stunnel - universal SSL tunnel
stunnel [filename] | -help | -version | -sockets
The stunnel program is designed to work as SSL encryption wrapper between remote clients and local (inetd-startable) or remote servers. The concept is that having non-SSL aware daemons running on your system you can easily set them up to communicate with clients over secure SSL channels.
stunnel can be used to add SSL functionality to commonly used Inetd daemons like POP-2, POP-3, and IMAP servers, to standalone daemons like NNTP, SMTP and HTTP, and in tunneling PPP over network sockets without changes to the source code.
This product includes cryptographic software written by Eric Young (firstname.lastname@example.org)
Each line of the configuration file can be either:
This is the directory in which stunnel will look for certificates when using the verify. Note that the certificates in this directory should be named XXXXXXXX.0 where XXXXXXXX is the hash value of the cert.
This file contains multiple CA certificates, used with the verify.
A PEM is always needed in server mode. Specifying this flag in client mode will use this certificate chain as a client side certificate chain. Using client side certs is optional. The certificates must be in PEM format and must be sorted starting with the certificate to the highest level (root CA).
chroot keeps stunnel in chrooted jail. CApath, pid and exec are located inside the jail and the patches have to be relative to the directory specified with chroot.
To have libwrap (TCP Wrappers) control effective in a chrooted environment you also have to copy its configuration files (/etc/hosts.allow and /etc/hosts.deny) there.
A colon delimited list of the ciphers to allow in the SSL connection. For example DES-CBC3-SHA:IDEA-CBC-MD5
default: no (server mode)
Level is a one of the syslog level names or numbers emerg (0), alert (1), crit (2), err (3), warning (4), notice (5), info (6), or debug (7). All logs for the specified level and all levels numerically less than it will be shown. Use debug = debug or debug = 7 for greatest debugging output. The default is notice (5).
The syslog facility 'daemon' will be used unless a facility name is supplied. (Facilities are not supported on Win32.)
Case is ignored for both facilities and levels.
Entropy Gathering Daemon socket to use to feed OpenSSL random number generator. (Available only if compiled with OpenSSL 0.9.5a or higher)
Stay in foreground (don't fork) and log to stderr instead of via syslog (unless output is specified).
default: background in daemon mode
Private key is needed to authenticate certificate owner. Since this file should be kept secret it should only be readable to its owner. On Unix systems you can use the following command:
chmod 600 keyfile
default: value of cert option
The parameter is the OpenSSL option name as described in the SSL_CTX_set_options(3ssl) manual, but without SSL_OP_ prefix. Several options can be used to specify multiple options.
For example for compatibility with erroneous Eudora SSL implementation the following option can be used:
options = DONT_INSERT_EMPTY_FRAGMENTS
If the argument is empty, then no pid file will be created.
Number of bytes of data read from random seed files. With SSL versions less than 0.9.5a, also determines how many bytes of data are considered sufficient to seed the PRNG. More recent OpenSSL versions have a builtin function to determine when sufficient randomness is available.
The SSL library will use data from this file first to seed the random number generator.
On Unix: inetd mode service name for TCP Wrapper library.
On NT/2000/XP: NT service name in the Control Panel.
setgid()to groupname in daemon mode and clears all other groups
setuid()to username in daemon mode
The values for linger option are l_onof:l_linger. The values for time are tv_sec:tv_usec.
socket = l:SO_LINGER=1:60 set one minute timeout for closing local socket socket = r:TCP_NODELAY=1 turn off the Nagle algorithm for remote sockets socket = r:SO_OOBINLINE=1 place out-of-band data directly into the receive data stream for remote sockets socket = a:SO_REUSEADDR=0 disable address reuse (enabled by default) socket = a:SO_BINDTODEVICE=lo only accept connections on loopback interface
level 1 - verify peer certificate if present level 2 - verify peer certificate level 3 - verify peer with locally installed certificate default - no verify
Each configuration section begins with service name in square brackets. The service name is used for libwrap (TCP Wrappers) access control and lets you distinguish stunnel services in your log files.
Note that if you wish to run stunnel in inetd mode (where it is provided a network socket by a server such as inetd, xinetd, or tcpserver) then you should read the section entitiled INETD MODE below.
