[PATCH] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member

linux-perf-users.vger.kernel.org archive mirror
 help / color / mirror / Atom feed

* [PATCH] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
@ 2025-06-19  3:19 tanze
  2025-06-19 14:33 ` Liang, Kan
  2025-09-28  9:44 ` [PATCH] " tanze
  0 siblings, 2 replies; 9+ messages in thread
From: tanze @ 2025-06-19  3:19 UTC (permalink / raw)
  To: peterz, kan.liang; +Cc: linux-perf-users, linux-kernel, x86, tanze

Perf's instructions event tests on Zhaoxin CPUs may exhibit:

  $perf stat -e instructions,cycles ls -l

  ......
  Performance counter stats for 'ls -l':

                 0      instructions                     #    0.00  insn per cycle
         9,488,278      cycles

       0.004365407 seconds time elapsed

       0.003303000 seconds user
       0.001099000 seconds sys

The absence of the fixedctr member leads to an incorrect hwc->event_base
value on Zhaoxin CPUs, causing a discrepancy in the instruction count
reported by perf stat. This commit resolves the instruction count issue
by properly initializing the fixedctr member.

Fixes: 149fd4712bcd ("perf/x86/intel: Support Perfmon MSRs aliasing")

Signed-off-by: tanze <tanze@kylinos.cn>
---
 arch/x86/events/zhaoxin/core.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/events/zhaoxin/core.c b/arch/x86/events/zhaoxin/core.c
index 4bdfcf091200..3fc3f9abece9 100644
--- a/arch/x86/events/zhaoxin/core.c
+++ b/arch/x86/events/zhaoxin/core.c
@@ -467,6 +467,7 @@ static const struct x86_pmu zhaoxin_pmu __initconst = {
 	.schedule_events	= x86_schedule_events,
 	.eventsel		= MSR_ARCH_PERFMON_EVENTSEL0,
 	.perfctr		= MSR_ARCH_PERFMON_PERFCTR0,
+	.fixedctr		= MSR_ARCH_PERFMON_FIXED_CTR0,
 	.event_map		= zhaoxin_pmu_event_map,
 	.max_events		= ARRAY_SIZE(zx_pmon_event_map),
 	.apic			= 1,
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* Re: [PATCH] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
  2025-06-19  3:19 [PATCH] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member tanze
@ 2025-06-19 14:33 ` Liang, Kan
  2025-06-20  1:51   ` [PATCH v2 v2] " tanze
                     ` (3 more replies)
  2025-09-28  9:44 ` [PATCH] " tanze
  1 sibling, 4 replies; 9+ messages in thread
From: Liang, Kan @ 2025-06-19 14:33 UTC (permalink / raw)
  To: tanze, peterz; +Cc: linux-perf-users, linux-kernel, x86



On 2025-06-18 11:19 p.m., tanze wrote:
> Perf's instructions event tests on Zhaoxin CPUs may exhibit:
> 
>   $perf stat -e instructions,cycles ls -l
> 
>   ......
>   Performance counter stats for 'ls -l':
> 
>                  0      instructions                     #    0.00  insn per cycle
>          9,488,278      cycles
> 
>        0.004365407 seconds time elapsed
> 
>        0.003303000 seconds user
>        0.001099000 seconds sys
> 
> The absence of the fixedctr member leads to an incorrect hwc->event_base
> value on Zhaoxin CPUs, causing a discrepancy in the instruction count
> reported by perf stat. This commit resolves the instruction count issue
> by properly initializing the fixedctr member.
> 
> Fixes: 149fd4712bcd ("perf/x86/intel: Support Perfmon MSRs aliasing")
> 

Please remove the empty line.

> Signed-off-by: tanze <tanze@kylinos.cn>

I have no idea Zhaoxin also have fixed counters.
If so, Yes, I think the patch is required.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>

Thanks,
Kan> ---
>  arch/x86/events/zhaoxin/core.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/x86/events/zhaoxin/core.c b/arch/x86/events/zhaoxin/core.c
> index 4bdfcf091200..3fc3f9abece9 100644
> --- a/arch/x86/events/zhaoxin/core.c
> +++ b/arch/x86/events/zhaoxin/core.c
> @@ -467,6 +467,7 @@ static const struct x86_pmu zhaoxin_pmu __initconst = {
>  	.schedule_events	= x86_schedule_events,
>  	.eventsel		= MSR_ARCH_PERFMON_EVENTSEL0,
>  	.perfctr		= MSR_ARCH_PERFMON_PERFCTR0,
> +	.fixedctr		= MSR_ARCH_PERFMON_FIXED_CTR0,
>  	.event_map		= zhaoxin_pmu_event_map,
>  	.max_events		= ARRAY_SIZE(zx_pmon_event_map),
>  	.apic			= 1,


