On 9/27/2019 11:21 PM, Paolo Bonzini wrote:
On 27/09/19 16:55, Liran Alon wrote:
Why is it confusing? Any exit-reason not specified in
kvm_vmx_exit_handlers[] is an exit-reason KVM doesn’t expect to be
raised from hardware. Whether it’s because VMCS is configured to not
raise that exit-reason or because it’s a new exit-reason only
supported on newer CPUs. (Which is kinda the same. Because a new
exit-reason should be raised only if hypervisor opt-in some VMCS
feature).
I agree that it's a bug compared to how other unhandled vmexits are
treated. I didn't want to rewrite kvm/next or have a revert, so I have
sent a pull request but this should be fixed. I'll wait for Liran's
patch or come up with one.
Paolo
I got it. Thanks.
