We have a custom board with an STM32MP151A and STPMIC1. The PMIC is connected to I2C4 over pins PD12 and PD13.
In the process of migrating our software from OpenSTLinux 6.0 to 6.2, I initially got stuck into an optee panic (attached optee-panic.log with optee log level set to 3); after some investigation I *think* I figured out that the problem was coming from I2C4 being assigned to the secure context, but pins D12/D13 not being secure.
I managed to go around that by making the I2C4 controller non-secure and removing all reference to it from the optee dts, but then the kernel was hanging forever (attached optee-pass-kernel-fail.log) until I also marked the RCC as non-secure for optee and changed the clock definitions in the kernel to not use SCMI (to do so I had to patch stm32mp15-scmi.dtsi to be empty).
I now have a booting configuration by essentially avoiding SCMI for clocks and regulators (I have attached the kernel and optee dts), but I am wondering whether this is necessary because we are using the 151A && we need to have the I2C4 controller in the non-secure context, or whether I went too far and there is a configuration in which I can have the RCC and regulators be owned by optee, with the kernel accessing them via SCMI.
