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, 1 Sep 2023 09:48:13 -0700
In-Reply-To: <20230902175431.2925-1-zeming@nfschina.com>
Mime-Version: 1.0
References: <20230902175431.2925-1-zeming@nfschina.com>
Message-ID: <ZPIVzccIAiQnG4IA@google.com>
Subject: Re: [PATCH] x86/kvm/mmu: =?utf-8?Q?Remove_?= =?utf-8?B?dW5uZWNlc3Nhcnkg4oCYTlVMTOKAmQ==?=
 values from sptep
From:   Sean Christopherson <seanjc@google.com>
To:     Li zeming <zeming@nfschina.com>
Cc:     pbonzini@redhat.com, tglx@linutronix.de, mingo@redhat.com,
        bp@alien8.de, dave.hansen@linux.intel.com, x86@kernel.org,
        kvm@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="us-ascii"
Precedence: bulk

On Sun, Sep 03, 2023, Li zeming wrote:
> sptep is assigned first, so it does not need to initialize the assignment.
> 
> Signed-off-by: Li zeming <zeming@nfschina.com>
> ---
>  arch/x86/kvm/mmu/mmu.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index ec169f5c7dce..95f745aec4aa 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -3457,7 +3457,7 @@ static int fast_page_fault(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault)
>  	struct kvm_mmu_page *sp;
>  	int ret = RET_PF_INVALID;
>  	u64 spte = 0ull;
> -	u64 *sptep = NULL;
> +	u64 *sptep;
>  	uint retry_count = 0;
>  
>  	if (!page_fault_can_be_fast(fault))

Hmm, this is safe, but there's some ugliness lurking.  Theoretically, it's possible
for spte to be left untouched by the walkers.  That _shouldn't_ happen, as it means
there's a bug somewhere in KVM.  But if that did happen, on the second or later
iteration, it's (again, theoretically) possible to consume a stale spte.

		if (tdp_mmu_enabled)
			sptep = kvm_tdp_mmu_fast_pf_get_last_sptep(vcpu, fault->addr, &spte);
		else
			sptep = fast_pf_get_last_sptep(vcpu, fault->addr, &spte);

		if (!is_shadow_present_pte(spte)) <=== could consume stale data
			break;

If we're going to tidy up sptep, I think we should also give spte similar treatment
and harden KVM in the process, e.g.

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 6325bb3e8c2b..ae2f87bbbf0a 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -3430,8 +3430,8 @@ static int fast_page_fault(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault)
 {
        struct kvm_mmu_page *sp;
        int ret = RET_PF_INVALID;
-       u64 spte = 0ull;
-       u64 *sptep = NULL;
+       u64 spte;
+       u64 *sptep;
        uint retry_count = 0;
 
        if (!page_fault_can_be_fast(fault))
@@ -3447,6 +3447,14 @@ static int fast_page_fault(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault)
                else
                        sptep = fast_pf_get_last_sptep(vcpu, fault->addr, &spte);
 
+               /*
+                * It's entirely possible for the mapping to have been zapped
+                * by a different task, but the root page is should always be
+                * available as the vCPU holds a reference to its root(s).
+                */
+               if (WARN_ON_ONCE(!sptep))
+                       spte = REMOVED_SPTE;
+
                if (!is_shadow_present_pte(spte))
                        break;