Re: [PATCH] libtracefs: Unit test for the tracing filter API

[sameeruddin shaik <sameeruddin.shaik8@gmail.com>]

On 02/04/21 8:26 pm, Steven Rostedt wrote:
> On Sat,  3 Apr 2021 07:31:00 +0530
> Sameeruddin shaik <sameeruddin.shaik8@gmail.com> wrote:
>
>> tracefs_function_filter();
> Need some more information in the change log. What is this testing?
>
>> Signed-off-by: Sameeruddin shaik <sameeruddin.shaik8@gmail.com>
>>
>> diff --git a/utest/tracefs-utest.c b/utest/tracefs-utest.c
>> index ed2693b..09e564f 100644
>> --- a/utest/tracefs-utest.c
>> +++ b/utest/tracefs-utest.c
>> @@ -11,6 +11,7 @@
>>   #include <time.h>
>>   #include <dirent.h>
>>   #include <ftw.h>
>> +#include <errno.h>
>>   
>>   #include <CUnit/CUnit.h>
>>   #include <CUnit/Basic.h>
>> @@ -1020,6 +1021,80 @@ static void test_custom_trace_dir(void)
>>   	free(dname);
>>   }
>>   
>> +static int test_instance_filter(struct tracefs_instance *instance,
>> +				const char **filters, const char *module,
>> +				int flags)
>> +{
>> +	int ret;
>> +	int i;
>> +
>> +	if (filters) {
>> +		for (i = 0; filters[i]; i++) {
>> +			ret = tracefs_function_filter(instance, filters[i],
>> +						      module, flags);
>> +			if (ret) {
>> +				if (errno == EINVAL)
>> +					printf("Filter %s did not match\n",
>> +					       filters[i]);
>> +				else
>> +					printf("Failed writing %s\n",
>> +					       filters[i]);
>> +			}
>> +		}
>> +	} else {
>> +		ret = tracefs_function_filter(instance, NULL, module, flags);
>> +	}
>> +	return ret;
>> +}
>> +
>> +static void test_tracefs_function_filter(void)
>> +{
>> +	const char *filter[] = {"run_init_process", "ufs*", "^ext4.*$", NULL};
> I ran this and it failed, because my box has no functions that start with
> "ufs" nor "ext4" (I use an xfs file system).
>
> You need to use core function names like "sched*" or "irq*", because ufs
> and ext4 are modules that do not exist in all machines that this may be
> executed on.
>
>> +	const char *future_filter[] = {"write_dummy", "read_dummy", NULL};
>> +	const char *future_module = "dummy";
>> +	struct tracefs_instance *instance;
>> +	const char *module = "btrtl";
> I also do not have the btrtl module. You may need to see what modules are
> loaded (reading /proc/modules will help you there).
>
>> +	int ret;
>> +
>> +	instance = tracefs_instance_create(TEST_INSTANCE_NAME);
>> +	CU_TEST(instance != NULL);
>> +
>> +	ret = tracefs_function_filter(instance, NULL, NULL,
>> +				      TRACEFS_FL_RESET | TRACEFS_FL_CONTINUE);
>> +	if (ret) {
>> +		printf("Failed to reset the filter\n");
> Should the above be a CU_TEST failure?
> +		if (ret < 0)
> +			tracefs_function_filter(instance, NULL, NULL, 0);
> +	}
> +	/* Test string, kernel glob and regex for specific_instance*/
> +	ret = test_instance_filter(instance, filter, NULL, TRACEFS_FL_CONTINUE);
> +	CU_ASSERT(ret == 0);
> +	/* Test Module only with no filters*/
> +	ret = test_instance_filter(instance, NULL, module, TRACEFS_FL_CONTINUE);
> +	CU_ASSERT(ret == 0);
> +	ret = test_instance_filter(instance, future_filter, future_module,
> +				   TRACEFS_FL_CONTINUE | TRACEFS_FL_FUTURE);
> +	CU_ASSERT(ret == 0);
> +	tracefs_function_filter(instance, NULL, NULL, 0);
> Should the above result also be tested?

Is it really needs to be tested?, because anyhow if we have any error, 
it will be thrown at the above lines

But anyhow, i will test the result here, since we are commiting the 
changes in the file.

>> +
>> +	ret = tracefs_function_filter(NULL, NULL, NULL,
>> +				      TRACEFS_FL_RESET | TRACEFS_FL_CONTINUE);
>> +	if (ret) {
> And here too.
okay.
>> +		printf("Failed to reset the filter\n");
>> +		if (ret < 0)
>> +			tracefs_function_filter(instance, NULL, NULL, 0);
>> +	}
>> +	/* Test top instance*/
>> +	ret = test_instance_filter(NULL, filter, NULL, TRACEFS_FL_CONTINUE);
>> +	CU_ASSERT(ret == 0);
>> +	ret = test_instance_filter(NULL, NULL, module, TRACEFS_FL_CONTINUE);
> As I stated above, I don't have the module you picked, and will most
> definitely fail this test as FUTURE is not set.

Yeah will try to add some standard modules.

Thanks,

sameer.