Front cover

EFS Systems on a Linux

Base: Gettingg Started

Planning for ThinkPad and Netfinity

bases

Installation, customization,
operation

Common techniques

Bill Ogden

ibm.com/redbooks
International Technical Support Organization

EFS Systems on a Linux Base: Getting Started

November 2003

SG24-7007-01
 Note: Before using this information and the product it supports, read the information in “Notices” on
page vii.

Second Edition (November 2003)

This edition is based on FLEX-ES Release 6.2.14, Red Hat Linux Release 9.0, and the z/OS Application
Development package for z/OS 1.4s.

© Copyright International Business Machines Corporation 2003. All rights reserved.

Note to U.S. Government Users Restricted Rights -- Use, duplication or disclosure restricted by GSA ADP Schedule
Contract with IBM Corp.
Contents

Notices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Trademarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
The team that wrote this redbook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Become a published author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Comments welcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .x

Chapter 1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Purpose of this redbook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Connectivity overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Basic EFS requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4.1 System selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 EFS limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Detailed FLEX-ES overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Chapter 2. Practical planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.1 General planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.1 Production or sandbox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.2 Selecting a Linux distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.3 Newer hardware and Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.4 Disk space needed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.5 Linux file system layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.6 Disk modes for FLEX-ES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.7 Network connections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 ThinkPad topics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.1 T40 ThinkPads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 BIOS settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.3 Device planning (Ultrabay usage) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.4 Choices for disk layouts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.5 Dual boot, alternate boot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Larger server topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.1 SMP, large memory, new LAN adapters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.2 BIOS and ServeRAID adapter updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.3 RAID configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Dual boot preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Chapter 3. Linux installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.1 Basic Red Hat Linux 9.0 installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Installation notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Kernel rebuild . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Additional kernel steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4.1 ServeRAID updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4.2 LVM updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5 Additional Linux steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.5.1 Red Hat Linux updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Chapter 4. FLEX-ES installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

© Copyright IBM Corp. 2003. All rights reserved. iii

 4.1 FLEX-ES installation steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.1.1 Default PATH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.1.2 Ownership of /s390 mount point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.1.3 Location of x3270 program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Installing the FLEX-ES license key and dongle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3 Reboot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Installing FLEX-ES upgrades . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

Chapter 5. S/390 operating system installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

5.1 AD CD-ROM distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.1.1 z/OS AD CD-ROM addresses and IPL data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.1.2 Installation tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.2 Dump/restore distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.3 The tar file distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.4 OMA/2 distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.5 File ownership (any distribution) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

Chapter 6. Operational details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

6.1 Linux desktop windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.2 FLEX-ES system and resource definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.2.1 Building a shell script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.2.2 IPL z/OS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.2.3 User terminal connection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.2.4 IODF requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.2.5 System performance monitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2.6 TCP/IP for z/OS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.3 Shutting down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.4 More comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Chapter 7. Basic discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

7.1 Resource definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.2 Linux windows and FLEX-ES operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.3 64-bit operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.4 Creating additional emulated volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.5 Operating systems messages console . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.6 x3270 client . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.7 S/390 identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.8 z/OS TCP/IP profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.9 Initial debugging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.10 VTAM APAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
7.11 Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.12 RAS discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.13 Useful Linux commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

Chapter 8. FAQs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

Appendix A. Sample FLEX-ES operation files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

Definitions for the Dallas DEMOpkg. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Minimal AD system definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Full AD system definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Customized Offerings Driver definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
z/VM OMA/2 definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Definition statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

iv EFS Systems on a Linux Base: Getting Started

Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Other publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Online resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
How to get IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Help from IBM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

Contents v
vi EFS Systems on a Linux Base: Getting Started
Notices

This information was developed for products and services offered in the U.S.A.

IBM may not offer the products, services, or features discussed in this document in other countries. Consult
your local IBM representative for information on the products and services currently available in your area. Any
reference to an IBM product, program, or service is not intended to state or imply that only that IBM product,
program, or service may be used. Any functionally equivalent product, program, or service that does not
infringe any IBM intellectual property right may be used instead. However, it is the user's responsibility to
evaluate and verify the operation of any non-IBM product, program, or service.

IBM may have patents or pending patent applications covering subject matter described in this document. The
furnishing of this document does not give you any license to these patents. You can send license inquiries, in
writing, to:
IBM Director of Licensing, IBM Corporation, North Castle Drive, Armonk, NY 10504-1785 U.S.A.

The following paragraph does not apply to the United Kingdom or any other country where such provisions are
inconsistent with local law: INTERNATIONAL BUSINESS MACHINES CORPORATION PROVIDES THIS
PUBLICATION "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED,
INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF NON-INFRINGEMENT,
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Some states do not allow disclaimer of
express or implied warranties in certain transactions, therefore, this statement may not apply to you.

This information could include technical inaccuracies or typographical errors. Changes are periodically made
to the information herein; these changes will be incorporated in new editions of the publication. IBM may make
improvements and/or changes in the product(s) and/or the program(s) described in this publication at any time
without notice.

Any references in this information to non-IBM Web sites are provided for convenience only and do not in any
manner serve as an endorsement of those Web sites. The materials at those Web sites are not part of the
materials for this IBM product and use of those Web sites is at your own risk.

IBM may use or distribute any of the information you supply in any way it believes appropriate without incurring
any obligation to you.

Information concerning non-IBM products was obtained from the suppliers of those products, their published
announcements or other publicly available sources. IBM has not tested those products and cannot confirm the
accuracy of performance, compatibility or any other claims related to non-IBM products. Questions on the
capabilities of non-IBM products should be addressed to the suppliers of those products.

This information contains examples of data and reports used in daily business operations. To illustrate them
as completely as possible, the examples include the names of individuals, companies, brands, and products.
All of these names are fictitious and any similarity to the names and addresses used by an actual business
enterprise is entirely coincidental.

COPYRIGHT LICENSE:
This information contains sample application programs in source language, which illustrate programming
techniques on various operating platforms. You may copy, modify, and distribute these sample programs in
any form without payment to IBM, for the purposes of developing, using, marketing or distributing application
programs conforming to the application programming interface for the operating platform for which the sample
programs are written. These examples have not been thoroughly tested under all conditions. IBM, therefore,
cannot guarantee or imply reliability, serviceability, or function of these programs. You may copy, modify, and
distribute these sample programs in any form without payment to IBM for the purposes of developing, using,
marketing, or distributing application programs conforming to IBM's application programming interfaces.

© Copyright IBM Corp. 2003. All rights reserved. vii

Trademarks
The following terms are trademarks of the International Business Machines Corporation in the United States,
other countries, or both:
Eserver® IBM® S/370™
ibm.com® IMS™ S/390®
xSeries® Multiprise® ServerGuide™
z/OS® MVS™ ServeRAID™
z/VM® Netfinity® ThinkPad®
zSeries® OS/2® Ultrabay™
CICS® OS/390® VM/ESA®
DB2® Parallel Sysplex® VSE/ESA™
DFSMSdss™ PartnerWorld® VTAM®
ESCON® Redbooks(logo) ™ WebSphere®
FICON™ Redbooks™
HiperSockets™ RMF™

The following terms are trademarks of other companies:

Intel, Intel Inside (logos), MMX and Pentium are trademarks of Intel Corporation in the United States, other
countries, or both.

Microsoft, Windows, Windows NT, and the Windows logo are trademarks of Microsoft Corporation in the
United States, other countries, or both.

Java and all Java-based trademarks and logos are trademarks or registered trademarks of Sun Microsystems,
Inc. in the United States, other countries, or both.

C-bus is a trademark of Corollary, Inc. in the United States, other countries, or both.

UNIX is a registered trademark of The Open Group in the United States and other countries.

SET, SET Secure Electronic Transaction, and the SET Logo are trademarks owned by SET Secure Electronic
Transaction LLC.

Other company, product, and service names may be trademarks or service marks of others.

viii EFS Systems on a Linux Base: Getting Started

Preface

This IBM® Redbook describes the basic installation of FLEX-ES (a product of Fundamental
Software, Inc., Fremont, California) with several z/OS® operating system packages. Much of
the material is in cookbook format. More advanced techniques and options are described in
S/390 PartnerWorld for Developers, ITSO/EFS Project EFS Systems on a Linux Base:
Additional Topics, SG24-7008. These books are intended primarily for members of the IBM
PartnerWorld® for Development organization and for internal IBM users of the ITSO/EFS
package.

This second edition includes the latest updates to the first edition of this book. It also contains
more information about using Red Hat 9.0 with FLEX-ES Release 6.2.14.

The team that wrote this redbook

Bill Ogden is a retired IBM Senior Technical Staff Member, still working on favorite projects
with the International Technical Support Organization, Poughkeepsie Center. He specializes
in small S/390® and z/OS systems, writes extensively, and teaches ITSO workshops relating
to these areas. He is the originator of the ITSO/EFS package for internal IBM use. Bill has
been with the ITSO since 1978. He completed this update for this book while working at St.
Michael’s Indian School, on the Navajo Reservation in Arizona.

We thank the following people from Fundamental Software, Inc. for their contributions to this
project:

Gary Eheman, who helped us understand many additional aspects of FLEX-ES and Linux,
and who was especially helpful with the updates in this book.

Michael Ryan, who helped with T40 ThinkPad®, performance, and upgrade questions.

Scott Carter, who is an eagle-eyed proofreader, corrector of errors, and a source of

knowledge for obscure FLEX-ES details.

Become a published author

Join us for a two- to six-week residency program! Help write an IBM Redbook dealing with
specific products or solutions, while getting hands-on experience with leading-edge
technologies. You'll team with IBM technical professionals, Business Partners and/or
customers.

Your efforts will help increase product acceptance and customer satisfaction. As a bonus,
you'll develop a network of contacts in IBM development labs, and increase your productivity
and marketability.

Find out more about the residency program, browse the residency index, and apply online at:
ibm.com/redbooks/residencies.html

© Copyright IBM Corp. 2003. All rights reserved. ix

Comments welcome
Your comments are important to us!

We want our Redbooks™ to be as helpful as possible. Send us your comments about this or
other Redbooks in one of the following ways:
򐂰 Use the online Contact us review redbook form found at:
ibm.com/redbooks
򐂰 Send your comments in an Internet note to:
[email protected]
򐂰 Mail your comments to:
IBM Corporation, International Technical Support Organization
Dept. HYJ Mail Station P099
2455 South Road
Poughkeepsie, NY 12601-5400

x EFS Systems on a Linux Base: Getting Started

 1

Chapter 1. Introduction
EFS systems provide small S/390 platforms. The EFS name is an IBM term and is derived
from “Enabled For S/390.” EFS systems may be used to run z/OS (and OS/390®), z/VM®
(and VM/ESA®), VSE/ESA™, and Linux for S/390. EFS systems may also be used for older
versions of these operating systems. This document discusses z/OS use, but the reader
should not conclude that the EFS is limited to z/OS.

EFS systems use FLEX-ES to provide S/390 emulation. FLEX-ES is a licensed product of
Fundamental Software, Incorporated (FSI) of Fremont, California. A high-level view of an EFS
system is presented in Figure 1-1.

Typically, there is no other significant workload

in Linux while FLEX-ES is running.

The emulated S/390 resides in Linux virtual

memory belonging to processes started by
z/OS FLEX-ES.

S/390

FLEX-ES FLEX-ES is a number of processes in Linux.

Red Hat Linux Only Red Hat Linux is supported.

A variety of PCs (mobile computers, desktops,

Intel Processor(s) servers) might be used.

Figure 1-1 High-level view of an EFS system

FLEX-ES runs on an Intel-based PC under Linux or UnixWare operating systems. This

redbook discusses only the Linux-based version.

FLEX-ES is a licensed product. The licenses are normally arranged through FSI Business
Partners.1 A FLEX-ES license specifies the number of PC processors (and their maximum

© Copyright IBM Corp. 2003. All rights reserved. 1

 speed) that can be used for S/390 emulation. A FLEX-ES license for Linux is keyed to a
specific serial number in a hardware dongle that is connected to the USB port of the PC and
runs only in a system that has this dongle connected.

An EFS system is an S/390. S/390 software licenses must be obtained as appropriate, both
for IBM software and other vendor software. These software licenses are arranged through
IBM, Business Partners, or other vendors. This document does not further address these
license requirements and assumes that any EFS user has all the appropriate S/390 software
licenses.

In this document we use the term S/390 emulation when the term S/390 and/or IBM
Eserver zSeries® and/or S/370™ and/or other emulations based on S/360 designs might be
more accurate. FLEX-ES can emulate zSeries architecture at the Architecture Level Set 3
(ALS 3) level. It can also emulate S/390 and older architectures. For the purposes of
readability we refer to all of these, collectively, as S/390 emulation and the resulting EFS
system as a S/390.

EFS systems can serve many purposes. These include application development,
demonstrations, education, and production work. While IBM discourages the use of MIPS
measurements and comparisons, it may be helpful to note that typical EFS systems are
roughly in the 12 to 25 MIPS range. The actual performance is directly related to the speed of
the underlying PC, and the number of processors on the PC. A FLEX-ES license can limit the
S/390 MIPS produced (assuming the PC is capable of faster operation) in order to meet
S/390 software license limitations.

1.1 Purpose of this redbook

This IBM Redbook describes only Linux-based EFS systems and describes the installation
and initial startup process for these systems. The specific combination described in much of
this document, that is, EFS systems running z/OS AD CD-ROM software, is available to
members of the IBM PartnerWorld for Developers (PWD) organization2 who have obtained
their systems through the PWD program.

EFS systems for PWD members are available only through IBM Business Partners and are
not available directly from IBM. In normal situations, the Business Partner performs the initial
machine setup—including hardware installation, basic Linux operating system installation,
initial disk configuration, and FLEX-ES installation. This redbook describes these same steps,
and, in a sense, describes steps that an owner may not need to perform.

However, we believe this material will be useful for several reasons:

򐂰 Understanding how a system is set up, even though someone else may have done it for
you, leads to a better understanding of the system.
򐂰 The system owner may accidently (or intentionally) delete or destroy part of the underlying
elements of his system. If this happens, he might ask his IBM Business Partner for
assistance, or he might prefer to rebuild the system himself.
򐂰 Significant upgrades to the underlying elements (hardware, Linux, FLEX-ES) may require
a partial or complete rebuilding of the system. Again, the system owner has the option of
asking his Business Partner to perform the upgrade, or of doing it himself.

This redbook updates and replaces an earlier version, EFS Systems on a Linux Base,
SG24-6834.

1
You can go to http://www.funsoft.com to contact FSI or locate an FSI Business Partner.
2 This organization was previously known as S/390 Partners in Development (PID).

2 EFS Systems on a Linux Base: Getting Started

1.2 Terminology
EFS descriptions can become confusing if the terminology is not well defined. These are
important terms:
򐂰 Processor means a PC processor. ThinkPads have a single processor, while some PC
servers have multiple processors.
򐂰 Processor(s) enabled for S/390 refers to the number of PC processors that the FLEX-ES
license allows to be used simultaneously for S/390 emulation.
򐂰 Server means the underlying PC hardware.
򐂰 Server operating system means Linux in the PC.
򐂰 S/390 CPU (or simply CPU) means an emulated single S/390 CPU engine.
򐂰 S/390 system means an emulated S/390 (in the EFS system) that might have more than
one emulated S/390 CPU engine.
򐂰 Instance (or FLEX-ES instance) means an emulated S/390 system. We can run several
FLEX-ES instances if we have sufficient PC resources.
򐂰 z/OS means the S/390 operating system. We could also use OS/390, z/VM, VM/ESA,
Linux for S/390, or VSE/ESA, but we elected to work only with z/OS.
򐂰 ThinkPad EFS and xSeries® EFS are generic names. Business Partners providing these
products often have their own names for their specific implementation of the products.
򐂰 IBM ~ xSeries is the name of an IBM product line. Older products used the name
Netfinity®.
򐂰 S/390, in this document only, is used to include both S/390 and zSeries functionality. This
abbreviated reference is solely for readability.

For any FLEX-ES instance, there is usually a one-to-one correspondence between

processors enabled for S/390 and S/390 CPUs. Running multiple instances is something like
running multiple z/OSs in separate logical partitions (LPARs), but it is not quite the same. We
can run multiple instances of single-CPU S/390s in a ThinkPad EFS system, for example.

Note: Throughout this book, we frequently refer to the IBM Redbook S/390 PartnerWorld
for Developers, ITSO/EFS Project EFS Systems on a Linux Base: Additional Topics,
SG24-7008.

1.3 Connectivity overview

An EFS system can be self-contained in a PC. That is, the system can provide the required
S/390 emulation plus several 3270 terminal sessions for a z/OS operator console and VTAM®
(TSO, CICS®, and so forth) terminals. It can also provide ASCII Telnet sessions for direct
logon to z/OS UNIX System Services. The number of terminal sessions (3270 and ASCII
Telnet) is limited by the screen space available for reasonable use on the PC display. For
practical purposes, this is a single-user system.

We can easily make a multi-user system by connecting the system to a LAN. This typically is
done with a single Ethernet interface. The single LAN connection, as illustrated in Figure 1-2
on page 4, can be used in a number of ways at the same time:
򐂰 Client workstations, using 3270 emulators, can connect to a TN3270e server running
under Linux in the EFS system. This server is the FLEX-ES Terminal Solicitor. It
transforms client TN3270e connections to appear as local, channel-attached, non-SNA
DFT 3270 terminals that can be used as z/OS operator consoles or VTAM (TSO, CICS)

Chapter 1. Introduction 3
 terminals. z/OS TCP/IP is not involved in these functions and need not be running. These
local 3270 terminals are the most basic z/OS terminal and z/OS is unaware they are
actually remote TN3270e terminals.
򐂰 Client workstations, using 3270 emulators, can connect to z/OS TCP/IP (assuming it is
configured and running). This interface can be used for normal VTAM purposes (such as
TSO) but not for traditional operator consoles.
򐂰 Other z/OS TCP/IP functions, such as FTP, Telnet, a Web server, and so forth, function
normally through this interface. If UNIX System Services Telnet is active, a remote user
could connect to this interface.

An important point is that all these connections can operate simultaneously through a single
physical LAN interface to the EFS system. If used as described, this interface has two IP
addresses: one for Linux and one for z/OS.

z/OS running Windows Windows

PCOM PCOM
under FLEX-ES
ThinkPad Desktop

LAN
EFS System

Desktop ThinkPad

Linux OS/2
Two IP addresses: PCOM, Telnet, or both
x3270
one for Linux TCP/IP,
Telnet
one for z/OS TCP/IP

Figure 1-2 Multi-user System

With the configuration shown in Figure 1-2, a typical client might use a Windows-based
mobile computer connected to the LAN. PCOM3 can provide 3270 sessions that appear as
locally-connected, non-SNA sessions to z/OS (if connected through the Terminal Solicitor) or
as z/OS TCP/IP TN3270 connections. The distinction is made by which IP address and port
number the client uses for the EFS connection. These sessions can be used for z/OS
operator consoles or VTAM applications such as TSO, CICS, and so forth. The Telnet
application provided by Windows can be used to connect to Linux on the ThinkPad EFS
system (if such access is enabled) or to z/OS UNIX System Services (if this access is
enabled). If properly enabled, the client systems can FTP to Linux or z/OS, connect to a Web
server under z/OS, and so forth.

The EFS host system in this illustration could be a ThinkPad, a desktop PC, or a large xSeries
server.

1.4 Basic EFS requirements

Key points in understanding hardware requirements are:
򐂰 No unique hardware is used for z/Series emulation. There is no “S/390 adapter.”
Everything is software, with these exceptions:

3
PCOM is the common name for the IBM eNetwork Personal Communications product. It provides TN3270e client
functions.

4 EFS Systems on a Linux Base: Getting Started

 – A dongle is required. This is a device (match-box size) with a cable that connects to
the USB port of your PC. This provides license control for the FLEX-ES software. You
must have a software license key file that matches the encoded serial number in your
dongle.
– Optional S/390 channel adapters are available, with the limitations mentioned below.
򐂰 The base operating system of the hardware platform is Linux. In particular, it is Red Hat
Linux at selected release levels. Descriptions in this redbook assume the use of Red Hat
Linux 9.0. Only Red Hat Linux is supported.
򐂰 A dual-boot system, probably with a version of Microsoft Windows as the alternate
operating system is possible. Microsoft Windows is not otherwise relevant to EFS
operation.
򐂰 Sufficient memory must be present. An extremely rough guideline is 256 MB plus the size
of the zSeries memory you want to emulate. For example, using a 512 MB PC, you could
emulate a zSeries machine with 256 MB of real memory. A PC with 1 GB memory could
safely emulate a S/390 with 768 MB memory.
򐂰 Sufficient disk space is required. We recommend about 6 GB for Linux, although the
actual requirements can considerably less. The remaining disk space required is whatever
is required for your S/390 volumes. A single 3390-3 requires 2.8 GB, for example. There is
a one-to-one correspondence between the size of a 3390 volume and the amount of PC
disk space required.
򐂰 While not absolutely required, you probably want to connect your system to a LAN. With a
Linux-based EFS system, using Ethernet, only a single LAN adapter is required. It can be
used by both Linux and z/OS (and each has a separate IP address).
򐂰 SCSI-connected tape drives may be used. These are commonly used with larger
server-based EFS systems. We recommend a separate SCSI adapter for these instead of
a SCSI adapter that is part of a RAID adapter.
򐂰 Larger servers (but not ThinkPads) can use unique hardware adapters from Fundamental
Software, Inc. At the time of writing, these are not supported under Linux; they are
supported only for the UnixWare-based FLEX-ES systems. We expect these adapters to
be supported for Linux-based FLEX-ES sometime in the future. These adapters include:
– S/390 parallel channel adapters. Two versions are available. One has a single parallel
channel, and the other has three parallel channels.
– ESCON® channel adapters. Two adapter versions will be available: a single ESCON
channel and four ESCON channels.4
– ICA (integrated communications adapter) cards, providing six communications
interfaces. These can run in bisync or SDLC modes. Except for a few special
instances, z/OS does not support these interfaces; they are often used by VSE/ESA
systems.
򐂰 If a PC server is involved, we recommend a ServeRAID™-4Mx (or later) Ultra160 SCSI
Controller as the RAID adapter.

