Using static code analysis to support DO-178b certification - Embedded.com

Using static code analysis to support DO-178b certification

In this Product How-To, Paul Anderson of GrammaTech takes you step by step through the DO-178B and how use his company’s static analysis tools to support the safety-critical software requirements of the specification.

Software is playing a growing role in airborne systems. These demanding environments require highly-reliable software, which is a challenge to develop. The DO-178B standard assists by providing production guidelines for software and equipment that must “comply with airworthiness requirements”.

This standard, also known as “Software Considerations in Airborne Systems and Equipment Certification” [1], is published by RTCA, Inc and developed jointly with EUROCAE, the European Organization for Civil Aviation Equipment.

Compliance with DO-178B involves meeting many requirements, yet the standard takes a flexible approach to meeting them. This enables developers of airborne software to employ a variety of techniques and technologies. Static analysis can be used as part of a comprehensive approach to assurance. This document shows how static analysis in general, and tools from GrammaTech in particular, can be applied as part of the DO-178B compliance process.

GrammaTech’s static analysis tools CodeSonar and CodeSurfer can be used together and separately to support an organization’s DO-178B activities. CodeSonar performs whole-program, interprocedural analysis on C and C++ source code, identifying programming bugs that can result in system crashes, memory corruption, and other serious problems.

It includes numerous workflow automation features, including an API for custom integrations and support for extensions that add custom checks. CodeSurfer is a program-understanding tool that does precise source code analysis, calculating a variety of representations that can be explored through the graphical user interface or accessed through an optional programming API. CodeSonar finds bugs automatically, while CodeSurfer makes manual review of code easier and faster.

The use of CodeSonar and CodeSurfer is most applicable to DO-178B chapters 6, 7, 11, and 12, so these are the chapters discussed in this article. It is worth noting, however, that GrammaTech static analysis tools can also support various activities and objectives from other chapters.

In particular, CodeSonar can provide value throughout the Software Development Process (Chapter 5), and the completion of a CodeSonar analysis – perhaps with a limit on the permissible number of warnings – is a useful transition criterion (Chapter 4) in many cases.

The following typographical conventions are used in this article.

* CodeSonar warning class names are rendered in italic, sans-serif font: Null Pointer Dereference . If the warning class is disabled by default, the class name is marked with an asterisk: Recursive Macro* .
* Software life cycle data artifacts defined in DO-178B Chapter 11 are capitalized: Software Design Standards.
* Direct quotes from the DO-178B document are formatted like the following: “The rapid increase in the use…”

DO-178B Chapter 6: Software Verification Process
“The purpose of the software verification process is to detect and report errors that may have been introduced during the software development processes.”

CodeSonar and CodeSurfer bring powerful static analysis capabilities to the Software Verification Process. DO178-B sections 6.3 (Software Reviews and Analyses) and 6.4 (Software Testing Process) describe verification objectives that are well supported by these tools, as described below.

6.3 Software Reviews and Analyses
“Reviews and analyses are applied to the results of the software development processes and software verification process. One distinction between reviews and analyses is that analyses provide repeatable evidence of correctness and reviews provide a qualitative assessment of correctness.”

It is useful to think of both CodeSonar and CodeSurfer as performing analyses whose outputs can be used to inform manual reviews.

6.3.3. Reviews & analysis of software architecture
CodeSonar and CodeSurfer provide complementary support for architecture review. CodeSurfer provides architecture visualization and sophisticated query functionality, while CodeSonar’s built-in checks and extension capabilities allow many architecture problems to be flagged automatically.

“a. Compatibility with the high-level requirements: The objective is to ensure that the software architecture does not conflict with the high-level requirements, especially functions that ensure system integrity, for example, partitioning schemes.”

Architecture reviewers can use CodeSurfer to inspect structural properties of the software at a number of levels of detail. The CodeSurfer GUI allows interactive browsing of the various structural elements, while the optional programming API provides automation capabilities and broad potential for customization. Both GUI and API provide powerful built-in queries, such as slicing, that can be used to check the integrity of partitioning schemes.

“b. Consistency: The objective is to ensure that a correct relationship exists between the components of the software architecture. This relationship exists via data flow and control flow.”

