This post is mostly to share information with NVIDIA and potential customers about how volatile the Jetson orin nano is.
I have a Jetson Orin Nano 8GB developer version. I’ve flashed it with Jetpack 6.0 DP on NVMe using SDK manager (Mar24). After ~4 months of routine use it won’t boot. It freezes on the NVIDIA logo with “Enter to continue boot”. I have had this issue before, and I fixed it by inserting a new SSD, flashing it, booting up. Then I reinstalled my old SSD and it seemed to work. I think the underlying issue is the bootloader gets corrupted, so reflashing a spare SSD also reflashes the bootloader?
This fix isn’t working this time around I assume because the boot loader has been updated to (rev2)? So, I can flash the spare SSD fine, but when I put the old one back in it attempts to boot 3 times and then freezes on a black screen.
Is there a way to recover my data from the SSD? I really need the 3 months of work I have on it. Would a USB → SSD adapter work if plugged into the newly flashed SSD?
(if relevant, the other hardware I have is a MacBook Pro M2 and iMac X86-64 flashed with Ubuntu)
(I do not have a TTL → USB cable for console log…)
Just spend to get one TTL cable. It would be helpful in debugging such issue.
That is the most precise way. Telling things from monitors are not precise.
Also, bootloader is not on the SSD. It is on the QSPI memory of the module.
So reflashing the SSD won’t recover the bootloader.
If you want to check the SSD, actually you could just put your SSD back to other host PC and check it. It does not matter whether jetson can still boot or not.
Incidentally, if you can put your SSD on another computer, then it is always easy to recover a partition. Or you could back it up as you go with something like rsync. If you have the SSD on another Linux system, try to identify which device it is in “lsblk -f” and post the part for the SSD.
I’m not interested in debugging for my own benefit, I’ve cut my losses. But, if it would be greatly beneficial to NVIDIA to improve future versions, I will get a TTL cable.
I understand the bootloader is not on the SSD, what I was saying is the fix was to reflash the Jetson with SDK. This flashed the SSD and the QSPI. Thus, reflashing did fix the Jetson. But because the bootloader has been updated, my old SSD (Jetpack 6.0 DP) will not longer boot. Does this make sense? Should I explain what I’m saying in more detail?
My host PC is an iMac, I cannot plug the SSD into it.
Thanks Linuxdev, I’ll find someone with a linux desktop to insert it into and get back to you.
(The original Jetson Nano and SD card were so clean and easy to use and didn’t require a Ubuntu desktop to be functional. NVIDIA please go back to a simpler solution. SDK manager is a mess - its incredible when it works, but a nightmare when it doesn’t)
It is hard to tell what is really going on there without log.
For example, if you accidentally crashed the filesystem, then changing to another bootloader may not be able to recover it.
It needs some log there so that we can tell what to fix or what to enhance. I am not saying we won’t fix anything here. The problem is just we need a direction to know what is going on there.
Sdkmanger has been there for both Jetson Nano and also Orin Nano. Actually, sdkamanger comes first and then the sdcard image.
Jetson Nano has chance to meet corrupted bootloader case too. For that kind of case, it would still need sdkmanager to flash it to recover.
And Orin Nano also has sdcard version too. The logic between these two kinds of platforms are similar. Just some software are different during boot flow.
For example, sdkmanager supports both. And sdcard image supports both too.