Deadlock is a potential problem in any operating system. It occurs when all the members of a set of processes are blocked waiting for an event that only other members of the same set can cause. This situation causes all the processes to wait forever. Commonly the event that the processes are waiting for is the release of some resource held by another member of the set. Another situation in which deadlock is possible is when a set of communicating processes are all waiting for a message and the communication channel is empty and no timeouts are pending.

Resource deadlock can be avoided by keeping track of which states are safe and which are unsafe. A safe state is one in which there exists a sequence of events that guarantee that all processes can finish. An unsafe state has no such guarantee. The banker’s algorithm avoids deadlock by not granting a request if that request will put the system in an unsafe state.

Resource deadlock can be structurally prevented by building the system in such a way that it can never occur by design. For example, by allowing a process to hold only one resource at any instant, the circular wait condition required for deadlock is broken. Resource deadlock can also be prevented by numbering all the resources and making processes request them in strictly increasing order.

Resource deadlock is not the only kind of deadlock. Communication deadlock is also a potential problem in some systems although it can often be handled by setting appropriate timeouts.

Livelock is similar to deadlock in that it can stop all forward progress, but it is technically different since it involves processes that are not actually blocked. Starvation can be avoided by a first-come, first-served allocation policy.