ssh-server-config

ssh-server-config — SSH Tectia Server configuration file format

The SSH Tectia Server configuration file ssh-server-config.xml is a valid XML file.

On Unix, the configuration related files are stored as follows:

ssh-server-config.xml file in directory /etc/ssh2/
XML DTD in directory /etc/ssh2/ssh-tectia/auxdata/ssh-server-ng.

On Windows, the configuration related files are stored as follows:

ssh-server-config.xml in directory "C:\Program Files\SSH Communications Security\SSH Tectia\SSH Tectia Server"
XML DTD in directory "C:\Program Files\SSH Communications Security\SSH Tectia\SSH Tectia AUX\ssh-server-ng".

If the configuration file cannot be found or some of the elements are missing, hardcoded default values are used. You can view the default values in the ssh-server-config-default.xml file that is stored in the same directory with the configuration file. However, this default configuration file is not actually read. It only shows the hardcoded system defaults.

In addition, an example configuration file ssh-server-config-example.xml can be found in the same directory with the configuration file. This file contains useful examples of the various configuration options.

The configuration file is divided in four blocks:

general server parameters (params)
connection rules and encryption methods (connections)
authentication rules and methods (authentication-methods)
service rules (services)

In the connections and authentication-methods blocks, different selectors can be used to set access rules to users based on the user parameters such as user name or location. Users can be divided to groups dynamically, for example, based on the authentication method they used when logging in. In the services block, each group can then be allowed or denied services such as tunneling, file transfer, and terminal access.

	Note
	The configuration file is read in top-down order during connection setup. If a connection is denied in one of the blocks, the connection setup phase ends immediately and the rest of the configuration file is not read.

This section describes the options available in the SSH Tectia Server configuration file. See Appendix B for more information on the syntax of the configuration file.

Selectors

The connection settings can be changed based on selectors in the configuration file. Using selectors makes it possible, for example, to:

allow/deny connections from certain IP addresses
require different authentication methods based on user name or group
restrict access based on a certificate field

Selectors are used in the connections, authentication-methods, and services blocks.

The different selector attributes are specified as sub-elements of the selector element. The following sub-elements are available:

certificate: Matches to a pattern in a specified field of the user certificate.
host-certificate: Matches to a pattern in a specified field of the client host certificate.
interface: Matches to the server listener interface ID or address/port.
ip: Matches to the IP address or FQDN of the client.
publickey-passed: Matches if the authentication is passed using a public key.
user: Matches to a user name or ID.
user-group: Matches to a user group name or ID.
user-password-change-needed (Unix): Matches if the user password has expired and should be changed.
user-privileged: Matches based on the user privilege status (yes/no).

In the connections block, only the interface and ip selectors can be used. In the authentication-methods and services blocks, all selectors can be used.

When a parent element contains multiple child elements with selectors, the first functional child element matched is used, and the rest are ignored. For example, if the connections element has multiple connection child elements, but the first one has an empty selector, or no selectors at all, that connection element always matches and the remaining ones are never used.

On Windows domain environment, the user and user-group selectors have a length limitation. For more information, see the description of option User in Configuration Tool (Windows).

Wild Cards in Selectors

Simple wild cards can be used inside selector values.

An asterisk (*) matches any string. A question mark (?) matches any character. A dash (-) can be used to specify a range.

For example, the following selector matches to user names jdoe and jdox, but not to jdoel:

<selector>
  <user name="jdo?" />
</selector>

For example, the following selector matches to IP addresses between 192.168.0.42 and 192.168.0.82:

<selector>
  <ip address="192.168.0.42-192.168.0.82" />
</selector>

Selector Processing

Multiple selector elements are in an OR relation (one of the selector elements must match for the parent element to match). For example, the following block matches if either the IP address is 192.168.0.3 or the user ID is 1001:

<selector>
  <ip address="192.168.0.3" />
</selector>
<selector>
  <user id="1001" />
</selector>

Selector attributes in the same selector element are normally in an AND relation (all attributes must match for the element to match). For example, the following block matches if both the IP address is 192.168.0.3 and the user ID is 1001:

<selector>
  <ip address="192.168.0.3" />
  <user id="1001" />
</selector>

However, selector attributes in the same selector element matching to the same attribute type are in an OR relation to each other. The following three examples produce the same result, either the user name exa or mple matches:

<selector>
  <user name="exa" />
  <user name="mple" />
</selector>

<selector>
  <user name="exa" />
</selector>
<selector>
  <user name="mple" />
</selector>

<selector>
  <user name="exa,mple" />
</selector>

An empty selector always matches:

<selector />

Also, typically, if an element accepts selectors, but none are given, the element is assumed to have an empty selector, which will then always match.

Selectors and Undefined Data

Normally when the server tries to match to a selector attribute for which the respective data has not been defined (the data is not available to the server), the selector matching process ends in error, effectively terminating the connection attempt. This happens, for example, in the following cases:

Other selectors than ip and interface are erroneously used in the connections block. Only the IP address of the client and the connected listener interface are available to the server in that stage of connection. For example, the user name is not yet known.
The certificate selector is erroneously used without previously requiring public-key authentication. The server will not have user certificate data unless it has received it first during public-key authentication.
The host-certificate selector is erroneously used without previously requiring host-based authentication. The server will not have host certificate data unless it has received it first during host-based authentication.
The certificate or host-certificate selector is used to match to a field that does not exist in the certificate.
The user-privileged selector is used in the authentication-methods block on a Windows server and the user is logging in using a domain account and does not yet have an access token allocated.

The allow-undefined attribute can be used in all selector sub-elements to control this behavior. Its value must be yes or no. If set to yes, the undefined data is treated as non-matched and the matching continues to other elements. The default is no (trying to match undefined data results in termination of the connection).

For example, encountering the following selector causes the connection attempt to end in failure if the certificate is not available or does not contain the altname-email field:

<selector>
  <certificate 
   field="altname-email" 
   pattern="%username%@ssh.com" />
</selector>

The following selector simply does not match when the certificate does not exist or does not contain the altname-email field, and the processing continues with the next block:

<selector>
  <certificate 
   field="altname-email" 
   pattern="%username%@ssh.com" 
   allow-undefined="yes" />
</selector>

Document Type Declaration and the Root Element

The server configuration file is a valid XML file and starts with the Document Type Declaration.

The root element in the configuration file is secsh-server. It can include params, connections, authentication-methods, and services elements.

An example of an empty configuration file is shown below:

<!DOCTYPE secsh-server SYSTEM
    "../auxdata/ssh-server-ng/ssh-server-ng-config-1.dtd">

<secsh-server>

  <params />
  <connections>
    <connection />
  </connections>
  <authentication-methods />
  <services>
    <rule />
  </services>

</secsh-server>

	Note
	It is not mandatory to include all elements in the configuration file. If an element is missing the default values shown in the `ssh-server-config-default.xml` file are used. However, the configuration file should always contain the `DOCTYPE` declaration. If the Document Type Declaration is omitted, the server will not be able to parse the configuration properly.

The `params` Block

The params block defines the general server parameters, such as the location of the host key file, the listen address, logging, connection limits, and certificate validation settings.

crypto-lib

This element selects the cryptographic library mode to be used. Either the standard version (standard) or the FIPS 140-2 certified version (fips) of the crypto library can be used. The library name is given as a value of the mode attribute. By default, standard crypto libraries are used.

<crypto-lib mode="standard" />

In the FIPS mode, the cryptographic operations are performed according to the rules of the FIPS 140-2 standard. The FIPS library includes the 3des-cbc, aes128-cbc, aes192-cbc, and aes256-cbc ciphers and the hmac-sha1 MAC.

