Overview of an SSL Application


Chapter 1
Overview of an SSL Application

SSL and related APIs allow compliant applications to configure sockets for authenticated, tamper-proof, and encrypted communications. This chapter introduces some of the basic SSL functions. Chapter 2, "Getting Started With SSL" illustrates their use in sample client and server applications.

An SSL application typically includes five parts:

Initialization
Configuration
Communication
Functions Used by Callbacks
Cleanup

Although the details differ somewhat for client and server applications, the concepts and many of the functions are the same for both.

WARNING: Some of the SSL header files provided as part of NSS 2.0 include both public APIs documented in the NSS 2.0 documentation set and private APIs intended for internal use by the NSS implementation of SSL. You should use only the SSL APIs (and related certificate, key, and PKCS #11 APIs) that are described in this document, the SSL Reference. Other APIs that may be exposed in the header files are not supported for application use.

Initialization

Initialization includes setting up configuration files, setting global defaults, and setting up callback functions. Functions used in the initialization part of an application can include the following:

Initializing Caches

SSL peers frequently reconnect after a relatively short time has passed. To avoid the overhead of repeating the full SSL handshake in situations like this, the SSL protocol supports the use of a session cache, which retains information about each connection, such as the master secret generated during the SSL handshake, for a predetermined length of time. If SSL can locate the information about a previous connection in the local session cache, it can reestablish the connection much more quickly than it can without the connection information.

By default, SSL allocates one session cache. This default cache is called the client session ID cache, (also known as the client session cache, or simply the client cache). The client cache is used for all sessions where the program handshakes as an SSL client. It is not configurable. You can initialize the client cache with the function SSL_ClearSessionCache.

If an application will use SSL sockets that handshake as a server, you must specifically create and configure a server cache, using either SSL_ConfigServerSessionIDCache or SSL_ConfigMPServerSIDCache. The server cache is used for all sessions where the program handshakes as an SSL server.

Configuration

The configuration portion of an SSL-enabled application typically begins by opening a new socket and then importing the new socket into the SSL environment:

It is also possible for an application to import a socket into SSL after the TCP connection on that socket has already been established. In this case, initial configuration takes place in the same way: pass the existing NSPR file descriptor to SSL_ImportFD and perform any additional configuration that has not already been determined by the model file descriptor.

Configuration functions control the configuration of an individual socket.

Callbacks and helper functions allow you to specify such things as how authentication is accomplished and what happens if it fails.

Communication

At this point the application has set up the socket to communicate using SSL. For simple encrypted and authenticated communications, no further calls to SSL functions are required. A variety of additional SSL functions are available, however. These can be used, for example, when interrupting and restarting socket communications, when the application needs to change socket parameters, or when an application imports a socket into SSL after the TCP connection on that socket has already been established.

Communication between SSL sockets always begins with the SSL handshake. The handshake occurs automatically the first time communication is requested with a socket read/write or send/receive call. It is also possible to force the handshake explicitly with SSL_ForceHandshake or repeat it explicitly with SSL_ReHandshake.

Once the SSL sockets have been configured, authentication and encryption happen automatically whenever you use the communication functions from the NSPR library.

A server application typically uses these functions to establish a connection:

PR_Bind
PR_Listen
PR_Accept
PR_GetSockName

A client application typically uses these functions to establish a connection:

PR_GetHostByName
PR_EnumerateHostEnt
PR_Connect
PR_GetConnectStatus

When an application imports a socket into SSL after the TCP connection on that socket has already been established, it must call SSL_ResetHandshake to determine whether SSL should behave like an SSL client or an SSL server. Note that this step would not be necessary if the socket weren't already connected. For an SSL socket that is configured before it is connected, SSL figures this out when the application calls PR_Connect or PR_Accept. If the socket is already connected before SSL gets involved, you must provide this extra hint.

Functions that can be used by both clients and servers during communication include the following:

PR_Send or PR_Write
PR_Read or PR_Recv
PR_GetError
PR_GetPeerName
PR_Sleep
PR_Malloc
PR_Free
PR_Poll
PR_Now
PR_IntervalToMilliseconds
PR_MillisecondsToInterval
PR_Shutdown
PR_Close
SSL_InvalidateSession

After establishing a connection, an application first calls PR_Send, PR_Recv, PR_Read, PR_Write, or SSL_ForceHandshake to initiate the handshake. The application's protocol (for example, HTTP) determines which end has responsibility to talk first. The end that has to talk first should call PR_Send or PR_Write, and the other end should call PR_Read or PR_Recv.

Use SSL_ForceHandshake when the socket has been prepared for a handshake but neither end has anything to say immediately. This occurs, for example, when an HTTPS server has received a request and determines that before it can answer the request, it needs to request an authentication certificate from the client. At the HTTP protocol level, nothing more is being said (that is, no HTTP request or response is being sent), so the server first uses SSL_ReHandshake to begin a new handshake and then call SSL_ForceHandshake to drive the handshake to completion.

Functions Used by Callbacks

An SSL application typically provides one or more callback functions that are called by the SSL or PKCS #11 library code under certain circumstances. Numerous functions provided by the NSS libraries are useful for such application callback functions, including these:

CERT_CheckCertValidTimes
CERT_GetDefaultCertDB
CERT_DestroyCertificate
CERT_DupCertificate
CERT_FindCertByName
CERT_FreeNicknames
CERT_GetCertNicknames
CERT_VerifyCertName
CERT_VerifyCertNow
PK11_FindCertFromNickname
PK11_FindKeyByAnyCert
PK11_SetPasswordFunc
PL_strcpy
PL_strdup
PL_strfree
PL_strlen
SSL_PeerCertificate
SSL_RevealURL
SSL_RevealPinArg

Cleanup

This portion of an SSL-enabled application consists primarily of closing the socket and freeing memory. After these tasks have been performed, call NSS_Shutdown to close the certificate and key databases opened by NSS_Init, and PR_Cleanup to coordinate a graceful shutdown of NSPR.