Return-Path: linux-nfs-owner@vger.kernel.org Received: from linode.aoot.com ([69.164.194.13]:51093 "EHLO linode.aoot.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751238AbaDYNes (ORCPT ); Fri, 25 Apr 2014 09:34:48 -0400 Message-ID: <535A6476.8080207@opengridcomputing.com> Date: Fri, 25 Apr 2014 08:34:46 -0500 From: Steve Wise MIME-Version: 1.0 To: Chuck Lever , "J. Bruce Fields" CC: Linux NFS Mailing List , linux-rdma@vger.kernel.org, tom@ogc.us Subject: Re: [PATCH] svcrdma: refactor marshalling logic References: <20140410183723.29724.18665.stgit@build.ogc.int> <20140425105838.GE22295@fieldses.org> <6B12FA09-0E8C-497A-9145-E2B7CAB16C53@oracle.com> In-Reply-To: <6B12FA09-0E8C-497A-9145-E2B7CAB16C53@oracle.com> Content-Type: text/plain; charset=windows-1252; format=flowed Sender: linux-nfs-owner@vger.kernel.org List-ID: On 4/25/2014 8:24 AM, Chuck Lever wrote: > Hi Bruce- > > On Apr 25, 2014, at 6:58 AM, J. Bruce Fields wrote: > >> On Thu, Apr 10, 2014 at 01:37:23PM -0500, Steve Wise wrote: >>> From: Tom Tucker >>> >>> This patch refactors the marshalling logic to remove the intermediary >>> map structures. It also fixes an existing bug where the NFSRDMA server >>> was not minding the device fast register page list length limitations. >>> >>> Signed-off-by: Tom Tucker >>> --- >>> >>> include/linux/sunrpc/svc_rdma.h | 3 >>> net/sunrpc/xprtrdma/svc_rdma_recvfrom.c | 551 +++++++++--------------------- >>> net/sunrpc/xprtrdma/svc_rdma_sendto.c | 230 +------------ >>> net/sunrpc/xprtrdma/svc_rdma_transport.c | 57 ++- >>> 4 files changed, 222 insertions(+), 619 deletions(-) >> Is it possible to make this change in more than one step? >> >> RDMA is pretty esoteric to most of us, so honestly this will probably >> get merged based just on your having tested it, but if it was possible >> to break this up into smaller patches you might give us at least a >> fighting chance of giving it some review?. > I agree it could be broken up. > > However, my testing revealed that the patch inadvertently breaks support > for HCAs that do not support FRMR. I?ve reported this to Steve and Tom. > IMO you should wait for a newer version of the refactoring patch. I'll figure out the regression, and work with tom to repost this as a series. Steve.