05:19 <narmstrong> xdarklight: oh weird thx for the pointer !

07:54 <warpme_> narmstrong: may You advice me how to avoid MAC address change on amlogic hw (MAC seems to be changed at every reboot and also at DHCP renewal)?. i.e for rpi there is kernel patch allowing to set MAC in kernel cmdline. Is this problem somehow addressed in AmLogic?

10:05 <narmstrong> warpme_: you can set it in the uboot env and the kernel will use it

10:07 <warpme_> narmstrong: well - I was trying to set this in uEnv.ini. I was trying following: bootargs=root=/dev/mmcblk0p2 rw console=ttyAML0,115200n8 console=tty0 logo.nologo vt.cur_default=1 mac=:11:22:33:44:55:66”. but mac is still random….

10:09 <narmstrong> warpme_: not the cmdline, the u-boot env: setenv ethaddr <your mac adress>

10:10 <warpme_> can this be added in uEnv.ini or rather it must be added in u-boot script?

10:11 <narmstrong> uEnv is ok, it must be set before running the kernel

10:20 <warpme_> hmm - my uEnv.ini is following: dtb_name=/s912_dtb

10:20 <warpme_> setenv ethaddr 5e:ec:31:26:53:6a

10:20 <warpme_> bootargs=root=/dev/mmcblk0p2 rw console=ttyAML0,115200n8 console=tty0 logo.nologo vt.cur_default=1

10:20 <warpme_> and mac is still random…

10:26 <warpme_> also I tried to add setenv ethaddr 5e:ec:31:26:53:6a in s905_autoscript. still not works

10:46 <warpme_> ok it looks like I need to add this in aml_autoscript. now works. milion thx!

10:48 <warpme_> oh btw: I’m trying to prepare single kernel config for rpi3, s905, s912 and rk3328. So far it works for rpi3, s905 and rk3328. but s912 has non-working eth. What exactly needs to be enabled additionally to get eth on s912?

15:17 <xdarklight> narmstrong: it seems that your G12A USB changes will make it into v5.2! the patches are showing up in today's -next and Kernel CI already used them: https://storage.kernelci.org/next/master/next-20190503/arm64/defconfig/gcc-7/lab-baylibre-seattle/boot-meson-g12a-x96-max.html :)

15:20 <narmstrong> Nice ! I really thought they would miss this release...

16:56 <xdarklight> warpme_: make sure you have CONFIG_DWMAC_MESON and CONFIG_REALTEK_PHY enabled. if that's already the case then please pastebin your .config and a kernel boot log

17:04 <warpme_> xdarklight: both CONFIG_DWMAC_MESON and CONFIG_REALTEK_PHY are set as „=m”. This exact config works ok with s905. to my surprice it is not working with s912. Just thinking how to get kernel boot log without network access. let me try by mounting /boot part with rw and copy demesg out. config I’m using is here: https://github.com/warpme/minimyth2/blob/master/script/kernel/linux-5.0/files/linux-5.0-arm64-armv8.config

17:05 <xdarklight> CONFIG_MDIO_BUS_MUX_MMIOREG is also set, which is good

17:06 <xdarklight> maybe copy the boot log to an USB thumb drive or use wifi to connect to your network (if your board supports that)

17:10 <warpme_> xdarklight: indeed. good idea with usb drive :-). Give me pls sometime as I need to recompile to mainline (now playing with panfrost from Neil git).

18:38 <warpme_> xdarklight: pls find dmesg output: https://pastebin.com/XKcxzyTq

19:37 <xdarklight> warpme_: thanks. does your userspace (ifconfig or ip) see eth0?

19:37 <xdarklight> do you see high packet loss or what happens if you "ifconfig eth0 up" and configure an IP address?

19:42 <warpme_> xdarklight: yes. I see in userspace eth0. but starting dhcpc blocks terminal….

19:44 <warpme_> xdarklight: is it possible You will try build 5.0.11 mainline kernel with my config and see how it goes?

19:48 <xdarklight> warpme_: we can try that. maybe you can try one more thing: use CONFIG_MDIO_BUS_MUX=y and CONFIG_MDIO_BUS_MUX_MMIOREG=y instead of =m for both

19:48 <xdarklight> I remember Neil recommended this to someone a while ago and it fixed some networking issues

19:50 <warpme_> xdarklight: sure. now rebuilding.

19:51 <warpme_> iterestng is that I have 0 RX data. It looks like i.e. interrupts problem….

19:52 <warpme_> also I have high rate of „error: can’t attach to PHY”. 4 boots have 2..3 sych werrors....

20:51 <warpme_> xdarklight: ok just give try with CONFIG_MDIO_BUS_MUX=y and CONFIG_MDIO_BUS_MUX_MMIOREG=y but this not helps. btw: is it ok that attached PHY driver [RTL8211F Gigabit Ethernet] (mii_bus:phy_addr=0.2009087f:00, reports irq=33 while ifconfig for eth0 reports irq=23?

20:53 <xdarklight> warpme_: IRQ mismatch is OK because what you see in the PHY kernel log line is the PHY's interrupt while ifconfig sees the MAC (Ethernet controller) interrupt

20:54 <warpme_> ah ok. thx!

20:54 <xdarklight> so your board is a Khadas VIM2, are you using Amlogic's u-boot or mainline u-boot?

20:56 <warpme_> oh - no it is BeeLink GT1. I’m using build-in AmLogic u-boot. I have all working with Armbian kernel config - so issue is definitelly in my kernel config....

21:01 <xdarklight> what kernel version is Armbian using?

21:03 <warpme_> xdarklight: 5.0.7 I think - but this doesn’t matther. I can do simple experiment: just only replace my kernel config to armbian one and rebuild: all works…

21:03 <xdarklight> ah I see

21:06 <warpme_> xdarklight: you know - I spend days trying to part-by-part merging/replacing my config with fragments from armbian config. so far - no success. geez - this I think must be kind of sublte bug….

21:13 <xdarklight> there was some Odroid-C1 user who had issues with CONFIG_HZ_250, he disabled that and enabled CONFIG_HZ_100 instead and that fixed it for him

21:13 <xdarklight> maybe there's a CONFIG_HZ_* diff between your .config and armbian?

21:14 <warpme_> yes - they are different

21:19 <warpme_> ok - I just sync all timers entries + CONFIG_HZ to armbian and rebuilding. lets see…

22:35 <warpme_> xdarklight: well - playing with CONFIG_HZ not helps. I think my issue exposes bug in kernel. I have 3 hypothessis: 1\ there is kernel module not stated as required for s912 - but if not enabled in config - it causes required by s912 networking modules not working. 2\ RTL8211+asociated sw have some additional dependency on other modules - but this dependency isn’t properly expressed in Kconfig and not enabled in my config. 3\ there i

22:35 <warpme_> conflict with rockchip networking aspects related to common with s912 hw. Both: a1 rk3328 and gt1 s912 are using the same RTL8211 (IIRC). I recall had quite similiar problem with rk3328: network was not working in quite the same way. Only way to get it working was use armbian rochchip kernel config and add required for amlogic entries. This was ok for s905 but look not enough for s912. All shows for RTL8211+associates….

22:44 <warpme_> most probable for me seems to be 3\

22:56 <warpme_> I mean bug like: init_code_for_s912_if_amlogic_enabled_by_config; then init_code_for_rk3328_if_rockchip_enabled_by_config; Maybe should be: if(hw=s912) then init_for_s912 elseif if(hw=rk3328) then init_for_rk3328. In other words: runtime hw dependency instead of compile-time….