On 27/05/21 16:11, Maxim Levitsky wrote:
Using 'nested_run_pending=1' perhaps? Or, we can get back to 'vm_bugged'
idea and kill the guest immediately if something forces such an exit.
Exactly, this is my idea. Set the nested_run_pending=1 always after the migration
It shoudn't cause any issues and it would avoid cases like that.
That variable can then be renamed too to something like 'nested_vmexit_not_allowed'
or something like that.
Paolo, what do you think?
(If it works :)) that's clever. It can even be set unconditionally on
the save side and would even work for new->old migration.
Paolo
