Add functions to allow extensions to be added to certificate requests.

[openssl.git] / doc / openssl.txt

This is some preliminary documentation for OpenSSL.

==============================================================================
                            BUFFER Library
==============================================================================

The buffer library handles simple character arrays. Buffers are used for
various purposes in the library, most notably memory BIOs.

The library uses the BUF_MEM structure defined in buffer.h:

typedef struct buf_mem_st
{
        int length;     /* current number of bytes */
        char *data;
        int max;        /* size of buffer */
} BUF_MEM;

'length' is the current size of the buffer in bytes, 'max' is the amount of
memory allocated to the buffer. There are three functions which handle these
and one "miscellaneous" function.

BUF_MEM *BUF_MEM_new()

This allocates a new buffer of zero size. Returns the buffer or NULL on error.

void BUF_MEM_free(BUF_MEM *a)

This frees up an already existing buffer. The data is zeroed before freeing
up in case the buffer contains sensitive data.

int BUF_MEM_grow(BUF_MEM *str, int len)

This changes the size of an already existing buffer. It returns zero on error
or the new size (i.e. 'len'). Any data already in the buffer is preserved if
it increases in size.

char * BUF_strdup(char *str)

This is the previously mentioned strdup function: like the standard library
strdup() it copies a null terminated string into a block of allocated memory
and returns a pointer to the allocated block.

Unlike the standard C library strdup() this function uses Malloc() and so
should be used in preference to the standard library strdup() because it can
be used for memory leak checking or replacing the malloc() function.

The memory allocated from BUF_strdup() should be freed up using the Free()
function.

==============================================================================
               OpenSSL X509V3 extension configuration
==============================================================================

OpenSSL X509V3 extension configuration: preliminary documentation.

INTRODUCTION.

For OpenSSL 0.9.2 the extension code has be considerably enhanced. It is now
possible to add and print out common X509 V3 certificate and CRL extensions.

BEGINNERS NOTE

For most simple applications you don't need to know too much about extensions:
the default openssl.cnf values will usually do sensible things.

If you want to know more you can initially quickly look through the sections
describing how the standard OpenSSL utilities display and add extensions and
then the list of supported extensions.

For more technical information about the meaning of extensions see:

http://www.imc.org/ietf-pkix/
http://home.netscape.com/eng/security/certs.html

PRINTING EXTENSIONS.

Extension values are automatically printed out for supported extensions.

openssl x509 -in cert.pem -text
openssl crl -in crl.pem -text

will give information in the extension printout, for example:

        X509v3 extensions:
            X509v3 Basic Constraints: 
                CA:TRUE
            X509v3 Subject Key Identifier: 
                73:FE:F7:59:A7:E1:26:84:44:D6:44:36:EE:79:1A:95:7C:B1:4B:15
            X509v3 Authority Key Identifier: 
                keyid:73:FE:F7:59:A7:E1:26:84:44:D6:44:36:EE:79:1A:95:7C:B1:4B:15, DirName:/C=AU/ST=Some-State/O=Internet Widgits Pty Ltd/Email=email@1.address/Email=email@2.address, serial:00
            X509v3 Key Usage: 
                Certificate Sign, CRL Sign
            X509v3 Subject Alternative Name: 
                email:email@1.address, email:email@2.address

CONFIGURATION FILES.

The OpenSSL utilities 'ca' and 'req' can now have extension sections listing
which certificate extensions to include. In each case a line:

x509_extensions = extension_section

indicates which section contains the extensions. In the case of 'req' the
extension section is used when the -x509 option is present to create a
self signed root certificate.

The 'x509' utility also supports extensions when it signs a certificate.
The -extfile option is used to set the configuration file containing the
extensions. In this case a line with:

extensions = extension_section

in the nameless (default) section is used. If no such line is included then
it uses the default section.

You can also add extensions to CRLs: a line

crl_extensions = crl_extension_section

will include extensions when the -gencrl option is used with the 'ca' utility.
You can add any extension to a CRL but of the supported extensions only
issuerAltName and authorityKeyIdentifier make any real sense. Note: these are
CRL extensions NOT CRL *entry* extensions which cannot currently be generated.
CRL entry extensions can be displayed.

NB. At this time Netscape Communicator rejects V2 CRLs: to get an old V1 CRL
you should not include a crl_extensions line in the configuration file.

As with all configuration files you can use the inbuilt environment expansion
to allow the values to be passed in the environment. Therefore if you have
several extension sections used for different purposes you can have a line:

x509_extensions = $ENV::ENV_EXT

and set the ENV_EXT environment variable before calling the relevant utility.

EXTENSION SYNTAX.

Extensions have the basic form:

extension_name=[critical,] extension_options

the use of the critical option makes the extension critical. Extreme caution
should be made when using the critical flag. If an extension is marked
as critical then any client that does not understand the extension should
reject it as invalid. Some broken software will reject certificates which
have *any* critical extensions (these violates PKIX but we have to live
with it).

There are three main types of extension: string extensions, multi-valued
extensions, and raw extensions.

String extensions simply have a string which contains either the value itself
or how it is obtained.

