CERT
Skip to end of metadata
Go to start of metadata

Although programmers often use integers and pointers interchangeably in C, pointer-to-integer and integer-to-pointer conversions are implementation-defined

Conversions between integers and pointers can have undesired consequences depending on the implementation. According to the C Standard, subclause 6.3.2.3 [ISO/IEC 9899:2011],

An integer may be converted to any pointer type. Except as previously specified, the result is implementation-defined, might not be correctly aligned, might not point to an entity of the referenced type, and might be a trap representation.

Any pointer type may be converted to an integer type. Except as previously specified, the result is implementation-defined. If the result cannot be represented in the integer type, the behavior is undefined. The result need not be in the range of values of any integer type.

Do not convert an integer type to a pointer type if the resulting pointer is incorrectly aligned, does not point to an entity of the referenced type, or is a trap representation.

Do not convert a pointer type to an integer type if the result cannot be represented in the integer type. (See undefined behavior 24.)

The mapping between pointers and integers must be consistent with the addressing structure of the execution environment. Issues may arise, for example, on architectures that have a segmented memory model.

Noncompliant Code Example

The size of a pointer can be greater than the size of an integer, such as in an implementation where pointers are 64 bits and unsigned integers are 32 bits. This code example is noncompliant on such implementations because the result of converting the 64-bit ptr cannot be represented in the 32-bit integer type:

Compliant Solution

Any valid pointer to void can be converted to intptr_t or uintptr_t and back with no change in value. (See INT36-EX2.) The C Standard guarantees that a pointer to void may be converted to or from a pointer to any object type and back again and that the result must compare equal to the original pointer. Consequently, converting directly from a char * pointer to a uintptr_t, as in this compliant solution, is allowed on implementations that support the uintptr_t type.

Noncompliant Code Example

In this noncompliant code example, the pointer ptr is converted to an integer value. The high-order 9 bits of the number are used to hold a flag value, and the result is converted back into a pointer. This example is noncompliant on an implementation where pointers are 64 bits and unsigned integers are 32 bits because the result of converting the 64-bit ptr cannot be represented in the 32-bit integer type.

A similar scheme was used in early versions of Emacs, limiting its portability and preventing the ability to edit files larger than 8MB.

Compliant Solution

This compliant solution uses a struct to provide storage for both the pointer and the flag value. This solution is portable to machines of different word sizes, both smaller and larger than 32 bits, working even when pointers cannot be represented in any integer type. 

Noncompliant Code Example

It is sometimes necessary to access memory at a specific location, requiring a literal integer to pointer conversion. In this noncompliant code, a pointer is set directly to an integer constant, where it is unknown whether the result will be as intended:

The result of this assignment is implementation-defined, might not be correctly aligned, might not point to an entity of the referenced type, and might be a trap representation.

Compliant Solution

Adding an explicit cast may help the compiler convert the integer value into a valid pointer. A common technique is to assign the integer to a volatile-qualified object of type intptr_t or uintptr_t and then assign the integer value to the pointer:

Exceptions

INT36-C-EX1: A null pointer can be converted to an integer; it takes on the value 0. Likewise, the integer value 0 can be converted to a pointer; it becomes the null pointer.

INT36-C-EX2: Any valid pointer to void can be converted to intptr_t or uintptr_t or their underlying types and back again with no change in value. Use of underlying types instead of intptr_t or uintptr_t is discouraged, however, because it limits portability.

Risk Assessment

Converting from pointer to integer or vice versa results in code that is not portable and may create unexpected pointers to invalid memory locations.

Rule

Severity

Likelihood

Remediation Cost

Priority

Level

INT36-C

Low

Probable

High

P2

L3

Automated Detection

Tool

Version

Checker

Description

Astrée17.04i

pointer-integral-cast

pointer-integral-cast-implicit

function-pointer-integer-cast

function-pointer-integer-cast-implicit

Fully checked
Clang3.9-Wint-to-pointer-cast, -Wint-conversionCan detect some instances of this rule, but does not detect all
CodeSonar4.4LANG.CAST.PC.CONST2PTR
LANG.CAST.PC.INT
Conversion: integer constant to pointer
Conversion: pointer/integer
Compass/ROSE   
Coverity2017.07PW.POINTER_CONVERSION_LOSES_BITSFully implemented
Klocwork2017MISRA.CAST.OBJ_PTR_TO_INT.2012 
LDRA tool suite9.5.6

