Jetson Orion NX board fails to boot after kernel update: “OTA work directory is not found” in NVMe boot mode

[subramaniyan.m]

Hello NVIDIA Team,

We are using a customized Orion NX carrier board with Jetson Orin module, and our system is configured to boot from NVMe.

Previously, the board was booting properly from NVMe without any issue. However, after updating the kernel, the device is no longer booting correctly. During boot, it appears to enter recovery mode and searches for the OTA work directory. The boot process fails because the OTA work directory is not found on either internal or external storage.

We are using NVMe boot mode, and the NVMe device is detected during boot. The log shows that /dev/nvme0n1p1 is found and mounted successfully, but /mnt/ota_work is missing.

Important log lines:

Linux version 5.15.185-rt-tegra
Kernel command line: root=/dev/initrd rw rootwait ...
Root device found: root=/dev/initrd
Mount initrd as rootfs and enter recovery mode

nvme0n1: p1 p2 p3 p4 p5 p6 p7 p8 p9 p10 p11 p12 p13 p14 p15

Looking for OTA work directory on the device(s): /dev/nvme0n1p1
mount /dev/nvme0n1p1 /mnt
EXT4-fs (nvme0n1p1): mounted filesystem with ordered data mode
OTA work directory /mnt/ota_work is not found on /dev/nvme0n1p1

Internal storage device(/dev/mmcblk0p1) does not exist
OTA work directory is not found on internal and external storage devices

From the log, it looks like the board is not continuing normal NVMe rootfs boot after the kernel update. Instead, it is booting with root=/dev/initrd and entering recovery/OTA-related flow. Since /mnt/ota_work is not present on the NVMe partition, the boot process stops.

Could you please help us understand:

1. Why is the board entering initrd/recovery/OTA mode after the kernel update?

2. Is this caused by an incorrect kernel update, bootloader configuration, extlinux.conf, initrd, or OTA metadata?

3. For NVMe boot mode, should the kernel command line point directly to the NVMe rootfs instead of root=/dev/initrd?

4. What is the correct recovery procedure to make the board boot normally again from NVMe?

Current status:

• Customized Orion X board

• NVMe boot mode

• Boot was working before kernel update

• Issue started only after kernel update

• NVMe is detected during boot

• /dev/nvme0n1p1 mounts successfully

• /mnt/ota_work is missing

• Internal eMMC device /dev/mmcblk0p1 does not exist

• Boot stops in recovery/initrd shell

Please let us know the recommended steps to recover the device without losing the existing NVMe rootfs, if possible.

I have attached the full serial boot log for reference.

kernel.log (129.8 KB)

[JerryChang]

hello subramaniyan.m,

since it’s a customize board, did you update l4t_generate_ota_package.sh and ota_board_specs.conf for creating customize board OTA payloads?
may I know what’s your commands to generate OTA payload? such failure usually due to an incorrect rootdev settings.