Re: [PATCH] direct-io: Submit bio after boundary buffer is added to it

[Jan Kara <jack@suse.cz>]

On Tue 09-04-13 13:47:58, Andrew Morton wrote:
> On Tue,  9 Apr 2013 17:42:49 +0200 Jan Kara <jack@suse.cz> wrote:
> 
> > Currently, dio_send_cur_page() submits bio before current page and
> > cached sdio->cur_page is added to the bio if sdio->boundary is set. This
> > is actually wrong because sdio->boundary means the current buffer is the
> > last one before metadata needs to be read. So we should rather submit
> > the bio after the current page is added to it.
> > 
> > Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com>
> > Signed-off-by: Jan Kara <jack@suse.cz>
> > ---
> >  fs/direct-io.c |   28 +++++++++++-----------------
> >  1 files changed, 11 insertions(+), 17 deletions(-)
> > 
> >  Andrew, can you add this patch the previous direct IO boundary handling fix?
> 
> What is the user-visible impact of these two patches?
  I've measured about 10% throughput improvement of direct IO reads on ext3
with SATA harddrive (from 90 MB/s to 100 MB/s) with these patches applied.
With ramdisk, the improvement was about 3-fold (from 350 MB/s to 1.2 GB/s).
For other filesystems (such as ext4), the improvements won't be as visible
because the frequency of BH_Boundary flag being set is much smaller...

								Honza
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR