From mboxrd@z Thu Jan 1 00:00:00 1970 From: Anthony Liguori Subject: Re: [Qemu-devel] [RFC PATCH 0/9] ACPI memory hotplug Date: Thu, 19 Apr 2012 09:49:31 -0500 Message-ID: <4F9025FB.2020900@codemonkey.ws> References: <1334844527-18869-1-git-send-email-vasilis.liaskovitis@profitbricks.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: qemu-devel@nongnu.org, kvm@vger.kernel.org, seabios@seabios.org, avi@redhat.com, gleb@redhat.com, Wanpeng Li To: Vasilis Liaskovitis Return-path: Received: from mail-ob0-f174.google.com ([209.85.214.174]:42152 "EHLO mail-ob0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752273Ab2DSOtf (ORCPT ); Thu, 19 Apr 2012 10:49:35 -0400 Received: by obbta14 with SMTP id ta14so8171321obb.19 for ; Thu, 19 Apr 2012 07:49:35 -0700 (PDT) In-Reply-To: <1334844527-18869-1-git-send-email-vasilis.liaskovitis@profitbricks.com> Sender: kvm-owner@vger.kernel.org List-ID: On 04/19/2012 09:08 AM, Vasilis Liaskovitis wrote: > This is a prototype for ACPI memory hotplug on x86_64 target. Based on some > earlier work and comments from Gleb. > > Memslot devices are modeled with a new qemu command line > > "-memslot id=name,start=start_addr,size=sz,node=pxm" Hi, For 1.2, I'd really like to focus on refactoring the PC machine as described in this series: https://github.com/aliguori/qemu/commits/qom-rebase.12 I'd like to represent the guest memory as a "DIMM" device. In terms of this proposal, I would then expect that the i440fx device would have a num_dimms property that controlled how many link's it had. Hotplug would consist of creating a DIMM at run time and connecting it to the appropriate link. One thing that's not clear to me is how the start/size fits in. On bare metal, is this something that's calculated by the firmware during start up and then populated in ACPI? Does it do something like take the largest possible DIMM size that it supports and fill out the table? At any rate, I think we should focus on modeling this in QOM verses adding a new option and hacking at the existing memory init code. Regards, Anthony Liguori