	Note
	Setting the FIPS mode does not prevent using algorithms from crypto plugins. For example, CryptiCore can be used even when the main crypto library is set in the FIPS mode. To enforce that only FIPS-compliant algorithms are used, disable the non-FIPS algorithms from the configuration. See cipher and mac.

For a list of platforms on which the FIPS library has been validated or tested, see SSH Tectia Client/Server Product Description.

settings

This element contains miscellaneous settings. It has the following attributes: proxy-scheme, xauth-path, ignore-aix-rlogin, record-ptyless-sessions and user-config-dir.

The proxy-scheme attribute defines rules for HTTP or SOCKS proxy servers that SSH Tectia Server uses when a client forwards a connection (local tunnel).

The format of the attribute value is a sequence of rules delimited by semicolons (;). Each rule has a format that resembles the URL format. In a rule, the connection type is given first. The type can be direct, socks, socks4, socks5, or http-connect (socks is a synonym for socks4). This is followed by the server address and port. If the port is not given, the default ports 1080 for SOCKS and 80 for HTTP are used.

After the address, zero or more conditions delimited by commas (,) are given. The conditions can specify IP addresses or DNS names.

direct:///[cond[,cond]...]
socks://server/[cond[,cond]...]
socks4://server/[cond[,cond]...]
socks5://server/[cond[,cond]...]
http-connect://server/[cond[,cond]...]

The IP address/port conditions have an address pattern and an optional port range:

ip_pattern[:port_range]

The ip_pattern may have one of the following forms:

a single IP address x.x.x.x
an IP address range of the form x.x.x.x-y.y.y.y
an IP sub-network mask of the form x.x.x.x/y

The DNS name conditions consist of a hostname which may be a regular expression containing the characters "*" and "?" and a port range:

name_pattern[:port_range]

An example proxy-scheme is shown below. It causes the server to access the callback address and the ssh.com domain directly, access *.example with HTTP CONNECT, and all other destinations with SOCKS4.

"direct:///127.0.0.0/8,*.ssh.com;
 http-connect://http-proxy.ssh.com:8080/*.example;
 socks://fw.ssh.com:1080/"

The xauth-path attribute contains a path to a supplementary XAuth binary used with X11 forwarding on Unix platforms.

The ignore-aix-rlogin attribute defines whether the server should ignore the remote login restriction on AIX. Local login permission is still honored. The value must be yes or no. The default is no.

The record-ptyless-sessions attribute can be used to control whether sessions without PTYs are recorded as user logins in the operating system. Sessions without PTYs are for example remote commands and SFTP sessions. By default, all sessions are recorded. However, some system utilities (such as finger on Solaris) do not allow sessions without PTYs to be recorded because these sessions do not have a valid TTY name. On these systems, only real shell logins should be recorded and others turned off by setting record-ptyless-sessions=no. The value must be yes or no. The default is yes.

The user-config-dir attribute can be used to specify a directory where user-specific public-key configuration data is found. With this, the administrator can control those options that are usually controlled by the user. The directory path can include a pattern string which is expanded by SSH Tectia Server.

The following pattern strings can be used:

%D or %homedir% is the user's home directory
%U or %username% is the user's login name
For Windows domain users, these strings are substituted differently:
- %U is expanded to domain.username
- %username% is expanded to domain\username
%IU or %userid% is the user's user ID (uid)
%IG or %groupid% is the user's group ID (gid)

Note that user ID and group ID are only supported on Unix, not on Windows.

The default is "%D/.ssh2".

Example of the settings element and its attributes:

<settings 
 proxy-scheme="direct:///10.0.0.0/8,localhost;socks5://fw.example.com:1080/" 
 xauth-path="/usr/X11R6/bin/xauth" 
 ignore-aix-rlogin="no" 
 record-ptyless-sessions="yes"
 user-config-dir="%D/.ssh2"/>

hostkey

The hostkey element defines the location of the private host key and optionally the location of the public key and/or certificate. The elements inside the element must be given in the right order (private key before public).

Inside one hostkey element either the public key or the certificate can be given, not both.

Giving the public key in the configuration file is not mandatory. It will be derived from the private key if it is not found otherwise. However, specifying the public key will decrease the start-up time for the software, as deriving the public key is a fairly slow operation.

private

This element gives the path to the private key file as a value of the file attribute.

The key file should be located on a local drive. Network or mapped drives should not be used, as the server program may not have proper access rights for them. The default is hostkey, in the /etc/ssh2 directory on Unix and in the "C:\Program Files\SSH Communications Security\SSH Tectia\SSH Tectia Server" directory on Windows.

On Unix, the private key file should be readable and writable only by root. The private key directory should be writable only by root

On Windows, the key file and directory should have full permissions for the Administrators group and the SYSTEM account and no other permissions.

public

This element gives the path to the public key file as a value of the file attribute.

The key file should be located on a local drive. Network or mapped drives should not be used, as the server program may not have proper access rights for them.

Alternatively, the public key can be specified as a base64-encoded ASCII element.

x509-certificate

This element gives the path to the X.509 user certificate file as a value of the file attribute.

Alternatively, the certificate can be specified as a base64-encoded ASCII element.

externalkey

This element defines an external host key. The type must be given as an attribute. The currently supported types are none, software, entrust, mscapi, pkcs11, and pkcs12. The init-info for the external key can also be given.

Sample hostkey elements are shown below:

<hostkey>      
  <private file="/etc/ssh2/hostkey_dsa" />
  <public file="/etc/ssh2/hostkey_dsa.pub" />
</hostkey>

<hostkey>      
  <private file="/etc/ssh2/hostcert_rsa" />
  <x509-certificate file="/etc/ssh2/hostcert_rsa.crt" />
</hostkey>

<hostkey>      
  <externalkey type="entrust" 
               init-info="profile-file(/etc/ssh2/hostcertprofile.epf)" />
</hostkey>

For PKCS#12, the <hostkey> settings are as follows:

<hostkey>      
  <externalkey type="software" 
               init-info="key_file(/etc/ssh2/server-cert.p12)
                          key_passphrase_file(/etc/ssh2/my-passphrase)" />
</hostkey>

In the PKCS#12 sample output, the hostkey setting reads the PKCS#12 file server-cert.p12 and if it needs a passphrase to open it, it will read the my-passphrase file and use the contents as the passphrase. The file can also contain additional certificates but they are ignored in SSH Tectia Server.

In the init-info string, the following keywords are supported:

directory(<directory_name>) - defines the directory to be polled for the keys. All files in the named directory are added to sshexternalkey. Note however, that this option lacks control over the actual server key and certificate.
polling_interval_ms(<time_ms>) - defines the polling interval for the option above.
key_files(<key_spec>) - defines that multiple comma-separated files are read. Loose grouping between files is expected so that public key, private key and certificate are assumed to be parts of the same key. Supported in SSH Tectia Server.
key_file(<file_name>) - defines that one key file is read. The same as key_files with one parameter.
key_passphrase(<passphrase>) - if a private key or certificate container is password-protected, the command tries to open it with the supplied passphrase first. In case the passphrase is not valid, the authentication callback is called normally. In the server, that means a failure to open the file as the server does not have an interactive prompt.
key_passphrase_file(<filename>) - defines that instead of giving the passphrase in the configuration file directly, it can be written to a separate file. This option is useful if server configuration file needs to be more widely readable. The private key and passphrase can still be with root access only.

listener

This element is used to specify where the Secure Shell server listens for connections. The element has three attributes: id, address, and port.

The id must be given as an attribute. The value must be unique. The value must begin with a letter, it can contain alphanumeric characters or underscore characters but no any whitespaces. Also the port and network interface address can be given. The default port for listeners is 22.

