On Thu, Nov 02, 2023 at 05:46:26PM +0530, Nikunj A. Dadhania wrote: > On 11/2/2023 5:37 PM, Nikunj A. Dadhania wrote: > > On 11/2/2023 4:03 PM, Kirill A. Shutemov wrote: > >> On Thu, Nov 02, 2023 at 11:23:34AM +0530, Nikunj A. Dadhania wrote: > >>> On 10/30/2023 10:48 PM, Dave Hansen wrote: > >>>> On 10/29/23 23:36, Nikunj A Dadhania wrote: > >>>> ... > >>>>> diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c > >>>>> index 15f97c0abc9d..b0a8546d3703 100644 > >>>>> --- a/arch/x86/kernel/tsc.c > >>>>> +++ b/arch/x86/kernel/tsc.c > >>>>> @@ -1241,7 +1241,7 @@ static void __init check_system_tsc_reliable(void) > >>>>> tsc_clocksource_reliable = 1; > >>>>> } > >>>>> #endif > >>>>> - if (boot_cpu_has(X86_FEATURE_TSC_RELIABLE)) > >>>>> + if (boot_cpu_has(X86_FEATURE_TSC_RELIABLE) || cc_platform_has(CC_ATTR_GUEST_SECURE_TSC)) > >>>>> tsc_clocksource_reliable = 1; > >>>> > >>>> Why can't you just set X86_FEATURE_TSC_RELIABLE? > >>> > >>> Last time when I tried, I had removed my kvmclock changes and I had set > >>> the X86_FEATURE_TSC_RELIABLE similar to Kirill's patch[1], this did not > >>> select the SecureTSC. > >>> > >>> Let me try setting X86_FEATURE_TSC_RELIABLE and retaining my patch for > >>> skipping kvmclock. > >> > >> kvmclock lowers its rating if TSC is good enough: > >> > >> if (boot_cpu_has(X86_FEATURE_CONSTANT_TSC) && > >> boot_cpu_has(X86_FEATURE_NONSTOP_TSC) && > >> !check_tsc_unstable()) > >> kvm_clock.rating = 299; > >> > >> Does your TSC meet the requirements? > > > > I have set TscInvariant (bit 8) in CPUID_8000_0007_edx and TSC is set as reliable. > > > > With this I see kvm_clock rating being lowered, but kvm-clock is still being picked as clock-source. > > Ah.. at later point TSC is picked up, is this expected ? > > [ 2.564052] clocksource: Switched to clocksource kvm-clock > [ 2.678136] clocksource: Switched to clocksource tsc On bare metal I see switch from tsc-early to tsc. tsc-early rating is equal to kvmclock rating after it gets lowered. Maybe kvmclock rating has to be even lower after detecting sane TSC? Sean, Paolo, any comments? -- Kiryl Shutsemau / Kirill A. Shutemov
