Re: [Xen-devel] [PATCH v12 for-xen-4.5 18/20] x86/VPMU: Add privileged PMU mode

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 25.09.14 at 21:28, <boris.ostrovsky@xxxxxxxxxx> wrote:
> --- a/xen/arch/x86/traps.c
> +++ b/xen/arch/x86/traps.c
> @@ -2579,6 +2579,10 @@ static int emulate_privileged_op(struct cpu_user_regs 
> *regs)
>          case MSR_AMD_FAM15H_EVNTSEL0...MSR_AMD_FAM15H_PERFCTR5:
>                  if ( vpmu_msr || (boot_cpu_data.x86_vendor == 
> X86_VENDOR_AMD) )
>                  {
> +                    if ( (vpmu_mode & XENPMU_MODE_ALL) &&
> +                         !is_hardware_domain(v->domain) )
> +                        break;
> +
>                      if ( vpmu_do_wrmsr(regs->ecx, msr_content, 0) )
>                          goto fail;
>                  }
> @@ -2701,6 +2705,14 @@ static int emulate_privileged_op(struct cpu_user_regs 
> *regs)
>          case MSR_AMD_FAM15H_EVNTSEL0...MSR_AMD_FAM15H_PERFCTR5:
>                  if ( vpmu_msr || (boot_cpu_data.x86_vendor == 
> X86_VENDOR_AMD) )
>                  {
> +                    if ( (vpmu_mode & XENPMU_MODE_ALL) &&
> +                         !is_hardware_domain(v->domain) )
> +                    {
> +                        /* Don't leak PMU MSRs to unprivileged domains */
> +                        regs->eax = regs->edx = 0;
> +                        break;
> +                    }
> +
>                      if ( vpmu_do_rdmsr(regs->ecx, &msr_content) )
>                          goto fail;
>  

Is ignoring writes and returning zeroes for reads really reasonable in
this case? I.e. is the guest validly being told that there is a (v)PMU?
Because if it's not, it has no business accessing these MSRs and
hence should probably get a #GP instead.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel