Concurrency Issues Past lectures Problem Safely coordinate access to shared resource Solutions - PDF document

Deadlock Concurrency Issues Past lectures:Problem: Safely coordinate access to shared resource Solutions: Solutions: Use semaphores, monitors, locks, condition variables What about coordinated access multiple objects?If you are not careful, it can lead to slecture: 2 Todays What is deadlock?How can we address deadlock? Motivating Examples processes share a buffer but use a different protocol for accessing the buffers Producer1() { P( Bff ) Producer2(){ dMLk A postscript interpreter and a visualization program compete for memory frames P( empty e r ) producerMutexLock emptyBuffer P() { l() { 3 P S_Interpreter () { process filerequest(frame_buffer, 1)draw file on screen Visua ize () { request(frame_buffer, 1) The TENEX Case If a process requests all systems buffers, operator console tries to print an error message lock the consolerequest a buffer 4 Deadlock Ready A set of processes is deadlocked when every process in the set is for an event that can onl y be g enerated b y some p rocess in Waiting Tailready queue Tail 5 gygyp the set Starvation vs. deadlockStarvation: threads wait indefinitely (e.g., because some other thread is Deadlock: circular waiting for resourcesstarvation, but not the other way The resource allocation graph ( Basic components of any resource allocation problem Model the state of a computer system as a directed graph G ( V E ) G = = the set of vertices = { i = the set of edges = { ed a resource to a p roces 6 i k request gfmp edges from a process to a resource Resource Allocation GraphsExamples A PostScript interpreter that is waiting for the frame buffer lock and a visualization process that is waiting for memoryvisualizationmemory framesframe buffer lock VisualizationProcess PostScript Interpreter Frame Buffer Interpreter Resource allocation graphs & deadlock Theorem If a resource allocation graph does not contain a cycle then no processes are deadlocked A cycle in a RAG is a necessary condition for deadlock A cycle in a RAG is a necessary condition for deadlock Is the existence of a cycle a sufficient condition? Game VisualizationProcessInterpreter Resource allocation graphs & deadlock Theorem If there is only a single unit of all resources then a set of processes are deadlocked iff there is a cycle in the resource allocationgraph allocation Visualization Memory Frames PostScript 9 Visualization Process PostScript Interpreter Using the TheoryAn operational definition of deadlock Visualization MemoryFrames PSi A set of processes are deadlocked the following conditions hold simultaneously P rocess Frame Buffer P ost t Interpreter 1.Mutual exclusion is required for resource usage (serially useable)2.A process is in a “hold-and-wait” state3.Preemption of resource usage is not allowed4.Circular waiting exists (a cycle exists in the Dealing With DeadlockDeadlock prevention & avoidance tion protocol that ensures deadlock can never occur Ddlkti/id D ea dl oc preven ti on Guarantee that deadlock will never occurGenerally breaks one of the following conditions:Hold-and-waitNo preemptionCircular wait *This is usually the weak link*Deadlock detection and recover y 11 Admit the possibility of deadlock occurring and periodically check for itOn detecting deadlock, abortBreaks the no-preemption conditionWhat does the RAG for a lock look like? Resource Ordering Recall this situation. How can we avoid it? emptyBuffer dMtLk ) producerMutexLock tBff ) p t } B u e r ) :} Eliminate circular waiting by ordering all locks (or semaphores, or resoruces). All code grabs locks in a predefined order. Problems?Maintaining global order is difficult, especially in a large project.Global order can f orce a client to g rab a lock earlier than it 12 would like, tying up a resource for too long.Deadlock is a global property, but lock manipulation is local. Deadlock Detection & RecoveryRecovering from deadlock R1R2R3R4 Abort all deadlocked processes & reclaim their resources Abort one process at a time until all cycles in the Where to start? P4 1 2 3 5P5 Select low priority processProcesses with most allocation of resources Caveat: ensure that system is in consistent state (e.g., transactions) Checkpoint processes periodically; rollback processes to checkpointed state Dealing With DeadlockDeadlock avoidance Examine each resource request and determine whether or not granting the request can lead to deadlock Define a set of vectors and matrices that characterize the resource allocation state matrix R1R2R3... RrP1P2 P 3 n... n1,n Define a set of vectors and matrices that characterize the current state of all resources and processesmaximum claim matrix Max ij = the maximum number of units the number of units of held by process n1, n2, n3, ..., nr� P 3 Pp n 3 , 1 npp,r.......... the maximum number of units of resource that the process ever require simultaneouslythe number of units ofthat are unallocated Dealing With DeadlockDeadlock detection & recovery What are some problems with the banker’s algorithm?Very slow O(n Tooslowtorunoneveryallocation.Whatelsecanwedo? Too What Deadlock prevention and avoidance:Develop and use resource allocation mechanisms and protocols that Deadlock detection and recovery: y stem deadlock and deal with it 15 Detect that a set of processes are deadlockedRecover from the deadlock