[dpdk-dev] [PATCH v16 7/9] dma/skeleton: add test cases

fengchengwen fengchengwen at huawei.com
Thu Aug 26 11:30:47 CEST 2021


The skeleton_dma mainly focus on dma framework ut, as it currently functions.

Agree add more general testcase which run with diffenernt hardware drivers, these
testcases mainly used to test drivers (not dma framework). In this
way, the driver's selftest ops can be implemented very easily.


On 2021/8/23 22:03, Bruce Richardson wrote:
> On Mon, Aug 23, 2021 at 11:31:32AM +0800, Chengwen Feng wrote:
>> Patch introduces dmadev unit testcase for validation against the
>> skeleton dmadev PMD implementation.
>>
>> Test cases are added along with the skeleton driver implementation.
>> It can be enabled by using vdev argument to any DPDK binary:
>>
>> 	--vdev="dma_skeleton,selftest=1"
>>
>> In case 'selftest=1' is not provided, autotest doesn't execute the
>> test cases but the vdev is still available for application use.
>>
>> Signed-off-by: Chengwen Feng <fengchengwen at huawei.com>
>> ---
> 
> Having self-tests was useful for rawdev implementations as each rawdev was
> (potentially) a completely different device type so no common set of tests
> could really be written for them. However, for dmadev, we have a common set
> of functions and APIs, so I think that the tests provided here should be in
> a general test-case set.
> 
> We are preparing to upstream a test suite for DMA devices, based off the
> work we have in progress on our own drivers, and we'll look to include some
> of the tests from here in that for consistency.  Please wait for that set -
> hopefully appearing the next day or two - before doing additional work on
> this set, as our set includes quite a comprehensive set of functional tests
> in it, which would go well with the API tests here.
> 
> In terms of the skeleton dmadev itself, I'm not fully convinced of its
> usefulness, but it does allow unit testing of the APIs in the absense of
> dma device hardware. Whether that utility is worth the maintenance cost,
> though, I'm not sure.
> 
> Regards,
> /Bruce
> .
> 


More information about the dev mailing list