On 13 June 2016 at 20:20, Andrea Bolognani <abologna@xxxxxxxxxx> wrote: > So, this has already been merged and it's definitely an > improvement over the existing situation (thanks to both you > and Michal, by the way!), but I was wondering whether > removing the hardware check altogether is a good idea. I wondered too, but opted in keeping the existing checks. They might still save someones time. > There is a difference between KVM not being built properly / > shipped / loaded and the host not supporting hardware > virtualization at all. We capture that difference on x86 > and s390, but not on aarch64 and ppc64. > Is figuring out whether hardware virtualization support is > present just not feasible on aarch64, eg. because each > reports the information in a different way, or because the > information is not exposed to userspace at all? The information isn't exposed to userspace. The closest thing we have is looking at kernel error messages when kvm is initialized. Lack of virtualization support on arm64 is usually a firmware issue. Riku -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list