Pointer arithmetic is appropriate only when the pointer argument refers to an array (see ARR37-C. Do not add or subtract an integer to a pointer to a non-array object), including an array of bytes. When performing pointer arithmetic, the size of the value to add to or subtract from a pointer is automatically scaled to the size of the type of the referenced array object. Adding or subtracting a scaled integer value to or from a pointer is invalid because it may yield a pointer that does not point to an element within or one past the end of the array. (See ARR30-C. Do not form or use out-of-bounds pointers or array subscripts.)

Adding a pointer to an array of a type other than character to the result of the sizeof operator or offsetof macro, which returns a size and an offset, respectively, violates this rule. However, adding an array pointer to the number of array elements, for example, by using the arr[sizeof(arr)/sizeof(arr[0])]) idiom, is allowed provided that arr refers to an array and not a pointer.

Noncompliant Code Example

In this noncompliant code example, sizeof(buf) is added to the array buf. This example is noncompliant because sizeof(buf) is scaled by int and then scaled again when added to buf.

enum { INTBUFSIZE = 80 };

extern int getdata(void);
int buf[INTBUFSIZE];
 
void func(void) {
  int *buf_ptr = buf;

  while (buf_ptr < (buf + sizeof(buf))) {
    *buf_ptr++ = getdata();
  }
}

Compliant Solution

This compliant solution uses an unscaled integer to obtain a pointer to the end of the array:

enum { INTBUFSIZE = 80 };

extern int getdata(void);
int buf[INTBUFSIZE];

void func(void) {
  int *buf_ptr = buf;

  while (buf_ptr < (buf + INTBUFSIZE)) {
    *buf_ptr++ = getdata();
  }
}

Noncompliant Code Example

In this noncompliant code example, skip is added to the pointer s. However, skip represents the byte offset of ull_b in struct big. When added to s, skip is scaled by the size of struct big.

#include <string.h>
#include <stdlib.h>
#include <stddef.h>
 
struct big {
  unsigned long long ull_a;
  unsigned long long ull_b;
  unsigned long long ull_c;
  int si_e;
  int si_f;
};

void func(void) {
  size_t skip = offsetof(struct big, ull_b);
  struct big *s = (struct big *)malloc(sizeof(struct big));
  if (s == NULL) {
    /* Handle malloc() error */
  }

  memset(s + skip, 0, sizeof(struct big) - skip);
  /* ... */
  free(s);
  s = NULL;
}

Compliant Solution

This compliant solution uses an unsigned char * to calculate the offset instead of using a struct big *, which would result in scaled arithmetic:

#include <string.h>
#include <stdlib.h>
#include <stddef.h>
 
struct big {
  unsigned long long ull_a;
  unsigned long long ull_b;
  unsigned long long ull_c;
  int si_d;
  int si_e;
};

void func(void) {
  size_t skip = offsetof(struct big, ull_b);
  unsigned char *ptr = (unsigned char *)malloc(
    sizeof(struct big)
  );
  if (ptr == NULL) {
     /* Handle malloc() error */
  }

  memset(ptr + skip, 0, sizeof(struct big) - skip);
  /* ... */
  free(ptr);
  ptr = NULL;
}

Noncompliant Code Example

In this noncompliant code example, wcslen(error_msg) * sizeof(wchar_t) bytes are scaled by the size of wchar_t when added to error_msg:

#include <wchar.h>
#include <stdio.h>
 
enum { WCHAR_BUF = 128 };
 
void func(void) {
  wchar_t error_msg[WCHAR_BUF];

  wcscpy(error_msg, L"Error: ");
  fgetws(error_msg + wcslen(error_msg) * sizeof(wchar_t), 
         WCHAR_BUF - 7, stdin);
  /* ... */
}

Compliant Solution

This compliant solution does not scale the length of the string; wcslen() returns the number of characters and the addition to error_msg is scaled:

#include <wchar.h>
#include <stdio.h>

enum { WCHAR_BUF = 128 };
const wchar_t ERROR_PREFIX[7] = L"Error: ";

void func(void) {
  const size_t prefix_len = wcslen(ERROR_PREFIX);
  wchar_t error_msg[WCHAR_BUF];

  wcscpy(error_msg, ERROR_PREFIX);
  fgetws(error_msg + prefix_len,
        WCHAR_BUF - prefix_len, stdin);
  /* ... */
}

Risk Assessment

Failure to understand and properly use pointer arithmetic can allow an attacker to execute arbitrary code.

Automated Detection

Tool	Version	Checker	Description
Astrée		scaled-pointer-arithmetic	Partially checked Besides direct rule violations, Astrée reports all (resulting) out-of-bound array accesses.
Axivion Bauhaus Suite		CertC-ARR39	Fully implemented
CodeSonar		LANG.MEM.BO LANG.MEM.BU LANG.MEM.TBA LANG.MEM.TO LANG.MEM.TU LANG.STRUCT.PARITH LANG.STRUCT.PBB LANG.STRUCT.PPE	Buffer overrun Buffer underrun Tainted buffer access Type overrun Type underrun Pointer Arithmetic Pointer before beginning of object Pointer past end of object
Coverity		BAD_SIZEOF	Partially implemented
Cppcheck Premium		premium-cert-arr39-c
Helix QAC		DF4955, DF4956, DF4957
Klocwork		CERT.ARR.PTR.ARITH
LDRA tool suite		47 S, 489 S, 567 S, 64 X, 66 X, 68 X, 69 X, 70 X, 71 X	Partially implemented
Parasoft C/C++test		CERT_C-ARR39-a CERT_C-ARR39-b CERT_C-ARR39-c	Avoid accessing arrays out of bounds Pointer arithmetic should not be used Do not add or subtract a scaled integer to a pointer
Polyspace Bug Finder		CERT C: Rule ARR39-C	Checks for incorrect pointer scaling (rule fully covered).
RuleChecker		scaled-pointer-arithmetic	Partially checked
TrustInSoft Analyzer		index_in_address	Exhaustively detects undefined behavior (see one compliant and one non-compliant example).

Related Vulnerabilities

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

Related Guidelines

Key here (explains table format and definitions)

Bibliography