^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH v2 v2] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
  2025-06-19 14:33 ` Liang, Kan
@ 2025-06-20  1:51   ` tanze
  2025-06-20  2:10   ` [PATCH 0/1] CANCELLED: [PATCH " tanze
                     ` (2 subsequent siblings)
  3 siblings, 0 replies; 9+ messages in thread
From: tanze @ 2025-06-20  1:51 UTC (permalink / raw)
  To: peterz, kan.liang; +Cc: linux-perf-users, linux-kernel, x86, tanze

Perf's instructions event tests on Zhaoxin CPUs may exhibit:

  $perf stat -e instructions,cycles ls -l

  ......
  Performance counter stats for 'ls -l':

                 0      instructions                     #    0.00  insn per cycle
         9,488,278      cycles

       0.004365407 seconds time elapsed

       0.003303000 seconds user
       0.001099000 seconds sys

The absence of the fixedctr member leads to an incorrect hwc->event_base
value on Zhaoxin CPUs, causing a discrepancy in the instruction count
reported by perf stat. This commit resolves the instruction count issue
by properly initializing the fixedctr member.

Fixes: 149fd4712bcd ("perf/x86/intel: Support Perfmon MSRs aliasing")

Signed-off-by: tanze <tanze@kylinos.cn>
---
 arch/x86/events/zhaoxin/core.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/events/zhaoxin/core.c b/arch/x86/events/zhaoxin/core.c
index 4bdfcf091200..3fc3f9abece9 100644
--- a/arch/x86/events/zhaoxin/core.c
+++ b/arch/x86/events/zhaoxin/core.c
@@ -467,6 +467,7 @@ static const struct x86_pmu zhaoxin_pmu __initconst = {
 	.schedule_events	= x86_schedule_events,
 	.eventsel		= MSR_ARCH_PERFMON_EVENTSEL0,
 	.perfctr		= MSR_ARCH_PERFMON_PERFCTR0,
+	.fixedctr		= MSR_ARCH_PERFMON_FIXED_CTR0,
 	.event_map		= zhaoxin_pmu_event_map,
 	.max_events		= ARRAY_SIZE(zx_pmon_event_map),
 	.apic			= 1,
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [PATCH 0/1] CANCELLED: [PATCH v2] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
  2025-06-19 14:33 ` Liang, Kan
  2025-06-20  1:51   ` [PATCH v2 v2] " tanze
@ 2025-06-20  2:10   ` tanze
  2025-06-20  2:19   ` tanze
  2025-06-20  2:37   ` [PATCH v3] " tanze
  3 siblings, 0 replies; 9+ messages in thread
From: tanze @ 2025-06-20  2:10 UTC (permalink / raw)
  To: peterz, kan.liang; +Cc: linux-perf-users, linux-kernel, x86, tanze

Hi all,

Please **disregard** my previous patch ([PATCH v2 perf/x86/zhaoxin: Fix instructions…]).  I will resend a corrected version asap.

Sorry for the noise.

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH v2] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
  2025-06-19 14:33 ` Liang, Kan
  2025-06-20  1:51   ` [PATCH v2 v2] " tanze
  2025-06-20  2:10   ` [PATCH 0/1] CANCELLED: [PATCH " tanze
@ 2025-06-20  2:19   ` tanze
  2025-06-20  2:37   ` [PATCH v3] " tanze
  3 siblings, 0 replies; 9+ messages in thread
From: tanze @ 2025-06-20  2:19 UTC (permalink / raw)
  To: peterz, kan.liang; +Cc: linux-perf-users, linux-kernel, x86, tanze

Perf's instructions event tests on Zhaoxin CPUs may exhibit:

  $perf stat -e instructions,cycles ls -l

  ......
  Performance counter stats for 'ls -l':

                 0      instructions                     #    0.00  insn per cycle
         9,488,278      cycles

       0.004365407 seconds time elapsed

       0.003303000 seconds user
       0.001099000 seconds sys

The absence of the fixedctr member leads to an incorrect hwc->event_base
value on Zhaoxin CPUs, causing a discrepancy in the instruction count
reported by perf stat. This commit resolves the instruction count issue
by properly initializing the fixedctr member.

