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;
Subject: Re: [PATCH v4 2/5] KVM: X86: Bail out of direct yield in case of
 under-committed scenarios
To:     Wanpeng Li <kernellwp@gmail.com>,
        Sean Christopherson <seanjc@google.com>
Cc:     LKML <linux-kernel@vger.kernel.org>, kvm <kvm@vger.kernel.org>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>
References: <1621339235-11131-1-git-send-email-wanpengli@tencent.com>
 <1621339235-11131-2-git-send-email-wanpengli@tencent.com>
 <YKQTx381CGPp7uZY@google.com>
 <CANRm+Cy_D3cBBEYQ9ApKMNC6p0dpTBQYQXs+dv5vrFedVkOy2w@mail.gmail.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <889a0a43-0641-70ce-d2a5-ed71bd54e59c@redhat.com>
Date:   Mon, 24 May 2021 15:42:30 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.8.1
MIME-Version: 1.0
In-Reply-To: <CANRm+Cy_D3cBBEYQ9ApKMNC6p0dpTBQYQXs+dv5vrFedVkOy2w@mail.gmail.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 19/05/21 04:57, Wanpeng Li wrote:
> Looks good. Hope Paolo can update the patch description when applying.:)
> 
> "In case of under-committed scenarios, vCPU can get scheduling easily,
> kvm_vcpu_yield_to add extra overhead, we can observe a lot of races
> between vcpu->ready is true and yield fails due to p->state is
> TASK_RUNNING. Let's bail out in such scenarios by checking the length
> of current cpu runqueue, it can be treated as a hint of under-committed
> instead of guaranteeing accuracy. 30%+ of directed-yield attempts can
> avoid the expensive lookups in kvm_sched_yield() in an under-committed
> scenario. "
> 

Here is what I used:

     In case of under-committed scenarios, vCPUs can be scheduled easily;
     kvm_vcpu_yield_to adds extra overhead, and it is also common to see
     when vcpu->ready is true but yield later failing due to p->state is
     TASK_RUNNING.
     
     Let's bail out in such scenarios by checking the length of current cpu
     runqueue, which can be treated as a hint of under-committed instead of
     guarantee of accuracy. 30%+ of directed-yield attempts can now avoid
     the expensive lookups in kvm_sched_yield() in an under-committed scenario.

Paolo