Hi NVIDIA,
Enabling the real-time patch in the kernel is likely to cause the terminal to freeze. I also replicated this issue by flashing the firmware of our custom carrier board onto the Thor development board.The suspicion is that it is a defect of the GPU hardware.
As shown in the log, there is a high probability that it will get stuck at this point of console switching (switch to display console).
Agx_thor_dev_board_with_my_kernel.txt (51.1 KB)
After encountering a problem, whether it’s a reboot or a power-off restart, there is a high probability that the issue will still persist and the terminal will get stuck again.
As long as the system is properly guided once, subsequent restarts will be successful continuously.
So I think it’s a hardware issue. Please help confirm this.
The log doesn’t say much for the specific issue (the only early on error might not actually be an error so much as early boot code notation before device setup; it could be the error, it just isn’t conclusive).
Does the same kernel and the (otherwise) default device tree do this on a dev kit carrier board? The customization of the carrier board and the device tree have to properly match, and when going back to the dev kit carrier board the device tree also has to be for that carrier board. Much of what the device tree does is to map drivers and physical addresses to specific hardware; any mistake could mean attempting to use a physical address (in kernel space) which is non-recoverable. The device tree not correctly mapping memory regions to devices and drivers is high in the probability of causing this. perhaps it is the GPU, but it might be the physical address mapping and not the actual hardware causing locking up.
Hi,
Please refer to Jetson AGX Thor FAQ
