I built a custom carrier board that can be used for Xavier, Jetson and Orin Nano versions. I have verified the HDMI works with a Jetson Nano I had laying around.
For the Orin Nano, I know that it is configured for DP out of the box (if I am using the stock kernel)
Is there some help online somewhere on what steps I should take to switch over to HDMI?
Is the first step getting the pinmux setup correctly?
Any help would be great, thanks!!!
Rick
p3509-a02-p3767 board config enabled the HDMI by default. You can try that first.
So what we tried was this:
ln -s p3509-a02+p3767-0000.conf jetson-orin-nano-devkit.conf
cd tools/
./jetson-disk-image-creator.sh -o ~/orinos-11hdmi.img -b jetson-orin-nano-devkit -r 100 -d SD
in
~/nvidia/nvidia_sdk/JetPack_5.1.1_Linux_JETSON_ORIN_NANO_TARGETS/Linux_for_Tegra/tools
And then burned that to an SD card.
On the custom board, nothing comes up (I have a second board coming that has the serial exposed so I can watch the boot, unfortunately, it’s not here yet.
If I take this same image and put it on th Nano Dev board, I see the boot loader screen, then I don’t see anything after that. Given that board does not have HDMI, I am guessing it’s still trying to use it.
Did I do something wrong here?
Hi,
Yes, you did it wrong.
The method goes wrong since you use sdcard image. Sdcard image does not really full flash to jetson board.
Please use the initrd flash tool to full flash the board instead of sdcard image.
If I take this same image and put it on th Nano Dev board, I see the boot loader screen, then I don’t see anything after that. Given that board does not have HDMI, I am guessing it’s still trying to use it.
BTW, I am totally not sure what you are trying to say in this comment.
Oh, so if I understand you correctly, these are changes that have to make modifications to the flash memory that is on the actual Nano compute module itself. Is that correct?
For the part that didn’t make sense, I think that is unrelated. In digging further, before we made all these changes, the person here compiled the images had set it to be barebones and in that configuration, nothing gets displayed after the boot screen (but you can log into it over the network). So ignore all that. (I wasn’t aware it disabled the video, but it definitely does).
I will work on flashing the board and get back to you.
Oh, so if I understand you correctly, these are changes that have to make modifications to the flash memory that is on the actual Nano compute module itself. Is that correct?
Yes, correct. Display related software is not only on the sdcard. It needs the update on the QSPI flash memory on the jetson module too.
BTW, if you have NV devkit board( xavier nx devkit + Orin Nano module), you can go through all of this first so that you would see it work.
I do have a Xavier NX kit as well! :) And a Jetson Nano too. Starting to get a collection.
On the orin nano dev, I assume we need a USB-C to hook to our machine with virtual ubuntu on it to do the flash?
I do have a Xavier NX kit as well! :) And a Jetson Nano too. Starting to get a collection.
Actually Orin nano and jetson nano are not same module. I hope this is just a typo in your comment.
On the orin nano dev, I assume we need a USB-C to hook to our machine with virtual ubuntu on it to do the flash?
We expect the host machine should be a native ubuntu instead of virtual machine in case of some usb detection problem.
I meant I had all three. Jetson Orin Nano, Jetson Xavier NX and my original Jetson Nano (all with three different dev boards).
Using this same soft link, we did this again, then we flashed using in recovery mode to the ORIN in the ORIN carrier board:
./flash.sh -S 16GiB jetson-orin-nano-devkit mmcblk0p1
If we did not specify -S, it would fail. Not sure why we had to put the size.
When it completes, I have to go into extlinunx.conf and change the root_fs to point to mmcblk1p1 instead, not sure why that was required. If I don’t, it says cannot find mmcblk0p1
After the flash completed (which takes awhile), I put the orin nano into the XavierNX carrier and the initial bootloader screen comes up, but then goes black after it boots into the SD card. We are able to log into the machine over ethernet. I grep hdmi in dmesg and only see a regulator message.
So, we took at look at the DTB out of the partition and saw that there isn’t any HDMI config like we saw in an earlier board (Jetson Nano). I only compared to that older DTB because I knew it worked (and just wanted to see what sort of info might be required in the new DTB)
So not sure what to do from this point. I went through the DTB pointed in p3509-a02+p3767-0000.conf and I don’t see any HDMI setting sections there either.
