Hi nv team,
I use our custom-made board because our hardware design differs from the devkit. Therefore, we have made some modifications to the device tree, as shown in the diagram below.
I can confirm that in this scenario, using the flash.sh script allows us to flash our board with the program. This is because when we enter recovery mode and run lsusb, we can see the NVIDIA device on host computer. In fact, we’ve been using this method for over a year.
Through forums, I found another link where someone had the same issue. With my custom system, I can’t see the acm0 device, but with the devkit system, I can see acm0.
Is it because of the modifications I made to the device tree that I can’t see the acm0 device, which in turn prevents me from using the l4t_initrd_flash.sh script?
If the answer to question 1 is yes, do I still have a way to flash my rootfs to an SSD?
For the timeout issue during flash, please check the serial console log.
It may be caused from the boot issue.
Please check with your HW team about if the debug UART port is available on your custom board.
For the devkit, debug Uart can be accessed through micro-USB port on the board with /dev/ttyACM0 recognized on the host.
For the custom board, you should have a custom board config and related changes according to your design.
I didn’t modify anything, and it flash to the ssd sucessfully. after that, I plug in the ssd on my custom board. after the system boot, I modify the /boot/extlinux/extlinux.conf.
- root=/dev/mmcblk0p1
+ root=/dev/nvme0n1p1
after a second reboot I can see the root is mount on /dev/nvme0n1p1
I don’t know if this is caused by an issue with my SSD read device. Or is it that the method of connecting the SSD to the host computer for flashing is inherently unstable?