Jetson NX 3668-0030 acr_ucode.bin ucode get fail for gv11b after successful boot

We bought some 3668-0030-000 Jetson NX modules ( 16GB) and upgraded our Linux to 32.7.2 bacause flashing with 32.4.3. was unsuccessful ( as discussed in other threads ).
We did a full make ( ./m.sh ) and flashed the system to the NX.
It came up with the start screen over HDMI, requested info on local, keyboard, name etc as normal.
Then it outputs several messages about configuring hardware at HDMI.
Finally it switches off the GUI and repeats the message block shown below on serial and on HDMI.
Nothing ( reboot, etc ) helps. Login to serial or SSH is impossible.
Are there any ideas what causes this problem ??

WARNING: at platform/drivers/pg/pg-gpu-t194.c:185
[ 21.087194] gk20a 17000000.gv11b: Direct firmware load for gv11b/acr_ucode.bin failed with error -2
[ 21.087394] gk20a 17000000.gv11b: Falling back to user helper
[ 21.090021] nvgpu: 17000000.gv11b gm20b_bootstrap_hs_acr:1279 [ERR] acr_ucode.bin ucode get fail for gv11b
[ 21.090222] nvgpu: 17000000.gv11b gk20a_finalize_poweron:323 [ERR] ACR bootstrap failed
WARNING: at platform/drivers/pg/pg-gpu-t194.c:185

Info: We helped ourselves with the 16GB XNX patch from L4T R32.4.4 archive | NVIDIA Developer ( https://developer.nvidia.com/xnx-16gb-overlay.tbz2 ) and continue with Linux 32.4.3 for now.

Hi,

Sorry for late reply. Could you provide the steps to reproduce this issue? Is this on devkit or your board?

Also, what does that mean “m.sh” in your comment?

Hi,
what we did was to install 32.7.2 ( in parallel to 32.4.3 ) on our development host.
We simply changed the path to the new Linux 32.7.2 directory ( downloaded from Nvidia ) and called our existing make-all script ( m.sh ) .
Make ends with a success and we can flash the XNX-16GB on a developer board.
Then it comes up as described and requires the configuration steps at the GUI over HDMI as described earlier in this post.
When all is done the XNX tries to configure its hardware and ends in a endless loop as described.
I’m sure it is somehow our fault, but we do not really understand the error message.
It seems to be something with GPU initialisation, doesn’t it ? Is there a missing software package which is not included in compilation and needs to be downloaded and can’t be found ?
Brgds, A.

Hi Andreas,

Just want to confirm, if you don’t use 32.4.3 but 32.7.2, would you still hit this error?

Also, does your rel-32.4.3 roofs have acr_ucode.bin?

Hi Wayne,
we ONLY have got this error in 32.7.2.
After patching our 32.4.3 ( my second post ) we can flash XNX-16GB using 32.4.3 without problems. This forced us to calm down a lot :-) .
The reason to start using 32.7.2 was the recommendation to use it when flashing XNX-16GB fails .
( see Flash.sh jetson-xavier-nx-devkit-emmc mmcblk0p1 with error! - Jetson & Embedded Systems / Jetson Xavier NX - NVIDIA Developer Forums )
Brgds, A.

Oh ok. So does your 32.7.2 rootfs have this acr_ucode.bin?

Exactly.

Could you attach the full log?

NX_Boot.txt (283.0 KB)

Is that file also existing in your /lib/firmware after device boot up? could you share us the result?

I will check, but it will take me some time.

It seems I know what is going on here. You said you are using rel-32.7.2, but your log tells you are using some earlier kernel which is definitely not 32.7.2.

Hi Wayne,
thanks.
I will check in this direction and keep you informed.
Brgds, A.

Have you managed to get issue resolved or still need the support?

Thanks

Hello kayyccc,

as we found the the patch for using Linux 32.4.3 with 16GB Jetson Xavier NX on the NVIDIA software site we are fine now.
No need ( no time at our side … ) to discuss about the reported problem during migration to 32.7.2 .

Maybe it could be a good option to add a hint to the blog(s) where they discuss about flashing problems with 16GB XNX production modules with 32.4.3 / 4 that there is a Linux-patch at your website available.
There are at least two blogs recommending upgrade to 32.7.2 which forced us to to try it with 32.2.7 which finally failed because we tried it the „easy way“ , I suppose.

Best regards

Andreas