man.bsd.lv manual page server

DESCRIPTION

PEM is the binary content encoding first defined in IETF RFC 1421. The content is a series of base64-encoded lines, surrounded by begin/end markers each on their own line. For example:

-----BEGIN PRIVATE KEY-----
MIICdg....
... bhTQ==
-----END PRIVATE KEY-----

Optional header line(s) may appear after the begin line, and their existence depends on the type of object being written or read.

PEM_write() writes to the file fp, while PEM_write_bio() writes to the BIO bp. The name is the name to use in the marker, the header is the header value or NULL, and data and len specify the data and its length.

The final data buffer is typically an ASN.1 object which can be decoded with the d2i_*() function appropriate to the type name; see d2i_X509(3) for examples.

PEM_read() reads from the file fp, while PEM_read_bio() reads from the BIO bp. Both skip any non-PEM data that precedes the start of the next PEM object. When an object is successfully retrieved, the type name from the "----BEGIN <type>-----" is returned via the name argument, any encapsulation headers are returned in header, and the base64-decoded content and its length are returned via data and len, respectively. The name, header, and data pointers should be freed by the caller when no longer needed.

The remaining functions are deprecated because the underlying PEM encryption format is obsolete and should be avoided. It uses an encryption format with an OpenSSL-specific key-derivation function, which employs MD5 with an iteration count of 1. Instead, private keys should be stored in PKCS#8 form, with a strong PKCS#5 v2.0 PBE; see PEM_write_PrivateKey(3) and d2i_PKCS8PrivateKey_bio(3).

PEM_get_EVP_CIPHER_INFO() can be used to determine the data returned by PEM_read() or PEM_read_bio() is encrypted and to retrieve the associated cipher and IV. The caller passes a pointer to a structure of type EVP_CIPHER_INFO via the cinfo argument and the header returned via PEM_read() or PEM_read_bio(). If the call is successful, 1 is returned and the cipher and IV are stored at the address pointed to by cinfo. When the header is malformed or not supported or when the cipher is unknown or some internal error happens, 0 is returned.

PEM_do_header() can then be used to decrypt the data if the header indicates encryption. The cinfo argument is a pointer to the structure initialized by a preceding call to PEM_get_EVP_CIPHER_INFO(). If that structure indicates the absence of encryption, PEM_do_header() returns successfully without taking any action. The data and len arguments are used both to pass in the encrypted data that was returned in the same arguments from the preceding call to PEM_read() or PEM_read_bio() and to pass out the decrypted data.

The callback function cb is used to obtain the encryption password; if cb is NULL, PEM_def_callback() is used instead. The password buffer needs to be at least size bytes long. PEM_def_callback() silently truncates the NUL-terminated byte string u to at most num bytes and copies it into password without a terminating NUL byte. If u is NULL, PEM_def_callback() instead prompts the user for the password with echoing turned off by calling EVP_read_pw_string_min(3) internally. In this case, the size is silently reduced to at most BUFSIZ and at most size - 1 bytes are accepted from the user and copied into the byte string buffer password. A callback function cb supplied by the application may use u for a different purpose than PEM_def_callback() does, e.g., as auxiliary data to use while acquiring the password. For example, a GUI application might pass a window handle. If the verify flag is non-zero, the user is prompted twice for the password to make typos less likely and it is checked that both inputs agree. This flag is not set by PEM_do_header() nor by other read functions.

If the data is a priori known to not be encrypted, then neither PEM_get_EVP_CIPHER_INFO() nor PEM_do_header() need to be called.