Hi, I put my Jetson TX2 in recovery mode, the host (laptop with ubuntu 18.04) recognised it, SDK started to flashing the os and after some time it showed ‘could not detect a board’.
I tried to skip it, as suggested here, but my jetson doesn’t show any hdmi output or can’t connect with ssh usb.
Previous I had problems with system flashing witch Jetson Orin and someone suggested Dell laptop as a host is the problem. I switched host to Sony and it flashed, but in this case I tried couple times with different hosts and the result is identical.
Does your Ubuntu 18.04 run bare-metal without any virtualisation of any kind directly on the hardware. Flashing from any virtual environment usually doesn’t work.
Does the Devkit USB connection show up with lsusb?
Suggestion: Use an USB-LVTTL serial cable and connect it to UART0 on the TX2. On the original devkit it is pin 8 for UART0_TX and Pin 10 for UART0_TX and Pin 9 for GND on the 40 pin connector. You need to cross TX and RX lines, TX from the usb adapter goes to RX on the TX2 and vice versa. You should have output there (115200 8N1, no handshake), and you should be able to login and perform the initial configuration. After this you may also use the TX2 Ethernet if the RNDIS-USB network connection doesn’t work.
microusb is required for flashing the EMMC throgh the recovery boot rom in the processor. This is the first step. After this the TX2 reboots, and the Linux drivers take over.
So you always need the microUSB connection.
Debug output is sent via UART0, the console UART. This debug output is essential to solve problems.
Do you use the lastest (!) Jetpack (4.6.4)? Newer TX2 modules don’t work with older Jetpacks since the hardware has changed oder time in the various module revisions, especially Flash and RAM chips. 4.6.4 is known to work with all TX2 revisions.
Ok, yes I use JetPack 4.6.4. One thing I did’t mention, when I skip this Default Ip problem, the installation ends with warning (sdk not installed on jetson), but two green LEDs (next to the buttons) are still on, so it means it is stuck in recovery mode?
There is a “EXPORT LOGS” button on sdkmanager which you can use to dump log.
We need the log to tell if the board really get flashed or not.
Flashing requires to use recovery mode.
But installing sdk has nothing to do with OS, so installing sdk does not need recovery mode.
For example, you don’t need to format your whole Windows OS on your PC to just install a CUDA driver. Similar case for Jetson.
That IP is asked because it needs to install sdk after flashing your OS.
Full logs above.
If I understand properly my system is flashed, but how can I use jetson if it’s not visible via USB and doesn’t produce any HDMI signals?
Incidentally, if you are having problems with first boot account setup, then in the “Linux_for_Tegra/” directory you can run either “sudo ./tools/l4t_create_default_user.sh” or “sudo ./l4t_create_default_user.sh” (in older releases it is not in the “tools/” subdirectory, so I listed both locations) to complete the account setup prior to flashing, and then flash again. You wouldn’t want to ship a commercial product with a default name/pass (especially to California), but this is quite useful if you are flashing for yourself and it skips first boot setup.
