On the Notions of Bounded Bypass, and How to Make any Deadlock-Free MUTEX Protocol Satisfy One of Them

Abstract: In the literature on mutual exclusion, bounded bypass has been used for a long time as a strengthening of starvation-freedom, but, to the best of our knowledge, it still lacks a satisfying definition as a liveness property on its own. Moreover, we have encountered MUTEX protocols for which this notion needs to be slightly weakened in order to be met. To solve these issues, we first provide a formal definition of bounded bypass (that also corrects a previous definition from Raynal) and then introduce the notions of post-doorway and intermittent bounded bypass, two liveness properties that lie between starvation-freedom and bounded bypass. Essentially, intermittent bounded bypass weakens bounded bypass by ignoring the possible bypasses that may happen during the execution of a certain finite set of write operations to shared registers. Orthogonally, post-doorway bounded bypass ignores the bypasses that may happen during a finite initial phase of the lock protocol. Furthermore, we study an algorithm proposed by Yoah Bar-David in 1998 to enhance the liveness properties of any deadlock-free MUTEX protocol and prove that: (1) in the setting of atomic registers, this algorithm upgrades any deadlock-free mutual exclusion protocol to a bounded bypass one, with a bound that is quadratic in the number of processes; and (2) in the setting of safe and regular registers, the very same algorithm ensures the intermittent version of bounded bypass, still with a quadratic (but slightly different) bound. Finally, we provide logical formulae for the different notions of bounded bypass defined in this paper and use them to confirm all claims made here, by using model checking. This had a positive impact on the theoretical development of the work, since it allowed us to identify and correct small mistakes/ambiguities in definitions and proofs.

• The paper presents formal definitions of bounded bypass and introduces new liveness conditions to bridge the gap between starvation-freedom and strict bypass bounds.
• It proposes a liveness-augmentation algorithm that ensures a process enters the critical section after at most n(n-1)-1 competitions in atomic memory settings.
• The study includes mechanized verification using mCRL2 and compares performance across atomic, regular, and safe memory models to address practical deployment challenges.

Formal Analysis and Augmentation of Liveness for Mutual Exclusion: Bounded Bypass Notions

Introduction

The paper "On the Notions of Bounded Bypass, and How to Make any Deadlock-Free MUTEX Protocol Satisfy One of Them" (2606.19003) presents an extensive study of liveness properties for mutual exclusion (MUTEX) protocols, focusing on the concept of bounded bypass (also known as bounded waiting). The authors provide a rigorous formalization of bounded bypass, introduce two new intermediate liveness conditions, and analyze a generic liveness-augmentation algorithm under various shared memory models. Their results clarify misconceptions in past literature and yield precise, mechanized bounds for important MUTEX algorithms.

Formalization of Bounded Bypass and Related Notions

Issues with Literature Definitions

The paper identifies significant ambiguities and errors in classical definitions of bounded bypass. Notably, previous formulations (e.g., by Raynal and Taubenfeld) do not guarantee starvation-freedom, as they permit pathological executions where processes can wait indefinitely without explicit progress. The authors pinpoint the missing requirement: initiation of bounded bypass must be tied to a process's visible indication of intent to enter the critical section, after which other processes must not indefinitely bypass it.

New Definitions

The authors propose refined liveness properties to systematically fill the spectrum between deadlock-freedom and strict bounded bypass:

• Bounded Bypass: After completing the first instruction of the lock entry section (a mandatory shared-memory write), a process is guaranteed to enter the critical section after losing at most $f(n)$ competitions to other processes. This definition precisely marks the point where the process's intent becomes visible to the system, avoiding protocol-specific ambiguities.
• Post-Doorway Bounded Bypass: Bypass bounds only start after completion of a protocol-defined "doorway"—an initial, bounded-length entry segment. This allows liveness bounds to ignore unbounded pre-doorway delays.
• Intermittent Bounded Bypass: Bypass counting is suspended during a finite set of "interrupting" assignments (typically certain writes), and only resumes outside these. This abstraction is motivated by weak memory models where certain atomicity cannot be strictly enforced.

