Re: [PATCH v2 1/2] KVM: nVMX: Fix VMX preemption timer migration

[Paolo Bonzini <pbonzini@redhat.com>]

On 20/05/20 20:53, Makarand Sonare wrote:
>>
>>> +
>>> +		if (get_user(vmx->nested.preemption_timer_deadline,
>>> +			     &user_vmx_nested_state->preemption_timer_deadline)) {
>> ... tt also seems that we expect user_vmx_nested_state to always have
>> all fields, e.g. here the offset of 'preemption_timer_deadline' is
>> static, we always expect it to be after shadow vmcs. I think we need a
>> way to calculate the offset dynamically and not require everything to be
>> present.
>>
> Would it suffice if I move preemption_timer_deadline field to
> kvm_vmx_nested_state_hdr?
> 

Yes, please do so.  The header is exactly for cases like this where we
have small fields that hold non-architectural pieces of state.

Also, I think you should have a boolean field, like
vmx->nested.has_preemption_timer_deadline.
nested_vmx_enter_non_root_mode would use it (negated) instead of
from_vmentry.  You can then set the field to true in
vmx_set_nested_state (if the incoming state has
KVM_STATE_NESTED_PREEMPTION_TIMER set) and in
nested_vmx_enter_non_root_mode; conversely, vmexit will set it to false
and vmx_get_nested_state can also use the field to decide whether to set
KVM_STATE_NESTED_PREEMPTION_TIMER.

This way, if you have an incoming migration where the field is not set,
nested_vmx_enter_non_root_mode will fall back as gracefully as possible.

Thanks,

Paolo