These considerations allow a range of systems to be considered. This list does not address
the speed of a PC to be used for EFS. We must note that the relative clock speeds of two PCs
(often expressed in Gigahertz) does not necessarily indicate the performance ratio between
the two systems. Memory bandwidth, bus speed, DMA operation, disk performance, and
cache operation have major effects on performance.

Assuming you want to run z/OS, we suggest a system with 1 GB memory, at least 1 GHz
clock speed, and 40 GB disk space as a reasonable starting point. Smaller and slower
4 We use the future tense here because the ESCON adapters were not yet available at the time of writing.

Chapter 1. Introduction 5
 systems also work. Larger and faster systems may be needed for specific applications or
workloads.

1.4.1 System selection

A ThinkPad EFS system is appropriate for many tasks, such as:
򐂰 Education, for a single owner or as a server running a classroom of clients.
򐂰 Small development and testing, typically for a single person. Proof-of-concept work is
often in this category.
򐂰 Demonstrations, in many different ways.

In our opinion, a ThinkPad EFS system is not suitable as a production server. Key factors
are:
򐂰 Non-parity memory is used. This memory simply lacks the reliability of ECC memory.
򐂰 RAID protection is not available for the disks. Furthermore, reliable backup arrangements
(to tape, for example) are typically more difficult to provide in a ThinkPad environment.
򐂰 ThinkPads are not fully designed for continuous operation. Heat dissipation is a concern,
although a docking station helps in this area.
򐂰 Disk operation is considerably slower than that of a fast RAID array.

A properly configured xSeries EFS system is much more robust and is suitable as a
production server. Important factors include:
򐂰 ECC memory (or equivalent) is used. The importance of this for long-term stable operation
cannot be overstated.
򐂰 RAID disks are used and assumed in all xSeries EFS discussions. RAID 5 is typically
used and provides automatic protection against any single disk failure. A hot spare is
always recommended and is considered normal in larger configurations.
򐂰 Multiple, redundant power supplies are available for the approved xSeries servers and
provide N+1 protection.
򐂰 Tape connections are more likely to be available in this environment, making routine
backups more practical.
򐂰 An xSeries EFS system normally has at least two processors. One might be enabled for
S/390 emulation, leaving the other for all the Linux processes involved with S/390 I/O
emulation, LAN operation, and so forth. Multiple processors can be enabled for S/390
emulation, making a much faster system.5

The presence of ECC memory, RAID disk protection, and N+1 power supplies covers the vast
majority of typical PC failures—providing a very stable hardware platform.

Each platform has its place. The outstanding characteristic of the ThinkPad is its portability, of
course. A secondary factor is the ease of installation of a ThinkPad EFS system (assuming
use of an Ethernet LAN).

5 This requires additional FLEX-ES license capabilities, with a corresponding increase in license fees.

6 EFS Systems on a Linux Base: Getting Started

1.5 EFS limitations
There are a number of functions not available for EFS systems. These include:
򐂰 Parallel Sysplex®, Coupling Facility processors, and Coupling Facility links are not
available.
򐂰 FICON™ channels are not available.
򐂰 Unique control unit hardware functions are generally not emulated, such as the cache and
compression functions of 3990-6 control units.
򐂰 Advanced zSeries (and S/390) hardware recovery functions, such as multiple I-units in
processors and the Application Preservation Feature, are not emulated.
򐂰 Cryptographic hardware adapters or coprocessors are not available.
򐂰 OSA-Express functions are not available.6
򐂰 HiperSockets™ are not available.
򐂰 LPARs are not available. However, if sufficient memory is available in the PC, we can
simultaneously emulate multiple z/Series machines. This is not quite the same as using
LPARs in a zSeries system, but it is similar in many ways. The multiple emulated systems
can communicate with each other over the LAN or by emulated CTC connections.
򐂰 The new and altered instructions in the zSeries z990 systems are not available at the time
of writing.
򐂰 Multiple paths to a device are not supported.

FLEX-ES has formal support only on specific xSeries models and specific ThinkPad models.
FSI adds to the list of supported models over time. This is discussed in more detail in the next
chapter.

1.6 Detailed FLEX-ES overview

This section provides a slightly more detailed description of FLEX-ES than was presented
earlier. FLEX-ES can be viewed as the following components:
򐂰 A S/390 instruction emulator, which is the heart of the system. It examines each S/390
operation code and emulates that operation, using the instructions of the underlying PC
processor.
򐂰 A resource manager that controls the interfaces between the emulated S/390 processors
and emulated I/O devices and connections.
򐂰 Emulators for various S/390 I/O devices.
򐂰 A FLEX-ES console for controlling FLEX-ES startup and operation. (This is not related to
the z/OS master console.)
򐂰 A Terminal Solicitor program that emulates local, channel-attached, non-SNA 3270
terminals. The actual terminals are TN3270e sessions that connect to this program
through normal TCP/IP protocols.
򐂰 A number of utility programs to help set up and run the FLEX-ES environment.

Figure 1-3 provides a simplified view of FLEX-ES operation. FLEX-ES is simply a process7
under Linux. While this illustration should not be taken too literally, it can be used to make a
number of basic points about FLEX-ES.
6
Basic OSA functions over Ethernet and token ring are available.
7 It is actually a number of interrelated processes and threads, but this detail can be ignored at the conceptual level.

Chapter 1. Introduction 7
 PC Memory

Emulated S/390
running z/OS
and its associated other Linux
applications processes

FLEX-ES
FLEX-ES processes Terminal Solicitor

TCP/IP
Linux operating system

PC Processor(s) in Server
Display and
Enet Adapt keyboard

PC disks

LAN

Figure 1-3 Conceptual view of FLEX-ES system

Key points include:

򐂰 FLEX-ES is a software product.8 All the hardware shown (in Figure 1-3) is standard PC
hardware. (The hardware dongle required by FLEX-ES is not shown in the figure.)
򐂰 The FLEX-ES program emulates a complete S/390 environment.
– FLEX-ES obtains sufficient Linux virtual memory to emulate the “real memory” for the
defined S/390 machine being emulated. While not detailed in the illustration, FLEX-ES
can also emulate S/390 expanded memory.
– FLEX-ES, as part of emulating a S/390, handles S/390 I/O instructions and emulates
the S/390 I/O devices as required. In the illustration, various 3390 volumes (containing
z/OS and so forth) are contained on the PC’s disks.
򐂰 The FLEX-ES license specifies how many PC processors may be used (at any one
instant) for S/390 emulation. For a ThinkPad EFS system, this is one processor. For an
xSeries EFS system, this can be multiple processors.
򐂰 The owner can elect to dedicate server processor(s) to FLEX-ES S/390 use. This
provides a performance boost. You cannot dedicate all the PC processors to S/390,
because other Linux functions need access to a processor. If processors are not
dedicated, then the normal Linux dispatching function dictates which processors are being
used by FLEX-ES at any instant. This number never exceeds the number permitted in the
FLEX-ES license. A ThinkPad EFS system has only a single processor, and it cannot be
dedicated for S/390 emulation.
򐂰 You can emulate a S/390 with a number of CPUs. This number cannot exceed the number
of server processors licensed for S/390 emulation. With the ThinkPad EFS system we are

8 The optional S/390 channel adapters available for use with FLEX-ES are ignored in this discussion.

8 EFS Systems on a Linux Base: Getting Started

 describing, this means that only a single-processor S/390 can be emulated. With an SMP
server a S/390 with multiple CPUs can be emulated.
򐂰 The emulated S/390 can use LAN adapter(s) on the PC. z/OS TCP/IP can run this way. A
single physical LAN adapter can be used both by z/OS TCP/IP and by Linux TCP/IP. This
differs from FLEX-ES under UnixWare, where a LAN adapter may not be shared by a
S/390 TCP/IP and UnixWare.9
򐂰 The FLEX-ES Terminal Solicitor is a Linux process. It provides a TN3270 server function
and listens (by default) on port 24 of Linux’s TCP/IP. An external user can connect a
TN3270 client to the Terminal Solicitor. The Terminal Solicitor and the FLEX-ES emulation
process then transform the TN3270 protocol so that it appears to be a local, non-SNA
3270 to the emulated S/390. This is the normal way to connect MVS™ consoles, TSO
terminals, CICS terminals, and so forth. There is no particular limit to the number of
terminals that can be connected this way.10 Each one must have appropriate VTAM
definitions for a local 3270, of course.
򐂰 If a LAN adapter is defined to z/OS TCP/IP, then TSO terminals may connect this way. If
z/OS TCP/IP and UNIX System Services are configured for it, ASCII Telnet sessions can
also use this connection. The Terminal Solicitor is not involved in these connections. The
two LAN interfaces (Linux and z/OS) would have separate IP addresses, of course, even
though they may share a common LAN adapter.
򐂰 The illustration shows a single emulated S/390. It is possible to emulate several S/390s at
the same time.11 The server processor(s) licensed for S/390 emulation are dispatched
among the several emulated S/390s. Only one Terminal Solicitor is used; it is shared by all
emulated S/390s.
The multiple emulated S/390s are known as multiple instances of FLEX-ES S/390
emulation. Multiple S/390 instances of z/OS are probably not very practical with a
ThinkPad EFS system, due to limited ThinkPad memory and due to there being a single
ThinkPad processor that must be shared among all the S/390 instances plus Linux.
However, multiple VSE/ESA or VM/ESA systems might be practical on a ThinkPad EFS
machine. Multiple S/390 instances are practical on larger PC systems.
򐂰 While it is not apparent in the illustration, the amount of real server memory available is a
key performance factor. The performance of an emulated S/390 degrades quickly if Linux
performs much paging (especially if the paging is triggered by FLEX-ES processes, or if
the memory being paged is used by FLEX-ES to emulate S/390 real memory). The
recommended paging level for Linux is zero, once steady-state S/390 emulation is
running. z/OS paging (in the emulated S/390) is a completely different issue and a
reasonable amount of paging has no drastic effects.
Do not confuse Linux paging with z/OS paging. The amount of real server memory
available is a key factor in setting the size of the emulated S/390 machine memory and in
deciding whether to use multiple S/390 instances.
򐂰 The terminal shown in the figure is usually the server display and keyboard. We assume
you use the X Windows interface to Linux. In this mode, several windows can be open.
Some of these might be x3270 sessions (TN3270 clients) connected to z/OS through the
Terminal Solicitor; one of these sessions might be the MVS master console. Another
window might have FLEX-ES command line interface (CLI) prompts used to control
FLEX-ES operation. CLI commands are used to IPL a S/390 operating system, for
example.

9
It also differs from P/390-based machines and the MP3000, none of which permit sharing of LAN adapters by
multiple TCP/IP stacks.
10 There is an overall FLEX-ES limitation of 2048 emulated devices. The actual limit is also controlled by the

FLEX-ES license being used and is typically 512.

11
A different FLEX-ES license is not required for this. The FLEX-ES license indicates how many server processors
may be used for S/390 operation, but does not limit how many instances of S/390s you can run.

Chapter 1. Introduction 9
 򐂰 The FLEX-ES console is actually a Linux process and might be described as a virtual
console. It might not have any terminal (window) connected to it, or it might have several
connections. The flexescli command can send commands from a Linux window to the
FLEX-ES console. Used another way, the flexescli command can create a full-time
connection between a Linux command window and the FLEX-ES console.

10 EFS Systems on a Linux Base: Getting Started

 2

Chapter 2. Practical planning

You must do some practical planning before starting a FLEX-ES installation. We can divide
the planning into several areas:
򐂰 General topics that are relevant to any FLEX-ES installation:
– Your goal: A production or sandbox system?
– The Linux distribution to be used
– Newer hardware and Linux support
– Disk space needed
– Linux file system layout
– Mode of disk usage for emulated S/390 disk volumes
– Network connections
򐂰 Topics that are especially relevant to ThinkPad installations:
– Device planning (Ultrabay™ usage)
– Choices for disk layouts
– Dual boot, alternate boot
򐂰 Topics that are especially relevant for larger server installations:
– SMP kernels, large memory, and new LAN devices
– BIOS and RAID adapter updates
– RAID configuration
– External hardware (SCSI tapes, channels, and so forth)

These topics interact with each other. We strongly suggest you read all the discussion in this
chapter rather than reading only isolated topics. The discussions here (and throughout this
document) assume that you use FLEX-ES Release 6.2.14 or later.

© Copyright IBM Corp. 2003. All rights reserved. 11

2.1 General planning
These topics are relevant to any platform used for FLEX-ES and you should understand the
following discussions.

2.1.1 Production or sandbox

By production, we mean a high-performance, highly stable system. By sandbox, we mean a
convenient, easier-to-install system. These two categories do not cover all situations, of
course, but they are convenient terms. In the sense used here, production implies some
combination of serious business, important work, maximum performance, or a
highly-stressed system. For example, a basic sysplex using multiple S/390 instances in a
ThinkPad falls in the production category, even if used only for self-education, because it
represents a highly-stressed system.

Production systems, in the sense used here, require a little more installation effort than a
sandbox system. For example, a production system should include:
򐂰 Superior hardware, including RAID disks and ECC or similar memory
򐂰 The latest BIOS updates and RAID firmware updates
򐂰 The latest Linux drivers and utilities for managing RAID adapters
򐂰 The latest Logical Volume Manager (LVM) updates (if LVM is used)
򐂰 The use of raw disk interfaces instead of simple Linux files for emulated S/390 volumes
򐂰 Installation of appropriate Linux fixes and updates
򐂰 Careful tuning of FLEX-ES disk buffers
򐂰 Selection of Linux partitions and file systems to avoid “file system full” exposures

For various redbook projects, we installed FLEX-ES systems using none of the above options
and other systems using most of these options. Both approaches produced useful, working
FLEX-ES systems. New users may start with a simple sandbox installation to gain experience
and confidence with FLEX-ES, while investing less time and effort than needed for a
production system. There is no firm dividing line between production and sandbox systems.
Your system may use some of the listed options and ignore others.

This redbook concentrates mostly on a sandbox installation, although “production-level”

facilities are briefly mentioned in various sections.

2.1.2 Selecting a Linux distribution

FLEX-ES is tested and supported only with selected releases of Red Hat Linux. It may work
with other Linux distributions, but there is no support if something goes wrong.

It is important to understand what is meant by a release of a Linux distribution:

򐂰 A Red Hat release contains a stock Linux kernel; that is, a pre-built kernel that is installed
as part of the standard installation process. (There may be more than one stock kernel;
perhaps one for uniprocessors and another for SMPs.)
򐂰 There is a kernel source tree that can be used to rebuild the kernel with different options.
򐂰 There are many packages that can be selected during installation. For example, you may
want the gcc compiler, a Web server, X Windows interfaces, and so forth. Each of these
packages has its own release level.

For example, Red Hat 9.0 provides a stock kernel at level 2.4.20 (with a corresponding kernel
source tree) and the gcc compiler at level 3.2.2. For reasons discussed later, you may replace
the kernel source tree with a different level and build a new kernel at the different level. Your

12 EFS Systems on a Linux Base: Getting Started

 system is still considered to be Red Hat 9.0, but with a different kernel. Alternately, you may
rebuild the kernel that was provided with your distribution but with different kernel options
selected.

In most cases, the stock kernel with the Red Hat release may be all you need. In other cases,
you must rebuild the kernel because the stock kernel does not provide the options you need.
We know of the following cases in which you must rebuild the stock Red Hat 9.0 kernel:
򐂰 If an SMP machine is used for FLEX-ES
򐂰 If some of the newer Ethernet adapters are used, especially 1000BaseT models
򐂰 If large server memory is to be used, where large means 1 GB or more
򐂰 If new RAID or LVM modules are to be integrated in the kernel

At the time of writing, we used or tried the following releases (with their stock kernels) with the
indicated results. In all cases, the kernel may need to be upgraded, rebuilt, or both for the
reasons just mentioned.
򐂰 Red Hat Linux 7.2: This release works well for older systems, including ThinkPads through
T21 models. The standard graphics driver does not work with later ThinkPads, or with
ThinkPad docking stations. This was a solid release that was used with many
Server-based FLEX-ES installations. If appropriate graphics drivers are installed, it is still
a good release.
򐂰 Red Hat Linux 7.3: The stock kernel does not work with FLEX-ES and we do not
recommend this release.
򐂰 Red Hat Linux 8.0: The stock kernel works well. We recommend that you use this release
unless your PC has USB 2.0 ports. This release does not work with USB 2.0 ports.
򐂰 Red Hat Linux 9.0: The stock kernel works well with FLEX-ES Release 6.2.11 and later.
(Earlier FLEX-ES releases require a kernel rebuild to use this Red Hat release.) This
release does not work correctly with some PCMCIA cards. (You can find more information
about this on Red Hat Web sites.) A kernel rebuild may be necessary for some ThinkPad
Ethernet adapters.

It is important to understand the criteria we use to recommend a Linux distribution and

release. The sole concern here is to build a small S/390, where Linux is one of the building
elements. We are not concerned with other uses of Linux, and these recommendations
would probably be different if we were interested in other Linux uses. We strongly suggest
that, if you want to work with Linux for purposes other than FLEX-ES, you build another Linux
system. For that system, you can install whatever distributions and releases you wish.

2.1.3 Newer hardware and Linux

PC hardware changes quickly. Commonly available Linux distributions and their stock kernels
may not support the latest hardware. Sometimes this can be fixed by rebuilding the kernel (or
a newer kernel) with different options.1 Sometimes it can be fixed by installing new device
drivers (if you can obtain them).

Particular areas of concern are: the processor (which must be absolutely compatible with
Pentium 3 or Pentium 4), graphics chips (the most common problem), USB 2.0 interfaces,
RAID adapters, LAN adapters (there are always new ones), and memory size (a relatively
new concern).

These are generally Linux issues. For the most part, if Linux is working properly with the
hardware, then these issues are transparent to FLEX-ES. However, FLEX-ES requires some
1
A frequent question is, “Why not include all the options in the kernel?” The answer is that some may be mutually
exclusive and, ignoring this, it would make the kernel quite large. Do not confuse the selection of kernel options with
the selection of Red Hat installation options.

Chapter 2. Practical planning 13

 specific Linux kernel options and FLEX-ES installation must access a kernel source tree that
accurately reflects the kernel being used.

FLEX-ES does not claim support for all PC hardware. FSI publishes a list of supported
machines, and this is a relatively small list. It is our experience that many other machines
work well, but that the very latest machines may have problems. (As mentioned, these are
usually Linux problems, but nevertheless may prevent you from using FLEX-ES successfully.)

Therefore, if you want to use FLEX-ES with the very latest, largest, fastest, super PC server
or mobile computer you must recognize that you are probably beyond the range of supported
systems. It may work, but you may need to learn something about Linux drivers, kernel
builds, and kernel levels and you may need a little more patience than users of more prosaic
machines. Restrictions (and fixes) discussed here are subject to frequent changes. If any of
these concerns apply to you, you should obtain the latest advice from whomever provides
your FLEX-ES support.

USB 2.0 ports

If your system has USB 2.0 ports, we suggest you use FLEX-ES Release 6.2.11 (or later) and
Red Hat 9.0.

1000BaseT LAN adapters

The newest mobile computers and server systems may include 1000BaseT Ethernet
adapters. These are not supported by the stock Red Hat 8.0 kernel. The kernel source
distributed by FSI and Red Hat 9.0 both support the more common 1000BaseT LAN
adapters.

Large memory
Servers with 1 GB or more memory require kernel changes to be most effective. This
requires rebuilding the kernel.

2.1.4 Disk space needed

PC disk drives are much larger than they were a few years ago. A FLEX-ES system running
z/OS requires a considerable amount of disk space. What is considered ample space for
normal PC usage may not be adequate. Consider the following:
򐂰 Red Hat Linux 9.0 needs 2 to 7 GB of disk space, depending on the installation options
selected. This includes a reasonable amount of work space and the space needed by
FLEX-ES modules.
򐂰 If you have a dual-boot system (probably with Microsoft Windows as the “other” system),
you need whatever disk space that system requires.
򐂰 Most S/390 (and zSeries) software is installed on 3390 volumes. FLEX-ES emulation of
3390 volumes uses almost exactly the same amount of space as the original device being
emulated. For example, a 3390-3 can contain about 2.8 GB of data and FLEX-ES
requires 2.8 GB of disk space on your PC to emulate a 3390-3.
򐂰 The amount of disk space needed for a z/OS system varies greatly. It might be as little as
a single 3390-3 (2.8 GB), although such a system would probably not be very useful. A
small TSO-oriented system might be four 3390-3 volumes and perhaps a 3390-1 for local
data sets. This would be slightly over 12 GB. A larger system, such as the prebuilt systems
available to certain users on CDs, if fully installed, might need 12 or 13 3390-3 volumes,
plus at least one local volume, with a total of about 40 GB. If a large amount of local data
is involved, space is needed for this.

14 EFS Systems on a Linux Base: Getting Started

 򐂰 You might or might not install the z/OS distribution library volumes (DLIBs). This can easily
be four 3390-3 volumes (and these are included in the 40 GB just mentioned). If you do
not need these, you save about 11 GB.

The question, “How much disk space do I need?” is almost impossible to answer unless you
have a fairly detailed idea of what volumes you need. A general answer is that, if you want to
use a recent z/OS release with many of the common subsystems and products, you should
have at least 40 GB disk space. More is better.

2.1.5 Linux file system layout

Our installation descriptions produce a very basic file system layout, with only four partitions:
򐂰 A small boot partition. (A separate partition might not be necessary, but almost all Linux
documentation seems to do this.)
򐂰 A small swap partition (because a Linux system used for FLEX-ES should not do much
paging).
򐂰 A root file system that contains all the normal Linux file systems, such as /usr, /tmp, /etc,
and so forth.
򐂰 A separate, large file system for emulated S/390 volumes. This can be a large LVM area
instead, if raw disk devices are used.

Experienced Linux (or UNIX) users may object to this, saying that it is better to have separate
partitions (file systems) for /usr and /tmp and /var and so forth. This is probably correct, but
we believe this is not needed for a basic FLEX-ES installation. This redbook concentrates on
getting started with FLEX-ES and attempts to simplify the process as much as possible.

If substantial other usage of the Linux system is anticipated or if exposures due to inadvertent
filling of file systems are a concern, then it is better to have separate partitions (or logical
volumes) for such things as /usr, /tmp, and so forth. If this is done, then you must decide on
partition sizes for these file systems, of course.

We could make the root file system very large and place all the emulated S/390 volumes
there. We elected to create a separate partition for the emulated volumes for two reasons:
򐂰 It is easier to “scratch and reinstall” Linux without disturbing the S/390 volumes.
򐂰 There are fewer potential fragmentation problems for the S/390 volumes.