For example:

nsComment="This is a Comment"

Multi-valued extensions have a short form and a long form. The short form
is a list of names and values:

basicConstraints=critical,CA:true,pathlen:1

The long form allows the values to be placed in a separate section:

basicConstraints=critical,@bs_section

[bs_section]

CA=true
pathlen=1

Both forms are equivalent. However it should be noted that in some cases the
same name can appear multiple times, for example,

subjectAltName=email:steve@here,email:steve@there

in this case an equivalent long form is:

subjectAltName=@alt_section

[alt_section]

email.1=steve@here
email.2=steve@there

This is because the configuration file code cannot handle the same name
occurring twice in the same section.

The syntax of raw extensions is governed by the extension code: it can
for example contain data in multiple sections. The correct syntax to
use is defined by the extension code itself: check out the certificate
policies extension for an example.

In addition it is also possible to use the word DER to include arbitrary
data in any extension.

1.2.3.4=critical,DER:01:02:03:04
1.2.3.4=DER:01020304

The value following DER is a hex dump of the DER encoding of the extension
Any extension can be placed in this form to override the default behaviour.
For example:

basicConstraints=critical,DER:00:01:02:03

WARNING: DER should be used with caution. It is possible to create totally
invalid extensions unless care is taken.

CURRENTLY SUPPORTED EXTENSIONS.

If you aren't sure about extensions then they can be largely ignored: its only
when you want to do things like restrict certificate usage when you need to
worry about them. 

The only extension that a beginner might want to look at is Basic Constraints.
If in addition you want to try Netscape object signing the you should also
look at Netscape Certificate Type.

Literal String extensions.

In each case the 'value' of the extension is placed directly in the
extension. Currently supported extensions in this category are: nsBaseUrl,
nsRevocationUrl, nsCaRevocationUrl, nsRenewalUrl, nsCaPolicyUrl,
nsSslServerName and nsComment.

For example:

nsComment="This is a test comment"

Bit Strings.

Bit string extensions just consist of a list of supported bits, currently
two extensions are in this category: PKIX keyUsage and the Netscape specific
nsCertType.

nsCertType (netscape certificate type) takes the flags: client, server, email,
objsign, reserved, sslCA, emailCA, objCA.

keyUsage (PKIX key usage) takes the flags: digitalSignature, nonRepudiation,
keyEncipherment, dataEncipherment, keyAgreement, keyCertSign, cRLSign,
encipherOnly, decipherOnly.

For example:

nsCertType=server

keyUsage=digitalSignature, nonRepudiation

Hints on Netscape Certificate Type.

Other than Basic Constraints this is the only extension a beginner might
want to use, if you want to try Netscape object signing, otherwise it can
be ignored.

If you want a certificate that can be used just for object signing then:

nsCertType=objsign

will do the job. If you want to use it as a normal end user and server
certificate as well then

nsCertType=objsign,email,server

is more appropriate. You cannot use a self signed certificate for object
signing (well Netscape signtool can but it cheats!) so you need to create
a CA certificate and sign an end user certificate with it.

Side note: If you want to conform to the Netscape specifications then you
should really also set:

nsCertType=objCA

in the *CA* certificate for just an object signing CA and

nsCertType=objCA,emailCA,sslCA

for everything. Current Netscape software doesn't enforce this so it can
be omitted.

Basic Constraints.

This is generally the only extension you need to worry about for simple
applications. If you want your certificate to be usable as a CA certificate
(in addition to an end user certificate) then you set this to:

basicConstraints=CA:TRUE

if you want to be certain the certificate cannot be used as a CA then do:

basicConstraints=CA:FALSE

The rest of this section describes more advanced usage.

Basic constraints is a multi-valued extension that supports a CA and an
optional pathlen option. The CA option takes the values true and false and
pathlen takes an integer. Note if the CA option is false the pathlen option
should be omitted. 

The pathlen parameter indicates the maximum number of CAs that can appear
below this one in a chain. So if you have a CA with a pathlen of zero it can
only be used to sign end user certificates and not further CAs. This all
assumes that the software correctly interprets this extension of course.

Examples:

basicConstraints=CA:TRUE
basicConstraints=critical,CA:TRUE, pathlen:0

NOTE: for a CA to be considered valid it must have the CA option set to
TRUE. An end user certificate MUST NOT have the CA value set to true.
According to PKIX recommendations it should exclude the extension entirely,
however some software may require CA set to FALSE for end entity certificates.

Extended Key Usage.

This extensions consists of a list of usages.

These can either be object short names of the dotted numerical form of OIDs.
While any OID can be used only certain values make sense. In partiular the
following PKIX, NS and MS values are meaningful:

Value                   Meaning
-----                   -------
serverAuth              SSL/TLS Web Server Authentication.
clientAuth              SSL/TLS Web Client Authentication.
codeSigning             Code signing.
emailProtection         E-mail Protection (S/MIME).
timeStamping            Trusted Timestamping
msCodeInd               Microsoft Individual Code Signing (authenticode)
msCodeCom               Microsoft Commercial Code Signing (authenticode)
msCTLSign               Microsoft Trust List Signing
msSGC                   Microsoft Server Gated Crypto
msEFS                   Microsoft Encrypted File System
nsSGC                   Netscape Server Gated Crypto

For example, under IE5 a CA can be used for any purpose: by including a list
of the above usages the CA can be restricted to only authorised uses.

