IND-COMMS-SDK: Profinet RT Test Issues

Hi Philip,

We have tested the ART test for AM64x and have not observed these issues. However we will review the shared logs and also run these tests for multiple iterations and provide an update.

Could you provide clarifications to few queries?

1) Have you made any changes in the application for the tests except for the patch for race condition on remanent IM callback?

2) Are all tests failing for every iteration?

Regards,
Laxman

Hi Laxman,

thank you for your reply.

Regarding your questions:

1) With changes to the application, do you mean the demo app? We are building our own application which is based off of the demo. However, in our usecase the fieldbus application is mainly implemented on the A53 Linux side where the stack interface is mostly handed through RpMsg to the application. Also the handling of remanent data is handled on Linux side. The reason for this is that we already had an generic profinet application which is now adapted to the Sitara.

2) I ran them several times and they look consistend. But one or two times the results of the mentioned case did change. I'm suspecting this might be because another prior test. 

Can you tell me if the values expeceted in the tests should be stored in the remanent PdevRecord?

Regards,
Philip 

Hi Philip,

Thanks for your message. We will take a closer look at this issue and the logs you provided and give you response as soon as possible.

Kind regards,
Kamil

Hi Kamil,

thank you! Just a quick note that the other fails in some of the logs regarding I&M encoding and config were unrelated and fixed by now. The described issues are unaffected by these.

Regards
Philip

Hi Kamil,

if there is any additional info I could provide to support with this, just let me know.

Kind Regards
Philip

Hi Philip,

this issue is being handled offline. If it isn't solved in Monday's delivery, we can discuss it further here.

Thanks.
Kind regards,
Kamil

Hi Kamil,

we did not receive a delivery yet. Did the timeline change?

Thank you and kind regards
Philip

Hi Philip,

please check your email. A link to the release is provided there.

Thank you.
Kind regards,
Kamil

Hi Kamil,

got it, thank you! I rerun the tests in question using the new version and see some improvements.

Pdev_Check_onePort - Scenario 1: Still same error
Pdev_Check_onePort - Scenario 2: Now Pass
Diagnosis - Scenario 1: Only Check LineDelay error in Step 9 remains
Different Access Ways: Still same error

Reports.zip

Kind Regards
Philip

As mentioned in the other post (regarding TEDCheck) I noticed that I had rerun the tests with TestBundle 2.44 instead of 2.45. 
With TestBundle 2.45 There are now 2 fails remaining:

- Pdev_Check_OnePort Scenario 2

- Diagnosis Scenario 1

Diagnosis Scenario 1 seems to be inconsistend where it sometimes fails and sometime passes, maybe hinting at a timing issue. With the Pdev_Check I also observed inconsistend results before.

In our application we have two additional tasks that handle the RpMsg Communication with the Linux host. Could there be an issue with priorities of these interfering with the stack? I played around with the prio values for them until I got something that seemed to work for both, RpMsg and Profinet, but maybe there is some info on what prios to choose or avoid?

7674.Reports.zip

Hi Philip,

we're currently adjusting some task priorities in our stack and this might have a direct impact (hopefully positive) on your application and tests. Let's discuss this again in a couple of days when we're done. However, it's still interesting to me to know what priorities are you giving to your application tasks?

Thanks.
Kind regards,
Kamil 

Hi Kamil,

for our application we create 3 tasks:

- MainTask, Prio 22, used for stack init, ends in endless loop with sleep(1000)
- AcyclicTask, Prio 24, used for exchange of acyclic RpMsg data between R5 and A53
- CyclicTask, Prio 25, used for exchange of the cyclic input/output data buffers

For testing I also had the CyclicTask at prio 29, seeing the same behavior.

Kind regards
Philip

Hi Philip,

We are currently using the following priorities.

- Main task at priority 22

- Cyclic Task at priority 29, but we are running few tests with priority 28 to check the impact.

Currently the Profinet stack expert is out of office, kindly expect a delay in response to further queries

Regards,
Laxman

Hi Philip,

we're done with the task priority adjustments and tests. Everything will be included in our next release in a few days. Is there any updates on your side? Have you managed to resolve this issue already?

Thank you.
Kind regards,
Kamil