Our descriptions use Linux ext3 file systems where appropriate. The ext3 file system offers
better recovery than other common file system types.

2.1.6 Disk modes for FLEX-ES

Emulated S/390 volumes (usually 3390 devices) can be handled in two ways:
򐂰 An emulated 33902 can be a single Linux file. It might be a large file, 2.8 GB to emulate a
3390-3, for example, but it is a simple Linux file.
򐂰 An emulated 3390 can be a Linux raw device. The raw device is then equated to a disk
partition (as created by the fdisk utility or something similar) or a logical volume (as
created by a LVM). A separate partition or logical volume is needed for each emulated
3390 volume.

2 We use 3390s in this discussion, but the same comments apply to any type of emulated S/390 disk drive.

Chapter 2. Practical planning 15

 Each method has advantages and disadvantages:
򐂰 The simple Linux files are easiest to set up, especially for a relatively unskilled Linux user.
Less planning is needed and disk management is more automatic.
򐂰 The simple Linux files are a little slower than the raw device method, and emulated disk
response is less predictable. (Normal Linux disk I/O processing, including caching, is
used for Linux files. This Linux processing uses instruction cycles. The Linux disk
caching and the FLEX-ES disk caching interact in ways that make response less
predictable.)
򐂰 The raw devices are more complex to set up. In practice LVM is used instead of “real” disk
partitions3 and the user must learn something about LVM usage.
򐂰 The raw devices offer smoother disk performance and use less processor overhead.
򐂰 Fragmentation is not an issue when raw devices are used.

When a z/OS application writes a block to disk, FLEX-ES writes to a FLEX-ES cache for that
volume. By default, each emulated 3390 has 15 tracks4 of cache.5 When FLEX-ES decides
to write the contents of a cached track to disk, it uses a normal write interface for Linux. If the
file is a normal Linux file, in a normal Linux file system, Linux then places the data in its own
cache and flushes this cache to disk sometime later. If the file is a raw device, the FLEX-ES
write operation causes a direct disk write. No Linux file system processing or caching takes
place.

FSI recommends the use of raw devices for all FLEX-ES installations. We, in this redbook,
offer slightly different recommendations:
򐂰 We recommend that you use a raw device for production systems or if every possible bit of
disk performance is needed.
򐂰 We recommend that you use simple Linux files for less demanding sandbox environments,
especially for ThinkPads used by relatively few users and used for self-education,
demonstrations, or minor development. In general, we believe the use of simple Linux files
is good enough for this environment and makes installation less complex.

The remainder of this redbook works with FLEX-ES installations using simple Linux files for
S/390 disk emulation. A chapter in a companion redbook, S/390 PartnerWorld for
Developers, ITSO/EFS Project EFS Systems on a Linux Base: Additional Topics, SG24-7008,
describes the use of raw devices.

It is possible to use both disk methods. Some emulated volumes can be in standard Linux
files and other volumes on raw devices using logical volumes.

FSI Business Partners installing commercial FLEX-ES systems usually work with raw
devices. This is appropriate because they have the skills to plan and install such systems and
these production systems usually have a relatively static emulated disk layout.

2.1.7 Network connections

Your system can be connected to the Internet or other LAN network. However, you should
consider a few issues before installation. These include:
򐂰 You can use an external DHCP server (already on your LAN) to obtain an IP address for
Linux. You cannot obtain an IP address for z/OS TCP/IP this way. (You are not required to

3 Only a limited number of “real partitions” can be placed on a disk. Also, subsequent changes with fdisk or a similar

utility can easily cause the loss of the contents in all the partitions on the disk.
4
That is, 15 3390 tracks, where a 3390 track is about 57 KB.
5 This is variable and a sophisticated FLEX-ES installation may tune the FLEX-ES cache options.

16 EFS Systems on a Linux Base: Getting Started

 use z/OS TCP/IP, of course. You need it if you want to FTP directly to z/OS, Telnet to UNIX
System Services, and so forth.) You need a static IP address in order to use z/OS TCP/IP.6
򐂰 You (or other users) can connect to your z/OS through the LAN by connecting to the
FLEX-ES Terminal Solicitor (which is a process under Linux) using your Linux IP address.
These connections appear as local 3270 terminals to z/OS, and do not use z/OS TCP/IP.
This can be done if Linux obtained its address via DHCP or if you assigned a static IP
address to Linux.

Our step-by-step instructions assume you are connected only to a small private LAN, with no
DHCP or DNS, or are not connected to any LAN. The examples use non-routable LAN
addresses in the 192.168.x.x address range. If you are connected to a larger LAN, we
suggest you install our way first (with no LAN, or with a small temporary LAN7). Once you
have Linux, FLEX-ES, and z/OS working in this basic environment, you can then extend (or
reinstall) your system for a larger environment.

2.2 ThinkPad topics

ThinkPads have a few unique issues, mostly related to I/O device combinations and
attachments. The considerations are not complicated, but we suggest that potential
ThinkPad users review this material.

2.2.1 T40 ThinkPads

You should be aware that T40 ThinkPads (and newer ThinkPads) have USB 2.0 ports. It is
best to use Red Hat 9.0 and FLEX-ES Release 6.2.11 or later with these ThinkPads.

2.2.2 BIOS settings

You should verify that the Boot device List F12 Option is enabled. This lets us temporarily
select the CD-ROM as a boot device. Our normal Boot List contained Removable Devices
(that is, a diskette) followed by Hard Drive.

We adjusted some of the power management options we found in the BIOS. This may not be
appropriate for you, but we did not want the screen to suddenly go blank when it contains the
z/OS console window, for example. These controls are found by pressing F1 during ThinkPad
startup and following Config->Power. Be certain to scroll down (using the down arrow key) in
this window to find all the controls.

z/OS cannot tolerate a ThinkPad Suspend or Hibernate operation. We disabled the

Hibernate by timer option. We did not disable the Suspend when LCD is closed function, but
we are careful not to close the ThinkPad cover when z/OS is running. We did leave a
3-minute HDD off timer active.

After exiting the BIOS setup function and performing power-off/power-on functions, we had an
initial prompt to provide a temporary boot device list. Pressing F12 provided this list. The
CD-ROM was the third element in the list, and this is needed to start Linux installation.

2.2.3 Device planning (Ultrabay usage)

Most ThinkPads have an internal disk drive bay and an Ultrabay. The Ultrabay can contain a
second hard disk, a diskette drive, or a CD-ROM (or CD-RW or DVD) drive, but only one
6
This discussion applies to any S/390 operating system run under FLEX-ES, of course, and is not unique to z/OS.
7 This can consist of a single Ethernet hub or token ring MAU.

Chapter 2. Practical planning 17

 device at a time. Furthermore, you can boot from any of these devices, by making
appropriate BIOS selections when you start the ThinkPad.

FLEX-ES systems tend to use a large amount of disk space, primarily for emulated S/390
volumes. Also, many owners want to use their ThinkPads for other purposes, such as
day-to-day Windows operation. These requirements can be met in a variety of ways.

A key factor here, considering z/OS use under FLEX-ES, is that you may need to access the
disk containing your emulated S/390 disk volumes and a CD-ROM drive at the same time.
This occurs when loading a new release of z/OS from CD to a hard disk, for example. You
may also need to access a diskette drive while running Linux.

These points may appear obvious, but please consider common situations:
򐂰 You want to keep your current internal disk, containing your day-to-day Windows functions,
and add Linux as a dual-boot function on this disk. You then obtain a second hard disk for
the Ultrabay and place all the large S/390 emulated volumes there.
򐂰 Another variation of this is that you place your complete Linux + FLEX-ES + S/390
volumes on a second hard disk and boot it from the Ultrabay.

In both cases, your internal disk drive and your Ultrabay are occupied. Now, if you need to
use a CD-ROM drive to install new S/390 volumes, where do you install the CD-ROM drive?
The same question exists for a diskette drive.

There are several solutions and you should think about these before deciding how to
configure your ThinkPad system for FLEX-ES:
򐂰 An attractive option is to obtain a ThinkPad docking station.8 This has another Ultrabay
slot where you can install a CD-ROM drive, a diskette drive, or another hard disk. An
additional advantage is that it appears to provide better air flow (cooling) for a long-running
system. A docking station is not as portable as a ThinkPad, but you probably do not need
to use the additional Ultrabay very often. An ability to borrow a docking station might be
an option.
򐂰 Some port replicators, USB attachments, or both support diskette and CD-ROM drives.
Provided that Linux can recognize and use these drives, these options may solve the
problem. We do not have a specific list of these attachments that work with Red Hat Linux
9.0.
򐂰 You can obtain an internal disk drive that is large enough to hold your Windows
partition(s), Linux partitions, and all the emulated S/390 volumes you need. This generally
means a 60 GB drive or larger.
򐂰 You can remove the internal Windows drive and install a Linux drive in the internal disk
slot. This is obviously not a recommended ThinkPad procedure although we have done it
many times without problems. It must be done carefully and gently, with particular
attention to the external plastic cover that is needed to “grab” the disk for removal.9
򐂰 You can stage S/390 volumes from a CD-ROM drive (in the Ultrabay) to space in a Linux
partition on the internal disk. You then remove the CD-ROM drive, install your second
hard disk in the Ultrabay, and copy the volume from the internal drive to the Ultrabay drive.
The amount of free Linux space on the internal drive determines how many volumes can
be staged at one time. A single 3390-3 volume requires 2.8 GB. You must stop and

8
The IBM Web pages indicate that docking stations are not supported for Linux. Based on this, FSI does not support
their use. If you are interested in docking stations, read the Redbook S/390 PartnerWorld for Developers, ITSO/EFS
Project EFS Systems on a Linux Base: Additional Topics, SG24-7008, for a brief discussion of our experience.
9
We noticed that some recent ThinkPad hard drives are shipped without a lightweight metal carrier on the bottom of
the drive. This carrier is needed to reasonably install the drive in the internal drive slot.

18 EFS Systems on a Linux Base: Getting Started

 reboot Linux for every Ultrabay change because Linux does not support hot swapping of
Ultrabay devices.10

2.2.4 Choices for disk layouts

The preceding discussion refers to a number of disk layout options. This section describes
these in more detail. Figure 2-1 roughly illustrates several options.

Windows
Windows 8G

/s390 40G

/s390 40G /s390 40G

swap .3G
Linux

/ (root) 3G

/boot 80M
Windows

swap .3G
swap .3G
Linux

Linux
/ (root) 3G
/ (root) 3G
/boot 80M Windows 8G
/boot 80M
Drive devoted Internal Ultrabay
Dual-boot single drive
to Linux Dual-boot setup, two disks

Figure 2-1 Three disk layout options

The sizes shown, such as 3 GB for a Linux root partition, are somewhat arbitrary and depend
on your particular requirements. The first drive in the illustration, devoted to Linux (and
FLEX-ES and S/390 emulation), can be used three ways:
򐂰 It can be the sole drive for the system and your ThinkPad is devoted to S/390 operation.
򐂰 It can be installed and booted in the Ultrabay, leaving your internal disk (which may contain
Windows) unused for the duration of your Linux session. Some adjustments are needed to
boot Linux from the Ultrabay. These are discussed in S/390 PartnerWorld for Developers,
ITSO/EFS Project EFS Systems on a Linux Base: Additional Topics, SG24-7008.
򐂰 It can be swapped with your normal internal hard disk. The internal drive bay in a
ThinkPad is not designed for frequent changes, and this limitation needs to be considered.

The second set of drives in the illustration has both Microsoft Windows and Linux on the
internal drive. The total space needed for Linux is less than 4 GB; all the remaining space
can be left for Windows. A dual boot function is used to select either Windows or Linux for
booting. A second hard disk, in the Ultrabay, contains all the emulated S/390 volumes. This
disk needs to be in the Ultrabay only when you are using S/390 emulation, of course.

The third option illustrated is the most straightforward, if your disk is large enough.
10 This was true at the time of writing, but may change in the future, of course.

Chapter 2. Practical planning 19

2.2.5 Dual boot, alternate boot
Dual boot means that two (or more) operating systems are present on a bootable disk.
Typically, a small program is loaded when the system is booted, and this program allows the
user to select which of the main operating systems to load.11

Alternate boot, in the context used here, means to boot from an alternate hard disk in an
Ultrabay slot. With a ThinkPad you can, for example, arrange BIOS options such that the
system boots from a hard disk in the Ultrabay if one is present. If a hard disk is not in the
Ultrabay, it boots from the internal disk.

Dual boot
There are many ways to provide dual boot functions. We describe one method in this
redbook. It involves:
򐂰 Starting with a hard disk that contains only a Microsoft Windows partition, we use
PartitionMagic (a product of PowerQuest) to “squeeze” the Windows partition into a
smaller area to make room for Linux partitions.
򐂰 The Linux installation process detects the presence of the Windows partition and offers to
install a dual boot function; we used grub and the Master Boot Record for this purpose.12
򐂰 When we boot the resulting system, we have a choice between Windows and Linux.

A concern with dual boot solutions is maintenance. The various methods of implementing
dual boot functions are not always compatible with all operating systems. For example, you
might apply maintenance to the non-Linux operating system that causes the dual-boot loader
to be overlaid. If this happens, you need to recover the dual boot loader somehow. (One
option is to create a Linux boot diskette and use this to begin recovery from such problems.)

The process we used to “squeeze” an existing partition (to make space for Linux partitions) is
briefly described in 2.4, “Dual boot preparation” on page 21. The relevant point during Red
Hat Linux 9.0 installation is mentioned in 3.1, “Basic Red Hat Linux 9.0 installation” on
page 24.

2.3 Larger server topics

By larger server we typically mean an SMP (multiprocessor) machine with RAID disks and
large memory. It might also have FSI channel adapter cards, SCSI tape drives, and other
peripherals attached to it.

A simple desktop PC falls between a ThinkPad and a larger server. In most cases, it can be
regarded as similar to a ThinkPad without the Ultrabay issues.

2.3.1 SMP, large memory, new LAN adapters

At the time of writing, you must rebuild the Linux kernel if you plan to run FLEX-ES on an SMP
system. You may need to rebuild the kernel if you have unusually large memory or newer LAN
adapters. This is discussed more in 3.3, “Kernel rebuild” on page 30.

11
There are other ways to do this, of course. We do not attempt to explore all the ways to implement dual-boot
systems.
12 grub is one of the loader programs available for Linux.

20 EFS Systems on a Linux Base: Getting Started

2.3.2 BIOS and ServeRAID adapter updates
IBM and FSI strongly recommend that you install BIOS and firmware upgrades before
installing FLEX-ES. (If you obtained your system from an FSI Business Partner, this work may
have been done for you.) You can obtain these updates by downloading the latest
ServerGuide™ CD, ServeRAID CD, or both from the IBM Web sites. The exact installation
procedures vary with different server models. We do not attempt to provide specific
installation instructions here.

2.3.3 RAID configuration

The ServerGuide CD allows you to customize the RAID configuration, using the ServeRAID
Manager program on the CD.13 This is a graphical user interface (GUI) program, and it is
difficult to describe the operation in a few words. The goal for our sandbox system was to
place all our disk drives in a single RAID 5 array, and then create a single logical drive using
all the space in the array.

The ServeRAID Manager displayed our initial, default configuration as:

Logical drv Size MB RAID Level Array Hot Spare
1 New 52071 5 A Yes

After working with the parameters, our final configuration was:

Logical drv Size MB RAID Level Array Hot Spare
1 New 69428 5 A No

To get to this point, we deleted the drives from the default array and then added all the drives
to a new array. (This left no drives for a hot spare. This was acceptable for our minimal test
system. However, we strongly recommend a hot spare for a production system.)

2.4 Dual boot preparation

There are many ways to install dual boot capabilities and we do not attempt to address the
general topic. The following description provides the base for a simple dual boot that has
worked well for us. This section describes only part of the setup. The actual dual boot code
(using Linux grub) is installed during Linux installation, as described in the next chapter.

This description is oriented to a ThinkPad, but you should have no difficulties adapting it to
other systems.

We started with a hard disk completely occupied by a single partition with a version of
Microsoft Windows installed. We used PartitionMagic (a product of PowerQuest) to reduce
the size of the Windows partition, making space for several Linux partitions. You can allocate
the Linux partitions using PartitionMagic, or you can allocate them during Linux installation.
We suggest allocating the Linux partitions while installing Linux (and while using Disk Druid),
but you need to take care not to disturb the Windows partition at that time.

In our particular example, we had a 15 GB internal disk drive on a ThinkPad. The drive
contained Microsoft Windows and a number of applications. Our goal was to install Linux
(with FLEX-ES) on this disk without destroying the existing contents. The disk is too small for
z/OS volumes. We intended to place them on another hard disk in the Ultrabay slot. With this
setup, we can boot either Windows or Linux from the internal ThinkPad disk. If we intended to
work with FLEX-ES, we would install the second hard disk in the Ultrabay before booting
Linux.
13 The ServeRAID CD should also have the same programs.

Chapter 2. Practical planning 21

 We obtained PartitionMagic on two diskettes. After booting it (and after it called for the second
diskette), we received a display something like that illustrated in Figure 2-2.

This reflects a 15 GB drive (14396 MB), with a single partition. 2385 MB are used in the
partition, and 12010 MB are free. We arbitrarily decided to reduce the Windows partition to
8 GB.

PartitionMagic Pro 6.0 by PowerQuest

General Disks Partitions Operations Tools Help

Disk 1 - 14396 MB C

Partition Type Size MB Used MB Unused MB Status Pri/Log

*:Win-2000 NTFS 14396.5 2385.9 12010.6 Active Primary

Apply Exit

Figure 2-2 PartitionMagic display

This is done by first selecting the line describing the Windows partition (and leaving this
selection highlighted), selecting Operations-> Resize, and entering 8000.0 as the New Size.
PartitionMagic then updates the display to show a smaller partition for Windows and a new
line for unallocated space.

PartitionMagic does not actually make disk changes until you click Apply. It queues the
changes until that time, but reflects them in the displayed information. At this point, you can
click Apply and make the change. You can then further partition the unallocated space while
you are installing Linux. Changing the disk layout takes time. After we clicked Apply, the
changes took 17 minutes in this example, using a T20 ThinkPad. Larger disks take longer; a
60 GB drive required 55 minutes in a T23 ThinkPad.

Alternately, you can define your Linux partitions now (before clicking Apply). This is done by
highlighting the unallocated space line and then selecting Operations-> Create. In the
following display, you can specify:
Primary Partition
Linux ext2
/boot
80 MB
OK

This creates a Linux /boot partition. You can then create a Linux root partition, a swap
partition, and so forth. While installing Linux, you can redefine the ext2 file type to be ext3.

The actual dual boot code is installed during the Linux installation, when you specify the boot
configuration. At that time, we elected to place the dual boot program in the Master Boot
Record (MBR).

22 EFS Systems on a Linux Base: Getting Started

 3

Chapter 3. Linux installation

This chapter describes in cookbook style how we installed Red Hat Linux 9.0. A new Linux
user may find the step-by-step instructions useful. An experienced Linux person should
review the comments about the choices we made. Some are arbitrary choices, but a few are
required.

This description is for installation on a ThinkPad. Installation on a server is similar, although

hardware related details (display selection and disk device names, for example) may be
different. This description and this whole redbook assume that you use simple Linux files for
S/390 volumes. The use of Logical Volume Manager (LVM) and raw devices for emulated
volumes is described in S/390 PartnerWorld for Developers, ITSO/EFS Project EFS Systems
on a Linux Base: Additional Topics, SG24-7008.

Red Hat Linux installation is easy and quick. If you do not like your first results, simply install it
again. If you do not change the disk partitioning during subsequent installation, the contents
of other partitions (such as, where emulated S/390 volumes are stored) are not disturbed.

© Copyright IBM Corp. 2003. All rights reserved. 23

3.1 Basic Red Hat Linux 9.0 installation
We booted from the first Linux CD (using the ThinkPad F12 option to select a temporary boot
device). This produced a Red Hat logo display and offered the choice of graphics mode or text
mode installation.1 We selected text mode, primarily because it was easier to document for
this redbook. Then we went through a number of installation prompts. Many of the prompts (in
text mode) allow you to use the Tab key to change choices and to move to the OK button.
linux text (To select installation mode)

Welcome to Red Hat Linux: OK

Language: English
Keyboard: US
Mouse: Generic 3-button mouse
Installation Type: Custom (Do not select mobile computer or server)
Select partition tool: Disk Druid

You can select fdisk instead of Disk Druid, and you should use whichever one you find most
comfortable. If you are not familiar with either, select Disk Druid.
Warning: Partition table unreadable......Initialize? Yes

You may get this message if you have a completely blank disk.

Be aware that Disk Druid (and fdisk) cannot resize partitions. They can only create or delete
partitions. If you need to resize partitions, you must do this before you start the Linux
installation process.

Our ThinkPad disk had no installed partitions and was 100% available for allocation. You may
have disk contents from a previous use or from a dual boot setup. You need to adapt your
disk layout to accomplish approximately what we outline here. Using Disk Druid, we allocated
four partitions, using ext3 file systems for root and /s390. Our final Red Hat 9.0 display was
like this:
Device Start End Size Type Mount Point
/dev/hda
/dev/hda1 1 10 73M ext3 /boot
/dev/hda2 11 417 3004M ext3 /
/dev/hda4 418 452 258M Extended
/dev/hda5 418 452 258M swap /
/dev/hda3 453 7752 53894M ext3 /s390

OK
(If it complains about the /boot size, say “Yes” to continue):

We obtained this layout (using a 60 GB disk) by requesting (in order):

򐂰 /boot, ext3, 76 MB
򐂰 / (for root), ext3, 3000 MB
򐂰 swap, 256 MB
򐂰 /s390, ext3, all the remaining space

Your hda numbers may differ, depending on the order in which you created the partitions. The
position of the Extended partition seems to vary, depending on the whims of Disk Druid, but it
does not matter. Remember that, if you have an operating system already installed (Microsoft
Windows, for example), you see that partition already in place when you start Disk Druid.

1
You can use the graphical user interface (GUI) installation. The steps are the same as in text mode and the input
required is the same as in text mode.

24 EFS Systems on a Linux Base: Getting Started

We selected a small, arbitrary size (256 MB) for a Linux swap partition. We may not need this
much, although Red Hat documentation recommends swap space equal to memory size
(which was 1 GB in our case). 256 MB seemed to be a reasonable compromise and has
worked without problems. Red Hat documentation recommends 75 MB for /boot, but the disk
allocation rounding method reduced our allocation below this point; we ignored the warning.

The creation of the /s390 partition assumes that standard Linux files are used for emulated
S/390 volumes. If you plan to use raw devices, you make this an LVM partition.

It may take some experimentation with Disk Druid to obtain the desired configuration. We
suggest that you simply work with the Disk Druid options (Add, Delete, Edit, scroll bars) until
you succeed. It may be inconvenient to change the partitions later, so we suggest you take
time to get it right now. Remember to scroll within some of the Disk Druid text windows
(especially the Filesystem type window). Also, Disk Druid automatically creates the extended
partition after you create a few other partitions, and move some of your partitions into the
extended space.

The /hda terminology is used for an IDE disk drive. If you have a SCSI drive, it is /sda. If your
SCSI drive is really a RAID array, and you have already created your RAID partition, you see
only a large SCSI drive. That is, a simple RAID array is not visible to Disk Druid (or to Linux).

If we selected Linux components (in a later step) to obtain a small Linux suitable for FLEX-ES,
we actually used about 1.9 GB in the root file system. If we elected to install everything, we
used 4.9 GB in the root file system. The file system should contain a reasonable amount of
free space. For this reason, we suggest that you make your root file system at least 3 GB to
6 GB.

If you have more Linux or UNIX experience, you may want to create several file systems for
Linux instead of placing everything in a single root file system. Other than having about 25
MB available in /usr (for later FLEX-ES installation and working files), there are no special
requirements for your Linux disk configuration.

We assigned all the remaining space on this drive to a partition with mount point /s390. This
name is arbitrary, but clearly indicates the intended purpose of the partition. Remember that
the disk layout we describe is not required; it is merely one way to arrange a system.

At this point, we had only a single hard disk installed. A second hard disk will eventually
replace the CD-ROM drive in the Ultrabay of the ThinkPad; we cannot have both the CD-ROM
drive and the second hard disk installed at the same time. We initialized the second drive
later.

The installation prompts continued with fairly basic items:

Boot Loader Configuration: grub
Boot Loader Configuration: (Possibly remove the hdc=ide-scsi parameter)
Boot Loader Configuration: (We did not use a GRUB password)
Boot Loader Configuration: Red Hat Linux /dev/hda2: OK
Boot Loader Configuration: /dev/hda Master Boot Record (MBR)

The hdc=ide-scsi parameter appears if you have a CD R/W drive in the Ultrabay (or an IDE
CD R/W unit in an xServer). This parameter causes the Ultrabay to be treated as a
pseudo-SCSI device. This is required to support CD writing. However, it does not allow a hard
disk to be used in the Ultrabay. You can easily change this later.

If you have a second operating system installed, additional Boot Loader Configuration panels
ask which system you want loaded by default and give you a chance to change the displayed
name for the other system.

Chapter 3. Linux installation 25

 Next you must specify a network address for your system:
Network Device: eth0
[ ] use bootp/dhcp (Probably deselect this)
[*] Activate on boot (You probably want this)
IP address: 192.169.0.110
Netmask: 255.255.255.0
Default gateway: (If you have one; we left this blank)
Primary name server: (If you have one; we left this blank)
Secondary name server:
Ternary name server:

If you have multiple LAN adapters, the system may not assign your IP address to the adapter
you want to use. You should verify your network setup after the basic Linux installation is
completed. These network device parameters are easy to change later.
Hostname Configuration:
Hostname: t23
Security level: No Firewall (Use the Spacebar to change it)
Language Support: English (USA) (As appropriate)
Time Zone: (As appropriate)
Root password: xxxxxxxx (Must be at least 6 characters)

The No Firewall security option was appropriate for our purposes, but may not match your
needs. As far as we know, this has no particular FLEX-ES security implications2. You can
select the options appropriate for your needs. If you need firewall protection, select high and
customize. In the customization window, specify 24:tcp,555:tcp.
Authentication Configuration:
[*] Use shadow passwords
[ ] Enable MD5 Passwords
[ ] Enable NIS
[ ] Enable LDAP
[ ] Enable Kerberos
OK

These authentication options are up to you. If you are not certain, we suggest that you use
the options shown here. Next you select the general packages you want installed. The easiest
option is to select Everything (which is the last option in the list). Be certain your root file
system is at least 5 GB. If you want to save a little disk space, you can use the choices we
used for Red Hat 9.0:
Package Group Selection: Customized (You need to scroll the following list)
X Windows System
Gnome Desktop Environment
Editors (We clicked Details and removed Emacs)
Graphical Internet
Text-based Internet
Office / Productivity
Server Configuration Tools
FTP Server
Network Servers
Development Tools
Kernel Development
Administration Tools

2 FLEX-ES needs to use TCP/IP port 24 on Linux for its Terminal Solicitor function. Almost all FLEX-ES installations

require this. FLEX-ES may also use port 555 if you have remote FLEX-ES resources installed. The typical ThinkPad
EFS user may not use this, especially for initial FLEX-ES use. If you install firewall functions, be certain that port 24
is not blocked.

26 EFS Systems on a Linux Base: Getting Started

 System Tools
Printing Support
OK

Gnome and KDE are competing implementations of an X Windows-based desktop

environment. You need at least one of them. Experienced Linux users often have strong
preferences for one or the other package. You must include Development Tools and Kernel
Development. You may select other options. As far as we know, additional selections have
no impact on FLEX-ES operation. The package groups we selected required 1.9 GB disk
space.

The installation took 10 to 15 minutes. It called for the second CD and third CD (if needed)
during the installation.
Boot Disk: No (Could not switch to diskette drive)

In our case, the Ultrabay was being used for a CD-ROM drive and we could not install a
diskette drive to create a boot diskette. If your environment allows you to write a diskette at
this time, we suggest that you answer Yes to this question.

The installation program then probes the video configuration. On our T23 ThinkPad (IBM
2647-3U5), Linux sensed the following:
Video card: S3 Savage/MX
Video RAM: 8192
OK

Monitor Configuration: Unprobed Monitor <Change> (Select Change)

Hsync Rate: 31.5 - 48.5 <Change>
VSync Rate: 50-70

We selected the Change option for the monitor. This produced a list with a large number of
monitors. We scrolled this list (using the down arrow key) and selected:
Generic Laptop Display Panel 1400x1050 (For our T23 model)
OK

Your requirements may be different, of course. We accepted the Hsync and VSync rates
produced by the system.
X Customization: High Color (16 Bit) <Change>
X Customization: Resolution: 1024x768 <Change> (Changed to 1400x1050)
X Customization: Default Login: Graphical

We changed the video resolution to 1400x1050 for our T23. You should select the correct
screen resolution to match your system, of course. We left the 16-bit color selection
unchanged, although you may want to optimize it to match your system.
Complete:
OK

Selecting OK at this point causes the system to reboot. You should remove the second Red
Hat CD while you are rebooting.

After the reboot, you are asked to add a new user. You should probably do this, but do not add
user ID flexes.

Select the Forward option in the graphics window and change your system date and time if
necessary. After selecting Forward again, our system offered to test our sound card. (This
produced no useful results.) Selecting Forward again, the system offered to register our
machine with the Red Hat Network.

Chapter 3. Linux installation 27

 Yes, I would like to register with Red Hat Network.
X No, I do not want to register my system.

If you elect to register your system, do not download updates and fixes unless you have
independent information that the updates will not affect FLEX-ES installation.

Selecting Forward again produced a panel that offered to install additional software. This is
not needed, and you can select Forward again. This produced the Finished Setup notice and
selecting Forward produced the standard login screen.

We logged into Linux as root and examined our filesystem sizes. After gnome starts, click the
right mouse button and select New Terminal to obtain a command-line window.)
# df -h (display file system usage)
Filesystem Size Used Avail Use% Mounted on
/dev/hda2 2.9G 1.4G 1.3G 51% /
/dev/hda1 71M 9.2M 58M 14% /boot
none 503M 0 503M 1% /dev/shm (swap)
/dev/hda3 52G 33M 49G 1% /s390

As seen, the Linux selections we chose used 1.4 GB in our root partition, leaving a
considerable amount of free space. There is nothing in the /s390 file system except the
inodes; it is not clear how the df command computes its displayed sizes.

3.2 Installation notes

The following notes may be of interest.

ThinkPad clock accuracy

An emulated S/390 under FLEX-ES on a Linux base on a ThinkPad can cause the ThinkPad
clock to lose time if the emulated S/390 CPU is heavily loaded. This was first observed with
Red Hat Linux 8.0 and is also present in Red Hat 9.0. This effect appears to be due to the
Linux software for ThinkPad power management. You can correct this by removing the power
management function in Linux. You need to use an apm=off parameter during Linux boot up.
One way to do this is to edit /boot/grub/grub.conf and use the line:
kernel /vmlinuz-2.4.20-8 ro root=LABEL=/ apm=off

Here, we added the apm=off parameter to the existing line. A side effect of this change is that
the system shutdown function of Linux no longer turns off the power for the ThinkPad.

Second hard disk on ThinkPad

If you install Red Hat Linux using a CD-RW (or DVD/CD-RW) drive in the Ultrabay, the
installation process detects that you are using a device with write capability. Linux uses a
pseudo-SCSI interface for CD write support. In this case, it automatically includes a boot
parameter that causes the Ultrabay to be treated as a pseudo-SCSI device. A normal
ThinkPad hard disk mounted in the Ultrabay does not work when the system is booted with
this parameter. The parameter is hdc=ide-scsi, and is located in /boot/grub/grub.conf. You
can remove the parameter, or create two boot selections so that you can choose, at boot time,
whether to keep the parameter (allowing CD write operations) or remove the parameter
(allowing use of a hard disk in the Ultrabay). Here is a grub.conf example to do this:
default=0
timeout=10
splashimage=(hd0,0)/grub/splash.xpm.gz
title Red Hat Linux (2.4.20-8) (CD rw)
root (hd0,0)

28 EFS Systems on a Linux Base: Getting Started

 kernel /vmlinuz-2.4.20-8 ro root=LABEL=/ apm=off hdc=ide-scsi
initrd /initrd-2.4.20-84.img
title Red Hat Linux (2nd HDD)
root (hd0,0)
kernel /vmlinuz-2.4.20-8 ro root=LABEL=/ apm=off
initrd /initrd-2.4.20-8.img

Your grub parameters may be slightly different. Notice that we included the apm=off
parameter in both cases. When you boot the system the next time, you can choose between
Red Hat Linux (2.4.18-14) (CD rw) and Red Hat Linux (2nd HDD), the first one being the
default (due to default=0). If you create an SMP kernel, as described in the following section,
you have more boot choices.

SMP kernel
If the installation process detects that you are installing on a server with multiple processors,
it installs an SMP kernel. If you intend to use FLEX-ES on an SMP machine, you must rebuild
the kernel as described in 3.3, “Kernel rebuild” on page 30. The stock SMP kernel does not
work with FLEX-ES.

Network services
After the installation outlined above, you can log into the system and ping another host on
your subnet.3 If the gateway IP system you specified (if any) is functional, you may be able to
access remote hosts.

The installation options we selected installed telnet and ftp. You can Telnet to another
machine but other systems cannot Telnet to your machine. Likewise, you can FTP to another
machine but they cannot FTP to you. To enable the Telnet and FTP daemons, select Main
menu -> System Settings -> Server Settings -> Services and enable telnet and vsftp in
the Services menu. (Red Hat installed vsftpd as the FTP server when we selected Everthing
in the Package Group Selection list.)

On an earlier Red Hat release, we discovered that an unattached LAN port may cause
generally sluggish operation of Linux. In one case, we used a token-ring PCMCIA card for
LAN connectivity and nothing was connected to the Ethernet port in the ThinkPad. System
operation was sluggish. On investigation, it appeared that Linux was starting the Ethernet
port and looking for a DHCP server. We disabled the port by deselecting the Activate device
when computer starts option in the Network configurator.

Be aware that Linux may configure and start all your LAN ports automatically.4 Check that the
IP address you specified during installation is associated with the correct LAN adapter.
Check that DHCP is not being used (unless you want it, of course). Use the Network
configuration tool to modify the setup. We found it necessary to reboot in order to make some
of our changes effective.

Token-ring support
A patch may be needed to Linux in order to use token-ring adapters. (This is for both
ThinkPad and xSeries systems.) The change is to net/802/tr.c in the Linux source tree. FSI
provides a fix in a file named linuxtokenringpatch, and this may be applied by the Business
Partner that supplies your system.

3
You may have problems if you ping systems on other subnets. The remote system must have a defined route back
to you in order for the ping to work. Testing with another system on your local subnet avoids routing problems.
4 While testing LAN operation on a server with multiple Ethernet interfaces, we noted that the installation process

had chosen a PCI Ethernet adapter instead of the integrated Ethernet port as the eth0 device. This was not a
problem, although it took a while to determine why we could not ping from a LAN connection to the integrated
Ethernet port. Linux detected the integrated Ethernet port and addressed it as eth1.

Chapter 3. Linux installation 29

 Exactly when is this patch needed? The best information we have is this:
A system lock-up may occur if the kernel needs to add routing information to the token ring
route tables. This can occur when a frame is received in which the source routing bit is on
and the destination MAC address is neither a broadcast nor a multicast address.

We have used token ring interfaces with a number of Linux systems (some with FLEX-ES,
and some without it) without problems (and without applying the patch).5

System freeze
We had a minor problem with earlier Red Hat releases used on T20 and T23 ThinkPads. We
do not know if the problem exists for other ThinkPads or other releases. By default, gnome
invokes a screen saver function after a period of no keyboard activity. This function uses a
random series of different screensaver programs for the screen. At least one of these patterns
caused the system to freeze.

We have not seen this freeze with Red Hat 9.0, but we prefer not to have a screen saver
function. We removed it by selecting the Start Here icon (on the gnome desktop). We then
selected Preferences, Screensaver, and Mode: Disable Screen Saver.

Boot diskette
If you did not create a boot diskette during Linux installation, we suggest you create one now.
It is useful if something overlays the master boot record on your hard disk. Log into Linux as
root, insert a blank diskette,6 and enter this command:
# /sbin/mkbootdisk 2.4.20-8

This command places the stock Red Hat 9.0 kernel on the diskette. It appears that this
command only works with vmlinuz kernels, not bzImage kernels. You can look in /boot to
determine exactly which vmlinuz kernel name to use. We had /boot/vmlinuz-2.4.20-8 and this
accounts for the operand for the mkbootdisk command.

Ownership of /s390
We created a filesystem with mount point /s390. This file system is intended to hold our S/390
emulated volumes when they are installed. For this plan to work with FLEX-ES, the /s390
mount point needs to be owned by user flexes. However, user flexes is not created until we
install FLEX-ES. We must remember to change ownership of /s390 after installing FLEX-ES.

3.3 Kernel rebuild

You may need to build a new Linux kernel for any of several reasons. These include:
򐂰 You have an SMP system. The stock SMP kernel with Red Hat releases cannot be used
with FLEX-ES.7
򐂰 You have large PC memory that requires special Linux handling, including selection of the
HIGHMEM I/O option.
򐂰 You have a LAN adapter not supported by your existing kernel.
򐂰 You have a stock kernel that does not run FLEX-ES.
򐂰 If you have updated the ips driver (for ServeRAID).
򐂰 If you want the latest LVM support.

5
This patch will be incorporated into the Linux kernel starting with release 2.4.21.
6
You may need to stop Linux, shut down your ThinkPad, and swap the CD-ROM drive for a diskette drive.
7 This may change in the future, of course, but this statement is true for all Red Hat releases up to the time of writing.

30 EFS Systems on a Linux Base: Getting Started

FSI provides Linux kernel source trees containing the options and FSI patches required for
FLEX-ES. FSI has provided these at the 2.4.17, 2.4.19, 2.4.20, and 2.4.21 kernel levels. The
following description assumes that you are using the 2.4.21 level, but should apply to other
levels with minor file name changes.

The FSI source tree already contains the FSI patches necessary for FLEX-ES SMP support.
It may not have the options you need for other cases, such as certain LAN adapters or large
memory. In these cases, you need to make additional kernel parameter changes as you work
through the installation steps. Kernel builds made with source trees not supplied by FSI are
not supported, especially for SMP systems.

Important: Your kernel source code must match your kernel or you will be unable to install
and use FLEX-ES.

The following material was developed by Gary Eheman and Michael Ryan, of Fundamental
Software, Inc.
1. If you plan to install the latest LVM support, obtain the necessary updates before starting
your kernel build. To obtain the updates, see 3.4.2, “LVM updates” on page 35.
2. If you plan to install the latest ServeRAID support, obtain the necessary modules before
you start your kernel build. This is briefly discussed in 2.3.2, “BIOS and ServeRAID
adapter updates” on page 21.
3. Obtain a copy of the 2.4.21 Linux kernel source tree, with the proper FSI patches and
options installed.8 This is typically in the form of a compressed tar file. We assume that it is
named linux-2.4.21.pset.tar.gz. For our examples, we assume it is on CD. Alternately, you
may download it.
4. Log into your Linux system as root. All the following work is done as root.
5. Untar the source tree:
# mount /dev/cdrom /mnt/cdrom Or let Linux automount do the mount
# cd /usr/src Put the source tree here
# tar -zxvf /mnt/cdrom/linux-2.4.21.pset.tar.gz Untar into current directory
6. Verify that a directory named linux-2.4.21.pset now exists, and change the “linux” symbolic
link to point to this directory. If the “linux” symbolic link does not exist, add it as shown (it
did not exist in our RH9.0 system):
# cd /usr/src Work in this directory
# ls -al See if linux-2.4.21.pset is there
# ls -ltr See if “linux” symbolic link exists
# rm linux Delete “linux” symbolic link if present
# rm linux-2.4 Delete this symbolic link
# ln -s linux-2.4.21.pset linux Make a new “linux” link
# ln -s linux-2-4-21.pset linux-2.4 Make a new link
# cd linux Move to this directory
# ls -al Verify that you see files and directories
7. Force a new kernel name to avoid duplicate names. The constant we use, flexes1, is
arbitrary. If you make multiple new kernels (and want to preserve all of them), you can
increment this name to flexes2, flexes3, and so forth:
# cd /usr/src/linux If not already in this directory
# vi Makefile Use a different editor if you prefer
...
VERSION = 2
PATCHLEVEL = 4

8 Your FSI Business Partner, or whomever you obtained your FLEX-ES software from, should help you with this.

Chapter 3. Linux installation 31

 SUBLEVEL = 21
EXTRAVERSION = flexes1 <== Change this line
Notice the space before and after the equal sign. Save the altered Makefile and exit from
the editor.
8. Save a copy of the .config file:
# cd /usr/src/linux If not already in this directory
# cp .config flexes.config Keep the FSI configuration parameters
# cat README | less (Might be interesting, but you are
not required to read it now.)
9. Begin the process of building a new kernel. Most of these steps produce large amounts of
screen output that you are not expected to read:
# make clean Expect about one screen of output
# make mrproper Expect several screens of output
10.At this point, make the kernel modifications for new ServeRAID modules. These steps are
described in 3.4.1, “ServeRAID updates” on page 35. If you use this step, you must
coordinate it with the updating of ServeRAID firmware as described in 2.3.2, “BIOS and
ServeRAID adapter updates” on page 21.
11.At this point, perform kernel modifications for new LVM modules. These steps are
described in 3.4.2, “LVM updates” on page 35.
12.Make any kernel configuration changes that are needed:
# make xconfig Several compiles, much output
13.This step should result in a GUI screen with many buttons:
a. Click Load Configuration from File (lower right part of the display).
b. For the file name, enter /usr/src/linux/flexes.config and click OK.
You need to load the FLEX-ES starting configuration file before you start looking at the
options. The following table has abbreviated titles in some of the entries to avoid using a
very small type font. Many of the buttons on this screen open windows that have additional
buttons that open other windows, and so forth. It is safe to “look around” in the options and
windows. Do not change anything unless you are fairly certain you know what you are
doing.

Code maturity level options SCSI support File systems

Loadable module support Fusion MPT device support Console drivers
Processor type and features IEEE 1394 (Firewire) support (EX Sound
General setup I2O device support USB support
Memory Technology Devices (MT Network device support Bluetooth support
Parallel port support Amateur Radio support Kernel hacking
Plug and Play configuration IrDA (infrared) support Library routines
Block devices ISDN subsystem
Multidevice support (RAID, LVM) Old CD-ROM drives (not SCSI) Save and Exit
Networking options Input core support Quit Without Saving
Telephony Support Character devices Load Configuration from File
ATA/IDE/MFM/RLL support Multimedia devices Store Configuration to File

14.This interface lets you make kernel option changes. Again we stress that you should not
change anything unless you know what you are doing. The FSI patches and options are
already applied, and the kernel is configured for FLEX-ES SMP use. A few options you
may need to change are:

32 EFS Systems on a Linux Base: Getting Started

 – Large memory. Select Processor type and features. Click the button labeled High
Memory Support. This should open a menu with three options: off, 4 GB, or 64 GB. If
your machine has less than 1 GB, select the off option. For 1 to 4 GB memory, select
the 4GB option. If your machine has more than 4 GB memory, select the 64 GB
option. The 64 GB option produces a kernel that works only on Intel IA32 processors.9
– HIGHMEM I/O support. If you select the 4 GB or 64 GB large memory option, you
should select this also, using the y option. It is on the Processor type and features
panel.
– Newer IBM processors may have 1000BaseT LAN adapters. These may not be
enabled with the default kernel options. Select Network device support on the GUI
kernel configuration panel, and then on the next panel, select Ethernet (1000 Mbit).
The final panel lists the 1000 megabit devices supported. Recent IBM systems may
have a Broadcom Tigon3 adapter (servers) or an Intel PRO/1000 Gigabit Ethernet
(T40 ThinkPads) and you see it listed on the panel. Select m and then OK for
support.10 You can select more than one adapter option.
– If you have a built-in Ethernet adapter using the Intel PRO/100 VE Network Connector
(used on T23 ThinkPads, for example) or the Dual-port IBM 10/100 adapter, select
Network device support and then select Ethernet (10 or 100Mbit). In this option, set
Ether Express pro/100 (e100, Alternate Intel Driver) to m.
15.After you make any kernel option changes, complete these steps:
a. Click Store Configuration to file and select a new name. We used
/usr/src/linux/flexes.config2. You may not need this file, but a backup may be useful
sometime.
b. Click Save and Exit.
16.Run the actual kernel-building steps. These steps take some time and produce many
screens of output, including various warning messages.
# make dep To make kernel dependencies
# make bzImage Use an upper case “i”
# make modules Longest step
# make modules_install
17.Copy the bzImage and System.map files to /boot:
# cp /usr/src/linux/arch/i386/boot/bzImage /boot/bzImage-2.4.21flexes1
# cp /usr/src/linux/System.map /boot/System.map-2.4.21flexes1
18.There is a System.map symbolic link in /boot. Make a new one that points to the new map
file. If the current System.map is a file instead of a symbolic link, then rename the file to
something like System.map.old.
# cd /boot
# ls -l System.map Verify old one is present
# rm System.map Remove old symbolic link
# ln -s System.map-2.4.21flexes1 System.map New symbolic link
19.Make a ram disk image.
# cd /boot If not already in this directory
# /sbin/mkinitrd -v initrd-2.4.21flexes1 2.4.21flexes1
This command looks in various standard directories for modules related to 2.4.21flexes1
(the second operand). Then it builds a ram disk using the name you specify in the first
operand. If you built a previous image with the same name, you must delete it before you
run mkinitrd again.
9
We used ThinkPads with 1 GB memory using kernels with the off option. This may produce slightly degraded
performance, but it was not observable.
10 This simply adds support to the kernel. It does not require you to have the adapter installed.

