1 / 25

Sequential reductions for verifying serializability

Sequential reductions for verifying serializability. Hagit Attiya Technion & EPFL G. Ramalingam MSR India Noam Rinetzky University of London. The goal. Verify concurrent data structures Pre-execution static analysis E.g., linked list with hand-over-hand locking

media
Télécharger la présentation

Sequential reductions for verifying serializability

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Sequential reductions for verifying serializability Hagit Attiya Technion & EPFL G. Ramalingam MSR India Noam Rinetzky University of London

  2. The goal Verify concurrent data structures • Pre-execution static analysis E.g., linked list with hand-over-hand locking • memory safety (no memory leaks) • shape invariants (it’s a list) • “atomicity” (serializability) Find sequential reductions • Consider only sequential executions • But conclude that properties hold in all executions

  3. Back-of-envelop estimate of gain Static analysis of a linked-list algorithm [Amit, Rinetzky, Reps, Sagiv, Yahav, CAV 2007] • Verifies e.g., memory safety, sortedness, pointed-to by a variable, heap sharing

  4. ~ ~ ~ ~ ~ ~ ~ ~ ~ Serializability [Papadimitriou ‘79] interleaved execution operation ~ thread-local views complete non-interleaved execution

  5. Serializability assists verification Concurrent code M Π = all executions of M cni-Π = all complete non-interleaved executions of M  Π φ = some thread-local property If M is serializable Then Π ⊨ φ cni-Π ⊨ φ

  6. How do we know that M is serializable, w/o considering all executions? I.e., by checking only cni-executions If M is serializable Then Π ⊨ φ cni-Π ⊨ φ

  7. Special (and common) case: Disciplined programming with locks Guard access to data with locks Access data only after getting a lock for it (well-lockedness) Follow a locking protocol that guarantees conflict serializability E.g., two-phase locking (2PL) or tree locking (TL)

  8. H Two-phase locking [Papadimitriou `79] • Locks acquire (grow) phase followed by locks release (shrink) phase • No lock is acquired after some lock is released t2 t1 t1 t1 t1

  9. H Tree (hand-over-hand) locking [Bayer & Scholnick ‘77] [Kedem & Sliberschatz ‘76] [Smadi ‘76] • Except for the first lock, acquire a lock only when holding the lock on its parent • No lock is acquired after being released t2 t1 t1 t1

  10. H Tree (hand-over-hand) locking [Bayer & Scholnick ‘77] [Kedem & Sliberschatz ‘76] [Smadi ‘76] • Except for the first lock, acquire a lock only when holding the lock on its parent • No lock is acquired after being released t2 t1 t1 t2

  11. void p() { acquire(B) B = 0 int b = B release(B) if (b) Y = 2 acquire(A) } void q() { acquire(B) B = 1 release(B) } Not two-phase locked But only in interleaved executions Yes! • for databases • concurrency control monitor ensures that M follows the locking policy at run-time  M is serializable No! • for static analysis • no central monitor How to (statically) verify that M follows a locking policy?

  12. Our approach Applies for local conflict-serializable (LCS) locking protocols • Use only thread-local information (see next) E.g., two phase locking, tree locking, (dynamic) DAG locking… But not protocols that rely on a concurrency control monitor!

  13. Thread-local properties A thread-owned view contains the values of thread’s local variables & global variables locked by it A property φ is thread-localif it • Can be expressed in terms of thread-owned views • Is prefix closed A thread-local property of an execution holds in every execution indistinguishable from it

  14. non-interleaved execution Our contribution: Easy step cni-Π = all complete non-interleaved executions of M  Π Two phase locking Tree locking Dynamic tree locking Dynamic DAG locking For any LCS locking policy LP Π ⊨ LP ni-Π ⊨ LP For any thread-local property φ Π ⊨ φ ni-Π ⊨ φ

  15. Ni-reduction: Proof sketch If all ni-executions follow LP  all executions follow LP σ is the shortest execution that does not follow LP • σ’ follows LP, guarantees conflict-serializability • there is a ni-execution that is “equivalent” to σ’ (t,e) σ σ’

  16. σ’ni (t,e) σ’ (t,e) Ni-reduction: Proof sketch (cont.) • there is a ni-execution that is “equivalent” to σ’ • there is a ni-execution that is “equivalent” to σwhere LP is violated σni

  17. almost complete non-interleaved execution Further reduction acni-Π = all almost-complete non-interleaved executions of M  Π For any LCS locking policy LP Π ⊨ LP acni-Π ⊨ LP

  18. Acni-reduction: A complication Need to argue about termination Observe Y == 1 & violates 2PL Y is set to 1 & the method enters an infinite loop int X=0, Y=0 void p() { acquire(Y) y = Y release(Y); if (y ≠ 0) acquire(X) X = 3 release(X) } void q() { if (random(5) == 3){ acquire(Y) Y = 1 release(Y) while (true) nop } }

  19. Acni-reduction: Termination  Can use sequential reduction to verify termination For any “terminating” LCS locking policy LP Π ⊨ LP acni-Π ⊨ LP

  20. Acni-reduction: Proof ideas Start from a ni-execution (and rely on the previous, ni-reduction to get there) Create its equivalent completion, if possible Not always possible, e.g., Does not access variables accessed by later threads v t1:lock(v), t1:lock(u), t2:lock(u) u

  21. Implications for statis analysis • Pessimistic analysis (over approximate) • Analyze a module from every possible state • Semi-optimistic analysis • Analyze a module only from states that occur after a sequence of modules ran one after the other • Optimistic analysis (precise) • Analyze a module only from states that occur after a sequence of modules ran to completion (one after the other) Ni-reduction Acni-reduction

  22. Initial analysis results Shape analysis of hand-over-hand lists * Does not verify sortedness of list and fails to verify linearizability in some cases Shape analysis of hand-over-hand trees(for the first time)

  23. What’s next? • Further extensions, esp., for software transactional memory • aborted transactions • non-locking non-conflict based serializability (e.g., using timestamps) • Combine with other reductions [Guerraoui, Henzinger, Jobstmann, Singh]

  24. And beyond… • The cost of verifying adherence to a locking policy in an acni- / ni-execution • Automatic insertion of lock acquire / release commands … Or fences?

  25. Thank you!

More Related