439 S, 440 S

Fully implemented
Parasoft C/C++test9.5MISRA2008-5_2_8Fully implemented
PRQA QA-C9.3305, 306, 309, 429, 432, 557, 563, 671, 674Partially implemented
RuleChecker17.04i

pointer-integral-cast

pointer-integral-cast-implicit

function-pointer-integer-cast

function-pointer-integer-cast-implicit

Fully checked
SonarQube C/C++ Plugin3.11S1767Partially implemented

Related Vulnerabilities

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

Related Guidelines

SEI CERT C++ Coding StandardINT11-CPP. Take care when converting from pointer to integer or integer to pointer
ISO/IEC TR 24772:2013Pointer Casting and Pointer Type Changes [HFC]
ISO/IEC TS 17961:2013Converting a pointer to integer or integer to pointer [intptrconv]
MITRE CWECWE-466, Return of Pointer Value Outside of Expected Range
CWE-587, Assignment of a Fixed Address to a Pointer

Bibliography

[ISO/IEC 9899:2011]6.3.2.3, "Pointers"

 


11 Comments

  1. uh, you should maybe rethink your identifiers and literals to be a little less colorful

    i think this could probably be broken into two recommendations. part 1: NULL and 0 overlaps with:

    EXP36-C. A pointer shall not be compared to NULL or assigned NULL, use plain 0 instead

    but I think this rule has been abandoned and your content is better. I would take what you have and move it to that rule.

    1. I would add to that rule but honestly I think it is wrong. There are many schools of thought on the NULL/0 usage but to be honest, it doesn't actually matter. NULL is effectively the same as 0. If I had to make a suggestion, NULL should be used in the pointer context to keep your code straight, but with the awareness that they aren't any different ... 

      I'll get on to fixing the names. It was kind of a joke.

  2. Isn't this really a pointer rule, not an integer rule?

    Wouldn't it be effectively the same rule if we said don't convert pointers to/from any other type?

    Seems strange to call it an integer rule.

  3. It's called an integer rule because of the old history of pointers being the same as integers. People are most likely to know they shouldn't go arbitrarly converting pointers to whatever, but they may not know that pointers are now different from integers.

     With that said it's not the best rationale in the world, the big issue is there's no actual plain pointer section... show me where to put it/feel free to move it.

     also I have not the best idea as to the vulnerability index so someone else might want to poke at that and see if it looks off.

  4. This rule should have a prerequisite that sizeof(unsigned int*) is equal to sizeof(unsigned int). Right? How does the intpoint union work on platform with different size of int and pointer, such as LP64?

    1. The recommendation has now been edited to take this into account.

  5. The second compliant solution is non-portable, because you are performing arithmetic on values of type intptr_t/uintptr_t.  All the C standard guarantees (7.18.1.4) is that you can convert void* values to intptr_t (or uintptr_t) and back again and end up with an equal value for the pointer.  Mathematical operations (including ==) are non-portable.

    1. There is nothing wrong with performing arithmetic on an intptr_t.  The problem lies in where the values came from.  Since they came from pointers, the two values may not have any particular relation to each other.  This example has been deleted.

  6. The bit about conversion okay for 0 should be omitted, and it isn't pertinent and is stated too strongly.  Consider

    intptr_t x = 0;
    double *p = (double *)x;  // result is not necessarily a null pointer!

    This example should have a cast:

    unsigned int *ptr = (unsigned int *)0xcfcfcfcf;

  7. This compliant solution

    Compliant Solution

    Adding an explicit cast may help the compiler convert the integer value into a valid pointer.

    unsigned int *ptr = (unsigned int *) 0xcfcfcfcf;

     

    Is almost identical to this noncompliant solution from  TS 17961: 

    Converting a pointer to integer or integer to pointer       [intptrconv]

     

    EXAMPLE 2 In this noncompliant example, a diagnostic is required because the conversion of the integer literal 0xdeadbeef to a pointer that is not known to point to an entity of the referenced type.

    unsigned int *g(void) {
      unsigned int *ptr = (unsigned int *)0xdeadbeef;  // diagnostic required
      /* ... */
      return ptr;
    }
  8. I feel that, with a little more work, this recommendation can be promoted to a rule.  The remaining issue for me would be changing the emphasis from portability to security.