After upgrading the kernel on our iMX6D system (from 3.10.17 to 5.4.8) we are experiencing instabilities on some of the modules (some modules crash every few minutes, others once a day and some others run without fault for a week or more).The system randomly generates kernel panics (virtual page pointer errors or scheduling stalls) and then has to restart.
After debugging and testing for quite some time the only clue we could come up with is that the problem must be related to SMP/Dual Core Use. The affected systems become stable if SMP is disabled in the kernel config.
During the problem investigation we upgraded further to kernel 5.4.123 and updated U-Boot/SPL (from 2014.04 to 2020.01[denx]) but those changes were without effect.
Does anyone have any idea what could be the root cause or how we could debug the problem further?
Update:
The system stays stable if SMP is activated but only 1 core is booted up (maxcpus=1). If the second core is started afterwards in Linux using
echo 1 > /sys/devices/system/cpu/cpu1/online
the system gets unstable pretty quickly.