From mboxrd@z Thu Jan 1 00:00:00 1970 From: Len Brown Subject: Re: [PATCH RESEND] tools: add power/x86/x86_energy_perf_policy to program MSR_IA32_ENERGY_PERF_BIAS Date: Mon, 22 Nov 2010 23:48:26 -0500 (EST) Message-ID: References: <8739r0rxlz.fsf@basil.nowhere.org> <20101122203359.GD21836@basil.fritz.box> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Return-path: In-reply-to: <20101122203359.GD21836@basil.fritz.box> Sender: linux-kernel-owner@vger.kernel.org To: Andi Kleen Cc: Greg Kroah-Hartman , linux-pm@lists.linux-foundation.org, linux-kernel@vger.kernel.org, linux-acpi@vger.kernel.org, x86@kernel.org List-Id: linux-acpi@vger.kernel.org On Mon, 22 Nov 2010, Andi Kleen wrote: > On Mon, Nov 22, 2010 at 03:13:24PM -0500, Len Brown wrote: > > Per the comments from Andrew and others, the concept of a > > "full tools build" doesn't actually exit (yet). > > > > So I guess the only assurance that somebody not on x86 would run > > make in this directory this utility lives in tools/power/x86/ > > > > Note that there are other utilities under tools > > which have no Makefile at all... > > I suspect this will need to be fixed at some point. > > e.g. kernel rpms probably don't want to hard code all of this > but just call some standard make file target. And the kernel > eventually needs a make install_user or similar. I agree, but I don't volunteer to set up such a build system as part of this particular patch. As I mentioned, supplying any Makefile is a step better than some of the peers... > > I'm not inclined to bother, as the use-case for this utility > > is to be invoked by another program, and the options available > > What other program? > > I could well imagine administrators sticking this > into their boot.locals to set the policy they want. right, and that would be a program. It is unlikely that users are going to be typing this command, except into an admin script. > > In the highly unlikely scenario that somebody uses > > the -r option to excerise the read-only code, > > and simultaneously invokes and completes a cpu hot remove > > FWIW there are setups where core offlining can happen > automatically in response to an error. Understood. I think it is fine if this utility simply exits if that error occurs while it is running. (turbostat, OTOH, may be long running, and it treats vanishing processors as a recoverable error) thanks, -Len Brown, Intel Open Source Technology Center