One minute MVS: LDAP defining resources

Having set up an LDAP server, you need to add information to the directory. This is not very well described in the TDS documentation.

Basic overview of data

To add information about a user use a file in USS like colin.ldif

dn: cn=colin, o=Your Company
objectclass: top
objectclass: person
objectclass: organizationalPerson
cn: LDAP Administrator
sn: Administrator


  • the “key” to identify an entry is the dn ..
  • objectclass is the sort of object, and what attributes it can have. It can have many object classes
  • cn: and sn: are attribute values
  • There is a blank line following to indicate end of definition. You can have many of these in a file, to allow you to do a bulk update.

How do I display the contents?

You need to issue a query. This comes in two parts, identifying the user, and the request.

To identify requestor you need something like

ldapsearch -h -D “cn=ibmuser, o=Your Company” -w ? …

and the query for example, to list all the information about cn=ibmuser, o=Your Company add the following to the ldapsearch request above

-b “cn=ibmuser, o=Your Company” “(objectclass=*)”

This gives

cn=ibmuser, o=Your Company

For all information under o=Your Company

-b “o=Your Company” “(objectclass=*)”

For only the list of sn for all users

-b “o=Your Company” “(objectclass=*)” sn

This gives

o=Your Company

cn=colinw, o=Your Company

cn=colin, o=Your Company

cn=LDAP Administrator, o=Your Company

cn=ibmuser, o=Your Company

How do I add information?

If I want to add a userid definition for ibmuser (above) so I can login with RACF, I need to add attribute

ibm-nativeId: COLIN

This attribute is in object type

objectclass: ibm-nativeAuthentication

So to be able to specify the ibm-native-ID: attribute, you need to tell specify the object class as well.

My definition is now

dn: cn=colin, o=Your Company
objectclass: top
objectclass: person
objectclass: organizationalPerson
objectclass: ibm-nativeAuthentication
cn: LDAP Administrator
sn: Administrator
ibm-nativeId: COLIN

Add it to the directory

You can add this to the directory using

ldapmodify -a -h … -p … -D “…” -w … -f colin.ldif


  • -a says add (instead of modify)
  • -f colin.ldif is the name of the file with the statements in it.

Modifying an entry

If you want to modify an existing entry, you can change the whole entry, or parts of it.

To add an entry

dn: cn=colin, o=Your Company
changetype: add
objectclass: top

To delete a whole entry

dn: cn=colin, o=Your Company
changetype: delete

To add an attribute to an entry

dn: cn=colin, o=Your Company
changetype: modify
add: attrccp
attrccp: value1
attrccp: value2…

This adds two attrccp values to the definition

To modify an existing attribute

dn: cn=colin, o=Your Company
changetype: modify
modify: ibm-nativeId
ibm-nativeId: PAICE

To delete an attribute

dn: cn=colin, o=Your Company
changetype: delete
delete: ibm-nativeId

This deletes all ibm-nativeID attributes.

If you want to delete a specific attribute specify it after the delete: line

dn: cn=colin, o=Your Company
changetype: delete
delete: attrccp
attrccp: value2

One minute MVS. Getting started with LDAP on ADCD.

LDAP is a standard protocol for accessing directory information over TCP/IP. For example the command

ldapsearch -h -D “cn=Admin, o=Your Company” -w secret -b “o=Your Company” “(objectclass=*)” aclEntry

This sends a request to IP address with userid cn=… and password “secret”, for information under the subtree of “o=Your Company” and requests is sends back information on any ACL entries.

z/OS implementation

LDAP on z/OS is also know as Tivoli Directory Server.

It can run with different backend databased from DB2 to a files in a USS directory. It can interface to RACF so you can use query userid and group information from RACF through LDAP.


You need to configure a schema of what fields there are, and the relationship. For example, for an organisation telephone directory you might have

dn: cn=LDAP Administrator, o=Your Company
objectclass: organizationalPerson
cn: LDAP Administrator
sn: Administrator
userPassword: ********


  • dn: cn=LDAP Administrator, o=Your Company This is the internal name of the object, and what part of the data tree it belongs to “o=Your Company
  • objectclass: organizationalPerson defines the object type
  • cn: LDAP Administrator This is the common name ( nick name) of the object
  • sn: Administrator This is the surname of the person
  • userPassword: ******** This is the user’s password. It has been defined that the value is not displayed
  • phoneNumber:1234567 This has been defined so that is can only take numbers and ‘-‘.

You can define your own attributes and properties. You just need to update the schema.

Which database is used?

A sample LDAP configuration might contain

suffix “o=Your Company”
databaseDirectory /var/ldap/ldbm

  • database LDBM says there is a database in a USS directory
  • GLDBLD31/GLDBLD64 are the names of the interface routines to use.
  • suffix “o=Your Company” is the root of the subtree in this database
  • databaseDirectory /var/ldap/ldbm is the name of the USS directory

You can configure LDAP to say for these names(o=someoneElsesCompany) go to another LDAP with at this address.

If I use a query like ldapsearch -h -D “cn=LDAP Administrator, o=Your Company” -w secret -b “o=Your Company” “(objectclass=*)” aclEntry…. the -D cn=LDAP Administrator, o=Your Company” says look for a userid with the given data in the o=Your Company subtree. With the above definitions it would look in the the USS file system under /var/ldap/ldbm, for a userid cn=LDAP Administrator, o=Your Company.

Configuring an LDAP server on ADCD.

ADCD is a preconfigured system which can on on zPDT and ZD&T. These provide a system 390 emulator. This system comes with a lot of software installed, and some subsystems such as z/OS, MQ, DB2, IMS, CICS and z/OSMF pre configured.

The software for LDAP(Tivoli Directory Server) is installed but not configured. The documentation is extensive, and the configuration file is very large (with lots of comments). You run a configuration script which produces some files.

However for a simple configuration you only need a few files to run.

Some of these files do not work – for example they try to define a userid with an existing Unix uid.

I’ve taken the updated files and put them on git hub.

The TDS documentation is here.

If you get into a mess you can just delete the /var/ldap/ldbm directory and start again!

Getting cipher keys to another site – the basics of Exporter and Importer keys in ICSF.

I’ve spent some time (weeks) exploring ICSF with the overall mission of sending an encrypted data set between two sites. Looking back it was like the saying when you’re up to your neck in alligators, it’s hard to remember that your initial objective was to drain the swamp.

I’ve explored many parts of ICSF. One area that confused me for a while was the use of Key Encrypting Keys, or Exporter and Importer keys, also known as transport keys. I’ll explain my current thoughts on it – bearing in mind these may not be 100 accurate.

If I want to encrypt data on one system, send the encrypted data to another system, and decrypt it on that system; the sender system needs encryption key, and the receiving system needs the decryption key.

Typically the data encryption is done using a symmetric key, where the same key is used to encrypt as decrypt. You can also use asymmetric keys where you encrypt with one key, but need a different key to decrypt it.

The first challenge is how to securely get the keys onto the systems.

  • You cannot just email the symmetric key to the remote site, because bad people monitoring your email will be able to get the symmetric key.
  • You could print the key, and sent it through the mail system, courier, or carrier pigeon to the remote site. This still means that bad guys could get the key (using a telephoto lens through a window, through the security cameras, or catch the pigeon).
  • A secure way using a technique called Diffi-Hellman can be used to create the same symmetric key at each end. It uses private/public keys and an agreed seed. No sensitive data is sent between the two systems.

When setting up ICSF to use cross system, you need to set up keys for A to B, and for B to A. You can use the same keys and seed for both, but the keys will be different.

If you are setting up several independent systems you will need keys A to B, B to A, A to C, C to A, B to C and C to B etc.

Setting up keys

You can set up a CIPHER key for encrypting data, and you can set up a MAC key for checking that what was sent is the same as what was received. (You hash the data, then encrypt the hashed value. Send the hash along with the data. If both ends do it, they should get the same answer!)

Decrypting and re-encypting

The keys are stored on disk in an encrypted format. The key for this is within the cryptographic hardware. If you want to send a key to another system, you need to encrypt it.

ICSF has a function in the hardware which says “here is some data encrypted with the hardware key, decrypt it, and re-encrypt it with this other key”. It has a matching function “here is some encrypted data, decrypt it, and re-encypt it with the hardware key”. This way the clear text of the key is never seen. To extract a key, you need to provide a key to re-encrypt it.

How to send the key to the remote system?

You have several choices depending on the level of security your enterprise has.

ICSF has a function to export a symmetric key using

  • an RSA public key. At the remote end you need the private key to be able to decrypt it. This needs an RSA key size of at least 2048.
  • an exporter key. At the remote end you need the matching importer key to decrypt it. Under the covers this is a symmetric (AES) key. The AES technique is faster, and “stronger” than RSA (it takes longer to break it). An AES 256 key is considered stronger than an RSA 4096 key.

I think of exporter and importer keys as a generalised public key and private key. The concept may be the same, but the implementation is very different.

The question “should I use RSA or Exporter/Importer” comes down to how secure do you want to be. If you export a key once a week the costs are small.

These keys used to encrypt keys, are known as key encrypting keys, and you often see KEK in the documentation.

What are the advantages of using an exporter and an importer key?

If you are using a symmetric key to encrypt your data, why is there exporter and importer keys, and not just one key?

If a purely symmetric key was used as a key encryption key, this means that if you are allowed to encrypt with it – you can decrypt with it. This then means you can decrypt other people’s data which used the same key.

By having a one key for encryption and another key for decryption you can isolate the authorities. If I have access to the exporter key, although I extract it and send it to my evil colleague at the remote end, they cannot use it to decrypt the data. If I have access to a symmetric key generated with Diffi-Hellman, the same key is used at each end, and so I could extract it and send it to my evil colleague who could use it.

Setting up my first importer or exporter key.

Use a pair of matching ECC public and private keys. The key type (eg Brain Pool) and key size must match. You use CSNBKTB2 to define a skeleton of type importer or exporter (or cipher). You then pass the skeleton into CSNDEDH along with the private and public keys. You can then use CSNBKRC2 to add it to the local CKDS.

You can do this at the remote end, just switch the public and private, and the importer to exporter. You do not need to send a key between the two systems.

Once you have defined your first exporter and importer key there is an alternative way of creating transport keys, using CSNBKGN2.

Alternative way of defining more transport keys.

There are three scenarios to consider when setting up transport keys between two systems.

  • On system A create an exporter key for system A, and create an importer key system for system B. The importer key will need to be encrypted as it need to be send to the system B. The system A (local) exporter key can then use the key to export a (cipher or MAC) key. Then write this key to a file and send it to the other system. At system B, use the importer key to import this data into the system B’s key store.
  • Do it from the “other end”. On system B create an importer key for system B, and create an exporter key for system A. This exporter key will need to be encrypted as it needs to be send to system A. Then write this key to a file and send it to the other system. At system A import it. System A can then use the new key to export a (cipher or MAC) key from that system, write it a file, and send it to system B which can import it.
  • Define them on another system. On system C create an exporter key for system A, and an importer key for system B. These both need to be encrypted (and use different keys to encrypt each one). Send the encrypted keys to system A and system B. They can import the key, and send keys from system A to System B.

You use CSNBKGN2 to do this. I’ll only cover the first two cases.

You need to specify one of

  • keytype1 = “IMPORTER”, keytype 2 = “EXPORTER” or
  • keytype1 = “EXPORTER”, keytype 2 = “IMPORTER”

Use RULE = “OPEX ” where

  • OP means store the first key in the local (OPerational) key store.
  • EX means keytype 2 is to be exported. You have to specify an AES exporter key in the field key_encrypting_key_identifier_2.

Note. To export the key you already need an AES exporter key! This means you cannot create your first transport key with this method.

The output of this function is

  1. a key which you can add to the local key store using CSNBKRC2,
  2. a key which you can write to a file and send it to the remote system.

At the remote system you use CSNDSYI2 and the Importer key to import the data and re-encrypt it with the local key. Then use CSNBKRC2 to add it to the CKDS.

Which technique is better?

I think using a pair of matching ECC public and private keys and Diffie-Hellman is simpler, as it does not involve sending a file between systems. As this activity is done infrequently it may not matter.

Using DH many not be as secure as the alternative.

How many transport keys do I need?

You could create a new transport key every week if you wanted to. It is only used to send a data key to a remote system, so transient. When you have created a cipher key to encrypt and decrypt your data you need to keep this as long as you need access to the data. Once you recreate it – you are unable to read data sets with the previous key.

Depending on your security requirements you might want to have more than one transport key for data isolation. For example test data and production data have different keys.

You could use Enterprise Key Management Foundation (EKMF) from IBM to manage your keys

EASY ICSF – making it easy to use the API to generate and export/import keys

I’ve put some code on GITHUB which has C and REXX code which have a simpler interface to ICSF. The code examples hide a lot of the complexity.

For example to generate an AES CIPHER key the high level C code is

// build the skeleton for C=CIPHER ( could be E for exporter or I for IMporter
//  It returns the skeleton and its length 
rc = skeletonAES("C",& pToken,& lToken); 
if ( rc != 0 ) return rc; 

// Generate the key - passing the skeleton and returning the Token
// input the skeleton 
// output the token 
rc = GENAES2(pToken,&lToken); 
if ( rc != 0 ) return rc; 

// Add this to the CKDS                                                        
rc = addCKDS(pKey,pToken       ,lToken,pReplace); 
if ( rc != 0 ) return rc; 

printf("GENAES %s successful\n",pKey); 
return rc; 

To export an AES key

// Pass in the name of the AES key pKey
// the name of the encryption key (AES EXPORT or PKI) pKek
// Get back the blob  of data
rc =exportAES (pKey,pKek,&pData, &lData); 
 if (rc > 0 ) return rc; 
Write the blob to a file specified by dd
 rc = writeKey("dd:TOKEN",pData,lData); 

It gives in //SYSPRINT

Exists: CSNBKRR2 read AESDHE CKDS rc 0 rs 0 No error found


Exists: CSNBKRR2 read PKDS2 CKDS rc 8 rs 10012 Key not found
Exists: CSNDKRR read PKDS2 PKDS rc 0 rs 0 No error found .



ExpAESK:CSNDSYX rc 8 rs 2055 The RSA public key is too small to encrypt the DES key


  • Exists: CSNBKRR2 read AESDHE CKDS rc 0 rs 0 No error found
    • It used the ICSF CSNBKRR2 to check AESDHE is in the CKDS
    • It reports some info on the key. It is a Symmetric (AES) Exporter and can do AES processing
  • Exists: CSNBKRR2 read PKDS2 CKDS rc 8 rs 10012 Key not found
    • This is ok — it looks in the CKDS first – but as this is a PKI – it will not be found
  • Exists: CSNDKRR read PKDS2 PKDS rc 0 rs 0 No error found .
    • It is found in the PKDS
    • This gives info about the Key Encryption Key. It is RSA and has a private key. The key size is 1024
    • This is the rule used
  • ExpAESK:CSNDSYX rc 8 rs 2055 The RSA public key is too small to encrypt the DES key
    • The size of the PKI key was too small.
    • As well as giving the return code and reason code, it gives the reason for some of the reason codes.
    • When I repeated this with a RSA key with a large enough key – it worked successfully.

There are also some macros such as

  • isRSAPRIV… is this token (blob of data) an RSA private key?
  • isEXPORTER … has this token been defined as an EXPORTER key?

I use these to check keys being used for operations to check that keys are valid for the ICSF operations.

The git hub code is work in progress. As I find problems I’m fixing them, but overall it should you what you can do with it.

z/OSMF autostart: how to stop it, and how to use it (or not)

I upgraded my z/OS from ADCD Z24A to ADCD Z24C. This has updates to lots of the software, including z/OSMF. This includes some performance fixes, so z/OSMF start up is much quicker and uses much less CPU. However the newer level of ADCD Z24C now starts z/OSMF automatically. It took a few attempts to stop this.

When z/OS starts, it takes configuration parameters from IEASYSxx. You can see which IEASYSxx you are using with the DISPLAY IPLINFO operator command. You can see which IZU parameter you are using with

d iplinfo,izu

With the DISPLAY PARMLIB command, you get the parmlib concatenation

    1   S   C4CFG1 USER.Z24C.PARMLIB
    2   S   C4CFG1 FEU.Z24C.PARMLIB
    3   S   C4SYS1 ADCD.Z24C.PARMLIB
    4   S   C4RES1 SYS1.PARMLIB

Where the ‘S’ means it came from a LOADxx parameter. A ‘D’ means Default SYS1.PARMLIB.

Look in each data set in turn for the IZUPRMxx member (xx=AS in my case).

Contents of the IZUPRMxx member

Within the member is SERVER_PROC(‘IZUSVR1’) This tells the IPL code which server to start.

Within the member is line with AUTOSTART(…). The value can be

  • CONNECT – I think of this as AUTOSTART(NO)
  • LOCAL – I think of this as AUTOSTART(MAYBE)

See here for a discussion.

It is a bit more complex than YES|NO. It has capability to allow one of a group of z/OSMF servers to start.


If you have AUTOSTART(LOCAL) and AUTOSTART_GROUP(COLIN) for more than one IZU servers. Then at IPL it checks to see if a Z/OSM server with AUTOSTART(LOCAL) and AUTOSTART_GROUP(COLIN) is already active. If so – the instance does not start.

The documentation says it checks by having an ENQ on the file system with the AUTOSTART_GROUP value. This implies you need the z/OSMF data directories to be on the same ZFS file system.

Should I use autostart?

This is a tough question. I cannot test it because I only have one LPAR, but I have some thoughts.

Single LPAR, single Z/OSMF instance

This is relatively easy. You can start z/OSMF automatically though commands at IPL, or you can use the z/OSMF IZUPRMxx method, or start it manually.

Multiple LPARs in a sysplex, single Z/OSMF instance.

If you have a shared file system, you can start the z/OSMF instance on any LPAR. If you start the instance more than once, it detects this and will only allow one instance to be active.

You have to plan to be able to starting an instance on different systems. For example the IP address and port for the base system will be different. You’ll need to set up a TCP/IP environment to support this. See HA Liberty web server – introduction to using VIPA to provide high availability connectivity and the z/OSMF documentation

Multiple LPARs in a sysplex, multiple z/OSMF instances.

This is where the autostart may be useful. The first LPAR to be started will start the z/OSMF instance. When other LPARs start, they detect that another z/OSMF in the group is active, and will not start the z/OSMF instance. As with starting a single z/OSMF instance in a multi LPAR environment, you need to plan the connectivity. See HA Liberty web server – introduction to using VIPA to provide high availability connectivity and the z/OSMF documentation.

