[PATCH i-g-t v2 5/5] tests/intel/xe_fault_injection: Do not assert if the captured error is not same as injected error.

[Bernatowicz, Marcin]

On 3/5/2025 12:53 PM, Francois Dugast wrote:
> On Wed, Feb 19, 2025 at 01:04:45PM +0530, Satyanarayana K V P wrote:
>> In the current implementation, test asserts if the captured error is not same as
>> injected error. It is a possible that the error received is translated to other
>> error which can be returned to application.
>>
>> Try to unbind the device at the end of inject_fault_probe() as this can be
>> executed in a loop if IGT_FAULT_INJECT_ITERATION environment variable is set.
>>
>> Signed-off-by: Satyanarayana K V P <satyanarayana.k.v.p at intel.com>
>> Cc: Michał Wajdeczko <michal.wajdeczko at intel.com>
>> Cc: Francois Dugast <francois.dugast at intel.com>
>> ---
>>   tests/intel/xe_fault_injection.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/tests/intel/xe_fault_injection.c b/tests/intel/xe_fault_injection.c
>> index cf0337733..82e338965 100644
>> --- a/tests/intel/xe_fault_injection.c
>> +++ b/tests/intel/xe_fault_injection.c
>> @@ -227,8 +227,8 @@ inject_fault_probe(int fd, char pci_slot[], const char function_name[])
>>   	injection_list_do(INJECTION_LIST_ADD, function_name);
>>   	set_retval(function_name, INJECT_ERRNO);
>>   	xe_sysfs_driver_do(fd, pci_slot, XE_SYSFS_DRIVER_TRY_BIND);
>> -	igt_assert_eq(-errno, INJECT_ERRNO);
>>   	injection_list_do(INJECTION_LIST_REMOVE, function_name);
>> +	xe_sysfs_driver_do(fd, pci_slot, XE_SYSFS_DRIVER_TRY_UNBIND);
> 
> Those should be 2 separate patches.
> 
> 1. Try driver unbind after fault injection during probe
> 2. Accept returned error different than injected one
> 
> I am fine with 1. as it is, just the one line.
> 
> However removing the assert in 2. is not acceptable. This assert *IS*
> the real test performed for existing fault injection tests. If we
> remove it then we will not be able to detect when probing does not
> behave as expected in case of failure.

@Francois, What is recommended on successful probe? Should we treat
successful probe as not an error ?
That happens when error injected function is not called at all (ex. 
xe_wopcm_init, xe_guc_log_init, xe_guc_ads_init on VF probe)

	xe_sysfs_driver_do(fd, pci_slot, XE_SYSFS_DRIVER_TRY_BIND);
	err = -errno;
	injection_list_do(INJECTION_LIST_REMOVE, function_name);
	if (err)
		igt_assert_eq(err, INJECT_ERRNO);
	else
		/* do additional checks like simple_workload or too complex for this 
test? */
		igt_kmod_unbind("xe", pci_slot);

--
marcin

> 
> We need a more flexible solution to your problem, such as adding one
> parameter to inject_fault_probe() with the expected returned error
> code.
> 
> Francois
> 
>>   }
>>   
>>   /**
>> -- 
>> 2.35.3
>>