Re: [PATCH] Ability to read the MKTME status from userspace

[Dave Hansen]

On 6/19/20 7:36 AM, Richard Hughes wrote:
> On Fri, 19 Jun 2020 at 15:23, Dave Hansen <dave.hansen@xxxxxxxxx> wrote:
>> Last night, I asked my kids if they brushed their teeth.  They said:
>> "Dad, my toothbrush was available."  They argued that mere availability
>> was a better situation than not *having* a toothbrush.  They were
>> logically right, of course, but they still got cavities.
> 
> I don't see how that's comparable, sorry. Surely Intel wants to sell
> hardware advertising TME as a security feature?

Of course!  Just like AVX-512 or VNNI or whatever, Intel will totally
tell you about the stuff baked into its silicon!  But, just like
AVX-512, there's a lot of work to do on top of mere presence in the
silicon to ensure it is providing benefit.

>>> So my take-away from that is that it's currently impossible to
>>> actually say if your system is *actually* using TME.
>> Not in a generic way, and it can't be derived from cpuid or MSRs alone.
> 
> Well, it seems not in any way at the moment.
> 
>> I'm pretty sure I'm using TME, but I didn't become sure from
>> poking at sysfs.
> 
> How do you know that Lenovo didn't disable TME without looking at
> dmesg? I don't think "pretty sure" is good enough when TME is
> considered a security feature.

You cut out the important part.  The "pretty sure" involves a bunch of
preconditions and knowing what your hardware configuration is in the
first place.

Let's take a step back.  We add read-only ABIs so that decisions can be
made.  What decision will somebody make from the ABI being proposed here?

Someone does 'cat /proc/mktme' (or whatever) and it says "1" or
whatever, which means yay, encryption is on.  What do they do?

What do they do differently when it says "0"?