Several listeners can be created to the same IP address to different ports. Each must have an unique ID. If the address is not specified, the server will listen to the given port on all interfaces.

Sample listener elements are shown below:

<listener id="internet" address="192.0.2.62" />
<listener id="intranet" address="10.0.0.1" />
<listener id="admin-private" port="222" />

logging

This element changes the logging settings that define the log event severities and logging facilities. The element contains one or more log-events elements.

log-events

This element sets the severity and facility of different logging events. The events have reasonable default values, which are used if no explicit logging settings are made. This setting allows customizing the default values.

For the events, facility and severity can be set as attributes. The events itself should be listed inside the log-events element.

The facility can be normal, daemon, user, auth, local0, local1, local2, local3, local4, local5, local6, local7, or discard. Setting the facility to discard causes the server to ignore the specified log events.

On Windows, only the normal and discard facilities are used.

The severity can be informational, notice, warning, error, critical, security-success, or security-failure.

Any events that are not specifically defined in the configuration file use the default values. The defaults can be overridden for all remaining events by giving an empty log-events element after all other definitions and setting a severity value for it.

For a complete list of log events, see Appendix D.

The following example sets the facility of the Auth_method_failure event to auth and the severity to notice. It also sets the facility of the Server_reconfig_started and Server_starting events to discard (the events will not be logged). All other events use the default settings.

<logging>
  <log-events facility="auth" severity="notice">
    Auth_method_failure
  </log-events>
  <log-events facility="discard">
    Server_reconfig_started Server_starting
  </log-events>
</logging>

limits

This element sets the maximum number of connections and processes the server will handle. SSH Tectia Server uses a distributed architecture where the master server process launches several servant server processes that handle the actual connections.

The max-processes attribute defines the maximum number of servant processes the master server will launch. The value range is 1-40. The default value is 40.

The max-connections attribute defines the maximum number of client connections allowed per servant. The value range is 1-256. The default value is 256.

This setting is useful in systems with low resources. The server has to be restarted to use the changed setting.

A sample limits element is shown below:

<limits max-connections="256" max-processes="40" />

cert-validation

This element contains the CA certificates used in validation of the host-based and public-key authentication certificates. The element can have two attributes: http-proxy-url and socks-server- url.

The http-proxy-url attribute defines a HTTP proxy and the socks-server-url attribute defines a SOCKS proxy for making LDAP or OCSP queries for certificate validity.

The address of the proxy is given as the value of the attribute. The format of the address is socks://username@socks_server:port/network/netmask,network/netmask ... (with a SOCKS proxy) or http://username@proxy_server:port/network/netmask,network/netmask ... (with an HTTP proxy).

For example, by setting socks-server-url to "socks://mylogin@socks.ssh.com:1080/192.168.0.0/16,10.100.23.0/24", the host socks.ssh.com and port 1080 are used as your SOCKS server for connections outside of networks 192.168.0.0 (16-bit domain) and 10.100.23.0 (8-bit domain). Those networks are connected directly.

The cert-validation element can contain multiple ldap-server and ocsp-responder elements, a cert-cache-file and crl-auto-update element, multiple crl-prefetch elements, a dod-pki element, and multiple ca-certificate elements.

The validity of a received certificate is checked separately using each of the defined ca-certificate elements in turn until they are exhausted (in which case the authentication fails), or a positive result is achieved. If the certificate is valid, the connections and authentication-methods elements determine whether the certificate allows the user to log in (of course, the correct signature generated by a matching private key is always required in addition to everything else).

ldap-server

This element specifies an LDAP server address and port used for fetching CRLs and/or subordinate CA certificates based on the issuer name of the certificate being validated. Several LDAP servers can be specified by using several ldap-server elements.

CRLs are automatically retrieved from the CRL distribution point defined in the certificate to be verified if the point exists.

The default value for port is 389.

ocsp-responder

This element specifies an OCSP (Online Certificate Status Protocol) responder service address in URL format (url). Several OCSP responders can be specified by using several ocsp-responder elements.

For the OCSP validation to succeed, both the end-entity certificate and the OCSP responder certificate must be issued by the same CA. If the certificate has an Authority Info Access extension with an OCSP Responder URL, it is only used if there are no configured OCSP responders. It is not used if any OCSP responders have been configured.

The validity-period in seconds for the OCSP data can be optionally defined. During this time, new OCSP queries for the same certificate are not made but the old result is used.

If an OCSP responder is defined in the configuration file or in the certificate, it is tried first; only if it fails, traditional CRL checking is tried, and if that fails, the certificate validation returns a failure.

cert-cache-file

This element specifies the name of the file where the certificates and CRLs are stored when the SSH Tectia Server service is stopped, and read back in when the service is restarted.

On Unix, the cache file should be writable only by root.

On Windows, the cache file should be writable only by the Administrators group and the SYSTEM account.

crl-auto-update

This element turns on automatic updating of certificate revocation lists. When it is on, SSH Tectia Server periodically tries to download the new CRL before the old one has expired. The update-before attribute can be used to specify how many seconds before the expiration the update takes place. The minimum-interval sets a limit for the maximum update frequency. The default minimum interval is 30 seconds.

crl-prefetch

This element instructs SSH Tectia Server to periodically download a CRL from the specified url. The url can be an LDAP or HTTP URL, or it can refer to a local file. The file format must be either binary DER or base64, PEM is not supported.

To download CRLs from the local file system, define the file URL in this format:

file:///absolute/path/name

To download CRLs from an LDAP, define the LDAP URL in this format:

ldap://ldap.server.com:389/CN=Root%20CA,OU=certification%20authorities,
DC=company,DC=com?certificaterevocationlist

Use the interval attribute to specify how often the CRL is downloaded. The default is 3600 (seconds).

dod-pki

This element defines whether the certificates are required to be compliant with the DoD PKI (US Department of Defense Public-Key Infrastructure). In practice, this means that the Digital Signature bit must be set in the Key Usage of the certificate. The enable attribute can have a value of yes or no. The default is no.

ca-certificate

This element enables user authentication using certificates. It can have four attributes: name, file, disable-crls, and use-expired-crls.

The name attribute must contain the name of the CA.

The element must either contain the path to the X.509 CA certificate file as a value of the file attribute, or include the certificate as a base64-encoded ASCII element.

CRL checking can be disabled by setting the disable-crls attribute to yes. The default is no.

	Note
	CRL usage should only be disabled for testing purposes. Otherwise it is highly recommended to always use CRLs.

Expired CRLs can be used by setting a numeric value (in seconds) for the use-expired-crls attribute. The default is 0 (do not use expired CRLs).

A sample cert-validation element is shown below:

<cert-validation http-proxy-url="http://proxy.example.com:800">
  <ldap-server address="ldap.example.com" port="389" />
  <ocsp-responder validity-period="60" url="http://ca.example.com/ocsp-1/" />
  <cert-cache-file file="/var/cert-cache.dat" />
  <crl-auto-update update-before="30" minimum-interval="600" />
  <crl-prefetch interval="1800" url="http://ca.example.com/default.crl" />
  <dod-pki enable="no" />
  <ca-certificate name="exa-ca1" file="/etc/ssh2/exa-ca1.crt" />
  <ca-certificate name="exa-ca2" file="/etc/ssh2/exa-ca2.crt"
     use-expired-crls="3600" />
  <ca-certificate name="testonly-ca" file="/etc/ssh2/testonly-ca.crt"
     disable-crls="yes" />
</cert-validation>

The `connections` Block

The connections block defines the basic rules for allowing and denying connections. The connections block includes one or more connection elements.

If a user does not match to any selectors in the connection elements, the server default connection settings are used.

connection

Each connection element specifies either an allow or deny rule for connections. The element can have three attributes: name, action, and tcp-keepalive.

The word allow or deny is given as a value of the action attribute. By default, if the action attribute is omitted, the connection is allowed.

The name attribute can be used to give an identifier to the connection rule. The value must be a valid XML name beginning with a letter and containing alphanumeric characters or the underscore character without any whitespace. The identifier can be used, for example, in auditing.

The tcp-keepalive attribute defines whether the system should send keepalive messages to the other side. If they are sent, a broken connection or crash of one of the machines will be properly noticed. However, this means that connections will die if the route is down temporarily, and this can be annoying in some situations. On the other hand, if keepalive messages are not sent, sessions may hang indefinitely on the server, leaving "ghost" users and consuming server resources. The value must be yes or no. The default is no (do not send keepalives).

The connection element can include one or more selectors, a rekey setting, and cipher and MAC definitions.

selector

The selectors define to which connections this connection rule applies to. Only the interface and ip selector attributes can be used in the connections block. Other information, for example the user name, is not yet available at this stage of the connection. See the section called “Selectors”.

interface

This selector matches to the listener interface id or address and/or port. At least one attribute must be given. If the id is defined, the others MUST NOT be given. If the id is not defined, either or both of address and port may be given.

ip

This selector matches to an IP address or fqdn (fully qualified domain name) of the client. Either address or fqdn can be given, not both.

The address can be in one of the following formats:

a single IP address x.x.x.x
an IP address range of the form x.x.x.x-y.y.y.y
an IP sub-network mask of the form x.x.x.x/y

The fqdn attribute matches to an FQDN pattern (case-insensitive). The attribute can include a comma-separated list of allowed FQDN patterns. These patterns may also contain "*" and "?" globbing characters.

rekey

This element specifies the number of seconds or transferred bytes after which the key exchange is done again.

If a value for both seconds and bytes is specified, rekeying is done whenever one of the values is reached, after which the counters are reset.

The defaults are 3600 seconds (1 hour) and 1000000000 bytes (~1 GB). The value 0 (zero) turns rekey requests off. This does not prevent the client from requesting rekeys.

cipher

This element selects a cipher name allowed by the server for data encryption. The supported ciphers are 3des-cbc, aes128-cbc, aes192-cbc, aes256-cbc, arcfour, blowfish-cbc, twofish-cbc, twofish128-cbc, twofish192-cbc, twofish256-cbc, crypticore128@ssh.com, seed-cbc@ssh.com, and none (no encryption).

Multiple ciphers can be specified by using multiple cipher elements. By default, the server allows 3des-cbc, aes128-cbc, aes192-cbc, aes256-cbc, seed-cbc@ssh.com, and crypticore128@ssh.com (on Windows and Linux x86).

The ciphers that can operate in the FIPS mode are 3des-cbc, aes128-cbc, aes192-cbc, and aes256-cbc.

Normally when a specified cipher is not found on the server, the configuration file reading fails and the server will not restart. The cipher element may optionally take an allow-missing attribute, which can have a value of yes or no. If a value of yes is given for this attribute and a specified cipher is not found during configuration reading (for example, CryptiCore on Solaris), the server logs a warning to the syslog but will restart normally. The default is no (a missing cipher is treated as fatal error and the server configuration reading fails).

Setting the allow-missing attribute to yes is useful when you want to use the same ssh-server-config.xml file on multiple servers and only some of the servers have, for example, CryptiCore available.

mac

This element selects a MAC name allowed by the server for data integrity verification. The supported MAC algorithms are hmac-md5, hmac-md5-96, hmac-sha1, hmac-sha1-96, crypticore-mac@ssh.com, and none (no data integrity verification).

The hmac-sha1 algorithm can operate in the FIPS mode.

Multiple MACs can be specified by using multiple mac elements. By default, the server allows hmac-md5, hmac-sha1, and crypticore-mac@ssh.com (on Windows and Linux x86).

Normally when a specified MAC is not found on the server, the configuration file reading fails and the server will not restart. The mac element may optionally take an allow-missing attribute, which can have a value of yes or no. If a value of yes is given for this attribute and a specified MAC is not found during configuration reading (for example, CryptiCore on Solaris), the server logs a warning to the syslog but will restart normally. The default is no (a missing MAC is treated as fatal error and the server configuration reading fails).

A sample connection element that allows connections from a specified IP address range is shown below:

<connection name="conn1" action="allow" tcp-keepalive="yes">
  <selector>
    <ip address="192.168.0.42-192.168.0.82" />
  </selector>
  <rekey seconds="600" bytes="500000000" />
  <cipher name="crypticore128@ssh.com" />
  <mac name="crypticore-mac@ssh.com" />
</connection>

A sample connection element that denies all connections is shown below. As the element does not contain any selectors, it matches always. This can be used as the last element in the connections element to deny all connections that were not explicitly allowed by the previous elements. (By default, non-matching connections would be allowed.)

<connection action="deny" />

The `authentication-methods` Block

The authentication-methods block defines the authentication methods that are allowed and required by the server. It can have one attribute: login-grace-time. It can contain a banner-message and an auth-file-modes element and multiple authentication elements.

The login-grace-time attribute is used to specify a time after which the server disconnects if the user has not successfully logged in. If the value is set to 0, there is no time limit. The default is 600 (seconds).

The authentication methods that are on the same level under one authentication element are considered allowed (one of them must succeed).

Several authentication methods can be set as required by nesting the authentication elements inside each other.

The server allows by default public-key, keyboard-interactive, and password authentication (one of them must succeed).

If the authentication-methods element is empty or missing from the ssh-server-config.xml file, the server does not allow any authentication methods and the configuration is essentially defunctional.

However, if you put inside authentication-methods an authentication element with no authentication methods defined, the matching users will be allowed to log in without authentication. This can be used in combination with selectors and/or nested authentication methods, but should never be used as the only authentication element.

	Caution
	Consider carefully before putting an empty `authentication` element in the `ssh-server-config.xml` file. It will allow the matching users (everyone, if no selectors are used) to log in without authentication.

banner-message

This element specifies the path to the message that is sent to the client before authentication. The path is given as a value of the file attribute. Alternatively, the banner message can be given as the contents of the banner-message element.

Note, however, that the client is not obliged to show this message.

<banner-message file="/etc/ssh2/banner-message">
  This is the server banner message. If file attribute is set, this
  inlined text is ignored, and the file is read instead
  (like in this example).
</banner-message>

auth-file-modes

This element specifies whether SSH Tectia Server on Unix platforms should check permissions and ownership of the user's key files used for public-key authentication.

The word yes or no is given as a value of the strict attribute. If set to yes, the permissions and ownership of the .ssh2 directory, the .ssh2/authorization file (if used), the .ssh2/authorized_keys directory (if used), and the keys listed in the authorization file or present in the authorized_keys directory are checked.

This is normally desirable because novices sometimes accidentally leave their directory or files world-writable, in which case anyone can edit the authorization and key files. The default is yes.

The mask-bits attribute can be used to specify the forbidden permission bits in octal format. The default is 022 (group and others must not have the write permission).

The ownership of the checked files and directories must be either root or the user.

<auth-file-modes strict="yes" mask-bits="022" />

authentication

Each authentication element specifies a chain of authentication methods. It can include one or more selectors and different authentication methods. It may also include other authentication elements.

Nesting authentication elements within each other sets the child elements as required (all must be passed for the authentication to be successful). Setting multiple authentication methods at the same level sets them as optional (one of the methods must be passed for the authentication to be successful).

The authentication elements are read in order. For elements on the same level, the first matching element is used and the remaining elements are ignored. If the element has nested child elements, they are matched next using the same procedure.

