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 v2] KVM: Check the allocation of pv cpu mask
To:     Vitaly Kuznetsov <vkuznets@redhat.com>
Cc:     "pbonzini@redhat.com" <pbonzini@redhat.com>,
        Sean Christopherson <sean.j.christopherson@intel.com>,
        "wanpengli@tencent.com" <wanpengli@tencent.com>,
        "jmattson@google.com" <jmattson@google.com>,
        "joro@8bytes.org" <joro@8bytes.org>, tglx@linutronix.de,
        mingo@redhat.com, "bp@alien8.de" <bp@alien8.de>,
        "hpa@zytor.com" <hpa@zytor.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "kvm@vger.kernel.org" <kvm@vger.kernel.org>,
        "x86@kernel.org" <x86@kernel.org>
References: <654d8c60-49f0-e398-be25-24aed352360d@gmail.com>
 <87y2lrnnyf.fsf@vitty.brq.redhat.com>
From:   Haiwei Li <lihaiwei.kernel@gmail.com>
Message-ID: <61e2fd6f-effd-64d7-148a-1b1f9fda1449@gmail.com>
Date:   Fri, 4 Sep 2020 09:01:59 +0800
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:68.0)
 Gecko/20100101 Thunderbird/68.12.0
MIME-Version: 1.0
In-Reply-To: <87y2lrnnyf.fsf@vitty.brq.redhat.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 20/9/3 18:39, Vitaly Kuznetsov wrote:
> Haiwei Li <lihaiwei.kernel@gmail.com> writes:
> 
>> From: Haiwei Li <lihaiwei@tencent.com>
>>
>> check the allocation of per-cpu __pv_cpu_mask. Initialize ops only when
>> successful.
>>
>> Signed-off-by: Haiwei Li <lihaiwei@tencent.com>
>> ---
>>    arch/x86/kernel/kvm.c | 24 ++++++++++++++++++++----
>>    1 file changed, 20 insertions(+), 4 deletions(-)
>>
>> diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
>> index 08320b0b2b27..d3c062e551d7 100644
>> --- a/arch/x86/kernel/kvm.c
>> +++ b/arch/x86/kernel/kvm.c
>> @@ -555,7 +555,6 @@ static void kvm_send_ipi_mask_allbutself(const
>> struct cpumask *mask, int vector)
>>    static void kvm_setup_pv_ipi(void)
>>    {
>>    	apic->send_IPI_mask = kvm_send_ipi_mask;
>> -	apic->send_IPI_mask_allbutself = kvm_send_ipi_mask_allbutself;
>>    	pr_info("setup PV IPIs\n");
>>    }
>>
>> @@ -654,7 +653,6 @@ static void __init kvm_guest_init(void)
>>    	}
>>
>>    	if (pv_tlb_flush_supported()) {
>> -		pv_ops.mmu.flush_tlb_others = kvm_flush_tlb_others;
>>    		pv_ops.mmu.tlb_remove_table = tlb_remove_table;
>>    		pr_info("KVM setup pv remote TLB flush\n");
>>    	}
>> @@ -767,6 +765,14 @@ static __init int activate_jump_labels(void)
>>    }
>>    arch_initcall(activate_jump_labels);
>>
>> +static void kvm_free_pv_cpu_mask(void)
>> +{
>> +	unsigned int cpu;
>> +
>> +	for_each_possible_cpu(cpu)
>> +		free_cpumask_var(per_cpu(__pv_cpu_mask, cpu));
>> +}
>> +
>>    static __init int kvm_alloc_cpumask(void)
>>    {
>>    	int cpu;
>> @@ -785,11 +791,21 @@ static __init int kvm_alloc_cpumask(void)
>>
>>    	if (alloc)
>>    		for_each_possible_cpu(cpu) {
>> -			zalloc_cpumask_var_node(per_cpu_ptr(&__pv_cpu_mask, cpu),
>> -				GFP_KERNEL, cpu_to_node(cpu));
>> +			if (!zalloc_cpumask_var_node(
>> +				per_cpu_ptr(&__pv_cpu_mask, cpu),
>> +				GFP_KERNEL, cpu_to_node(cpu)))
>> +				goto zalloc_cpumask_fail;
>>    		}
>>
>> +#if defined(CONFIG_SMP)
>> +	apic->send_IPI_mask_allbutself = kvm_send_ipi_mask_allbutself;
>> +#endif
>> +	pv_ops.mmu.flush_tlb_others = kvm_flush_tlb_others;
> 
> This is too late I'm afraid. If I'm not mistaken PV patching happens
> earlier, so .init.guest_late_init (kvm_guest_init()) is good and
> arch_initcall() is bad.

.init.guest_late_init (kvm_guest_init()) is called before 
arch_initcall() and kvm_flush_tlb_others && kvm_send_ipi_mask_allbutself 
rely on __pv_cpu_mask.  So, i can not put this assign in kvm_guest_init().

> 
> Have you checked that with this patch kvm_flush_tlb_others() is still
> being called?

yes. I add a printk and i get the log.

> 
> Actually, there is no need to assign kvm_flush_tlb_others() so late. We
> can always check if __pv_cpu_mask was allocated and revert back to the
> architectural path if not.
I am sorry i don't really understand. Can you explain in more detail? Thx.

> 
>>    	return 0;
>> +
>> +zalloc_cpumask_fail:
>> +	kvm_free_pv_cpu_mask();
>> +	return -ENOMEM;
>>    }
>>    arch_initcall(kvm_alloc_cpumask);
>>
>> --
>> 2.18.4
>>
>