Deadlocks

Detection and recovery

Noticing that there is a cycle of waiting processes, and killing one of them in order to break the deadlock. This requires the OS to wake up every once in a while, and check on all the threads which are waiting, to see if there is a deadlock cycle between waiting processes. If there is, action is taken, which may be as drastic as killing one of the offending processes, or, less drastically, to wake up one of the waiting threads with a failure message (i.e. to indicate that it is not allowed to have the resource for which it was waiting).


last updated 23 April 1998