I struggle to see why starting just one instance is useful. For availability I would want more than once instances to be running at the same time. With only one instance. If you stop it, and restart on a different LPAR, you have a period of a minute or more where you do not have z/OSMF running.

I would have a group_token, so each instance can register the “group name” is active. An application can then ask to be notified when a member of the group becomes active, using standard z/OS services.

Stateless z/OSMF instances

If you are using z/OSMF facilities which save state, the autostart of just one server will not work. For example if you are using any workflow facilities, state is saved in the file system. You need to logon to the same instance to be able to continue working on the workflow. If today you run on LPARA’s z/OSMF and tomorrow you run on LPARB’s z/OSMF you cannot do your workflow.

You need to plan your z/OSMF usage and plan to have “stateless” z/OSMF servers which can use AUTOSTART; and workflow servers – for which you have only one instance (which can be moved around) and do not use autostart.

How do I put today’s date in JCL?

I have a backup job which I run to take current copy of the file and save it with today’s date. For months I’ve been editing it to change today’s date. 10 minutes of browsing the internet showed me how easy it was!

For example


The procedure has


which gave me


I could have used (see here for a complete list)

  • &LYYMMDD for local date
  • &HHMMSS for time
  • &LHHMMSS for local time

Easy – I should have done this years ago!

ICSF: why do I need to have transport keys as well as data encryption keys.

As part of my scenario of encrypting a file and sending the encrypted file to another z/OS system, I struggled to understand why the documentation referred to transportation keys, key encryption keys (KEKs), import keys and export keys.

I found the subject very unclear. As I currently see it (and I’ve changed my view several times). You need the matching key on each system. If this is a symmetric key, it is the same key. If you are using PKI, they keys are asymmetric.

How do you get symmetric keys on both systems. I see there are two ways

  1. Generate the same key on both systems This can be done using private and public keys, and a technique called Diffie-Hellman.
  2. Generate a key on one system, and send it securely to the other system. For this you need to securely package the symmetric keys while they are in transit.

I was able to perform the setup and transfer a file securely to another system without the need for these additional keys. What was I missing?

The discussion about transport keys is for the second example where keys are sent over the network. You can use a CIPHER key to encrypt the key. It comes down to can I do it ? Yes. Should I do it ? No (well, maybe not, it depends on the scale of risk).

Within an IT environment the userid administration should be a different team to the systems programmers. This is to prevent any conflict of interest, fraud, and errors. The system programmers cannot give themselves access to sensitive data. In my small company (with just me in it) I have to do sysprog and userid administration.

IBM has similar guidelines for implementing cryptography. For example

  • Separation of the roles and responsibilities. The people who create keys are different from the people who give access to the keys, and from the people who use the keys.
  • Separation of encryption keys based on what they are used for. A key for encrypting datasets should not be used for encrypting a key to send to a remote system. If a data set encryption key is made public, the key-encryption-key should still be secure.

I could provide isolation of keys by having two keys, one is authorised only for data set encryption and the other authorised only for key encryption, but this separation may not be enough.

Creating exporter/importer using the API

I spent a couple of days trying to create an importer/exporter pair. I found one way of doing it – there may be other (more obscure) ways. It uses Diff-Hellman to create the same key on two sites without transferring sensitive material. I describe it here. It requires each side to have its own private key, and the public key of the other side.

There are three parts

  • Generate a skeleton
  • Use the skeleton, private key and public key to generate the Diffi-Hellman key
  • store it in the key store

Exporter:Generate a skeleton

I used CSNBKTB2 with rules ‘AES ‘||’INTERNAL’||’EXPORTER’.

Exporter:Generate the DH Key

I have a “helper” rexx function which has parameters, private key name, public key name, the completed skeleton.

It used CSNDEDH with

  • rule_array = ‘DERIV01 ‘||’KEY-AES ‘
  • party_identifier (a string both sides agree) = ‘COLINS-id’
  • KEK_key_identifier_length = 0. This is used when the private key is not stored in the PKDS, but passed in encrypted. I think of this as acting as a proxy. “Here is the private key to use – but it has been encrypted with the KEK which is in your local key store”. Setting the length to zero says the private key>is< in the local key store. Definitely an advanced topic!
  • Name of side A’s private key in the PKDS
  • Name of side B’s public key (from the other side) in the PKDS
  • key_bit_length = 256.

It returns a blob encrypted with the local master key.


This is another rexx helper. It takes the name of the key to generate, the encrypted blob, and “replace=Y|N”

This uses

  • CSNBKRC2 to add to the CKDS
  • if it gets record found, and needs to delete it,
    • it invokes delckds which uses CSNBKRD to delete it
    • it tries the add again

Importer ( on the remote system)

The steps are the same, except

  • I used CSNBKTB2 with rules ‘AES ‘||’INTERNAL’||’IMPORTER’.
  • Generate the DH key, you use the other keys, side B’s private, and side A’s public.

To export a key using exporter/importer

If you are using an AES exporter key to encrypt the data you need to use CSNDSYX with

  • The name of the key you want to export
  • The label of the AES exporter key
  • rule_array = ‘AES ‘||’AESKW ‘

It returns a blob which you can write to a data set.

To import the key using exporter/importer

read the data into a buffer

Use CSNDSYI2 with

  • rule_array = ‘AES ‘||’AESKW ‘
  • the name of the importer key

It returns a blob of data.

Use the helper addckds passing the new label name, the blob of data, and replace=yes|no.

  • This uses CSNBKRC2 to add the record, with rule_array = ”
  • If the record exists and replace=yes then
    • use delckds with CSNBKRD and rule_array = ‘LABEL-DL’
    • re-add it

To export a key using PKI public/private keys

If you are using an PKI cipher key to encrypt the data you need to use CSNDSYX with

  • The name of the key you want to export
  • The label of the PKI public key
  • rule_array = ‘AES ‘||’PKOAEP2 ‘

To import the key using pki private key

read the data into a buffer

Use CSNDSYI2 with

  • rule_array = ‘AES ‘||’PKOAEP2 ‘, matching the exporter
  • the name of the private key

It returns a blob of data.

Use the helper addckds passing the new label name, the blob of data, and replace=yes|no.

