Re: [PATCH 1/2] nVMX: Ensure host's PAT is loaded at the end of all VMX tests

[Yan Zhao]

On Mon, Jun 10, 2024 at 06:47:55PM -0700, Sean Christopherson wrote:
> On Tue, Jun 11, 2024, Yan Zhao wrote:
> > On Wed, Jun 05, 2024 at 03:45:26PM -0700, Sean Christopherson wrote:
> > > @@ -7274,6 +7275,8 @@ static void test_pat(u32 field, const char * field_name, u32 ctrl_field,
> > >  					error = 0;
> > >  
> > >  				test_vmx_vmlaunch(error);
> > > +				wrmsr(MSR_IA32_CR_PAT, pat_msr_saved);
> > > +
> > >  				report_prefix_pop();
> > >  
> > >  			} else {	// GUEST_PAT
> > 
> > Is it possible that ENT_LOAD_PAT of GUEST_PAT is tested when there's no support of
> > EXI_LOAD_PAT of HOST_PAT?
> > Then
> > 	wrmsr(MSR_IA32_CR_PAT, pat_msr_saved);
> > is also required in this case?
> 
> Heh, in theory, yeah, a nested setup could create such a monstrosity.  It's easy
> enough to handle, so I guess why not?  Though I'm tempted to assert instead,
> because practically speaking this code will never be hit, and thus never validated.
> 
> diff --git a/x86/vmx_tests.c b/x86/vmx_tests.c
> index 2063ee90..ffe7064c 100644
> --- a/x86/vmx_tests.c
> +++ b/x86/vmx_tests.c
> @@ -7288,6 +7288,9 @@ static void test_pat(u32 field, const char * field_name, u32 ctrl_field,
>                                 error = (i == 0x2 || i == 0x3 || i >= 0x8);
>                                 test_guest_state("ENT_LOAD_PAT enabled", !!error,
>                                                  val, "GUEST_PAT");
> +
> +                               if (!(ctrl_exit_rev.clr & EXI_LOAD_PAT))
> +                                       wrmsr(MSR_IA32_CR_PAT, pat_msr_saved);
>                         }
>  
>                 }
Tested passed with EXI_LOAD_PAT intentionally cleared in ctrl_exit_rev.clr.

Reviewed-by: Yan Zhao <yan.y.zhao@xxxxxxxxx>