Re: [libvirt] [PATCH 0/6] Support for per-guest-node binding

On Thu, Jun 26, 2014 at 01:50:22AM +0000, chen.fan.fnst(a)cn.fujitsu.com wrote: >On Wed, 2014-06-04 at 16:56 +0200, Martin Kletzander wrote: >> Currently we are only able to bind the whole domain to some host nodes >> using the /domain/numatune/memory element. Numerous requests were >> made to support host<->guest numa node bindings, so this series tries >> to pinch an idea on how to do that using /domain/numatune/memnode >> elements. >> >> That is incompatible with automatic numa placement (numad) since that >> makes no sense. Also this disables any live changes to numa >> parameters (the /domain/numatune/memory settings) since we cannot >> change the settings given to qemu. >Hi Martin, > > Sorry for that I have not observed this patch. I made a duplicated >work about this recently. and I found this patch has not been updated >for several days, but since the QEMU have extra supported "memory-file" >and some flags/properties, this patches should be refactored. > Do you plan to send a new version ? If not, Can I take over them? > I'm completely re-factoring the numatune parsing code and reworking few other things for this patch. For memory-file, that will be automatically supported as well, but with Michal's patches. We already have an option that says "use hugepages" and we would like to re-use that instead of creating new device(s). But we will greatly value your input on these patches (both mine and Michal's) when these hit the list. So if there's something else you find missing or wrong, that should be added or fixed, let me know.

Martin