Re: [PATCH v2] KVM: VMX: Enable Notify VM exit

[Sean Christopherson]

On Fri, Sep 10, 2021, Xiaoyao Li wrote:
> On 9/10/2021 2:47 AM, Sean Christopherson wrote:
> > On Tue, Sep 07, 2021, Xiaoyao Li wrote:
> > > On 9/3/2021 12:29 AM, Sean Christopherson wrote:
> > > > > After syncing internally, we know that the internal threshold is not
> > > > > architectural but a model-specific value. It will be published in some place
> > > > > in future.
> > > > 
> > > > Any chance it will also be discoverable, e.g. via an MSR?
> > > 
> > > I also hope we can expose it via MSR. If not, we can maintain a table per
> > > FMS in KVM to get the internal threshold. However, per FMS info is not
> > > friendly to be virtualized (when we are going to enable the nested support).
> > 
> > Yeah, FMS is awful.  If the built-in buffer isn't discoverable, my vote is to
> > assume the worst, i.e. a built-in buffer of '0', and have the notify_window
> > param default to a safe value, e.g. 25k or maybe even 150k (to go above what the
> > hardware folks apparently deemed safe for SPR).  It's obviously not idea, but
> > it's better than playing FMS guessing games.
> > 
> > > I'll try to persuade internal to expose it via MSR, but I guarantee nothing.
> > 
> > ...
> > 
> > > > On a related topic, this needs tests.  One thought would be to stop unconditionally
> > > > intercepting #AC if NOTIFY_WINDOW is enabled, and then have the test set up the
> > > > infinite #AC vectoring scenario.
> > > > 
> > > 
> > > yes, we have already tested with this case with notify_window set to 0. No
> > > false positive.
> > 
> > Can you send a selftest or kvm-unit-test?
> > 
> 
> Actually we implement the attacking case of CVE-2015-5307 with
> kvm-unit-test, while manually disabling the intercept of #AC.
> 
> First, it requires modification of KVM that only posting the kvm-unit-test
> doesn't help.

It helps in that hacking KVM to disable #AC interception is a lot easier than
re-writing a test from scratch.

> Second, release the attacking case is not the correct action.

As in it's irresponsible to provide code that can be used to DoS a hypervisor?
The CVE is six years old, IMO security-through-obscurity is unnecessary at this
point.