JP6.1 and Assertion bug fix

[waldman]

HI folks -

TL;DR: Does UEFI release 2502.0-updates have the fix for the “Assertion error”?

Details:
We are using Jetson AGX Orin Industrial with a custom carrier board. We have seen several instances of the ASSERTION error referenced here: Assertion issue in uefi during boot.

Our system is fixed to Jetpack 6.1 for the next few months, so we are unable to upgrade to Jetpack 6.2 that is reported to have the fix.

Following the directions on that ticket, we built new uefi_Jetson_RELEASE.bin and uefi_StandaloneMmOptee_RELEASE.bin with release tag uefi-202502.0-updates, the last UEFI commit compatible with 36.4 according to Combos · NVIDIA/edk2-nvidia Wiki · GitHub.

We also built a new tos-t234.img using arm_trusted_firmware and optee and ./nv_public_src_build_tos.sh. However, the resulting tos-t234.img seemed to be the same as the original image from JP6.1.

Can you confirm that this process (new uefi_jetson.bin, new standalonemmoptee.bin for 202502.0-updates same tos-t234.img) will fix the assertion error?

Thanks!

[DaneLLL]

Hi,
It should be included. The patch is shared in

Jetson/L4T/r36.4.x patches - eLinux.org

You may check the source code to confirm it.

[waldman]

Great, thank you. Can you confirm that only the uefi_jetson.bin and uefi_standaloneMmOptee.bin needed to be replaced?

Thank you,
sam

[jameskuo]

Hi:
Please refer to Assertion issue in UEFI during boot

tos-optee_t234.img, uefi_standaloneMmOptee.bin should be replaced.

[waldman]

James -

When we recompiled tos_t234.img, it’s md5sum matched the previous file. That suggests the inputs didn’t change. Either we did something wrong (likely!) or the inputs to the nv script for that image didn’t change and so we got the same output (maybe?). We will work the procedure and try again - maybe it was just user error. Thanks for your help,
Sam

[jameskuo]

Hi:
Same sha256sum is pretty strange. Do you point UEFI_STMM_PATH to your newly built uefi_StandaloneMmOptee_RELEASE.bin before you build optee?

You can also verify if the patch is applied by resetting UEFI settings, as the link I mentioned above.

[waldman]

Maybe we don’t point UEFI_STMM_PATH – that is where the user error comes in.
If you assert that all three products should be changed – uefi_jetson, standalone, and tos-t234.img – then we did something incorrectly and we will bang at it until we get it right. We will also try the verification – that is definitely a piece we missed that is very useful.

sam

[jameskuo]

Hi:
Precisely, in nv_public_src_build_tos.sh:

and

Seems like you need to call the script using -u flag to point out your STMM image. UEFI_STMM_PATH is used for optee_src_build.sh.

[waldman]

We called it with (mildly redacted):

./nv_public_src_build_tos.sh -p t234 -u ~/jp-6.1/edkrepo/nvidia-uefi/images/uefi_StandaloneMmOptee_RELEASE.bin

-s ~/jp-6.1/Linux_for_Tegra/nv_tegra/tos-scripts/gen_tos_part_img.py

It is entirely possible that call was mistaken or taken out of order, or otherwise screwed up. That’s why we were surprised to get an output in src_out that looked the same. We’ll repeat the process and pay even more attention.

More importantly, we’ll run your verification and “prove” that the patch “took”.

Thanks for your time.