Return-Path: Received: from userp1040.oracle.com ([156.151.31.81]:18926 "EHLO userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752175AbdHPUDt (ORCPT ); Wed, 16 Aug 2017 16:03:49 -0400 Content-Type: text/plain; charset=us-ascii Mime-Version: 1.0 (Mac OS X Mail 9.3 \(3124\)) Subject: Re: [PATCH v2 5/6] SUNRPC: Add a function to allow waiting for RPC transmission From: Chuck Lever In-Reply-To: <1502913733.5613.5.camel@primarydata.com> Date: Wed, 16 Aug 2017 16:03:44 -0400 Cc: Linux NFS Mailing List Message-Id: <146ED3FC-ED3C-4153-A410-FC57B1C4A161@oracle.com> References: <20170816161929.11271-1-trond.myklebust@primarydata.com> <20170816161929.11271-2-trond.myklebust@primarydata.com> <20170816161929.11271-3-trond.myklebust@primarydata.com> <20170816161929.11271-4-trond.myklebust@primarydata.com> <20170816161929.11271-5-trond.myklebust@primarydata.com> <20170816161929.11271-6-trond.myklebust@primarydata.com> <1502913733.5613.5.camel@primarydata.com> To: Trond Myklebust Sender: linux-nfs-owner@vger.kernel.org List-ID: > On Aug 16, 2017, at 4:02 PM, Trond Myklebust wrote: > > On Wed, 2017-08-16 at 12:43 -0400, Chuck Lever wrote: >>> On Aug 16, 2017, at 12:19 PM, Trond Myklebust >> arydata.com> wrote: >>> >>> Sometimes, we only want to know that the RPC message is in the >>> process >>> of being transmitted. Add a function to allow that. >> >> The last time I tried these last two, they resulted in a significant >> performance regression. Is there anything I can do to help track >> that down? > > If they are the root cause of a performance regression, then I'll just > drop them. I've been rebasing to ensure that I keep them at the end of > the queue precisely because I was unsure of their effect on overall > performance. My experience was applying the whole stack of the original 20 resulted in some performance degradations (fio). Popping these two from that stack restored performance across the suite of tests I was doing. -- Chuck Lever