Region Analysis for Race Detection

  • Authors:
  • Helmut Seidl;Vesal Vojdani

  • Affiliations:
  • Lehrstuhl für Informatik II, Technische Universität München, Garching b. München, Germany D-85748;Lehrstuhl für Informatik II, Technische Universität München, Garching b. München, Germany D-85748

  • Venue:
  • SAS '09 Proceedings of the 16th International Symposium on Static Analysis
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Automatic race detection of C programs requires fast, yet sufficiently precise analysis of dynamic memory. Therefore, we present a region-based pointer analysis which seeks to identify disjoint regions of dynamically allocated objects to ensure that write accesses to the same region are always protected by the same mutexes. Our approach has been implemented within the interprocedural analyzer of concurrent C programs GobLint and we have successfully applied it on code from the Linux kernel, such as the access vector cache. This code relies on a synchronized hash table where an array of doubly linked lists is protected by an array of locks.