From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dennis Dalessandro Subject: Re: [PATCH v2 19/20] IB/rdmavt, IB/qib, IB/hfi1: Make percpu refcount optional for user MRs Date: Wed, 5 Apr 2017 15:46:48 -0400 Message-ID: References: <20170321001900.28538.38175.stgit@scvm10.sc.intel.com> <20170321002631.28538.2121.stgit@scvm10.sc.intel.com> <1491417489.2923.6.camel@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1491417489.2923.6.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Doug Ledford Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Mike Marciniszyn List-Id: linux-rdma@vger.kernel.org On 04/05/2017 02:38 PM, Doug Ledford wrote: > On Mon, 2017-03-20 at 17:26 -0700, Dennis Dalessandro wrote: >> +static unsigned int hfi1_no_user_mr_percpu; >> +module_param_named(no_user_mr_percpu, hfi1_no_user_mr_percpu, uint, >> + S_IRUGO); >> +MODULE_PARM_DESC(no_user_mr_percpu, >> + "Avoid percpu refcount for user MRs (default 0)"); >> + > > Does this have to be a module parameter? Those are frowned upon now a > days... > Yeah I don't like it either really, but there is a pretty big tradeoff between memory deregistration improvement and cost to the data path. Most use cases care about the data path but some uses register and deregister memory a lot and this helps them. Is there a another way we should be looking at for setting things like this? -Denny -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html