On Wed, Aug 10, 2022, Manali Shukla wrote:
> static void svm_nm_test(void)
IMO, this should be renamed to svm_no_nm_test().
> {
> handle_exception(NM_VECTOR, guest_test_nm_handler);
> write_cr0(read_cr0() & ~X86_CR0_TS);
> test_set_guest(svm_nm_test_guest);
>
> - vmcb->save.cr0 = vmcb->save.cr0 | X86_CR0_TS;
> - report(svm_vmrun() == SVM_EXIT_VMMCALL && nm_test_counter == 1,
> - "fnop with CR0.TS set in L2, #NM is triggered");
> -
> - vmcb->save.cr0 = (vmcb->save.cr0 & ~X86_CR0_TS) | X86_CR0_EM;
> - report(svm_vmrun() == SVM_EXIT_VMMCALL && nm_test_counter == 2,
> - "fnop with CR0.EM set in L2, #NM is triggered");
> -
> vmcb->save.cr0 = vmcb->save.cr0 & ~(X86_CR0_TS | X86_CR0_EM);
> - report(svm_vmrun() == SVM_EXIT_VMMCALL && nm_test_counter == 2,
> + report(svm_vmrun() == SVM_EXIT_VMMCALL && nm_test_counter == 0,
> "fnop with CR0.TS and CR0.EM unset no #NM excpetion");
No need to keep the #NM handler, just rely on the VMMCAL assertion and the fact
that an unexpected #NM will also cause a test failure.
