Hey @WayneWWW I appreciate someone else looking at it since we’ve had multiple people from your side telling that it works, despite there being multiple reports that it doesn’t. I had multiple questions about 5.1:
1. How to set a slot to bootable again after it was marked unbootable (In order to try an update again) Answered by @JerryChang Setting bootable and unbootable AB rootFS slots for L4T r35.1 - #36 by JerryChang
Current rootfs slot: B
Active rootfs slot: A
slot: 0, retry_count: 0, status: unbootable
slot: 1, retry_count: 2, status: normal
Set a slot to unbootable in the UEFI. Afterwards set it to active with nvbootctrl → reboot loop
- Why Jetpack 5.1 does not reboot automatically to attempt booting when the current slot is not able to boot:
Jetpack 5.1 Kernel Panic does not lead to reboot with A/B System → Talking around the topic for days
Setting bootable and unbootable AB rootFS slots for L4T r35.1 - #38 by Max_Dichler → Read the whole topic here, you will see that they both are not able to get it to reboot.
- When updating from 5.0.2 to 5.1 I need to update the bootloader before the rootfs, so I need to use the old bootloader update mechanism. Unfortunately the update works, but afterwarts nvbootctrl tells me that A/B is disabled…I can still switch slots with the tool, but something seems to be wrong.
Now the story why all this together seems to be a big issue for me and others.
The use case of A/B Redundancy is to have a second slot to boot from if a slot gets corrupted. There can be various reasons for the slot to be corrupted, unexpected power loss, deleted files or an update attempt which failed.
In the current 5.1 the system will not automatically reboot to the other slot, if the active one is broken. Taking power away does not fix that, as the scratch register is not updated then. So you’re stuck with an unbootable system forever. Changing the slot to unbootable is not a workable solution in production, but even if it was…it does not work, the system will still not use the other slot.
Now we have been told that in case it WOULD reboot to the other slot, we would be left with an unbootable slot and a bootable slot. In that case an update would have to be placed on the unbootable slot. We have no tool, except the uefi, which we can’t access from the running system, to set the unbootable slot to bootable again.
So the whole idea of using A/B redundancy would not work anymore as soon as something goes wrong 1 single time…you can not update anymore then.
I have not been able to test if the nvbootctrl would set the slot to bootable again if I set it active since we are not able to get the system to mark the slot unbootable when it fails.
The issue about me not being able to switch slots from the topic post seems to be on my side, but the rest seems to affect everyone. Edit: My personal issue that switching A/B was not possible has been resolved after flashing it again…strange. The other issues regarding the A/B failover and the lack of a possibility to make an unbootable slot bootable again still exist.