CodeSurfer allows users to examine both control flow and data flow from a number of perspectives. Among those most useful for inspecting the relationship between software architecture components are the control flow graph and call graph.

The results of control- and data-related queries can be superimposed on these graphs, providing a straightforward depiction of the dependencies involved.

“c. Compatibility with the target computer: The objective is to ensure that no conflicts exist, especially initialization, asynchronous operation, synchronization and interrupts, between the software architecture and the hardware/software features of the target computer.”

Several built in CodeSonar checks can detect potential compatibility problems, including Uninitialized Variable, Double Lock, Double Unlock, Try-lock that will never succeed, and Shift Amount Exceeds Bit Width.

“d. Verifiability: The objective is to ensure that the software architecture can be verified, for example, there are no unbounded recursive algorithms.”

While demonstrating the complete absence of unbounded recursion is of course provably impossible in general, CodeSonar offers several checks that help reduce the likelihood that infinite recursion will occur.

If the system architects choose to forbid recursion entirely, the Recursion* and Recursive Macro* checks can be enabled in order to detect violations. If recursion is to be permitted, the Excessive Stack Depth* check can identify potential cases of runaway recursion.

Similarly, CodeSonar’s Potential Unbounded Loop* check does not and cannot detect all infinite iterative loops in all programs, but will identify many loops whose boundedness cannot be established.

“e. Conformance to standards: The objective is to ensure that the Software Design Standards were followed during the software design process and that deviations to the standards are justified, especially complexity restrictions and design constructs that would not comply with the system safety objectives.”

CodeSurfer computes many standard complexity metrics, whose values can be compared against limits set out in the Software Design Standards. The metrics are computed at the project, file, and function levels, allowing designers to easily identify problem areas.

If system safety objectives are such that specific design constructs must be restricted or forbidden, custom CodeSonar checks can be written to issue warnings whenever those constructs are encountered.

“f. Partitioning integrity: The objective is to ensure that partitioning breaches are prevented or isolated.”

CodeSurfer allows users to trace both control and data through the various possible executions of a program. Users can obtain answers to such questions as “can data in region A influence execution in region B?”, “is there an execution path between region C and region D?”, and “what are the callers of function F?”

The optional Path Inspector extension to CodeSurfer can be invoked to confirm that execution flow from one part of the program cannot reach another part (and, if necessary, vice versa). Custom CodeSonar checks can also be used to detect partitioning breaches automatically.

Explicit control and data flow may not be the only avenues for breaching partitioning requirements. In some cases other channels, such as the file system, must be considered.

Designers may elect to forbid use of the file system entirely, in which case custom CodeSonar checks for uses of file-related functions and data types will be extremely useful.

Alternatively, file usage may be permitted under restricted circumstances, in which case the custom checks will need to verify that the appropriate conditions hold whenever files are used.

6.3.4. Reviews and Analyses of the Source Code
Source code analysis is the most significant strength of both CodeSonar and CodeSurfer. A full accounting of the reviews and analyses supported by these products is not possible in this space; instead we outline some key features.

“b. Compliance with the software architecture: The objective is to ensure that the Source Code matches the data flow and control flow defined in the software architecture.”

CodeSurfer provides powerful functionality for visualizing data and control flow in software, and for answering complex queries about both. Users can compare the flow computed by CodeSurfer to that defined in the software architecture to identify any discrepancies. CodeSurfer’s pointer analysis means that pointer aliasing and indirect function calls are accurately accounted for.

If control flow requirements specify a particular operation sequence that must be adhered to, CodeSonar can be extended with a custom check that issues warnings whenever the operations appear out of order.

“c. Verifiability. The objective is to ensure that the Source Code does not contain statements and structures that cannot be verified and that the code does not have to be altered to test it.”

Both CodeSonar and CodeSurfer observe the normal build process for a software project and use the information thus gained to construct an internal representation that is then subjected to various analyses. The production code does not need to be altered in any way.

“d. Conformance to standards: The objective is to ensure that the Software Code Standards were followed during the development of the code, especially complexity restrictions and code constraints that would be consistent with the system safety objectives….”

