> > +static void atomic_switch_msr_limit_test_guest(void)
> > +{
> > + vmcall();
>
> I finally dug into the weird double-enter_guest(). Rather than re-enter
> the guest to cleanup, just remove this vmcall() so that the first VM-Enter
> invokes hypercall() with HYPERCALL_VMEXIT to set guest_finished.
>
> enter_guest() will verify VM-Enter succeeded, and the guest_finished check
> verifies the guest did a VMCALL. I don't see any added value in an extra
> VMCALL.
Done. To be consistent with Krish's review, I utilized the existing
v2_null_test_guest() function, which is empty.
> > + if (count <= max_allowed) {
> > + enter_guest();
> > + skip_exit_vmcall();
>
> If vmcall() is removed, this skip and the one in the else{} can be dropped.
Done.
> > + /*
> > + * Re-enter the guest with valid counts
> > + * and proceed past the single vmcall instruction.
> > + */
>
> Nit: "Re-enter the guest" should either be "Retry VM-Enter" or simply
> "Enter". The reason this code exists is that we never actually
> entered the guest :-)
>
> E.g. if you drop the vmcall():
>
> /* Enter the guest (with valid counts) to set guest_finished. */
Done.