Chapter 3. Linux installation 33

 20.Update /boot/grub/grub.conf to include the new kernel. Ignoring the comment lines, our
original /boot/grub/grub.conf looked like this:
default=0
timeout=10
splashimage=(hd0,0)/grub/splash.xpm.gz
title Red Hat Linux (2.4.20-8)
root (hd0,0)
kernel /vmlinuz-2.4.20-8 ro root=LABEL=/
initrd /initrd-2.4.20-8.img
We edited the file to look like this:
default=0
timeout=10
splashimage=(hd0,0)/grub/splash.xpm.gz
title Red Hat Linux (2.4.20-8)
root (hd0,0)
kernel /vmlinuz-2.4.20-8 ro root=LABEL=/
initrd /initrd-2.4.20-8.img
title New kernel (2.4.21)
root (hd0,0) (Copy numbers from original grub entry)
kernel /bzImage-2.4.21flexes1
initrd /initrd-2.4.21flexes1
Save the file and exit from your editor. Your hd0,0 parameters may be different, especially
if you have a dual boot system. Copy the parameters from the first title line in
/boot/grub/grub.conf. (Note that there may be symbolic links from other names to
/boot/grub/grub.conf; you can work through any of these links.)
21.Shut down your Linux cleanly and reboot.
22.When the Red Hat splash screen appears, you should see two options: one for the original
kernel and one for your new kernel. Use the down arrow key to select the new kernel. You
need to do this fairly quickly because your grub parameters specify a 10 second timeout
with the default selection of the original kernel. You can change these in
/boot/grub/grub.conf:
default=1 (Make second kernel the default)
timeout=12 (Give user 12 seconds to manually select)
Note that the “title” sections in the grub configuration are counted starting with zero.
23.Verify that you are running with your new kernel:
# cd /proc
# cat version
The text in this line contains “2.4.21flexes1” if you are running with your new kernel.11

You can build and test an SMP kernel on a single processor system. We performed the
complete kernel rebuild just described on ThinkPads and on an xSeries SMP.

Rebuild FLEX-ES module

If you previously installed FLEX-ES, you must rebuild one of the modules for your new kernel.
(Be certain you are running with the new kernel when you do this!) Do the following under the
flexes user ID:
$ cd /usr/flexes/modules Do this as flexes, not root
$ make clean
$ make all
(reboot to pick up the new msgmgr module)

11 The “flexes1” portion of the name reflects whatever constant you specified in the EXTRAVERSION name.

34 EFS Systems on a Linux Base: Getting Started

3.4 Additional kernel steps
You can insert these optional steps into the appropriate place when you rebuild the kernel.

3.4.1 ServeRAID updates

For our sandbox projects used when writing Redbooks, we do not use the Linux ServeRAID
management programs. We build a RAID array and create a single logical partition by using
the stand-alone programs on the ServerGuide CD. We then install Linux with no additional
steps. This has worked for our sandbox projects, but does not provide the full RAID
management needed for a production FLEX-ES system. A production installation should
include the most current ServeRAID updates and Linux support.

You should have a ServeRAID CD ROM and install BIOS and ServeRAID firmware updates
from the same CD. That is, the BIOS, firmware, and Linux updates should all be at the same
level. There are extensive instructions on the ServeRAID CD for installing the upgrades and
Linux commands. If you follow the instructions, you can add these upgrades and commands
after Linux is installed.

You can shorten the process by installing the upgrades at the same time you are rebuilding
the kernel. FSI uses the following steps, at the appropriate time, to do this:
# mount /dev/cdrom /mnt/cdrom (Or let automount do it for you)
# cd /mnt/cdrom/linux
# cp cmdline/ipssend /usr/local/sbin (Get the command line utility)
# chmod 500 /usr/local/sbin/ipssend (Only root can use it)
# rpm --install manager/RaidMan-* (Install ServeRAID manager program)
# mkdir /root/ips/ (Make a working directory)
# cp /mnt/cdrom/linux/driver/ips-610.tgz /root/ips
# cd /root/ips
# tar -zxf ips-610.tgz
(We need only the ips.c and ips.h files. Ignore the patch files)
# mv /usr/src/linux/drivers/scsi/ips.c /usr/src/linux/drivers/scsi/ips.c.orig
# mv /usr/src/linux/drivers/scsi/ips.h /usr/src/linux/drivers/scsi/ips.h.orig
# cp /root/ips/ips.? /usr/src/linux/drivers/scsi/
# ls -l /usr/src/linux/drivers/scsi/ips*
(Verify that ips.c, ips.h, ips.c.org, and ips.h.org are present)
# cd /usr/src/linux (Correct place for next kernel step)

After you do this, return to the normal kernel building process, where the next step is to make
xconfig.

3.4.2 LVM updates

The LVM modules in Red Hat 9.0 work well. However, for the best possible operation of a
production FLEX-ES system, should install the latest LVM version. At the time of writing, this
was 1.0.7. You must integrate the new modules into a new Linux kernel as part of the kernel
building process. To install the latest version, obtain the material before you start building a
new Linux kernel:
1. With a browser (which need not be running under Linux), go to:
http://www.sistina.com
2. Click Products.
3. Click Logical Volume Manager for Linux (LVM).
4. Click Download.
5. Click Download LVM 1.0.7 or later, if available.
6. Select lvm_1.0.7.tar.gz (372 KB). We copied this to a DOS diskette.

Chapter 3. Linux installation 35

 Unpack the tar file in the /root directory of your Linux system:
# mcopy a:lvm_1.0.7.tar.gz /tmp/lvm_1.0.7.tar.gz (Assume it is on a DOS diskette)
# cd /root
# tar -zxvf /tmp/lvm_1.0.7.tar.gz

At the appropriate place in your kernel build (after make clean and make mrprogram), run the
following commands:
# cd /root/LVM/1.0.7
# ./configure
# make
# cd /root/LVM/1.0.7/PATCHES
# make
# cd /usr/src/linux
# patch -p1 < /root/LVM/1.0.7/PATCHES/lvm-1.0.7-2.4.21flexes1.patch
patching file include/linux/lvm.h
...
patching file drivers/md/lvm-fs.c
# cd /root/LVM/1.0.7
# make install (Install the new LVM)
# cd /usr/src/linux (Correct place for next kernel step)

The file name in the PATCHES directory reflects the name of the kernel that you are building
and may differ from the name shown here. After doing these steps, return to the normal
kernel building process, where the next step is to make xconfig.

3.5 Additional Linux steps

Red Hat 9.0 may not automatically use DMA for disk accesses on the ThinkPad. This varies
for different ThinkPad models. This has a drastic impact on performance. You should perform
the following actions after logging into the system as root:
# hdparm -d /dev/hda (Check DMA usage)
using_dma = 0 (off) (This is bad)
using_dma = 1 (on) (This is good)
(If using_dma = 0, then change it with the following command)
# hdparm -d1 /dev/hda (Change to dma = on)
# hdparm -d /dev/hda (Verify that it worked)

This should make an immediate change in disk performance. You can use hdparm -d0
/dev/hda to turn off DMA again. You can run a little benchmark (and try it with DMA on and
off) by entering:
# hdparm -t -T /dev/hda

You can make the change permanent by editing /etc/rc.d/rc.local and including this line in the
file:
hdparm -d1 /dev/hda

We checked our CD drive in the Ultrabay (hpdarm -d /dev/hdc) and found that it already had
DMA turned on.

The only diskette drive we found for a T40 ThinkPad was connected through the USB ports.
The simple DOS diskette commands (mdir, mcopy) could not find the drive.12 We used it via a
mount command:
# mount /dev/sda /mnt/floppy (Mount diskette as a file system)

12 You may be able to use these commands by editing /etc/mtools.conf and making appropriate changes.

36 EFS Systems on a Linux Base: Getting Started

 Remember to umount a diskette when you are finished with it.

3.5.1 Red Hat Linux updates

Red Hat offers many fixes and updates through the Red Hat Network. The Linux up2date tool
provides an interface to obtain these updates. We make the following recommendations:
򐂰 Do not obtain kernel updates or replacements through this process unless you are willing
to verify that the new kernel works properly with FLEX-ES. Remember that you must keep
the kernel source tree in your system synchronized with the working kernel. You must also
reinstall the FLEX-ES msgmgr module after any kernel change. Also remember that the
FSI kernel source includes FSI patches that are needed for use of FLEX-ES on an SMP
system.
򐂰 Obtaining and applying fixes and updates for various Linux packages is acceptable
provided they have no kernel dependencies that may require a kernel change.

The owner of a production system may want to work with up2date and obtain updates in an
orderly manner followed by appropriate system testing. For sandbox systems, we suggest
that you do not obtain updates unless needed for a specific problem.

Chapter 3. Linux installation 37

38 EFS Systems on a Linux Base: Getting Started
 4

Chapter 4. FLEX-ES installation

This chapter describes the installation process for FLEX-ES. System definitions, resource
definitions, and operation are described later. The same installation process is used for
ThinkPad and xSeries systems.

FLEX-ES is very easy to install. However, we must stress the importance of having a good
Linux installation before you start working with FLEX-ES. There are areas of special concern:
򐂰 Is the X Window System working correctly? Can you log into the system with the graphics
interface? Does gnome (or kde) appear to work correctly? Can you open and close terminal
windows with proper responsiveness? If not, you need to resolve these problems now.
򐂰 If you are using an SMP system, have you built a new kernel using a source tree provided
by FSI?
򐂰 Have you installed the required Red Hat components, such as Development Tools and
Kernel Development? (If you elect to install Everything during Red Hat installation, this is
not an issue.)

This chapter describes the installation of FLEX-ES Release 6.2.14. The general process
should match any recent FLEX-ES release.

Attention: You must install and use FLEX-ES with a Linux kernel that matches the Linux
kernel source tree in your system. If you have not installed new kernel source, or have not
downloaded anything that changes the stock kernel, this should not be a concern. (The
msgmgr module of FLEX-ES is sensitive to this pitfall.)

© Copyright IBM Corp. 2003. All rights reserved. 39

4.1 FLEX-ES installation steps
FLEX-ES can be delivered a number of ways:
򐂰 FTP over the Internet
򐂰 CD-ROM
򐂰 Preinstalled by a Business Partner

The FLEX-ES package is not large (about 4 MB), and an FTP download is easy. In addition
to the FLEX-ES code, you need a FLEX-ES license key file and dongle. The license key file is
a few hundred bytes and is normally shipped with your system. For our examples, we
assume that you have FLEX-ES on a CD.

The FLEX-ES distribution file is normally in tar format. You want to untar the file and place the
results in the /tmp directory:
# mount /dev/cdrom /mnt/cdrom (Linux may do the CD mount automatically)
# cd /mnt/cdrom
# ls
flexes-6.2.14.i386.rpm.tar (Or a name similar to this)
# cd /tmp (Untar into /tmp)
# tar xvf /mnt/cdrom/flexes-6.2.10.i386.rpm.tar (If working from CD)
# umount /dev/cdrom (CD not needed any more)

If you are downloading FLEX-ES, simply download the tar file into /tmp and the rest of the
instructions here should apply. Your exact file names may be different, depending on the
FLEX-ES release you are using. The tar command should produce several files in /tmp:
# ls
flexes-6.2-14.i386.rpm (Primary FLEX-ES modules)
msgmgr-6.2-14.i386.rpm (FLEX-ES kernel communication)
ftlib-6.2-14.i386.rpm (Source for creating programs using FakeTape)

Again, your exact file names may differ but the general names are as shown here. You must
be root to use these rpm commands.
# cd /tmp (Location of FLEX-ES rpm files)
# rpm -i flexes-6.2-14.i386.rpm
/var/tmp/rpm -tmp.66303: /usr/bin/expect: No such file.....
(This message means the installation process failed to set a password for user flexes.
You can later set the password manually)
# rpm -i msgmgr-6.2-14.i386.rpm
gcc ......... (Inspect any messages carefully)
ld .......... (Inspect any messages carefully)
# rpm -i ftlib-6.2-14.i386.rpm

These three rpm commands perform the basic FLEX-ES installation. Installation of msgmgr
involves compilation and link steps. These must run without errors. You see a long gcc
command and a long ld command displayed. There should be no error messages associated
with these commands. If there is an error here, FLEX-ES is unable to load or start msgmgr
later.

If you chose to install “Everything” when you installed Linux, the system has the
/usr/bin/expect file and password “abcdef1” is automatically set by the rpm processing. If you
selectively installed a smaller Linux system, the /usr/bin/expect file is probably not present
and you need to manually set a password for user ID flexes. This is easy to do and no side
effects are involved.

A userid and a groupid named flexes are created automatically during FLEX-ES installation.
All normal FLEX-ES operation is done under this userid. The FLEX-ES installation process

40 EFS Systems on a Linux Base: Getting Started

 attempts to assign the initial password “abcdef1” for userid flexes. In some cases the
FLEX-ES install process cannot set the password. If this happens, you can use:
# passwd flexes
abcdef1 (After prompt to enter password)
BAD PASSWORD: it is too simplistic/systematic (You can ignore this warning)
abcdef1 (After second prompt to enter password)

The password is accepted, regardless of the message.

FLEX-ES installation creates /usr/flexes and makes this the home directory for userid flexes.

4.1.1 Default PATH

We recommend a minor step that we think makes operation easier:
# cd /usr/flexes
# mkdir rundir (Conventional location for FLEX-ES files)
# chown flexes:flexes rundir (Make flexes the owner)
# vi /root/.bash_profile
...
...
USERNAME=”root”
PATH=$PATH:/usr/flexes/bin <=== Add this line
export USERNAME BASH_ENV PATH

Adding /usr/flexes/bin (where all the FLEX-ES executables are located) to PATH makes
operation easier.1 IBM Redbooks dealing with FLEX-ES assume that you create directory
/usr/flexes/rundir and store your FLEX-ES operational files in this directory. This is not
required, but we assume this directory location in the remainder of this redbook.

4.1.2 Ownership of /s390 mount point

For our disk allocation plan, we created a file system with mount point /s390. This file system
is used for our emulated S/390 disk volumes. This mount point must be owned by user flexes,
and this userid was created automatically by the FLEX-ES installation process. Following our
instructions, the /s390 mount point was created during Linux installation and is owned by root
at that time. This needs to be changed. Working as root, enter:
# chown flexes:flexes /s390

4.1.3 Location of x3270 program

Most Red Hat installations contain an x3270 program. FLEX-ES installs a different x3270
program that is found in /usr/flexes/bin. Using the wrong x3270 results in this message when
you attempt to start x3270 when the flexes program2 is running:
Error: app-defaults version mismatch: want 3.2.18, got 3.1.1.0

We need to ensure that the correct x3270 program is used. A way to do this is to always use
a full path name for it, but this is awkward. Another way is to rename the Red Hat x3270 to a
different name. If /usr/flexes/bin is added to the standard search path, this causes the
FLEX-ES version of x3270 to be found. You can enter:
# cd /usr/X11R6/bin Where old x3270 is found
# mv x3270 x3270RH Rename x3270

1
The /usr/flexes/.bash_profile already contains PATH=$PATH:$HOME/bin where $HOME is /usr/flexes.
2 This is the main FLEX-ES program that provides S/390 emulation.

Chapter 4. FLEX-ES installation 41

 The FSI version of x3270 was modified to allow session binding via the device names that
you place in the FLEX-ES definitions (or in a FLEX-ES mount command).

4.2 Installing the FLEX-ES license key and dongle

Connect the dongle to the USB port. You can check the dongle with the following command:
# cat /proc/bus/usb/devices

On our system, the cat results were as follows:

T: Bus=01 Lev=00 Prnt=00 Port=00 Cnt=00 Dev#= 1 Spd=12 MxCh=2

B: Alloc = 0/900 us ( 0%), #Int= 0, #Iso= 0
D: Ver= 1.00 Cls=09(hub ) Sub=00 Prot=00 MxPS= 8 #Cfgs= 1
P: Vendor=0000 ProdID=0000 Rev=0.00
S: Product=USB UHCI Root Hub
S: SerialNumber=1860
C:* #Ifs= 1 Cfg#= 1 Atr=40 MxPwr= 0mA
I: If#= 0 Alt= 0 #EPs= 1 Cls=09(hub ) Sub=00 Prot=00 Driver=hub
E: Ad=81(I) Atr=03(Int.) MxPS= 8 Ivl=255ms
T: Bus=01 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#= 2 Spd=12 MxCh= 0
D: Ver= 1.10 Cls=00(>ifc ) Sub=00 Prot=00 MxPS=16 #Cfgs= 1
P: Vendor=1404 ProdID=cddc Rev= 1.00
S: Manufacturer=Fundamental Software Inc.
S: Product=FSI Dongle
C:* #Ifs= 1 Cfg#= 1 Atr=80 MxPwr=100mA
I: If#= 1 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=serial
E: Ad=81(I) Atr=02(Bulk) MxPS= 16 Ivl= 0ms
E: Ad=01(0) Atr=02(Bulk) MxPS= 16 Ivl= 0ms

If your output is something like this and includes lines about the FSI dongle, you can assume
the dongle is being recognized. The following command is needed to make the dongle
available to FLEX-ES software:
# /sbin/modprobe usbserial vendor=0x1404 product=0xcddc

Your system should already include this command. Look in file /etc/init.d/FLEXES and see if
this command is present in the file. If it is not, add it. In an early system, we added it just after
the “loadable drivers” comments:
....
#
# Install our loadable drivers
#
/sbin/modprobe usbserial vendor=0x1404 product=0xcddc <== add this line if needed
#
if [ -f /usr/flexes/modules/msgmgr.o ]; then
....
....

You need a FLEX-ES license key, which is a Linux file of a few hundred bytes. The key must
match the dongle (via an encoded serial number in the key and dongle). Copy the license key
file to /var/adm/flexes/.flexeslicense. (Note that the target file name begins with a period.) If
the license key is provided on diskette, you need to copy it from there, for example:
(shut down Linux if you need to swap Ultrabay devices)
(remove the CD-ROM drive from the Ultrabay of the ThinkPad)
(insert the diskette drive in the Ultrabay)
(boot Linux and log in as root)

42 EFS Systems on a Linux Base: Getting Started

 # mdir a:
(if the diskette is in DOS format, you should see the file with your key)
# mcopy a:yourkey987.6.2.14.key /var/adm/flexes/.flexeslicense
(done)
(if the diskette is not in DOS format, try the following)
# mount /dev/fd0 /mnt/floppy (For a T40, it is /dev/sda)
# ls -al /mnt/floppy
(if the key file is displayed, copy it with the following command)
# cp /mnt/floppy/yourkey987.6.2.14.key /var/adm/flexes/.flexeslicense

ThinkPad models are updated and changed frequently. The exact method of changing from
CD-ROM use to diskette use depends on your ThinkPad model. In some models, you can
have both a CD-ROM drive and a diskette drive active at the same time. In this case, there is
no need to shut down Linux to swap drives.

If you have multiple FLEX-ES licenses, be certain to identify the correct license with the
correct dongle. There is no visible way to associate a license file with a specific dongle. Each
dongle has a serial number on it, but you do not see this serial number in the license file. If
you have multiple EFS systems, you need to devise a way to keep track of which dongle goes
with which license file.

The Linux kernel that is provided with the Red Hat 9.0 system should recognize and use the
USB dongle, as described. However, if you build your own kernel, it may not use the dongle
correctly. If you build a kernel, configure the USB Serial Converter as a module and say yes to
USB Generic Serial Driver support. The kernel source trees provided by FSI already have this
configured.

The license key file is an ASCII file, with the contents masked (scrambled). Do not attempt to
edit it. Also, the current license key files are sensitive to the FLEX-ES release level. If the
license key file (before you copy it to the /var/adm/flexes directory) has a name like xxxx6.2.14
then that license is for FLEX-ES Release 6.2.14. It can be used for any earlier FLEX-ES
release but not for a later release. If you want to use a later FLEX-ES release, you need a
new license key file.

4.3 Reboot
Installing (or updating) FLEX-ES changes modules that are loaded during Linux booting. You
should reboot after you install or update FLEX-ES.

4.4 Installing FLEX-ES upgrades

Installing a new FLEX-ES release under Linux involves a few rpm commands. You must be
root to use these rpm commands.
# mount /dev/cdrom /mnt/cdrom (Or let automount do it)
# ls /mnt/cdrom
flexes-6.2.14.i386.rpm.tar (Or a name similar to this)
# cd /tmp (Untar into /tmp)
# tar xvf /mnt/cdrom/flexes-6.2.14.i386.rpm.tar (If working from CD)
(This untars the new release into your /tmp directory)
# umount /dev/cdrom (Not needed any more)
# rpm -e ftlib (Remove existing FLEX-ES installation)
# rpm -e msgmgr
# rpm -e flexes
# cd /tmp (If not already there)
# rpm -i flexes-6.2-14.i386.rpm

Chapter 4. FLEX-ES installation 43

 # rpm -i msgmgr-6.2-14.i386.rpm
(Check these messages!)
# rpm -i ftlib-6.2-14.i386.rpm

(Reboot Linux)
(Use cfcomp to recompile your FLEX-ES definitions)

The old FLEX-ES packages are removed (with the -e option) in the order shown, and the new
packages installed in the order shown. Your package names may be different (to reflect new
release numbers). When removing the old packages, you can ignore a message about not
removing the /usr/flexes directory.