  • This uses CSNBKRC2 to add the record, with rule_array = ”
  • If the record exists and replace=yes then
    • use delckds with CSNBKRD and rule_array = ‘LABEL-DL’
    • re-add it

ICSF: exploiting Rexx

ICSF provides APIs and commands to manage cryptographic keys. For example to encrypt a datasets you need to define the key that will be used.

You can use Rexx to use the API’s and make your own commands.

There are some Rexx samples provided with ICSF, and there are others on the internet if you search for the API function and Rexx. These tend to be a large Rexx exec written to do one function.

You can use the power of Rexx to allow significant reuses of these execs, by having one Rexx exec to generate a key, another Rexx exec to add it to the keystore, another Rexx exec to export it, and another Rexx exec to import it.


Rexx Address linkpgm facility

With TSO Rexx there is the “address linkpgm” command environment. This allows you to call z/OS functions with Rexx parameters.

For example

rc = 0
z= 16
address linkpgm “ZOSPROG myrc Y Z”

generates the standard low level request

call ZOSPROG(addr(myrc),addr(y),addr(z));

It returns a variable ‘RC’ for example -3 if the program is not found, or the return code from the program.

Be careful not to specify ‘RC’ as a parameter as it may override it.

If does what you tell it. If you are expecting a string to be returned, then the variable you give it must be big enough to hold the data, it cannot allocate a bigger string.

If you want to create a variable of a fixed size you can use

token         = copies(’00’x,3500);

If you are passing a number or hex string, you have to convert it to the internal value.

For example on input

myInt = ‘00000000’x
mylen = C2D(length(“ABCDEFG”),4) /* the 4 says make field 4 (int) wide */

on output, convert the hex return code to a readable hex code

myrc = c2x(myrc)

To create an internal format length you can use either of

lToken = ‘00001964’x /* 6500 */
lToken = d2c(6500,4); /* of size 4*/

Passing parameters to external Rexx programs

You can call external Rexx programs and get a returned data. For example

with the program mycode

parse arg a,b
return 0 “COLINS”||A b||”xxx”

and call it using

zz = mycode(“AA”,”B”)
say zz
parse var zz rc x y
say rc
say x
say y



Using this you can have an external function which generates an AES key, which returns the return code, reason code and the data.

Using hex strings

Many of the ICSF functions return a hex structure. You can convert this from internal using the Rexx function c2x. This takes a string and creates the hex version of it. When you want to use it in another ICSF function you convert it back again using x2c().

x = ‘ABC’
y = c2x(x)
say ‘y:’y /* gives y:C1C2C3 */

When an ICSF function returns data, you can convert it to the hex string, and return it to the caller.

Using lengths

If a hex length has been returned, you can convert it to a Rexx number using C2D

x = ‘00000000c’x
say ‘x:’c2d(x) /* prints x:12 */

Converting from Rexx to internal format

x = 14
y = d2c(x,4) /* a 4 byte field */
say ‘x:’c2x(y) /* display in hex gives x:0000000E */

Using ICSF from Rexx

Using the program

/* Generate a 256-bit AES DATA key to export */ 
rc = genAES() /* this returns several bits of data*/
say "CPBKGN " rc 

parse var rc myrc myrs key 
if myrc <> 0 then return rc 
/* Store the AES DATA key in the CKDS       */ 
/* just return code */ 
rc= addCKDS("REXXLABEL",key) 
say "CPBkrc2" rc 
return 0 


say "In GenAES" 
parse arg a  /* no parameters passed in */ 
/* Generate a 256-bit AES DATA key to export*/ 
key_form               = 'OP  ' 
key_length             = 'KEYLN32 ' 
key_type_1             = 'AESDATA ' 
key_type_2             = '' 
kek_id_1               = COPIES('00'x,64) 
kek_id_2               = '' 
generated_key_id_1 = COPIES('00'x,64) 
generated_key_id_2 = '' 
myrc             = 'FFFFFFFF'x 
myrs              = 'FFFFFFFF'x 
exit_length = d2c(0,4)
exit_data       = '' 
   'myrc'               'myrs'          , 
   'exit_data_length'   'exit_data'     , 
   'key_form'           'key_length'    , 
   'key_type_1'         'key_type_2'    , 
   'kek_id_1'           'kek_id_2'      , 
   'generated_key_id_1' 'generated_key_id_2' 
myrc = c2d(myrc)
myrs = c2d(myrs)                                                                 
IF (myc <> 0 ) THEN 
    SAY 'KGN Failed   (rc='myrc' rs='myrs')' 
    Return  myrc myrs 
Return  myrc myrs c2x(generated_key_id_1)


/* -------------------------------------------*/ 
/*  Add CKDS : label and data                 */ 
/* CSNBKRC2 - Key Record Create2              */ 
/* -------------------------------------------*/ 
parse arg label, token 
say "CPBKRC2 " label token 
myrc = 'FFFFFFFF'x 
myrs = 'FFFFFFFF'x 
exit_length =d2c(0,4)
exit_data = '' 
rule_count = d2c(0,4)
rule_array = '' 
token_length = d2c(64,4)
token =x2c(token) 
LEFT(data,64) /* Make sure string length = 64 */ 
   'myrc'          'myrs'            , 
   'exit_length'   'exit_data'       , 
   'rule_count'    'rule_array'      , 
   'label'         'token_length'   , 
myrc = c2d(myrc)
myrs = c2d(myrs)                                                                 
IF (myrc <> 0 ) THEN 
    /* print the return code and description text */
    SAY 'KRC2 Failed   (rc='myrc' rs='myrs')',
    RETURN  myrc myrs 
RETURN   myrc myrs 

and the printable reason code

/* exec to give back reason code string from passed value */ 
parse arg rc,rs 

v.= "Not listed" rs 
v.762="The key values structure for CSNDPKB has a field in error"||, 
            "A length or format is not correct" 
v.2012="The rule_array_count parameter contains a" ||, 
           " number that is not valid." 
v.2016="Rule Array wrong" 
v.2040="Wrong key type.   For example PKI when Importer was expected" 
v.2054="RSA:OAEP optional encoding parameters failed validation" 
v.2089="The algorithm does not match the algorithm"||, 
           " of the key identifier" 
v.10012="Key not found" 


I converted from a string to a hex representation of the string when passing data around because the hex data could have a blanks in it. Using the Rexx parse var x a b c parses on blank delimited words, and imbedded blanks could cause a mis-parse.

