Deadlock Detection and Recovery

It's a method of permitting the system to enter a deadlock state, detect it and then recover.

Deadlock detection :

• Is the process of in fact determining that a deadlock exists and identifying the processes and resources involved in the deadlock.
• The fundamental idea is to check allocation against resource availability for all possible allocation sequences to determine if the system is in deadlocked state. 
• As well a Wait-For-Graph (WFG) may be used if there is only a single instance of each resource. In the WFG the nodes denotes processes and the edges denotes the blockages or dependencies.  Therefore if process A is waiting for a resource deem by process B there is an edge in the WFG from the node for process A to the node for process B.A cycle in the graph point to a deadlock.

Deadlock recovery:

Formerly a deadlock is detected, there wants to be a way to recover several alternatives exists:

• Temporarily protect resources from deadlocked processes.
• Back off a process to some check point permitting preemption of a needed resource and restarting the process at the checkpoint later.
• Successively termination processes until the system is deadlock free.