Re: [PATCH v2 1/3] x86/vpmu: Add get/put_vpmu() and VPMU_AVAILABLE

From: Boris Ostrovsky <boris.ostrovsky@oracle.com>
To: "Tian, Kevin" <kevin.tian@intel.com>,
	"xen-devel@lists.xen.org" <xen-devel@lists.xen.org>
Cc: "andrew.cooper3@citrix.com" <andrew.cooper3@citrix.com>,
	"Nakajima, Jun" <jun.nakajima@intel.com>,
	"jbeulich@suse.com" <jbeulich@suse.com>
Subject: Re: [PATCH v2 1/3] x86/vpmu: Add get/put_vpmu() and VPMU_AVAILABLE
Date: Wed, 22 Feb 2017 09:06:28 -0500	[thread overview]
Message-ID: <9f77089f-f656-2af0-e684-b737665d6748@oracle.com> (raw)
In-Reply-To: <AADFC41AFE54684AB9EE6CBC0274A5D190C42EEC@SHSMSX101.ccr.corp.intel.com>

>>>> +
>>>> +void vpmu_initialise(struct vcpu *v)
>>>> +{
>>>> +    get_vpmu(v);
>>>> +
>>>> +    /*
>>>> +     * Guests without LAPIC (i.e. PV) call vpmu_arch_initialise()
>>>> +     * from pvpmu_init().
>>>> +     */
>>> implication is that PV VPMU is not counted then?
>> No. get_vpmu() is what does the counting now. Since we do
>> vcpu_initialise() -> vpmu_initialise() for all type of VCPUs both PV and
>> HVM VPMUs are counted here. But we defer arch-specific intialization
>> (which doesn't do the counting) for PV until the hypercall.
>>
> earlier comment said vpmu_count is to count active VPMUs.
> what's the definition of 'active' here?  An uninitialized pv VPMU
> is also considered active?

Yes. Whenever a VCPU with appropriately defined CPUID leaf  0xa is
initialized we should assume that it eventually will be used. I can
clarify the comment.

(In the second patch I claim that "appropriately defined" is version=0
but perhaps there is a better definition.)

-boris

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel