Allegro CL Socket Library

This document contains the following sections:

1.0 Introduction and background

Sockets are a mechanism for interprocess communication designed at U.C. Berkeley for use in their version of Unix. Sockets have been added to many other versions of Unix and there is an implementation of sockets for Windows called Winsock. This document describes the Allegro interface to sockets. This interface works on Unix and on Windows.

Symbols naming objects in the socket utility are in the acl-socket package. It has the nickname socket.

The socket module is not included in all versions of Allegro CL. If it is present, it is (by default) included in a development image (one built with the include-devel-env argument to build-lisp-image specified true). To load the socket module if it is not present in an image, evaluate

(require :sock)

Note that runtime images cannot include the development environment (so include-devel-env must be specified nil when a runtime image is being built). If the socket module is needed, it must be loaded when the image is built. See runtime.htm, building-images.htm and delivery.htm for more information.

2.0 Support for Internet Protocol version 6 (IPv6)

Allegro CL supports Internet Protocol version 6 sockets (IPv6 sockets). As part of this support, several new functions have been added and several functions have been modified. The new functions are ipv6, get-ip-interfaces, ipaddrp, ipaddr-equalp, ipv6-address-p, and (added in a June, s006 update) dotted-address-p. The modified functions are dotted-to-ipaddr, dns-query, lookup-hostname, make-socket, and send-to. There is also a new variable *ipv6*.

The feature :ipv6 is added to the *features* list to indicate IPv6 support.

IPv6 support may be unusable on MacOS X 64-bit

Because of apparent bugs in Mac OS X 64-bit, certain IPv6 functionality may be restricted or unusable. In particular:

• The get-ip-interfaces function will signal an error.
• Name-based scope ids will not work.

3.0 IP addresses in Allegro CL

Throughout the socket documentation, we make use of the term IP address. But what exactly is an IP address? Unless further clarified in the context in which it is used, an IP address is either an unsigned 32-bit integer or an ipv6 address structure.

The function socket:ipaddrp returns true when passed an IP adress. That function can be used to identify an object as an IP address. (Unsigned 32 bit integers obviously have other uses that representing IP addresses. The function simply determines whether the type and form of its argument is suitable as an IP address.)

4.0 Characteristics

There are three independent characteristics of sockets:

Note on bivalent format:

Starting in release 5.0.1, the bivalent format is accepted for stream sockets. Bivalent means that the stream will accept text and binary stream functions. That is, you can write-byte or write-char, read-byte or read-char. A bivalent stream is useful in the http protocol (used between web browsers and web servers) since in that protocol the header data is sent in text format and the body can be in binary data (image files, for example).

Internally a bivalent socket stream is configured like a binary socket stream with 8 bit bytes. Character position is not maintained.

Bivalent socket streams have very efficient read-sequence and write-sequence implementations (as long as the sequence is either a vector of element-type character, (unsigned-byte 8) or (signed-byte 8)).

Bivalent socket streams also support the chunking protocol found in http/1.1. This protocol allows the sender to signal end of file without closing down the stream.

5.0 Stream Sockets

5.1 Connections

Stream sockets have a fourth characteristic called connect, with a value :active or :passive. In order to use stream sockets you have to set up a link between two of them. That link is called a connection. You set up a connection in this way:

1. Machine A: create a passive socket at port port-b:

   (setq s-a (make-socket :connect :passive :local-port port-b))
   

2. Machine B: create an active socket telling it to connect to Machine A, port port-b:

   (setq s-b (make-socket :remote-host "machine-a" 
                          :remote-port port-b))
   

3. Machine A: wait for a connect request from anyone and when it occurs return a stream for I/O:

   (setq str-a (accept-connection s-a))
   

4. When the accept-connection returns, machine A can use stream str-a to send messages to machine B and machine B can use stream s-b to send messages to machine A.

Note that steps 2 and 3 can occur in either order.

Note the asymmetry: a passive socket is not a Lisp stream (you can't do read and write to it). An active socket is a Lisp stream.

When accept-connection is called on a passive socket, it does not return until a connection is made to the passive socket. The value accept-connection returns is a stream.

As long as the passive socket is not closed, new connections can still be made to the port of that socket.

An active socket can be used for only one connection. Once that connection has been made, the socket should be closed and a new active socket created.

5.2 Host Naming

Host naming conventions: this package supports three conventions for naming a host:

6.0 Variables

The variables defined by the interface are:

*socket-version*

Please provide the value of this variable when asking for technical support with sockets as it tells us whether you have the latest version.

*print-hostname-in-stream*

This variable controls whether the socket printing code converts the ip address of a socket into a hostname. This is usually what you want, however this can be a slow process (taking up to a minute to accomplish). The default value for this variable is t. See the full description for a discussion of the causes of the possible slowdown when the value is t.

*ipv6*

Specifies the default value of the ipv6 keyword argument to lookup-hostname and make-socket.

7.0 Functions

The first table shows general functions defined by the interface and the second shows accessors.

Socket Accessors

These functions retrieve slot values from socket instances. The values of these slots are set when the socket is created.

8.0 Errors

When errors are raised by the socket interface, Lisp conditions are signaled. This section describes those conditions.

A condition is a CLOS class and thus fits into the hierarchy of CLOS classes. The condition socket-error is a subclass of the condition stream-error.

socket-error is the superclass for all socket related errors. See More on cl:stream-error in errors.htm.

socket-error denotes operating system detected socket errors. It has the following slots:

Handling socket error is difficult because the error returned in exceptional situations can depend on the operating system and the address of the other side of the connection. For example, attempting to make a connection to a machine that is down may result in a "Connection Timed Out" or a "Host Unreachable" error, or maybe something else on certain systems.

The error codes assigned to socket errors vary from operating system to operating system. We translate a large set of the common error codes from a machine dependent number to a symbol which we call the identifier to make it easier for you to write portable code. Condition handling code should check the identifier field (using stream-error-identifier) If the identifier value is :unknown then this is not a common socket error and the operating system dependent code value of the condition must be used.

Possible identifier values and their meanings:

9.0 Sockets and SMP

An SMP (symmetric multiprocessing) Lisp runs one of more OS threads which can use more than one processor (see multiprocessing.htm). When a socket is opened in one SMP process, do not close it from another process.

If you have a process which is reading from a socket and that socket fails so the reading process hangs, you can wake the process up using a process interrupt. So assuming hung-process is the reading process and hung-socket is the socket, the following form executed in another process breaks the read (with an error) and closes the socket:

   (mp:process-interrupt hung-process 'close hung-socket)

You can also wrap the read in a (catch 'dead-partner ...) and the from another process do:

   (mp:process-run-function hung-process
                            #'(lambda ()
                                 (throw 'dead-partner t)))

10.0 Examples

USER(1): (let ((s (make-socket :remote-host "vapor" :remote-port "chargen")))
           (dotimes (i 5) (print (read-char s))) (close s))
#\space
#\!
#\"
#\#
#\$

USER(1): (print (read (accept-connection
                       (make-socket :connect :passive :local-port 9933))))
.. this hangs ...

USER(1): (let ((s (make-socket :remote-host "vapor" :remote-port 9933)))
    (format s "Secret-message~%") (close s)) 

Secret-message
Secret-message
USER(2): 

user(2): (setq s (make-socket :type :datagram :local-port 9999))
#<text datagram socket waiting for connection at */9999 @ #x20664e82>
user(3):  

user(10): (setq x (make-socket :type :datagram))
#<text datagram socket waiting for connection at */45602 @ #x20717fb2>
user(11): (send-to x "foo-the-bar" 11 :remote-host "vapor" :remote-port 9999)
11
user(12): 

user(3): (receive-from s 100 :extract t)

"foo-the-bar"

11 ;; length of result

3229900653 ;; frisky's IP address

45602 ;; the port number chosen for the socket by frisky

user(4): 

11.0 Secure Socket Layer (SSL)

Allegro CL supports Secure Socket layers as described in this section. See also aserve/aserve.html, which describes Webserver support in Allegro CL. Using any https feature in aserve triggers loading of the :ssl module (i.e. evaluates (require :ssl)).

Starting with a patch released in June, 2016, Allegro CL loads OpenSSL libraries dynamically when the :ssl module is loaded. Also loaded is the aclssl or aclissl shared library (which depends on the type of Allegro CL being run). Prior to the patch, the OpenSSL library was linked into the aclsslaclissl library (except on Windows, which has always worked like all platforms now work). This made installation and usage easier but meant that a patch and an update was required whenever a new version of the OpenSSL library was released. Now, newly installed versions of OpenSSL libraries will be used when installed on your computer without (usually) requiring an Allegro CL update. However, it is now the user's responsibility to ensure that OpenSSL libraries are available, properly updated, and findable.

This documentation has been updated to reflect behavior assuming the new SSL patch has been installed. See sys:update-allegro for information on getting patches.

Users must themselves download and install OpenSSL libraries for their computers, and make sure these are updated when necessary. On some platforms, such as the Mac and some Linux machines, this is often done automatically. Further, Users must ensure that Allegro CL knows where to find the OpenSSL libraries. On some platforms, there are standard locations where Allegro CL will look. On all platforms, environment variables or Lisp variables can be used to specify the library location.

Here is how to set things up for using OpenSSL:

• If your machine automatically installs OpenSSL libraries, it is likely everything will work without further action. Try (require :ssl) in Lisp. If it works, you are done. Allegro CL will always use the latest OpenSSL library installed on your machine.
• If your machine does not automatically install and update OpenSSL libraries, you must get them and update them yourself. See Installing OpenSSL libraries in installation.htm for information on installating OpenSSL libraries. If you install them in the standard location on your platform, nothing further should be required. Try (require :ssl). If the library cannot be found, you must set the appropriate environment variable (always prior to starting Allegro CL).
• If you want to use a version of the OpenSSL libraries not in the standard location and perhaps not with their standard names, that is supported on Windows and Linux. Simply set the appropriate environment variable and/or Lisp variable appropriately. See *ssl-library-names*.

In all cases, if (require :ssl) loads the SSL fasl file and the aclssl/aclissl library and loads the OpenSSL library without error, you are ready to use SSL.

The function openssl-version returns information about the version of the OpenSSL library which is loaded and the version used to build the aclssl/acl1ssl library.

Most Allegro CL platforms support SSL. All that do have :ssl-support included on the *features* list.

The SSL functionality is in the ssl module. To ensure it is loaded, evaluate (require :ssl). Calling either of the two SSL functions, make-ssl-client-stream and make-ssl-server-stream, automatically loads that module.

If you are including the SSL facility in an application intended for delivery, be sure to include the :ssl module by adding the keyword :ssl to the list which is the value of the input-files argument to generate-application. If the copy-shared-libraries argument to that function is false, then the aclssl/aclissl shared library must be copied explicitly from the Allegro CL installation directory to the generated application directory. When the generated application is run, it must be possible to locate the the OpenSSL shared libraries in the same way that is described above.

OpenSSL version

Our aclssl and aclissl shared libraries depend on the OpenSSL version specified in Required versions of OpenSSL needed for the :ssl module in the release-notes. SSL library releases often have lettered suffixes corresponding to fixes, and you should always be running the latest, patched version. For some operating systems, such as RHEL/CentOS, Fedora and Ubuntu, patches usually come out quickly after vulnerabilities are found, and the fixed versions may and usually do appear older than the versions available from http://openssl.org. For example, when this was written, at which time the current OpenSSL version is 1.0.1t, CentOS 6.8 had a patched 1.0.1e, which had the same fixes for vulnerabilities as 1.0.1t from http://openssl.org.

If your application uses foreign libraries that use OpenSSL, those foreign libraries must be compatible with a version of OpenSSL compatible with our :ssl module. Otherwise, our :ssl module cannot be used.

11.1 The algorithm for loading the OpenSSL library

In this section, we give the exact steps on the various platforms for finding and loading the OpenSSL library.

There are two library files. Their names (without the file extensions) are in a list which is the value of *ssl-library-names*. The initial value of that variable is the standard library names on each platform (so the initial value may be different on different platforms). It is usually not necessary to change the value of that variable. The system will not look for libraries with names other than specified by that variable.

Here are the steps for finding the libraries on the various supported platforms:

Finding OpenSSL library on Windows:

• If the environment variable ACL_SSL_LIBRARY_NAMES is set, its value will be a string concatenating two filenames separated by a single space. Load those files (PATH is set to modify the search used by the system loader).
• If the environment variable is not set, the value of *ssl-library-names* should be a list of two items, each a string nameing a file with an extension. Search the directories in the PATH environment variable for a directory containing the first file.
• If a directory is not found, signal an error, otherwise, merge each filename with the directory and load it.

After the OpenSSL library files are loaded, load aclssl or aclissl (depending on what version of Allegro CL is running).

Finding OpenSSL library on Linux:

• If the environment variable ACL_SSL_LIBRARY_NAMES is set, its value must be a string concatenating two filenames separated by a single space. Load those files (LD_LIBRARY_PATH is set to modify the search used by the system loader).
• If the environment variable is not set, the value of *ssl-library-names* should be a list of two items, each a string or a list of strings. If the element is a string, it should name a file complete with extension. Load that file. If the value is a list, it should be a list of filenames with extensions. Try to load each until one succeeds. LD_LIBRARY_PATH is set to modify the search used by the system loader.

After the OpenSSL library files are loaded, load aclssl or aclissl (depending on what version of Allegro CL is running).

Finding OpenSSL library on Mac OS X, Freebsd, and Solaris:

Simply load aclssl or aclissl and assume that the system loader will find the OpenSSL libraries. The system loader looks in the directory specified in DYLD_LIBRARY_PATH (as well as other places) on the Mac and in LD_LIBRARY_PATH (as well as other places) on Freebsd, and Solaris.

11.2 SSL History

In 1994 Netscape Corporation designed the Secure Socket Layer (SSL) protocol to provide a means of safely and securely transporting private data (such as credit card numbers) between a Web Browser and a Web Server. Rather than tie SSL to the http protocol, Netscape wrote it as a protocol for making any TCP/IP connection secure.

At the end of 1994 version 2 of SSL was introduced and this was the first version shipped with a commercial web browser (Netscape Navigator (r)). In 1995 version 3 of SSL was introduced. At that point an international standards organization (IETF) took over work on SSL and introduced Transport Layer Security (or TLS) protocol (which is based on SSL but has a different handshake protocol). The IETF introduced TLS version 1.0 in 1999.

Allegro CL, starting in release 6.0, provides an interface that supports SSL version 2, SSL version 3 and TLS version 1. When we use the name SSL, we mean SSL or TLS.

11.3 Secure connections

A secure TCP connection exists between two processes when both agree on the following:

• A cryptographic algorithm (called a cipher) to convert data (plaintext) to encrypted data (ciphertext) and back to plaintext.
• A secret key to use in the cryptographic algorithm. A secret key is essential since the cryptographic algorithms themselves are well known. There is a distinct key for each direction of data transmission (client to server and server to client).
• An authentication method used in each packet to ensure that the contents of the packets have not been altered in transit in some way.

These three items are determined via negotiation when the connection is made and the first data is to be sent.

11.4 Client/Server

In an SSL connection, one side is the client and the other side is the server. In the http environment, the web browser is the client and the web server is the server.

When a secure connection is started, the client starts the negotation by telling the server all the possible ways that it can communicate securely. The server then chooses one of the possible ways and informs the client.

Then the server sends its certificate and possibly other certificates if they are needed to prove that its certificate can be trusted. The important item in the certificate is the public key for the server. The client will use this public key to encrypt a random value which will be used by both the client and server to create the keys needed for the cipher chosen for data transmission.

In theory a certificate isn't necessary if both the client and server side support a key exchange algorithm that can generate a public key on the fly. The SSL libraries we use do not have this capability, thus you must always supply a server certificate.

Once both sides know the keys the other side will use to transmit, the secure data transmission can occur.

11.5 Authentication

The SSL protocol also permits each side of the connection to declare who they are. This is done by the exchange of certificates. The server must send a certificate describing itself to the client. The server can request that the client send a certificate to the server (although in the use of SSL on the web this is never done).

11.6 Certificates

A certificate is a digital document that stores information about an entity in such a way that it can be verified to be true. The primary use of certificates is to store the public key that can be used to send encrypted messages to the entity.

In the SSL protocol certificates have two uses:

1. Encrpytion - by providing a public key they enable encrypted messages to be sent.
2. Authentication - the certificate proves that the entity on the other end of the socket is who it claims to be.

Strictly speaking a certificate isn't required for SSL communication if both sides support a certain key exchange protocol. The OpenSSL libraries we use do not support this protocol thus whenever you create a server SSL stream you must supply a certificate (if you don't have your own we supply one in <Allegro directory>/examples/ssl/server.pem that you can use).

While certificates support authentication, the SSL protocol doesn't require that you take advantage of this facility.

A certificate contains the following:

1. A Subject Identifier: a set of fields describing where the subject is geographically and its role within an organization.
2. A Subject Public Key: the key that can be used to encrypt messages that only the Subject can decrypt since only the Subject has the associated private key.
3. A Valid Time Interval: the interval of time during which this certificate is valid.
4. An Issuer Identifier: just like the Subject Identifier but describing the entity that certifies that the Subject is who it says it is and that the public key is the correct one for the subject.
5. An Issuer signature: a value which can be used by anyone to verify that the Issuer and only the Issuer signed this document testifying to it being correct.
6. Various other fields: like serial numbers, version numbers, and other minor things.

A certificate is a combination of text and binary data and in order to make it easy to transport certificates they are usually encoded in a form called PEM which turns them into a sequence of printable characters.

When a web browser connects to a site via SSL (which is caused by the use of the 'https:' at the beginning of the url), it checks three things about the certificate:

1. Does it know the Issuer and did the Issuer sign the certificate? A web browser knows about a set of Issuers (called Certificate Authorities) when it's installed on the machine (the Issuer certificates are part of the files that make up the web browser).
2. Is the certificate valid right now or has it expired?
3. Is the certificate for the machine we've contacted? If the url was https://www.foo.com/whatever then the certificate must be for www.foo.com. The convention used is to store the name of the server machine in the CommonName slot of the Subject Identifier field of the certificate.

If all three tests pass then the web browser silently accepts the certificate and does a secure web page access. If any of the tests fail then the web browser notifies the user and waits for a response. Each browser displays the failure differently. For example, the Microsoft Internet Explorer (r) shows which of the three tests passed and which failed while the Netscape Navigator (r) just says that it received an invalid certificate. In both cases the person using the web browser is given the option of continuing with the web access. Transmission will still be secure if it is elected to continue. The only issue in doubt is the authenticity of the web server.

11.7 CRL support

The SSL implementation includes certificate revocation list (CRL) support. CRL checking is controlled by the crl-check and crl-file keyword arguments to make-ssl-client-stream and make-ssl-server-stream.

If you enable CRL checking, you must supply a proper PEM-encoded CRL, even if it contains zero revocations. If you do not supply a CRL, peer verification will never succeed.

11.8 The Allegro CL SSL API

The following operators, variable and class comprise the SSL API. make-ssl-client-stream and make-ssl-server-stream create the streams that are used for communication. All symbols naming these objects are in the acl-socket (nicknamed socket) package.

• get-ssl-peer-certificate (generic function)
• get-ssl-verify-result (generic function)
• make-ssl-server-context (function)
• make-ssl-client-context (function)
• make-ssl-client-stream (function)
• make-ssl-server-stream (function)
• ssl-do-handshake (generic function)
• x509-certificate (class)
• x509-certificate-issuer (generic function)
• x509-certificate-not-after (generic function)
• x509-certificate-not-before (generic function)
• x509-certificate-serial-number (generic function)
• x509-certificate-issuer (generic function)
• x509-certificate-subject (generic function)
• x509-certificate-subject-alt-name (generic function)
• x509-certificate-version (generic function)

The following function, variable, and classes (naming conditions) are related to loading and using OpenSSL libraries. The conditions are signaled when there are problems loading the OpenSSL library. All these symbols are in the excl package.

• *ssl-library-names*
• openssl-version
• openssl-load-error
• aclssl-load-error

The file <Allegro directory>/examples/ssl/server.pem is a sample certificate and private key file. You can use this file when starting the server side of an SSL connection. The AllegroServe facility uses SSL. It is described in aserve/aserve.html.

Copyright (c) 1998-2019, Franz Inc. Oakland, CA., USA. All rights reserved.
This page has had significant revisions compared to the 8.2 page.
Created 2019.8.20.