ceph: set up readahead size when rsize is not passed
authorYehuda Sadeh <yehuda@hq.newdream.net>
Fri, 22 Jul 2011 18:12:28 +0000 (11:12 -0700)
committerSage Weil <sage@newdream.net>
Tue, 26 Jul 2011 18:29:14 +0000 (11:29 -0700)
This should improve the default read performance, as without it
readahead is practically disabled.

Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
fs/ceph/super.c

index d4757c98ec1400f9a2328e8e17c1c8e2a8a8c836..d47c5ec7fb1ff0500ebbeb7f403debfa865c87a0 100644 (file)
@@ -779,6 +779,10 @@ static int ceph_register_bdi(struct super_block *sb,
                fsc->backing_dev_info.ra_pages =
                        (fsc->mount_options->rsize + PAGE_CACHE_SIZE - 1)
                        >> PAGE_SHIFT;
+       else
+               fsc->backing_dev_info.ra_pages =
+                       default_backing_dev_info.ra_pages;
+
        err = bdi_register(&fsc->backing_dev_info, NULL, "ceph-%d",
                           atomic_long_inc_return(&bdi_seq));
        if (!err)