[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1bfecd6b-e05a-c470-ef09-e398de8db521@intel.com>
Date: Mon, 28 Jun 2021 15:53:40 +0800
From: "Zhu, Lingshan" <lingshan.zhu@...el.com>
To: "Wang, Wei W" <wei.w.wang@...el.com>,
Liuxiangdong <liuxiangdong5@...wei.com>,
"peterz@...radead.org" <peterz@...radead.org>,
"pbonzini@...hat.com" <pbonzini@...hat.com>
Cc: "bp@...en8.de" <bp@...en8.de>,
"seanjc@...gle.com" <seanjc@...gle.com>,
"vkuznets@...hat.com" <vkuznets@...hat.com>,
"wanpengli@...cent.com" <wanpengli@...cent.com>,
"jmattson@...gle.com" <jmattson@...gle.com>,
"joro@...tes.org" <joro@...tes.org>,
"Yang, Weijiang" <weijiang.yang@...el.com>,
"kan.liang@...ux.intel.com" <kan.liang@...ux.intel.com>,
"ak@...ux.intel.com" <ak@...ux.intel.com>,
"eranian@...gle.com" <eranian@...gle.com>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"x86@...nel.org" <x86@...nel.org>,
"kvm@...r.kernel.org" <kvm@...r.kernel.org>,
"like.xu.linux@...il.com" <like.xu.linux@...il.com>,
"Fangyi (Eric)" <eric.fangyi@...wei.com>,
Xiexiangyou <xiexiangyou@...wei.com>
Subject: Re: [PATCH V7 00/18] KVM: x86/pmu: Add *basic* support to enable
guest PEBS via DS
On 6/28/2021 3:49 PM, Wang, Wei W wrote:
> On Friday, June 25, 2021 5:46 PM, Zhu, Lingshan wrote:
>>> Only on the host?
>>> I cannot use pebs unless try with "echo 0 > /proc/sys/kernel/watchdog"
>>> both on the host and guest on ICX.
>> Hi Xiangdong
>>
>> I guess you may run into the "cross-map" case(slow path below), so I think you
>> can disable them both in host and guest to make PEBS work.
>>
> Hi Lingshan, could we also reproduce this issue?
>
> If the guest's watchdog takes away the virtual fixed counter, this will schedule the guest PEBS to use virtual PMC0. With the fast path (1:1 mapping), I think physical PMC0 is likely to be available for the guest PEBS emulation if no other host perf events are running.
I think it is possible, even a virtual counter need a perf event
scheduled on the host. This depends on the guest / host workloads.
Thanks,
Zhu Lingshan
> Best,
> Wei
Powered by blists - more mailing lists