From: Paolo Bonzini <pbonzini@redhat.com>
To: Wanpeng Li <kernellwp@gmail.com>
Cc: "Wanpeng Li" <wanpeng.li@hotmail.com>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"kvm@vger.kernel.org" <kvm@vger.kernel.org>,
"Radim Krčmář" <rkrcmar@redhat.com>
Subject: Re: [PATCH v3 1/2] KVM: LAPIC: Tune lapic_timer_advance_ns smoothly
Date: Mon, 16 Sep 2019 09:49:47 +0200 [thread overview]
Message-ID: <82ff90b6-f518-e2a8-c4f5-ef4b294af15e@redhat.com> (raw)
In-Reply-To: <CANRm+Cy+5pompcDDS2C9YnxvE_-87i24gbBfc53Qa1tcWNck2Q@mail.gmail.com>
On 16/09/19 06:02, Wanpeng Li wrote:
> Hi Paolo,
>
> Something like below? It still fluctuate when running complex guest os
> like linux. Removing timer_advance_adjust_done will hinder introduce
> patch v3 2/2 since there is no adjust done flag in each round
> evaluation.
That's not important, since the adjustment would be continuous.
How much fluctuation can you see?
> I have two questions here, best-effort tune always as
> below or periodically revaluate to get conservative value and get
> best-effort value in each round evaluation as patch v3 2/2, which one
> do you prefer? The former one can wast time to wait sometimes and the
> later one can not get the best latency. In addition, can the adaptive
> tune algorithm be using in all the scenarios contain
> RT/over-subscribe?
I prefer the former, like the patch below, mostly because of the extra
complexity of the periodic reevaluation.
Paolo
>
> ---8<---
> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> index 685d17c..895735b 100644
> --- a/arch/x86/kvm/lapic.c
> +++ b/arch/x86/kvm/lapic.c
> @@ -69,6 +69,7 @@
> #define LAPIC_TIMER_ADVANCE_ADJUST_INIT 1000
> /* step-by-step approximation to mitigate fluctuation */
> #define LAPIC_TIMER_ADVANCE_ADJUST_STEP 8
> +#define LAPIC_TIMER_ADVANCE_FILTER 5000
>
> static inline int apic_test_vector(int vec, void *bitmap)
> {
> @@ -1479,29 +1480,28 @@ static inline void
> adjust_lapic_timer_advance(struct kvm_vcpu *vcpu,
> s64 advance_expire_delta)
> {
> struct kvm_lapic *apic = vcpu->arch.apic;
> - u32 timer_advance_ns = apic->lapic_timer.timer_advance_ns;
> - u64 ns;
> + u32 timer_advance_ns = apic->lapic_timer.timer_advance_ns, ns;
> +
> + if (abs(advance_expire_delta) > LAPIC_TIMER_ADVANCE_FILTER ||
> + abs(advance_expire_delta) < LAPIC_TIMER_ADVANCE_ADJUST_DONE) {
> + /* filter out random fluctuations */
> + return;
> + }
>
> /* too early */
> if (advance_expire_delta < 0) {
> ns = -advance_expire_delta * 1000000ULL;
> do_div(ns, vcpu->arch.virtual_tsc_khz);
> - timer_advance_ns -= min((u32)ns,
> - timer_advance_ns / LAPIC_TIMER_ADVANCE_ADJUST_STEP);
> + timer_advance_ns -= ns/LAPIC_TIMER_ADVANCE_ADJUST_STEP;
> } else {
> /* too late */
> ns = advance_expire_delta * 1000000ULL;
> do_div(ns, vcpu->arch.virtual_tsc_khz);
> - timer_advance_ns += min((u32)ns,
> - timer_advance_ns / LAPIC_TIMER_ADVANCE_ADJUST_STEP);
> + timer_advance_ns += ns/LAPIC_TIMER_ADVANCE_ADJUST_STEP;
> }
>
> - if (abs(advance_expire_delta) < LAPIC_TIMER_ADVANCE_ADJUST_DONE)
> - apic->lapic_timer.timer_advance_adjust_done = true;
> - if (unlikely(timer_advance_ns > 5000)) {
> + if (unlikely(timer_advance_ns > 5000))
