[RFC,1/4] driver/perf: Add identifier sysfs file for CMN

  To allow userspace to identify the specific implementation of the device,
add an "identifier" sysfs file.

The perf tool can match the arm CMN metric through the identifier.

Signed-off-by: Jing Zhang <renyu.zj@linux.alibaba.com>
---
 drivers/perf/arm-cmn.c | 43 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)
  

On 27/03/2023 03:46, Jing Zhang wrote:
> To allow userspace to identify the specific implementation of the device,
> add an "identifier" sysfs file.
> 
> The perf tool can match the arm CMN metric through the identifier.
> 
> Signed-off-by: Jing Zhang <renyu.zj@linux.alibaba.com>
> ---
>   drivers/perf/arm-cmn.c | 43 +++++++++++++++++++++++++++++++++++++++++++
>   1 file changed, 43 insertions(+)
> 
> diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c
> index c968986..0c138ad 100644
> --- a/drivers/perf/arm-cmn.c
> +++ b/drivers/perf/arm-cmn.c
> @@ -1168,10 +1168,53 @@ static ssize_t arm_cmn_cpumask_show(struct device *dev,
>   	.attrs = arm_cmn_cpumask_attrs,
>   };
>   
> +static ssize_t arm_cmn_identifier_show(struct device *dev,
> +		struct device_attribute *attr, char *buf)
> +{
> +	struct arm_cmn *cmn = to_cmn(dev_get_drvdata(dev));
> +	if (cmn->model == CMN700) {
> +		return sysfs_emit(buf, "%s\n", "CMN700");

Is it possible to have a pointer to this string in struct arm_cmn, such 
that we don't have to do this model to identifier lookup here? If-else 
chains like this are not scalable.

BTW, does this HW have some HW identifier register, like iidr? I think 
that using that may be preferable.

> +	}
> +	else if (cmn->model == CMN650) {
> +		return sysfs_emit(buf, "%s\n", "CMN650");

I'd use lowercase names

> +	}
> +	else if (cmn->model == CMN600) {
> +		return sysfs_emit(buf, "%s\n", "CMN600");
> +	}
> +	else if (cmn->model == CI700) {
> +		return sysfs_emit(buf, "%s\n", "CI700");
> +	}
> +	return sysfs_emit(buf, "%s\n", "UNKNOWN");

can we have a "is_visble" attr to just no show this when unknown?

> +}
> +
> +static umode_t arm_cmn_identifier_attr_visible(struct kobject *kobj,
> +		struct attribute *attr, int n)
> +{
> +	struct device *dev = kobj_to_dev(kobj);
> +	struct arm_cmn *cmn = to_cmn(dev_get_drvdata(dev));
> +	if (cmn->model <= 0)
> +		return 0;
> +	return attr->mode;
> +};
> +
> +static struct device_attribute arm_cmn_identifier_attr =
> +__ATTR(identifier, 0444, arm_cmn_identifier_show, NULL);
> +
> +static struct attribute *arm_cmn_identifier_attrs[] = {
> +	&arm_cmn_identifier_attr.attr,
> +	NULL,

nit: no need for trailing ',' on a sentinel

> +};
> +
> +static struct attribute_group arm_cmn_identifier_attr_group = {
> +	.attrs = arm_cmn_identifier_attrs,
> +	.is_visible = arm_cmn_identifier_attr_visible,
> +};
> +
>   static const struct attribute_group *arm_cmn_attr_groups[] = {
>   	&arm_cmn_event_attrs_group,
>   	&arm_cmn_format_attrs_group,
>   	&arm_cmn_cpumask_attr_group,
> +	&arm_cmn_identifier_attr_group,
>   	NULL
>   };
>
  

在 2023/3/27 下午3:55, John Garry 写道:
> On 27/03/2023 03:46, Jing Zhang wrote:
>> To allow userspace to identify the specific implementation of the device,
>> add an "identifier" sysfs file.
>>
>> The perf tool can match the arm CMN metric through the identifier.
>>
>> Signed-off-by: Jing Zhang <renyu.zj@linux.alibaba.com>
>> ---
>>   drivers/perf/arm-cmn.c | 43 +++++++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 43 insertions(+)
>>
>> diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c
>> index c968986..0c138ad 100644
>> --- a/drivers/perf/arm-cmn.c
>> +++ b/drivers/perf/arm-cmn.c
>> @@ -1168,10 +1168,53 @@ static ssize_t arm_cmn_cpumask_show(struct device *dev,
>>       .attrs = arm_cmn_cpumask_attrs,
>>   };
>>   +static ssize_t arm_cmn_identifier_show(struct device *dev,
>> +        struct device_attribute *attr, char *buf)
>> +{
>> +    struct arm_cmn *cmn = to_cmn(dev_get_drvdata(dev));
>> +    if (cmn->model == CMN700) {
>> +        return sysfs_emit(buf, "%s\n", "CMN700");
> 
> Is it possible to have a pointer to this string in struct arm_cmn, such that we don't have to do this model to identifier lookup here? If-else chains like this are not scalable.
> 
Will do.

