Jetson Thor unexpectedly reboots

Robotics & Edge Computing Jetson Systems Jetson Thor
1

Hi,

We are using a Jetson AGX Thor Developer Kit, and we are seeing intermittent unexpected reboots/shutdowns.

The issue does not seem to depend on workload. It can happen while the system is under heavy CPU/GPU/ROS workload, but it can also happen when the system is almost idle with no major processes running.

The system does not appear to perform a clean shutdown. The previous boot journal ends abruptly during normal application logs.

According to journalctl --list-boots, the previous boot ended unexpectedly, and the next boot started later. The last lines of the previous boot only show normal backend HTTP 200 responses such as:

[backend] INFO: "GET /status HTTP/1.1" 200 OK
[backend] INFO: "GET /timesync/diff HTTP/1.1" 200 OK
gpg-agent: can't connect to the daemon /usr/lib/gnupg/scdaemon

There is no clean shutdown sequence near the end of the previous boot. I do not see messages such as systemd stopping services, reboot target reached, poweroff, or a normal shutdown sequence.

I also checked the previous boot kernel journal:

journalctl -k -b -1 --no-pager -o short-monotonic

but it returned:

-- No entries --

So I could not find any previous-boot kernel panic, watchdog, thermal shutdown, or USB over-current messages from journalctl -k -b -1.

I then checked the current boot dmesg. There are some important messages:

EXT4-fs (nvme0n1p1): recovery complete

and:

systemd-journald[281]: File /var/log/journal/.../system.journal corrupted or uncleanly shut down, renaming and replacing.

This suggests that the previous shutdown was not clean.

I also checked pstore/ramoops. dmesg shows that ramoops is enabled:

pstore: Registered ramoops as persistent store backend
ramoops: using 0x200000@0x1fbf880000, ecc: 0

However, /sys/fs/pstore is empty after the reboot:

sudo ls -al /sys/fs/pstore
total 0
drwxr-x--- 2 root root 0 Nov 26  2025 .
drwxr-xr-x 9 root root 0 Jan  1  1970 ..

So I could not find any persistent kernel panic/oops log.

I also checked the PMC reset information:

cat /sys/devices/platform/bus@0/c800000.pmc/reset_reason
UNKNOWN

cat /sys/devices/platform/bus@0/c800000.pmc/reset_level
WARM

So it looks like the system experienced a warm reset rather than a clean software shutdown or a full cold power cycle. However, the reset reason is reported as UNKNOWN.

I also grepped dmesg for power/thermal/watchdog/reset related messages:

sudo dmesg -T | grep -Ei "thermal|throttle|over|voltage|power|pmic|brown|watchdog|wdt|shutdown|reset|panic|oops|error|fail"

The relevant lines include:

EXT4-fs (nvme0n1p1): recovery complete
systemd[1]: Using hardware watchdog 'NVIDIA Tegra186 WDT', version 0, device /dev/watchdog0
systemd[1]: Watchdog running with a timeout of 2min.
pstore: Registered ramoops as persistent store backend
ramoops: using 0x200000@0x1fbf880000, ecc: 0

I do not see a clear thermal shutdown, kernel panic, explicit watchdog reset, undervoltage, brownout, or PMIC shutdown message.

There are also some NVIDIA driver warnings/errors during boot, for example:

NVRM: devm_reset_control_get failed, err: -2
WARNING: CPU: 2 PID: 1402 at drivers/reset/core.c:766 __reset_control_get_internal+0x68/0x16c
NVRM: rpcRmApiControl_dce: NVRM_RPC_DCE: Failed RM ctrl call cmd:0x731341 result 0xffff: Failure: Generic Error [NV_ERR_GENERIC]

These happen during the current boot, so I am not sure if they are related to the unexpected reset.

Summary of what we found:

  • The previous boot journal ends abruptly during normal application HTTP 200 logs.

  • There is no clean shutdown sequence.

  • journalctl -k -b -1 has no previous boot kernel entries.

  • /sys/fs/pstore is empty.

  • dmesg shows EXT4 recovery on the root filesystem.

  • dmesg shows that the previous system journal was corrupted or uncleanly shut down.

  • PMC reset information shows:

    reset_reason: UNKNOWN
reset_level: WARM

  • I do not see clear evidence of kernel panic, thermal shutdown, explicit watchdog reset, undervoltage, brownout, or PMIC shutdown in the logs.

Questions:

  1. What does reset_reason: UNKNOWN with reset_level: WARM usually indicate on Jetson AGX Thor?
  2. Is there any additional Jetson-specific log, register, or debugfs/sysfs entry that can explain an UNKNOWN/WARM reset?
  3. Could this indicate a watchdog reset, PMIC/power rail glitch, carrier-board-level reset, or firmware-level reset?
  4. Since pstore is empty, is there any other way to capture the cause of a sudden warm reset on Jetson Thor?
  5. Are the NVRM reset_control warnings or DCE RM control failure during boot known issues, or are they unrelated to the unexpected reset?

We can attach the following logs if needed:

  • previous boot journal tail
  • current dmesg
  • dmesg power/reset/thermal filter output
  • journalctl boot list
  • PMC reset_reason/reset_level output

Thanks.
current_dmesg.txt (148.1 KB)
prev_boot_monotonic_end_300.txt (33.0 KB)
dmesg_power_filter.txt (18.2 KB)

and our power >>

IMG2772
IMG27721920×2560 704 KB

4

There is no update from you for a period, assuming this is not an issue anymore.
Hence, we are closing this topic. If need further support, please open a new one.
Thanks ~0708

Could you use the serial console log from UART to check the board? syslog or other text file on your rootfs may not capture the reason of the sudden reboot.