EXP35-C. Do not access or modify an array in the result of a function call after a subsequent sequence point

C99 Section 6.5.2.2 says [[ISO/IEC 9899:1999]],

If an attempt is made to modify the result of a function call or to access it after the next sequence point, the behavior is undefined.

C forbids functions from returning arrays; however it is possible to return structs or unions that contain arrays.

Consequently, if a function call's return value contains an array, that array should never be accessed or modified within the expression.

Non-Compliant Code Example

The following non-compliant code attempts to retrieve an array from a struct that is returned by a function call.

#include <stdio.h>

struct X { char a[6]; };

struct X addressee(void) {
  struct X result = { "world" };
  return result;
}

int main(void) {
  printf("Hello, %s!\n", addressee().a);
  return 0;
}

This solution is problematic because of three inherent properties of C:
1. In C, the lifetime of a return value ends at the next sequence point. Therefore by the time printf() is called, the struct returned by the addressee() call is no longer considered valid, and may have been overwritten.
1. C function arguments are passed by value. Thus, copies are made of all objects generated by the arguments. For example, a copy is made of the pointer to "Hello, %s!\n". Under most circumstances, these copies protect one from the effects of sequence points described above.
1. Finally, C implicitly converts arrays to pointers when passing them as function arguments. This means that in the previous point, a copy is made of the pointer to the addresee().a array, and that pointer copy is passed to printf(). But the array data itself is not copied, and no longer exists when printf() is called.

Therefore when printf() tries to dereference the pointer passed as its 2nd argument, it will likely find garbage.

Implementation Details

This code compiles cleanly and runs without error under Microsoft Visual C++ Version 8.0. On GCC Compiler Version 4.2, the program compiles with a warning when the -Wall switch is used, and execution on Linux results in a segmentation fault. However, if one passes the flag --std=c99 to GCC, the program compiles with no warning and runs with no error.

Compliant Solution

This compliant solution does not have undefined behavior because the structure returned by the call to addressee() is stored as the variable my_x before calling the printf() function.

#include <stdio.h>

struct X { char a[6]; };

struct X addressee(void) {
  struct X result = { "world" };
  return result;
}

int main(void) {
  struct X my_x = addressee();
  printf("Hello, %s!\n", my_x.a);
  return 0;
}

Risk Assessment

Attempting to access or modify an array inside the result of a function call after a subsequent sequence point may result in unexpected and perhaps unintended program behavior.

Automated Detection

Splint Version 3.1.1 can detect violations of this rule.

Related Vulnerabilities

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

References

[[ISO/IEC 9899:1999]] Section 6.5.2.2, "Function calls"
[[ISO/IEC PDTR 24772]] "DCM Dangling references to stack frames" and "SAM Side-effects and order of evaluation"

EXP34-C. Ensure a null pointer is not dereferenced      03. Expressions (EXP)       EXP36-C. Do not convert pointers into more strictly aligned pointer types