08:30 <naoki> Kwiboo: I trid rk3582 patch with RockyLinux 9.5 Desktop userland. interestingly gdm wayland session doesn't work ;)

08:30 <naoki> I thought it works with software rendering

08:30 <naoki> X session works

08:45 <naoki> Nov 16 00:00:07 rocky-desktop gnome-shell[699]: Adding device '/dev/dri/card0' (rockchip) using atomic mode setting.

08:45 <naoki> Nov 16 00:00:08 rocky-desktop gnome-shell[699]: Failed to initialize accelerated iGPU/dGPU framebuffer sharing: Do not want to use software renderer (llvmpipe (LLVM 18.1.8, 128 bits)), falling back to CPU copy path

08:45 <naoki> Nov 16 00:00:08 rocky-desktop gnome-shell[699]: Created gbm renderer for '/dev/dri/card0'

08:46 <naoki> Nov 16 00:00:10 rocky-desktop gnome-shell[699]: meta_drm_buffer_gbm_new_lock_front failed: gbm_surface_lock_front_buffer failed

08:47 <naoki> I don't understand what happened ;)

09:08 <warpme> guys: i'm a bit confused - so some help for better understanding is needed: I have nanopc-t6. It has 2xRTL8125 nic by pcie. My issue is random MAC on those nics. I have also rock5b with the same nic@pcie and this box has always the same mac. Now: I was thinking mainline kernel for 8125 gets mac from eeprom (or otp) so boards with this nic have by default unique&constant mac. But my nanopc-t6 case shows differently. How this?

11:08 <sigmaris> warpme: I also have random MAC addresses with the RTL8125s on NanoPC-T6. I don't think the PCIe devices actually have any EEPROM or OTP to store a MAC address

11:09 <sigmaris> IIRC the kernel driver logs some messages about the MAC address during boot

11:11 <warpme> sigmaris : indeed - hypothesis that: t6 has no eeprom/otp programmed (and rock5b has eeprom/otp programmed) is my most probable one.

11:13 <sigmaris> last time I looked, there were two possible solutions - have userspace set a persistent MAC address when bringing up the interface, or have firmware (u-boot) derive a MAC address from the CPU serial number and add it to the devicetree

11:14 <warpme> as i'm developing stateless appliance (no systemd, etc) where state is restored from remote server - mac is used as unique id of device.....

11:15 <warpme> still think what will be best solution here....

11:15 <dlg> can you use the cpu serial instead?

11:16 <warpme> option with uboot will be best one as appliance is(can be) diskless so i'm doing network boot.

11:20 <warpme> on boards with soc build-in emac/gmac usually alias in uboot dt helps. but t6 has nics on pcie - so i'm not sure is alias solution possible/will work at all....

11:22 <sigmaris> there should be a way to create DT nodes for the PCIe devices and add MAC address to them, in u-boot's DT fixup stage

11:22 <sigmaris> Like this example for the wifi device on a chromebook: https://github.com/torvalds/linux/blob/master/arch/arm64/boot/dts/rockchip/rk3399-gru-chromebook.dtsi#L509

11:23 <sigmaris> I'm just not sure what are the right values for `reg` for the PCie devs, I couldn't quite follow the bindings

22:42 <naoki> Unable to handle kernel NULL pointer dereference at virtual address 0000000000000028

22:42 <naoki> pc : rk_hdptx_phy_runtime_resume+0x18/0x84 [phy_rockchip_samsung_hdptx]

22:42 <naoki> sometimes I got this oops during boot