ARR00-C. Understand how arrays work

The incorrect use of arrays has traditionally been a source of exploitable vulnerabilities. Elements referenced within an array using the subscript operator [] are not checked unless the programmer provides adequate bounds checking. As a result, the expression array [pos] = value can be used by an attacker to transfer control to arbitrary code.

If the attacker can control the values of both pos and value in the expression array [pos] = value, the attacker can perform an arbitrary write (which is when the attacker overwrites other storage locations with different content). The consequences range from changing a variable used to determine what permissions the program grants to executing arbitrary code with the permissions of the vulnerable process. Arrays are also a common source of buffer overflows when iterators exceed the bounds of the array.

An array is a series of objects, all of which are the same size and type. Each object in an array is called an array element. The entire array is stored contiguously in memory (that is, there are no gaps between elements). Arrays are commonly used to represent a sequence of elements where random access is important, but there is little or no need to insert new elements into the sequence (which can be an expensive operation with arrays).

Arrays containing a constant number of elements can be declared as follows:

enum { ARRAY_SIZE = 12 };
int dis[ARRAY_SIZE];

These statements allocate storage for an array of 12 integers referenced by dis. Arrays are indexed from 0..n-1 (where n represents an array bound). Arrays can also be declared as follows:

int ita[];

This is called an incomplete type because the size is unknown. If an array of unknown size is initialized, its size is determined by the largest indexed element with an explicit initializer. At the end of its initializer list, the array no longer has incomplete type.

int ita[] = { 1, 2 };

While these declarations work fine when the size of the array is known at compilation time, it is not possible to declare an array in this fashion when the size can be determined only at runtime. The C99 standard adds support for variable-length arrays or arrays whose size is determined at runtime. Before the introduction of variable-length arrays in C99, however, these "arrays" were typically implemented as pointers to their respective element types allocated using malloc(), as shown in this example.

int *dat = (int *)malloc(ARRAY_SIZE * sizeof(int));

Always check that malloc() returns a non-NULL pointer, as per MEM32-C. Detect and handle memory allocation errors.

It is important to retain any pointer value returned by malloc() so that the referenced memory may eventually be deallocated. One possible way of preserving such a value is to use a constant pointer.

int * const dat = (int * const)malloc(
  ARRAY_SIZE * sizeof(int)
);
/* ... */
free(dat);

Both dis and dat arrays can be initialized as follows:

for (i = 0; i < ARRAY_SIZE; i++) {
   dis[i] = 42; /* Assigns 42 to each element; */
   /* ... */
}

The dis array can also be initialized as follows:

for (i = 0; i < ARRAY_SIZE; i++) {
   *dis = 42;
   dis++;
}
dis -= ARRAY_SIZE;

The dat identifier cannot be incremented, so the expression dat++ results in a fatal compilation error. Both arrays can be initialized as follows:

int *p = dat;
for (i = 0; i < ARRAY_SIZE; i++)  {
  *p = 42; /* Assigns 42 to each element*/
  p++;
}

The variable p is declared as a pointer to an integer and then incremented in the loop. This technique can be used to initialize both arrays and is a better style of programming than incrementing the pointer to the array because it does not change the pointer to the start of the array.

Obviously, there is a relationship between array subscripts [] and pointers. The expression dis[i] is equivalent to *(dis+i). In other words, if dis is an array object (equivalently, a pointer to the initial element of an array object) and i is an integer, dis[i] designates the ith element of dis (counting from zero). In fact, because *(dis+i) can be expressed as *(i+dis), the expression dis[i] can be represented as i[dis], although doing so is not encouraged.

The initial element of an array is accessed using an index of zero; for example, dat[0] references the first element of dat array. The dat identifier points to the start of the array, so adding zero is inconsequential because *(dat+i) is equivalent to *(dat+0), which is equivalent to *(dat).

Risk Assessment

Arrays are a common source of vulnerabilities in C language programs because they are frequently used but not always fully understood.

Related Vulnerabilities

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

Related Guidelines

CERT C++ Secure Coding Standard: ARR00-CPP. Understand when to prefer vectors over arrays

ISO/IEC 9899:1999 Section 6.7.5.2, "Array declarators"

MITRE CWE: CWE-119, "Failure to Constrain Operations within the Bounds of an Allocated Memory Buffer"

MITRE CWE: CWE-129, "Unchecked Array Indexing"

Bibliography

06. Arrays (ARR)      06. Arrays (ARR)