Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Wanpeng Li <kernellwp@gmail.com>
To:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org
Cc:     Paolo Bonzini <pbonzini@redhat.com>,
        =?UTF-8?q?Radim=20Kr=C4=8Dm=C3=A1=C5=99?= <rkrcmar@redhat.com>
Subject: [PATCH 2/3] KVM: LAPIC: lapic timer is injected by posted interrupt
Date:   Wed,  5 Jun 2019 18:09:10 +0800
Message-Id: <1559729351-20244-3-git-send-email-wanpengli@tencent.com>
In-Reply-To: <1559729351-20244-1-git-send-email-wanpengli@tencent.com>
References: <1559729351-20244-1-git-send-email-wanpengli@tencent.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

From: Wanpeng Li <wanpengli@tencent.com>

Dedicated instances are currently disturbed by unnecessary jitter due 
to the emulated lapic timers fire on the same pCPUs which vCPUs resident.
There is no hardware virtual timer on Intel for guest like ARM. Both 
programming timer in guest and the emulated timer fires incur vmexits.
This patchset tries to avoid vmexit which is incurred by the emulated 
timer fires in dedicated instance scenario. 

When nohz_full is enabled in dedicated instances scenario, the unpinned 
timer will be moved to the nearest busy housekeepers after commit 444969223c8
("sched/nohz: Fix affine unpinned timers mess"). However, KVM always makes 
lapic timer pinned to the pCPU which vCPU residents, the reason is explained 
by commit 61abdbe0 (kvm: x86: make lapic hrtimer pinned). Actually, these 
emulated timers can be offload to the housekeeping cpus since APICv 
is really common in recent years. The guest timer interrupt is injected by 
posted-interrupt which is delivered by housekeeping cpu once the emulated 
timer fires. 

3%~5% redis performance benefit can be observed on Skylake server.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
---
 arch/x86/kvm/lapic.c | 22 +++++++++++++++++++++-
 1 file changed, 21 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index 8c9c14d..e9db086 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -1465,6 +1465,23 @@ static void apic_timer_expired(struct kvm_lapic *apic)
 }
 
 /*
+ * On APICv, lapic timer is injected by posted interrupt
+ * to dedicated instance.
+ */
+static void apic_timer_expired_pi(struct kvm_lapic *apic)
+{
+	struct kvm_timer *ktimer = &apic->lapic_timer;
+
+	kvm_apic_local_deliver(apic, APIC_LVTT);
+	if (apic_lvtt_tscdeadline(apic))
+		ktimer->tscdeadline = 0;
+	if (apic_lvtt_oneshot(apic)) {
+		ktimer->tscdeadline = 0;
+		ktimer->target_expiration = 0;
+	}
+}
+
+/*
  * On APICv, this test will cause a busy wait
  * during a higher-priority task.
  */
@@ -2297,7 +2314,10 @@ static enum hrtimer_restart apic_timer_fn(struct hrtimer *data)
 	struct kvm_timer *ktimer = container_of(data, struct kvm_timer, timer);
 	struct kvm_lapic *apic = container_of(ktimer, struct kvm_lapic, lapic_timer);
 
-	apic_timer_expired(apic);
+	if (unlikely(posted_interrupt_inject_timer(apic->vcpu)))
+		apic_timer_expired_pi(apic);
+	else
+		apic_timer_expired(apic);
 
 	if (lapic_is_periodic(apic)) {
 		advance_periodic_target_expiration(apic);
-- 
2.7.4