Model-Checking the Linux Virtual File System

Galloway, Andy; Lüttgen, Gerald; Mühlberg, Jan Tobias; Siminiceanu, Radu

doi:10.1007/978-3-540-93900-9_10

Cited by 20 publications

(21 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a slightly recent work [29], a small specification language PL that models business process, is translated into linear temporal formulas, to check the deadlock freedom of interacting business processes of an airline ticket reservation system. In another work [30], We have also used symbolic model checking to check the locking mechanism of the Linux Virtual File System (VFS) by extracting abstract models from the Linux kernel. Our work analyses specifications based on access permissions and typestates, that according to best knowledge of authors is not the subject of previous work.…”

Section: Related Workmentioning

confidence: 99%

A case study on the lightweight verification of a multi-threaded task server

Cataño

Ahmed²,

Siminiceanu

et al. 2014

Science of Computer Programming

Self Cite

View full text Add to dashboard Cite

We present a case study of verifying the design of a commercial multi-threaded task server (MTTS), developed by the Novabase company, used for massively parallelising computational tasks. In a first stage, we employed the Plural tool, which is designed to perform lightweight verification of Java programs using a Data Flow Analysis (DFA) framework, to specify and verify the MTTS. We wrote the Plural specification for the MTTS based on the code developed by Novabase, its informal documentation, and our discussions with Novabase engineers, who validated our understanding of the MTTS application. The Plural specification language is based on typestates and access permissions. In a second stage, we developed the Pulse tool that enhances the analysis performed by Plural, and used the tool on the MTTS specifications. Pulse translates Plural specifications into an abstract state-machine model that captures the semantics of all the possible concurrent programs implementing the given specifications, and uses the evmdd-smc symbolic model-checker to verify the machine model. The experimental results on the MTTS specification show that the exhaustive model-checking approach scales reasonably well and is efficient at finding errors in specifications that were not previously detected with the Data Flow Analysis (DFA) capabilities of Plural.

show abstract

Section: Related Workmentioning

confidence: 99%

A case study on the lightweight verification of a multi-threaded task server

Cataño

Ahmed²,

Siminiceanu

et al. 2014

Science of Computer Programming

Self Cite

View full text Add to dashboard Cite

show abstract

“…Galloway et al [GLMS09] verify the existing Linux Virtual File System (VFS) by extracting and validating a model from an available implementation of VFS. Yang et al [YTEM06] build their own model checker "FiSC" to find serious file system errors.…”

Section: Related Workmentioning

confidence: 99%

Formalizing a hierarchical file system

Hesselink

Lali

2012

Form. Asp. Comput.

View full text Add to dashboard Cite

Abstract. An abstract file system is defined here as a partial function from (absolute) paths to data. Such a file system determines the set of valid paths. It allows the file system to be read and written at a valid path, and it allows the system to be modified by the Unix operations for creation, removal, and moving of files and directories. We present abstract definitions (axioms) for these operations. This specification is refined towards a pointer implementation. The challenge is to have a natural abstraction function from the implementation to the specification, to define operations on the concrete store that behave exactly in the same way as the corresponding functions on the abstract store, and to prove these facts. To mitigate the problems attached to partial functions, we do this in two steps: first a refinement towards a pointer implementation with total functions, followed by one that allows partial functions. These two refinements are proved correct by means of a number of invariants. Indeed, the insights gained consist, on the one hand, of the invariants of the pointer implementation that are needed for the refinement functions, and on the other hand of the precise enabling conditions of the operations on the different levels of abstraction. Each of the three specification levels is enriched with a permission system for reading, writing, or executing, and the refinement relations between these permission systems are explored. Files and directories are distinguished from the outset, but this rarely affects our part of the specifications. All results have been verified with the proof assistant PVS, in particular, that the invariants are preserved by the operations, and that, where the invariants hold, the operations commute with the refinement functions.

show abstract

“…During the last years, three verification environments were build in order to define verification tasks from Linux drivers: the Linux Driver Verification project 1 [23], the Avinux project [27], and the DDVerify project 2 [32]. The Linux code base is a popular source for verification tasks [17,22,24,25]. Linux drivers provide a unique combination of specific characteristics that attract researchers and practitioners to challenge their tools.…”

Section: Overviewmentioning

confidence: 99%

Linux Driver Verification

Beyer

Petrenko²

2012

Leveraging Applications of Formal Methods, Verification and Validation. Applications and Case Studies

View full text Add to dashboard Cite

Abstract. Linux driver verification is a large application area for software verification methods, in particular, for functional, safety, and security verification. Linux driver software is industrial production code -IT infrastructures rely on its stability, and thus, there are strong requirements for correctness and reliability. This implies that if a verification engineer has identified a bug in a driver, the engineer can expect quick response from the development community in terms of bug confirmation and correction. Linux driver software is complex, low-level systems code, and its characteristics make it necessary to bring to bear techniques from program analysis, SMT solvers, model checking, and other areas of software verification. These areas have recently made a significant progress in terms of precision and performance, and the complex task of verifying Linux driver software can be successful if the conceptual state-of-the-art becomes available in tool implementations.

show abstract

Model-Checking the Linux Virtual File System

Cited by 20 publications

References 12 publications

A case study on the lightweight verification of a multi-threaded task server

A case study on the lightweight verification of a multi-threaded task server

Formalizing a hierarchical file system

Linux Driver Verification

Contact Info

Product

Resources

About