const char *atoaddr(const char *src, size_t srclen,
struct in_addr *addr);
size_t addrtoa(struct in_addr addr, int format,
char *dst, size_t dstlen);
const char *atosubnet(const char *src, size_t srclen,
struct in_addr *addr, struct in_addr *mask);
size_t subnettoa(struct in_addr addr, struct in_addr mask,
int format, char *dst, size_t dstlen);
An address is specified in ASCII as a dotted-decimal address (e.g. 1.2.3.4), an eight-digit network-order hexadecimal number with the usual C prefix (e.g. 0x01020304, which is synonymous with 1.2.3.4), an eight-digit host-order hexadecimal number with a 0h prefix (e.g. 0h01020304, which is synonymous with 1.2.3.4 on a big-endian host and 4.3.2.1 on a little-endian host), a DNS name to be looked up via gethostbyname(3), or an old-style network name to be looked up via getnetbyname(3).
A dotted-decimal address may be incomplete, in which case ASCII-to-binary conversion implicitly appends as many instances of .0 as necessary to bring it up to four components. The components of a dotted-decimal address are always taken as decimal, and leading zeros are ignored. For example, 10 is synonymous with 10.0.0.0, and 128.009.000.032 is synonymous with 128.9.0.32 (the latter example is verbatim from RFC 1166). The result of addrtoa is always complete and does not contain leading zeros.
The letters in a hexadecimal address may be uppercase or lowercase or any mixture thereof. Use of hexadecimal addresses is strongly discouraged; they are included only to save hassles when dealing with the handful of perverted programs which already print network addresses in hexadecimal.
DNS names may be complete (optionally terminated with a ``.'') or incomplete, and are looked up as specified by local system configuration (see resolver(5)). The h_addr value returned by gethostbyname(3) is used, so with current DNS implementations, the result when the name corresponds to more than one address is difficult to predict. Name lookup resorts to getnetbyname(3) only if gethostbyname(3) fails.
A subnet specification is of the form network/mask. The network and mask can be any form acceptable to atoaddr. In addition, the mask can be a decimal integer (leading zeros ignored) giving a bit count, in which case it stands for a mask with that number of high bits on and all others off (e.g., 24 means 255.255.255.0). In any case, the mask must be contiguous (a sequence of high bits on and all remaining low bits off). As a special case, the subnet specification %default is a synonym for 0.0.0.0/0.
Atosubnet ANDs the mask with the address before returning, so that any non-network bits in the address are turned off (e.g., 10.1.2.3/24 is synonymous with 10.1.2.0/24). Subnettoa generates the decimal-integer-bit-count form of the mask, with no leading zeros, unless the mask is non-contiguous.
The srclen parameter of atoaddr and atosubnet specifies the length of the ASCII string pointed to by src; it is an error for there to be anything else (e.g., a terminating NUL) within that length. As a convenience for cases where an entire NUL-terminated string is to be converted, a srclen value of 0 is taken to mean strlen(src).
The dstlen parameter of addrtoa and subnettoa specifies the size of the dst parameter; under no circumstances are more than dstlen bytes written to dst. A result which will not fit is truncated. Dstlen can be zero, in which case dst need not be valid and no result is written, but the return value is unaffected; in all other cases, the (possibly truncated) result is NUL-terminated. The freeswan.h header file defines constants, ADDRTOA_BUF and SUBNETTOA_BUF, which are the sizes of buffers just large enough for worst-case results.
The format parameter of addrtoa and subnettoa specifies what format is to be used for the conversion. The value 0 (not the ASCII character '0', but a zero value) specifies a reasonable default, and is in fact the only format currently available. This parameter is a hedge against future needs.
The ASCII-to-binary functions return NULL for success and a pointer to a string-literal error message for failure; see DIAGNOSTICS. The binary-to-ASCII functions return 0 for a failure, and otherwise always return the size of buffer which would be needed to accommodate the full conversion result, including terminating NUL; it is the caller's responsibility to check this against the size of the provided buffer to determine whether truncation has occurred.
Fatal errors in atosubnet are: no / in src; atoaddr error in conversion of network or mask; bit-count mask too big; mask non-contiguous.
Fatal errors in addrtoa and subnettoa are: unknown format.
Ignoring leading zeros in dotted-decimal components and bit counts is arguably the most useful behavior in this application, but it might occasionally cause confusion with the historical use of leading zeros to denote octal numbers.
It is barely possible that somebody, somewhere, might have a legitimate use for non-contiguous subnet masks.
Getnetbyname(3) is a historical dreg.
The restriction of ASCII-to-binary error reports to literal strings (so that callers don't need to worry about freeing them or copying them) does limit the precision of error reporting.
The ASCII-to-binary error-reporting convention lends itself to slightly obscure code, because many readers will not think of NULL as signifying success. A good way to make it clearer is to write something like:
const char *error;
error = atoaddr( /* ... */ );
if (error != NULL) {
/* something went wrong */