CodeSonar provides significant support for upholding coding standards, both through built-in checks for standard rule sets such as Power of Ten[4] and through a rich API that allows users to implement custom checks in support of local code standards. 11.7 Software Design Standards and 11.8 Software Code Standards detail the ways in which CodeSonar can support and enforce a wide range of software standards.

“f. Accuracy and consistency: The objective is to determine the correctness and consistency of the Source Code, including stack usage, fixed point arithmetic overflow, resource contention, worst-case execution timing, exception handling, use of uninitialized variables or constants, unused variables or constants, and data corruption due to task or interrupt conflicts.”

CodeSonar performs a large number of correctness and consistency checks, including those for the following warning classes:

* Excessive Stack Depth
* Buffer Overrun, Buffer Underrun
* Type Overrun, Type Underrun
* Cast Alters Value, Integer Overflow of Allocation Size
* Uninitialized Variable
* Unused Value
* Deadlock, File System Race Condition

6.4 Software Testing Process
CodeSonar’s static analysis can detect several of the problems listed as identifiable through requirements-based testing methods, and can help resolve questions arising from test coverage analysis.

6.4.3. Requirements-Based testing methods .
Static analysis is not a substitute for testing but a complement to it, and in some cases will find problems that tests miss.

“a. Requirements-Based Hardware/Software Integration Testing”
“b. Requirements-Based Software Integration Testing”.
“c. Requirements-Based Low-Level Testing”

CodeSonar can detect many of the “typical errors revealed by [these] testing method[s]” early in the software development process. The relevant CodeSonar warning classes are shown in the following table below.

6.4.4. Test Coverage Analysis
“Test coverage analysis is a two step process, involving requirements-based coverage analysis and structural coverage analysis. The first step analyzes the test cases in relation to the software requirements to confirm that the selected test cases satisfy the specified criteria. The second step confirms that the requirements-based test procedures exercised the code structure.”

CodeSonar and CodeSurfer can be useful in the resolution phase of test coverage analysis, if this is required.

6.4.4.3 Structural Coverage Analysis Resolution
Structural coverage analysis may reveal code structure that was not exercised during testing. Resolution would require additional software verification process activity. This unexecuted code may be the result of:…

“…c. Dead code”
“…d. Deactivated code”

In some cases, a test suite may be unable to exercise a particular code region because the region is in fact unreachable. If structural coverage analysis resolution is to include a diagnosis of dead or deactivated code, it can be useful to support this diagnosis with static analysis results.

CodeSonar detects unreachable code as part of its standard analysis suite. Distinct warning classes Unreachable Call, Unreachable Computation, Unreachable Conditional, Unreachable Control Flow, and Unreachable Data Flow identify the most important element present in each unreachable region, providing the user with additional information about the consequences of not executing the code.

In some cases code is unreachable because it is guarded by a conditional statement that can only ever evaluate one way. CodeSonar’s Redundant Condition warning class is designed to identify such situations.

CodeSurfer provides complementary information about code reachability at several levels of detail. Users can determine whether there are entire functions that are never called using the program call graph. At the statement level, a point-mode forward slice from the program entry point will compute all code that can be reached; any code not in this set is therefore unreachable.

Note that both CodeSonar and CodeSurfer work by analyzing a specific build of a software project. To establish that code is unreachable under all applicable build configurations (for example, unreachable on all target platforms, or under all permitted preprocessor settings), the analyses must be applied to all these builds and the results compared.

DO-178B Chapter 7: Software Configuration Management Process
While CodeSonar and CodeSurfer are not themselves software configuration management tools, they can interact with the Software configuration management process in several useful ways. GrammaTech tools can be used in support of the following software configuration management process activities.

7.2.2 Baselines and Traceability
Where unambiguous labels have been assigned to configuration items, CodeSonar provides several powerful mechanisms for applying corresponding labels to analysis artifacts generated for those items. When a labeled configuration item is analyzed with CodeSonar:

* The analysis can be annotated with the same label, either directly or retroactively.

* A user can manually annotate all warnings issued by the analyses with the label (all at the same time, if desired).

* A custom warning processor can automatically annotate each warning issued by the analysis with the label. With further customization, annotation can be restricted to a particular subset of the warnings, for example only those that were issued for the first time (that is, not seen in any previous analysis).

