If a relation exists between constants, you should encode the relationship in the definitions. Do not give two independent definitions, because a maintainer may fail to preserve that relationship when modifying the code. As a corollary, do not encode an impermanent or false relationship between constants, because future modifications may result in an incorrect definition for the dependent constant.

Noncompliant Code Example

In this noncompliant code example, the definition for OUT_STR_LEN must always be two greater than the definition of IN_STR_LEN. The following definitions fail to embody this relationship:

enum { IN_STR_LEN=18, OUT_STR_LEN=20 };

A programmer performing maintenance on this program would need to identify the relationship and modify both definitions accordingly. Although this sort of error appears relatively benign, it can easily lead to serious security vulnerabilities, such as buffer overflows.

Compliant Solution

The declaration in this compliant solution embodies the relationship between the two definitions:

enum { IN_STR_LEN=18, OUT_STR_LEN=IN_STR_LEN+2 };

As a result, a programmer can reliably modify the program by changing the definition of IN_STR_LEN.

Noncompliant Code Example

In this noncompliant code example, a relationship is established between two constants where none exists:

enum { ADULT_AGE=18 };

/* Misleading; relationship established when none exists */
enum { ALCOHOL_AGE=ADULT_AGE+3 };

A programmer performing maintenance on this program may modify the definition for ADULT_AGE but fail to recognize that the definition for ALCOHOL_AGE has also been changed as a consequence.

Compliant Solution

This compliant solution does not assume a relationship where none exists:

enum { ADULT_AGE=18 };
enum { ALCOHOL_AGE=21 };

Risk Assessment

Failing to properly encode relationships in constant definitions may lead to the introduction of defects during maintenance. These defects could potentially result in vulnerabilities, for example, if the affected constants were used for allocating or accessing memory.

Recommendation

Severity

Likelihood

Remediation Cost

Priority

Level

DCL08-C

Low

Unlikely

High

P1

L3

Automated Detection

Tool

Version

Checker

Description

Related Vulnerabilities

Search for vulnerabilities resulting from the violation of this rule on the CERT website.

Related Guidelines

[Plum 1985]Rule 1-4

 


5 Comments

  1. This is common when encoding the limit of enumerations:

    enum { TYPE_READ=1, TYPE_WRITE, TYPE_SYMBOL, TYPE_BLAH, TYPE_MAX = 5 }; // bad 
    
    enum { TYPE_READ=1, TYPE_WRITE, TYPE_SYMBOL, TYPE_BLAH, TYPE_MAX = TYPE_BLAH }; // better
    
    1. TYPE_MAX = TYPE_BLAH is subject to being wrong if someone inserts another type past BLAH and doesn't notice they need to adjust MAX. My typical enumeration is like this:

      typedef enum {
        TYPE_INVALID = 0,
        TYPE_FOO,
        /* ... */
        TYPE_BLAH,
        TYPE_COUNT  /\* MUST BE LAST\! \*/
      } myType_t;
      

      Of course, COUNT isn't always an accurate count, as some values may be skipped, but it will generally do for such purposes, and of course range-checking (valid must be > INVALID and < COUNT).

      The assignment of 0 to INVALID achieves the same thing that I suspect you're trying to do, but makes it explicit. (That is, avoid valid values in uninitialized memory. This can't be completely achieved, but at least making the most common uninitialized value invalid, will help.)

  2. In the compliant solution, 3*OUT_STR_LEN will not be the same as OUT_STR_LEN*3 since the expression is expanded and then evaluated.  A dependent definition with arithmetic needs to be enclosed in parentheses.

    1. What you are suggesting would be true if we had used preprocessor macros rather than enumeration types. (Which is why we recommend parentheses around macro expansions). But enumerator initializers are compile-time constants; that is, they are evaluated by the compiler and the resulting value is used in subsequent locations. See C99 Section 6.6 and 6.7.2.2 for more background information.

  3. Minor point: there is no descriptive text in the risk assessment section.