From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeremy Fitzhardinge Subject: Re: [Xen-devel] [PATCH][pvops_dom0][2/4] Introduce the external control operation interface for domain0 ACPI parser Date: Thu, 30 Jul 2009 15:04:52 -0700 Message-ID: <4A721904.7040406@goop.org> References: <4D05DB80B95B23498C72C700BD6C2E0B315D64CC@pdsmsx502.ccr.corp.intel.com> <4D05DB80B95B23498C72C700BD6C2E0B315D671A@pdsmsx502.ccr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: Received: from claw.goop.org ([74.207.240.146]:33882 "EHLO claw.goop.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751828AbZG3WEw (ORCPT ); Thu, 30 Jul 2009 18:04:52 -0400 In-Reply-To: Sender: linux-acpi-owner@vger.kernel.org List-Id: linux-acpi@vger.kernel.org To: Len Brown Cc: "Yu, Ke" , "linux-acpi@vger.kernel.org" , "Tian, Kevin" On 07/30/09 09:29, Len Brown wrote: > Unclear that the power management partitioning between xen hypervisor > and dom0 is fully baked. > > Uncear (to me) what xen is doing internally with these power management > objects, and how that differs from what Linux would do. > Yes. The key thing is that Xen is the only entity which really knows about physical CPUs and how they're being used, and so is the only thing which can correctly apply the chosen policy. If any particular guest domain did it, it would only take into account that particular domain's CPU use, and ignore everyone else (or have the extra complexity of extracting system-wide usage from Xen then applying that to its own policy). If I understand correctly, the code currently relies on Linux running the _PSD method with its AML interpreter, and then feeding the results to Xen as it doesn't have an AML interpreter. And putting AML into Xen would be an all-or-nothing proposition, because the entity which runs AML maintains a lot of state which can't be separated between Xen and Linux, and can't be shared. J