You can use the rpm -U option for upgrades, instead of the separate rpm -e and rpm -i
commands. We show the separate commands to more clearly illustrate that an upgrade is a
complete replacement of the FLEX-ES packages. Using the -U option is a more normal way
to install an upgrade.

Also, you may need to recompile your system and resource definitions (using cfcomp) after
installing a new FLEX-ES release.3 This takes a few seconds and we suggest that you do it
after you install any new FLEX-ES release. You may also need to reinstall any x3270
keyboard modifications you made.

After upgrading FLEX-ES, you should reboot Linux. One of the FLEX-ES modules is
automatically loaded during booting and you want to have the new version of this module
active.

3
If your system starts resadm automatically during the boot processes, you should reboot after recompiling your
resource definitions. The setup described in this redbook does not start resadm automatically, but many production
FLEX-ES systems installed by FSI Business Partners do start resadm automatically.

44 EFS Systems on a Linux Base: Getting Started

 5

Chapter 5. S/390 operating system

installation
In principle, you can install a S/390 operating system on your EFS system using any of the
distribution media and logical packaging available within IBM (for IBM internal users) or from
IBM.1 These include, but are not limited to:
򐂰 Tapes in dump/restore format, provided you also have an appropriate restore utility. This
utility might reside on an already-installed S/390 operating system or you may be able to
IPL it from a stand-alone tape. Of course, you need a tape drive on your EFS system to
use this method.
򐂰 Tapes in typical IBM distribution formats, such as ServerPacs. These assume you already
have a working S/390 operating system on your system. This working system might have
come from a stand-alone restore operation.
򐂰 CDs containing images of S/390 volumes that contain a S/390 operating system. IBM has
made several of these available to members of the S/390 PartnerWorld for Development
organization who obtained their systems through this program. Other CD-resident
distributions have been created for IBM internal use. There are three general formats of
these CDs (all of which are compressed in some way so that a 3390 volume fits on a CD):
– Zipped AWSCKD: This format was originally created for IBM P/390 systems and has
been used for IBM Multiprise® 3000 and other systems. It is the current format for the
AD CD-ROM systems described below. FLEX-ES provides a utility to convert this
format into its own emulated disk format.
– Compressed tar or gzip: These are copies of emulated S/390 volumes from a working
Linux-based FLEX-ES system. The Linux file containing the emulated volume is tarred,
compressed, or both and then written to CD. These are easy to create but are not
directly suitable for restoring to raw disk devices.
– Compressed ckdbackup files: These can be restored as simple Linux files or as raw
disk device files. This is the recommended format for future distributions.

1
This statement assumes that you have a tape drive on your EFS system that is compatible with the IBM-provided
media, of course.

© Copyright IBM Corp. 2003. All rights reserved. 45

 򐂰 Emulated tapes (on CD or other media) can be used for distributions or stand-alone
utilities. FLEX-ES supports FakeTape,2 AWSTAPE, and OMA/2 formats. OMA/2 is an
optional distribution format for z/VM.
򐂰 It is possible to FTP or otherwise transfer images of S/390 volumes from other
Linux-based FLEX-ES systems. This can be a practical method for a classroom setup, for
example.

This chapter describes installation based on four of these methods, with emphasis on the AD
CD-ROM systems because that appears to be the most common distribution used by EFS
owners. An expanded description of ckdbackup for creating and installing distributions
appears in Chapter 5 in S/390 PartnerWorld for Developers, ITSO/EFS Project EFS Systems
on a Linux Base: Additional Topics, SG24-7008.

Required documentation
We assume that an operating system distribution provides some documentation. For
example, a usable z/OS system contains an IODF (or several IODFs) that determine which
S/390 addresses (device numbers) are used for various types of devices. If we place our
3390 IPL volume at address 0A80, this is because the IODF specifies address 0A80 is for a
3390 device type. There is no way to guess such addresses. IODFs are typically built with a
large number of addresses for various device types. You can select among appropriate
addresses for your EFS operation, but you require a list of addresses “generated” in your
system.

z/VM is more flexible and automatically senses the addresses of more modern device types.

You may also need to know the IPL parameter needed by your S/390 operating system.

Linux file names

An emulated S/390 volume is a simple Linux file.3 As such, it has a Linux file name. These
files can be renamed with the mv command, copied to other locations with cp, and treated as
normal (but large) Linux files. All our examples place the emulated volumes in a file system
with mount point /s390. We use the volser of the emulated 3390 volume (in upper case) as
the Linux file name. For example, if we have a 3390 with volser SYSRES, the Linux file
containing this volume would be /s390/SYSRES. We find this convenient, but it is completely
arbitrary. You can place the files anywhere in the Linux file system, using any file names you
like. You must have the correct file names in the FLEX-ES definition files.

5.1 AD CD-ROM distributions

In practice, most of the IBM PartnerWorld for Developers (PWD) members use prepackaged
systems distributed on CD-ROMs. The only recent package available is the z/OS AD
CD-ROM system. This is available only to PWD members who obtained systems through the
PWD organization.

An AD system provides a very easy way to install a useful z/OS system. You could build a
system starting with a ServerPac, in the same way most z/OS customers build their systems.
However, this requires considerably more time and effort. It would also require a hardware
system with tape drives compatible with the ServerPac media.

In general, a z/OS AD system is a rather straightforward implementation of z/OS. The

experience and results of using it on an EFS system is about the same as using any other

2
FakeTape is a trademark of Fundamental Software, Inc.
3 This section does not apply if you are using raw devices for emulated volumes.

46 EFS Systems on a Linux Base: Getting Started

 straightforward z/OS implementation. The AD CD-ROM systems, as the name implies, are
distributed on CD-ROMs. The CD-ROMs are not seen by the S/390; they are processed by
Linux programs.

Basic CD-ROM formats

The z/OS volumes (on the AD CD-ROM) are zipped files in AWSCKD format. AWSCKD is a
P/390 device manager program that emulates 3380 and 3390 devices. In general, a
complete 3390 volume is in one or two PC files. All the AD releases of OS/390 and z/OS to
date are in this format.

The AWSCKD format is, in essence, a complete image of a S/390 disk volume. Within an
AWSCKD file CKD tracks, cylinders, R0s, and so forth are defined. There is a standard label,
a VTOC, probably a VTOC index, and whatever data sets appear on that volume in a z/OS
context. The data is in S/390 format. Text contained on a S/390 volume is EBCDIC and
executables are S/390 binary files suitable for execution by z/OS. To a PC program, an
AWSCKD file is simply a large binary file that is not useful to typical PC programs.

AWSCKD files on the CD-ROM are in ZIP format simply to save space. It is usually possible
to ZIP an AWSCKD 3390-3 (2.8 GB) so that it fits on a CD-ROM (about 600 MB). There is no
basic requirement that ZIP files be used and, in rare cases, disk images might not be zipped.

The AWSCKD format was developed for P/390s, where the underlying operating system used
to emulate CKD drives is OS/2. OS/2 is a 32-bit operating system and has the restriction that
a single file cannot be larger than 2 GB. An AWSCKD-emulated 3390-3 requires more than
2 GB, and is split into two AWSCKD files, the first is 2 GB and the second is about .8 GB4.
The two files are usually placed in a single ZIP file.

FLEX-ES formats
FLEX-ES emulates CKD drives, but the internal format (in the Linux files used for emulation)
is different than the AWSCKD format. FLEX-ES provides a utility (ckdconvaws) to convert the
AWSCKD format to the FLEX-ES format for emulated CKD drives.

5.1.1 z/OS AD CD-ROM addresses and IPL data

The AD CD-ROM documentation provides the addresses (device numbers) used for various
devices. A large number of addresses are provided. We use the following address ranges:
Device type Addresses
3390 DASD 0A80-0A9F
3270 MVS console 0700
3270 VTAM terminals 0701-071F
3172 (LAN) 0E20, 0E21
3480 tape drive 0560

Based on the documentation and addresses we selected, the IPL information is:
IPL address = 0A80 IPL parameter = 0A82CS

The AD system documentation indicates that four volumes are required for a minimum
IPLable system; the volsers are S4RES1, S4RES2, OS39M1, and S4USS1.5 To make our
system match existing AD system documentation, we installed these volumes at addresses
0A80, 0A81, 0A82, and 0A87. Additional volumes can be installed containing DLIBs, DB2®,

4 The two files contain the characters _1 and _2 as the last two characters of each file name. P/390 utilities

recognize that the second file is a continuation of the first. Unlike FLEX-ES, the P/390 does not emulate 3390-9
drives. If it did, it would use four 2 GB files plus a 1 GB file to equal the 9 GB contained on a 3390-9 drive.
5 These volsers are for the z/OS AD release current at the time of writing. Other releases will have different volsers.

Chapter 5. S/390 operating system installation 47

 CICS, IMS™, and WebSphere®. You should consult the AD CD-ROM documentation for
more information about device addresses, volumes available, and IPL options.

5.1.2 Installation tasks

If you let the Linux automount function mount CDs, you can perform the complete AD
installation under the flexes userid. Installation consists of:
򐂰 Inserting a CD in the drive (and waiting for Linux automount to recognize it)
򐂰 Using an unzip and ckdconvaws command, piped together
򐂰 Issuing an umount command to remove the CD
򐂰 Repeating this process for each CD

The automount window is convenient for confirming the contents of the CD. You might close
this window before you umount the CD.

If loading a CD on your system does not trigger an automount function you can issue the
appropriate commands. You may need to do this from a root window, and in this case using
two windows (one as root and one as flexes) is convenient:
# mount /dev/cdrom /mnt/cdrom (Issue from root)
(Process the CD from a flexes window)
# umount /dev/cdrom
(Remove the CD)

Installing a volume
The output of the unzip command can be piped directly to the ckdconvaws command:
(Insert CD and wait for automount to recognize it)
$ unzip -p /mnt/cdrom/zos14/s4res1.zip | ckdconvaws -r - /s390/S4RES1 3390-3
FSIDC146 Max head = 14, cyl = 3343, blks = 57
FSIDC180 Cylinder nnn completed in mmm milliseconds
FSIDC190 CKD Conversion Completed (3339 cyls copied)
(Close the automount window)
$ umount /dev/cdrom

The ckdconvaws command converts the AWSCKD format (used on the AD systems) to the
FLEX-ES format, and writes to the Linux file indicated.6 In this example, the volume being
unzipped is a 3390-3 (and we must tell the ckdconvaws program this). The second operand,
/s390/S4RES1, is the target location for the Linux file that will emulate the S/390 volume.
Enter the command as shown; notice the space before and after the single dash in the
second part of the command. Use the correct pipe symbol (long vertical line, sometimes
shown with two sections in the vertical line).

Command speed may appear erratic, with fast cylinder copies followed by long pauses. This
is due to Linux buffering large amounts of disk I/O in memory, and then flushing it to disk.
(This effect is not seen if you are using raw disk interfaces.)

3390-3 volumes are held in two files in the original AWSCKD format. Both these files are
usually placed in the same zip file. The ckdconvaws program handles this transition
automatically.

Repeat the same process for the other CDs that are to be loaded. For the z/OS 1.4s system,
we used:
CD CD File................. Linux File Disk Model
2 /mnt/cdrom/zos14/s4res2.zip /s390/S4RES2 3390-3

6 The -r option indicates that ckdconvaws should create the output file, if it does not already exist.

48 EFS Systems on a Linux Base: Getting Started

 3 /mnt/cdrom/zos14/os39m1.zip /s390/OS39M1 3390-3
4 /mnt/cdrom/zos14/s4uss1.zip /s390/S4USS1 3390-3

This produces an IPLable system.

Loading the remaining volumes

There were 14 volumes distributed with the z/OS 1.4s AD system we used. Only four volumes
are needed to IPL the system and use TSO. The remaining volumes are:
CD# volser 3390 model assumed address content
5 S4DIS1 3390-3 A85 system DLIBs
6 S4DIS2 3390-3 A86 system DLIBs
7 S4DIS3 3390-3 A88 system DLIBs
8 S4DB21 3390-3 A83 DB2 (including DLIBs)
9 S4CIC1 3390-2 A84 CICS (including DLIBs)
10 S4IMS1 3390-2 A89 IMS (including DLIBs)
11 S4WAS1 3390-3 A8A WebSphere
12 S4WAS2 3390-3 A8B WebSphere
13 SARES1 3390-3 A8C Stand-alone system
14 S4DIS4 3390-3 A8D system DLIBs

You may elect to install any or all of these additional volumes. If you want to install system
PTFs, for example, you almost certainly need the DLIB volumes. The SARES1 volume is also
a z/OS 1.4 system, but in a very minimal form that requires only one volume. It is meant as an
emergency recovery system or as a platform for customizing the main volumes while they are
not active.

Later AD releases may have different volumes (and have different volsers involved), but the
general principles should remain the same.

FLEX-ES definitions
Two FLEX-ES definitions (and matching shell scripts) are listed in Appendix A, “Sample
FLEX-ES operation files” on page 91. The defR14A definition includes only the four volumes
needed for an IPLable z/OS AD 1.4s system. The defR14B definition includes all the volumes
in the AD 1.4s distribution, plus several local volumes. You may need to modify these
definitions to match your volumes, IP address, or naming conventions.

Staged loading
For a ThinkPad with two hard disks and where the Ultrabay must be shared between a
CD-ROM drive and the second hard disk, you may need a few staging steps:
1. Install the CD-ROM drive. This may require shutting down and rebooting Linux.
2. Determine how much free space you have on the internal hard disk. A df -h command
can be used. Look in the /s390 file system, if you used our suggested disk layout.
Determine how many emulated volumes can be loaded in this space.
3. Install (mount, unzip/ckdconvaws, umount) volumes from CDs to fill the free space.
4. Remove the CD-ROM drive and install the second hard disk. This may involve shutting
down and rebooting Linux.
5. Copy the volumes from the internal hard disk to the second hard disk; for example:
$ cp /s390/S2DIS1 /s391/S2DIS1
$ cp /s390/S2DIS2 /s391/S2DIS2
This example assumes you are using our naming conventions. These copy commands
take several minutes for each file.
6. Delete the volumes from the internal hard disk, for example:

Chapter 5. S/390 operating system installation 49

 $ rm /s390/S2DIS1
$ rm /s390/S2DIS2
7. Repeat the process with more volumes from the CDs, if necessary.
8. After all the additional volumes are on the second hard disk, use an ls -al /s391
command to verify that they are owned by userid flexes.

Another way to stage volumes would be to copy zipped files from CD onto the first hard disk.
You could then unzip them onto the second hard disk.

5.2 Dump/restore distribution

We installed the Customized Offering Driver system (IBM product 5665-343, release 1.17).
This is often known as the COD system and it replaces what was formerly known as the
Starter System. It is a working z/OS system on two 3390-3 volumes. Only limited functions are
available. The COD system is a sufficient base for installing a z/OS ServerPac. That is, if you
have no other z/OS (or OS/390) system installed, you could first install the COD (using a
stand-alone restore) and then use the COD to customize and install a ServerPac. This
process is independent of any AD systems and is one of the startup processes that might be
followed by new z/OS customers who are not using AD CD-ROM packages.

For this installation we used a PC that had a SCSI-attached 3480/90 tape drive.

Although this section describes the installation of the COD system, the same general steps
apply to installing backup tapes of any system. In essence, the COD package is simply a
small z/OS system that has been backed up to tape (plus an additional tape containing two
stand-alone utility programs).

Our copy of the COD included:

򐂰 Four tapes containing 3390-3 volume D9ESY1
򐂰 Five tapes containing 3390-3 volume D9ECAT
򐂰 One tape with stand-alone utilities

We first studied the COD documentation (IBM order number GI10-0615-04, or later) to
determine what device addresses we could use.7 We found the following should work:
3390 disk drives at addresses 320, 321 (up through 33F, if needed)
3490 tape drive at address 390
3270 console for NIP and z/OS at address 0A1
3270 terminals for TSO at addresses 0F00-0F03

The COD has a large number of devices and addresses (“device numbers”) in its IODF and
our choices were somewhat arbitrary. Based on our choices, we created appropriate
FLEX-ES definitions and a startup shell script, shown in “Customized Offerings Driver
definitions” on page 96.

Included with the COD system is a tape with two stand-alone utilities. A stand-alone utility is
one that can be IPLed directly--without any operating system involved. The first utility on the
tape is Device Support Facilities (DSF) that is used to initialize a disk volume. The second
utility on the tape is DFSMSdss™, which is used to restore a volume from tape to disk.

7 If you are restoring some other system (other than the COD), you must determine the device numbers used by that

system. That is, you must know the devices (and device numbers) included in the IODF that will be used when you
IPL the new system. You must find this information from external sources; there is no easy way to determine it by
examining the tapes.

50 EFS Systems on a Linux Base: Getting Started

The general process we used is this:
1. Use a FLEX-ES utility to create two 3390 volumes on our EFS system.
2. Compile the FLEX-ES definitions and start a FLEX-ES instance.
3. IPL and use DSF (from the stand-alone utility tape) to initialize the volumes.
4. IPL and use DFSMSdss (from the tape) to restore the COD tapes.
5. IPL the COD system.

We started by creating two empty 3390-3 volumes:

$ ckdfmt -n -r /s390/TEMP01 3390-3
$ ckdfmt -n -r /s390/TEMP02 3390-3

These steps create two new 3390 volumes for FLEX-ES. The file names and locations are
arbitrary. Each file is about 2.8 GB. The following steps compile the FLEX-ES definitions,
start the FLEX-ES resource manager, and start S/390 emulation. These functions are further
explained in the next chapter.
$ cfcomp defCOD (Compile the FLEX-ES definitions)
$ su (Switch to root)
(password)
# resadm -s RCOD.rescf (Start the resource manager)
(pause for about 5 seconds)
# resadm -r (Verify that the resources started)
# exit (Leave root)
$ sh shCOD (Start a new FLEX-ES instance)
flexes> (Waiting for an IPL command)

The FLEX-ES definitions in file defCOD use the resource name RCOD and the system name
SCOD. At this point, we mounted the stand-alone utility tape in our SCSI tape drive:
flexes> ipl 390 (IPL from the tape)

(After tape activity stops, press Enter on any of the 3270 sessions started
by the shell script. Do not touch the tape drive. Do not rewind it.)
(The following messages are on the 3270 screen.)

CLEAR SCREEN WHEN READY (Should appear on the 3270 screen)

(Alt-c is the default x3270 combination for clearing the screen)
(You should receive a number of title messages from DSF)

ICK005E DEFINE INPUT DEVICE, REPLY ‘DDDD,CCUU’ or ‘CONSOLE’

(simply press Enter)8
ICK006E DEFINE OUTPUT DEVICE, REPLY ‘DDDD,CCUU’ or ‘CONSOLE’
(simply press Enter)
ENTER INPUT/COMMAND:
init unit(320) nvfy devtyp(3390) volid(D9ESY1) nomap
(more messages)
U (Reply U to initialize the volume)

ENTER INPUT/COMMAND:
init unit(321) nvfy devtyp(3390) volid(D9ECAT) nomap
(more messages)
U (Reply U to initialize the volume)

The two init commands you entered perform the volume initialization for the 3390 volumes.
At this point, switch to the window with the flexes prompt and IPL the tape again. If you did not

8
By Enter, we mean the 3270 Enter function. You may have customized your x3270 sessions to move this function to
the right Ctrl key. By default, it is the large Enter key on the PC keyboard.

Chapter 5. S/390 operating system installation 51

 alter it after IPLing the first time, the tape should already be positioned so that the second
utility is loaded.
flexes> ipl 390

This produces startup messages from DFSMSdss (on the 3270 session) that are very similar
to the startup messages from DSF. Respond to them in the same way. Remove the tape from
the drive and load the first tape to be restored. This is tape 1 of 4 for volume D9ESY1 in our
example.
ENTER INPUT/COMMAND:
restore fromdev(tape) fromaddr(390) toaddr(320) verify(D9ESY1)

(The tape should be read and you should see disk activity as the volume is restored.)
(The tape will be unloaded automatically after it is read, and
a message to load the next tape will appear on the screen.)
(Load tapes when requested. Also, clear the screen when requested
because the utility waits until you clear the screen before continuing operation.)
(After the last tape you should see the message FUNCTION COMPLETE
HIGHEST CONDITION CODE WAS 0.)
(You can now restore the second volume. Load the first tape of
the second disk volume.)

restore fromdev(tape) fromaddr(390) toaddr(321) verify(D9ECAT)

(Follow the same process for all the tapes of this volume.)

The COD system is now ready to use.

flexes> ipl 320 (No IPL parameter is required)

You should see the SPECIFY SYSTEM PARAMETERS message on the 3270 session at
address 0A1. The COD documentation provides the specific startup steps for its version of
z/OS.

The same general process (and stand-alone utilities) can be used to restore dumps of any
z/OS or OS/390 system. The stand-alone utilities adapt themselves to any device addresses.
However, the devices in your FLEX-ES definitions should match the addresses known in the
z/OS IODF in order to IPL a system after it is restored.

5.3 The tar file distributions

The example we use here is an IBM internal package known as the Dallas DEMOpkg and is
intended for IBM internal training. However, it is typical of the manner in which S/390
operating systems could be distributed in tar format. This package had one S/390 3390
volume per CD, in the form of a single compressed tar file. The file name (on the CD) was the
same as the S/390 volser plus “tar” as a suffix. The first CD had a few additional files
containing sample FLEX-ES definition and operational files.

The base file contained in a tar file is a complete image of a S/390 disk volume. Within it are
emulated CKD tracks, cylinders, R0s, and so forth. There is a standard label, a VTOC,
probably a VTOC index, and whatever data sets appear on that volume in a z/OS context. The
data is in S/390 format. Text contained on a S/390 volume is usually EBCDIC and
executables are S/390 binary files suitable for execution by z/OS. Due to the size of the
emulated 3390 files, it is usually not practical to place several of these files in a single tar file.

