Return-Path: Received: from daytona.panasas.com ([67.152.220.89]:39663 "EHLO daytona.panasas.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750706Ab1EYQ5F (ORCPT ); Wed, 25 May 2011 12:57:05 -0400 Message-ID: <4DDD34DE.4080409@panasas.com> Date: Wed, 25 May 2011 19:57:02 +0300 From: Benny Halevy To: Boaz Harrosh CC: Trond Myklebust , linux-nfs@vger.kernel.org Subject: Re: [PATCH] SQUASHME: pnfs-obj: pg_test check for max_io_size References: <4DDA8C3D.5080706@panasas.com> <1306168699-11685-1-git-send-email-bhalevy@panasas.com> <4DDD0675.9080900@panasas.com> <4DDD06F6.7070508@panasas.com> In-Reply-To: <4DDD06F6.7070508@panasas.com> Content-Type: text/plain; charset=UTF-8 Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 On 2011-05-25 16:41, Boaz Harrosh wrote: > > This should be squashed into: > pnfs-obj: osd raid engine read/write implementation > > Implement pg_test vector to test for max IO sizes. We calculate > a max_io_size member only once, and cache it in lseg so to not > do so on every page insert. Looks good (except one minor nit, see inline below). I prefer to commit it as a separate patch rather than squash it if ok with you. Benny > > Signed-off-by: Boaz Harrosh > --- > fs/nfs/objlayout/objio_osd.c | 14 +++++++++++++- > 1 files changed, 13 insertions(+), 1 deletions(-) > > diff --git a/fs/nfs/objlayout/objio_osd.c b/fs/nfs/objlayout/objio_osd.c > index 0dcffc2..0dba6ec 100644 > --- a/fs/nfs/objlayout/objio_osd.c > +++ b/fs/nfs/objlayout/objio_osd.c > @@ -129,6 +129,8 @@ struct objio_segment { > u64 group_depth; > unsigned group_count; > > + unsigned max_io_size; > + > unsigned comps_index; > unsigned num_comps; > /* variable length */ > @@ -366,6 +368,11 @@ extern int objio_alloc_lseg(struct pnfs_layout_segment **outp, > objio_seg->group_count = 1; > } > > + /* Cache this calculation it will hit for every page */ > + objio_seg->max_io_size = (BIO_MAX_PAGES_KMALLOC * PAGE_SIZE - > + objio_seg->stripe_unit) * > + objio_seg->group_width; > + > *outp = &objio_seg->lseg; > return 0; > > @@ -1000,7 +1007,12 @@ int > objlayout_pg_test(struct nfs_pageio_descriptor *pgio, struct nfs_page *prev, > struct nfs_page *req) > { > - return 1; > + struct objio_segment *layout = OBJIO_LSEG(pgio->pg_lseg); > + > + if ((pgio->pg_count + req->wb_bytes) > layout->max_io_size) > + return 0; > + else > + return 1; why not simply do this? return (pgio->pg_count + req->wb_bytes) <= layout->max_io_size; Benny > } > > static struct pnfs_layoutdriver_type objlayout_type = {