Re: [PATCH v12 15/19] tsc: Upgrade TSC clocksource rating

["Nikunj A. Dadhania" <nikunj@amd.com>]

On 10/9/2024 9:46 PM, Sean Christopherson wrote:
> On Wed, Oct 09, 2024, Nikunj A Dadhania wrote:
>> In virtualized environments running on modern CPUs, the underlying
>> platforms guarantees to have a stable, always running TSC, i.e. that the
>> TSC is a superior timesource as compared to other clock sources (such as
>> kvmclock, HPET, ACPI timer, APIC, etc.).
>>
>> Upgrade the rating of the early and regular clock source to prefer TSC over
>> other clock sources when TSC is invariant, non-stop and stable.
>>
>> Suggested-by: Thomas Gleixner <tglx@linutronix.de>
>> Signed-off-by: Nikunj A Dadhania <nikunj@amd.com>
>> ---
>>  arch/x86/kernel/tsc.c | 17 +++++++++++++++++
>>  1 file changed, 17 insertions(+)
>>
>> diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
>> index c83f1091bb4f..8150f2104474 100644
>> --- a/arch/x86/kernel/tsc.c
>> +++ b/arch/x86/kernel/tsc.c
>> @@ -1264,6 +1264,21 @@ static void __init check_system_tsc_reliable(void)
>>  		tsc_disable_clocksource_watchdog();
>>  }
>>  
>> +static void __init upgrade_clock_rating(struct clocksource *tsc_early,
>> +					struct clocksource *tsc)
>> +{
>> +	/*
>> +	 * Upgrade the clock rating for TSC early and regular clocksource when
>> +	 * the underlying platform provides non-stop, invaraint and stable TSC.
>> +	 */
>> +	if (boot_cpu_has(X86_FEATURE_CONSTANT_TSC) &&
>> +	    boot_cpu_has(X86_FEATURE_NONSTOP_TSC) &&
>> +	    !tsc_unstable) {
> 
> Somewhat of a side topic, should KVM (as a hypervisor) be enumerating something
> to guests to inform them that the TSC is reliable, i.e. that X86_FEATURE_TSC_RELIABLE
> can be forced?  

Xen does something similar by advertising TSC related information as part of 
a CPUID leaf (Leaf 4 (0x40000x03))

> Or, should KVM (as the guest) infer X86_FEATURE_TSC_RELIABLE if
> INVARIANT_TSC is advertised by KVM (the hyperivosor)?

I am not sure about this though.
 
> Also, why on earth is 0x8000_0007.EDX manually scattered via x86_power?

Are you referring to CPU capabilty settings in early_init_amd() dependent
on x86_power?

> 
>> +		tsc_early->rating = 499;
>> +		tsc->rating = 500;
>> +	}
>> +}

Regards
Nikunj