single 8K process stack vs 4K process stack and a seperate 4K interrupt stack
 

Why was the decision to split the initially 8K process kernel stack into separate stacks for process and interrupt made?
Doesn't single 8K stack give more effective utilization over 2 different 4K stacks for process and interrupt?

I'd google for this question. Have you read this? http://kerneltrap.org/Linux/Moving_4K_Stacks_Forward