From mboxrd@z Thu Jan 1 00:00:00 1970 From: Keir Fraser Subject: Re: [RFC][PATCH 1/2] export NUMA topology from xen Date: Tue, 10 Apr 2007 10:12:22 +0100 Message-ID: References: <20070410010929.GF28281@us.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20070410010929.GF28281@us.ibm.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Ryan Harper , xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org On 10/4/07 02:09, "Ryan Harper" wrote: > nr_nodes : 4 > mem_chunks : node0:0x0000000000000000-0x0000000190000000 > node1:0x0000000190000000-0x0000000300000000 > node2:0x0000000300000000-0x0000000470000000 > node3:0x0000000470000000-0x0000000640000000 > node_to_cpu : node0:0-7 > node1:8-15 > node2:16-23 > node3:24-31 > > I've also reworked the the physinfo call to contain an array of > cpu_to_node elements rather than node_to_cpu to support machines larger > than 64-way. I convert the array back to node_to_cpu for brevity in > xm info display. The same would make sense for memory regions (i.e., have a list of memory-regions and include a node identifier for each one, rather than mapping node-id to memory-region) as this would allow to have multiple memory regions per node quite easily. But actually I'm not convinced that allowing dom0 to read out the physical addresses of memory regions is at all useful -- why would anyone care which particular physical address ranges belong to a particular node? The hypercall to find amount of free memory per node seems more useful, and probably sufficient by itself. -- Keir