Re: [PATCH v7 2/3] KVM: vmx: Emulate MSR IA32_UMWAIT_CONTROL

[Tao Xu <tao3.xu@intel.com>]

On 7/17/2019 9:17 AM, Tao Xu wrote:
> On 7/17/2019 12:03 AM, Eduardo Habkost wrote:
>> On Fri, Jul 12, 2019 at 04:29:06PM +0800, Tao Xu wrote:
>>> UMWAIT and TPAUSE instructions use IA32_UMWAIT_CONTROL at MSR index E1H
>>> to determines the maximum time in TSC-quanta that the processor can 
>>> reside
>>> in either C0.1 or C0.2.
>>>
>>> This patch emulates MSR IA32_UMWAIT_CONTROL in guest and differentiate
>>> IA32_UMWAIT_CONTROL between host and guest. The variable
>>> mwait_control_cached in arch/x86/power/umwait.c caches the MSR value, so
>>> this patch uses it to avoid frequently rdmsr of IA32_UMWAIT_CONTROL.
>>>
>>> Co-developed-by: Jingqi Liu <jingqi.liu@intel.com>
>>> Signed-off-by: Jingqi Liu <jingqi.liu@intel.com>
>>> Signed-off-by: Tao Xu <tao3.xu@intel.com>
>>> ---
>> [...]
>>> +static void atomic_switch_umwait_control_msr(struct vcpu_vmx *vmx)
>>> +{
>>> +    if (!vmx_has_waitpkg(vmx))
>>> +        return;
>>> +
>>> +    if (vmx->msr_ia32_umwait_control != umwait_control_cached)
>>> +        add_atomic_switch_msr(vmx, MSR_IA32_UMWAIT_CONTROL,
>>> +            vmx->msr_ia32_umwait_control,
>>> +            umwait_control_cached, false);
>>
>> How exactly do we ensure NR_AUTOLOAD_MSRS (8) is still large enough?
>>
>> I see 3 existing add_atomic_switch_msr() calls, but the one at
>> atomic_switch_perf_msrs() is in a loop.  Are we absolutely sure
>> that perf_guest_get_msrs() will never return more than 5 MSRs?
>>
> 
> Quote the code of intel_guest_get_msrs:
> 
> static struct perf_guest_switch_msr *intel_guest_get_msrs(int *nr)
> {
> [...]
>      arr[0].msr = MSR_CORE_PERF_GLOBAL_CTRL;
>      arr[0].host = x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_guest_mask;
>      arr[0].guest = x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_host_mask;
>      if (x86_pmu.flags & PMU_FL_PEBS_ALL)
>          arr[0].guest &= ~cpuc->pebs_enabled;
>      else
>          arr[0].guest &= ~(cpuc->pebs_enabled & PEBS_COUNTER_MASK);
>      *nr = 1;
> 
>      if (x86_pmu.pebs && x86_pmu.pebs_no_isolation) {
> [...]
>          arr[1].msr = MSR_IA32_PEBS_ENABLE;
>          arr[1].host = cpuc->pebs_enabled;
>          arr[1].guest = 0;
>          *nr = 2;
> [...]
> 
> There are most 2 msrs now. By default umwait is disabled in KVM. So by 
> default there is no MSR_IA32_UMWAIT_CONTROL added into 
> add_atomic_switch_msr().
> 
> Thanks.

And for old hardware, kvm use core_guest_get_msrs, but umwait is for now 
hardware, and if hardware in host doesn't have the cpuid, there is no 
MSR_IA32_UMWAIT_CONTROL in kvm as well.

>>
>>> +    else
>>> +        clear_atomic_switch_msr(vmx, MSR_IA32_UMWAIT_CONTROL);
>>> +}
>>> +
>>>   static void vmx_arm_hv_timer(struct vcpu_vmx *vmx, u32 val)
>>>   {
>>>       vmcs_write32(VMX_PREEMPTION_TIMER_VALUE, val);
>> [...]
>>
>>
>