CTR52-CPP. Guarantee that library functions do not overflow

Copying data into a container that is not large enough to hold that data results in a buffer overflow. To prevent such errors, data copied to the destination container must be restricted on the basis of the destination container's size, or, preferably, the destination container must be guaranteed to be large enough to hold the data to be copied.

Vulnerabilities that result from copying data to an undersized buffer can also involve null-terminated strings. Consult STR50-CPP. Guarantee that storage for strings has sufficient space for character data and the null terminator for specific examples of this rule that involve strings.

Copies can be made with the std::memcpy() function. However, the std::memmove() and std::memset() functions can also have the same vulnerabilities because they overwrite a block of memory without checking that the block is valid. Such issues are not limited to C standard library functions; standard template library (STL) generic algorithms such as std::copy(), std::fill(), and std::transform() also assume valid output buffer sizes [ISO/IEC 14882-2014].

Noncompliant Code Example

STL containers can be subject to the same vulnerabilities as array data types. The std::copy() algorithm provides no inherent bounds checking and can lead to a buffer overflow. In this noncompliant code example, a vector of integers is copied from src to dest using std::copy(). Because std::copy() does nothing to expand the dest vector, the program will overflow the buffer on copying the first element.

#include <algorithm>
#include <vector>

void f(const std::vector<int> &src) {
  std::vector<int> dest;
  std::copy(src.begin(), src.end(), dest.begin());
  // ...
}

This hazard applies to any algorithm that takes a destination iterator, expecting to fill it with values. Most of the STL algorithms expect the destination container to have sufficient space to hold the values provided.

Compliant Solution (Sufficient Initial Capacity)

The proper way to use std::copy() is to ensure the destination container can hold all the elements being copied to it. This compliant solution enlarges the capacity of the vector prior to the copy operation:

#include <algorithm>
#include <vector>
void f(const std::vector<int> &src) {
  // Initialize dest with src.size() default-inserted elements
  std::vector<int> dest(src.size());
  std::copy(src.begin(), src.end(), dest.begin());
  // ...
}

Compliant Solution (Per-Element Growth)

An alternative approach is to supply a std::back_insert_iterator as the destination argument. This iterator expands the destination container by one element for each element supplied by the algorithm, which guarantees the destination container will become sufficiently large to hold the elements provided.

#include <algorithm>
#include <iterator>
#include <vector>

void f(const std::vector<int> &src) {
  std::vector<int> dest;
  std::copy(src.begin(), src.end(), std::back_inserter(dest));
  // ...
}

Compliant Solution (Assignment)

The simplest solution is to construct dest from src directly, as in this compliant solution:

#include <vector>

void f(const std::vector<int> &src) {
  std::vector<int> dest(src);
  // ...
}

Noncompliant Code Example

In this noncompliant code example, std::fill_n() is used to fill a buffer with 10 instances of the value 0x42. However, the buffer has not allocated any space for the elements, so this operation results in a buffer overflow.

#include <algorithm>
#include <vector>

void f() {
  std::vector<int> v;
  std::fill_n(v.begin(), 10, 0x42);
}

Compliant Solution (Sufficient Initial Capacity)

This compliant solution ensures the capacity of the vector is sufficient before attempting to fill the container.

#include <algorithm>
#include <vector>

void f() {
  std::vector<int> v(10);
  std::fill_n(v.begin(), 10, 0x42);
}

However, this compliant solution is inefficient. The constructor will default-construct 10 elements of type int, which are subsequently replaced by the value 0x42 by the call to std::fill_n(), meaning that each element in the container is initialized twice.

Compliant Solution (Fill Initialization)

This compliant solution initializes v to 10 elements whose values are all 0x42.

#include <algorithm>
#include <vector>

void f() {
  std::vector<int> v(10, 0x42);
}

Risk Assessment

Copying data to a buffer that is too small to hold that data results in a buffer overflow. Attackers can exploit this condition to execute arbitrary code.

Automated Detection

Related Vulnerabilities

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

Related Guidelines

Bibliography