PPS not synching on imx6ul devices

[Kensou]

I have two Variscite boards, each consisting of a module and a carrier- A VAR-SOM-6UL with Symphony Board- A VAR-DART-6UL with Customboard6UL
Both are equipped with NXP imx6ul.

I have created two Yocto images by modifying the device tree so that the EVENT_OUT signals of the ptp "come out": for the VAR-SOM channel 2 comes out, while for the VAR-DART channel 0 comes out. In both I have modified the driver to have a square-wave PPS instead of a pulse PPS, as per pdf nxp: https://community.nxp.com/pwmxy87654/attachments/pwmxy87654/imx-processors%40tkb/5709/1/i.MX%20IEEE1...
The system starts up correctly, and by configuring the two Ethernet interfaces and starting the pps with:
echo 1 > /sys/class/ptp/ptp0/pps_enable

The square waves of the two boards start up correctly on EVENT_OUT (EVENT0_OUT on VAR-DART and EVENT2_OUT on VAR-SOM, both INET1) and I detect them with the oscilloscope.
Now, our goal is for the waveforms of the two boards to be aligned with each other, and to do this we synchronise the ptp's of the two boards:

ptp4l -A -4 -H -m -i eth0
ptp4l -A -4 -H -m -i eth -s

One master and one slave.
From what I can see, however, the two waveforms do not deviate. In other words, they maintain the mutual deviation from the moment they start with echo 1 > /sys/class/ptp/ptp0/pps_enable

Now I'm wondering: am I missing any steps? Do I need to do anything else?

[Kensou]

Dear Sirs,

we saw in a NXP community topic that INETx_EVENTy_IN is not supported by your fec_ptp.c driver.

https://community.nxp.com/t5/i-MX-Processors/i-MX6s-dl-input-PPS-to-FEC-for-PTP-Grand-master-ENET-15...

we could circumvent the problem by configuring the period and time of the start of the PPS, i.e:

echo "0 $timestamp_ptp_start 0 1 0" > /sys/class/ptp/ptp0/period

In any case, we did not encounter the following problems:

1) Starting the ptp at the same time, there is no alignment, a phase is maintained between the two PPSs of the two boards. In theoria, with the above command with period 1 second and having 0 nanoseconds as the start time (timestamp_ptp_start seconds and 0 nanoseconds), they should align, which is not the case.

2) When any change takes place, such as ptp being switched off or restarted and the Ethernet cable being switched on or off, the PPS signal of both cards collapses, remaining at a stable low logic value, requiring both cards to be rebooted.

How can we solve this?

[jimmychan]

Do you mean the PPS signals do not align between master and slave?

[Kensou]

Exactly

[jimmychan]

Could you show me the waveform?