Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
From:   Lai Jiangshan <jiangshanlai@gmail.com>
To:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        Sean Christopherson <seanjc@google.com>
Cc:     Vitaly Kuznetsov <vkuznets@redhat.com>,
        Maxim Levitsky <mlevitsk@redhat.com>,
        David Matlack <dmatlack@google.com>,
        Lai Jiangshan <jiangshan.ljs@antgroup.com>
Subject: [PATCH V3 03/12] KVM: X86/MMU: Reduce a check in using_local_root_page() for common cases
Date:   Sat, 21 May 2022 21:16:51 +0800
Message-Id: <20220521131700.3661-4-jiangshanlai@gmail.com>
In-Reply-To: <20220521131700.3661-1-jiangshanlai@gmail.com>
References: <20220521131700.3661-1-jiangshanlai@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

For most cases, mmu->root_role.direct is true and mmu->root_role.level
is not PT32E_ROOT_LEVEL which means using_local_root_page() is often
checking for all the three test which is not good in fast paths.

Morph the conditions in using_local_root_page() to an equivalent one
to reduce a check.

Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
---
 arch/x86/kvm/mmu/mmu.c | 45 ++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 43 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 624b6d2473f7..240ebe589caf 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -1716,11 +1716,52 @@ static void drop_parent_pte(struct kvm_mmu_page *sp,
  *
  * (There is no "mmu->root_role.level > PT32E_ROOT_LEVEL" here, because it is
  *  already ensured that mmu->root_role.level >= PT32E_ROOT_LEVEL)
+ *
+ * But mmu->root_role.direct is normally true and mmu->root_role.level is
+ * normally not PT32E_ROOT_LEVEL.  To reduce a check for the fast path of
+ * fast_pgd_switch() in mormal case, mmu->root_role.direct is checked first.
+ *
+ * The return value is:
+ * 	mmu->root_role.level == PT32E_ROOT_LEVEL ||
+ * 	(!mmu->root_role.direct && mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL)
+ * =>
+ * 	(mmu->root_role.direct && mmu->root_role.level == PT32E_ROOT_LEVEL) ||
+ * 	(!mmu->root_role.direct && mmu->root_role.level == PT32E_ROOT_LEVEL) ||
+ * 	(!mmu->root_role.direct && mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL)
+ * =>
+ * 	(mmu->root_role.direct && mmu->root_role.level == PT32E_ROOT_LEVEL) ||
+ * 	(!mmu->root_role.direct &&
+ * 	 (mmu->root_role.level == PT32E_ROOT_LEVEL ||
+ * 	  mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL))
+ * => (for !direct, mmu->root_role.level == PT32E_ROOT_LEVEL implies
+ * 	mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL)
+ * =>
+ * 	(mmu->root_role.direct && mmu->root_role.level == PT32E_ROOT_LEVEL) ||
+ * 	(!mmu->root_role.direct && mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL)
+ *
+ * In other words:
+ *
+ * For the first and third cases, it is
+ * 	mmu->root_role.direct && mmu->root_role.level == PT32E_ROOT_LEVEL
+ * And if this condition is true, it must be one of the two cases.
+ *
+ * For the 2nd, 4th and 5th cases, it is
+ * 	!mmu->root_role.direct && mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL
+ * And if this condition is true, it must be one of the three cases although
+ * it is not so intuitive.  It can be split into:
+ * 	mmu->root_role.level == PT32E_ROOT_LEVEL &&
+ * 	(!mmu->root_role.direct && mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL)
+ * which is for the 2nd and 4th cases and
+ * 	mmu->root_role.level > PT32E_ROOT_LEVEL &&
+ * 	!mmu->root_role.direct && mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL
+ * which is the last case.
  */
 static bool using_local_root_page(struct kvm_mmu *mmu)
 {
-	return mmu->root_role.level == PT32E_ROOT_LEVEL ||
-	       (!mmu->root_role.direct && mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL);
+	if (mmu->root_role.direct)
+		return mmu->root_role.level == PT32E_ROOT_LEVEL;
+	else
+		return mmu->cpu_role.base.level <= PT32E_ROOT_LEVEL;
 }
 
 static struct kvm_mmu_page *kvm_mmu_alloc_page(struct kvm_vcpu *vcpu, int direct)
-- 
2.19.1.6.gb485710b