2023-09-14 11:33:54

by Yicong Yang

[permalink] [raw]
Subject: [PATCH v2 5/5] hwtracing: hisi_ptt: Add dummy callback pmu::read()

From: Junhao He <[email protected]>

When start trace with perf option "-C $cpu" and immediately stop it
with SIGTERM or others, the perf core will invoke pmu::read() while
the driver doesn't implement it. Add a dummy pmu::read() to avoid
any issues.

Signed-off-by: Junhao He <[email protected]>
Signed-off-by: Yicong Yang <[email protected]>
Acked-by: Jonathan Cameron <[email protected]>
---
drivers/hwtracing/ptt/hisi_ptt.c | 5 +++++
1 file changed, 5 insertions(+)

diff --git a/drivers/hwtracing/ptt/hisi_ptt.c b/drivers/hwtracing/ptt/hisi_ptt.c
index 62a444f5228e..c1b5fd2b8974 100644
--- a/drivers/hwtracing/ptt/hisi_ptt.c
+++ b/drivers/hwtracing/ptt/hisi_ptt.c
@@ -1184,6 +1184,10 @@ static void hisi_ptt_pmu_del(struct perf_event *event, int flags)
hisi_ptt_pmu_stop(event, PERF_EF_UPDATE);
}

+static void hisi_ptt_pmu_read(struct perf_event *event)
+{
+}
+
static void hisi_ptt_remove_cpuhp_instance(void *hotplug_node)
{
cpuhp_state_remove_instance_nocalls(hisi_ptt_pmu_online, hotplug_node);
@@ -1227,6 +1231,7 @@ static int hisi_ptt_register_pmu(struct hisi_ptt *hisi_ptt)
.stop = hisi_ptt_pmu_stop,
.add = hisi_ptt_pmu_add,
.del = hisi_ptt_pmu_del,
+ .read = hisi_ptt_pmu_read,
};

reg = readl(hisi_ptt->iobase + HISI_PTT_LOCATION);
--
2.24.0


2023-09-15 13:09:39

by Suzuki K Poulose

[permalink] [raw]
Subject: Re: [PATCH v2 5/5] hwtracing: hisi_ptt: Add dummy callback pmu::read()

On 14/09/2023 12:22, Yicong Yang wrote:
> From: Junhao He <[email protected]>
>
> When start trace with perf option "-C $cpu" and immediately stop it
> with SIGTERM or others, the perf core will invoke pmu::read() while
> the driver doesn't implement it. Add a dummy pmu::read() to avoid
> any issues.

What issues are we talking about here ? Shouldn't the core perf
skip the call, if pmu::read() is not available ?

Suzuki

>
> Signed-off-by: Junhao He <[email protected]>
> Signed-off-by: Yicong Yang <[email protected]>
> Acked-by: Jonathan Cameron <[email protected]>
> ---
> drivers/hwtracing/ptt/hisi_ptt.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/drivers/hwtracing/ptt/hisi_ptt.c b/drivers/hwtracing/ptt/hisi_ptt.c
> index 62a444f5228e..c1b5fd2b8974 100644
> --- a/drivers/hwtracing/ptt/hisi_ptt.c
> +++ b/drivers/hwtracing/ptt/hisi_ptt.c
> @@ -1184,6 +1184,10 @@ static void hisi_ptt_pmu_del(struct perf_event *event, int flags)
> hisi_ptt_pmu_stop(event, PERF_EF_UPDATE);
> }
>
> +static void hisi_ptt_pmu_read(struct perf_event *event)
> +{
> +}
> +
> static void hisi_ptt_remove_cpuhp_instance(void *hotplug_node)
> {
> cpuhp_state_remove_instance_nocalls(hisi_ptt_pmu_online, hotplug_node);
> @@ -1227,6 +1231,7 @@ static int hisi_ptt_register_pmu(struct hisi_ptt *hisi_ptt)
> .stop = hisi_ptt_pmu_stop,
> .add = hisi_ptt_pmu_add,
> .del = hisi_ptt_pmu_del,
> + .read = hisi_ptt_pmu_read,
> };
>
> reg = readl(hisi_ptt->iobase + HISI_PTT_LOCATION);

2023-09-19 13:27:37

by Yicong Yang

[permalink] [raw]
Subject: Re: [PATCH v2 5/5] hwtracing: hisi_ptt: Add dummy callback pmu::read()

On 2023/9/15 20:53, Suzuki K Poulose wrote:
> On 14/09/2023 12:22, Yicong Yang wrote:
>> From: Junhao He <[email protected]>
>>
>> When start trace with perf option "-C $cpu" and immediately stop it
>> with SIGTERM or others, the perf core will invoke pmu::read() while
>> the driver doesn't implement it. Add a dummy pmu::read() to avoid
>> any issues.
>
> What issues are we talking about here ? Shouldn't the core perf
> skip the call, if pmu::read() is not available ?
>

Actually no, the core doesn't check it. So I think that's why some PMUs
like SPE implements a dummy pmu::read() callback. Otherwise we'll
dereference a NULL pointer.