Note: software packages may place additional interpretations on certificate 
use, in particular some usages may only work for selected CAs. Don't for example
expect just including msSGC or nsSGC will automatically mean that a certificate
can be used for SGC ("step up" encryption) otherwise anyone could use it.

Subject Key Identifier.

This is really a string extension and can take two possible values. Either
a hex string giving details of the extension value to include or the word
'hash' which then automatically follow PKIX guidelines in selecting and
appropriate key identifier. The use of the hex string is strongly discouraged.

Example: subjectKeyIdentifier=hash

Authority Key Identifier.

The authority key identifier extension permits two options. keyid and issuer:
both can take the optional value "always".

If the keyid option is present an attempt is made to copy the subject key
identifier from the parent certificate. If the value "always" is present
then an error is returned if the option fails.

The issuer option copies the issuer and serial number from the issuer
certificate. Normally this will only be done if the keyid option fails or
is not included: the "always" flag will always include the value.

Subject Alternative Name.

The subject alternative name extension allows various literal values to be
included in the configuration file. These include "email" (an email address)
"URI" a uniform resource indicator, "DNS" (a DNS domain name), RID (a
registered ID: OBJECT IDENTIFIER) and IP (and IP address).

Also the email option include a special 'copy' value. This will automatically
include and email addresses contained in the certificate subject name in
the extension.

Examples:

subjectAltName=email:copy,email:my@other.address,URL:http://my.url.here/
subjectAltName=email:my@other.address,RID:1.2.3.4

Issuer Alternative Name.

The issuer alternative name option supports all the literal options of
subject alternative name. It does *not* support the email:copy option because
that would not make sense. It does support an additional issuer:copy option
that will copy all the subject alternative name values from the issuer 
certificate (if possible).

CRL distribution points.

This is a multi-valued extension that supports all the literal options of
subject alternative name. Of the few software packages that currently interpret
this extension most only interpret the URI option.

Currently each option will set a new DistributionPoint with the fullName
field set to the given value.

Other fields like cRLissuer and reasons cannot currently be set or displayed:
at this time no examples were available that used these fields.

If you see this extension with <UNSUPPORTED> when you attempt to print it out
or it doesn't appear to display correctly then let me know, including the
certificate (mail me at steve@openssl.org) .

Examples:

crlDistributionPoints=URI:http://www.myhost.com/myca.crl
crlDistributionPoints=URI:http://www.my.com/my.crl,URI:http://www.oth.com/my.crl

Certificate Policies.

This is a RAW extension. It attempts to display the contents of this extension:
unfortunately this extension is often improperly encoded.

The certificate policies extension will rarely be used in practice: few
software packages interpret it correctly or at all. IE5 does partially
support this extension: but it needs the 'ia5org' option because it will
only correctly support a broken encoding. Of the options below only the
policy OID, explicitText and CPS options are displayed with IE5.

All the fields of this extension can be set by using the appropriate syntax.

If you follow the PKIX recommendations of not including any qualifiers and just
using only one OID then you just include the value of that OID. Multiple OIDs
can be set separated by commas, for example:

certificatePolicies= 1.2.4.5, 1.1.3.4

If you wish to include qualifiers then the policy OID and qualifiers need to
be specified in a separate section: this is done by using the @section syntax
instead of a literal OID value.

The section referred to must include the policy OID using the name
policyIdentifier, cPSuri qualifiers can be included using the syntax:

CPS.nnn=value

userNotice qualifiers can be set using the syntax:

userNotice.nnn=@notice

The value of the userNotice qualifier is specified in the relevant section.
This section can include explicitText, organization and noticeNumbers
options. explicitText and organization are text strings, noticeNumbers is a
comma separated list of numbers. The organization and noticeNumbers options
(if included) must BOTH be present. If you use the userNotice option with IE5
then you need the 'ia5org' option at the top level to modify the encoding:
otherwise it will not be interpreted properly.

Example:

certificatePolicies=ia5org,1.2.3.4,1.5.6.7.8,@polsect

[polsect]

policyIdentifier = 1.3.5.8
CPS.1="http://my.host.name/"
CPS.2="http://my.your.name/"
userNotice.1=@notice

[notice]

explicitText="Explicit Text Here"
organization="Organisation Name"
noticeNumbers=1,2,3,4

TECHNICAL NOTE: the ia5org option changes the type of the 'organization' field,
according to PKIX it should be of type DisplayText but Verisign uses an 
IA5STRING and IE5 needs this too.

Display only extensions.

Some extensions are only partially supported and currently are only displayed
but cannot be set. These include private key usage period, CRL number, and
CRL reason.

==============================================================================
                X509V3 Extension code: programmers guide
==============================================================================

The purpose of the extension code is twofold. It allows an extension to be
created from a string or structure describing its contents and it prints out an
extension in a human or machine readable form.

1. Initialisation and cleanup.

X509V3_add_standard_extensions();

This function should be called before any other extension code. It adds support
for some common PKIX and Netscape extensions. Additional custom extensions can
be added as well (see later).

void X509V3_EXT_cleanup(void);

This function should be called last to cleanup the extension code. After this
call no other extension calls should be made.

2. Printing and parsing extensions.

The simplest way to print out extensions is via the standard X509 printing
routines: if you use the standard X509_print() function, the supported
extensions will be printed out automatically.