Hi,
You need to learn how to really flash orin nano/nx first. Orin nano and nx does not have internal emmc on it… so your flash.sh command basically flash nothing to your board…
I mean, your board has no correct kernel and rootfs to boot up at all… it is just uses previous device tree and kernel stored in your sdcard…
https://docs.nvidia.com/jetson/archives/r35.3.1/DeveloperGuide/text/IN/QuickStart.html
Also, orin won’t have anything to check in dmesg for HDMI anymore. Basically you can only check xrandr.
Thank you for the pointer, somehow I got onto another instructions page for something earlier than r25.3.1 and it wasn’t clear what commands I should be using. Now looking at the page you sent, I see:
$ sudo ./tools/kernel_flash/l4t_initrd_flash.sh --external-device mmcblk1p1 \
-c tools/kernel_flash/flash_l4t_external.xml -p "-c bootloader/t186ref/cfg/flash_t234_qspi.xml" \
--showlogs --network usb0 jetson-orin-nano-devkit internal
So we’ll try that. And thanks for the heads up on what we should see in dmesg. Will report back once we try.
Hi,
Just a reminder, board config needs to change to p3509 one if you want to use that one xavier nx devkit carrier board…
Okay, we flashed with the proper method, using (which completed successfully):
sudo ./tools/kernel_flash/l4t_initrd_flash.sh --external-device mmcblk1p1 \
-c tools/kernel_flash/flash_l4t_external.xml -p "-c bootloader/t186ref/cfg/flash_t234_qspi.xml" \
--showlogs --network usb0 jetson-orin-nano-devkit internal
We ran with the link I mentioned earlier to map to the p3509:
ln -s p3509-a02+p3767-0000.conf jetson-orin-nano-devkit.conf
Once finished, we took over to the JavierNX carrier and booted. Once it goes to the SD and starts loading the OS, the monitor shuts off. We logged in remotely and can get into it.
The image we flashed doesn’t have all the packages, so we had to install them like I read on some of the pages with the apt command. We verified it installed the Nvidia x11 l4t version. We rebooted and it did not start X. We started X with “startx” but it seems to just create a virtual desktop.
What seems suspicious is that after it starts booting, the screen still shuts off. We have another Orin with DP connected with a similar image (without the p3509 board, since we want it to use DP in that case, on an Orin board). This boots and then shows a the console login on the monitor attached by DP.
So it sounds like something might still not be configured right. We must still be doing something wrong. Given so few steps here (point to p3509 conf, flash, reboot, install l4t packages, reboot, nothing)
We also tried to use:
nvidia-xconfig
To see if we could query the GPU, but I couldn’t tell if this command was supported on these modules. The results I saw were:
sudo nvidia-xconfig --query-gpu-info
WARNING: error opening libnvidia-cfg.so.1: libnvidia-cfg.so.1: cannot open shared object file: No such file or directory.
ERROR: Unable to query GPU information
Hi,
This is a common mistake that you want to change lots of variables all at once. It will just lead you to don’t know which one is causing the problem.
What I suggest here is you just put your module on 3509 and uses default BSP which includes all the software first and boot up. Make sure HDMI can work fine with this setup first. And this setup is your base line.
You need to get a baseline first otherwise you may think the lower level (p3509 device tree/ kernel) may have problem.
There is likely confusion on our end what is the right things to be pulling down. Can you point me to the instructions for using the default BSP to build an image just to make sure we are following all the proper procedures.
Hi,
There is no need to “build image” because we already put everything on the website or you can use sdkmanager to download the file to you.
My question here is I don’t know what difficulty you hit here. This is basically just download some files, doing nothing and run the flash command.
I mean you don’t need to “ln -s” anything or run startx or install nvidia-xconfig or try DP.
These are just making chaos as I am not on your side and I won’t know how many things you’ve done. Some of them could be just dummy but some of them could be fatal.
You just need to flash your orin nano on xavier p3509 devkit. If you hit problem, do not try anything. Just come back to me and share the log or the situation you saw.
Thanks for all these details, this is exactly what we needed to know. It wasn’t apparent to me in this discussion that we should be using the sdkmanager, but we have all that installed so it will be easy to test. We were clearly making things more complicated then we needed to at this stage.
We will put the Orin in the Xavier p3509 devkit board, attach with USB, and use the sdkmanager to flash. Will report back.