From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759445Ab0I0VRK (ORCPT ); Mon, 27 Sep 2010 17:17:10 -0400 Received: from ogre.sisk.pl ([217.79.144.158]:47363 "EHLO ogre.sisk.pl" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757032Ab0I0VRI (ORCPT ); Mon, 27 Sep 2010 17:17:08 -0400 From: "Rafael J. Wysocki" To: Nigel Cunningham Subject: Re: [PATCH 02/23] Record & display i/o speed post resume. Date: Mon, 27 Sep 2010 23:16:30 +0200 User-Agent: KMail/1.13.5 (Linux/2.6.36-rc5-rjw+; KDE/4.4.4; x86_64; ; ) Cc: Linux PM , LKML , "TuxOnIce-devel" References: <1285566238-10966-1-git-send-email-nigel@tuxonice.net> <201009272249.59966.rjw@sisk.pl> <4CA10729.5010001@tuxonice.net> In-Reply-To: <4CA10729.5010001@tuxonice.net> MIME-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-2" Content-Transfer-Encoding: 7bit Message-Id: <201009272316.30718.rjw@sisk.pl> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Monday, September 27, 2010, Nigel Cunningham wrote: > Hi. > > On 28/09/10 06:49, Rafael J. Wysocki wrote: > > On Monday, September 27, 2010, Nigel Cunningham wrote: ... > >> Well then, for now, let's drop that patch then. At least it's got us the > >> numbers we needed to see these patches are useful. Longer term, I'll do > >> what I do in TuxOnIce and allocate a page while writing the image, store > >> it's location in the image header and use it the transfer information > >> from the boot kernel to the resumed kernel. (Essentially the same thing, > >> except without the special section). > > > > The problem is that on x86_64 the boot kernel may be completely different > > from the image kernel (different version, different configuration), so there is > > no guarantee that your page frame allocated while writing the image would be > > available to the boot kernel (it may contain the kernel code, for example). > > > > You'd really need to replace one of the image pages with something prepared > > while the image was being read. For that, you'd need to allocate a page before > > creating the image and pass its PFN in the image header. > > Yes - that's what I'm doing. Maybe that's why the __nosavedata changes > you mentioned above haven't bitten me. Well, other than that, the majority of people don't know about that change and they still use the same kernel for boot/image. :-) Thanks, Rafael