Jetson TX2 crashing due to Ethernet connection

Yes, by internal network, I just mean our office network environment.

Although, I feel I should mention, we have had some other issues lately: we can not connect to our own VPN while in the office (which might not be an issue, but I don’t know enough); we can not see our own website while in the office. Someone mentioned it might be due to the DNS. Again, I don’t know if these are related to the issue, and if they’re somehow messing with the TX2, but I felt like I should mention it anyway.

I can try, but will take multiple days, as I have no way to carry it home today. Is there anything else that you would recommend me trying in the meantime?

I need to discuss with some experts about what and how to check this issue.

Since this issue would crash the device, it is hard to “stop and check” some register or packets result.

I don’t know what to check either, but some observations:

  • The sk buffer is a general network data structure. This isn’t tied to a particular interface.
  • The skb errors do not point at a particular interface.
  • The included output from ifconfig shows activity on the native eth0, but no error of any kind.
  • No interface with activity showed any error.
  • Several interfaces had no activity, and thus they were also unlikely to be related.
  • Perhaps this is a data driven issue. Is there anything unusual about your network, e.g., special hardware?
  • Is there any custom software consuming network data?

After some more testing, our TX2i devkit crashes on a fresh flash of Jetpack 4.5.1, with the network switch removed out of the equation. On the other hand, the original TX2 that I was working on has not crashed after being connected to Wi-Fi, on the same network.

A coworker of mine will try connecting it to his home network, and see if any similar behaviour happens.

What seemed odd to me though, was that most of the crashes happened during office hours. One single crash happened recently at midnight, but usually none between 6pm local time and 9am.
As far as I’m aware, no special hardware than just our own work computers is connected to our network, and no custom software that would cause issues of this sort.

I am now trying to get logs from the TX2i, while it has had absolutely nothing else installed on it.

fresh.cap (119.0 KB)
Attached are minicom serial interface logs from the board crashing at least twice.

No crashes recorded on it in the timeframe when it was at home. There is nothing else that comes to mind to test for. Any luck with your experts? Someone recommended me to try wireshark, but I don’t know how to use it/don’t know what to look for, when analyzing packets.

Since rel-32.6 just released, can you also try with rel-32.6?

I would rather wait on that as that would then require reinstalling my entire setup, and it may introduce new issues with the actual project.

That does remind me to ask this: Which L4T release is currently being used? See “head -n 1 /etc/nv_tegra_release”.

R32 (release), REVISION: 5.1, GCID: 27362550, BOARD: t186ref, EABI: aarch64, DATE: Wed May 19 18:16:00 UTC 2021

Updating the OS on the Jetson TX2 through the SDK Manager has been attempted. The only issue is that the framos data package is not functional on it. I will get logs from the serial console soon, but it fails to boot now. The installed files linked and are the framos common driver, and the driver for the IMX530 sensor respectively.

Updating the OS resulted in the OS to not boot properly and the screen freezes. The last few lines are present in the following logs, being Created slice User Slice of mircea. \ Started Session 1 of user mircea. \ Starting User Manager for UID 1000... \ Started User Manager for UID 1000. There is one failed task present on the screen as well, throughout the many lines Failed to start nvpmodel service. But being logged in through serial console, it boots just fine, and even allowed me to log in to my user on the TX2. Although typing sudo shutdown now made it freeze entirely, and the screen was unchanged. It then required me holding down the power button to shutdown. After typing the shutdown command, these lines come up, and nothing happens past this point.

[  152.301046] tegra-xudc-new 3550000.xudc: exiting ELPG
[  152.302761] android_work: did not send uevent (0 0           (null))
[  152.317500] tegra-xudc-new 3550000.xudc: exiting ELPG done
[  152.323039] tegra-xudc-new 3550000.xudc: ep 0 disabled
[  152.328326] tegra-xudc-new 3550000.xudc: entering ELPG
[  152.339881] tegra-xudc-new 3550000.xudc: entering ELPG done
[  152.355576] device usb0 left promiscuous mode
[  152.359982] l4tbr0: port 2(usb0) entered disabled state
[  152.434744] device rndis0 left promiscuous mode
[  152.439320] l4tbr0: port 1(rndis0) entered disabled state

Holding down the power button does not have anything else show up on screen, nor on the serial logs. Attached are the serial logs.

update.cap (58.0 KB)

@WayneWWW, I have just updated to Jetpack 4.6, and the issue persists.

head -n 1 /etc/nv_tegra_release
R32 (release), REVISION: 6.1

The TX2i development kit has crashed two times since it has been flashed today.

Hi mircea.claudiu.01,

This error is similar to another issue, please check below patch if can help:

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.