11:37 <diederik> I get "r8169 0000:01:00.0: can't read MAC address, setting random one" on my NanoPi R5S. Where should I look for a solution: kernel or u-boot?

13:31 <robmur01> diederik: depends what solution you want - I guess the NIC doesn't have an EEPROM to store its MAC, so if you need a consistent MAC at boot time then you probably want U-Boot to program it, otherwise you can always reconfigure it from userspace post-boot

13:34 <diederik> ok, thanks :) I had *assumed* friendlywrt would not have this problem and wanted to know in which repo to look deeper. Turns out, I was partly wrong. Instead of a warning on both (2.5GbE) NICs, it 'only' had it for 1 (which is 1 more then expected)

14:09 <robmur01> indeed the R5S schematic doesn't show any EEPROMs, so I guess U-Boot is probably poking at some of the network interfaces already, just not all of them?

14:45 <diederik> I guess, but that's something I need to investigate

15:29 <macromorgan> diederik: I think U-Boot has the provision to generate unique MAC addresses based on some value from EFUSE or OTP... that said, I never quite figured out how to describe the PCIE device in devicetree so I could pass it a mac address. I don't think the driver itself supports a MAC address kernel module argument either.

15:30 <macromorgan> I have the same issue with the Indiedroid Nova (rk3588s based) but haven't devoted effort to a solution just yet. I'm currently looking at high suspend power usage for my rk3566 devices...

15:31 <diederik> macromorgan: yeah, I did look into how it was done/fixed for rk3328-rock64 and there EFUSE was enabled which I guess made for unique (and stable) addresses

15:32 <macromorgan> right... I just didn't settle on a good way to pass that value to the hardware

15:33 <diederik> I have some vague memory that one could define it in DT ('somewhere'), but then you'd have to create a separate DT for each device or sth like that

15:37 <diederik> I don't plan to invest a big amount of effort to it as (at least for the time being) I have no use for those 2.5GbE ports. (I bought it for NVMe storage and the case)

15:38 <diederik> But while I have the board, might as well see if I can improve some things upstream :)

15:38 <diederik> s/while/now/

15:40 <qschulz> diederik: you can patch the device-tree before passing it to Linux to set the mac-address

15:41 <diederik> good to know, thanks!

15:41 <robmur01> here's one example of using the PCI DT bindings for this: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/freescale/imx8mm-venice-gw72xx.dtsi#n160

15:41 <qschulz> i remember that there are multiple ways to pass mac addresses to the kernel and most of them are deprecated IIRC

17:39 <CounterPillow> Since my RK3576 HWRNG submission only featured one meme in the cover letter, here's a second one https://overviewer.org/~pillow/up/8ee58e255d/rockchip-hwrng.png