From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alex Elder Subject: Re: [PATCH 1/3] rbd: skip the copyup when an entire object writing Date: Tue, 11 Mar 2014 23:39:31 -0500 Message-ID: <531FE503.7030000@ieee.org> References: <1394598285-17225-1-git-send-email-lucienchao@gmail.com> <1394598285-17225-2-git-send-email-lucienchao@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: Received: from mail-ie0-f171.google.com ([209.85.223.171]:55691 "EHLO mail-ie0-f171.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751315AbaCLEjb (ORCPT ); Wed, 12 Mar 2014 00:39:31 -0400 Received: by mail-ie0-f171.google.com with SMTP id ar20so10145461iec.30 for ; Tue, 11 Mar 2014 21:39:31 -0700 (PDT) In-Reply-To: <1394598285-17225-2-git-send-email-lucienchao@gmail.com> Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Guangliang Zhao , ceph-devel@vger.kernel.org Cc: sage@inktank.com On 03/11/2014 11:24 PM, Guangliang Zhao wrote: > It need to copyup the parent's content when layered writing, > but an entire object write would overwrite it, so skip it. This is a pretty reasonable optimization. Has anyone found how often this case is hit? Is it a big win? I think the patch looks good but I have something for you to consider, below. Either way: Reviewed-by: Alex Elder > Signed-off-by: Guangliang Zhao > --- > drivers/block/rbd.c | 8 +++++++- > 1 file changed, 7 insertions(+), 1 deletion(-) > > diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c > index b365e0d..965b9b9 100644 > --- a/drivers/block/rbd.c > +++ b/drivers/block/rbd.c > @@ -2629,12 +2629,14 @@ static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) > struct rbd_img_request *img_request; > struct rbd_device *rbd_dev; > bool known; > + u64 obj_size; > > rbd_assert(obj_request_img_data_test(obj_request)); > > img_request = obj_request->img_request; > rbd_assert(img_request); > rbd_dev = img_request->rbd_dev; > + obj_size = (u64) 1 << rbd_dev->header.obj_order; > > /* > * Only writes to layered images need special handling. > @@ -2644,11 +2646,15 @@ static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) > * simple object requests. Finally, if the target object is > * known to already exist, its parent data has already been > * copied, so a write to the object can also be handled as a > - * simple object request. > + * simple object request. Another type: if the obj_request > + * aligns with the boundary and equals to the size of an object, > + * it doesn't need copyup, because the obj_request will overwrite > + * it finally. > */ This test is awfully big, and it's now even bigger. I think it would be good to encapsulate it into a helper function. Something like rbd_obj_request_simple() (or figure out a better name). So it might then look like: /* * If it's a simple image object request, no special * handling is required. */ if (rbd_img_obj_request_simple(obj_request)) { struct rbd_device *rbd_dev; struct ceph_osd_client *osdc; return rbd_obj_request_submit(osdc, obj_request); } The called function could probably be made a little easier to understand by splitting up the conditions and returning more than once. E.g.: /* Only layered writes need special handling. */ img_request = obj_request->img_request; rbd_dev = img_request->rbd_dev; if (!img_request_write_test(obj_request)) return true; if (!img_request_layered_test(img_request)) return true; /* * Layered writes that start beyond the end of the overlap * with the parent have no parent data, so they too are * simple object requests. */ if (rbd_dev->parent_overlap <= obj_request->img_offset) return true; And so on. > if (!img_request_write_test(img_request) || > !img_request_layered_test(img_request) || > rbd_dev->parent_overlap <= obj_request->img_offset || > + ((!obj_request->offset) && (obj_request->length == obj_size)) || > ((known = obj_request_known_test(obj_request)) && > obj_request_exists_test(obj_request))) { > >