Currently we only met this on emulated platforms with very slow CPUs,
follow the instructions in the commit above.

> Suzuki
>
>>
>> Signed-off-by: Junhao He <[email protected]>
>> Signed-off-by: Yicong Yang <[email protected]>
>> Acked-by: Jonathan Cameron <[email protected]>
>> ---
>>   drivers/hwtracing/ptt/hisi_ptt.c | 5 +++++
>>   1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/hwtracing/ptt/hisi_ptt.c b/drivers/hwtracing/ptt/hisi_ptt.c
>> index 62a444f5228e..c1b5fd2b8974 100644
>> --- a/drivers/hwtracing/ptt/hisi_ptt.c
>> +++ b/drivers/hwtracing/ptt/hisi_ptt.c
>> @@ -1184,6 +1184,10 @@ static void hisi_ptt_pmu_del(struct perf_event *event, int flags)
>>       hisi_ptt_pmu_stop(event, PERF_EF_UPDATE);
>>   }
>>   +static void hisi_ptt_pmu_read(struct perf_event *event)
>> +{
>> +}
>> +
>>   static void hisi_ptt_remove_cpuhp_instance(void *hotplug_node)
>>   {
>>       cpuhp_state_remove_instance_nocalls(hisi_ptt_pmu_online, hotplug_node);
>> @@ -1227,6 +1231,7 @@ static int hisi_ptt_register_pmu(struct hisi_ptt *hisi_ptt)
>>           .stop        = hisi_ptt_pmu_stop,
>>           .add        = hisi_ptt_pmu_add,
>>           .del        = hisi_ptt_pmu_del,
>> +        .read        = hisi_ptt_pmu_read,
>>       };
>>         reg = readl(hisi_ptt->iobase + HISI_PTT_LOCATION);
>
>
> .

2023-09-19 17:10:04

by Suzuki K Poulose

[permalink] [raw]
Subject: Re: [PATCH v2 5/5] hwtracing: hisi_ptt: Add dummy callback pmu::read()

On 19/09/2023 14:03, Yicong Yang wrote:
> On 2023/9/15 20:53, Suzuki K Poulose wrote:
>> On 14/09/2023 12:22, Yicong Yang wrote:
>>> From: Junhao He <[email protected]>
>>>
>>> When start trace with perf option "-C $cpu" and immediately stop it
>>> with SIGTERM or others, the perf core will invoke pmu::read() while
>>> the driver doesn't implement it. Add a dummy pmu::read() to avoid
>>> any issues.
>>
>> What issues are we talking about here ? Shouldn't the core perf
>> skip the call, if pmu::read() is not available ?
>>
>
> Actually no, the core doesn't check it. So I think that's why some PMUs
> like SPE implements a dummy pmu::read() callback. Otherwise we'll
> dereference a NULL pointer.
>
> Currently we only met this on emulated platforms with very slow CPUs,
> follow the instructions in the commit above.

Ok, then it calls for a Fixes tag. Please tag it to the commit that
introduced the PMU.

Suzuki

2023-09-20 07:19:32

by Yicong Yang

[permalink] [raw]
Subject: Re: [PATCH v2 5/5] hwtracing: hisi_ptt: Add dummy callback pmu::read()

On 2023/9/20 1:01, Suzuki K Poulose wrote:
> On 19/09/2023 14:03, Yicong Yang wrote:
>> On 2023/9/15 20:53, Suzuki K Poulose wrote:
>>> On 14/09/2023 12:22, Yicong Yang wrote:
>>>> From: Junhao He <[email protected]>
>>>>
>>>> When start trace with perf option "-C $cpu" and immediately stop it
>>>> with SIGTERM or others, the perf core will invoke pmu::read() while
>>>> the driver doesn't implement it. Add a dummy pmu::read() to avoid
>>>> any issues.
>>>
>>> What issues are we talking about here ? Shouldn't the core perf
>>> skip the call, if pmu::read() is not available ?
>>>
>>
>> Actually no, the core doesn't check it. So I think that's why some PMUs
>> like SPE implements a dummy pmu::read() callback. Otherwise we'll
>> dereference a NULL pointer.
>>
>> Currently we only met this on emulated platforms with very slow CPUs,
>> follow the instructions in the commit above.
>
> Ok, then it calls for a Fixes tag. Please tag it to the commit that
> introduced the PMU.
>

Sure. I'll add the tag in v3.

Thanks.