The flexible search in CodeSonar includes functionality that allows users to find all warnings that have a particular annotation, or that were issued by an analysis with a particular annotation. The annotation capability thus provides direct traceability between a particular configuration item and the analysis results for that item.

From time to time it may be necessary to analyze projects in which different components have different configuration identifiers. CodeSonar can provide traceability even in this case: the only differences are that the analysis annotation will need to include all relevant labels, and the warning processor (or person) responsible for annotating warnings will need to take each warning’s location into account when assigning its label.

Similarly, when a baseline is established for a configuration item, CodeSonar annotations can be used to identify analysis artifacts associated with that baseline.

7.2.3 Problem Reporting, Tracking, and Corrective Action
Problem reporting and tracking are key roles of CodeSonar. Each analysis of a project issues a number (possibly zero) of warnings. As a project undergoes a cycle of modification and re-analysis over time, the CodeSonar hub builds up a historical record. Information is kept about each analysis, including the warnings issued, files analyzed, build settings on which the project was based, and annotations made.

At a finer-grained level, the “same” warning can be tracked over time, determining the analysis in which it first appeared and (if the underlying problem is successfully fixed) the analysis in which it first went away. CodeSonar keeps all annotations for each warning (except for cases where an analysis has been explicitly deleted), so a history of notes, priority determinations, owners, and assessments is built up over time.

The CodeSonar hub is not restricted to warnings issued by the CodeSonar analysis. For example, CodeSonar ships with an example script that uploads the warnings issued by a FindBugs[5] analysis to a CodeSonar hub; scripts that upload analysis results from other tools with different output formats are similarly possible.

Conversely, warnings issued by the CodeSonar analysis are not restricted to the CodeSonar hub. Custom warning processors can send analysis results to external tools, and can be applied either at analysis time or retroactively to individual warnings and groups of warnings. CodeSonar ships with a processor that creates a bug report concerning one or more CodeSonar warnings and submits it to a Bugzilla[2] database.

7.2.4 Change Control
“e. Throughout the change activity, software life cycle data affected by the change should be updated and records should be maintained for the change control activity.”

As described in 7.2.3 Problem Reporting, Tracking, and Corrective Action , CodeSonar by default keeps substantial records of past and present analysis results and the relationships between them. In cases where these records must be expanded upon, they can readily be exported in a variety of formats.

When traceability between a particular change and the analysis of the changed software is important, CodeSonar provides several mechanisms for making the connection explicit; these are described in 7.2.2 Baselines and Traceability .

7.2.5 Change Review
Just as CodeSonar warning processors can be used to apply annotations to warnings (7.2.2 Baselines and Traceability ) or to submit them to third-party trackers (7.2.3 Problem Reporting, Tracking, and Corrective Action ), they can be used to assign warnings to specified reviewers.

During the early development stages, a warning processor might be configured to assign each incoming warning to the engineer who last modified the source file containing the warning. Later, a warning processor might use the software level of the affected code to assign a priority to each incoming warning and then assign it to an engineer of appropriate seniority.

DO-178B Chapter 11: Software Life Cycle Data
“Data is produced during the software life cycle to plan, direct, explain, define, record, or provide evidence of activities.”

The Software Life Cycle Data defined by DO-178B include Plans with activities that can be accomplished using GrammaTech static analysis tools. In addition, several of the specified Standards can be supported or partially supported by these tools.

11.7 Software Design Standards
GrammaTech static analysis tools can be used to support and enforce a number of the Software Design Standards listed in DO-178B section 11.7.

“b. Naming conventions to be used.”

As described in11.8 Software Code Standards,

“c. Conditions imposed on permitted design methods, for example, scheduling, and the use of interrupts and event-driven architectures, dynamic tasking, re-entry, global data, and exception handling, and rationale for their use.”

It is straightforward to write custom CodeSonar checks on the use of global data or exception handling: such checks would inspect the internal representation and issue warnings whenever the relevant artifacts were observed (or observed outside permitted contexts).

“e. Constraints on design, for example, exclusion of recursion, dynamic objects, data aliases, and compacted expressions.”

CodeSonar ships with several checks that support design constraints, and can be extended to add others. The Recursion* and Recursive Macro* warning classes behave as one might expect. CodeSonar issues warnings whenever functions (or macros, respectively) are directly or indirectly recursive.

