The invention relates to a method in a hardware semaphore lock (L1 LN) intended for a multi-processor system, which semaphore lock (L1 LN) protects a shared resource (R1 RN) in connection with the system in such a way that only a process which has reserved the semaphore lock (L1 LN) and has thus become a holder of the lock, has access to use the resource protected by the lock. The semaphore lock (L1 LN) is reserved by a single read operation of a memory location representing the semaphore lock by the process software. The read operation returns to the process the number of vacant holder positions, i.e. keyholes vacant at the time of the reservation of the lock. The semaphore lock does not require the support of the system for atomic read/write operations.