[libvirt-users] SEV machines and memory pinning

Daniel P. Berrangé berrange at redhat.com
Thu Apr 4 09:58:17 UTC 2019


On Wed, Apr 03, 2019 at 07:59:13PM +0200, Boris Bobrov wrote:
> Hello,
> 
> I am working on implementing SEV support in OpenStack. There are some
> questions that came up in the discussion of the spec [0]
> 
> [0] https://review.openstack.org/#/c/641994/
> 
> As far as i understand, the memory for SEV machines need to be pinned so
> that it doesn't migrate to swap and page migration. ROMS, UEFI pflash and
> video RAM should be pinned too.
> 
> Initially we planned to use hard_limit of <memtune> element to pin the
> memory. However, from the discussion in the spec it seems that there is no
> way to determine a good enough value and that hard_limit should not be used
> at all.
> 
> What should be used then?
> 
> There is a suggestion to use something like this:
> 
>   <memoryBacking>
>     <hugepages>
>       <page size="2" unit="M" nodeset="1"/>
>     </hugepages>
>     <nosharepages/>
>     <locked/>
>     <source type="file"/>
>     <access mode="shared"/>
>     <allocation mode="immediate"/>
>   </memoryBacking>
> 
> Will it work? Are there any caveats we should remember about with this
> config? If we can use it, is there anything that would be redundant or not
> necessary for our case?

AFAIK from the SEV pov only the <locked/> element is required.

This extra things where just suggested on the openstack review
becasue it was suggested to make it dependant on use of huge
pages in Nova.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|




More information about the libvirt-users mailing list