The following functions allow finer control over extension display:

int X509V3_EXT_print(BIO *out, X509_EXTENSION *ext, int flag, int indent);
int X509V3_EXT_print_fp(FILE *out, X509_EXTENSION *ext, int flag, int indent);

These two functions print out an individual extension to a BIO or FILE pointer.
Currently the flag argument is unused and should be set to 0. The 'indent'
argument is the number of spaces to indent each line.

void *X509V3_EXT_d2i(X509_EXTENSION *ext);

This function parses an extension and returns its internal structure. The
precise structure you get back depends on the extension being parsed. If the
extension if basicConstraints you will get back a pointer to a
BASIC_CONSTRAINTS structure. Check out the source in crypto/x509v3 for more
details about the structures returned. The returned structure should be freed
after use using the relevant free function, BASIC_CONSTRAINTS_free() for 
example.

3. Generating extensions.

An extension will typically be generated from a configuration file, or some
other kind of configuration database.

int X509V3_EXT_add_conf(LHASH *conf, X509V3_CTX *ctx, char *section,
                                                                 X509 *cert);
int X509V3_EXT_CRL_add_conf(LHASH *conf, X509V3_CTX *ctx, char *section,
                                                                 X509_CRL *crl);

These functions add all the extensions in the given section to the given
certificate or CRL. They will normally be called just before the certificate
or CRL is due to be signed. Both return 0 on error on non zero for success.

In each case 'conf' is the LHASH pointer of the configuration file to use
and 'section' is the section containing the extension details.

See the 'context functions' section for a description of the ctx paramater.


X509_EXTENSION *X509V3_EXT_conf(LHASH *conf, X509V3_CTX *ctx, char *name,
                                                                 char *value);

This function returns an extension based on a name and value pair, if the
pair will not need to access other sections in a config file (or there is no
config file) then the 'conf' parameter can be set to NULL.

X509_EXTENSION *X509V3_EXT_conf_nid(char *conf, X509V3_CTX *ctx, int nid,
                                                                 char *value);

This function creates an extension in the same way as X509V3_EXT_conf() but
takes the NID of the extension rather than its name.

For example to produce basicConstraints with the CA flag and a path length of
10:

x = X509V3_EXT_conf_nid(NULL, NULL, NID_basic_constraints,"CA:TRUE,pathlen:10");


X509_EXTENSION *X509V3_EXT_i2d(int ext_nid, int crit, void *ext_struc);

This function sets up an extension from its internal structure. The ext_nid
parameter is the NID of the extension and 'crit' is the critical flag.

4. Context functions.

The following functions set and manipulate an extension context structure.
The purpose of the extension context is to allow the extension code to
access various structures relating to the "environment" of the certificate:
for example the issuers certificate or the certificate request.

void X509V3_set_ctx(X509V3_CTX *ctx, X509 *issuer, X509 *subject,
                                 X509_REQ *req, X509_CRL *crl, int flags);

This function sets up an X509V3_CTX structure with details of the certificate
environment: specifically the issuers certificate, the subject certificate,
the certificate request and the CRL: if these are not relevant or not
available then they can be set to NULL. The 'flags' parameter should be set
to zero.

X509V3_set_ctx_test(ctx)

This macro is used to set the 'ctx' structure to a 'test' value: this is to
allow the syntax of an extension (or configuration file) to be tested.

X509V3_set_ctx_nodb(ctx)

This macro is used when no configuration database is present.

void X509V3_set_conf_lhash(X509V3_CTX *ctx, LHASH *lhash);

This function is used to set the configuration database when it is an LHASH
structure: typically a configuration file.

The following functions are used to access a configuration database: they
should only be used in RAW extensions.

char * X509V3_get_string(X509V3_CTX *ctx, char *name, char *section);

This function returns the value of the parameter "name" in "section", or NULL
if there has been an error.

void X509V3_string_free(X509V3_CTX *ctx, char *str);

This function frees up the string returned by the above function.

STACK_OF(CONF_VALUE) * X509V3_get_section(X509V3_CTX *ctx, char *section);

This function returns a whole section as a STACK_OF(CONF_VALUE) .

void X509V3_section_free( X509V3_CTX *ctx, STACK_OF(CONF_VALUE) *section);

This function frees up the STACK returned by the above function.

Note: it is possible to use the extension code with a custom configuration
database. To do this the "db_meth" element of the X509V3_CTX structure should
be set to an X509V3_CTX_METHOD structure. This structure contains the following
function pointers:

char * (*get_string)(void *db, char *section, char *value);
STACK_OF(CONF_VALUE) * (*get_section)(void *db, char *section);
void (*free_string)(void *db, char * string);
void (*free_section)(void *db, STACK_OF(CONF_VALUE) *section);

these will be called and passed the 'db' element in the X509V3_CTX structure
to access the database. If a given function is not implemented or not required
it can be set to NULL.

