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;
Subject: Re: [PATCH v2 6/7] locking/spinlocks, paravirt, hyperv: Correct the
 hv_nopvspin case
To:     Sasha Levin <sashal@kernel.org>
Cc:     linux-kernel@vger.kernel.org, tglx@linutronix.de, mingo@kernel.org,
        bp@alien8.de, hpa@zytor.com, boris.ostrovsky@oracle.com,
        jgross@suse.com, sstabellini@kernel.org, peterz@infradead.org,
        srinivas.eeda@oracle.com, Waiman Long <longman@redhat.com>,
        "K. Y. Srinivasan" <kys@microsoft.com>,
        Haiyang Zhang <haiyangz@microsoft.com>,
        Stephen Hemminger <sthemmin@microsoft.com>,
        Ingo Molnar <mingo@redhat.com>, linux-hyperv@vger.kernel.org
References: <1561377779-28036-1-git-send-email-zhenzhong.duan@oracle.com>
 <1561377779-28036-7-git-send-email-zhenzhong.duan@oracle.com>
 <20190627222851.GC11506@sasha-vm>
From:   Zhenzhong Duan <zhenzhong.duan@oracle.com>
Organization: Oracle Corporation
Message-ID: <eaeb5e52-5fa8-20c9-bc5c-090572718511@oracle.com>
Date:   Fri, 28 Jun 2019 08:53:00 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:60.0) Gecko/20100101
 Thunderbird/60.7.2
MIME-Version: 1.0
In-Reply-To: <20190627222851.GC11506@sasha-vm>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Content-Language: en-US
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 2019/6/28 6:28, Sasha Levin wrote:
> On Mon, Jun 24, 2019 at 08:02:58PM +0800, Zhenzhong Duan wrote:
>> With the boot parameter "hv_nopvspin" specified a Hyperv guest should
>> not make use of paravirt spinlocks, but behave as if running on bare
>> metal. This is not true, however, as the qspinlock code will fall back
>> to a test-and-set scheme when it is detecting a hypervisor.
>>
>> In order to avoid this disable the virt_spin_lock_key.
>>
>> Same change for XEN is already in Commit e6fd28eb3522
>> ("locking/spinlocks, paravirt, xen: Correct the xen_nopvspin case")
>>
>> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@oracle.com>
>> Cc: Waiman Long <longman@redhat.com>
>> Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
>> Cc: "K. Y. Srinivasan" <kys@microsoft.com>
>> Cc: Haiyang Zhang <haiyangz@microsoft.com>
>> Cc: Stephen Hemminger <sthemmin@microsoft.com>
>> Cc: Sasha Levin <sashal@kernel.org>
>> Cc: Thomas Gleixner <tglx@linutronix.de>
>> Cc: Ingo Molnar <mingo@redhat.com>
>> Cc: Borislav Petkov <bp@alien8.de>
>> Cc: linux-hyperv@vger.kernel.org
>> ---
>> arch/x86/hyperv/hv_spinlock.c | 3 +++
>> 1 file changed, 3 insertions(+)
>>
>> diff --git a/arch/x86/hyperv/hv_spinlock.c 
>> b/arch/x86/hyperv/hv_spinlock.c
>> index 07f21a0..d90b4b0 100644
>> --- a/arch/x86/hyperv/hv_spinlock.c
>> +++ b/arch/x86/hyperv/hv_spinlock.c
>> @@ -64,6 +64,9 @@ __visible bool hv_vcpu_is_preempted(int vcpu)
>>
>> void __init hv_init_spinlocks(void)
>> {
>> +    if (unlikely(!hv_pvspin))
>> +        static_branch_disable(&virt_spin_lock_key);
>
> This should be combined in the conditional under it, which already
> attempts to disable PV spinlocks, note how hv_pvspin is checked there.
> hc_pvspin isn't the only reason we would disable PV spinlocks on hyperv.

In virt_spin_lock() there is a comment as below. The test-and-set spinlock

is an optimization to hypervisor platform when PV spinlock is unsupported.

         /*
          * On hypervisors without PARAVIRT_SPINLOCKS support we fall
          * back to a Test-and-Set spinlock, because fair locks have
          * horrible lock 'holder' preemption issues.
          */


So my understanding is:

If hv_pvspin=0 by command line, we want to behave as if running on bare 
metal(the fair locks path).

Though there is performance regression, but it's not that important when 
we use hv_pvspin=0.

If PV spinlock is disabled by other reasons, we prefer the optimization 
path.

>
> Also, there's no need for the unlikely() here, it's only getting called
> once...

Ok, I'll removed it.


Thanks

Zhenzhong