Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Sender: "seanjc via sendgmr" <seanjc@seanjc798194.pdx.corp.google.com>
Reply-To: Sean Christopherson <seanjc@google.com>
Date:   Thu, 25 Feb 2021 12:47:48 -0800
In-Reply-To: <20210225204749.1512652-1-seanjc@google.com>
Message-Id: <20210225204749.1512652-24-seanjc@google.com>
Mime-Version: 1.0
References: <20210225204749.1512652-1-seanjc@google.com>
Subject: [PATCH 23/24] KVM: x86/mmu: Use low available bits for removed SPTEs
From:   Sean Christopherson <seanjc@google.com>
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     Sean Christopherson <seanjc@google.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, Ben Gardon <bgardon@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Use low "available" bits to tag REMOVED SPTEs.  Using a high bit is
moderately costly as it often causes the compiler to generate a 64-bit
immediate.  More importantly, this makes it very clear REMOVED_SPTE is
a value, not a flag.

Cc: Ben Gardon <bgardon@google.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/spte.c | 11 ++++++++++-
 arch/x86/kvm/mmu/spte.h | 11 +++++++----
 2 files changed, 17 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c
index e07aabb23b8a..66d43cec0c31 100644
--- a/arch/x86/kvm/mmu/spte.c
+++ b/arch/x86/kvm/mmu/spte.c
@@ -277,7 +277,16 @@ void kvm_mmu_set_mmio_spte_mask(u64 mmio_value, u64 mmio_mask, u64 access_mask)
 				  SHADOW_NONPRESENT_OR_RSVD_MASK_LEN)))
 		mmio_value = 0;
 
-	WARN_ON((mmio_value & mmio_mask) != mmio_value);
+	/*
+	 * The masked MMIO value must obviously match itself and a removed SPTE
+	 * must not get a false positive.  Removed SPTEs and MMIO SPTEs should
+	 * never collide as MMIO must set some RWX bits, and removed SPTEs must
+	 * not set any RWX bits.
+	 */
+	if (WARN_ON((mmio_value & mmio_mask) != mmio_value) ||
+	    WARN_ON(mmio_value && (REMOVED_SPTE & mmio_mask) == mmio_value))
+		mmio_value = 0;
+
 	shadow_mmio_value = mmio_value;
 	shadow_mmio_mask  = mmio_mask;
 	shadow_mmio_access_mask = access_mask;
diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
index 2fad4ccd3679..b53036d9ddf3 100644
--- a/arch/x86/kvm/mmu/spte.h
+++ b/arch/x86/kvm/mmu/spte.h
@@ -174,13 +174,16 @@ extern u64 __read_mostly shadow_nonpresent_or_rsvd_mask;
  * non-present intermediate value. Other threads which encounter this value
  * should not modify the SPTE.
  *
- * This constant works because it is considered non-present on both AMD and
- * Intel CPUs and does not create a L1TF vulnerability because the pfn section
- * is zeroed out.
+ * Use a semi-arbitrary value that doesn't set RWX bits, i.e. is not-present on
+ * bot AMD and Intel CPUs, and doesn't set PFN bits, i.e. doesn't create a L1TF
+ * vulnerability.  Use only low bits to avoid 64-bit immediates.
  *
  * Only used by the TDP MMU.
  */
-#define REMOVED_SPTE (1ull << 59)
+#define REMOVED_SPTE	0x5a0ULL
+
+/* Removed SPTEs must not be misconstrued as shadow present PTEs. */
+static_assert(!(REMOVED_SPTE & SPTE_MMU_PRESENT_MASK));
 
 static inline bool is_removed_spte(u64 spte)
 {
-- 
2.30.1.766.gb4fecdf3b7-goog