On Mon, Apr 25, 2022, Sean Christopherson wrote: > On Sun, Apr 24, 2022, SU Hang wrote: > > Why? As gross as it is, I actually think INTn is a better option because it > doesn't require writing multiple MSRs, and can work for both 64-bit and 32-bit KUT. > The latter is currently a moot point since this code is 64-bit only, but the UMIP > test _does_ support 32-bit, and it's do_ring3() should really be rolled into this > framework. > > Furthermore, we really should have a test to verify that KVM correctly emulates > SYSCALL at CPL3 with EFER.SCE=0, and forcing EFER.SCE=1 just to get to CPL3 would > make it impossible to utilize this framework for such a test. And a concrete reason not to apply this patch: it causes the nVMX #AC test to fail: FAIL: #AC handled by L2 FAIL: #AC handled by L1