Understanding ICSF fixed and variable format keys and how not to get confused when using them.

This is part of the project to set up ICSF so I can create an encrypted dataset on one system, and use it on a different system. This involves setting up keys for encryption, public/private keys and sending stuff between the two systems.

Dataset encryption keys are symmetric and reside in the CKDS.

It took me a while to understand the implications of the fixed and variable format keys, and lots of head scratching when things did not work as expected.

This post is a very simplistic view of the topic – but it should give you enough information to get started with ICSF.

One thing you need to know about (and then quickly forget) is that there are two format of keys.

  1. Fixed length – an example of this is a DATA key.
  2. Variable length – an example of this is is a CIPHER key.

I just think of there being two versions of APIs – Version 1 for Data and version 2 for Cipher. Any new project should use variable length keys.

You can use either DATA or CIPHER to encrypt a dataset.

Each type of key has its own APIs, so you cannot use a fixed key in an API designed for variable length keys.

Most ICSF APIs have “rules” which are like passing parameters to a command. This is an array of 8 character strings such as “AES “,”PKCS-1.2”,OP “. This string indicates

  • AES – encrypt an AES keyType
  • PKCS-1.2 using PKCS
  • OP “Operational” for use on this system.

When using the APIs’ to export keys, both ends must have matching configuration

For example with a DATA key any one of the following.

  1. AES,PKCS-12

I think the last set of parameters is the strongest.

For CIPHER, this worked

  1. AES, PKOAEP2, SHA-256

When programming I found it easier to create some helper routines to reduced the complexity of the APIs. For example I created a ADDKEY routine to specify PKDS|CKDS, the key name, and the buffer contents. As a result I had a high amount of reuse, and my main programs were very compact.

Below I give the API calls for

  • Using variable length Cipher keys
    • create the skeleton using CSNBKTB2
    • create the data key using CSNBKGN2
    • export the data key using CSNDSYX
    • import the data key (on another system) using CSNDSYI2
  • Using fixed length data keys
    • create the data key using CSNDSYG
    • export the data key using CSNDSYX
    • import the data key (on another system) using CSNDSYI

Using variable length Cipher keys

I see the variable length keys as an evolution in key management from the fixed length keys.

For example you can store the name of the key within the key (though I do not know when or how this is used).

Create the skeleton

I built a skeleton using CSNBKTB2. You can optionally pass in

  • a key name
  • user data

I did not set these (I set the lengths to zero (&zero)).

I passed in the rule

  • char 8 rule[4] = “INTERNAL”,”AES “,”CIPHER “, “ANY-MODE”

The ANY-MODE was required for the encryption to work.

Create the cipher key

I then created the key using CSNBKGN2 and passed in

  • char8 rule[2] = {“AES “,”OP “};
  • keyLength = 256 ; why use a weaker key?
  • char8 keyType1 = {“TOKEN “}; This says use the data passed in from the skeleton token.
  • char8 keyType2 = {” “};
  • The skeleton data

This returns a block of data (the AES Cipher key encrypted with the local key).

I added it to the local keystore using CSNBKRC2.

Export the cipher key

I exported the key using a public certificate with CSNDSYX. I passed

  • the name of the public key.
  • char8 rule_array[3] = { “AES “,”PKOAEP2 “,”SHA-256 “}

It returns a block of data containing the AES cipher key encrypted with the public key.

I wrote this data to a file (in binary) and sent it to my remote system.

Import the cipher key at the remote system

I read the file into a buffer, and used CSNDSYI2 to decrypt the contents using the private key, and encrypting it with the local key.

I passed

  • char8 rule_array[2] ={“AES “,”PKOAEP2 “}
  • the name of the private key
  • the buffer

it returned a buffer containing the re-encrypted key

I added it to the keystore on the remote system using CSNBKRC2.


If you want to use fixed length keys(why do you?)…

Using fixed length data keys

You can create a data key, and export it at a later date, or you can generate it and export it at the same time.

I feel more comfortable about exporting it when it is needed, in case the “old” copy is out of data.

If you export it from the live system you know it is current.

Create the data key

I used use CSNDSYG


  • a char 64 key name
  • char8 rule[3] = {“AES “, “PKCS-1.2″| PKCSOAEP, “OP “}; If you use PKCSOAEP you can specify SHA-512

This creates a data key in the local repository.

I think PKCSOAEP is better than PKCS-1.2 as is is more recent.

You can also get it to create an RSA enciphered version of the key using a public certificate (as part of key creation) This can be written to a file, and the file sent to the remote system. I tend not to use this, but export the key at the time when wanted. This way you can be sure you have the correct key.

Export the data key

Use CSNDSYX with similar parameters as for the create (CSNDSYG) to create an encrypted version of the Data key.

CSNDSYX returns a block of storage with the encrypted key in it. You can write this to a file in binary or create a base64 encoding of it.

Import the data key

Use CSNDSYI with

  • The block of encrypted data (which you can read from a file)
  • rule = “AES “, “PKCS-1.2″|| “PKCSOAEP”. If you use PKCSOAEP you can specify SHA-512
  • The parameters must match the sending end
  • The name of the private key in the PKDS to be used to decrypt the data

This returns a block of data which is the data key, encrypted with the local system’s key.

You can then add it to the CKDS using CSNBKRC2

ICSF: My first C application.

As part of setting up data set encryption to send data sets between two sites. I wrote some programs to help securely send the symmetric key exchange between two systems using private and public keys.

Basis flow to get a symmetric key on two systems to allow data set encryption and decryption, using private key and public key.

The steps to allow me to send an encrypted dataset to a remote system, using PKI are as follows

  1. On the remote system generate a private/public key in the PKDS
  2. Extract the public key and send it to the local system
  3. Import the public key into the local PKDS
  4. On the local system, generate a symmetric AES key for encrypting data sets.
  5. Export the key, encrypt it using the public key, into a buffer.
  6. Write the buffer to a file
  7. Send the file to the remote system.
  8. Read the file, decrypt the buffer using the private key, re-encrypt with the local master key, store it in the local CKDS.
  9. Displayed the key at each end and compared the
  10. Use the symmetric key to decrypt a dataset.

The programs

I wrote some helper programs for example read a key from the CKDS returning the data; create an entry in the PKDS with a specified key and a blob of data containing the token.

