Thanks Michal for this calculation.
I will send separate patch after some more analysis to accommodate this implementation of accessing message buffers from ipi-id.
However, for this series this isn't required. For this series, I will just split this patch into three different patches. I hope it's okay.