Static Analysis of Bottom Tier Sensor Networks

In CENS ARR , 1 pages , CENS , May 2007.

NESL Technical Report #: TR-UCLA-NESL-200705-01

Developing software for the lowest tier of sensor networks is difficult. Directly manipulating low level hardware motivates using languages notorious for providing minimal safety to developers. Nontrivial software logic, ranging from routing protocols to to sensor driver state, are also common in these bottom tier systems. Current sensor network research is exploring the use of traditional static analysis to help minimize these challenges. Static analysis of sensor network systems is often limited by the reactive nature of these systems. A senor node's physical ties to the environment create a reactive system where one of many different events can fire and any time. Direct application of a static analysis to these reactive systems requires overly conservative analysis to handle the nondeterministic event ordering. This work explores extending standard static analysis with event ordering assumptions built into the system being analyzed. We present an extended version of the Lighthouse memory checker for the SOS operating system. Lighthouse uses dataflow analysis to look for incorrect use of dynamic memory within a single event handler. Our work uses event ordering specifications provided by system developers to augment this analysis. The refined analysis is able to reason about dynamic memory usage across event firings. We apply the extended Lighthouse memory analysis to both user code and kernel code distributed with the SOS operating system.

• Roy S Shea
• Todd Millstein
• Rupak Majumdar
• Mani B Srivastava

• PDF file of presentation

• Lighthouse : Lighthouse

Poster