Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Message-ID: <4a8beb6d-1589-7f53-881f-8faaeb52f7ba@gmail.com>
Date:   Tue, 8 Feb 2022 18:10:23 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:91.0)
 Gecko/20100101 Thunderbird/91.5.0
Subject: Re: [PATCH kvm/queue v2 3/3] KVM: x86/pmu: Setup the
 {inte|amd}_event_mapping[] when hardware_setup
Content-Language: en-US
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     Sean Christopherson <seanjc@google.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Joerg Roedel <joro@8bytes.org>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, Jim Mattson <jmattson@google.com>
References: <20220117085307.93030-1-likexu@tencent.com>
 <20220117085307.93030-4-likexu@tencent.com>
 <13de6271-61bc-7138-15b3-9241508d94fa@redhat.com>
From:   Like Xu <like.xu.linux@gmail.com>
Organization: Tencent
In-Reply-To: <13de6271-61bc-7138-15b3-9241508d94fa@redhat.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 1/2/2022 8:28 pm, Paolo Bonzini wrote:
> On 1/17/22 09:53, Like Xu wrote:
>> +
>> +    for (i = 0; i < PERF_COUNT_HW_MAX; i++) {
>> +        config = perf_get_hw_event_config(i) & 0xFFFFULL;
>> +
>> +        kernel_hw_events[i] = (struct kvm_event_hw_type_mapping){
>> +            .eventsel = config & ARCH_PERFMON_EVENTSEL_EVENT,
>> +            .unit_mask = (config & ARCH_PERFMON_EVENTSEL_UMASK) >> 8,
>> +            .event_type = i,
>> +        };
> 
> Should event_type be PERF_COUNT_HW_MAX if config is zero?

Emm, we do not assume that the hardware event encoded with "eventsel=0 && 
unit_mask=0"
(in this case, config is zero) are illegal.

If perf core puts this encoded event into "enum perf_hw_id" table as this part 
is out of the scope of
KVM, we have to setup with a valid event_type value instead of PERF_COUNT_HW_MAX.

In this proposal, the returned perf_hw_id from kvm_x86_ops.pmu_ops->pmc_perf_hw_id()
is only valid and used if "pmc->eventsel & 0xFFFFULL" is non-zero, otherwise the
reprogram_gp_counter() will fall back to use PERF_TYPE_RAW type.

Please let me know if you need more clarification on this change.

> 
> Thanks,
> 
> Paolo
>