Create a private public key

I could not find a batch utility to define a private key, but could use the ISPF panels.

I started writing my own program to do this…

I used API function CSNDPKB. I successfully created a Private/Public key using Elliptic Curve (ECC).

I could display this using the ISPF panels 5 UTILITY->6 PKDS KEYS → 1. This lists all of the keys. Use the line command K to display it. ( D is for delete!). It displays attributes like Algorithm: ECC, size 512, Sections: PRIVATE PUBLIC.

I later found out that ECC cannot be used to encrypt a symmetric AESkey. “Use the Symmetric Key Export callable service to transfer an application-supplied AES, DES or variable length symmetric key token key from encryption under a master key to encryption under an application supplied RSA public key or AES EXPORTER key. I could not find how to generate an EXPORTER key with public certificate.

I tried using CSNDPKB to create an RSA private key. You have to complete a “Key Value Structure”. I struggled to complete this for RSA because I needed to specify “prime number p”,and “prime number q”. Ive since found how to do this. I tried an alternate approach of using the ISPF panels.

On the remote system, I used the ISPF panels to create a PKI key. 5 UTILITY → 6 PKDS KEYS → 6 Generate PKA keys. I created a new RSA key with bit length 4096, as bit length 512 was too weak.

Having created my private/public key, I used the ISPF panels to export the public key to a data set, sent it to my local system and import it using the ISPF panels. This public key does not need to be sent securely.

  • You can extract the contents of this file using RACDCERT ADD(‘COLIN.EXPORT.ECC1’) ID(COLIN). This gave me a key with
Certificate ID: 2QXD1tPJ1dPBwsXT8PDw8PDw8PFA
Status: TRUST
Start Date: 2021/08/29 09:42:44
End Date: 2041/08/29 09:42:44
Serial Number:
Issuer's Name: >CN=ECCB512Z< 
Subject's Name: >CN=ECCB512Z<
Signing Algorithm: sha256ECDSA
Key Type: Brainpool ECC
Key Size: 512
Private Key: NO
Ring Associations:

I had now had a private key at the remote end, and the public certificate at the local end.

I can encrypt at the local end using the public key, and decrypt it at the remote end using the private key.

Create the AES symmetric key on the local end.

I found I could not use AES type DATA, and so had to use AES type CIPHER.

If I used KGUP utility with


This could not be used to encrypt because it gave IEC143I 213-85 RC=X’00000008′, RSN=X’0000085E’

85E (2142) The key usage attributes of the variable-length key token does not allow the requested operation. For example, the request might have been to encrypt data, but encryption is not allowed, or the request might have been to use the ECB cipher mode, but that mode is not allowed.

When I displayed through the ISPF panels (5.5.1) it it showed


If I generated an AES key using the API, I found that ENCRYPT DECRYPT ANY-MODE worked, but I could not see how to set ANY-MODE using KGUP.

Program to generate an AES key

 // build the skeleton.   It is returned in pToken
 rc = skeletonAES(&pToken,& lToken); 
 if ( rc != 0 ) return rc; 
 // input: the skeleton 
 // output: the token 
 rc = GENAES2(pToken,&lToken); 
 if ( rc != 0 ) return rc; 

 // add it to the CKDS                                                           
 rc = doAdd2("C",pKey,pToken,lToken,pDelete); 
 if ( rc != 0 ) return rc; 
 return rc; 

skeletonAES program

// this program allocates storage, uses it, and passes it back
// to the caller. 
int skeletonAES(char ** pData, int * lData) 
  int rc; 
  int rs; 
  int zero = 0; 
  int rule_count                 =  4 ; 
  char8 rule[ 4] ={ 
   "INTERNAL","AES     ","CIPHER  ",    "ANY-MODE" 
  char * pRule = & rule[0][0];
  int lKeyToken=725; 
  char * pKeyToken; 
  pKeyToken = (char *) malloc(lKeyToken); 
           &rc,         /* return code             */ 
           &rs,         /* reason code             */ 
           &zero,       /* exit data length        */ 
           0,           /* exit data[]             */ 
           &rule_count, /* rule array count        */ 
           pRule,       /* rule array[]            */ 
           &zero,       /* clear key bit length    */ 
           0,           /* key value[]             */ 
           &zero,       /* key name length         */ 
           0,           /* key name[64]            */ 
           &zero,       /* user assoc data length  */ 
           0,           /* user associated data[]  */ 
           &zero,       /* token data length       */ 
           0,           /* token data[]            */ 
           &zero,       /* service data length     */ 
           0,           /* service data[]          */ 
           &lKeyToken,  /* target key token length */ 
           pKeyToken ); /* target key token[]      */ 
 if ( rc > 0) 
    printf("CSNBKTB2 rc %i rs %i %s\n",rc,rs,csfgetrc(rs)); 
    return rc; 
   free (pData); // get rid of the passed in block 
// return the length
 *lData = lKeyToken;
//pass back the address of the token
 *pData = pKeyToken; 
 printf("skeletonAES key length %i\n",lKeyToken); 
// printHex(stdout,&keyToken,lKeyToken); 
   printAES((char *) pKeyToken,lKeyToken); 
   return 0; 

Create program keygenerate

// this takes the data passed in, and uses it
// The length is updated.
int keyGenerate2( char * pData, int *  lData) 
  int rc; 
  int rs; 
  int zero = 0; 
  int rule_count                 =  2; 
  // key type AES and used for OP (on this system 
  char8 rule[2]  = {"AES     ","OP      "}; 
  char * pRule = &rule[0][0];
  int keyLength = 256 ;//  AES 256  - why use any other? 
  char8  keyType1 = {"TOKEN   "}; 
  char8  keyType2 = {"        "}; 
  int i64 = 64; 
  int l725 = 725; 

