From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: [kvm-ppc-devel] RFC/patch portability: split kvm_vm_ioctl v2 Date: Fri, 26 Oct 2007 12:36:32 +0200 Message-ID: <4721C330.3030302@qumranet.com> References: <1192192452.7630.16.camel@cotte.boeblingen.de.ibm.com> <1193327325.8345.9.camel@cotte.boeblingen.de.ibm.com> <1193327326.3284.2.camel@izike-woof.qumranet.com> <1193329322.28279.21.camel@basalt> <472114B6.6080000@qumranet.com> <1193345465.28279.47.camel@basalt> <4721A185.1070808@qumranet.com> <4721A3A3.8020504@de.ibm.com> <4721A82A.2040402@de.ibm.com> <4721B1ED.6040006@qumranet.com> <4721C1DB.3040207@de.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Cc: "kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org" , kvm-ppc-devel , carsteno-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org, Hollis Blanchard , "Zhang, Xiantao" To: carsteno-tA70FqPdS9bQT0dZR+AlfA@public.gmane.org Return-path: In-Reply-To: <4721C1DB.3040207-tA70FqPdS9bQT0dZR+AlfA@public.gmane.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: kvm-devel-bounces-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org Errors-To: kvm-devel-bounces-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org List-Id: kvm.vger.kernel.org Carsten Otte wrote: > Avi Kivity wrote: >> I don't really see a big difference between what we have now (sparse >> userspace, sparse guest) and Izik's idea (contiguous userspace, >> sparse guest). In both cases you need something like memory slots to >> describe the different sections. > We don't on s390: we receive a page fault by the guest once it > accesses a sparse hole in its address space. We check the user space's > VMA and either page it in or submit an addressing exception to the guest. I was talking about x86. On x86, you need contiguous userspace, contiguous guest, but again, what's the problem with one memory slot? > >> Moreover, on x86 you may want different properties for different >> sections (4K pages for the framebuffer, large pages for main memory), >> so you can't allocate memory in one big chunk. > That's right. On s390, we can live with whatever properties a section > has with regard to page size, backing device and such. So memory may > well come to live by different alloations, and different allocation > methods. All we need is a permanent contiguous mapping of the guest > physical addresses to host user addresses. So Izik's idea would work > for us even if we have different sections. So would the current memory slot thing, no? -- error compiling committee.c: too many arguments to function ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/