Jetson Nano: GUI not working on HDMI out put

Hello,
I have Jetson Nano production module with custom carrier board from seeed studio. Which was working fine for the first boot , after that when I power cycled the board, the GUI is not coming up I only see the some log messages for which I added the photo in attachment ,
The two errors I see there are: “[ERROR] failed to start nvpmodel service” & “[ERROR] Failed to start Docker Application Container Engine”.

I am able to login through serial console, also attached complete log file from the console gtkterm.log (25.7 KB).
what could be the issue?

In the next step I want to flash the Jetson Nano with additional driver integration, just want to make sure that this issue will not create any trouble during flashing.

Thanks,
Krishna

Log is not complete due to kernel log is in silent mode. Please type “dmesg” in your console and attach it.

Here is the complete log with dmesg
gtkterm1.log (116.8 KB)

please also share /var/log/Xorg.0.log.

cat_ var_log_Xorg_0_log (7.9 KB)

Your gpu driver seems gone.

[ 257.302] (II) NVIDIA GLX Module 32.6.1 Release Build (integ_stage_rel) (buildbrain@mobile-u64-5496-d5000) Mon Jul 26 12:11:24 PDT 2021
[ 257.302] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA graphics device!
[ 257.302] (EE) NVIDIA(0): Failing initialization of X screen 0

Do you remember what was the last thing you did before hitting this issue?

What is the “lsmod” result?

I tried to patch the custom driver , but only took backup of kernel Image and the script which then looks for device tree backup but it failed and return without doing any modifications
But i did revert the changes i.e put back the original kernel Image back in /boot/Image.
for reference attached the script.
install_binaries.sh (13.2 KB)

“lsmod” returns nothing.

ok, then this is nothing special and nothing new. You build the kernel but system cannot find the modules according to the kernel local version.

lsmod just indicates that.

Please follow the steps mentioned in the document to build kernel.

https://docs.nvidia.com/jetson/archives/l4t-archived/l4t-3261/index.html#page/Tegra%20Linux%20Driver%20Package%20Development%20Guide/kernel_custom.html#

Especially this one.

$ export LOCALVERSION=-tegra

The search path of modules are under /lib/modules/{uname -r}. Your current issue is probably this path gives no modules.

1 Like

Ok , Thank you. I will do that and check if it’s working

Hello,

If I re-flash with the command line without building the kernel (using L4T driver package (BSP) and sample rootfs) it should work , right? as building the kernel on Jetson Nano is not feasible I have limited memory ,only internal eMMC.

Thank you,
Krishna

Yes, that is the base case as what got flashed by sdkmanager.

Has the kernel already been modified? I ask because apparently some of the “/sys” files used for nvpmodel are missing. These are not actual files, but are instead kernel drivers or features pretending to be files, and those files are missing, then it tends to mean the kernel was compiled without that content.

EDIT: Sorry, did not see other replies until after I posted this. Kernel modules missing are the same as whole groups of kernel features missing. Note that modules are searched for at “/lib/modules/$(uname -r)/kernel”. What is your “uname -r”? If not correct, then you probably did not set CONFIG_LOCALVERSION.

1 Like

I flash the board with new image.

“uname -r” is 4.9.253-gcca4bf6ca, instead it should be something like 4.9.253-tegra

Yes, unless you installed all new modules against the new “uname -r” all modules will fail to be found. Try rebuilding with “CONFIG_LOCALVERSION=-tegra”.

1 Like