External documentation listed five volumes that were needed to have a workable system.
(Seven more volumes were included, containing optional products that we did not elect to

52 EFS Systems on a Linux Base: Getting Started

 install.) The base volsers were: DMTRES, DMTCAT, DMTOS1, DMTOS2, and DMTOS3. We
installed these as follows:
򐂰 Insert first CD in the CD-ROM drive and wait for Linux to open a window displaying the
contents of the CD. You could also use an ls command to display the contents:
$ls /mnt/cdrom (To display the contents of the CD)
dmtres.tar flexes.rsc flexes.sys ipl300 README.TXT
sg24-6834-00 startFlex startRes
򐂰 After inspecting the contents, we wanted two FLEX-ES definition files from the CD for later
use. We copied them into the rundir directory:
$ cp /mnt/cdrom/flexes.rsc /usr/flexes/rundir/DDrsc
$ cp /mnt/cdrom/flexes.sys /usr/flexes/rundir/DDsys
򐂰 Untar the emulated volume into our /s390 directory:
$ cd /s390
$ tar -zxvf /mnt/cdrom/dmtres.tar
(Close the CD-ROM automount window, if it is open)
$ umount /dev/cdrom
(Remove the CD)
򐂰 Install the next CD and untar the contents, following the same pattern.

We then customized the FLEX-ES definitions that were on the first CD and created a shell
script to help start this system. Our FLEX-ES definitions and the shell script are listed in
“Definitions for the Dallas DEMOpkg” on page 92. The original definitions (from the CD) were
in files DDrsc and DDsys; we merged these into a single file, DDdef. We determined the
addresses (S/390 device numbers) to use by studying the sample definitions and the
documentation provided with the CDs.

We compiled our definitions, started the FLEX-ES resource manager, started S/390
emulation (with the shell script), and IPLed our system:
$ cd /usr/flexes/rundir
$ cfcomp DDdef (Our FLEX-ES definitions)
# su (Must be root to start resource manager)
(password)
# resadm -s DDR.rescf (Start resource manager)
(pause) (Wait about 5 seconds)
# resadm -r (To confirm that it started OK)
# exit (Leave root)
$ sh shDD (Start the shell script)
flexes> ipl 0300 0301DP.1 (IPL the Dallas DEMOpkg system)
flexes>

We then logged onto TSO as SYSPRG1 (with the initial password the same as the userid).
The functions described here (cfcomp, resadm, shell scripts, the flexes> prompt) are described
in the next chapter. They are presented here for completeness. (Please note that our choice
of DDR as a resource file name is not related to the DDR program often used with z/VM.)

5.4 OMA/2 distributions

The most likely use of an OMA/2 distribution is for z/VM. The following material describes the
initial steps for this process.

We started with two CDs in a package labeled Product Package 3390 DDR for z/VM Version
4 Release 3.0. Examination of the first CD disclosed two directories: tapes and 9999. The

Chapter 5. S/390 operating system installation 53

 tapes directory contained a file named uaa925.tdf. This is a typical OMA/2 arrangement and a
tdf file is always present on an OMA/2 distribution.

We first created two empty 3390-3 volumes in the /s390 directory:

$ ckdfmt -n -r /s390/VM1000 3390-3
$ ckdfmt -n -r /s390/VM2000 3390-3

The file names (VM1000, VM2000) are completely arbitrary. We then created the FLEX-ES
definitions listed in “z/VM OMA/2 definitions” on page 97, along with the shell script shown.
We compiled the definitions with the cfcomp command. (This was all done in the
/usr/flexes/rundir directory.) The definitions provide two 3390 volumes at addresses 400 and
401, an emulated 3422 tape drive at address 500, and three 3270 sessions at addresses 600,
601, and 602. The 3422-type tape drive is typically used for OMA/2 emulation.

We installed the first CD in the CD-ROM drive and let Linux automount find it. (If this does not
work, a mount /dev/cdrom /mnt/cdrom command can be used.) We used resadm to start the
resource manager:
# mount /dev/cdrom /mnt/cdrom (If needed)
# ls -al /mnt/cdrom (Inspect the CD)
# cd /usr/flexes/rundir (Our FLEX-ES definitions)
# resadm -s VMR.rescf
(pause)
# resadm -r (Verify the resources started)

We then switched to userid flexes (that is, we exited from the root window):
$ sh shVM (Start the shell script)
(This should open a 3270 window)
flexes> mount 500 /mnt/cdrom:/mnt/cdrom/tapes/uaa925.tdf
flexes> ipl 500 (IPL ICKDSF)
flexes>

The FLEX-ES definitions made the 3422 tape drive at address 500 offline. The mount
command9 makes the device ready and mounts the indicated OMA/2 TDF file. The format of
the FLEX-ES OMA/2 mount command may appear odd. Use it exactly as shown here; there
are no spaces before or after the colon in the command. We then IPLed from the OMA/2 tape
drive.

Go to the 3270 window and press Enter. This should produce the message Clear Screen
When Ready. With an x3270 emulator using an unmodified keymap, you can clear the screen
with the Alt-c key combination, using the left Alt key. You can then proceed with the following
sequence in the 3270 window. Remember to clear the screen whenever requested:
ICK005E DEFINE INPUT DEVICE, REPLY’DDDD,CUU’ or ‘CONSOLE’
Enter (Press the Enter key)
ICK006E DEFINE OUTPUT DEVICE, REPLY ‘DDDD,CUU’ or ‘CONSOLE’
Enter
-----DEVICE SUPPORT FACILITY 16.0X-----
ENTER INPUT/COMMAND:
cpvolume format unit(400) novfy volid(430RES) mode(ESA) nofiller
ICK003D REPLY U TO ALTER VOLUME 0400 CONTENTS, ELSE T
U
....takes about 10 minutes. Clear screen as necessary for messages....
cpvolume format unit(401) novfy volid(430W01) mode(ESA) nofiller
U

9 Do not confuse the FLEX-ES mount command with the Linux mount command.

54 EFS Systems on a Linux Base: Getting Started

At this point the two 3390 volumes are initialized in a VM format. Now IPL the OMA/2 tape
again to load the DDR restore program:
flexes> ipl 500

Wait about 10 seconds and press Enter on the 3270 screen. If the 3270 keyboard is locked,
the keyboard Reset sequence for an unmodified x3270 is Alt-r.
Enter
z/VM DASD Dump/Restore Program
......
Enter:
sysprint cons
Enter:
input 500 tape (skip 1 rew
Enter:
output 400 dasd 430RES
Enter:
restore all
(This takes about one minute)
HOLDING
Enter
Enter:
Enter
END OF JOB

We can now IPL the z/VM volume. We must include the address of the 3270 console as the
IPL parameter:
flexes> ipl 400 600
flexes>

The IPL address is 400 and the IPL parameter is the address of the 3270 console. This
should produce the Stand-Alone Loader screen for the Initial Installation System (IIS) for
z/VM. We then followed the instructions in the z/VM Installation Guide, GC24-5992.10
Remember to use the keyboard combinations Alt-c and Alt-r as needed. While following the
instructions, we noted that the instdir command took at least a minute (and the system
appeared hung during this time). We used the VM command attach 500 * 181 at the
appropriate place in the instructions, and assigned virtual tape address 181 for both tape
volumes later in the installation process.

The installation process eventually issues the following messages:

INSTALL PROCESSING CONTINUES
(long pause)
PLEASE MOUNT VOLUME 2 ON TAPE DRIVE 181 THEN PRESS ENTER TO CONTINUE

At this point, do the following:

from flexescli flexes> mount 500 OFFLINE
from root # umount /mnt/cdrom
(Remove old CD, insert second CD)
(Wait for Linux automount or issue a mount /dev/cdrom /mnt/cdrom command)
from flexescli flexes> mount 500 /mnt/cdrom:/mnt/cdrom/tapes/uaa926.tdf
flexes>
on VM 3270 Enter

Processing should continue with the second volume. Note that the CD file name was
uaa926.tdf for the second volume. This name will be different for other releases or products,
of course.

10 Be sure to use the “instvm cd” EXEC when installing VM.

Chapter 5. S/390 operating system installation 55

 The z/VM installation process asks to process any RSU (service) tapes. We had a single CD
labeled 4301RSU Stacked. It had two TDF files in the tapes directory: uaa931.tdf and
uaa932.tdf. This represents two service tapes. We followed the same procedure to process
these tapes:
from flexescli flexes> mount 500 OFFLINE
from root # umount /mnt/cdrom
(Remove old CD, insert second CD)
(Wait for Linux automount or issue a mount /dev/cdrom /mnt/cdrom command)
from flexescli flexes> mount 500 /mnt/cdrom:/mnt/cdrom/tapes/uaa931.tdf
flexes>
(The attach 500 * 181 VM command was still in effect)

Following the installation instructions, we entered a service command in the VM session.

When this finished (almost 20 minutes, many screens of output), it issued the message RSU
tape 2 required on 0181. Since the second tape was on the same OMA/2 CD, we issued the
command:
from flexescli flexes> mount 500 /mnt/cdrom:/mnt/cdrom/tapes/uaa932.tdf

Mounting the new tape created a device ready interrupt and z/VM continued the service
processing automatically. The final step in the z/VM basic installation used the command
put2prod. This required several minutes and produced multiple screens of output.

5.5 File ownership (any distribution)

All of the files containing emulated S/390 volumes, however you received and installed them,
must be owned by userid flexes and groupid flexes. You can verify this, as follows:
# ls -al /s390 (Check the owner and group names)
# ls -al /s391 (If you have a second disk installed)
(If you need to change ownership, do the following)
# su (Change to root, if not already root)
# chown flexes:flexes /s390/Z1RES1 (Or whatever file needs to be changed)

The “flexes:flexes” parameter format causes both the owner and the group name to be
changed. These examples assume you place the emulated volumes in /s390 or /s391. If you
have a different arrangement, you need to display the appropriate directories.

56 EFS Systems on a Linux Base: Getting Started

 6

Chapter 6. Operational details

Installing and running an EFS system requires a number of steps, some of which have
already been described:
򐂰 Install the system hardware.
򐂰 Install Linux as the server operating system.
򐂰 Install FLEX-ES.
򐂰 Install a S/390 operating system.

We have completed the steps listed above. The following steps are required to run a S/390
instance:
1. Create a FLEX-ES definition file.
2. Compile the definition file, creating rescf and syscf files.
3. Start the resource manager, using resadm, pointing to a rescf file.1
4. Start flexes, the S/390 emulator, pointing to a syscf file.
5. Start the flexescli program, which provides a command line interface for controlling a
FLEX-ES instance.
6. Start at least one TN3270e session (for the MVS master console, or a similar purpose).
7. Issue an ipl command through the flexescli interface.
8. Use the S/390 operating system in the normal way.
9. Shut down the FLEX-ES system when finished.

These steps are described in this chapter. In general, most of the setup described here is
done under the flexes userid. By convention, all the FLEX-ES configuration files, resource
definitions, shell scripts, and so forth, are kept in /usr/flexes/rundir. This is not required, but it
is common practice and the following examples use this directory.

Sample FLEX-ES definition files were referenced in the previous chapter and are listed in
Appendix A, “Sample FLEX-ES operation files” on page 91. We do not discuss the syntax
and construction of these definitions here. This is briefly covered in the Appendix and much of
the construction is apparent by inspecting the definitions. The formal FSI documentation is
1
resadm is a command used to manage the FLEX-ES resource manager. The Linux process name for the resource
manager is resmanager. However, it is a common shortcut to refer to the resource manager as resadm.

© Copyright IBM Corp. 2003. All rights reserved. 57

 very detailed and covers all the syntax and construction. The companion redbook, S/390
PartnerWorld for Developers, ITSO/EFS Project EFS Systems on a Linux Base: Additional
Topics, SG24-7008, covers the basic syntax of common FLEX-ES definition statements.

There are many ways to set up and operate a FLEX-ES system. This chapter discusses the
way we routinely do this, but there are many other valid approaches and styles. We assume
you are using z/OS. If not, you can modify the following details to match your operating
system.

6.1 Linux desktop windows

When using our EFS system, we routinely log into the system as flexes and open two terminal
windows on the Linux desktop. We change to the /usr/flexes/rundir directory in both windows.
In one window we su to root. We use this window for starting and stopping the FLEX-ES
resource manager (using resadm). The other window is used for running a shell script and is
then used to enter flexescli commands.

The shell scripts we use automatically start two x3270 sessions. We then have a total of four
terminal sessions on the desktop:
򐂰 One (su to root) for resadm commands
򐂰 One for flexescli commands
򐂰 One x3270 session for the MVS console
򐂰 One x3270 session for TSO

There are many other ways you can arrange your desktop and FLEX-ES operation. However,
we suggest you follow our pattern for your initial usage.

6.2 FLEX-ES system and resource definitions

Before a newly installed/restored z/OS system can be used, we must define the FLEX-ES
system and resource elements required. “Minimal AD system definitions” on page 93 shows
the file that we used for our initial AD CD-ROM system with four system volumes. We placed
this definition file in /usr/flexes/rundir/defR14A. (File name defR14A is completely arbitrary.)
We specify the name of the definition file as an argument for the FLEX-ES configuration
compiler:
$ cd /usr/flexes/rundir
$ cfcomp defR14A
Start FLEX-ES Configuration Utility
Configuration processing *SUCCEEDED* with no errors
Data Space Manager Terminated

We could issue the cfcomp command directly because we earlier placed /usr/flexes/bin in our
Linux PATH. If you did not do this, you would need to use a full path name with the command:
$ /usr/flexes/bin/cfcomp defR14A

Our FLEX-ES definition file consists of two sections: the system section and the resources
section. The compilation creates files S14A.syscf and R14A.rescf, based on the S14A and
R14A names contained in our definitions.2 We can then invoke the FLEX-ES resource
administrator to activate our resources:
$ su
Password: (<--- Enter root password when this prompt is shown)
# cd /usr/flexes/rundir (If you are not already there)
# resadm -s R14A.rescf (<--- To activate our resources. Must be root)

58 EFS Systems on a Linux Base: Getting Started

 (wait a few seconds)
# resadm -r (<--- To check that resources are started)

Again, we could issue the resadm command directly because /usr/flexes/bin is in our PATH.
You must be root to issue start or terminate resource manager functions with the resadm
command. This is the only root operation required for normal EFS use.

Comments
The steps just shown, to compile a system and resource definition file and to activate the
resources (but not the system), require some explanations.

Our definition file source had been placed in /usr/flexes/rundir. We created the rundir directory
just after we installed FLEX-ES; it was not created automatically. Two separate definitions are
needed for a FLEX-ES emulated S/390. One defines the system being emulated:
򐂰 Central memory size available to this system
򐂰 Expanded memory size (if any)
򐂰 Mode of operation (ESA, zArchitecture)
򐂰 Number and type of CPUs
򐂰 Number and usage of channels
򐂰 Control units for all system devices

The other defines resources for a single system or multiple systems, such as:
򐂰 Interfaces for all the control units defined in the system section
򐂰 Devices for all the control units defined in the system section

The FLEX-ES configuration file(s) (resources and systems) must be compiled, using the
command cfcomp. Compilation produces two files. Parts of the output file names are taken
from the names in the definition files, and part of the names (syscf and rescf) are fixed. In our
case, we have S14A.syscf and R14A.rescf, based on the names coded in our definitions.

Once compiled, the resource portion (the rescf file) can be activated with the resadm
command:
$ cd /usr/flexes/rundir
$ su (Switch to root)
# resadm -s R14A.rescf (Must be root to use -s)

You can list all active resources with the command:

$ resadm -r
Resource: CPU Flags: READY Type: CPU Port: 32772 Pid: 17483
Resource: CU3990 Flags: READY Type: CU Port: 32778 Pid: 17485
Resource: CU3991 Flags: READY Type: CU Port: 32780 Pid: 17486
Resource: CU3174 Flags: READY Type: CU Port: 32782 Pid: 17487
Resource: CU3172 Flags: READY Type: CU Port: 32776 Pid: 17488
Resource: CU3480 Flags: READY Type: CU Port: 32784 Pid: 17489
Resource: NETCU Flags: READY Type: NETCU Port: 32786 Pid: 17492
Resource: TS3270 Flags: READY Type: TS3270 Port: 32774 Pid: 17493

If you try the resadm -r command very quickly after the resadm -s command, many of the
resources may be NOTREADY. They should become READY within a few seconds. You

2
We could have used two definition files, one for the system definitions and one for the resource definitions. The use
of separate system and resource definition files is most common when multiple instances of FLEX-ES are used
(meaning multiple S/390s being emulated at the same time). This is unlikely to be the case with a ThinkPad EFS
system. If we did use separate system and resource definition files, we would need to compile both of them in order
to obtain the syscf and rescf files we need to actually use a S/390 emulation instance. We found it more convenient to
always combine the definitions in one source file.

Chapter 6. Operational details 59

 normally have no use for the port and pid numbers, but these can be useful in debugging
situations.

More options of the resadm command are discussed in S/390 PartnerWorld for Developers,
ITSO/EFS Project EFS Systems on a Linux Base: Additional Topics, SG24-7008.

Once the resources are active, the system portion is ready to be started and IPLed. This is
easier to do with a shell script.

6.2.1 Building a shell script

We used vi to create the following shell script file in /usr/flexes/rundir/sh14A. (The file name is
arbitrary.) The contents are:
flexes S14A.syscf
xmodmap -e ‘keysym Alt_L = Alt_L Meta_L‘
xset fp+ /usr/flexes/fonts
xset fp rehash
x3270 -model 3 -keymap pc -port tn3270 localhost:mstcon &
x3270 -model 3 -keymap pc -port tn3270 localhost:L701 &
flexescli localhost S14A

The flexes command (in the shell script) starts an emulated S/390, using the indicated
compiled system definition file.

The next three commands3 prepare X Windows parameters, and the x3270 commands start
3270 client windows on the Linux screen. We elected to identify specific 3270 sessions
(mstcon and L701) for these two 3270 client windows.4 These session names match names
in our resource definitions. There is no requirement to start these x3270 sessions here (in the
shell script), but the terminal for the z/OS master console or consoles should start before
IPLing z/OS. Using the shell script this way provides a convenient, automatic way to establish
the z/OS console.5 We normally use the second 3270 session as a TSO terminal.

The last command of this shell script, flexescli localhost S14A, starts the FLEX-ES
Command Line Interface (CLI) program in interactive mode, with a flexes> prompt replacing
the default Linux prompt. You can enter flexescli commands here,6 or enter a quit command
to return the session to a Linux prompt.

Assuming that we have already started the resource manager (with a resadm -s S14A.rescf
command), we can now invoke the shell script to start S/390 operation:
$ sh sh14A
flexes> ipl a80 0a82cs
flexes>

Here we started the sh14A shell script. The last command in the shell script starts flexescli
(the command line interpreter, or CLI) in interactive mode, and this program provides the
flexes> prompt. At this point we can enter FLEX-ES CLI commands. We entered an ipl
command to start z/OS. We could have included the ipl command in the sh14A script, but we
found it more convenient to enter it manually because we often change the IPL parameter
values. (The “0a82cs” in the example is an IPL parameter value.)

3
These are not required, but are suggested by FSI to help make smoother x3270 operation.
4
If we omitted the session identification, we would obtain the Terminal Solicitor selection menu on these sessions.
5 If you do not start the x3270 sessions this way, you would need to access the Terminal Solicitor and start the

TN3270 session that will be used for the z/OS console before starting an IPL.
6
For example, you could enter an IPL command at this point. Some users prefer to code an IPL command in their
shell script, while others prefer to enter it at the flexes prompt.

60 EFS Systems on a Linux Base: Getting Started

If we wanted to include an ipl command in the shell script, it can look like this:
flexes S14A.syscf
xmodmap -e ‘keysym Alt_L = Alt_L Meta_L‘
xset fp+ /usr/flexes/fonts
xset fp rehash
x3270 -model 3 -keymap pc -port tn3270 localhost:mstcon &
x3270 -model 3 -keymap pc -port tn3270 localhost:L701 &
echo ‘ipl a80 0a82cs’ | flexescli localhost S14A
flexescli localhost S14A

This illustrates the two ways in which flexescli can be used. If a command is piped to it (with
echo), then flexescli executes that command and quits. If nothing is piped to it (as in the last
line of the shell script), then flexescli starts in interactive mode and issue the flexes> prompt.
The two operands for flexescli are the IP name/address of the system running FLEX-ES
(localhost) and the name of the FLEX-ES instance (S14A). The requirement for these two
operands becomes more apparent if you are running multiple FLEX-ES images with
networked channels.

Note that we hardcoded the name of our FLEX-ES instance (S14A) in the shell script. If we
intended to use several different FLEX-ES defined systems, we would need multiple shell
scripts. There are many ways you might decide to customize this operation once you are
more familiar with it.

The default x3270 keyboard parameters may not be what you want. See 7.6, “x3270 client” on
page 71, for more information about the Enter, New Line, and Reset key assignments.

The IPL parameters in this example (address A80 and parameter 0A82CS) are suitable for
the first IPL of the z/OS AD CD-ROM 1.4 system. You should read the AD documentation for
additional parameters.

You can place CLI commands (typically mount commands) in a Linux file and supply the file
name as the third parameter for the flexescli command. The commands in the file are
executed before the flexes> prompt appears.

Terminal Solicitor
The Terminal Solicitor is a standard FLEX-ES component that operates as a Linux process. It
is a TN3270 server that converts TN3270 client sessions to appear as local, non-SNA 3270
terminals. A remote user, for example, could connect his client TN3270 session to the
Terminal Solicitor (using the Linux TCP/IP stack, not the z/OS TCP/IP function) and log onto
TSO.

Starting the resource manager, using resadm, automatically starts the FLEX-ES Terminal
Solicitor. In order to connect to the FLEX-ES Terminal Solicitor, we connect a TN3270e client
to our Linux IP address using port 24. Port 24 is the default port for the FLEX-ES Terminal
Solicitor. The client system can be connected to any LAN that is connected to our Linux7. The
Terminal Solicitor presents a panel to us with the names of available8 3270 devices (these
are often made available by the FLEX-ES CLI mount command, although we elected to
hardcode them in the resource definitions).

7
It can also be another window on our Linux desktop.
8
Once a 3270 terminal device is selected by a user, it is removed from the Terminal Solicitor panel. When the device
is freed by the user, it reappears on the Terminal Solicitor panel and can be selected by another user.

Chapter 6. Operational details 61

 Welcome to the FLEX-ES Terminal Solicitor (node: t20)

Please select (X) the desired service and press enter

(PA1 to exit; CLEAR to refresh)

_ L702 _ L703
_ L704 _ L705 _ L706 _ L707
_ L708 _ L709 _ L70A _ L70B
_ L70C _ L70D _ L70E _ L70F
_ L710 _ L711 _ L712 _ L713

Figure 6-1 Terminal Solicitor panel (on a TN3270e screen)

The Terminal Solicitor display in Figure 6-1 might result from the definitions in “Full AD system
definitions” on page 94. If we had not started the x3270 sessions for mstcon and L701 in our
shell script, then these names would also appear in the Terminal Solicitor list. We would need
to select mstcon from the list to activate that session before IPLing z/OS, since it is used for
the MVS master console.9 When we select a terminal (by marking it with an X and pressing
Enter10), the Terminal Solicitor screen (on our client TN3270e session) is replaced by a 3270
connection to the indicated terminal address. After z/OS is up, this is normally result in the
USSTAB logo screen.

