Fa setting controls which hashing method to use, and also supplies various parameters to the chosen method, most importantly a random ``salt'' which ensures that no two stored hashes are the same, even if the Fa phrase strings are the same.
The Fa data argument to crypt_r is a structure of type Vt struct crypt_data . It has at least these fields:
struct crypt_data {
char output[CRYPT_OUTPUT_SIZE];
char setting[CRYPT_OUTPUT_SIZE];
char phrase[CRYPT_MAX_PASSPHRASE_SIZE];
char initialized;
};
Upon a successful return from crypt_r the hashed passphrase will be stored in Fa output . Applications are encouraged, but not required, to use the Fa phrase and Fa setting fields to store the strings that they will pass as Fa phrase and Fa setting to crypt_r This will make it easier to erase all sensitive data after it is no longer needed.
The Fa initialized field must be set to zero before the first time a Vt struct crypt_data object is first used in a call to Fn crypt_r . We recommend zeroing the entire object, not just Fa initialized and not just the documented fields, before the first use. (Of course, do this before storing anything in Fa setting and Fa phrase . )
The Fa data argument to crypt_rn should also point to a Vt struct crypt_data object, and Fa size should be the size of that object, cast to Vt int . When used with crypt_rn the entire Fa data object (except for the Fa phrase and Fa setting fields) must be zeroed before its first use; this is not just a recommendation, as it is for crypt_r Otherwise, the fields of the object have the same uses that they do for crypt_r
On the first call to crypt_ra Fa data should be the address of a Vt void * variable set to NULL, and Fa size should be the address of an Vt int variable set to zero. crypt_ra will allocate and initialize a Vt struct crypt_data object, using malloc(3), and write its address and size into the variables pointed to by Fa data and Fa size . These can be reused in subsequent calls. After the application is done hashing passphrases, it should deallocate the Vt struct crypt_data object using free(3).
crypt places its result in a static storage area, which will be overwritten by subsequent calls to crypt It is not safe to call crypt from multiple threads simultaneously.
crypt_r crypt_rn and crypt_ra place their result in the Fa output field of their Fa data argument. It is safe to call them from multiple threads simultaneously, as long as a separate Fa data object is used for each thread.
Upon error, crypt_r crypt_rn and crypt_ra write an invalid hashed passphrase to the Fa output field of their Fa data argument, and crypt writes an invalid hash to its static storage area. This string will be shorter than 13 characters, will begin with a `* ' and will not compare equal to Fa setting .
Upon error, crypt_rn and crypt_ra return a null pointer. crypt_r and crypt may also return a null pointer, or they may return a pointer to the invalid hash, depending on how libcrypt was configured. (The option to return the invalid hash is for compatibility with old applications that assume that crypt cannot return a null pointer. See Sx PORTABILITY NOTES below.)
All four functions set errno when they fail.
POSIX does not specify any hashing methods, and does not require hashed passphrases to be portable between systems. In practice, hashed passphrases are portable as long as both systems support the hashing method that was used. However, the set of supported hashing methods varies considerably from system to system.
The behavior of crypt on errors isn't well standardized. Some implementations simply can't fail (except by crashing the program), others return a null pointer or a fixed string. Most implementations don't set errno but some do. POSIX specifies returning a null pointer and setting errno but it defines only one possible error, Er ENOSYS , in the case where crypt is not supported at all. Some older applications are not prepared to handle null pointers returned by crypt The behavior described above for this implementation, setting errno and returning an invalid hashed passphrase different from Fa setting , is chosen to make these applications fail closed when an error occurs.
Due to historical restrictions on the export of cryptographic software from the USA, crypt is an optional POSIX component. Applications should therefore be prepared for crypt not to be available, or to always fail (setting errno to Er ENOSYS ) at runtime.
POSIX specifies that crypt is declared in In unistd.h , but only if the macro _XOPEN_CRYPT is defined and has a value greater than or equal to zero. Since libcrypt does not provide In unistd.h , it declares crypt crypt_r crypt_rn and crypt_ra in In crypt.h instead.
On a minority of systems (notably recent versions of Solaris), crypt uses a thread-specific static storage buffer, which makes it safe to call from multiple threads simultaneously, but does not prevent each call within a thread from overwriting the results of the previous one.
Vt struct crypt_data may be quite large (32kB in this implementation of libcrypt; over 128kB in some other implementations). This is large enough that it may be unwise to allocate it on the stack.
Some recently designed hashing methods need even more scratch memory, but the crypt_r interface makes it impossible to change the size of Vt struct crypt_data without breaking binary compatibility. The crypt_rn interface could accommodate larger allocations for specific hashing methods, but the caller of crypt_rn has no way of knowing how much memory to allocate. crypt_ra does the allocation itself, but can only make a single call to malloc(3).
| Interface | Attribute | Value |
| crypt | Thread safety | MT-Unsafe race:crypt |
| crypt_r crypt_rn crypt_ra | Thread safety | MT-Safe |
crypt_r originates with the GNU C Library. There's also a crypt_r function on HP-UX and MKS Toolkit, but the prototypes and semantics differ.
crypt_rn and crypt_ra originate with the Openwall project.