5. String helper functions.

There are several "i2s" and "s2i" functions that convert structures to and
from ASCII strings. In all the "i2s" cases the returned string should be
freed using Free() after use. Since some of these are part of other extension
code they may take a 'method' parameter. Unless otherwise stated it can be
safely set to NULL.

char *i2s_ASN1_OCTET_STRING(X509V3_EXT_METHOD *method, ASN1_OCTET_STRING *oct);

This returns a hex string from an ASN1_OCTET_STRING.

char * i2s_ASN1_INTEGER(X509V3_EXT_METHOD *meth, ASN1_INTEGER *aint);
char * i2s_ASN1_ENUMERATED(X509V3_EXT_METHOD *meth, ASN1_ENUMERATED *aint);

These return a string decimal representations of an ASN1_INTEGER and an
ASN1_ENUMERATED type, respectively.

ASN1_OCTET_STRING *s2i_ASN1_OCTET_STRING(X509V3_EXT_METHOD *method,
                                                   X509V3_CTX *ctx, char *str);

This converts an ASCII hex string to an ASN1_OCTET_STRING.

ASN1_INTEGER * s2i_ASN1_INTEGER(X509V3_EXT_METHOD *meth, char *value);

This converts a decimal ASCII string into an ASN1_INTEGER.

6. Multi valued extension helper functions.

The following functions can be used to manipulate STACKs of CONF_VALUE
structures, as used by multi valued extensions.

int X509V3_get_value_bool(CONF_VALUE *value, int *asn1_bool);

This function expects a boolean value in 'value' and sets 'asn1_bool' to
it. That is it sets it to 0 for FALSE or 0xff for TRUE. The following
strings are acceptable: "TRUE", "true", "Y", "y", "YES", "yes", "FALSE"
"false", "N", "n", "NO" or "no".

int X509V3_get_value_int(CONF_VALUE *value, ASN1_INTEGER **aint);

This accepts a decimal integer of arbitrary length and sets an ASN1_INTEGER.

int X509V3_add_value(const char *name, const char *value,
                                                STACK_OF(CONF_VALUE) **extlist);

This simply adds a string name and value pair.

int X509V3_add_value_uchar(const char *name, const unsigned char *value,
                                                STACK_OF(CONF_VALUE) **extlist);

The same as above but for an unsigned character value.

int X509V3_add_value_bool(const char *name, int asn1_bool,
                                                STACK_OF(CONF_VALUE) **extlist);

This adds either "TRUE" or "FALSE" depending on the value of 'ans1_bool'

int X509V3_add_value_bool_nf(char *name, int asn1_bool,
                                                STACK_OF(CONF_VALUE) **extlist);

This is the same as above except it adds nothing if asn1_bool is FALSE.

int X509V3_add_value_int(const char *name, ASN1_INTEGER *aint,
                                                STACK_OF(CONF_VALUE) **extlist);

This function adds the value of the ASN1_INTEGER in decimal form.

7. Other helper functions.

<to be added>

ADDING CUSTOM EXTENSIONS.

Currently there are three types of supported extensions. 

String extensions are simple strings where the value is placed directly in the
extensions, and the string returned is printed out.

Multi value extensions are passed a STACK_OF(CONF_VALUE) name and value pairs
or return a STACK_OF(CONF_VALUE).

Raw extensions are just passed a BIO or a value and it is the extensions
responsiblity to handle all the necessary printing.

There are two ways to add an extension. One is simply as an alias to an already
existing extension. An alias is an extension that is identical in ASN1 structure
to an existing extension but has a different OBJECT IDENTIFIER. This can be
done by calling:

int X509V3_EXT_add_alias(int nid_to, int nid_from);

'nid_to' is the new extension NID and 'nid_from' is the already existing
extension NID.

Alternatively an extension can be written from scratch. This involves writing
the ASN1 code to encode and decode the extension and functions to print out and
generate the extension from strings. The relevant functions are then placed in
a X509V3_EXT_METHOD structure and int X509V3_EXT_add(X509V3_EXT_METHOD *ext);
called.

The X509V3_EXT_METHOD structure is described below.

strut {
int ext_nid;
int ext_flags;
X509V3_EXT_NEW ext_new;
X509V3_EXT_FREE ext_free;
X509V3_EXT_D2I d2i;
X509V3_EXT_I2D i2d;
X509V3_EXT_I2S i2s;
X509V3_EXT_S2I s2i;
X509V3_EXT_I2V i2v;
X509V3_EXT_V2I v2i;
X509V3_EXT_R2I r2i;
X509V3_EXT_I2R i2r;

void *usr_data;
};

The elements have the following meanings.

ext_nid         is the NID of the object identifier of the extension.

ext_flags       is set of flags. Currently the only external flag is
                X509V3_EXT_MULTILINE which means a multi valued extensions
                should be printed on separate lines.

usr_data        is an extension specific pointer to any relevant data. This
                allows extensions to share identical code but have different
                uses. An example of this is the bit string extension which uses
                usr_data to contain a list of the bit names.

All the remaining elements are function pointers.

ext_new         is a pointer to a function that allocates memory for the
                extension ASN1 structure: for example ASN1_OBJECT_new().

ext_free        is a pointer to a function that free up memory of the extension
                ASN1 structure: for example ASN1_OBJECT_free().

d2i             is the standard ASN1 function that converts a DER buffer into
                the internal ASN1 structure: for example d2i_ASN1_IA5STRING().

i2d             is the standard ASN1 function that converts the internal
                structure into the DER representation: for example
                i2d_ASN1_IA5STRING().