> timer_advance_ns = LAPIC_TIMER_ADVANCE_ADJUST_INIT;
> - apic->lapic_timer.timer_advance_adjust_done = false;
> - }
> apic->lapic_timer.timer_advance_ns = timer_advance_ns;
> }
>
> @@ -1521,7 +1521,7 @@ static void __kvm_wait_lapic_expire(struct kvm_vcpu *vcpu)
> if (guest_tsc < tsc_deadline)
> __wait_lapic_expire(vcpu, tsc_deadline - guest_tsc);
>
> - if (unlikely(!apic->lapic_timer.timer_advance_adjust_done))
> + if (lapic_timer_advance_ns == -1)
> adjust_lapic_timer_advance(vcpu,
> apic->lapic_timer.advance_expire_delta);
> }
>
> @@ -2298,10 +2298,8 @@ int kvm_create_lapic(struct kvm_vcpu *vcpu, int
> timer_advance_ns)
> apic->lapic_timer.timer.function = apic_timer_fn;
> if (timer_advance_ns == -1) {
> apic->lapic_timer.timer_advance_ns = LAPIC_TIMER_ADVANCE_ADJUST_INIT;
> - apic->lapic_timer.timer_advance_adjust_done = false;
> } else {
> apic->lapic_timer.timer_advance_ns = timer_advance_ns;
> - apic->lapic_timer.timer_advance_adjust_done = true;
> }
>
>
> diff --git a/arch/x86/kvm/lapic.h b/arch/x86/kvm/lapic.h
> index 50053d2..2aad7e2 100644
> --- a/arch/x86/kvm/lapic.h
> +++ b/arch/x86/kvm/lapic.h
> @@ -35,7 +35,6 @@ struct kvm_timer {
> s64 advance_expire_delta;
> atomic_t pending; /* accumulated triggered timers */
> bool hv_timer_in_use;
> - bool timer_advance_adjust_done;
> };
>
> struct kvm_lapic {
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 93b0bd4..4f65ef1 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -141,7 +141,7 @@
> * advancement entirely. Any other value is used as-is and disables adaptive
> * tuning, i.e. allows priveleged userspace to set an exact advancement time.
> */
> -static int __read_mostly lapic_timer_advance_ns = -1;
> +int __read_mostly lapic_timer_advance_ns = -1;
> module_param(lapic_timer_advance_ns, int, S_IRUGO | S_IWUSR);
>
> static bool __read_mostly vector_hashing = true;
> diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h
> index 6594020..2c6ba86 100644
> --- a/arch/x86/kvm/x86.h
> +++ b/arch/x86/kvm/x86.h
> @@ -301,6 +301,7 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu,
> unsigned long cr2,
>
> extern bool enable_vmware_backdoor;
>
> +extern int lapic_timer_advance_ns;
> extern int pi_inject_timer;
>
> extern struct static_key kvm_no_apic_vcpu;
>
next prev parent reply other threads:[~2019-09-16 7:49 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-28 8:19 [PATCH v3 1/2] KVM: LAPIC: Tune lapic_timer_advance_ns smoothly Wanpeng Li
2019-08-28 8:19 ` [PATCH v3 2/2] KVM: LAPIC: Periodically revaluate to get conservative lapic_timer_advance_ns Wanpeng Li
2019-09-11 15:45 ` [PATCH v3 1/2] KVM: LAPIC: Tune lapic_timer_advance_ns smoothly Paolo Bonzini
[not found] ` <TY2PR02MB4160421A8C88D96C8BCB971180B00@TY2PR02MB4160.apcprd02.prod.outlook.com>
2019-09-12 12:37 ` Paolo Bonzini
2019-09-16 4:02 ` Wanpeng Li
2019-09-16 7:49 ` Paolo Bonzini [this message]
2019-09-16 8:09 ` Wanpeng Li
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=82ff90b6-f518-e2a8-c4f5-ef4b294af15e@redhat.com \
--to=pbonzini@redhat.com \
--cc=kernellwp@gmail.com \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=rkrcmar@redhat.com \
--cc=wanpeng.li@hotmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).