200 likes | 356 Views
Scalable Multi-core Model Checking Fairness Enhanced Systems. Yang Liu, Jun Sun and Jin Song Dong School of Computing National University of Singapore. Outline. General Introduction to PAT http://pat.comp.nus.edu.sg Motivation: Parallel Model Checking Model Checking with Fairness
E N D
Scalable Multi-core Model Checking Fairness Enhanced Systems Yang Liu, Jun Sun and Jin Song Dong School of Computing National University of Singapore
Outline • General Introduction to PAT • http://pat.comp.nus.edu.sg • Motivation: Parallel Model Checking • Model Checking with Fairness • Experiment Results • Conclusion
Model Checking • Determining whether a model satisfies a property by the means of exhaustive searching. Model Model Checker Counterexample! Property
Model Checking Works! • Three researchers won Turing Award 2007 for their pioneer work on model checking! • Intel Core i7 processor is verified by symbolic model checking completely without executing a single test case! • The Slam project from Microsoft successfully detected many bugs in many driver software!
PAT: Motivation • We aim to develop a self-contained framework for formal specification and verification of compositional systems which involves, • concurrency, • real-time, • complex data structures and operations, • complicated control flows, • and etc.
Motivation • Model checking is limited by state space explosion. • We do have multiple cores nowadays!
Safety • “Something bad never happens” • Reachability analysis • Depth-first search • Breadth-first search
Liveness • “Something good eventually happens” • Liveness checking = bad loop searching • Nested depth-first-search • SCC-based algorithms
Fairness • Liveness often requires fairness! • Process level weak/strong fairness • Event level weak/strong fairness • Strong global fairness
Verification under Fairness • Automata-based LTL model checking • weak fairness: SCC search • strong fairness: strongly connected sub-graph search • strong global fairness = terminal SCC search
Verification under Fairness • A lasso is counterexample if and only if the loop is fair and it fails the liveness property. • It is (process-level) weak fair iff there is NO process which is always enabled during the loop and never made any progress. • It is (process-level) strong fair iff there is NO process which is enabled during the loop and never made any progress. • It is strong global fairness iff …
Sequential Algorithm A: Find SCC-0 True Is Not Fair B: Check if SCC-0 is fair Is Fair C: Generate Counterexample
Parallel Algorithm A0 B0 A1 B1 A2 B2 Thread 1 Thread 2 Thread 3 Thread 4
Parallel Algorithm A0 A1 A2 Thread 1 B0 B1 Thread 2 B2 Thread 3 Thread 4
Effectiveness • Overhead – negligible • Based on shared-memory architecture. • Depends on how expensive checking whether a SCC satisfies the fairness constraint • Weak fairness: linear in the number of transitions • Strong fairness: bounded by #states * #transitions • Global fairness: linear in the number of transitions
Conclusion • A simple way of making use of multi-cores for model checking with fairness. • The technique is available in PAT. • http://pat.comp.nus.edu.sg • Related work • Spin’s liveness checking algorithm for dual-core systems • Barnat et al. multi-core LTL model checking • MAP, OWCTY, NEGC, OBF