Comment 12 for bug 1767857

Hmm,
so it is not even burning CPU for anything neither in the Host (usr ~=qemu, sys ~=kernel) nor in the guest itself (~=guest).
The KVM exits confirm that, it doesn't do a a lot entry/exit is the pass in/out of guest context and the only meaningful exit means it emulates an instruction.

So if it is not hogged on a CPU it is waiting for something. Unfortunately this is too windows and x86-arch specific for me to read, even if you get the info. BTW getting the info what instructions it is spinning on can be done via [1] - this might be helpful for others coming by to help. To bad that I still can't reproduce that locally :-/, but at least I haven't heard of anyone else hitting this yet so it must be sort of special to maybe your guest?

Especially I can't see yet how a host kernel update would trigger this :-/
Sorry, but this is up for the kernel Team to consider the changes that happened between those versions in regard to this issue.

[1]: https://lkml.org/lkml/2012/6/27/241