The remaining functions are depend on the type of extension. One i2X and
one X2i should be set and the rest set to NULL. The types set do not need
to match up, for example the extension could be set using the multi valued
v2i function and printed out using the raw i2r.

All functions have the X509V3_EXT_METHOD passed to them in the 'method'
parameter and an X509V3_CTX structure. Extension code can then access the
parent structure via the 'method' parameter to for example make use of the value
of usr_data. If the code needs to use detail relating to the request it can
use the 'ctx' parameter.

A note should be given here about the 'flags' member of the 'ctx' parameter.
If it has the value CTX_TEST then the configuration syntax is being checked
and no actual certificate or CRL exists. Therefore any attempt in the config
file to access such information should silently succeed. If the syntax is OK
then it should simply return a (possibly bogus) extension, otherwise it
should return NULL.

char *i2s(struct v3_ext_method *method, void *ext);

This function takes the internal structure in the ext parameter and returns
a Malloc'ed string representing its value.

void * s2i(struct v3_ext_method *method, struct v3_ext_ctx *ctx, char *str);

This function takes the string representation in the ext parameter and returns
an allocated internal structure: ext_free() will be used on this internal
structure after use.

i2v and v2i handle a STACK_OF(CONF_VALUE):

typedef struct
{
        char *section;
        char *name;
        char *value;
} CONF_VALUE;

Only the name and value members are currently used.

STACK_OF(CONF_VALUE) * i2v(struct v3_ext_method *method, void *ext);

This function is passed the internal structure in the ext parameter and
returns a STACK of CONF_VALUE structures. The values of name, value,
section and the structure itself will be freed up with Free after use.
Several helper functions are available to add values to this STACK.

void * v2i(struct v3_ext_method *method, struct v3_ext_ctx *ctx,
                                                STACK_OF(CONF_VALUE) *values);

This function takes a STACK_OF(CONF_VALUE) structures and should set the
values of the external structure. This typically uses the name element to
determine which structure element to set and the value element to determine
what to set it to. Several helper functions are available for this
purpose (see above).

int i2r(struct v3_ext_method *method, void *ext, BIO *out, int indent);

This function is passed the internal extension structure in the ext parameter
and sends out a human readable version of the extension to out. The 'indent'
paremeter should be noted to determine the necessary amount of indentation
needed on the output.

void * r2i(struct v3_ext_method *method, struct v3_ext_ctx *ctx, char *str);

This is just passed the string representation of the extension. It is intended
to be used for more elaborate extensions where the standard single and multi
valued options are insufficient. They can use the 'ctx' parameter to parse the
configuration database themselves. See the context functions section for details
of how to do this.

Note: although this type takes the same parameters as the "r2s" function there
is a subtle difference. Whereas an "r2i" function can access a configuration
database an "s2i" function MUST NOT. This is so the internal code can safely
assume that an "s2i" function will work without a configuration database.

==============================================================================
                            PKCS#12 Library
==============================================================================

This section describes the internal PKCS#12 support. There are very few
differences between the old external library and the new internal code at
present. This may well change because the external library will not be updated
much in future.

This version now includes a couple of high level PKCS#12 functions which
generally "do the right thing" and should make it much easier to handle PKCS#12
structures.

HIGH LEVEL FUNCTIONS.

For most applications you only need concern yourself with the high level
functions. They can parse and generate simple PKCS#12 files as produced by
Netscape and MSIE or indeed any compliant PKCS#12 file containing a single
private key and certificate pair.

1. Initialisation and cleanup.

No special initialisation is needed for the internal PKCS#12 library: the 
standard SSLeay_add_all_algorithms() is sufficient. If you do not wish to
add all algorithms (you should at least add SHA1 though) then you can manually
initialise the PKCS#12 library with:

PKCS12_PBE_add();

The memory allocated by the PKCS#12 library is freed up when EVP_cleanup() is
called or it can be directly freed with:

EVP_PBE_cleanup();

after this call (or EVP_cleanup() ) no more PKCS#12 library functions should
be called.

2. I/O functions.

i2d_PKCS12_bio(bp, p12)

This writes out a PKCS12 structure to a BIO.

i2d_PKCS12_fp(fp, p12)

This is the same but for a FILE pointer.

d2i_PKCS12_bio(bp, p12)

This reads in a PKCS12 structure from a BIO.

d2i_PKCS12_fp(fp, p12)

This is the same but for a FILE pointer.

3. Parsing and creation functions.

3.1 Parsing with PKCS12_parse().

int PKCS12_parse(PKCS12 *p12, char *pass, EVP_PKEY **pkey, X509 **cert,
                                                                 STACK **ca);

This function takes a PKCS12 structure and a password (ASCII, null terminated)
and returns the private key, the corresponding certificate and any CA
certificates. If any of these is not required it can be passed as a NULL.
The 'ca' parameter should be either NULL, a pointer to NULL or a valid STACK
structure. Typically to read in a PKCS#12 file you might do:

p12 = d2i_PKCS12_fp(fp, NULL);
PKCS12_parse(p12, password, &pkey, &cert, NULL);        /* CAs not wanted */
PKCS12_free(p12);

3.2 PKCS#12 creation with PKCS12_create().

PKCS12 *PKCS12_create(char *pass, char *name, EVP_PKEY *pkey, X509 *cert,
                        STACK *ca, int nid_key, int nid_cert, int iter,
                                                 int mac_iter, int keytype);

This function will create a PKCS12 structure from a given password, name,
private key, certificate and optional STACK of CA certificates. The remaining
5 parameters can be set to 0 and sensible defaults will be used.