> BTW, does this HW have some HW identifier register, like iidr? I think that using that may be preferable.
> 

I didn't find the relevant identifier register.

Do Illka and Robin know that there is such a register that can identify different CMN versions? Looking forward to your suggestions.


>> +    }
>> +    else if (cmn->model == CMN650) {
>> +        return sysfs_emit(buf, "%s\n", "CMN650");
> 
> I'd use lowercase names
> 
Ok.

>> +    }
>> +    else if (cmn->model == CMN600) {
>> +        return sysfs_emit(buf, "%s\n", "CMN600");
>> +    }
>> +    else if (cmn->model == CI700) {
>> +        return sysfs_emit(buf, "%s\n", "CI700");
>> +    }
>> +    return sysfs_emit(buf, "%s\n", "UNKNOWN");
> 
> can we have a "is_visble" attr to just no show this when unknown?
> 

Ok.

>> +}
>> +
>> +static umode_t arm_cmn_identifier_attr_visible(struct kobject *kobj,
>> +        struct attribute *attr, int n)
>> +{
>> +    struct device *dev = kobj_to_dev(kobj);
>> +    struct arm_cmn *cmn = to_cmn(dev_get_drvdata(dev));
>> +    if (cmn->model <= 0)
>> +        return 0;
>> +    return attr->mode;
>> +};
>> +
>> +static struct device_attribute arm_cmn_identifier_attr =
>> +__ATTR(identifier, 0444, arm_cmn_identifier_show, NULL);
>> +
>> +static struct attribute *arm_cmn_identifier_attrs[] = {
>> +    &arm_cmn_identifier_attr.attr,
>> +    NULL,
> 
> nit: no need for trailing ',' on a sentinel
> 

Ok, Will do.

>> +};
>> +
>> +static struct attribute_group arm_cmn_identifier_attr_group = {
>> +    .attrs = arm_cmn_identifier_attrs,
>> +    .is_visible = arm_cmn_identifier_attr_visible,
>> +};
>> +
>>   static const struct attribute_group *arm_cmn_attr_groups[] = {
>>       &arm_cmn_event_attrs_group,
>>       &arm_cmn_format_attrs_group,
>>       &arm_cmn_cpumask_attr_group,
>> +    &arm_cmn_identifier_attr_group,
>>       NULL
>>   };
>>
  

On 2023-03-29 12:53, Jing Zhang wrote:
> 
> 
> 在 2023/3/27 下午3:55, John Garry 写道:
>> On 27/03/2023 03:46, Jing Zhang wrote:
>>> To allow userspace to identify the specific implementation of the device,
>>> add an "identifier" sysfs file.
>>>
>>> The perf tool can match the arm CMN metric through the identifier.
>>>
>>> Signed-off-by: Jing Zhang <renyu.zj@linux.alibaba.com>
>>> ---
>>>    drivers/perf/arm-cmn.c | 43 +++++++++++++++++++++++++++++++++++++++++++
>>>    1 file changed, 43 insertions(+)
>>>
>>> diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c
>>> index c968986..0c138ad 100644
>>> --- a/drivers/perf/arm-cmn.c
>>> +++ b/drivers/perf/arm-cmn.c
>>> @@ -1168,10 +1168,53 @@ static ssize_t arm_cmn_cpumask_show(struct device *dev,
>>>        .attrs = arm_cmn_cpumask_attrs,
>>>    };
>>>    +static ssize_t arm_cmn_identifier_show(struct device *dev,
>>> +        struct device_attribute *attr, char *buf)
>>> +{
>>> +    struct arm_cmn *cmn = to_cmn(dev_get_drvdata(dev));
>>> +    if (cmn->model == CMN700) {
>>> +        return sysfs_emit(buf, "%s\n", "CMN700");
>>
>> Is it possible to have a pointer to this string in struct arm_cmn, such that we don't have to do this model to identifier lookup here? If-else chains like this are not scalable.
>>
> Will do.
> 
>> BTW, does this HW have some HW identifier register, like iidr? I think that using that may be preferable.
>>
> 
> I didn't find the relevant identifier register.
> 
> Do Illka and Robin know that there is such a register that can identify different CMN versions? Looking forward to your suggestions.

In principle the "part number" fields from CFGM_PERIPH_ID_0/1 are 
supposed to identify the model, but for various reasons I'm suspicious 
of that being unreliable (not least that no actual values are 
documented, only "configuration-dependent"). That's why I went down the 
route of making sure we have explicit ACPI/DT identifiers for every model.

However, the model alone seems either too specific or not specific 
enough for a jevents identifier. The defined metrics are pretty trivial 
and should have no real reason not to be common to *any* CMN PMU where 
the underlying events are present. On the other hand, if we want to get 
down to the level of specific events in JSON then we'd need to consider 
the revision as well, since there are several events which only exist on 
certain revisions of a given model (but often are also common to later 
models).

