Re: [PATCH] x86/bugs/multihit: Fix mitigation reporting when KVM is not in use

[Pawan Gupta <pawan.kumar.gupta@linux.intel.com>]

On Tue, Jul 14, 2020 at 12:54:26PM -0700, Dave Hansen wrote:
> On 7/14/20 12:17 PM, Pawan Gupta wrote:
> > On Tue, Jul 14, 2020 at 07:57:53AM -0700, Dave Hansen wrote:
> >> Let's stick to things which are at least static per reboot.  Checking
> >> for X86_FEATURE_VMX or even CONFIG_KVM_INTEL seems like a good stopping
> >> point.  "Could this kernel run a naughty guest?"  If so, report
> >> "Vulnerable".  It's the same as Meltdown: "Could this kernel run
> >> untrusted code?"  If so, report "Vulnerable".
> > 
> > Thanks, These are good inputs. So what I need to add is a boot time
> > check for VMX feature and report "Vulnerable" or "Not
> > affected(VMX disabled)".
> > 
> > Are you suggesting to not change the reporting when KVM deploys the
> > "Split huge pages" mitigation? Is this because VMX can still be used by
> > other VMMs?
> > 
> > The current mitigation reporting is very specific to KVM:
> > 
> > 	- "KVM: Vulnerable"
> > 	- "KVM: Mitigation: Split huge pages"
> > 
> > As the kernel doesn't know about the mitigation state of out-of-tree
> > VMMs can we add VMX reporting to always say vulnerable when VMX is
> > enabled:
> > 
> > 	- "VMX: Vulnerable, KVM: Vulnerable"
> > 	- "VMX: Vulnerable, KVM: Mitigation: Split huge pages"
> > 
> > And if VMX is disabled report:
> > 
> > 	- "VMX: Not affected(VMX disabled)"
> 
> I see three inputs and four possible states (sorry for the ugly table,
> it was this or a spreadsheet :):
> 
> X86_FEATURE_VMX	CONFIG_KVM_*	hpage split  Result	   Reason
> 	N			x	    x	     Not Affected  No VMX
> 	Y			N	    x	     Not affected  No KVM
> 	Y			Y	    Y	     Mitigated	   hpage split
> 	Y			Y	    N	     Vulnerable

Thank you.

Just a note... for the last 2 cases kernel wont know about "hpage split"
mitigation until KVM is loaded. So for these cases reporting at boot
will be "Vulnerable" and would change to "Mitigated" once KVM is loaded
and deploys the mitigation. This is the current behavior.

Thanks,
Pawan