Classes Dynamic Allocation After Initialization* and Pointer Type Inside Typedef* are useful if the design constraints limit but do not fully exclude the use of dynamic objects.

If dynamic objects are to be fully excluded, users can construct custom CodeSonar checks that issue warnings whenever allocators are used. Similarly, the CodeSonar extension mechanisms allow users to write checks for violations of restrictions on data aliasing and on expression side-effects.

“f. Complexity restrictions, for example, maximum level of nested calls or conditional structures, use of unconditional branches, and number of entry/exit points of code components.”

As with the design constraints, CodeSonar has built in checks in support of some forms of complexity restriction and can be extended to add checks for others.

Excessive Stack Depth* warnings are issued when the function call stack exceeds a specified size: this expands on the notion of restricting call nesting by taking into account the size of the execution record for each call.

Several built in CodeSonar checks cover various forms of unconditional branch: Empty if Statement, Empty switch Statement, Redundant Condition, Goto Statement* , and Use of longjmp* . Users can create custom CodeSonar checks for violations of other complexity restrictions that might be included in a project’s Software Design Standards.

11.8 Software Code Standards
CodeSonar provides significant support for upholding coding standards, both through built-in checks for standard rule sets such as Power of Ten[4] and through a rich API that allows users to implement custom checks in support of local code standards.

“a. Programming language(s) to be used and/or defined subset(s).”

The CodeSonar API provides access to the abstract syntax trees (ASTs) generated for an analyzed software project. Users can leverage this access to author custom checks that report errors when forbidden program features are used.

“c. Naming conventions for components, subprograms, variables, and constants.”

Because the CodeSonar API provides full programmatic access to the internal representation generated for its analysis, users can readily create custom checks for naming convention violations and have these checks carried out as part of the CodeSonar analysis.

Checks can be targeted to specific program entities so can easily handle naming conventions that impose, for example, one set of rules governing the permissible names for variables and another for functions.

“d. Conditions and constraints imposed on permitted coding convention, such as the degree of coupling between software components and the complexity of logical or numerical expressions and rationale for their use.”

Various code complexity metrics have become standard in software engineering practice. CodeSurfer computes and reports a number of these standard metrics, including the McCabe[6] and Halstead[3] metrics. CodeSonar will incorporate this functionality in the near future.

The AST access provided by the CodeSonar API is useful here, as in previous cases. Given an expression complexity limit specified by the Software Code Standards, a user can create a custom check that issues a warning whenever the analysis encounters an expression whose complexity exceeds this limit.

DO-178B Chapter 12: Additional Considerations
Of the additional considerations discussed in DO-178B Chapter 12, CodeSurfer and CodeSonar are particularly applicable to section 12.1 (Use of Previously Developed Software ).

The understanding, use, and modification of existing software is a fundamental part of any software development effort, and GrammaTech’s static analysis tools provide advanced capabilities for accomplishing such tasks.

12.1 Use of Previously Developed Software
“The guidelines of this subsection discuss the issues associated with the use of previously developed software,…”

Two significant issues for users of previously developed software are understanding the software and trusting that it does not contain bugs and vulnerabilities.

CodeSurfer aids understanding of previously developed software and its interaction with the overall system at multiple levels, including data and control flow, call graphs, and effects on global values.

To help with establishing trust, CodeSonar can be applied to previously developed source code just as easily as it can be applied to code under current development. Problems, including security vulnerabilities such as buffer overflows – whether introduced accidentally or deliberately – can thus be identified and marked for elimination.

12.1.1 Modifications to Previously Developed Software
“This guidance discusses modifications to previously developed software where the outputs of the previous software life cycle processes comply with this document. Modification may result from requirement changes, the detection of errors, and/or software enhancements.”

Both CodeSonar and CodeSurfer are designed to support the iterative nature of real-world software development.

“d. The area affected by a change should be determined.”

In general, modifications to a program can affect areas far removed from the site of the modification, and sometimes in unexpected ways. CodeSurfer provides powerful tools for identifying and understanding these effects. Given a program region that has been modified, added, or removed, CodeSurfer users can obtain answers to questions such as:

* Which statements are affected by the changed region, and which statements affect it?
* How do the modifications affect the program call graph?
* How do control and data flow to and from the modified region?