The hierarchy:

$$\text{Bounded Bypass} \implies \text{Post-Doorway Bounded Bypass} \implies \text{Intermittent Bounded Bypass} \implies \text{Starvation-Freedom} \implies \text{Deadlock-Freedom}$$

is established, with strictness demonstrated via classic protocols (e.g., Dekker, Anderson).

Memory Models: Atomic, Regular, and Safe

The analysis is sensitive to the strength of the memory abstraction:

• Atomic registers: Reads/writes are linearizable; suitable for many textbook analyses.
• Regular registers: Reads concurrent with writes may return either the old or new value.
• Safe registers: The weakest, where overlapping reads and writes can return arbitrary in-domain values.

The distinction is critical because certain liveness properties—particularly variants of bounded bypass—cannot be guaranteed in safe/regular memory, as demonstrated by precise protocol counterexamples.

Augmenting Deadlock-Free Protocols: The Bar-David Construction

A central technical contribution is a rigorous analysis of the liveness-augmentation protocol attributed to Yoah Bar-David and popularized by Taubenfeld and Raynal. Given any deadlock-free protocol (DLF), this wrapper protocol enhances it by:

1. Adding process-local "intent" flags and a global, shared "turn" pointer.
2. Requiring that each process waits either for its turn or until the current turn-holder withdraws.
3. Delegating to the original DLF for actual mutual exclusion once the augmented conditions are satisfied.

Analysis for Atomic Memory

In atomic registers, the augmented protocol guarantees strong bounded bypass. The authors show, through detailed worst-case executions and inductive invariants, that a process can be bypassed by at most $n(n-1)-1$ other critical section entries after setting its intent flag (where $n$ is the number of processes). This bound is proven tight, displayed through an execution trace construction. Formally:

Theorem: After a process sets its entry flag, it will enter the critical section after losing at most $n(n-1)-1$ competitions.

This is a substantial strengthening over mere starvation-freedom and allows reasoning about tight response-time bounds in fixed-process settings.

Safe and Regular Memory

When shared memory is weaker, e.g., safe registers, bounded bypass is strictly unattainable, as reads during concurrent writes can delay a process for an unbounded time if adversarial. However, the authors show the protocol satisfies intermittent bounded bypass: after a bounded ($n$) number of "interrupting" writes, a process suffers at most $n^2-2$ effective bypasses. This matches the best possible in this model and is again proven tight.

Mechanized Verification and Logical Characterizations

To solidify these claims, the authors develop $\mu$-calculus formulae capturing the various bypass properties, parameterized by the protocol structure and memory model. These enable model checking (using mCRL2) of MUTEX protocol implementations for small $n$, verifying both classic and new claims about bypass bounds. The formalizations not only confirm hand proofs but also lead to corrections of earlier (non-tight) results and highlight subtle behavioral corner cases (e.g., precise identification of "interrupting assignments" in non-atomic models).

Implications and Future Directions

This work resolves longstanding ambiguities surrounding liveness strengthening properties in MUTEX algorithms. The formalization of intermediate bypass notions and their mechanized verification bridges a gap between practical protocol engineering and theoretical guarantees.

• Theoretical impact: The clarified hierarchy enables clean separation and selection of liveness conditions, facilitating compositional reasoning and reduction-based design in concurrent algorithm research.
• Practical impact: Verified bounds offer practitioners precise, implementation-agnostic guarantees for performance-critical mutual exclusion components, especially under hardware with weak consistency.

Further research directions include:

• Extending the formalization and augmentation approach for more general synchronization primitives (e.g., condition variables, semaphores).
• Exploring tighter lower bounds and protocol synthesis for circumventing bounded bypass impossibility in ultra-weak memory.
• Investigating parameterized verification techniques scalable beyond 3 processes.

Conclusion

The paper delivers a comprehensive, correction-oriented treatment of bounded bypass and its relatives in MUTEX protocol design. By pinpointing formal weaknesses in previous definitions, introducing intermediate liveness conditions, and providing tight, mechanized bounds for a canonical liveness-augmentation protocol, it substantially advances the formal understanding and automated verification of progress properties in concurrent algorithms. These results directly inform both theoreticians and implementers working on the correctness and performance of synchronization under realistic memory models.