Functions that have an array as a parameter should also have an additional parameter that indicates the maximum number of elements that can be stored in the array. This additional That parameter is required to ensure that the function does not access memory outside the bounds of the array and adversely influence program execution. This additional That parameter should be present for each array parameter (in other words, the existence of each array parameter implies the existence of a complementary parameters representing parameter that represents the maximum number of elements in the array).
Note that the word array is used in this recommendation to mean array, string, or any other pointer to a contiguous block of memory in which one or more elements of a particular type are (potentially) stored. These terms are all effectively synonymous and represent the same potential for error.
Also note that this recommendation suggests that the parameter accompanying array parameters indicate the maximum number of elements that can be stored in the array and not the maximum size, in bytes, of the array. This is because,
- It does not make sense to think of array sizes in bytes in all cases. For example, it makes little sense to think ; for example, in the case of an array of integers in its size in bytes.
- If the size in bytes of the array is required, it can be derived from the number of elements in the array.
- It is better not to add to the cognitive load of the function user of the function and it is probably easier for the user to think of the size of the array in terms of its elements.
In most cases, the distinction between the number of elements and number of bytes is moot because there : There is a clear mapping between the two, and it is easier to think in terms of number of elements anyway. Unfortunately, this issue can become muddled when working with multibyte strings as the logical entity being manipulated differs from that of the type being used to implement it. Here, it is important to remind oneself remember that the type of the array is a character and not a multibyte character. Accordingly, the number of elements in the array is represented as a number of characters.
...
It is not necessary to go beyond the standard C library to find examples that violate this recommendation. This is , because the C language often prioritizes performance at the expense of robustness. The following are two examples from §7.21 of the C standard.
...
The problem with these functions is twofold. First, there is no indication of the size of the first array, s1. Thus, it is not possible to discern within the function how large s1 is and how many elements may be written into it. Second, while it appears that a size for s2 is supplied, this is, in fact, a the number of elements to copy. Thus, there is no guarantee that the elements being copied are members of array s2.
Compliant Solution
The above functions could be improved by adding element count parameters as follows:
| Code Block | ||
|---|---|---|
| ||
char *strncpy(char * restrict s1, size_t s1count, const char * restrict s2, size_t s2count, size_t n); char *strncat(char * restrict s1, size_t s1count, const char * restrict s2, size_t s2count); |
Note that for strncpy(), it makes sense to maintain the n parameter as the caller may only wish to copy a subset of the total number of elements. For strncat(), the n parameter is no longer required.
Compliant Solution (TR24731-1)
...
There are two notable differences between the compliant solution above and the secure versions from TR24731-1. First, the TR24731-1 versions use rsize_t instead of size_t. This allows the size to be compared against the reasonable limit for a single object, RSIZE_MAX. Second, the TR24731-1 versions do not require an element count for the second array. Consequently, these functions have limited ability to validate the input for s2. However, a size value for s1 is required, so memory outside of the range for s1 should not be overwritten.
...