This document provides programming aids relevant to EntireX Security programming. It assumes you are familiar with the basics of EntireX Broker ACI programming. See EntireX Broker ACI Programming. It covers the following topics:
Note:
ACI versions 1-7 apply to the communication model client and server
only. ACI version 8 and above apply to the communication models client and
server and publish and subscribe.
See Overview of EntireX Security for overview of concepts and installation.
If your applications are using ACI versions 1 to 7, you will decide at installation time whether they are to communicate with a secured Broker. Your administrator will probably have installed components of EntireX Security into the Broker stub environment(s) and into the Broker kernel.
If your environment is configured using components of EntireX Security, your applications can communicate only with secured Broker kernels. If you attempt to communicate with both secured and non-secured Broker kernels, you will receive ACI response code 00200379, indicating "inconsistent security installation".
To achieve greater flexibility, particularly when migrating applications
from development to production, ACI version 8 introduces the new functionality
described in the following table. For ACI version 8 and above, the application
may assign to the broker control block field KERNELSECURITY
one of the following values:
| Value | Description |
|---|---|
| N | Application does not intend to communicate with a secured Broker kernel. |
| Y | Application intends to communicate with a Broker kernel which is secured using EntireX Security. |
| U | Application intends to communicate with a Broker kernel which is secured with the customer's own security exits. |
This information indicates the application's intention and ensures that
the correct execution occurs in the Broker stub and the Broker kernel. If the
stub and the field KERNELSECURITY do not match,
the application will receive ACI response code
00200379.
If an improper value is assigned, it is treated as a blank. To make this
assignment seamless, use an initial KERNELVERS
command when communicating with each Broker kernel so that the field is
assigned automatically.
Note:
The default value (binary zero or space) specified in this field
will result in the behavior being determined by the security configuration
rather than programmatically. It is therefore possible to communicate either
with a secure or non-secure Broker.
Issuing a KERNELVERS
command will return information in the
KERNELSECURITY
field of the broker control block structure to indicate whether the
application is communicating with a secure or non-secure Broker Kernel. This
information can be important for ensuring the security of transactions and when
making decisions such as prompting for USER-ID and
PASSWORD values.
The following values are returned in the
KERNELSECURITY field for ACI version 8 and above:
| Value | Description |
|---|---|
| N | This is not a secured Broker kernel. |
| Y | This is a secured Broker kernel which is using EntireX Security. |
| U | This is a secured Broker kernel which is using the customer's own written security exits. |
By issuing a KERNELVERS
command, the appropriate value of KERNELSECURITY is
automatically assigned to the control block structure; the user
application does not need to take any further action other than supplying the
correct USER-ID and
PASSWORD. The application
must maintain the contents of the control block structure for
the duration of communication with the Broker kernel in order to retain the
correct value of the
KERNELSECURITY
field. See Broker ACI Control Block Layout for
Assembler |
C |
COBOL |
Natural |
PL/I |
RPG.
Notes:
USER-ID and TOKEN if
specified). Furthermore, if the application communicates with different Broker
kernels, a separate copy of the Broker control block must be maintained for
each user and each Broker ID.
The application is responsible for assigning the correct
USER-ID and
PASSWORD values to the
control block structure. This information is normally
communicated through the LOGON
command, since this command initiates the user's session with the Broker
kernel. Where the attribute file contains
AUTOLOGON=YES the first
command issued by a user does not have to be LOGON,
in which case the application must supply USER-ID and PASSWORD credentials for
the commands SEND or
REGISTER.
Supplying the USER-ID and PASSWORD could subsequently be required if the
user times out due to expiration of either
CLIENT-NONACT,
SERVER-NONACT,
PUBLISHER-NONACT
or SUBSCRIBER-NONACT
time limits. If the user context has timed out due to these inactivity limits
being exceeded, one of the following events will occur when the application
attempts to issue the next command.
Application must perform another explicit
LOGON with
correct credentials in the USER-ID and
PASSWORD fields:
AUTOLOGON=NO in the attribute file, or AUTOLOGON=YES and
FORCE-LOGON=YES.
Application must supply correct credentials in USER-ID and PASSWORD
fields:
AUTOLOGON=YES
in attribute file, FORCE-LOGON=YES not specified in the
control block.
Subsequent commands do not require explicit
LOGON to be issued.
Application has attempted to transfer control to a different thread,
or process, without correctly transferring the necessary values of USER-ID, TOKEN
and STOKEN:
The application transferring control must make values of USER, TOKEN
and STOKEN available to the application that is delegated to continue thread of
execution.
Application has not correctly maintained the value of security token
(STOKEN) in the control block structure:
The application must maintain the value of STOKEN in order to
communicate securely with Broker kernel without sending PASSWORD with
each command.
The application is able to change the password by assigning both
PASSWORD and
NEWPASSWORD
fields of the control block structure. This must be done at the
time of initial authentication or at a subsequent time when authentication is
repeated due to timeout. It cannot be done at an arbitrary time by assigning
the NEWPASSWORD field.
The PASSWORD and
NEWPASSWORD fields are always communicated in an
encrypted format.