In the authentication element, the action attribute takes values allow or deny. By default, if the action attribute is omitted, authentication is allowed.

If an authentication chain ends in a deny action, or if the user does not match to any selectors in the authentication elements, the user is not allowed to log in.

	Note
	Note that the behavior has changed in SSH Tectia Server 5.1. In SSH Tectia Server 5.0, a non-matching user was allowed the default authentication methods.

In a nested chain of authentication elements, it is possible, for example, to set the parent method to deny authentication and a child element with a selector to allow authentication. If the user matches the selector and successfully completes the authentication method(s), login is allowed.

For more information on using authentication chains, see Configuring User Authentication Chains.

The authentication element can take a repeat-block attribute. This can be used when the authentication element contains a nested element with certificate selectors.

Set the value of repeat-block to yes when you expect the user to have several certificates of which only some allow login (that is, match the selectors in the child authentication element). In this case, SSH Tectia Server will try all offered certificates until a match is found.

If repeat-block is set to no (the default), SSH Tectia Server will try to match only the first certificate offered by the client.

The authentication element can additionally take a set-group attribute, which sets a group for the users that pass the particular authentication chain. The group definition can be later used in the services element.

If set-group is used here, it overrides any group definitions in the services element. See the section called “The services Block”.

The authentication name can be optionally given as an attribute. The value of name must be a valid XML name beginning with a letter and containing alphanumeric characters or the underscore character without any whitespace. The authentication name can be used, for example, in auditing.

	Caution
	Consider carefully before putting an empty `authentication` element in the `ssh-server-config.xml` file. It will allow the matching users (everyone, if no selectors are used) to log in without authentication.

selector

The selectors define to which connections this authentication method applies to. All selectors can be used in the authentication-methods block. See the section called “Selectors”.

certificate

This selector matches to a pattern in a specified field of the user certificate used in public-key authentication. Using this selector requires that the parent element in the authentication chain contains an auth-publickey element.

The field can be either ca-list, issuer-name, subject-name, serial-number, altname-email, altname-upn, altname-ip, or altname-fqdn.

The format of the pattern depends on the type of the field. The ca-list field contains a list of CA names separated by commas. The names that are defined in the ca-certificate element are used. The issuer-name and subject-name fields contain distinguished names, serial-number a positive integer. The altname-fqdn field contains a hostname and altname-ip an IP address or a range. The altname-email field contains an email address and altname-upn the principal name.

The altname-fqdn, altname-upn, altname-email, subject-name, and issuer-name selectors may contain the %username% keyword which is replaced with the user's login name before comparing with the actual certificate data. On Windows, the %username-without-domain% keyword can be used and it is replaced by the user's login name without the domain part. The %hostname% keyword can be used in the same way and it is replaced by the client's FQDN. These patterns may also contain "*" and "?" globbing characters.

Patterns are normally matched case-insensitively. Alternatively, the pattern can be specified using the pattern-case-sensitive attribute.

For the subject-name selector, you can also define if the pattern has to match the subject name completely or only partly. Use the ignore-prefix="yes" attribute to match only the end of subject name, and the ignore-suffix="yes" attribute to match only the beginning of the subject name. The ignore options are optional.

You can also define both of the ignore options on simultaneously in which case the pattern has to match with some point in the subject name. For example: when both ignore settings are defined on, pattern O=SSH,OU=*,CN=example matches with:

C=FI, O=SSH, OU=RandD, CN=example, CN=UID12345

The allow-undefined attribute can be used to control the behavior of the selector when the required certificate data is not defined (certificates have not been used at all, or the certificate does not contain the fields to be matched). Its value must be yes or no. The default is no. For more information, see the section called “Selectors and Undefined Data”.

Caution

When creating the certificate selectors, make sure that every selector element ties the user name to the certificate, either by including a user sub-element, or by putting the special substitution string %username% or %username-without-domain% to a field used to match the correcponding field in the certificate.

Failing to do this may cause unintended consequences, for example authentication succeeding with many different user names with a single certificate.

host-certificate

This selector matches to a pattern in a specified field of the client host certificate used in host-based authentication. Using this selector requires that the parent element in the authentication chain contains an auth-hostbased element.

The field can be either ca-list, issuer-name, subject-name, serial-number, altname-email, altname-upn, altname-ip, or altname-fqdn.

Patterns are normally matched case-insensitively. Alternatively, the pattern can be specified using the pattern-case-sensitive attribute.

For the subject-name selector, you can also define if the pattern has to match the subject name completely or only partly. Use the ignore-prefix="yes" attribute to match only the end of the subject name, and the ignore-suffix="yes" attribute to match only the beginning of the subject name. Both attributes can be used together in which case the pattern has to match with some point in the subject name. The ignore options are optional.

interface

ip

This selector matches to an IP address or fqdn (fully qualified domain name) of the client. Either address or fqdn can be given, not both.

The address can be in one of the following formats:

a single IP address x.x.x.x
an IP address range of the form x.x.x.x-y.y.y.y
an IP sub-network mask of the form x.x.x.x/y

The fqdn attribute matches to an FQDN pattern (case-insensitive). The attribute can include a comma-separated list of allowed FQDN patterns.

user

This selector matches to a user name or id. A list of user names or IDs can be given as a comma- separated list.

Names are normally matched case-insensitively. Alternatively, the names can be specified using the name-case-sensitive attribute.

On Windows domain environment, the user and user-group selectors have a length limitation. For more information, see the description of option User in Configuration Tool (Windows).

user-group

This selector matches to a user group name or id. A list of user-group names or IDs can be given as a comma-separated list.

Names are normally matched case-insensitively. Alternatively, the names can be specified using the name-case-sensitive attribute.

On Windows domain environment, the user and user-group selectors have a length limitation. For more information, see the description of option User in Configuration Tool (Windows).

user-privileged

This selector matches to a privileged user (administrator or root) or to a non-privileged user. The value can be yes (match to a privileged user) or no (match to a normal user).

The allow-undefined attribute can be used to control the behavior of the selector when the required data is not defined (user-privilege level is not known). Its value must be yes or no. The default is no. For more information, see the section called “Selectors and Undefined Data”.

Note

On a Windows server, the user-privilege level is not available during the authentication phase when the user is logging in using a domain account and does not yet have an access token allocated. To get the user-privilege status for domain users, the user should first pass password or GSSAPI authentication.

If the privilege level needs to be checked for local accounts, the allow-undefined attribute should be set to yes or else connection fails for users logging in using domain accounts. However, this means that the user-privilege status will not be verified for Windows domain users.

To check the privilege level of domain accounts on a Windows server in the authentication-methods block, the user-privileged selector should be used in a nested authentication block when password or GSSAPI authentication has already been passed.

publickey-passed

This selector matches if authentication is passed using a normal public key (without a certificate). Using this selector requires that the authentication chain contains an auth-publickey element.

Optionally, the length range of the public key can be given as an attribute, for example "1024-2048" (keys from 1024 to 2048 bits match). The range can also be left open, for example "1536-" (keys over 1536 bits match).

The allow-undefined attribute can be used to control the behavior of the selector when the required data is not defined (public-key authentication has not been used). Its value must be yes or no. The default is no. For more information, see the section called “Selectors and Undefined Data”.

user-password-change-needed

On Unix platforms, this selector matches if the user password has expired and should be changed. The actual password change can be done by defining a rule with a forced passwd command. See sample rules in the section called “The services Block” and the section called “Forcing Password Change”.

The allow-undefined attribute can be used to control the behavior of the selector when the required data is not defined. Its value must be yes or no. The default is no. For more information, see the section called “Selectors and Undefined Data”.

auth-publickey

This element sets the public-key authentication method. The element can have three attributes: authorization-file, authorized-keys-directory, and openssh-authorized-keys-file.

The authorization-file attribute can be used to specify a path to the file that lists the user public keys that are authorized for login. The path can contain a pattern string that is expanded by SSH Tectia Server.

The following pattern strings can be used:

%D or %homedir% is the user's home directory
%U or %username% is the user's login name
For Windows domain users, these strings are substituted differently:
- %U is expanded to domain.username
- %username% is expanded to domain\username
%IU or %userid% is the user's user ID (uid)
%IG or %groupid% is the user's group ID (gid)

Note that user ID and group ID are only supported on Unix, not on Windows.

The default is %D/.ssh2/authorization.

The authorized-keys-directory attribute can be used to specify a path to the directory that contains the user public keys that are authorized for login. As above, the path can contain a pattern string that is expanded by SSH Tectia Server. The default is %D/.ssh2/authorized_keys.

The openssh-authorized-keys-file attribute can be used to specify a path to an OpenSSH-style authorized_keys file that contains the user public keys that are authorized for login. As above, the path can contain a pattern string that is expanded by SSH Tectia Server.

	Note
	These attributes override the `user-config-dir` attribute in the `settings` element.

For more information, see User Authentication with Public Keys.

auth-hostbased

This element sets the host-based authentication method. The element can have one attribute: require-dns-match.

If the require-dns-match attribute is set to yes, host-based authentication will require the host name given by the client to match the one found in DNS. If the hostname does not match, the authentication will fail. The default is no (exact match not required).

auth-password

This element sets the password authentication method.

The delay between failed attempts in seconds (failure-delay) and the maximum number of attempts (max-tries) can be given as attributes. The default delay is 2 seconds and default maximum is 3 attempts.

auth-keyboard-interactive

This element sets the keyboard-interactive authentication method.

The delay between failed attempts in seconds (failure-delay) and the maximum number attempts (max-tries) can be given as attributes. The default delay is 2 seconds and default maximum is 3 attempts.

The keyboard-interactive submethods are given as child elements. The supported methods are submethod-password, submethod-pam, submethod-securid, and submethod-radius, and submethod-generic.

If no submethods are configured, all available submethods are allowed by default (however, the server may not be able find the necessary libraries for SecurID and PAM, for example). If some of the submethods are configured, the rest of the submethods are implicitly disabled.

submethod-pam

This element sets the keyboard-interactive PAM submethod in use. PAM is supported on Unix platforms.

The dll-path to the PAM DLL can be given in an attribute. The path must point to the operating-system specific PAM module, for example, "/lib/libpam.so.1" on Red Hat Linux, or "/lib/libpam.a(shr.o)" on AIX.

On AIX, the path should include the archive file, unless the library is a shared object or has been extracted from the shared object.

submethod-password

This element sets the keyboard-interactive password submethod in use.

submethod-securid

This element sets the keyboard-interactive SecurID submethod in use.

The dll-path to the SecurID DLL can be given in an attribute. The path must point to the operating-system specific SecurID module, for example, "/usr/lib/libaceclnt.so" on Solaris.

submethod-radius

This element sets the keyboard-interactive RADIUS submethod in use.

The element can contain multiple radius-server child elements.

radius-server

This element defines a RADIUS server. The element has four attributes: address, port, timeout, and client-nas-identifier.

The address is the IP address of the RADIUS server. The port is the RADIUS server port. The default port is 1812.

The timeout is the time in seconds after which the RADIUS query is terminated if no response is gained. The default is 10 seconds.

The client-nas-identifier attribute defines the network access server (NAS) identifier to be used when talking to the RADIUS server.

The element must contain one radius-shared-secret child element.

radius-shared-secret

This element defines the RADIUS shared secret file.

The path to the secret file can be given as a value of the file attribute.

Alternatively, the secret can be included as the contents of the radius-shared-secret element.

submethod-generic

This element sets the generic submethod in use. This element can be used to add custom submethods to keyboard-interactive authentication.

The name on the method must be given in the attribute.

Optional params for the submethod can be given as well.

auth-gssapi

This element sets the GSSAPI authentication method.

The dll-path can be given as an attribute. This specifies where the necessary GSSAPI libraries are located. If this attribute is not specified, the libraries are searched for in a number of common locations. The full path to the libraries should be given, for example, "/usr/lib/libkrb5.so,/usr/lib/libgssapi_krb5.so".

On AIX, the dll-path should include the archive file, if applicable, for example, "<path>/libgssapi_krb5.a(libgssapi_krb5.a.so)". The archive(shared_object) syntax is not necessary if the library is a shared object or has been extracted from the shared object.

On Windows, the dll-path attribute is ignored. SSH Tectia Server locates the correct DLL automatically.

The allow-ticket-forwarding attribute defines whether the server allows forwarding the Kerberos ticket over several connections. The attribute can have a value of yes or no. The default is no.

Normally if a specified authentication method is not found on the server, the configuration file reading fails and the server will not restart. The auth-gssapi element may optionally take an allow-missing attribute, which can have a value of yes or no. If a value of yes is given for this attribute and GSSAPI plugin is not found during configuration reading, the server logs a warning to the syslog but will restart normally. The default is no (if GSSAPI is specified but not found, it is treated as fatal error and the server configuration reading fails).

Setting the allow-missing attribute to yes is useful when you want to use the same ssh-server-config.xml file on multiple servers and only some of the servers have Kerberos/GSSAPI available.

authentication

The authentication elements can be nested within each other. The method(s) in the child element(s) must be passed in addition to the method in the parent element.

Authentication Examples

A sample authentication-methods element is shown below:

<authentication-methods>
  <authentication>
    <selector>
      <user-group name="staff" />
    </selector>
    <auth-publickey authorized-keys-directory="%IG/ssh2_authorized_keys" />
    <auth-password />
  </authentication>
</authentication-methods>

In this simple authentication example, the users who belong to group "staff" are allowed to log in using either public-key or password authentication. The user public keys are checked from an alternate location. As there are no other authentication blocks, all users that do not match to the selector are implicitly denied authentication.

	Note
	Note that the behavior has changed in SSH Tectia Server 5.1. In SSH Tectia Server 5.0, a non-matching user was allowed the default authentication methods.

Another sample authentication-methods element is shown below:

<authentication-methods>
  <authentication action="deny">
    <auth-publickey />
    <authentication action="allow" set-group="local-user">
      <selector>
        <ip address="10.1.55.14-10.1.55.99" />
        <user name="johnd" />
        <user name="janed" />
      </selector>
      <auth-keyboard-interactive max-tries="4"/>
        <submethod-radius>
          <radius-server address="10.1.61.128">
            <radius-shared-secret file="&configdir;/radius-secret-file"/>
          </radius-server>
        </submethod-radius>
      </auth-keyboard-interactive>
    </authentication>
    <authentication action="allow" set-group="finance-inspector">
      <selector>
        <user-group name="finance" />
      </selector>
      <auth-password />
    </authentication>
  </authentication>
</authentication-methods>

In the example above, all users are first required to authenticate using public-key authentication. Based on selector matching, also a second method needs to be passed (RADIUS via keyboard-interactive or password). A group is set based on the matched and passed authentication methods. If a user does not match to either of the child authentication elements, access is denied (the parent authentication element has the action set to deny).

If the action of the parent authentication element would be allow, the non-matching users would be let in after having passed public-key authentication.

See Configuring User Authentication Chains for more examples of configuring authentication chains.

