st20 port statistics reporting of err_packets appears to be broken with redundancy

[cwhite102]

MTL 25.02.1

We're calling the function st20p_rx_get_port_stats() to fill in the st20_rx_port_status data structure.

When running with redundancy (primary and redundant ports), these calls seem to produce nonsense information.
Sometimes both ports show a lot of errors, sometimes one port shows many errors and not the other.

The stream is received fine using either port individually or with redundancy, so I highly doubt any of these error numbers are correct. No errors are reported without redundancy (from either port).
In addition, no corrupted or reconstructed frames are indicated.

Example of errors being reported mostly on the primary port..

I looked at the MTL code where err_packets is set, and it looked rather questionable...

(st_rx_video_session.c - rv_handle_mbuf)

The 'ret' code is being incremented by 'handler_ret' in a loop. I'm guessing to track any negative value? But then 'ret' is used in the if statement to count err_packets. I think the 'if' statement probably should be using handler_ret but I don't know this code and it isn't well commented.

However, this can't be the only issue, as the error counter is incrementing like crazy when it probably shouldn't be incrementing at all in this test.

[Magda-Pytel]

Hi,

Thank you for noticing this issue. Could you please check in debug mode which function is actually assigned to pkt_handler? During the initialization phase, different conditions may set this pointer to different handler functions.

You are correct that handler_ret should be validated instead of the ret variable, as the loop iterates and checks each specific case individually.

Could you please verify if PR#1191 resolves the issue for you?

[cwhite102]

Hi,
No, it looks like the issue still persists. I'm sure PR 1191 is necessary, but it isn't the only bug at play here.
As per the original bug report, errors are still being reported when they (most likely) shouldn't be.

Again, receiving either the primary or secondary feed alone has no errors. But in redundant mode, Rx errors are being constantly reported. (Sometimes on one interface, sometimes both)

My colleague Diana checked on which function is assigned to pkt_handler... she reported that it is: assigned to rv_handle_frame_pkt()

[Magda-Pytel]

Hi, thank you for your massage, I'll try to investigate it further. Could you please provide the configuration files that you used?

[cwhite102]

This is being seen in our application. st20p video Rx using a primary and redundant port. We use external frames, user pacing, user timestamping.
We can try and get it reproduced in the sample application if you need us to.

[Magda-Pytel]

Yes, it would be very helpful. You can post the results here or send to our support e-mail.

[DianaEs2]

Hi Magda, clarification is needed about stat_pkts_redundant_dropped that results in an increase of port_user_stats[s_port].err_packets.

The problem can be reproduced with RxTxApp without any modifications. All you need to do is enable redundancy on Video Rx. For your reference, here is my config file

local_tx_rx_1v.json

And the resulting log

July09_redundancy.log

[Magda-Pytel]

Thank you, I'll let you know about investigation findings soon.