Question about TSN API operation

Autonomous Vehicles DRIVE AGX Orin DRIVE AGX Orin General
1

DRIVE OS Version: 6.0.6

Issue Description:

Hello NVIDIA Team.

We are testing TSN through TSN API Sample application.

We made nvether_sample_app using ether_api.c file.

But, we have several issue.

  1. avb settint does not operate normally.

    • tcinx does not chage from 0 to setting value

    image
    image681×127 50.3 KB

  2. GCL does not operate normally.
    #./nvether_sample_app eqos_0 fpe 0x0 1 //disable fpe
    #./nvether_sample_app eqos_0 est 1 0 0 0 10 0x101d5ff 0 1 0xffffff //close all gate

    After fpe disable and close all gate setting, we sent test packet. But, whole packet transmitted even though all gates are closed. And, hlbs_q count does not increse at all.

    image
    image259×308 41.7 KB

    Is there a way to check which part is the problem?

Regards.

6

And, is there any limation to use eqos_0 interface?

7

Dear @taesik.kim ,
Is there any dependency on DOS 6.0.6 release?
Let me check internally on this issue and get back to you..

8

Yes, we are using 6.0.6 version.

I have one more question.

In ehter_api.c file, “0 - default (strict priority for EQOS and EST for MGBE)” means eqos_0 does not support EST?

And, “Use the primary interface i.e eqos_0, mgbe2_0, etc. for configuration” means mgbe1_0 also does not support TSN feature in Sample Application | NVIDIA Docs page?

And, please explain alog=2 value throug gavb option. In guide page, there is no explain about value 2. (In case of , if we try to set algo 0, setting value 2 is read.

Regards.

9

Hello NVIDIA Team.

drive-linux/kernel/source/oss_src/nvidia/drivers/net/ethernet/nvidia/nvethernet/nvethernetrm/osi/core/mgbe_core.c (line 2693~2720)

Due to above code implementation, it appears that hlbs_q does not increase beyond 1 and EST is disabled. Is this the intended implementation? Why is EST being disabled?

Reagrds.

10

Dear @taesik.kim ,
Could you share details about your requirement to share commands and feasibility?

You have given BTR_offset as 10 sec, so this GCL switch happen after 10 sec, till that switch earlier GCL will be used. If you start packet transfer before new GCL active, you will packet transmitted as per earlier configuration. Also, You are setting ctr(nsec) = 0x101d5ff nsec where GCL entry (only 1) is for 0xffffff, which means all gate closed for 16,777,215ns and for 120,320ns all gate are opened**, if packets are able to transmit in 120,320ns, you will not hit HLBS.
If you use ./nvether_sample_app eqos_0 est 1 0 0 0 1 0x101d5ff 0 2 0xffffff 0x1d600, it allows all gates are closed in entire cycle time (i.e 0x101d5ff = 0xffffff + 0x1d600).

Yes. when there is HLBF/HLBS, as per design we don’t drop packets but disable EST to make data flow continue with a error message, to set GCL correctly

Could you share steps on how to repro locally.

12

Could you please provide any update for this topic?