Adding OpenWrt support for Xiaomi AX3600 (Part 1)

For Developers
1063

Should it print some kind of return code or panic?
It prints absolutely nothing.

1 Like
1064

they should be module right? rmmod and insmod

1065

Yeah, I tried also rmmoding and insmoding it, nothing is printed.
Quite literaly like its not triggered of compatible.

Could it be this stupid?

/*
 * nss_dp_init()
 */
int __init nss_dp_init(void)
{
	int ret;

	/*
	 * Bail out on not supported platform
	 * TODO: Handle this properly with SoC ops
	 */
	if (!of_machine_is_compatible("qcom,ipq807x") &&
			!of_machine_is_compatible("qcom,ipq8074") &&
			!of_machine_is_compatible("qcom,ipq6018") &&
			!of_machine_is_compatible("qcom,ipq5018"))
		return 0;
1 Like
1066

totally that AHAHAH (one of the reason NOT to use the qsdk code...)

anyway about the flags defining the Compile skip the autoconf part that's why the config file was ignored

1067

And yeah, just added the "qcom,ipq8074" as second compatible and now it works.
Like seriously, why?

BTW, it works.
I get DHCP just fine.

4 Likes
1068

ow wow... could be that the problem I had ages ago was about missing clock?

1069

Yes, because I had the same issue on IPQ60xx.
It would throw an error on setting clock but it will continue on, and then clocks for the port wont be UP and you will get no traffic.

1070

so you are telling me they fail on model compatible but they don't care if the clk is not setup correctly? WTF what a joke...

1071

Looks like it, they just have a warning message.
Compatible thing is just so stupid because it throws no error, but returns 0 like it suceed loading.

QCA at its finest.

1 Like
1072

if you want to have fun you can try to disable the swconfig part and use switchdev

1 Like
1073

swconfig is already disabled as that requires patching into OpenWrt, they just present interfaces as standalone ones.
I will try switchdev as it may compile with 5.10 after all, it fails on 5.12.
Enabled switchev in the kernel and it compiled, appears to work.

Tommorow I will work on frequency scaling and CPU regulators.
PCI is working like 50% of the time.

4 Likes
1074

what problem do you are having with pci ?

1075

It is working like 50% of attempts when ath10k goes probing and initializing, then 25% it will hang once ath10k tries to switch the card state and reboot, and last 25% it will switch the card but fail on everything else and quit starting.

It's potentially an issue with backporting as the patch for the 2.9.0 DWC PCI was picked from a series for IPQ6018(Its exactly the same PCI hw) is based on 5.13 where the DWC drivers were reworked so that core takes care of a lot of things instead of vendor drivers.
So I had to backport a lot of stuff which happened in 5.11 and 5.12.

2 Likes
1076

wonder if kernel 5.13 have the same problem... if you want I can try

1 Like
1077

Feel free, I just dont have the energy to do an 5.13 tree with buildroot

1078

I added SPMI regulators, APSS clock and finally frequency scaling for IPQ8071 as each model has different OPP table.

4 Likes
1079

Off topic but a little message of encouragement. I think many of us are waiting for a 100% OpenWRT port.

4 Likes
1080

main problem for now is using qsdk source is problematic and ath11k is still shit...

1 Like
1081

ath11k works fine on IPQ6018, but it's most likely due to the newer kernels as 5.12 and 5.13 have tons of fixes for it.

1 Like
1082

ow didn't know... so it's really only a big effort in backporting tons of patch...

Do you think I should test if pci with 5.13 works better to check if something is wrong with the backport?

2 Likes