Kernel KVM virtualization development
 help / color / mirror / Atom feed
* [kvm-unit-tests PATCH] x86/pmu: Verify all available GP counters in check_counters_many()
@ 2025-06-11  7:58 Dapeng Mi
  2025-06-12  7:35 ` Sandipan Das
  2025-06-25 22:25 ` Sean Christopherson
  0 siblings, 2 replies; 3+ messages in thread
From: Dapeng Mi @ 2025-06-11  7:58 UTC (permalink / raw)
  To: Sean Christopherson, Paolo Bonzini
  Cc: kvm, linux-kernel, Jim Mattson, Mingwei Zhang, Zide Chen,
	Das Sandipan, Shukla Manali, Dapeng Mi, Dapeng Mi

The intent of check_counters_many() is to verify all available counters
can count correctly at the same time. So an alternative event should be
picked to verify the avaialbe GP counter instead of skiping the counter
if the initial event is not available.

Signed-off-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
---
Please notice this patch is based on Sean's "x86: Add CPUID properties,
 clean up related code" v2 patchset (https://lore.kernel.org/all/20250610195415.115404-1-seanjc@google.com/).

 x86/pmu.c | 22 +++++++++++++++++++---
 1 file changed, 19 insertions(+), 3 deletions(-)

diff --git a/x86/pmu.c b/x86/pmu.c
index 3987311c..a6b0cfcc 100644
--- a/x86/pmu.c
+++ b/x86/pmu.c
@@ -457,18 +457,34 @@ static void check_fixed_counters(void)
 	}
 }
 
+static struct pmu_event *get_one_event(int idx)
+{
+	int i;
+
+	if (pmu_arch_event_is_available(idx))
+		return &gp_events[idx % gp_events_size];
+
+	for (i = 0; i < gp_events_size; i++) {
+		if (pmu_arch_event_is_available(i))
+			return &gp_events[i];
+	}
+
+	return NULL;
+}
+
 static void check_counters_many(void)
 {
+	struct pmu_event *evt;
 	pmu_counter_t cnt[48];
 	int i, n;
 
 	for (i = 0, n = 0; n < pmu.nr_gp_counters; i++) {
-		if (!pmu_arch_event_is_available(i))
+		evt = get_one_event(i);
+		if (!evt)
 			continue;
 
 		cnt[n].ctr = MSR_GP_COUNTERx(n);
-		cnt[n].config = EVNTSEL_OS | EVNTSEL_USR |
-			gp_events[i % gp_events_size].unit_sel;
+		cnt[n].config = EVNTSEL_OS | EVNTSEL_USR | evt->unit_sel;
 		n++;
 	}
 	for (i = 0; i < fixed_counters_num; i++) {
-- 
2.43.0


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [kvm-unit-tests PATCH] x86/pmu: Verify all available GP counters in check_counters_many()
  2025-06-11  7:58 [kvm-unit-tests PATCH] x86/pmu: Verify all available GP counters in check_counters_many() Dapeng Mi
@ 2025-06-12  7:35 ` Sandipan Das
  2025-06-25 22:25 ` Sean Christopherson
  1 sibling, 0 replies; 3+ messages in thread
From: Sandipan Das @ 2025-06-12  7:35 UTC (permalink / raw)
  To: Dapeng Mi, Sean Christopherson, Paolo Bonzini
  Cc: kvm, linux-kernel, Jim Mattson, Mingwei Zhang, Zide Chen,
	Shukla Manali, Dapeng Mi

On 6/11/2025 1:28 PM, Dapeng Mi wrote:
> The intent of check_counters_many() is to verify all available counters
> can count correctly at the same time. So an alternative event should be
> picked to verify the avaialbe GP counter instead of skiping the counter
> if the initial event is not available.
> 
> Signed-off-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
> ---
> Please notice this patch is based on Sean's "x86: Add CPUID properties,
>  clean up related code" v2 patchset (https://lore.kernel.org/all/20250610195415.115404-1-seanjc@google.com/).
> 
>  x86/pmu.c | 22 +++++++++++++++++++---
>  1 file changed, 19 insertions(+), 3 deletions(-)
> 

For AMD processors
Tested-by: Sandipan Das <sandipan.das@amd.com>

> diff --git a/x86/pmu.c b/x86/pmu.c
> index 3987311c..a6b0cfcc 100644
> --- a/x86/pmu.c
> +++ b/x86/pmu.c
> @@ -457,18 +457,34 @@ static void check_fixed_counters(void)
>  	}
>  }
>  
> +static struct pmu_event *get_one_event(int idx)
> +{
> +	int i;
> +
> +	if (pmu_arch_event_is_available(idx))
> +		return &gp_events[idx % gp_events_size];
> +
> +	for (i = 0; i < gp_events_size; i++) {
> +		if (pmu_arch_event_is_available(i))
> +			return &gp_events[i];
> +	}
> +
> +	return NULL;
> +}
> +
>  static void check_counters_many(void)
>  {
> +	struct pmu_event *evt;
>  	pmu_counter_t cnt[48];
>  	int i, n;
>  
>  	for (i = 0, n = 0; n < pmu.nr_gp_counters; i++) {
> -		if (!pmu_arch_event_is_available(i))
> +		evt = get_one_event(i);
> +		if (!evt)
>  			continue;
>  
>  		cnt[n].ctr = MSR_GP_COUNTERx(n);
> -		cnt[n].config = EVNTSEL_OS | EVNTSEL_USR |
> -			gp_events[i % gp_events_size].unit_sel;
> +		cnt[n].config = EVNTSEL_OS | EVNTSEL_USR | evt->unit_sel;
>  		n++;
>  	}
>  	for (i = 0; i < fixed_counters_num; i++) {


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [kvm-unit-tests PATCH] x86/pmu: Verify all available GP counters in check_counters_many()
  2025-06-11  7:58 [kvm-unit-tests PATCH] x86/pmu: Verify all available GP counters in check_counters_many() Dapeng Mi
  2025-06-12  7:35 ` Sandipan Das
@ 2025-06-25 22:25 ` Sean Christopherson
  1 sibling, 0 replies; 3+ messages in thread
From: Sean Christopherson @ 2025-06-25 22:25 UTC (permalink / raw)
  To: Sean Christopherson, Paolo Bonzini, Dapeng Mi
  Cc: kvm, linux-kernel, Jim Mattson, Mingwei Zhang, Zide Chen,
	Das Sandipan, Shukla Manali, Dapeng Mi

On Wed, 11 Jun 2025 07:58:42 +0000, Dapeng Mi wrote:
> The intent of check_counters_many() is to verify all available counters
> can count correctly at the same time. So an alternative event should be
> picked to verify the avaialbe GP counter instead of skiping the counter
> if the initial event is not available.

Applied to kvm-x86 next, thanks!

[1/1] x86/pmu: Verify all available GP counters in check_counters_many()
      https://github.com/kvm-x86/kvm-unit-tests/commit/525bdb5d65d5

--
https://github.com/kvm-x86/kvm-unit-tests/tree/next

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2025-06-25 22:27 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-06-11  7:58 [kvm-unit-tests PATCH] x86/pmu: Verify all available GP counters in check_counters_many() Dapeng Mi
2025-06-12  7:35 ` Sandipan Das
2025-06-25 22:25 ` Sean Christopherson

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox