Few programmers consider the issues around formatted I/O and type definitions. A programmer-defined integer type might be any type supported by the implementation, even a type larger than unsigned long long
. For example, given an implementation that supports 128-bit unsigned integers and provides a uint_fast128_t
type, a programmer may define the following type:
typedef uint_fast128_t mytypedef_t;
Furthermore, the definition of programmer-defined types may change, which creates a problem when these types are used with formatted output functions, such as printf()
, and formatted input functions, such as scanf()
. (See FIO47-C. Use valid format strings.)
The C intmax_t
and uintmax_t
types can represent any value representable by any other integer types of the same signedness. (See INT00-C. Understand the data model used by your implementation(s).) This capability allows conversion between programmer-defined integer types (of the same signedness) and intmax_t
and uintmax_t
:
mytypedef_t x; uintmax_t temp; temp = x; /* Always secure if mytypedef_t is unsigned*/ /* ... Change the value of temp ... */ if (temp <= MYTYPEDEF_MAX) { x = temp; }
Formatted I/O functions can be used to input and output greatest-width integer typed values. The j
length modifier in a format string indicates that the following d
, i
, o
, u
, x
, X
, or n
conversion specifier will apply to an argument with type intmax_t
or uintmax_t
. C also specifies the z
length modifier for use with arguments of type size_t
and the t
length modifier for arguments of type ptrdiff_t
.
In addition to programmer-defined types, there is no requirement that an implementation provide format-length modifiers for implementation-defined integer types. For example, a machine with an implementation-defined 48-bit integer type may not provide format-length modifiers for the type. Such a machine still must have a 64-bit long long
, with intmax_t
being at least that large.
Noncompliant Code Example (printf()
)
This noncompliant code example prints the value of x
as an unsigned long long
value even though the value is of a programmer-defined integer type:
#include <stdio.h> mytypedef_t x; /* ... */ printf("%llu", (unsigned long long) x);
There is no guarantee that this code prints the correct value of x
, as x
may be too large to represent as an unsigned long long
.
Compliant Solution (printf()
)
The C intmax_t
and uintmax_t
can be safely used to perform formatted I/O with programmer-defined integer types by converting signed programmer-defined integer types to intmax_t
and unsigned programmer-defined integer types to uintmax_t
, then outputting these values using the j
length modifier. Similarly, programmer-defined integer types can be input to variables of intmax_t
or uintmax_t
(whichever matches the signedness of the programmer-defined integer type) and then converted to programmer-defined integer types using appropriate range checks.
This compliant solution guarantees that the correct value of x
is printed, regardless of its length, provided that mytypedef_t
is an unsigned type:
#include <stdio.h> #include <inttypes.h> mytypedef_t x; /* ... */ printf("%ju", (uintmax_t) x);
Compliant Solution (Microsoft printf()
)
Visual Studio 2012 and earlier versions do not support the standard j
length modifier and do not have a nonstandard analog. Consequently, the programmer must hard code the knowledge that intmax_t
is int64_t
and uintmax_t
is
for Microsoft Visual Studio versions.uint64_t
#include <stdio.h> #include <inttypes.h> mytypedef_t x; /* ... */ #ifdef _MSC_VER printf("%llu", (uintmax_t) x); #else printf("%ju", (uintmax_t) x); #endif
A feature request has been submitted to Microsoft to add support for the j
length modifier to a future release of Microsoft Visual Studio.
Noncompliant Code Example (scanf()
)
This noncompliant code example reads an unsigned long long
value from standard input and stores the result in x
, which is of a programmer-defined integer type:
#include <stdio.h> mytypedef_t x; /* ... */ if (scanf("%llu", &x) != 1) { /* Handle error */ }
This noncompliant code example can result in a buffer overflow if the size of mytypedef_t
is smaller than unsigned long long
, or it might result in an incorrect value if the size of mytypedef_t
is larger than unsigned long long
. Moreover, scanf()
lacks the error checking capabilities of alternative conversion routines, such as strtol()
. For more information, see INT06-C. Use strtol() or a related function to convert a string token to an integer.
Compliant Solution (strtoumax()
)
This compliant solution guarantees that a correct value in the range of mytypedef_t
is read, or an error condition is detected, assuming the value of MYTYPEDEF_MAX
is correct as the largest value representable by mytypedef_t
: The strtoumax()
function is used instead of scanf()
as it provides enhanced error checking functionality. The fgets()
function is used to read input from stdin
.