The Path Inspector extension provides further support for impact analysis, allowing users to issue complex queries about execution flow in the modified program.

For example, a user might invoke Path Inspector to confirm that execution flow between two parts of the program does not pass through the modified region, or that the modified region conforms to initialization requirements.

“e. Areas affected by the change should be reverified considering the guidelines of section 6.”

CodeSonar and CodeSurfer support reverification of areas affected by changes in several important ways. Firstly, both products can be used to analyze partial programs. Once the areas affected by a change have been identified, those in charge of verification can thus choose to concentrate their resources on reverifying only those affected areas.

Secondly, CodeSonar and CodeSurfer both support incremental analysis, in which only those parts of the internal representation affected by changes in the code base are rebuilt and reanalyzed. This can offer substantial time savings when analyzing large projects.

Thirdly, the CodeSonar hub database provides a historical record of the analyses for a software project and the warnings issued by the analyses. Given a particular warning, CodeSonar users can identify the analysis that first issued the “same” warning (or a closely related one), and, if applicable, the analysis at which the warning stopped being issued (because the underlying problem was fixed).

12.1.3 Change of Application or Development Environment
“Use and modification of previously developed software may involve a new development environment, a new target processor or other hardware, or integration with other software than that used for the original application.”

The internal representations constructed by CodeSonar and CodeSurfer take into account many aspects of the software being analyzed and the build process used to construct that software, including the compiler or compilers, compiler options, preprocessor settings, and the platform for which the software is being built.

The analyses carried out on these internal representations will likewise reflect these factors. When a change in application or development environment leads to a change in the project and its representation, CodeSonar and CodeSurfer can be used to analyze the scope and impact of these changes in the same ways described in 12.1.1 Modifications to Previously Developed Software .

12.1.4 Upgrading a Development Baseline
“Guidelines follow for software whose software life cycle data from a previous application are determined to be inadequate or do not satisfy the objectives of this document, due to the safety objectives associated with a new application… Reverse engineering may be used to regenerate software life cycle data that is inadequate or missing in satisfying the objectives of this document.”

Just as CodeSonar and CodeSurfer can be used in support of the requirements of DO-178B Chapter 6 for software currently under development, they can be applied to previously developed software to bring software life cycle data up to the level required.

Conclusion
Static analysis tools can make a significant contribution to DO-178B activities. Sophisticated analyses in both tools provide multiple points of leverage for verification and reverification.

In addition, the extensive reporting and record-keeping mechanisms provide by CodeSonar support for various other DO-178B objectives, as do the program understanding capabilities of CodeSurfer.

References
[1] DO-178B, Software Considerations in Airborne Systems and Equipment Certification. 1992, RTCA Inc.
[3] Halstead,M.H., Elements of Software Science. 1977, New York, NY: Elsevier. 142.
[4] Holzmann,G.J., The Power of 10: Rules for Developing Safety-Critical Code. IEEE Computer, 2006. 39(6): pp. 95-97.
[5] Pugh,W., FindBugs – Find Bugs in Java Programs, http://findbugs.sourceforge.net/.
[6] Watson,A.H. and McCabe,T.J., Structured Testing: A Testing Methodology Using the Cyclomatic Complexity Metric. 1996, Computer Systems Laboratory, National Institute of Standards and Technology, Gaithersburg, MD. NIST Special Publication 500-235.

Paul Anderson is VP of Engineering at GrammaTech, a spin-off of Cornell University that specializes in static analysis. He received his B.Sc. from Kings College, University of London and his Ph.D. in computer science from City University London. Paul manages GrammaTech’s engineering team and is the architect of the company’s static-analysis tools.
He has helped a wide variety of organizations, including NASA, the FDA, the FAA, MITRE, Draper Laboratory, GE, Lockheed Martin, and Boeing, apply automated code analysis to critical projects. Paul has worked in the software industry for 16 years, with most of his experience focused on developing static-analysis, automated-testing, and program-transformation tools. A significant portion of his work has involved applying program analysis to improve security. His research on static analysis tools and techniques has been reported in numerous articles, journal publications, book chapters, and international conferences.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.