This actually foreshadows a question I was planning to bring up in the 
context of another driver I'm working on - for this one I would rather 
like to try using jevents rather than have to maintain another sprawl of 
event tables in a driver, but it's still going to have the same thing of 
wanting model/revision matching along the lines of what 
arm_cmn_event_attr_is_visible() is doing for CMN events. AFAICS this 
would need jevents to grow a rather more flexible way of encoding and 
matching identifiers, since having dozens of almost-identical copies of 
event definitions for every exact identifier value is clearly 
unworkable. Does anyone happen to have any thoughts or preferences 
around how that might be approached?

>>> +    }
>>> +    else if (cmn->model == CMN650) {
>>> +        return sysfs_emit(buf, "%s\n", "CMN650");
>>
>> I'd use lowercase names
>>
> Ok.
> 
>>> +    }
>>> +    else if (cmn->model == CMN600) {
>>> +        return sysfs_emit(buf, "%s\n", "CMN600");
>>> +    }
>>> +    else if (cmn->model == CI700) {
>>> +        return sysfs_emit(buf, "%s\n", "CI700");
>>> +    }
>>> +    return sysfs_emit(buf, "%s\n", "UNKNOWN");
>>
>> can we have a "is_visble" attr to just no show this when unknown?

No need - it will never be unknown unless someone goes out of their way 
to break the probing code and/or match_data.

>>
> 
> Ok.
> 
>>> +}
>>> +
>>> +static umode_t arm_cmn_identifier_attr_visible(struct kobject *kobj,
>>> +        struct attribute *attr, int n)
>>> +{
>>> +    struct device *dev = kobj_to_dev(kobj);
>>> +    struct arm_cmn *cmn = to_cmn(dev_get_drvdata(dev));
>>> +    if (cmn->model <= 0)
>>> +        return 0;
>>> +    return attr->mode;
>>> +};

As above, "cmn->model <= 0" can never be true.

Thanks,
Robin.

>>> +
>>> +static struct device_attribute arm_cmn_identifier_attr =
>>> +__ATTR(identifier, 0444, arm_cmn_identifier_show, NULL);
>>> +
>>> +static struct attribute *arm_cmn_identifier_attrs[] = {
>>> +    &arm_cmn_identifier_attr.attr,
>>> +    NULL,
>>
>> nit: no need for trailing ',' on a sentinel
>>
> 
> Ok, Will do.
> 
>>> +};
>>> +
>>> +static struct attribute_group arm_cmn_identifier_attr_group = {
>>> +    .attrs = arm_cmn_identifier_attrs,
>>> +    .is_visible = arm_cmn_identifier_attr_visible,
>>> +};
>>> +
>>>    static const struct attribute_group *arm_cmn_attr_groups[] = {
>>>        &arm_cmn_event_attrs_group,
>>>        &arm_cmn_format_attrs_group,
>>>        &arm_cmn_cpumask_attr_group,
>>> +    &arm_cmn_identifier_attr_group,
>>>        NULL
>>>    };
>>>
  

On 29/03/2023 18:47, Robin Murphy wrote:
>> Do Illka and Robin know that there is such a register that can 
>> identify different CMN versions? Looking forward to your suggestions.
> 
> In principle the "part number" fields from CFGM_PERIPH_ID_0/1 are 
> supposed to identify the model, but for various reasons I'm suspicious 
> of that being unreliable (not least that no actual values are 
> documented, only "configuration-dependent"). That's why I went down the 
> route of making sure we have explicit ACPI/DT identifiers for every model.
> 
> However, the model alone seems either too specific or not specific 
> enough for a jevents identifier. The defined metrics are pretty trivial 
> and should have no real reason not to be common to *any* CMN PMU where 
> the underlying events are present. On the other hand, if we want to get 
> down to the level of specific events in JSON then we'd need to consider 
> the revision as well, since there are several events which only exist on 
> certain revisions of a given model (but often are also common to later 
> models).
> 
> This actually foreshadows a question I was planning to bring up in the 
> context of another driver I'm working on - for this one I would rather 
> like to try using jevents rather than have to maintain another sprawl of 
> event tables in a driver, but it's still going to have the same thing of 
> wanting model/revision matching along the lines of what 
> arm_cmn_event_attr_is_visible() is doing for CMN events. AFAICS this 
> would need jevents to grow a rather more flexible way of encoding and 
> matching identifiers, since having dozens of almost-identical copies of 
> event definitions for every exact identifier value is clearly 
> unworkable.

This sort of problem has not occurred yet as perf tool only supports 
"system" events for a handful of SoCs so far :)

> Does anyone happen to have any thoughts or preferences 
> around how that might be approached?
> 

Currently the perf tool will just match system events based on the exact 
HW identifier and PMU name.

However, if you consider PMCG PMU support as an example of possible area 
of improvement, it has a number of fixed events and a number of IMPDEF 
events. There should be no reason to need to describe in a separate JSON 
those fixed events for every instance of that PMU.

So a simple change would be to teach perf tool that for certain fixed 
events we only need to match based on the PMU name. For others we need 
to match based on some identifier.

If matching based on an identifier still leads to unwieldy amounts of 
tables, then maybe HW identifier wildcard matching may suit, like what 
is done for CPU events.

Thanks,
John