#include <stdio.h> #include <inttypes.h> #include <errno.h> mytypedef_t x; uintmax_t temp; /* ... */ if (fgets(buff, sizeof(buff), stdin) == NULL) { if (puts("EOF or read error\n") == EOF) { /* Handle error */ } } else { /* Check for errors in the conversion */ errno = 0; temp = strtoumax(buff, &end_ptr, 10); if (ERANGE == errno) { if (puts("number out of range\n") == EOF) { /* Handle error */ } } else if (end_ptr == buff) { if (puts("not valid numeric input\n") == EOF) { /* Handle error */ } } else if ('\n' != *end_ptr && '\0' != *end_ptr) { if (puts("extra characters on input line\n") == EOF) { /* Handle error */ } } /* No conversion errors, attempt to store the converted value into x */ if (temp > MYTYPEDEF_MAX) { /* Handle error */ } else { x = temp; } }
Risk Assessment
Failure to use an appropriate conversion specifier when inputting or outputting programmer-defined integer types can result in buffer overflow and lost or misinterpreted data.
Recommendation | Severity | Likelihood | Remediation Cost | Priority | Level |
---|---|---|---|---|---|
INT15-C | High | Unlikely | Medium | P6 | L2 |
Automated Detection
Tool | Version | Checker | Description |
---|---|---|---|
Axivion Bauhaus Suite | 7.2.0 | CertC-INT15 | |
Compass/ROSE | Can catch violations of this rule by scanning the | ||
LDRA tool suite | 9.7.1 | 586 S | Enhanced Enforcement |
Parasoft C/C++test | 2023.1 | CERT_C-INT15-a | Use intmax_t or uintmax_t for formatted IO on programmer-defined integer types |
Related Vulnerabilities
Search for vulnerabilities resulting from the violation of this rule on the CERT website.
Related Guidelines
SEI CERT C++ Coding Standard | VOID INT15-CPP. Use intmax_t or uintmax_t for formatted IO on programmer-defined integer types |
MITRE CWE | CWE-681, Incorrect conversion between numeric types |
Bibliography
[Saks 2007c] | Standard C's Pointer Difference Type |
11 Comments
David Svoboda
I don't like the phrase 'user-defined'. I believe you are referring to nonstandard integer types supported by hardware (eg your hypothetical 48-bit integer). Suggest using 'implementation-defined' instead.
Hallvard Furuseth
No, I think it means type(def)s defined by the program, and that the first typedef has gotten reversed and should be
typedef uint_fast128_t mytypedef_t;
Robert Seacord (Manager)
The typedef was wrong, that's for sure. The rule is definitely about user or programmer-defined types (maybe programmer is a better, more specific term than user?) But perhaps this should also applied to implementation-defined types for which no length modifier is defined by the implementation. Of course, even if a length modifier is defined it will not be portable, so perhaps this idiom of casting to max integer types should be preferred even in this case.
David Svoboda
Several touch-ups to the rule, added "#include <inttypes.h>", and s/user-defined/programmer-defined/g;
Alex Volkovitsky
what is the first greyed out example trying to accomplish? i see
temp=x
andx=temp
...which is a nopDavid Svoboda
it was trying to illustrate how to convert x to temp (automatic) and temp to x (range check required). I made the code example more clear (hopefully)
Robert Seacord (Manager)
The j length modifier is not supported in Visual C++, so we probably need an alternate solution for Windows. I'm not sure how these values are printed on this platform, however.
Robert Seacord
I think I fixed this, although the solution is not very elegant. Please check if you have the chance.
Steven Stewart-Gallus
It might be nice to note that for standard numeric types there are defines in inttypes.h such as PRId8 that define strings for outputting them. Also, the provider of a type can provide their own defines as well.
Gabor Buella
This seems to contradict with another recommendation INT5-C . Should this suggest using strtoimax or strtoumax instead of scanf?
David Svoboda
This rule and INT05-C. Do not use input functions to convert character data if they cannot handle all possible inputs cover different, though similar, territory. This rule specifically addreses user-defined integer types, not standard types like
long
. INT05-C covers more issues with inputting standard integer types (orintmax_t
t ypes). The only overlap is in using*scanf()
to input a user-defined integer type, in which case both recommendations should be followed.