[PATCH v4 04/13] coresight: Hide unused ETMv3 format attributes

James Clark james.clark at linaro.org
Tue Nov 18 06:56:33 PST 2025



On 14/11/2025 2:55 pm, Leo Yan wrote:
> On Wed, Nov 12, 2025 at 03:22:10PM +0000, James Clark wrote:
>> ETMv3 only has a few attributes, and setting unused ones results in an
>> error, so hide them to begin with.
>>
>> Signed-off-by: James Clark <james.clark at linaro.org>
>> ---
>>   drivers/hwtracing/coresight/coresight-etm-perf.c | 19 +++++++++++++++++++
>>   1 file changed, 19 insertions(+)
>>
>> diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c
>> index 17afa0f4cdee..91132abca244 100644
>> --- a/drivers/hwtracing/coresight/coresight-etm-perf.c
>> +++ b/drivers/hwtracing/coresight/coresight-etm-perf.c
>> @@ -106,8 +106,27 @@ static struct attribute *etm_config_formats_attr[] = {
>>   	NULL,
>>   };
>>   
>> +static umode_t etm_format_attr_is_visible(struct kobject *kobj,
>> +					  struct attribute *attr, int unused)
>> +{
>> +	/* ETM4 has all attributes */
>> +	if (IS_ENABLED(CONFIG_CORESIGHT_SOURCE_ETM4X))
>> +		return attr->mode;
>> +
>> +	/* ETM3 only has these attributes */
>> +	if (attr == &format_attr_cycacc.attr ||
>> +	    attr == &format_attr_timestamp.attr ||
>> +	    attr == &format_attr_retstack.attr ||
>> +	    attr == &format_attr_sinkid.attr ||
>> +	    attr == &format_attr_configid.attr)
> 
> Do we support configid for ETM3?
> 

They're "activated" in etm_setup_aux() so this would be for both ETM3 
and 4. But they're only subsequently "enabled" for ETM4. So no, they're 
not supported. I'll hide this attribute too.

I can leave it for a later fix if we want to stop activating them, but I 
don't think it does any harm.

>> +		return attr->mode;
> 
> It is good to give a bit information why only these attributes
> supported, e.g.,
> 
>    /*
>     * ETM3 only support subset attributes (see ETM3X_SUPPORTED_OPTIONS),
>     * keep 'sinkid' attr for specifying sink ID.
>     */
> 
> Two side topics but not issues caused this patch:
> 
> - Since CTXTID/CTXTID2 is only used for ETM4x, we can remove "#if
>    IS_ENABLED(CONFIG_CORESIGHT_SOURCE_ETM4X)" from the
>    format_attr_contextid_show() function.

Will change to CONFIG_ARM64 as mentioned on the other thread.

> 
> - etm_parse_event_config() does not check attr->config3, thus user
>    sets 'cc_threshold' it will slip without any error report.

Will ignore this comment as mentioned on patch 7, we're not doing any 
errors for unsupported arguments anymore.

> 
> Thanks,
> Leo
> 
>> +
>> +	return 0;
>> +}
>> +
>>   static const struct attribute_group etm_pmu_format_group = {
>>   	.name   = "format",
>> +	.is_visible = etm_format_attr_is_visible,
>>   	.attrs  = etm_config_formats_attr,
>>   };
>>   
>>
>> -- 
>> 2.34.1
>>




More information about the linux-arm-kernel mailing list