If no host specified, defaults to all IP addresses for the local host.
If no host specified, defaults to localhost.
Quoting is currently not supported. Arguments are speparated with arbitrary number of whitespaces.
currently supported: smtp, pop3, nntp
Re-write address to appear as if wrapped daemon is connecting from the SSL client machine instead of the machine running stunnel. This option is only available in local mode (exec option) by LD_PRELOADing env.so shared library or in remote mode (connect option) on Linux 2.2 kernel compiled with transparent proxy option and then only in server mode. Note that this option will not combine with proxy mode (connect) unless the client's default route to the target machine lies through the host running stunnel, which cannot be localhost.
stunnel returns zero on success, non-zero on error.
In order to provide SSL encapsulation to your local imapd service, use
[imapd] accept = 993 exec = /usr/sbin/imapd execargs = imapd
If you want to provide tunneling to your pppd daemon on port 2020, use something like
[vpn] accept = 2020 exec = /usr/sbin/pppd execargs = pppd local pty = yes
If you want to use stunnel in inetd mode to launch your imapd process, you'd use this stunnel.conf. Note there must be no [service_name] section.
exec = /usr/sbin/imapd execargs = imapd
Option execargs does not support quoting.
stunnel cannot be used for the FTP daemon because of the nature of the FTP protocol which utilizes multiple ports for data transfers. There are available SSL enabled versions of FTP and telnet daemons, however.
The most common use of stunnel is to listen on a network port and establish communication with either a new port via the connect option, or a new program via the exec option. However there is a special case when you wish to have some other program accept incoming connections and launch stunnel, for example with inetd, xinetd, or tcpserver.
For example, if you have the following line in inetd.conf:
imaps stream tcp nowait root /usr/sbin/stunnel stunnel /etc/stunnel/imaps.conf
In these cases, the inetd-style program is responsible for binding a network socket (imaps above) and handing it to stunnel when a connection is received. Thus you do not want stunnel to have any accept option. All the Service Level Options should be placed in the global options section, and no [service_name] section will be present. See the EXAMPLES section for example configurations.
Each SSL enabled daemon needs to present a valid X.509 certificate to the peer. It also needs a private key to decrypt the incoming data. The easiest way to obtain a certificate and a key is to generate them with the free OpenSSL package. You can find more information on certificates generation on pages listed below.
Two things are important when generating certificate-key pairs for stunnel. The private key cannot be encrypted, because the server has no way to obtain the password from the user. To produce an unencrypted key add the -nodes option when running the req command from the OpenSSL kit.
The order of contents of the .pem file is also important. It should contain the unencrypted private key first, then a signed certificate (not certificate request). There should be also empty lines after certificate and private key. Plaintext certificate information appended on the top of generated certificate should be discarded. So the file should look like this:
-----BEGIN RSA PRIVATE KEY----- [encoded key] -----END RSA PRIVATE KEY----- [empty line] -----BEGIN CERTIFICATE----- [encoded certificate] -----END CERTIFICATE----- [empty line]
stunnel needs to seed the PRNG (pseudo random number generator) in order for SSL to use good randomness. The following sources are loaded in order until sufficient random data has been gathered:
With recent (>=OpenSSL 0.9.5a) version of SSL it will stop loading random data automatically when sufficient entropy has been gathered. With previous versions it will continue to gather from all the above sources since no SSL function exists to tell when enough data is available.
Note that on Windows machines that do not have console user interaction (mouse movements, creating windows, etc) the screen contents are not variable enough to be sufficient, and you should provide a random file for use with the RNDfile flag.
Note that the file specified with the RNDfile flag should contain random data -- that means it should contain different information each time stunnel is run. This is handled automatically unless the RNDoverwrite flag is used. If you wish to update this file manually, the openssl rand command in recent versions of OpenSSL, would be useful.
One important note -- if /dev/urandom is available, OpenSSL has a habit of seeding the PRNG with it even when checking the random state, so on systems with /dev/urandom you're likely to use it even though it's listed at the very bottom of the list above. This isn't stunnel's behaviour, it's OpenSSLs.