[Pkg-virtualbox-devel] Bug#836366: trying to run nested does bad things with the host

[Gianfranco Costamagna]

Hi Marc,


>I am trying to do nested virtualization and wanted to use virtualbox

>as the "inner layer" of the nest. This is needed because I want to
>test software using test-kitchen, which does not have a KVM driver.
>Because test-kitchen is not in Debian, I decided to move the entire
>setup in a scratch VM that I can easily snapshot and scrap.
>
>There are a few bug reports in the upstream tracker that virtualbox
>does not support nested virtualization and probably never will. As the
>bug reports are not clear whether they are talking about virtualbox as
>inner or outer layer of the virtualization, I decided to try it
>anyway. Actually, I do not know whether the _inner_ layer of a nested
>virtualization does need to know that it is running virtualized in the
>first place.
>
>That being said, virtualbox used as the inner layer of nested
>virtualization hangs at "Starting virtual machine (2/2) 20%". Clicks on
>the window close button are ignored, but it is possible to kill off
>the VirtualBox process.
>
>Virtualbox should properly detect situations it will not run in and
>report this to not confuse people. It should also document more
>clearly that "does not support nested virtualization" even applies to
>the inner layer of a nest.


I'm not really sure about your test case, but it should be working with the iner

virtualization
(I successfully did some time ago)
Ubuntu trusty amd64 running vbox and a machine Ubuntu trusty amd64, running a Debian i386 VM

anyway, quoting upstream:
- First, "does not support nested virtualization and probably never will" is wrong. Second, if he is using VirtualBox as inner layer and it does not properly run then I would ask help from the people of the outer layer first. Perhaps VirtualBox runs fine on bare metal.
- If I understood him correctly then he uses some VMM as outer layer (KVM?) and tries to run VirtualBox inside.

I would open an upstream ticked and link it here,

thanks,

G.