swupdate/capsule update, how to mark-good?

[AndersGnistrup]

I have a working system based on orin/nano, meta-tegra and swupdate
It works just fine and I can update the system via swupdate -i
I have read this guide:
https://docs.nvidia.com/jetson/archives/r36.2/DeveloperGuide/SD/Bootloader/UpdateAndRedundancy.html#manually-sync-bootloader-a-b-slots-with-the-capsule-update
From what I can read the process is:

• copy capsule and notify the bootloader setting bit2 of the OsIndications UEFI variable.
• reboot and check the status the capsule update using nvbootctrl

But this only indicates if the update was OK, not if the SW in the packages is OK.
The rauc has the operation mark-good to indicate that SW is good to go after the update.

How is the "mark good" handled in swupdate?

[AndersGnistrup]

Ok, after some further reading it seems to be related to a swupdate progress and a swupdate service.
I will post my findings.

[AndersGnistrup]

I was wrong, swupdate-progress has nothing to do with mark-good, only the current state/progress of the file transfer/copy.

I am using origin/scarthgap-l4t-r35.x on meta-tegra
There is a script nv_update_verifier that calls nvbootctrl verify but this returns:
variable BootChainFwStatus is not found.
And I gees the reason is that the uefi bootloader does not create this variable in the uefi variables.
So, I think I need to update my distro a clean scarthgap branch,

overall conclusion: mark good does not work in jetson 35.x, you need 36.3
Is this correct?

I have used the source for debug and these pages:
https://docs.nvidia.com/jetson/archives/r36.3/DeveloperGuide/SD/RootFileSystem.html#rootfs-selection
#1428
https://docs.nvidia.com/jetson/archives/l4t-archived/l4t-3275/index.html#page/Tegra%20Linux%20Driver%20Package%20Development%20Guide/rootfs_custom.html#wwpID0E5HA

[dwalkes]

Not sure if you found it already but there is some discussion about BootChainFwStatus for marking good status in mendersoftware/meta-mender-community#394 and OE4T/meta-mender-community#19. I'm not completely sure where this ended up with the current mender implementation but this might be something to look at. Ideally we'd upstream something all update frameworks could use.

This would be something great to add to the reference swupdate demo at https://github.com/OE4T/tegra-demo-distro?tab=readme-ov-file#update-image-demo. There's a Jetpack 4 swupdate implementation in Matt's test distro which used https://github.com/OE4T/tegra-boot-tools/blob/059ddc8187ae0e9f7ba0acdb672bdabaa7559434/doc/tegra-bootinfo.md for this. You can find more details at about that original implementation from Matt which I used for the tegra-demo-distro implementation at https://github.com/madisongh/tegra-test-distro/wiki/swupdate-integration-notes

[AndersGnistrup]

Thanks for the reply.
I will try to make a solution that could be used in the swupdate demo. But that will take some time. At lease I have a known path.

[AndersGnistrup]

Now for an update.
First, nvbootctrl verify does work, but emits a warning about the BootChainFwStatus is missing. I am not sure but this seems to be a left over since the efi var is gone in the kernel in jetson 36.3
If I delete/disable the service nv-update_verifyer.service and boots three times after a swupdate, bootloader switches back back to A which is expected.

Current rootfs slot: A
Active rootfs slot: A
num_slots: 2
slot: 0,             retry_count: 3,             status: normal
slot: 1,             retry_count: 0,             status: unbootable

So stuff actually seems to work just fine.
The problem is that the rootfs slot B is marked unbootable so new firmware will be discarded by the bootloader. These are the flags.

cat /sys/firmware/efi/efivars/RootfsStatusSlotB-781e084c-a330-417c-b678-38e696380cb9 | hexdump -C
00000000  07 00 00 ff 00 00 00 00                           |........|
00000008

There are several ways to clear the flag before a swupdate, see https://forums.developer.nvidia.com/t/rootfs-a-b-redundancy-fail-over-mechanism-in-jetpack5-1/247290

And now I am a bit confused.
It seems that the BootChainFwStatus is obsolete/replaced with something else. Is this correct?
Most, if not all. seems to be covered by the efi vars present in 36.3.

This is right after a swupdate.

# nvbootctrl -t rootfs dump-slots-info
Current rootfs slot: B
Active rootfs slot: B
num_slots: 2
slot: 0,             retry_count: 3,             status: normal
slot: 1,             retry_count: 2,             status: normal
# nvbootctrl verify
Info: variable BootChainFwStatus is not found.
# nvbootctrl -t rootfs dump-slots-info
Current rootfs slot: B
Active rootfs slot: B
num_slots: 2
slot: 0,             retry_count: 3,             status: normal
slot: 1,             retry_count: 3,             status: normal