         &rc  ,        /* return code             */ 
         &rs  ,        /* reason code             */ 
         &zero,        /* exit data length        */ 
         0,            /* exit data[]             */ 
         &rule_count,  /* rule array count        */ 
         pRule,        /* rule array[]            */ 
         &keyLength,   /* clear key bit length    */ 
(char *) &keyType1   , /* key type1[8]            */ 
(char *) &keyType2   , /* key type2[8]            */ 
         &zero,        /* key name1 length        */ 
         0,            /* key name1[64]           */ 
         &zero,        /* key name2 length        */ 
         0,            /* key name2[64]           */ 
         &zero,        /* user assoc data1 length */ 
         0,            /* user associated data1[] */ 
         &zero,        /* user assoc data2 length */ 
         0,            /* user associated data2[] */ 
         &zero,        /* KEK id1 length          */ 
         0,            /* KEK id1[]               */ 
         &zero,        /* KEK id2 length          */ 
         0,            /* KEK id2[]               */ 
         &l725,        /* output key id1 length   */ 
 (char *)pData,        /* output key id1[]        */ 
         &zero    ,    /* output key id2 length   */ 
         0          ); /* output key id2[]        */ 
  if ( rc > 0) 
    printf("CSNBKGN2 rc %i rs %i %s\n",rc,rs,csfgetrc(rs)); 
    return rc; 
  *lData = l725; 
  // the same buffer is used,so no need to set pData
  return rc; 

Using these and my add2(…) functions I could create my AES for CIPHER with ANY-MODE. (See ANY-MODE above).

Export the AES key

To export the AES key, you need the name of the public key.

You need a different set of rules for a DATA key and a CIPHER key.

 int   lData; 
 char * pData; 
 rc =doExportAES (pKey,pPublic,pType,&pData, &lData); 
 if (rc > 0 ) return rc; 
 rc = writeKey(dd,pData,lData); 
 if (rc > 0 ) return rc; 

I created a routine which read the key, and one to write it to a file.

The parameters to doExportAES are

  • dd name “dd:CERT would use //CERT .. in the JCL
  • key is a char64 left justified key name in the CKDS
  • pPublic a char 64 left justified name of the public key to use. It must exit in the PKDS.
  • A pointer to the string to containing the data
  • The length of the returned data.
int doExportAES (char * pKey, char * pPublic, char * pType, 
                 char ** pData, int * lData)  
  int rc; 
  int rs; 
  int i64 = 64; 
  int zero = 0; 
  int rule_count                 =  3; 
  char8 rule_Cipher[3] = 
                     {"AES     ", 
                      "PKOAEP2 ", 
                      "SHA-256 "}; 
  char * pRule =   rule_Cipher[0][0];
  int lOut = 900; 
  char * pOut; 
  pOut = (char * ) malloc(lOut); 
  printf("Source Key user   %64.64s.\n",key  ); 
  printf("Transport key     %64.64s.\n",encrypt); 

            &rc,                /* return code            */ 
            &rs ,               /* reason code            */ 
            &zero,              /* exit data length       */ 
            0    ,              /* exit data[]            */ 
            &rule_count,        /* rule array count       */ 
            pRule,              /* rule array[]           */ 
            &i64   ,            /* source key    length,  */ 
            pKey  ,             /* source key             */ 
            &i64,               /* RSA public key length  */ 
            pPublic,            /* RSA public key token   */ 
            &lOut,              /* exported key length    */ 
            pOut           );   /* exported key           */ 

if( rc > 0) 
  printf("CSNDSYX  rc %i rs %i %s\n",rc,rs,csfgetrc(rs)); 
if (rc != 0) return rc; 

// return the exported data

 *pData = pOut; 
* lData = lOut; 
   return 0; 

Import the certificate at the remote end

I created IMPAES

  rc=  read(dd,&pData,&lData);         // returns a buffer  from the file
  printf("readCert rc %i data length %i\n",rc,lData ); 
  if ( rc != 0) return 8; 

  rc = doImportAES(pKey,pPrivateKey,&pData,&lData); 
  if ( rc != 0) return 8; 
  printf("Returned buffer size %i\n",lData); 
  rc = doAdd2("C",  pKey,pData,lData); 
  if ( rc != 0 ) return rc; 
  return 0; 

read opens the ddname , and gets back the (binary) data in pData of length lData.

This get passed to doImportAES which uses the private key pointed to by pPrivateKey (a 64 char left justified string). This then returns the AES key (encrypted with the local key) which is then added using do ADD2.


/* --------------------------------------------------------------- *
/* Import public key                                               *
/* --------------------------------------------------------------- *
int doImportAES (char * pKey, char * pDecryptKey, 
    char ** pData, int * lData) 
  int rc; 
  int rs; 
  int zero = 0; 
  int i64 = 64; 
  int rule_count                 =  2; 
  char8 rule_array[2]            ; 
  memcpy(&rule_array[0],"AES     ",8); 
  memcpy(&rule_array[1],"PKOAEP2 ",8); 
  int lOutput = 725 ; 
  char * pOutput = (char * ) malloc(lOutput); 
  if ( pOutput == 0) 
    printf("malloc for %i failed in IMPAES\n",lOutput); 
  printf("Key %64.64s.\n",pKey); 
  printf("Decrypt Key %64.64s.\n",pDecryptKey); 
  printf("lInput  %i\n",*lData     ); 
//printHex(stdout,* pData,32); 
              &rc,           /* return code            */ 
              &rs ,          /* reason code            */ 
              &zer           /* exit data length       */ 
              0    ,         /* exit data[]            */ 
              &rule_count,   /* rule array count       */ 
   (char *)   &rule_array[0],/* rule array[]           */ 
              lData     ,    /* length of input data   */ 
            * pData,         /* input data             */ 
              &i64,          /* cca DES token length,  */ 
   (char *)   pDecryptKey ,  /* cca DES token[64]      */ 
              &i64,          /* key name length        */ 
   (char *)   pKey,          /* key name[]             */ 
              &lOutput    ,  /* Imported key length    */ 
   (char *)   pOutput   );   /* Imported key           */ 
// if( rc > 0) 
   printf("CSNDSYI2 rc %i rs %i %s\n",rc,rs,csfgetrc(rs)); 
 if (rc != 0) return rc; 

 // get rid of input buffer 
   printf("CSNDSYI2 data lengt %i\n",lOutput); 
   free (*pData); 
// and update with the new data
   *pData =  pOutput; 
   *lData = lOutput; 
 return 0; 

All this creates a key in the CKDS on the remote system which I could use to decrypt a data set.


It took a long time to write these programs, because I did not know the path to take, and went down many dead ends. Once you know the concepts and know which ICSF functions and options you need, it is not too difficult.

Having these helper routines, I was able to create a program to generate a symmetric key using Diffi-Hellman in about 2 hours! (Most of this time was reading the documentation).