SDK Manager Flashing Problems

[Aleks-CST]

Hi, I’m having some issues re-flashing my Jetson. Error messages from Terminal as follows:

09:36:38 SUMMARY: Drivers for Jetson - target_image: Install completed successfully.
09:44:15 ERROR: File System and OS - target_image: /etc/kernel/postinst.d/apt-auto-removal:
09:44:15 ERROR: File System and OS - target_image: qemu: Unsupported syscall: 276
09:44:21 SUMMARY: File System and OS - target_image: Install completed successfully.
09:44:36 ERROR: Can not find matching device in recovery mode.
09:44:54 ERROR: Flash Jetson AGX Xavier - flash: — Parsing board ID (ERROR: calculated CRC8 0xfd != stored CRC8 0x8e) succeeded.
09:44:54 ERROR: Flash Jetson AGX Xavier - flash: — Parsing board version (ERROR: calculated CRC8 0xfd != stored CRC8 0x8e) succeeded.
09:44:54 ERROR: Flash Jetson AGX Xavier - flash: — Parsing board SKU (ERROR: calculated CRC8 0xfd != stored CRC8 0x8e) succeeded.
09:44:54 ERROR: Flash Jetson AGX Xavier - flash: — Parsing board REV (ERROR: calculated CRC8 0xfd != stored CRC8 0x8e) succeeded.
09:44:54 ERROR: Flash Jetson AGX Xavier - flash: Error: Target board not found.
09:44:54 ERROR: Flash Jetson AGX Xavier - flash: command terminated with error
09:44:54 SUMMARY: Flash Jetson AGX Xavier - flash: First Error: Installation failed.
09:44:54 SUMMARY: NVIDIA Container Runtime with Docker integration (Beta) - target: Depends on failed component
09:44:54 SUMMARY: Multimedia API - target: Depends on failed component
09:44:54 SUMMARY: TensorRT on Target - target: Depends on failed component
09:44:54 SUMMARY: VPI on Target - target: Depends on failed component
09:44:54 SUMMARY: CUDA Toolkit for L4T - target: Depends on failed component
09:44:54 SUMMARY: cuDNN on Target - target: Depends on failed component
09:44:54 SUMMARY: OpenCV - target: Depends on failed component
09:44:54 SUMMARY: DateTime Target Setup - target: Depends on failed component

I’ve checked using i2cdetect/get and found that CRC8 is indeed 0x8e. Not sure what this means, and if it’s safe to set this to 0xfd instead. Any help appriciated!

[WayneWWW]

We have eeprom on the module and every time we flash the board, the flash script will read the eeprom content, calculate the crc8 checksum and compare it to the last field of eeprom.

Current issue on your side is somehow these bits do not match each other. If this module is not new one, it means somehow you use i2c tool to put some value to the eeprom and you didn’t update the checksum bit.

Please check the eeprom layout document and see if you got anything wrong there.

https://docs.nvidia.com/jetson/archives/l4t-archived/l4t-3271/index.html#page/Tegra%20Linux%20Driver%20Package%20Development%20Guide/jetson_eeprom_layout.html

If most of the field is correct, then you can just directly update the checksum bit to 0xfd.

[Aleks-CST]

Hi Wayne, thanks for your reply. The module is not a new one. I’ve flashed it a few times with different JetPack versions. I suspect I’ve deleted something in Ubuntu that I shouldn’t have. I’m new to Linux… Could it be there’s an old kernel that is now bugged and interrupting the flashing process somehow?

I have another module which I’m currently comparing the i2c dump with, and indeed there’s a few differences. I will try to change those (except MAC adresses) and try again.

[WayneWWW]

No, this issue has nothing to do with kernel version. Just pure eeprom content.

Our flash process won’t care about what is previous installed on the board. For example, if your board is totally software crashed, sdkamanger is still able to flash it.

[Aleks-CST]

Changed a few bytes according to another Jetson board. I left MAC addresses and the ‘manufacturing major revision’ byte unedited. Had to go through another flash which resulted in an error because of the CRC8. After changing that once more it worked. Thank you very much Wayne!