EntireX Security automatically generates a non-repeated security token,
which is placed in the ACI control block of the calling application. A unique
security token is generated on behalf of all Broker participants only after
successful authentication has occurred, and is used to ensure nobody can
"tap in" to a participant's session. The calling application is
responsible for maintaining the contents of the control block
structure for the duration of its communication with the Broker kernel in order
to ensure the correct value of security token is available on subsequent
commands. An incorrect value of security token will cause access to be denied.
Security token avoids the need for applications to supply a password except for
presenting this once during the LOGON command, or
the first command (excluding KERNELVERSION), if
AUTOLOGON=YES is defined.
If a LOGOFF command is issued or a participant is
timed out, the password must be reentered so that a new unique security token
can be generated.
An additional benefit of the security token is that it enables an
application to transfer its execution to a different thread or even to a
different process. This requires the application to make available the
following fields of the control block structure to the program
which is delegated to continue the thread of execution:
USER, TOKEN and
STOKEN. However, it is not necessary for the program
transferring control to make its password available.
Note:
If an application is unwilling or does not want to maintain the
security token field (STOKEN) in the control block structure, it is possible for the systems administrator to
configure the following field in the EntireX Security configuration module:
BKISTK=Y. See Ignore Security Token.
This mechanism is available where at least one application and Broker kernel are executing on z/OS and communication is through Entire Net-Work (Adabas SVC).
Trusted User
ID is an optional mechanism with which EntireX Security determines
the identity under which the application is executing, without the application
having to provide the USER-ID and
PASSWORD in the Broker control block. See Trusted User ID.
The benefit of this mechanism is that application components executing on z/OS never have to provide credentials for authentication. This is because the identity under which execution occurs has already been verified when initially accessing the machine in each of these cases:
online users
batch jobs or started tasks.
All subsequent security authorization checks - for example
SEND
or REGISTER -
are then performed under the known user ID under which the application
executes.
Application components intending to utilize Trusted User ID must assign
the Broker control block field USER-ID only. The value
assigned to this field is arbitrary for security purposes but required in order
to satisfy execution the stub. The application is not allowed to assign any
value to PASSWORD if Trusted User ID is used. The
following example is given:
USER-ID = 'SERVER123' /* arbitrary value: used by Broker but not
significant for security purposes */
PASSWORD = ' ' /* password field must be
set to blanks or binary zeros */
If the application does not clear the PASSWORD
field, EntireX Security will assume that the application does not want to use
Trusted User ID. Therefore valid credentials must be supplied to the
USER-ID and PASSWORD fields in
order to perform conventional authentication.
See also Trusted User ID.
Server applications are able to determine the user ID under which the
partner client is executing by examining the content of the
CLIENT-USERID field exposed in the Broker control block.
Specifically, the CLIENT-USERID field should be examined
on the first RECEIVE command of each new
conversation to obtain the identity of the client. When EntireX Security is
active, the server application is able to rely on the accuracy of the client
user identity since it is derived from the user ID and password credentials
supplied by the client.
See also Trusted User ID and Verified Client User ID.
This topic does not apply to the publish-and-subscribe communication
model since this requires an explicit logon and cannot use
AUTOLOGON.
FORCE-LOGON
is used to override the AUTOLOGON feature of the
Broker, with the result that the user does not log on to the Broker kernel
implicitly with the first command issued but instead requires an explicit
LOGON. When
this option is used, it is necessary for the client and server to issue
explicit LOGON function calls - even after the
expiration of a client timeout
CLIENT-NONACT
or server timeout
SERVER-NONACT.
See Timeout Parameters.
FORCE-LOGON can be useful in cases where an
implicit logon would be undesirable, for example when attempting to
authenticate a user. Specifically, unless the password was communicated with
every command, an implicit logon - after a period of inactivity - would fail
because of a missing PASSWORD.
When FORCE-LOGON is set - and in the case of
a client/server inactivity timeout - error
00200134
is returned instead of an implicit logon being performed automatically.
Therefore, the specification of FORCE-LOGON can be
used to give the programmer the opportunity to provide the ACI field
PASSWORD, which is needed for successful
authentication.
Note:
Caution should be taken when repeating a failed authentication
attempt for both an explicit and an implicit logon. Repeating the attempt
several times can lead to a revocation of the user ID, depending on the
configuration of your security system.
Applications that create publications are subject to authorization
requests under EntireX Security. For every new publication, an authorization
check is performed based on the
TOPIC.
Publications are transmitted to subscriber applications only if the
authorization check is successful; otherwise an ACI response is returned to the
application issuing the SEND_PUBLICATION
command.
Subscriber applications are subject to an authorization check if
security is installed for EntireX Broker. An authorization check based on the
topic is performed when the subscriber application issues a
SUBSCRIBE
command. The application is allowed to subscribe only if the authorization
check is successful; otherwise an ACI response code is returned to the
subscriber. Similarly, if the administrator performs third-party subscription
or unsubscription on behalf of a subscriber using command and information
services (CIS), an
authorization check is made, based on the topic. See Writing Applications: Command and Information Services.
