Further stabilize ethernet of rockpi-4b

[fraz0815]

This pull 1736 was a good start for improving network stability.
snps,txpbl = <0x4>;

But in my case, or I guess with a typical german ISP with DS-lite (Native IPv6 + IPv4 tunneling through IPv6 - it is a nightmare in general). I did not use the board much because it was only usuable with
ethtool -K eth0 tx off rx off. Turns out the above pull request improved situation, but I was still unable to complete a simple apt update. After checking some other fixes around I also increased rx_delay = <0x20>; and all my problems are gone for the first time and I am actually using this board. The setting could possibly use some fine tuning, maybe 16 is already enough.

[chwe17]

please check:
#546
and/or
https://github.com/ayufan-rock64/linux-build/blob/02fae982767ca242009baaf995f7ce64b5d82469/recipes/gmac-delays-test/range-test

so that we come up with settings based on proper testing. @ThomasKaiser script might need some adjustments cause it seems we deal with much higher rx/tx delays as as we did on H64 for what it was written in the first time.

[fraz0815]

Okay thanks, interesting scripts. Will start testing tomorrow after work

[piter75]

You mention Rock Pi 4 but you actually modify the device tree of NanoPC/Pi family.
Does the change really fix the issue for you?

Would you mind verifying and providing similar patch for "-dev" while you are at it?

[piter75]

Could you please make the patch name consistent with the ones that are already there.
board-rockpi4-dts-rx-delay-tweak.patch would fit better what we already have IMHO ;-)

[piter75]

I see that you have opened another PR with eMMC fixes for NanoPC T4... Maybe this one should also addresses issues with NanoPC T4.

If so the patch name should be corrected and the change should be addressed at board's level device tree similarly to this one.

[fraz0815]

I own both boards, nanopct4 and rockpi-4b, and both have some quirks (nanopct4 -> emmc, rockpi-4b -> broken ethernet packets) out of the box.

I just messed the patch totally up, should have gone sleeping, wtf.

Please excuse me, and let me doublecheck what happened here and why I picked that one and why it actually helped :-/

[fraz0815]

I am closing this one, as the patch was messed up (wrong targeted dtsi).
I am starting to understand the test scripts mentionend and making the needed changes.

This will take some time to test and interpret the results.
Data to discuss will follow in a few days.