Re: [PATCH v3 4/4] platform/x86: intel_tdx_attest: Add TDX Guest attestation interface driver

From: Kai Huang <kai.huang@intel.com>
To: Sathyanarayanan Kuppuswamy 
	<sathyanarayanan.kuppuswamy@linux.intel.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	x86@kernel.org, Hans de Goede <hdegoede@redhat.com>,
	Mark Gross <mgross@linux.intel.com>
Cc: "H . Peter Anvin" <hpa@zytor.com>,
	"Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>,
	Tony Luck <tony.luck@intel.com>, Andi Kleen <ak@linux.intel.com>,
	linux-kernel@vger.kernel.org,
	platform-driver-x86@vger.kernel.org
Subject: Re: [PATCH v3 4/4] platform/x86: intel_tdx_attest: Add TDX Guest attestation interface driver
Date: Wed, 20 Apr 2022 10:38:09 +1200	[thread overview]
Message-ID: <5371d2df8940226674f27a7ce950e9ae1468a951.camel@intel.com> (raw)
In-Reply-To: <07ef65c4-708e-1bcf-9a7e-f804acefcc7c@linux.intel.com>

On Tue, 2022-04-19 at 07:00 -0700, Sathyanarayanan Kuppuswamy wrote:
> 
> On 4/19/22 1:16 AM, Kai Huang wrote:
> > In fact after slightly thinking more, I think you can split TDREPORT TDCALL
> > support with GetQuote/SetupEventNotifyInterrupt support.  The reason is as I
> > said, GetQuote isn't mandatory to support attestation.  TD attestation agent can
> > use i.e. vsock, tcp/ip, to communicate to QE directly.  Whether kernel needs to
> > support GetQuote is actually arguable.
> 
> IMO, we should not use a usage model to categorize "GetQuote" support
> as a mandatory or non-mandatory requirement.
> 
> For customers who use VSOCK, they can get away without GetQuote
> TDVMCALL support. But for customers who do not want to use
> VSOCK model, this is a required support. AFAIK, our current customer
> requirement is to use TDVMCALL approach for attestation support.
> 
> If your suggestion is to split GetQuote support as separate
> patch to make it easier for review, I am fine with such
> suggestion.
> 

I am not saying we should get rid of GetQuote support.  If there's customer
wants this with a good reason, we can certainly support it.  I understand that
some customer wants to deploy QE in host and don't want additional communication
channel (i.e. vsock) between guest and host, which may add additional attack
window and/or customer's validation resource.

My point is regardless whether we need to support GetQuote, logically this
driver can be split to two parts as I said: 1) basic TDREPORT support to
userspace; 2) additional GetQuote support.  And I think there are many benefits
if you do in this way as I commented below.

> > 
> > So IMHO you can split this attestation driver into two parts:
> > 
> > 1) A "basic" driver which supports reporting TDREPORT to userspace
> > 2) Additional support of GetQuote/SetupEventNotifyInterrupt.
> > 
> > The 1) can even be in a single patch (I guess it won't be complicated).  It is
> > easy to review (and i.e. can be merged separately), and with it, you will
> > immediately have one way to support attestation.
> > 
> > 2) can be reviewed separately, perhaps with one additional Kconfig option (i.e.
> > CONFIG_INTEL_TDX_ATTESTATION_GET_QUOTE).  I think this part has most of the
> 
> 
> GetQuote IOCTL support is a very simple feature support, so, IMO, we
> don't need to complicate it with additional config.
> 
> > 

Additional Kconfig can reduce attack window by turning it off for people don't
need it.  Anyway no strong opinion here.

-- 
Thanks,
-Kai