitehouse
Date: Mon Oct 24 2011 - 08:52:40 EST


We do not need to start a transaction unless the atime
check has proved positive. Also if we are going to flush
the complete ail list anyway, we might as well skip the
writeback for this specific inode's metadata, since that
will be done as part of the ail writeback process in an
order offering potentially more efficient I/O.

Signed-off-by: Steven Whitehouse <swhiteho@xxxxxxxxxx>

diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index b7beadd..afb8761 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -768,30 +768,30 @@ static int gfs2_write_inode(struct inode *inode, struct writeback_control *wbc)
goto do_flush;
unlock_required = 1;
}
- ret = gfs2_trans_begin(sdp, RES_DINODE, 0);
- if (ret)
- goto do_unlock;
ret = gfs2_meta_inode_buffer(ip, &bh);
if (ret == 0) {
di = (struct gfs2_dinode *)bh->b_data;
atime.tv_sec = be64_to_cpu(di->di_atime);
atime.tv_nsec = be32_to_cpu(di->di_atime_nsec);
if (timespec_compare(&inode->i_atime, &atime) > 0) {
- gfs2_trans_add_bh(ip->i_gl, bh, 1);
- gfs2_dinode_out(ip, bh->b_data);
+ ret = gfs2_trans_begin(sdp, RES_DINODE, 0);
+ if (ret == 0) {
+ gfs2_trans_add_bh(ip->i_gl, bh, 1);
+ gfs2_dinode_out(ip, bh->b_data);
+ gfs2_trans_end(sdp);
+ }
}
brelse(bh);
}
- gfs2_trans_end(sdp);
-do_unlock:
if (unlock_required)
gfs2_glock_dq_uninit(&gh);
do_flush:
if (wbc->sync_mode == WB_SYNC_ALL)
gfs2_log_flush(GFS2_SB(inode), ip->i_gl);
- filemap_fdatawrite(metamapping);
if (bdi->dirty_exceeded)
gfs2_ail1_flush(sdp, wbc);
+ else
+ filemap_fdatawrite(metamapping);
if (!ret && (wbc->sync_mode == WB_SYNC_ALL))
ret = filemap_fdatawait(metamapping);
if (ret)
--
1.7.4.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/