Xavier NX eth0 won't come up after commanded reboot

[jbalme]

Hi there,
I’m seeing an interesting issue where the network interface on my Xavier NX won’t come up after I issue a sudo shutdown -r now command. Everything works fine if I do a full power-cycle, but if I reboot via the shutdown command, this is what I see in dmesg afterwards:

[   54.297201] [eqos_poll_for_swr][598][type:0x4][loga-0x0] poll_for_swr: timeout
[   54.297473] net eth0: failed to poll MAC Software reset

When I log in via the serial console and try to manually bring eth0 up I get the exact same errors:
$ sudo ifconfig eth0 up
[ 961.703194] [eqos_poll_for_swr][598][type:0x4][loga-0x0] poll_for_swr: timeout
[ 961.703393] net eth0: failed to poll MAC Software reset
ifconfig: SIOCSIFFLAGS: Operation not permitted

I’m using Jetpack 5.0.2. Any advice you have on how to further troubleshoot this would be much appreciated!

[KevinFFF]

Hi jbalme,

Are you using the devkit or custom board?

@jbalme

ifconfig: SIOCSIFFLAGS: Operation not permitted

Could you help to try entering super user mode first and bring eth0 up later?

$ sudo su -
# ifconfig eth0 up 

Please also help to provide the serial console log before you reboot with shutdown command until running command to bring eth0 up after boot up.

[jbalme]

I’m using a custom board. I tried entering super user mode and got the same result:

$ sudo su -
# ifconfig eth0 up
[ 1313.864459] [eqos_poll_for_swr][598][type:0x4][loga-0x0] poll_for_swr: timeout
[ 1313.864669] net eth0: failed to poll MAC Software reset
ifconfig: SIOCSIFFLAGS: Operation not permitted

I’ve attached two dmesg logs - the first one (dmesg_before_commanded_reboot.txt) is after a power cycle, when everything comes up cleanly. Then I did a sudo shutdown -r now to reproduce the bad state. The second log (dmesg_after_commanded_reboot.txt) contains the dmesg output after that commanded reboot, when the network interface fails to come up.
dmesg_before_commanded_reboot.txt (51.5 KB)

[jbalme]

Sorry, forum rules only allowed me one attachment on that post because I’m new :) Here’s the second file:
dmesg_after_commanded_reboot.txt (52.1 KB)

[KevinFFF]

After checking the logs you provided, I found the boot up process are almost the same and only one difference about if ethernet could Link Up.

I could not reproduce this issue on Xavier NX devkit with l4t-r35.1.
So, do you follow the adaptation guide for your custom board?
Jetson Module Adaptation and Bring-Up: Jetson Xavier NX Series — Jetson Linux
Developer Guide 34.1 documentation (nvidia.com)

I’ve also found two topics about the similar issue as yours.
Please refer them to make sure your RGMII configuration (pinmux, dts) is right for ethernet phy.
Does Orin support 88E1512 PHY chip? - Jetson & Embedded Systems / Jetson AGX Orin - NVIDIA Developer Forums
RGMII with switch can’t work - Jetson & Embedded Systems / Jetson AGX Orin - NVIDIA Developer Forums

[jbalme]

Yes, we did follow the adaptation guide (although we could always have made a mistake!). If it’s a pinmux issue, though, I would expect that the ethernet phy would not work at all, rather than working fine after power-cycle but failing to come up after a commanded reboot. Do you agree with that assessment?

[KevinFFF]

It’s not exactly.
Ethernet may work with generic ethernet phy driver even though not using specific vendor driver for that phy.
Do you make sure that ethernet phy and driver are matched?