Re: [PATCH v2 07/25] x86/sgx: Initialize virtual EPC driver even when SGX driver is disabled

From: Kai Huang <kai.huang@intel.com>
To: Jarkko Sakkinen <jarkko@kernel.org>
Cc: Sean Christopherson <seanjc@google.com>,
	kvm@vger.kernel.org, linux-sgx@vger.kernel.org, x86@kernel.org,
	linux-kernel@vger.kernel.org, luto@kernel.org,
	dave.hansen@intel.com, rick.p.edgecombe@intel.com,
	haitao.huang@intel.com, pbonzini@redhat.com, bp@alien8.de,
	tglx@linutronix.de, mingo@redhat.com, hpa@zytor.com
Subject: Re: [PATCH v2 07/25] x86/sgx: Initialize virtual EPC driver even when SGX driver is disabled
Date: Thu, 18 Mar 2021 13:04:35 +1300	[thread overview]
Message-ID: <20210318130435.4cf6142368bb1837c585b447@intel.com> (raw)
In-Reply-To: <YFCojQmyM8fdGmnl@kernel.org>

On Tue, 16 Mar 2021 14:46:05 +0200 Jarkko Sakkinen wrote:
> On Mon, Mar 15, 2021 at 06:13:53PM -0700, Sean Christopherson wrote:
> > On Tue, Mar 16, 2021, Jarkko Sakkinen wrote:
> > > On Tue, Mar 16, 2021 at 01:05:05AM +0200, Jarkko Sakkinen wrote:
> > > > The way I've understood it is that given that KVM can support SGX
> > > > without FLC, vEPC should be available even if driver cannot be
> > > > enabled.
> > > > 
> > > > This is also exactly what the short summary states.
> > > > 
> > > > "Initialize virtual EPC driver even when SGX driver is disabled"
> > > > 
> > > > It *does not* state:
> > > > 
> > > > "Initialize SGX driver even when vEPC driver is disabled"
> > > > 
> > > > Also, this is how I interpret the inline comment.
> > > > 
> > > > All this considered, the other direction is undocumented functionality.
> > > 
> > > Also:
> > > 
> > > 1. There is *zero* good practical reasons to support the "2nd direction".
> > 
> > Uh, yes there is.  CONFIG_KVM_INTEL=n and X86_FEATURE_VMX=n, either of which
> > will cause vEPC initialization to fail.  The former is obvious, the latter is
> > possible via BIOS configuration.
> 
> Hmm... So you make the checks as if ret != -ENODEV? That's the sane way to
> deal with that situation IMHO.
> 
> /Jarkko

OK. I actually wrote the code to show your idea:

-       ret = sgx_drv_init();
-       if (ret)
+       /*
+        * Only continue to initialize SGX driver when SGX virtualization
+        * initialization is successful, or is not supported (-ENODEV), since
+        * any other initialization failure means SGX driver is unlikely to be
+        * initialized successfully.
+        */
+       ret = sgx_vepc_init();
+       if (ret && ret != -ENODEV)
                goto err_kthread;

+       if (sgx_drv_init()) {
+               /*
+                * Cleanup when *both* SGX virtualization and SGX driver are
+                * not enabled, due to either not supported (-ENODEV), or
+                * somehow fail to initialize.
+                */
+               if (ret && ret != -ENODEV)
+                       goto err_kthread;
+       }
+
        return 0;

 err_kthread:

Is this the code you want?