On Thu, Feb 07, 2019 at 02:05:33PM -0500, Krish Sadhukhan wrote: > According to section "Checks on VMX Controls" in Intel SDM vol 3C, the > following check is performed on vmentry of L2 guests: > > On processors that support Intel 64 architecture, the IA32_SYSENTER_ESP > field and the IA32_SYSENTER_EIP field must each contain a canonical > address. > > Signed-off-by: Krish Sadhukhan <krish.sadhukhan@xxxxxxxxxx> > Reviewed-by: Mihai Carabas <mihai.carabas@xxxxxxxxxx> > Suggested-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx> Reviewed-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx> > --- > x86/unittests.cfg | 6 ++++++ > x86/vmx_tests.c | 42 ++++++++++++++++++++++++++++++++++++++++++ > 2 files changed, 48 insertions(+) > > diff --git a/x86/unittests.cfg b/x86/unittests.cfg > index d7975e9..580dd8c 100644 > --- a/x86/unittests.cfg > +++ b/x86/unittests.cfg > @@ -542,6 +542,12 @@ extra_params = -cpu host,+vmx -m 2560 -append vmx_controls_test > arch = x86_64 > groups = vmx > > +[vmx_host_state_area] > +file = vmx.flat > +extra_params = -cpu host,+vmx -m 2560 -append vmx_host_state_area_test This test doesn't actually need 2gb of memory, i.e. "-m 2560" isn't required. Only the ept_access_* tests allocate a 1gb page. I wouldn't hold up this series for that though, there are a bunch of other VMX tests that specify "-m 2560" when they don't need to, and in most cases the only real impact is that the tests won't run under 32-bit KVM, which is irrevelant here since this is x86_64 only. Cleaning up the issues across all VMX tests would also be a good opportunity to reduce the amount of copy+paste boilerplate in unittests.cfg. > +arch = x86_64 > +groups = vmx > + > [vmx_vmentry_movss_shadow_test] > file = vmx.flat > extra_params = -cpu host,+vmx -m 2560 -append vmentry_movss_shadow_test
