From mboxrd@z Thu Jan  1 00:00:00 1970
From: Radim Krcmar <rkrcmar@redhat.com>
Subject: Re: [patch 2/3] KVM: x86: add option to advance tscdeadline hrtimer
 expiration
Date: Wed, 17 Dec 2014 15:58:13 +0100
Message-ID: <20141217145805.GA29368@potion.brq.redhat.com>
References: <20141216140813.493421022@redhat.com>
 <20141216140853.687723255@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: kvm@vger.kernel.org, Luiz Capitulino <lcapitulino@redhat.com>,
	Rik van Riel <riel@redhat.com>,
	Paolo Bonzini <pbonzini@redhat.com>
To: Marcelo Tosatti <mtosatti@redhat.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:47818 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1750905AbaLQO6R (ORCPT <rfc822;kvm@vger.kernel.org>);
	Wed, 17 Dec 2014 09:58:17 -0500
Received: from int-mx10.intmail.prod.int.phx2.redhat.com (int-mx10.intmail.prod.int.phx2.redhat.com [10.5.11.23])
	by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id sBHEwHlT007791
	(version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=FAIL)
	for <kvm@vger.kernel.org>; Wed, 17 Dec 2014 09:58:17 -0500
Content-Disposition: inline
In-Reply-To: <20141216140853.687723255@redhat.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

2014-12-16 09:08-0500, Marcelo Tosatti:
> For the hrtimer which emulates the tscdeadline timer in the guest,
> add an option to advance expiration, and busy spin on VM-entry waitin=
g
> for the actual expiration time to elapse.
>=20
> This allows achieving low latencies in cyclictest (or any scenario=20
> which requires strict timing regarding timer expiration).
>=20
> Reduces average cyclictest latency from 12us to 8us
> on Core i5 desktop.
>=20
> Note: this option requires tuning to find the appropriate value=20
> for a particular hardware/guest combination. One method is to measure=
 the=20
> average delay between apic_timer_fn and VM-entry.=20
> Another method is to start with 1000ns, and increase the value
> in say 500ns increments until avg cyclictest numbers stop decreasing.
>=20
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Reviewed-by: Radim Kr=C4=8Dm=C3=A1=C5=99 <rkrcmar@redhat.com>

> +++ kvm/arch/x86/kvm/lapic.c
> @@ -1087,11 +1089,64 @@ static void apic_timer_expired(struct kv
[...]
> +/*
> + * On APICv, this test will cause a busy wait
> + * during a higher-priority task.
> + */

(A bit confusing ... this test doesn't busy wait.)

> +
> +static bool lapic_timer_int_injected(struct kvm_vcpu *vcpu)
[...]
> +void wait_lapic_expire(struct kvm_vcpu *vcpu)
> +{
[...]
> +	tsc_deadline =3D apic->lapic_timer.expired_tscdeadline;
> +	apic->lapic_timer.expired_tscdeadline =3D 0;
> +	guest_tsc =3D kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> +
> +	while (guest_tsc < tsc_deadline) {
> +		int delay =3D min(tsc_deadline - guest_tsc, 1000ULL);

Why break the __delay() loop into smaller parts?

> +
> +		__delay(delay);

(Does not have to call delay_tsc, but I guess it won't change.)

> +		guest_tsc =3D kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> +	}
>  }
> =20

Btw. simple automatic delta tuning had worse results?