The parameters nid_key and nid_cert are the key and certificate encryption
algorithms, iter is the encryption iteration count, mac_iter is the MAC
iteration count and keytype is the type of private key. If you really want
to know what these last 5 parameters do then read the low level section.

Typically to create a PKCS#12 file the following could be used:

p12 = PKCS12_create(pass, "My Certificate", pkey, cert, NULL, 0,0,0,0,0);
i2d_PKCS12_fp(fp, p12);
PKCS12_free(p12);

LOW LEVEL FUNCTIONS.

In some cases the high level functions do not provide the necessary
functionality. For example if you want to generate or parse more complex
PKCS#12 files. The sample pkcs12 application uses the low level functions
to display details about the internal structure of a PKCS#12 file.

Introduction.

This is a brief description of how a PKCS#12 file is represented internally:
some knowledge of PKCS#12 is assumed.

A PKCS#12 object contains several levels.

At the lowest level is a PKCS12_SAFEBAG. This can contain a certificate, a
CRL, a private key, encrypted or unencrypted, a set of safebags (so the
structure can be nested) or other secrets (not documented at present). 
A safebag can optionally have attributes, currently these are: a unicode
friendlyName (a Unicode string) or a localKeyID (a string of bytes).

At the next level is an authSafe which is a set of safebags collected into
a PKCS#7 ContentInfo. This can be just plain data, or encrypted itself.

At the top level is the PKCS12 structure itself which contains a set of
authSafes in an embedded PKCS#7 Contentinfo of type data. In addition it
contains a MAC which is a kind of password protected digest to preserve
integrity (so any unencrypted stuff below can't be tampered with).

The reason for these levels is so various objects can be encrypted in various
ways. For example you might want to encrypt a set of private keys with
triple-DES and then include the related certificates either unencrypted or
with lower encryption. Yes it's the dreaded crypto laws at work again which
allow strong encryption on private keys and only weak encryption on other
stuff.

To build one of these things you turn all certificates and keys into safebags
(with optional attributes). You collect the safebags into (one or more) STACKS
and convert these into authsafes (encrypted or unencrypted).  The authsafes
are collected into a STACK and added to a PKCS12 structure.  Finally a MAC
inserted.

Pulling one apart is basically the reverse process. The MAC is verified against
the given password. The authsafes are extracted and each authsafe split into
a set of safebags (possibly involving decryption). Finally the safebags are
decomposed into the original keys and certificates and the attributes used to
match up private key and certificate pairs.

Anyway here are the functions that do the dirty work.

1. Construction functions.

1.1 Safebag functions.

M_PKCS12_x5092certbag(x509)

This macro takes an X509 structure and returns a certificate bag. The
X509 structure can be freed up after calling this function.

M_PKCS12_x509crl2certbag(crl)

As above but for a CRL.

PKCS8_PRIV_KEY_INFO *PKEY2PKCS8(EVP_PKEY *pkey)

Take a private key and convert it into a PKCS#8 PrivateKeyInfo structure.
Works for both RSA and DSA private keys. NB since the PKCS#8 PrivateKeyInfo
structure contains a private key data in plain text form it should be free'd
up as soon as it has been encrypted for security reasons (freeing up the
structure zeros out the sensitive data). This can be done with
PKCS8_PRIV_KEY_INFO_free().

PKCS8_add_keyusage(PKCS8_PRIV_KEY_INFO *p8, int usage)

This sets the key type when a key is imported into MSIE or Outlook 98. Two
values are currently supported: KEY_EX and KEY_SIG. KEY_EX is an exchange type
key that can also be used for signing but its size is limited in the export
versions of MS software to 512 bits, it is also the default. KEY_SIG is a
signing only key but the keysize is unlimited (well 16K is supposed to work).
If you are using the domestic version of MSIE then you can ignore this because
KEY_EX is not limited and can be used for both.

PKCS12_SAFEBAG *PKCS12_MAKE_KEYBAG(PKCS8_PRIV_KEY_INFO *p8)

Convert a PKCS8 private key structure into a keybag. This routine embeds the
p8 structure in the keybag so p8 should not be freed up or used after it is
called.  The p8 structure will be freed up when the safebag is freed.

PKCS12_SAFEBAG *PKCS12_MAKE_SHKEYBAG(int pbe_nid, unsigned char *pass, int passlen, unsigned char *salt, int saltlen, int iter, PKCS8_PRIV_KEY_INFO *p8)

Convert a PKCS#8 structure into a shrouded key bag (encrypted). p8 is not
embedded and can be freed up after use.

int PKCS12_add_localkeyid(PKCS12_SAFEBAG *bag, unsigned char *name, int namelen)
int PKCS12_add_friendlyname(PKCS12_SAFEBAG *bag, unsigned char *name, int namelen)

Add a local key id or a friendlyname to a safebag.