A sample authentication-methods element that sets requirements for certificate authentication is shown below:

<authentication-methods>
  <authentication action="allow" repeat-block="yes">
    <auth-publickey />
    <authentication action="allow" set-group="admin">
      <selector>
        <user-privileged value="yes" allow-undefined="yes" />
        <certificate field="ca-list" pattern="exa-ca1,exa-ca2" 
                     allow-undefined="yes" />
        <certificate field="subject-name" pattern="C=FI, O=SSH, OU=*, CN=%username%" 
                     ignore-suffix="yes" allow-undefined="yes" />
      </selector>
    </authentication>
    <authentication action="allow">
      <selector>
        <publickey-passed length="1024-2048" />
      </selector>
    </authentication>
    <authentication action="deny" />
  </authentication>
</authentication-methods>

In the example above, privileged users (administrators) are required to pass certificate authentication and the certificate must contain the correct fields. Other users are allowed to log in using a plain public key of a size from 1024 to 2048 bits.

In this example, the allow-undefined attribute has to be used in the selectors of the first nested authentication block. Otherwise, the authentication will end in error for users with plain public keys. When the user uses a plain public key, the server will not have the certificate fields to be matched defined. For more information, see the section called “Selectors and Undefined Data”.

	Note
	Specifying an explicit `deny` action last is necessary in a restrictive policy, as otherwise a non-matching connection would use the `allow` action of the parent element (if it passed public-key authentication with any key length). A better way to achieve the same result is to set the action of the parent authentication element to `deny` (as in the previous example).

The `services` Block

The services block defines the policy for the various services the server offers.

The services block contains one or more rules (rule) and optionally defines groups (group).

group

Creates a group that can be used as a basis for restricting services. Groups are defined based on selectors.

The name must be given as an attribute. The value of name must be a valid XML name beginning with a letter and containing alphanumeric characters or the underscore character without any whitespace.

If the user was already put to a group during authentication using the set-group attribute, the group definitions in the services element are ignored.

selector: The selectors define the users that belong to the group. The same selectors can be used as in the authentication-methods block. See the section called “Selectors” and the section called “The authentication-methods Block”.

Sample group elements are shown below:

<!-- Remote access. -->
<group name="remote-access">
  <selector>
    <interface address="192.0.2.62" />
  </selector>
</group>

<!-- Backup. -->
<group name="backup">
  <selector>
    <user name="backup" />
  </selector>
</group>

<!-- Password change needed. -->
<group name="passwd-change">
  <selector>
    <user-password-change-needed />
  </selector>
</group>

rule

This element defines a rule for the specified group of users. Rules can be used to restrict the services and commands the server allows to the users. The element can have three attributes: group, idle-timeout, and print-motd.

The rules are read in order, and the first rule that matches the user's group is used. The match must be exact. No wild cards are allowed in the group attribute. If no group is specified, the rule matches to all users.

The idle-timeout attribute sets the idle timeout limit in seconds. If the connection (all channels) has been idle this long, the connection is closed. The default is 0 (zero), which disables idle timeouts.

The print-motd attribute defines whether the message of the day (/etc/motd) is printed when a user logs in interactively to a Unix server. The value must be yes or no. The default is yes.

Each rule can contain environment, terminal, subsystem, tunnel-agent, tunnel-x11, tunnel-local, tunnel-remote, and command elements.

An empty rule allows the specified group to perform all actions.

A default unnamed rule allows all users access to all services. The unnamed rule should be kept as the last rule, so it will match to users that are not set in any group, but you should edit the rule according to your security policy.

environment

This element defines the environment variables the user group is allowed to set at the client side. The variables are given in the allowed attribute as a comma-separated list. By default, the user can set the TERM, PATH, TZ, LANG, and LC_* variables.

Do not use * (asterisk), as it will allow any and all variables, and that can be a security risk.

Allowed variables are normally matched case-insensitively. If case-sensitive variables are needed, specify them using the allowed-case-sensitive attribute.

terminal

This element defines whether terminal access is allowed or denied for the user group. The word allow or deny can be given as the value of the action attribute. By default, terminal access is allowed.

On Unix systems, the chroot attribute can be optionally used to define a directory where the user is chrooted during the terminal session. For more information on chrooting, see Chrooting (Unix).

If terminal access is denied, also shell commands are denied, unless commands are explicitly allowed or set as forced by the command element. See command below.

subsystem

This element defines a subsystem. The element can take the following attributes: type, action, audit (optional), exec-directly (optional), application (optional), and chroot (optional).

The type attribute must be given. It defines the subsystem, for which the settings are made. For example sftp.

The action attribute defines whether the use of the subsystem is allowed or denied. The possible values are allow and deny. The default is allow.

	Note
	Denying the SFTP subsystem denies both SFTP and SCP2 operations to the server, but it does not deny OpenSSH-style SCP operations. To deny OpenSSH SCP, you should restrict remote commands. See command below.

The optional audit attribute can be used to define whether the audit messages of the subsystem are recorded in the system log. Possible values are yes and no. The default is yes. The audit attribute can be used only with the SFTP subsystem.

On Unix, the optional exec-directly attribute can be used to define that the subsystem is launched directly, instead of using the user's default shell. Setting this attribute to yes will allow the use of the subsystem even if the shell defined for the user account prevents login or command execution (for example: /bin/false or /bin/nologin). The default is no.

An example configuration:

<subsystem type="sftp" action="allow" audit="no" exec-directly="yes" />

The optional application attribute can be used to define the executable of the subsystem. This attribute is not necessary with the SFTP subsystem if the SFTP binary is in the default location. Example setting:

<subsystem type="sftp" application="sft-server-g3.exe" action="allow" />

On Unix, the optional chroot attribute can be used to define a directory where the user is chrooted when running the subsystem. For more information on chrooting SFTP, see Chrooting SFTP (Unix).

The element can contain multiple attribute child elements.

attribute

This element defines an attribute of a subsystem.

The name must be given as an attribute. The value can be optionally given.

This can be used, for example, on Windows platforms to set the user home directory and virtual folders for SFTP.

<subsystem type="sftp" application="sft-server-g3.exe">
  <attribute name="home" value="%USERPROFILE%" />
</subsystem>

For more information, see Defining SFTP Virtual Folders (Windows).

command

This element defines a shell command as allowed, denied, or forced. The element can have four attributes, however, all of them are not used at the same time: action, application, application-case-sensitive, and chroot.

The value of the action attribute can be either allow, deny, or forced. The default is allow.

If the deny action is set, all shell commands are denied and no further attributes should be specified. Commands are also denied if terminal access is denied in the rule and the command element is omitted.

For the allow action, the application can be optionally specified as an attribute. When the allow action is set and the application attribute is specified, running the specified application(s) is allowed and all other applications are implicitly denied. If the application is not given, running all commands is allowed.

For the forced action, the application must be given as an attribute. When the forced action is set, the specified application is run automatically when the user logs in successfully, instead of the application the user is trying to run. All other applications are implicitly denied.

If the terminal element is omitted from the rule and the command element specifies a forced command, terminal is implicitly denied. If the user requests a shell, the forced command is run instead.

If the terminal is explicitly allowed in the rule, the forced action of the command element applies only when the user tries to run remote commands. If the user requests a shell, he can get it normally and the forced command is not run.

If the SFTP subsystem is allowed, the user can also use the scp2/scpg3 and sftp2/sftpg3 programs normally. However, if the SFTP subsystem is denied, trying to use it will not cause the forced command to be run, but gives an error message.

	Note
	Support for legacy OpenSSH SCP in SSH Tectia Server is implemented using a command called scp1-compat-srv. When a client uses OpenSSH SCP to connect to SSH Tectia Server, the server invokes this command. Restrictions on remote commands apply also to OpenSSH-style SCP operations to the server.

