From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Kasagar, Srinidhi" Subject: Re: [PATCH v4] cpufreq: Add SFI based cpufreq driver support Date: Tue, 18 Nov 2014 11:09:58 +0530 Message-ID: <20141118053958.GA28487@intel-desktop> References: <20141104170031.GA25991@intel-desktop> <20141110105606.GA11060@intel-desktop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mga14.intel.com ([192.55.52.115]:48377 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750808AbaKRFjE (ORCPT ); Tue, 18 Nov 2014 00:39:04 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-pm-owner@vger.kernel.org List-Id: linux-pm@vger.kernel.org To: Len Brown Cc: "Rafael J. Wysocki" , Viresh Kumar , Linux PM list , vishwesh.m.rudramuni@intel.com, "Brown, Len" , srinidhi.kasagar@intel.com On Mon, Nov 10, 2014 at 02:10:44PM -0500, Len Brown wrote: > > Not sure if in understood correctly. Do you mean I should not > > even read from PERF_CTL in get_cur_freq_on_cpu, rather just > > return the cached last requested frequency? I'm not sure > > of the behavior in case if we offline and online a cpu. > > Best to simply cache the last written value in software, > that way it can be read quickly w/o even switching to > a processor to do the local MSR read. But it will be wrong after offline-online sequence. I have sent v5 version of the patch. Please check. > > The cpufreq get-current-frequency interface is somewhat bogus. > It generally doesn't make sense on a lot of modern hardware for the > driver to actually know the current frequency -- as the driver is only making > a request and does not have total control over the frequency... > But we need to answer something b/c there are people still > using the legacy cpufreq sysfs interface, and they expect a number. Agree. Srinidhi