[PATCH] perf/x86/lbr: Reset LBR_SELECT during vlbr reset

[PATCH] perf/x86/lbr: Reset LBR_SELECT during vlbr reset

[Wanpeng Li]

From: Wanpeng Li <wanpengli@tencent.com>

lbr_select in kvm guest has residual data even if kvm guest is poweroff.
We can get residual data in the next boot. Because lbr_select is not
reset during kvm vlbr release. Let's reset LBR_SELECT during vlbr reset.

Cc: Like Xu <likexu@tencent.com>
Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
---
 arch/x86/events/intel/lbr.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c
index 9e6d6eaeb4cb..1076de93a2f5 100644
--- a/arch/x86/events/intel/lbr.c
+++ b/arch/x86/events/intel/lbr.c
@@ -279,6 +279,8 @@ void intel_pmu_lbr_reset(void)
 
 	cpuc->last_task_ctx = NULL;
 	cpuc->last_log_id = 0;
+	if (!static_cpu_has(X86_FEATURE_ARCH_LBR) && cpuc->lbr_select)
+		wrmsrl(MSR_LBR_SELECT, 0);
 }
 
 /*
-- 
2.25.1

Re: [PATCH] perf/x86/lbr: Reset LBR_SELECT during vlbr reset

[Like Xu]

On 5/11/2021 3:20 pm, Wanpeng Li wrote:
> From: Wanpeng Li <wanpengli@tencent.com>
> 
> lbr_select in kvm guest has residual data even if kvm guest is poweroff.
> We can get residual data in the next boot. Because lbr_select is not
> reset during kvm vlbr release. Let's reset LBR_SELECT during vlbr reset.

This is because when a new vlbr event is first scheduled in the current CPU,
the lbr_select is not reset, which causes the previous lbr_select value
to be visible to the next vcpu.

Considering that the requested lbr_select will be reconfigured
when the event is enabled. It is trivial, but a valid fix. Thanks.

> 
> Cc: Like Xu <likexu@tencent.com>
> Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
> ---
>   arch/x86/events/intel/lbr.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c
> index 9e6d6eaeb4cb..1076de93a2f5 100644
> --- a/arch/x86/events/intel/lbr.c
> +++ b/arch/x86/events/intel/lbr.c
> @@ -279,6 +279,8 @@ void intel_pmu_lbr_reset(void)
>   
>   	cpuc->last_task_ctx = NULL;
>   	cpuc->last_log_id = 0;
> +	if (!static_cpu_has(X86_FEATURE_ARCH_LBR) && cpuc->lbr_select)
> +		wrmsrl(MSR_LBR_SELECT, 0);
>   }
>   
>   /*
> 

[tip: perf/urgent] perf/x86/lbr: Reset LBR_SELECT during vlbr reset

[tip-bot2 for Wanpeng Li]

The following commit has been merged into the perf/urgent branch of tip:

Commit-ID:     0fe39a3929ac7af980347814d552a734b51adacf
Gitweb:        https://git.kernel.org/tip/0fe39a3929ac7af980347814d552a734b51adacf
Author:        Wanpeng Li <wanpengli@tencent.com>
AuthorDate:    Fri, 05 Nov 2021 00:20:51 -07:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Thu, 11 Nov 2021 13:09:34 +01:00

perf/x86/lbr: Reset LBR_SELECT during vlbr reset

lbr_select in kvm guest has residual data even if kvm guest is poweroff.
We can get residual data in the next boot. Because lbr_select is not
reset during kvm vlbr release. Let's reset LBR_SELECT during vlbr reset.

Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/1636096851-36623-1-git-send-email-wanpengli@tencent.com
---
 arch/x86/events/intel/lbr.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c
index 6b72e9b..8043213 100644
--- a/arch/x86/events/intel/lbr.c
+++ b/arch/x86/events/intel/lbr.c
@@ -265,6 +265,8 @@ void intel_pmu_lbr_reset(void)
 
 	cpuc->last_task_ctx = NULL;
 	cpuc->last_log_id = 0;
+	if (!static_cpu_has(X86_FEATURE_ARCH_LBR) && cpuc->lbr_select)
+		wrmsrl(MSR_LBR_SELECT, 0);
 }
 
 /*