The ACI error response codes encountered for authorization failures are: 00080009 | 00080010.
See also publish-and-subscribe example under Resource Profiles in EntireX Security and Writing Applications: Client and Server.
Client applications are automatically subject to authorization requests if security is installed for EntireX Broker. For UNIX and Windows, see Authorization Rules.
An authorization check based on class, server and service is performed
for the first SEND of a
conversation and on every SEND if there is only one
message in the conversation (CONV-ID).
Messages are transmitted through to the server application only if the
authorization check is successful; otherwise an ACI response is given to the
client.
Server applications are automatically subject to authorization requests
if security is installed for EntireX Broker. For UNIX
and Windows, see Authorization Rules. An authorization
check based on class/server/service is performed when the server application
issues a REGISTER
command. The server is allowed to register only if the authorization check is
successful; otherwise an ACI response code is returned to the server
application. In a similar way, if the administrator terminates a server through
Command and Information Services (CIS), an authorization check is made based on
the class/server/service.
The ACI error response codes encountered for authorization failures are: 00080009 | 00080010.
See also client-and-server example under Resource Profiles in EntireX Security and Writing Applications: Client and Server.
For UNIX and Windows, Broker authorization checks are made using a set of definitions maintained in an LDAP repository (Lightweight Directory Access Protocol). Authorization rules are the mechanism by which authorization checks are performed for UNIX and Windows. For more information see Configuring Authorization Rules under UNIX | Windows.
Encryption of the message data with the EntireX Security encryption is configured in the broker and in client and server applications.
The ENCRYPTION-LEVEL attribute configured in the broker attribute file is a service-specific attribute. Allowed values are 0, 1, and 2. Using
ENCRYPTION-LEVEL 1 or 2 requires SECURITY=YES in the DEFAULTS = BROKER section.
| Encryption Level | Description |
|---|---|
| 0 | No encryption requested, but allowed. |
| 1 | Encryption for server requested, client can encrypt. |
| 2 | Encryption for client and server requested. |
Example:
DEFAULTS = SERVICE CONV-NONACT = 5M DEFERRED = YES SERVER-NONACT = 10M ENCRYPTION-LEVEL = 2 CLASS = RPC, SERVER = SRV1, SERVICE = CALLNAT, TRANSLATION = SAGTCHA
Applications can configure the encryption level by setting the broker ACI field ENCRYPTION-LEVEL or using the methods of the Java ACI. Allowed values are 0, 1, and 2. 0 does not encrypt the message, 1 encrypts the message
to the broker, while 2 encrypts the message to the broker and requests that the partner application also encrypts the message.
To guarantee end-to-end encryption from client to server, use encryption level 2.
The broker controls the encryption and issues one of the error codes 00200401 | 00200419 | 00200420 | 00200421 | 00200422 if encryption levels do not match.
In all cases encryption requires broker connections secured with EntireX Security. Broker returns error code 00210402 if this is violated.
Encryption behavior is slightly different depending on whether the application is Java-based or not. For non-Java-based applications, the behavior depends on the ACI version. These differences are described below.
For non-Java applications, Encryption behavior depends on the ACI version:
Note:
In the sections below, "EL" refers to the field ENCRYPTION-LEVEL in the ACI control block.
Encryption is configured by the application per message. The table shows allowed combinations of ENCRYPTION-LEVEL and EL.
ENCRYPTION-LEVEL in Broker Attribute File
|
Server | Client |
|---|---|---|
| 0 (or undefined) | EL= 0 or 1 | EL= 0 or 1 |
| 1 | EL= 1 | EL= 0 or 1 |
| 2 | EL= 2 | EL= 2 |
Encryption is configured by installing the security exit for the application (security exit is secuexit.dll on windows, secuexit.so on UNIX, SECUEXI0 on z/OS) and the environment variable NA2_BKPRIV (or ETB_ENCRPYT) is set to 1. The table shows allowed combinations of ENCRYPTION-LEVEL and EL.
ENCRYPTION-LEVEL in Broker Attribute File
|
Server | Client |
|---|---|---|
| 0 (or undefined) | EL= 0 | EL= 0 or 1 |
| 1 | EL= 1 | EL= 0 or 1 |
| 2 | EL= 2 | EL= 2 |
Encryption is configured by installing the security exit for the application (security exit is secuexit.dll on windows, secuexit.so on UNIX, SECUEXI0 on z/OS) and the environment variable NA2_BKPRIV (or ETB_ENCRPYT) is set to 1. Only ENCRYPTION-LEVEL in the broker attribute file is available. Field EL for the stub is not available.
On the Broker object, use the methods setSecurity(...) or useEntireXSecurity(...) to enable EntireX Security and set the encryption level.
See Broker in the Javadoc for details.
For RPC servers, the encryption level is set in the configuration. See documentation of the configuration for the specific RPC server.
C-based RPC Clients
Use the structure ERX_CLIENT_IDENTIFICATION to set the encryption level.
See ERX_CLIENT_IDENTIFICATION for details.
Natural RPC Clients
Natural Clients use user exit USR4009N to set the encryption level. See Operating a Natural RPC Environment in the Natural documentation.