Fixes: 149fd4712bcd ("perf/x86/intel: Support Perfmon MSRs aliasing")
Signed-off-by: tanze <tanze@kylinos.cn>
---
 arch/x86/events/zhaoxin/core.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/events/zhaoxin/core.c b/arch/x86/events/zhaoxin/core.c
index 4bdfcf091200..3fc3f9abece9 100644
--- a/arch/x86/events/zhaoxin/core.c
+++ b/arch/x86/events/zhaoxin/core.c
@@ -467,6 +467,7 @@ static const struct x86_pmu zhaoxin_pmu __initconst = {
 	.schedule_events	= x86_schedule_events,
 	.eventsel		= MSR_ARCH_PERFMON_EVENTSEL0,
 	.perfctr		= MSR_ARCH_PERFMON_PERFCTR0,
+	.fixedctr		= MSR_ARCH_PERFMON_FIXED_CTR0,
 	.event_map		= zhaoxin_pmu_event_map,
 	.max_events		= ARRAY_SIZE(zx_pmon_event_map),
 	.apic			= 1,
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* [PATCH v3] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
  2025-06-19 14:33 ` Liang, Kan
                     ` (2 preceding siblings ...)
  2025-06-20  2:19   ` tanze
@ 2025-06-20  2:37   ` tanze
  2025-09-30  2:49     ` Mi, Dapeng
  3 siblings, 1 reply; 9+ messages in thread
From: tanze @ 2025-06-20  2:37 UTC (permalink / raw)
  To: peterz, kan.liang; +Cc: linux-perf-users, linux-kernel, x86, tanze

Perf's instructions event tests on Zhaoxin CPUs may exhibit:

  $perf stat -e instructions,cycles ls -l

  ......
  Performance counter stats for 'ls -l':

                 0      instructions                     #    0.00  insn per cycle
         9,488,278      cycles

       0.004365407 seconds time elapsed

       0.003303000 seconds user
       0.001099000 seconds sys

The absence of the fixedctr member leads to an incorrect hwc->event_base
value on Zhaoxin CPUs, causing a discrepancy in the instruction count
reported by perf stat. This commit resolves the instruction count issue
by properly initializing the fixedctr member.

Fixes: 149fd4712bcd ("perf/x86/intel: Support Perfmon MSRs aliasing")
Signed-off-by: tanze <tanze@kylinos.cn>
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
---
 arch/x86/events/zhaoxin/core.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/events/zhaoxin/core.c b/arch/x86/events/zhaoxin/core.c
index 4bdfcf091200..3fc3f9abece9 100644
--- a/arch/x86/events/zhaoxin/core.c
+++ b/arch/x86/events/zhaoxin/core.c
@@ -467,6 +467,7 @@ static const struct x86_pmu zhaoxin_pmu __initconst = {
 	.schedule_events	= x86_schedule_events,
 	.eventsel		= MSR_ARCH_PERFMON_EVENTSEL0,
 	.perfctr		= MSR_ARCH_PERFMON_PERFCTR0,
+	.fixedctr		= MSR_ARCH_PERFMON_FIXED_CTR0,
 	.event_map		= zhaoxin_pmu_event_map,
 	.max_events		= ARRAY_SIZE(zx_pmon_event_map),
 	.apic			= 1,
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* Re: [PATCH v3] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
  2025-06-20  2:37   ` [PATCH v3] " tanze
@ 2025-09-30  2:49     ` Mi, Dapeng
  2025-10-15  8:55       ` tanze
  0 siblings, 1 reply; 9+ messages in thread
From: Mi, Dapeng @ 2025-09-30  2:49 UTC (permalink / raw)
  To: tanze, peterz, kan.liang; +Cc: linux-perf-users, linux-kernel, x86


On 6/20/2025 10:37 AM, tanze wrote:
> Perf's instructions event tests on Zhaoxin CPUs may exhibit:
>
>   $perf stat -e instructions,cycles ls -l
>
>   ......
>   Performance counter stats for 'ls -l':
>
>                  0      instructions                     #    0.00  insn per cycle
>          9,488,278      cycles
>
>        0.004365407 seconds time elapsed
>
>        0.003303000 seconds user
>        0.001099000 seconds sys
>
> The absence of the fixedctr member leads to an incorrect hwc->event_base
> value on Zhaoxin CPUs, causing a discrepancy in the instruction count
> reported by perf stat. This commit resolves the instruction count issue
> by properly initializing the fixedctr member.
>
> Fixes: 149fd4712bcd ("perf/x86/intel: Support Perfmon MSRs aliasing")
> Signed-off-by: tanze <tanze@kylinos.cn>
> Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
> ---
>  arch/x86/events/zhaoxin/core.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/arch/x86/events/zhaoxin/core.c b/arch/x86/events/zhaoxin/core.c
> index 4bdfcf091200..3fc3f9abece9 100644
> --- a/arch/x86/events/zhaoxin/core.c
> +++ b/arch/x86/events/zhaoxin/core.c
> @@ -467,6 +467,7 @@ static const struct x86_pmu zhaoxin_pmu __initconst = {
>  	.schedule_events	= x86_schedule_events,
>  	.eventsel		= MSR_ARCH_PERFMON_EVENTSEL0,
>  	.perfctr		= MSR_ARCH_PERFMON_PERFCTR0,
> +	.fixedctr		= MSR_ARCH_PERFMON_FIXED_CTR0,
>  	.event_map		= zhaoxin_pmu_event_map,
>  	.max_events		= ARRAY_SIZE(zx_pmon_event_map),
>  	.apic			= 1,

