LS1028 eno0 QBV test when meet large time jump.

[xinliwang]

Hi ,community @yipingwang @khushbur 

I've been testing ls1028ardb board qbv function on eno0 port and I found a very strange thing want to discuss with you guys. I don't know if it is a bug or just in normal.

According to the reference menu, by testing the qbv function we don't have to do the ptp at first. On the other hand the qbv function and ptp are independent.

So, I have completed the following setup

cat > qbv5.txt << EOF

t0 11111111b 1000000 t1 00000000b 1000000

EOF

qbvset --device eno0 --entryfile qbv5.txt

/home/root/samples/pktgen/pktgen_twoqueue.sh -i eno0 -q 3 -n 0

#The stream would get about half line rate.

As I have done,  I didn't do the ptp synchronization in the process of qbv test, so the qbv would run according to the default time which is 1970.0.0.0... 

Here I have another board which we called device B as a Master clock device.

When I sync ls1028 eno0 port with device B(master clock device) which ptp time is 7/24/2023, the eno0 port lose connection with device B and its qbv would not work any more.

I suppose the qbv function appeals to not working when meeting a large time jump(from 1970 to 2023).

It seems not normal for the qbv. Please tell me if I'm wrong or it is a bug 

Thank you 

[yipingwang]

Which SDK are you using? Could you please try the latest RTEdge release and see if the issue still exist?  https://www.nxp.com/design/software/development-software/real-time-edge-software:REALTIME-EDGE-SOFTW...

[xinliwang]

Hi @yipingwang 

I'm using the latest one yocto v2.5.

[yipingwang]

I can reproduce your issue, will investigate it.

"According to the reference menu, by testing the qbv function we don't have to do the ptp at first. On the other hand the qbv function and ptp are independent."

Above statement is NOT correct. Actually the base time of gate control list is derived from PTP clock, regardless of whether the PTP clock is synchronize or not. I suspect the schedule of the gate control list also depends on the PTP clock, when there is a significant step adjust of the PTP clock, the schedule will be affected.

[xinliwang]

Hi @yipingwang 

Thank for your reply.

For your statement, I think you are correct that the schedule will be affected. But I think the effect would not last long but a instant effect. On the contrary, the ls1028 eno0 port can not recover the impact which resulted in losing connection.

Anyway, I'm looking forward to your reply and maybe new patch.

Thank you

[yipingwang]

Confirmed with R&D team this is a software bug, I have created a ticket for R&D team to address this issue and will provide you the fix patch once it is ready.

[yipingwang]

Please try the attached patch.

[xinliwang]

One more question, can I apply this patch on linux 6.1.22 (real-time-edge v2.6)? 

[xinliwang]

Hi @yipingwang 

Than you and can you give me the link of the patch thread? I want to check the code.

Thank you