I am having significant difficulties with my Jetson Xavier. After a period of sitting on the shelf due to other obligations, I set about working with my (early release) Xavier Dev kit.
When set up normally (attached to an HDMI monitor) the device seemed to fail to boot. I had gone through the process of installing at least one, maybe two Jetpack releases since I received it, and all seemed well as far as I went with things. I never got time to do any serious work however, but just explored the environment and perhaps some demos.
I realized that it was appearing with the original hostname on my network. So I tried using Putty to connect to it, and logged in without problem. Just the video configuration seemed to be messed up.
I decided to install the latest Jetpack (4.3) using SDK Manager (1.0.0.5517). That seemed to go OK, though I didn’t watch the terminal. It ended by saying “INSTALLATION COMPLETED SUCCESSFULLY”. It still would not come up graphically. Looking at the network, it was being handed an IP address by DHCP, but the hostname was now ‘localhost.localhost’ and it would not accept SSH login.
I repeated this entire process a couple more times and realized that I was consistently getting write errors during the early parts of the flash process. This does not seem to be noticed by the SDK manager.
To me it looks as if critical ranges of the eMMC flash have failed, probably taking my display configuration and who knows what else with them. Reinstalling the OS is not correcting the problem.
As SSH is not set up, I do not know how to get in to do more investigation. I gather that there is a serial port way in, but trying to get in through the /dev/ttyACM0 that appears on my host does not seem to work (I don’t know the port parameters to use).
Thanks for the info about serial debug ports and great documentation pages!
Unfortunately for me the four devices (/dev/ttyUSB0 etc.) appear when I plug in the USB2 cable but give me a frozen blank console using either minicom or Putty. This is exactly the behavior as with SSH as well. The connection is valid, but no prompt of any kind.
And no, at this point I do not even get the Nvidia Logo on the monitor. HDMI recognizes that it is active, but nothing gets displayed at all.
The first device in the list should be the correct device. Monitor “dmesg --follow” and look for the first name during plug-in. Settings should be “115200 8N1”. I recommend “gtkterm” (“sudo apt-get install gtkterm”). Using this, and assuming ttyUSB0:
sudo gtkterm -b 8 -t 1 -s 115200 -p /dev/ttyUSB0
Notice that unless your user is a member of group “dialout” that “sudo” will be required. If your user is a member of dialout, then you can exclude “sudo”.
Unfortunately while the port devices appear in /dev, connecting to any of them with the suggested settings gives a blank screen with no prompt. Same with gtkterm, minicom, putty. My account is a member of ‘dialout’ but using sudo yields the same result. The Xavier allows the connection, but whatever normally responds with a login prompt simply does not proceed.
I did exactly that twice today. I followed the directions for a manual flash from your documentation pages. Same results.
These are my ‘notes to self’ for this:
Manually flashing Xavier with L4T
===================================
Power down - unplug.
Connect monitor, kbd, mouse ETH0 to Xavier.
power on.
plug USB-C from Host into FRONT USB-C port.
press 'recover' button (Middle) and 'reset' (most towards center).
release 'reset', 2 seconds later release 'recover'.
Type:
------
cd ~/nvidia/nvidia_sdk/JetPack_4.3_Linux_P2888/Linux_for_Tegra
sudo ./flash.sh jetson-xavier mmcblk0p1 | tee flash_2020-01-15.out
Unplug USB-C cable
When prompted:
---------------
press reset
System should boot normally.
I thought that this was coming from the flashing process, but could not find these errors in any log.
I do not see these errors when I manually use the flash.sh script (see flash_xxxxx. out above). Are these part of some earlier stage of preparation of the system image? Is stuff messed up before I even try to flash it?
After this configuration, reflash the board and keep the type-c usb cable connected and see if there is a new network interface (you could check it through ifconfig) on your host with ip 192.168.55.100.
“Sorry for the log you saw in #9. It was a stupid mistake that we put it as “stderr” so it outputs as “ERROR”. Actually, it is not.”
Those 'ERROR’s are what have been scaring me the most.
#7:“In the meanwhile, the uart should also have flash log spew. Please check.” #9:“I saw the flash process is done in your log. Are you able to see any uart log during flash?”
I’m not clear what you mean by uart log. A uart.log file under ~/.nvsdkm/logs? or output from a /dev/ttyUSBx device while doing the flash process? (not sure how), or something else?