Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 30 Jun 2023 08:37:06 -0700
In-Reply-To: <20230630050157.kurgzdcpjngs6w4k@yy-desk-7060>
Mime-Version: 1.0
References: <20230628071217.71126-1-ishiir@g.ecc.u-tokyo.ac.jp>
 <ZJxTTZzZnfbyMVIH@google.com> <ZJ0w5pKk/41Zv26i@yzhao56-desk.sh.intel.com>
 <ZJ3FyLUYrlr6+HLw@google.com> <ZJ3p5wBwyQZ+aQOV@google.com> <20230630050157.kurgzdcpjngs6w4k@yy-desk-7060>
Message-ID: <ZJ72omPGZcdjjWbo@google.com>
Subject: Re: [PATCH] KVM: nVMX: Prevent vmlaunch with EPTP pointing outside
 assigned memory area
From:   Sean Christopherson <seanjc@google.com>
To:     Yuan Yao <yuan.yao@linux.intel.com>
Cc:     Yan Zhao <yan.y.zhao@intel.com>,
        Reima Ishii <ishiir@g.ecc.u-tokyo.ac.jp>,
        shina@ecc.u-tokyo.ac.jp, Paolo Bonzini <pbonzini@redhat.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,
        "H. Peter Anvin" <hpa@zytor.com>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, yuan.yao@intel.com
Content-Type: text/plain; charset="us-ascii"
Precedence: bulk

On Fri, Jun 30, 2023, Yuan Yao wrote:
> On Thu, Jun 29, 2023 at 01:30:31PM -0700, Sean Christopherson wrote:
> > @@ -3834,8 +3822,8 @@ static int mmu_alloc_shadow_roots(struct kvm_vcpu *vcpu)
> >  			if (!(pdptrs[i] & PT_PRESENT_MASK))
> >  				continue;
> >
> > -			if (mmu_check_root(vcpu, pdptrs[i] >> PAGE_SHIFT))
> > -				return 1;
> > +			if (kvm_vcpu_is_visible_gfn(vcpu, pdptrs[i] >> PAGE_SHIFT))
> > +				pdptrs[i] = 0;
> 
> Hi Sean,
> 
> Should this be "!kvm_vcpu_is_visible_gfn(vcpu, pdptrs[i] >> PAGE_SHIFT)" and

Yep, typo that inverted the check.  Thanks for saving me some debug time!

> turn the pae_root[i] to dummy root yet ?

No, zeroing the PDPTR is sufficient.  Unlike CR3, which is always "present", PDPTRs
have a present bit and so KVM can communicate to hardware that the entry isn't
valid simply by clearing the PDPTPR.