Users can also define forced commands for public keys in their authorization files or OpenSSH-style authorized_keys files. However, if a command is defined in the ssh-server-config.xml file, it overrides any commands defined in the authorization or authorized_keys files. For more information, see the section called “Authorization File Options”.

Applications are normally matched case-insensitively. Alternatively, the application can be specified using the application-case-sensitive attribute.

On Unix systems, the chroot attribute can be optionally used to define a directory where the user is chrooted when running the command. For more information on chrooting, see Chrooting (Unix).

tunnel-agent

This element defines whether agent tunneling (forwarding) is allowed or denied by the server.

The word allow or deny can be given as the value of the action attribute. By default, agent forwarding is allowed.

For more information on agent forwarding, see Agent Forwarding (Unix).

tunnel-x11

This element defines whether X11 tunneling (forwarding) is allowed or denied by the server.

The word allow or deny can be given as the value of the action attribute. By default, X11 forwarding is allowed.

For more information on X11 forwarding, see X11 Forwarding (Unix).

tunnel-local

This element defines a rule for local TCP tunnels (port forwarding). There can be several of these rules. When a user attempts tunneling, the rules are read in order and the first matching rule is used. See the section called “Tunneling Rule Processing” for details.

The word allow or deny can be given as the value of the action attribute. By default, local tunnels are allowed.

Tunneling restrictions can be further defined with the src and dst elements.

src

This element defines source address(es) and port(s) for local tunnels.

The address or the fqdn (not both) can be given as an attribute.

The address can be in one of the following formats:

a single IP address x.x.x.x
an IP address range of the form x.x.x.x-y.y.y.y
an IP sub-network mask of the form x.x.x.x/y

dst

This element defines destination address(es) and port(s) for local tunnels.

The address or the fqdn (not both) can be given as an attribute. Also the port can be given.

The address and the fqdn can be in the formats described above for the src element.

The port attribute can specify a single port or a port range (for example, 2000-9000).

For more information on local tunnels, see Local Tunnels.

tunnel-remote

This element defines a rule for remote TCP tunnels (port forwarding). There can be several of these rules. When a user attempts tunneling, the rules are read in order and the first matching rule is used. See the section called “Tunneling Rule Processing” for details.

The word allow or deny can be given as the value of the action attribute. By default, remote tunnels are allowed.

Tunneling restrictions can be further defined with the src and listen elements.

src

This element defines source address(es) and port(s) for remote tunnels.

The address or the fqdn (not both) can be given as an attribute.

The address can be in one of the following formats:

a single IP address x.x.x.x
an IP address range of the form x.x.x.x-y.y.y.y
an IP sub-network mask of the form x.x.x.x/y

listen

This element defines listen address(es) and port(s) for remote tunnels.

The address and the port can be given as attributes.

The address can be in the formats described above for the src element.

The port attribute can specify a single port or a port range (for example, 2000-9000).

For more information on remote tunnels, see Remote Tunnels.

Sample rule elements are shown below:

<!-- Administrators are allowed to do anything. -->
<rule group="admin" />

<!-- The finance inspector. -->
<rule group="finance-inspector" print-motd="no">
  <tunnel-local action="allow">
    <!-- Microsoft SQL ports. -->
    <dst fqdn="finance-db.example.com" port="1433" />
    <dst fqdn="finance-db.example.com" port="1434" />
  </tunnel-local>
  <tunnel-remote action="deny" />
  <!-- Can execute commands and shells, as no overriding behavior is defined. -->
</rule>

<!-- Remote access. -->
<rule group="remote-access" idle-timeout="600">
  <!-- Setting terminal action to "deny" also denies shell commands, 
       unless they are specifically allowed.     -->
  <terminal action="deny" />
  <subsystem type="sftp" application="sft-server-g3" chroot="%homedir%" />
  <!-- The listed local tunnels are allowed, other local tunnels are 
       denied. -->
  <tunnel-local action="allow">
    <!-- IMAP. -->
    <dst fqdn="imap.example.com" port="143" />
    <dst fqdn="imap.example.com" port="993" />
    <!-- POP. -->
    <dst fqdn="mail.example.com" port="109" />
    <dst fqdn="mail.example.com" port="110" />
    <dst fqdn="mail.example.com" port="995" />
  </tunnel-local>
  <tunnel-remote action="deny" />
</rule>

<rule group="backup">
  <terminal action="deny" />
  <!-- This account is only used to back up the disk drive. -->
  <command application="dd if=/dev/hda" action="forced" />
  <tunnel-local action="deny" />
  <tunnel-remote action="deny" />
</rule>

<!-- This rule is used to force password change. -->
<rule group="passwd-change">
  <terminal action="deny"/>
  <subsystem type="sftp" application="sft-server-g3" action="deny" />
  <command application="/usr/bin/passwd" action="forced" />
  <tunnel-local action="deny" />
  <tunnel-remote action="deny" />
</rule>

<!-- The default rule, used if the user has not been put to any group. -->
<rule>
  <!-- The listed environment variables are allowed and all others are 
       denied. There is no "denied" setting. -->
  <environment allowed-case-sensitive="TERM,PATH,TZ,LANG,LC_*" />
  <terminal action="deny" />
  <subsystem type="sftp" application="sft-server-g3" chroot="%homedir%" />
  <!-- Only the date command is allowed. Other commands are denied. -->
  <command application="date" action="allow" />
  <tunnel-local action="allow" />
  <tunnel-remote action="deny" />
</rule>

Forcing Password Change

On Unix, the Server default configuration includes settings for forcing the users to change expired passwords.

When users whose passwords have expired attempt to login, the settings in <rule group="passwd-change"> are applied to them. The passwd-change group settings deny all services and force the user to enter a new password.

<!-- This rule is used to force password change. -->
<rule group="passwd-change">
  <terminal action="deny"/>
  <subsystem type="sftp" application="sft-server-g3" action="deny" />
  <command application="/usr/bin/passwd" action="forced" />
  <tunnel-local action="deny" />
  <tunnel-remote action="deny" />
</rule>

Note that the user needs an ssh2/sshg3 terminal client to be able to change the password.

In case you start creating a configuration from scratch, you need to manually add a rule for handling expired passwords. Make sure you do not accidentally allow any services for users who are directed to the passwd- change group after their password has expired. You can use the ssh-server-config-example.xml settings as a model.

Tunneling Rule Processing

The tunneling rules defined by the tunnel-local and tunnel-remote elements operate in the same way as selectors. Inside a tunnel rule, elements of different type are in AND relation to each other. If a tunnel rule contains several items of the same type, they are in OR relation to each other.

For example, the following rule matches if both the listen port and the source FQDN match:

<rule>
  <tunnel-remote action="allow">
    <listen port="1-9000" />
    <src fqdn="trusted.example.com" />
  </tunnel-remote>
...
</rule>

For example, the following rule matches if either of the source addresses match:

<rule>
  <tunnel-remote action="deny">
    <src address="192.168.23.1" />
    <src address="10.1.0.1" />
  </tunnel-remote>
...
</rule>

With several tunneling rules of the same type, the first matching rule is used. For example, the following configuration allows local tunnels to all other addresses in network 192.168.14.0/24 except 192.168.14.21-192.168.14.30:

<rule>
  <tunnel-local action="deny">
    <dst address="192.168.14.21-192.168.14.30" />
  </tunnel-remote>

  <tunnel-local action="allow">
    <dst address="192.168.14.0/24" />
  </tunnel-remote>
...
</rule>

If the tunnel-local elements were in different order, tunnels to the whole 192.168.14.0/24 network would be allowed as any tunneling attempts would match the first (allow) rule and the second (deny) rule would not be read.

For more examples of tunneling rules, see Chapter 8.