[PATCH] x86/acrn: Set X86_FEATURE_TSC_KNOWN_FREQ

[PATCH] x86/acrn: Set X86_FEATURE_TSC_KNOWN_FREQ

[Fei Li]

Avoiding recalibration by setting X86_FEATURE_TSC_KNOWN_FREQ.
This patch also removes `inline` for acrn_get_tsc_khz() since
it doesn't make sense.

Signed-off-by: Fei Li <fei1.li@intel.com>
Reviewed-by: Yin, Fengwei <fengwei.yin@intel.com>
---
 arch/x86/include/asm/acrn.h | 5 -----
 arch/x86/kernel/cpu/acrn.c  | 6 ++++++
 2 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/arch/x86/include/asm/acrn.h b/arch/x86/include/asm/acrn.h
index 1dd14381bcb6..aa12c74ea959 100644
--- a/arch/x86/include/asm/acrn.h
+++ b/arch/x86/include/asm/acrn.h
@@ -30,11 +30,6 @@ static inline u32 acrn_cpuid_base(void)
 	return 0;
 }
 
-static inline unsigned long acrn_get_tsc_khz(void)
-{
-	return cpuid_eax(ACRN_CPUID_TIMING_INFO);
-}
-
 /*
  * Hypercalls for ACRN
  *
diff --git a/arch/x86/kernel/cpu/acrn.c b/arch/x86/kernel/cpu/acrn.c
index 485441b7f030..c5ff75b6a949 100644
--- a/arch/x86/kernel/cpu/acrn.c
+++ b/arch/x86/kernel/cpu/acrn.c
@@ -24,6 +24,12 @@ static u32 __init acrn_detect(void)
 	return acrn_cpuid_base();
 }
 
+static unsigned long acrn_get_tsc_khz(void)
+{
+	setup_force_cpu_cap(X86_FEATURE_TSC_KNOWN_FREQ);
+	return cpuid_eax(ACRN_CPUID_TIMING_INFO);
+}
+
 static void __init acrn_init_platform(void)
 {
 	/* Setup the IDT for ACRN hypervisor callback */

base-commit: 1a2dcbdde82e3a5f1db9b2f4c48aa1aeba534fb2
-- 
2.34.1

Re: [PATCH] x86/acrn: Set X86_FEATURE_TSC_KNOWN_FREQ

[Peter Zijlstra]

On Tue, Oct 25, 2022 at 04:41:47PM +0800, Fei Li wrote:
> Avoiding recalibration by setting X86_FEATURE_TSC_KNOWN_FREQ.
> This patch also removes `inline` for acrn_get_tsc_khz() since
> it doesn't make sense.

Can we have a few words on how this is supposed to work at all?

Supposedly the HV is providing host rate through this ACRN CPUID leaf?
How does it deal with systems where TSC isn't known stable?

Re: [PATCH] x86/acrn: Set X86_FEATURE_TSC_KNOWN_FREQ

[Fei Li]

On 2022-10-25 at 15:22:58 +0200, Peter Zijlstra wrote:
> On Tue, Oct 25, 2022 at 04:41:47PM +0800, Fei Li wrote:
> > Avoiding recalibration by setting X86_FEATURE_TSC_KNOWN_FREQ.
> > This patch also removes `inline` for acrn_get_tsc_khz() since
> > it doesn't make sense.
> 
Hi Peter

Thanks for your comments.

> Can we have a few words on how this is supposed to work at all?

If the TSC frequency is known from the acrn_get_tsc_khz(),
the TSC frequency does not need to be recalibrated.

We can avoid recalibration by setting X86_FEATURE_TSC_KNOWN_FREQ.
> 
> Supposedly the HV is providing host rate through this ACRN CPUID leaf?
ACRN would provide the calibrated TSC frequency, not just the
host rate through this CPUID leaf.

> How does it deal with systems where TSC isn't known stable?
>