Re: [PATCH] KVM: x86: hyper-v: Avoid writing to TSC page without an active vCPU

[Vitaly Kuznetsov <vkuznets@redhat.com>]

Sean Christopherson <seanjc@google.com> writes:

> On Thu, Apr 07, 2022, Vitaly Kuznetsov wrote:
>> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
>> index 676705ad1e23..3460bcd75bf2 100644
>> --- a/arch/x86/include/asm/kvm_host.h
>> +++ b/arch/x86/include/asm/kvm_host.h
>> @@ -979,10 +979,10 @@ enum hv_tsc_page_status {
>>  	HV_TSC_PAGE_GUEST_CHANGED,
>>  	/* TSC page MSR was written by KVM userspace, update pending */
>>  	HV_TSC_PAGE_HOST_CHANGED,
>> +	/* TSC page needs to be updated due to internal KVM changes */
>> +	HV_TSC_PAGE_KVM_CHANGED,
>
> Why add KVM_CHANGED?  I don't see any reason to differentiate between userspace
> and KVM, and using KVM_CHANGED for the kvm_vm_ioctl_set_clock() case is wrong as
> that is very much a userspace initiated update, not a KVM update.

Indeed, there seems to be no benefit in differentiating between
HV_TSC_PAGE_HOST_CHANGED and HV_TSC_PAGE_KVM_CHANGED. Let me retest
without it, I'll be sending v2 shortly.

>
>>  	/* TSC page was properly set up and is currently active  */
>>  	HV_TSC_PAGE_SET,
>> -	/* TSC page is currently being updated and therefore is inactive */
>> -	HV_TSC_PAGE_UPDATING,
>>  	/* TSC page was set up with an inaccessible GPA */
>>  	HV_TSC_PAGE_BROKEN,
>>  };
>

-- 
Vitaly