1.2 Authsafe functions.

PKCS7 *PKCS12_pack_p7data(STACK *sk)
Take a stack of safebags and convert them into an unencrypted authsafe. The
stack of safebags can be freed up after calling this function.

PKCS7 *PKCS12_pack_p7encdata(int pbe_nid, unsigned char *pass, int passlen, unsigned char *salt, int saltlen, int iter, STACK *bags);

As above but encrypted.

1.3 PKCS12 functions.

PKCS12 *PKCS12_init(int mode)

Initialise a PKCS12 structure (currently mode should be NID_pkcs7_data).

M_PKCS12_pack_authsafes(p12, safes)

This macro takes a STACK of authsafes and adds them to a PKCS#12 structure.

int PKCS12_set_mac(PKCS12 *p12, unsigned char *pass, int passlen, unsigned char *salt, int saltlen, int iter, EVP_MD *md_type);

Add a MAC to a PKCS12 structure. If EVP_MD is NULL use SHA-1, the spec suggests
that SHA-1 should be used.

2. Extraction Functions.

2.1 Safebags.

M_PKCS12_bag_type(bag)

Return the type of "bag". Returns one of the following

NID_keyBag
NID_pkcs8ShroudedKeyBag                 7
NID_certBag                             8
NID_crlBag                              9
NID_secretBag                           10
NID_safeContentsBag                     11

M_PKCS12_cert_bag_type(bag)

Returns type of certificate bag, following are understood.

NID_x509Certificate                     14
NID_sdsiCertificate                     15

M_PKCS12_crl_bag_type(bag)

Returns crl bag type, currently only NID_crlBag is recognised.

M_PKCS12_certbag2x509(bag)

This macro extracts an X509 certificate from a certificate bag.

M_PKCS12_certbag2x509crl(bag)

As above but for a CRL.

EVP_PKEY * PKCS82PKEY(PKCS8_PRIV_KEY_INFO *p8)

Extract a private key from a PKCS8 private key info structure.

M_PKCS12_decrypt_skey(bag, pass, passlen) 

Decrypt a shrouded key bag and return a PKCS8 private key info structure.
Works with both RSA and DSA keys

char *PKCS12_get_friendlyname(bag)

Returns the friendlyName of a bag if present or NULL if none. The returned
string is a null terminated ASCII string allocated with Malloc(). It should 
thus be freed up with Free() after use.

2.2 AuthSafe functions.

M_PKCS12_unpack_p7data(p7)

Extract a STACK of safe bags from a PKCS#7 data ContentInfo.

#define M_PKCS12_unpack_p7encdata(p7, pass, passlen)

As above but for an encrypted content info.

2.3 PKCS12 functions.

M_PKCS12_unpack_authsafes(p12)

Extract a STACK of authsafes from a PKCS12 structure.

M_PKCS12_mac_present(p12)

Check to see if a MAC is present.

int PKCS12_verify_mac(PKCS12 *p12, unsigned char *pass, int passlen)

Verify a MAC on a PKCS12 structure. Returns an error if MAC not present.


Notes.

1. All the function return 0 or NULL on error.
2. Encryption based functions take a common set of parameters. These are
described below.

pass, passlen
ASCII password and length. The password on the MAC is called the "integrity
password" the encryption password is called the "privacy password" in the
PKCS#12 documentation. The passwords do not have to be the same. If -1 is
passed for the length it is worked out by the function itself (currently
this is sometimes done whatever is passed as the length but that may change).

salt, saltlen
A 'salt' if salt is NULL a random salt is used. If saltlen is also zero a
default length is used.

iter
Iteration count. This is a measure of how many times an internal function is
called to encrypt the data. The larger this value is the longer it takes, it
makes dictionary attacks on passwords harder. NOTE: Some implementations do
not support an iteration count on the MAC. If the password for the MAC and
encryption is the same then there is no point in having a high iteration
count for encryption if the MAC has no count. The MAC could be attacked
and the password used for the main decryption.

pbe_nid
This is the NID of the password based encryption method used. The following are
supported.
NID_pbe_WithSHA1And128BitRC4
NID_pbe_WithSHA1And40BitRC4
NID_pbe_WithSHA1And3_Key_TripleDES_CBC
NID_pbe_WithSHA1And2_Key_TripleDES_CBC
NID_pbe_WithSHA1And128BitRC2_CBC
NID_pbe_WithSHA1And40BitRC2_CBC

Which you use depends on the implementation you are exporting to. "Export
grade" (i.e. cryptographically challenged) products cannot support all
algorithms. Typically you may be able to use any encryption on shrouded key
bags but they must then be placed in an unencrypted authsafe. Other authsafes
may only support 40bit encryption. Of course if you are using SSLeay
throughout you can strongly encrypt everything and have high iteration counts
on everything.

3. For decryption routines only the password and length are needed.

4. Unlike the external version the nid's of objects are the values of the
constants: that is NID_certBag is the real nid, therefore there is no 
PKCS12_obj_offset() function.  Note the object constants are not the same as
those of the external version. If you use these constants then you will need
to recompile your code.

5. With the exception of PKCS12_MAKE_KEYBAG(), after calling any function or 
macro of the form PKCS12_MAKE_SOMETHING(other) the "other" structure can be
reused or freed up safely.