Myths About the Mutual Exclusion Problem

Recently in these pages appeared a report by Doran and Thomas [2] which gave partially simplified versions of Dekker-like solutions to the two process mutual exclusion problem with busy-waiting. This report presents a truly simple solution to the problem and attempts in a small way to dispel some myths that seem to have arisen concerning the problem. Briefly, the mutual exclusion problem for two processes is to fmd sections of code (trying protocol, exit protocol) for each of two asynchronous processes to use when trying to enter and upon exiting their designated critical sections. The protocols must preserve mutual exclusion and not have deadlock or lockout. Mutual exclusion means that both processes can never be in their critical sections at the same time. No deadlock or lockout means that no process waits forever inside a protocol. More formal definitions can be found in [S] and elsewhere. The original solution due to Dekker is discussed at length by Dijkstra in [l]. GE the many reformulations given since, perhaps the best appears in 131. (Unfortumtely the authors believe their correct solution is incorrect.) The solutions of Doran and Thomas are . slight improvements which eliminate the ‘loop inside a loop’ structure of the previously published solutions. The solution presented here has an extremely simple structure and, as shown later, is easy to prove correct.