LGTM.

Reviewed-by: Dapeng Mi <dapeng1.mi@linux.intel.com>



^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH v3] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
  2025-09-30  2:49     ` Mi, Dapeng
@ 2025-10-15  8:55       ` tanze
  0 siblings, 0 replies; 9+ messages in thread
From: tanze @ 2025-10-15  8:55 UTC (permalink / raw)
  To: Mi, Dapeng, peterz, kan.liang; +Cc: linux-perf-users, linux-kernel, x86

Hi Dapeng,

在 2025/9/30 10:49, Mi, Dapeng 写道:
> 
> On 6/20/2025 10:37 AM, tanze wrote:
>> Perf's instructions event tests on Zhaoxin CPUs may exhibit:
>>
>>    $perf stat -e instructions,cycles ls -l
>>
>>    ......
>>    Performance counter stats for 'ls -l':
>>
>>                   0      instructions                     #    0.00  insn per cycle
>>           9,488,278      cycles
>>
>>         0.004365407 seconds time elapsed
>>
>>         0.003303000 seconds user
>>         0.001099000 seconds sys
>>
>> The absence of the fixedctr member leads to an incorrect hwc->event_base
>> value on Zhaoxin CPUs, causing a discrepancy in the instruction count
>> reported by perf stat. This commit resolves the instruction count issue
>> by properly initializing the fixedctr member.
>>
>> Fixes: 149fd4712bcd ("perf/x86/intel: Support Perfmon MSRs aliasing")
>> Signed-off-by: tanze <tanze@kylinos.cn>
>> Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
>> ---
>>   arch/x86/events/zhaoxin/core.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/arch/x86/events/zhaoxin/core.c b/arch/x86/events/zhaoxin/core.c
>> index 4bdfcf091200..3fc3f9abece9 100644
>> --- a/arch/x86/events/zhaoxin/core.c
>> +++ b/arch/x86/events/zhaoxin/core.c
>> @@ -467,6 +467,7 @@ static const struct x86_pmu zhaoxin_pmu __initconst = {
>>   	.schedule_events	= x86_schedule_events,
>>   	.eventsel		= MSR_ARCH_PERFMON_EVENTSEL0,
>>   	.perfctr		= MSR_ARCH_PERFMON_PERFCTR0,
>> +	.fixedctr		= MSR_ARCH_PERFMON_FIXED_CTR0,
>>   	.event_map		= zhaoxin_pmu_event_map,
>>   	.max_events		= ARRAY_SIZE(zx_pmon_event_map),
>>   	.apic			= 1,
> 
> LGTM.
> 
> Reviewed-by: Dapeng Mi <dapeng1.mi@linux.intel.com>

Thank You
Ze Tan

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member
  2025-06-19  3:19 [PATCH] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member tanze
  2025-06-19 14:33 ` Liang, Kan
@ 2025-09-28  9:44 ` tanze
  1 sibling, 0 replies; 9+ messages in thread
From: tanze @ 2025-09-28  9:44 UTC (permalink / raw)
  To: peterz, kan.liang; +Cc: linux-perf-users, linux-kernel, x86

Hi, peter，Kan，

Just a gentle ping on this patch.

Thanks for your time.

--
Best regards,
Ze Tan

^ permalink raw reply	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2025-10-15  8:55 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-06-19  3:19 [PATCH] perf/x86/zhaoxin: Fix instructions error by missing fixedctr member tanze
2025-06-19 14:33 ` Liang, Kan
2025-06-20  1:51   ` [PATCH v2 v2] " tanze
2025-06-20  2:10   ` [PATCH 0/1] CANCELLED: [PATCH " tanze
2025-06-20  2:19   ` tanze
2025-06-20  2:37   ` [PATCH v3] " tanze
2025-09-30  2:49     ` Mi, Dapeng
2025-10-15  8:55       ` tanze
2025-09-28  9:44 ` [PATCH] " tanze

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).