6.2.2 IPL z/OS

The z/OS IPL starts when you execute a flexescli ipl command. Once our z/OS was started,
we watched NIP messages, responded as necessary, and watched normal z/OS startup on
the x3270 window. Once you have z/OS IPLed, operation is the same as for any other z/OS
installation.

6.2.3 User terminal connection

A TSO user, for example, would normally connect to the system by connecting a TN3270e
session to the Linux IP address (or name, if you have a Domain Name Server service), port
24. This should produce the Terminal Solicitor display, as shown in Figure 6-1 on page 62.
The user would then select one of the available sessions (in the Terminal Solicitor display)
and would be connected to this S/390 address. When the user eventually drops the TN3270e
session, the S/390 3270 address is restored to the Terminal Solicitor display for others to use.

If you start z/OS TCP/IP, a user could connect directly to it by using TN3270 and the IP
address you assigned to z/OS TCP/IP. In this case, the user would not be connected to the
Terminal Solicitor, but would be connected directly to z/OS TCP/IP.

6.2.4 IODF requirements

z/OS requires an IODF data set that defines the I/O configuration seen by the software. This
normally matches the IOCDS defined for the S/390 I/O hardware configuration. An EFS
platform does not have an IOCDS. All resources are defined in FLEX-ES system and
resource files, compiled with the FLEX-ES resource compiler cfcomp, and then activated by
the FLEX-ES resource manager resadm command.

9 The AD system expects a 3270 terminal at address 700 for its master console. The name mstcon is arbitrary, and
matches the name chosen for a 3270 defined at address 700 when the resource definition file was created.
10
Many 3270 emulator users set up the right Ctrl key as the logical 3270 Enter key, since this most closely matches
a “real” 3270 keyboard. We did this and, using the IBM PCOM emulator, we pressed the right Ctrl key.

62 EFS Systems on a Linux Base: Getting Started

 An IODF is still required within a z/OS system, but the HCD input to generate it does not need
to define control unit details. That is, a simple device definition (device number, type, optional
features) is all that is required. The z/OS dynamic I/O redefinition capability is not available.
You can perform software dynamic configuration changes via z/OS Hardware Configuration
Definition (HCD) dialogs or the z/OS ACTIVATE command, provided the affected devices are
included in the FLEX-ES definition files.

In response to a z/OS d ios,config(all) command, we received:

- 11.38.06 d ios,config(all)
11.38.06 IOS506I 11.38.06 I/O CONFIG DATA 656
ACTIVE IODF DATA SET = SYS1.IODF99
CONFIGURATION ID = OS390 EDT ID = 00
HARDWARE SYSTEM AREA DATA COULD NOT BE OBTAINED
ELIGIBLE DEVICE TABLE LATCH COUNTS
0 OUTSTANDING BINDS ON PRIMARY EDT

6.2.5 System performance monitors

Because I/O sub-channel blocks are not maintained by FLEX-ES software emulation, the
z/OS Resource Measurement Facility (RMF™) is not fully supported. You can run RMF, or
other system performance monitors, but some of the reporting (especially when it comes to
I/O activity) is not complete.

When we started RMF, it reported the absence of an IOCDS in this environment and
automatically terminated I/O queuing activity reporting:
- 11.22.34 s rmf
- 11.22.35 STC00439 $HASP373 RMF STARTED
11.22.36 STC00439 ERB100I RMF: ACTIVE
11.22.36 STC00439 ERB265I RMF: IOCDS INFORMATION UNAVAILABLE TO RMF.
RESPONSE CODE 01F0
11.22.37 STC00439 ERB260I ZZ : I/O QUEUING ACTIVITY RMF REPORT TERMINATED
11.22.38 STC00439 ERB100I ZZ : ACTIVE

SYS1.LOGREC may not have hardware error information. Consequently, any report
produced by the Environmental Recording and Editing Program (EREP) has limited value
when it comes to hardware detected errors. We have not investigated this area in more detail.

6.2.6 TCP/IP for z/OS

It is possible to share a single Ethernet or token ring LAN adapter between Linux (including
the FLEX-ES Terminal Solicitor) and z/OS TCP/IP. This requires some coordination between
the FLEX-ES definitions and the z/OS TCP/IP profile data set parameters. Our goal was to
assign IP address 192.168.0.111 to z/OS TCP/IP, using the z/OS device addresses E20 and
E21 for the interface. (Our Linux IP address was 192.168.0.110.)

We included the following FLEX-ES definitions:

System definition:

channel (3) local

cu devad(0xE20,2) path(3) resource(R10A3172)

Resource definitions:

R10A3172: cu 3172 (for token ring: cu 3272TR)

interface local(1)
options ‘ipaddress=192.168.0.111’

Chapter 6. Operational details 63

 device(00) 3172 eth0 (for token ring: tr0)
device(01) OFFLINE
end R10A3172

We changed the following lines in the TCPIP.PROFILE.TCPIP data set of z/OS:

...
DEVICE LCS1 LCS E20
LINK ETH1 ETHERNET 0 LCS1 (LINK TR1 IBMTR 0 LCS1 for token ring)
...
HOME 192.168.0.111 ETH1
...
BEGINRoutes
; Destination Subnet Mask First Hop Link Packet Size
ROUTE 192.168.0.0 255.255.255.0 = ETH1 MTU 1500
ROUTE DEFAULT 192.168.0.1 ETH1 MTU DEFAULTSIZE
ENDRoutes
START LCS1

The fourth element in the LINK statement specifies the use of LAN adapter 0 (sometimes
known as the MPTS adapter number, from the OS/2 MPTS program). Adapter 0 is the default
LAN adapter number for FLEX-ES. This number is not related to any Linux LAN number. In
particular, it is not related to the ethn, or trn numbers used in Linux. A more complete listing of
our z/OS TCP/IP profile is in 7.8, “z/OS TCP/IP profile” on page 73.

With the specifications described here and using the z/OS 1.4s AD system, we have these
connection options from another host in the network:
򐂰 TN3270 to 192.168.0.110, port 24: This connects to the Terminal Solicitor (assuming the
resource manager is started). We can then select a session and should receive the
USSTAB logo screen. The Terminal Solicitor is a Linux program, provided with FLEX-ES.
򐂰 TN3270 to 192.168.0.111, port 23: This connects directly to z/OS TCP/IP. In recent AD
systems (where a USSTAB function is defined for TCP/IP) this should produce the logo
screen.
򐂰 telnet to 192.168.0.110, port 23: This connects to Red Hat Linux (if Telnet connections
are enabled).
򐂰 telnet to 192.168.0.111, port 1023: This connects to z/OS UNIX System Services, as an
ASCII terminal (assuming you are using a recent AD system).
򐂰 telnet to 192.168.0.111, port 23: This produces a line-mode connection (via VTAM) to
TSO and is not very useful.

6.3 Shutting down

Shutting down the system involves the following steps:
򐂰 Shut down z/OS. (With the AD systems we would enter s shutdown on the MVS console
to start a VTAMAPPL script that issues various commands.)
򐂰 Using the flexes window:
flexes> shutdown (Terminates S/390 emulation)
$
򐂰 Using the root window:
# resadm -T (Terminate all resources)
# resadm -k (Kill resource manager)
#

64 EFS Systems on a Linux Base: Getting Started

 To start the S/390 system again, we would use the resadm -s xxx.rescf command (using the
name of our compiled resource file), start the shell script, and then issue an ipl command
from the flexes prompt. When you shut down Linux, the resource manager is also stopped.
You are not required to use the resadm -k command.

6.4 More comments

Some users (and FSI Business Partners) prefer to start the FLEX-ES resource manager
automatically. This is done by placing a resadm command in the /etc/init.d/flexes file.11 We
have not done this (in the setup described in this redbook) for the following reasons:
򐂰 The resadm command must be hardcoded if it is included in the /etc/init.d/flexes file; that is,
it must specify the name of the resource file to be started. In our case, we had several
different resource files that we used at different times. (These include R14A.rescf and
R14B.rescf that are used in our examples.) We preferred to enter our resadm -s command
manually, and select which rescf file to use at that time.
򐂰 We sometimes used our Linux system for other purposes, and did not want any of the
FLEX-ES functions active.

While customizing new FLEX-ES definitions, we often need to stop the current resources
definitions and start a new set. There are two ways to do this:
# resadm -T (Stop all the current FLEX-ES resources)
# resadm -k (Kill resource manager)
# resadm -s new.rescf (Start a new resource file)
# resadm -r (Verify that it looks OK)
----- Better method --------
# resadm -T (Stop all the current FLEX-ES resources)
# resadm -x new.rescf (Start a new resource file)
# resadm -r (Verify that it looks OK)

In both examples, we assume you used the rescf command to compile a new set of
resources, creating new.rescf. The point is that it is better not to completely stop the resource
manager (with the -k option).

Tip: We must note that FLEX-ES is very unforgiving about errors. Any error, whatsoever, in
the definition files, the startup script, or the emulated volume files prevent FLEX-ES
emulation from starting. Once it starts, it is very stable. However, you must have everything
exactly right in order to get it started.

11
Starting with FLEX-ES Release 6.1.15, this is the /etc/init.d/FLEXES file. You can use the symbolic name
S90FLEXES instead because this name remains constant.

Chapter 6. Operational details 65

66 EFS Systems on a Linux Base: Getting Started
 7

Chapter 7. Basic discussions

Most of this redbook is concerned with the installation of a basic, entry-level EFS system.
This chapter discusses additional topics that may be relevant for such systems. The
companion IBM Redbook S/390 PartnerWorld for Developers, ITSO/EFS Project EFS
Systems on a Linux Base: Additional Topics, SG24-7008, discusses more advanced topics.

© Copyright IBM Corp. 2003. All rights reserved. 67

7.1 Resource definitions
Typical FLEX-ES system and resource definitions are shown in the Appendix. We routinely
place both system definitions and resource definitions in the same file. (This is not necessary
and may not be desirable for more complex operations. However, we recommend it for
getting started.)

Creating or changing FLEX-ES definitions for an emulated S/390 requires several steps:
򐂰 Edit the FLEX-ES definition file to make the changes. You can edit the same file you used
to create the resources that are currently running, for example:
$ vi defR14B (Or you can use another editor)
򐂰 Use the cfcomp command to compile the new definitions:
$ cfcomp defR14B
򐂰 Bring down z/OS, if it is running.
򐂰 Stop the emulated S/390:
flexes> shutdown
(You probably want to end any x3270 sessions on your screen)
򐂰 Stop the current FLEX-ES resources that are running:
# resadm -T (Only if resource manager is already running)
򐂰 Start the newly compiled resources:
# resadm -x R14B.rescf (If resource manager is already running)
# resadm -s R14B.rescf (If resource manager is not already running)
򐂰 Start your shell script (which starts FLEX-ES emulation) and IPL your system:
$ sh sh14B
flexes> ipl a80 0a8200

The resadm -T command stops any running FLEX-ES resources without stopping the
resource manager itself. Note that you should edit and compile your FLEX-ES definitions
while running as flexes, but you must be root to start, stop, or modify the resource manager
using appropriate resadm operands.

You must recompile your definitions (with the cfcomp command) after you make any changes.
You must then restart the resource manager with the newly compiled rescf file to make the
changes active. Editing a definition file has no effect on FLEX-ES. Compiling a definition file
has no effect until you restart the resource manager with the newly compiled definitions.

7.2 Linux windows and FLEX-ES operation

We found that we often have four windows open on the Linux desktop:
򐂰 A Linux window running flexescli (with the flexes prompt)
򐂰 A Linux window with su to root
򐂰 An x3270 session for the z/OS master console
򐂰 An x3270 session for TSO

The root window is useful for issuing resadm commands.1 The window with the flexes prompt
is generally required for FLEX-ES operation. We find that having two 3270 sessions is our
most useful arrangement. Our typical startup process goes like this:
1
This is not necessary, of course. Some users prefer to open a window, su to root, start the resource manager, exit
from root, and run flexescli in that window.

68 EFS Systems on a Linux Base: Getting Started

 Log into Linux with userid flexes (Working directory is /usr/flexes)
Start two terminal windows on the Linux desktop
Window 1
$ su (Enter root password)
# cd rundir (Change to /usr/flexes/rundir)
# resadm -s R14A.rescf (Start FLEX-ES resources)
# resadm -r (Verify resources started)
Window 2
$ cd rundir (Change to /usr/flexes/rundir)
$ sh sh14A (Start our shell script)
Wait for 3270 sessions to start
flexes> ipl a80 0a8200 (Note that prompt changes to flexes)

Operate z/OS as normal, using 3270 sessions. Shut down z/OS when finished.

Window 2
flexes> shutdown (Stop S/390 emulation)
Window 1
# resadm -T (Stop FLEX-ES resources cleanly)
# resadm -k (Terminate resource manager; optional)
# exit
$ exit

This process can be automated or abbreviated in a number of different ways. We like to keep
a straightforward, manual control over FLEX-ES startup and use the steps shown here.

Another window (not root) is sometimes useful for running vmstat commands and similar
utility functions. (This can be done from the root window if you ignore the usual advice to
avoid root usage except when specifically required. Or, you could exit from root in Window 1
and work from there.)

7.3 64-bit operation

FLEX-ES emulates an Architecture Level Set 3 (ALS 3) system if the instset parameter is set
to z. This parameter is in the system definition portion of the FLEX-ES definitions. ALS 3
includes 64-bit operation. For example:
instset(esa) (Emulates a 31-bit system)
instset(z) (Emulates an ALS 3 (64-bit capable) system)

Remember that this refers only to the emulated S/390.2 The base Intel system, running Red
Hat Linux, is still in 32-bit mode. The largest S/390 memory that can be emulated is 2 GB.
(We say that Linux is a 32-bit system because it is using 32-bit registers and integers for most
functions. With the proper kernel options, Linux can use more than 2 GB of real PC memory.
This has no direct bearing on the size of S/390 memory that can be emulated.)

7.4 Creating additional emulated volumes

You can create a new emulated DASD volume in two ways:
򐂰 You can restore it from a FLEX-ES (or AWSCKD) archived copy. This has the advantage
that the internal structure of the volume (label, VTOC, catalogs, data sets) is already in
place.
򐂰 You can create an uninitialized volume and then run jobs to initialize it.

2 We should probably say zSeries here, but we use S/390 consistently throughout this book.

Chapter 7. Basic discussions 69

 In either case you must have Linux disk space for the new volume. In our examples, we
assume that directory /s390 has sufficient space. The Linux command df -h can be used to
survey the available space in your file systems. In practice, there is no special need to have all
the emulated volumes in the same Linux file system.

The general steps involved are:

1. Create or restore a 3390 volume as a Linux file.
2. Update the FLEX-ES definitions.
3. Compile the new FLEX-ES definitions.
4. Start FLEX-ES operation and IPL z/OS.
5. Run an ICKDSF job to initialize the volume, if needed.
6. Vary the volume online to z/OS, if needed.
7. Possibly update the VATLST00 member in PARMLIB.

You can create a new emulated volume with the following command:
$ ckdfmt -n -r /s390/WORK02 3390-1

This example creates an emulated 3390-1 volume, creating the Linux file name WORK02 in
directory /s390. Many users select a file name (WORK02) that matches the volser of the
emulated volume (after they initialize it to contain a volser), but there is no requirement to do
this. The -r flag in the command indicates that the file should be created if it does not already
exist. The ckdfmt command formats the file with 3390 cylinders and tracks but does not
create any other internal structure. FLEX-ES need not be running for you to use the ckdfmt
command.

You need to then add the new volume to your FLEX-ES resource definitions. Your FLEX-ES
definitions must assign an address (device number) that corresponds to the correct device
type for your S/390 operating system.3 Remember that FLEX-ES definitions assign S/390
addresses (device numbers) through the cu devad parameter in the system definition. Also
remember that the number of devices specified in the cu statement in the system section
must match the number of devices defined for the matching control unit in the resource
section. If you want to access the volume, you must compile (cfcomp) the new definitions and
use resadm to restart the resource manager using the new rescf file.

It is possible to update an operational resource file while z/OS, for example, is running.
However, this involves a number of techniques and is not recommended for a new user. For
practical purposes (until you become a FLEX-ES expert), you must shut down your S/390
emulation to add new emulated devices.

If z/OS is IPLed with a newly created emulated drive, it detects an uninitialized volume and
takes it offline. We need to run an ICKDSF job to create a label and VTOC on the volume. A
typical z/OS job is:
//ICKDSF01 JOB 1,OGDEN,MSGCLASS=X
// EXEC PGM=ICKDSF,REGION=4096K
//SYSPRINT DD SYSOUT=*
//SYSIN DD *
INIT UNIT(A90) NOVAL NVFY VOLID(WORK02) PURGE VTOC(0,1,29)
/*

A job similar to this can be found in recent AD systems in SYS1.P390.CNTL(DSFINIT). Do

not include a STORAGEGROUP parameter in the INIT statement unless you are certain you
want the volume controlled by DFSMS. Adjust the UNIT parameter to the address you
assigned to the volume via FLEX-ES definitions. When you run this job, you need to reply U
3
This depends on your S/390 operating system and its documentation. z/VM, for example, automatically senses
devices and their addresses while z/OS has preassigned addresses in whatever IODF file is used for IPLing. For the
AD system, the IODFs include 3390 devices in the A80-A9F range and in other ranges.

70 EFS Systems on a Linux Base: Getting Started

 to a z/OS console message. When the job has finished, you can VARY A90,ONLINE on the
z/OS console and begin using the volume.

We usually modify the VATLST00 member in the z/OS PARMLIB to control the mount
attributes of new volumes, but this is not required.

7.5 Operating systems messages console

Current S/390 and zSeries machines have an Operating Systems Messages console function
that is provided through the Support Element (SE) or a Hardware Management Console
(HMC). This console function is sometimes known as the “system console” or the “hardware
system console”. z/OS attempts to use it if all other MVS consoles fail.

FLEX-ES emulates this console through the CLI window--the window with the flexes prompt.
Messages written from the S/390 to the Operating Systems Messages console appear after
the flexes prompt. You need to press Enter (with the desktop focus in this window) to restore
the flexes prompt. You can reply or enter commands through the Operating System
Messages console by using the CLI command hwc:
flexes> SPECIFY SYSTEM PARAMETERS (Message from z/OS)
(Press Enter to get flexes prompt)
flexes> hwc R 00,CLPA
flexes>

If you are flooded with messages, you can issue the following command to remove the
pending output:
flexes> clear messages

Remember to press Enter (to return to the flexes> prompt) after any S/390 messages are
displayed.

7.6 x3270 client

The FLEX-ES product distribution includes a copy of x3270. This is based on the freely
distributable x3270 package, but with a few modifications. In general, it is intended only for
use on the EFS system itself, under Linux. Users connecting from other platforms may use
whatever TN3270e clients are available for their platforms. For this redbook project, we used
the IBM PCOM package—and worked from OS/2 and Microsoft Windows machines.

An x3270 window (on the Linux console) is started with the command:
$ x3270 -model 3 -keymap pc -port tn3270 localhost &
| | | | | +-- Run in background
| | | | +--------- Local system (from /etc/hosts)
| | | +-------------------- Port tn3270 (from /etc/services)
| | +------------------------------- In /usr/X/lib/app-defaults/X3270
| +----------------------------------------- Models 2, 3, 4, 5 available
+------------------------------------------------- Command

You can optionally identify the FLEX-ES terminal session you want to connect to like this:
$ x3270 -model 3 -keymap pc -port tn3270 localhost:L701 &

In this case, your resource definition (or a prior FLEX-ES mount command) must have defined
a 3270 device named L701. If you do not identify a specific terminal connection, the x3270

Chapter 7. Basic discussions 71

 session is connected to the Terminal Solicitor. In a FLEX-ES system, port tn3270 normally
resolves (via /etc/services) to port 24, the default port for the Terminal Solicitor.

The x3270 program must be in the current search path, of course. For FLEX-ES, it is in the
/usr/flexes/bin directory, and we suggested placing this directory in the current PATH. The
model parameter refers to 3270 models:
Model Characteristics
2 24 lines x 80 columns
3 32 lines x 80 columns
4 43 lines x 80 columns
5 27 lines x 132 columns

These are standard 3270 family model designations. Model 3 is perhaps the most common
and that is what we used.

Default keyboard mapping for x3270 (when started as shown) includes:

Function Keys Used
PA1 Left-Alt + 1 (Hold down the left Alt key and press 1)
PA2 Left-Alt + 2
Clear Left-Alt + c or Left-Alt + Home
Reset Left-Alt + r
Enter Enter (Large Enter key performs 3270 Enter function)
NewLine Left-Alt + Enter (Performs 3270 New Line function)
SysRq Left-Alt + s (The System Request key (usually for SNA))
F13 Left-Alt + F1 (And so forth for F14 - F24)
EraseEOF Left-Alt + f
Insert Left-Alt + i
Del Delete
BackTab Shift + Tab (Tab backwards)

This key mapping may be disconcerting to experienced 3270 users (especially ISPF users)
because it uses the large Enter key (on the PC keyboard) as the 3270 Enter key. On a “real”
3270, the large key in this position performs a New Line function, and a smaller key (where
the right Cntrl key is positioned on a PC keyboard) performs the 3270 Enter function. You can
change x3270 key assignments by doing the following:
# cd /usr/lib/X11/app-defaults
# cp X3270 X3270old (Keep copy of original, just in case)
# vi X3270
.....
x3270.keymap.pc: \ (Scroll until you find this line)
Meta<Key>d: Redraw()\n\
Alt<Key>d: Redraw()\n\
Meta<Key>Return: Newline()\n\
Alt<Key>Return: Newline()\n\
<Key>Return: Newline()\n\ (Change this line, as shown)
!Shift<Key>Tab: BackTab()\n\
.....
..... (Near end of ‘x3270.keymap.pc’ section)
Alt<Key>Home: Clear()\n\
<Key>KP_Enter: Enter()\n\ (Insert this line)
<Key>Control_R: Enter()\n\ (Insert this line)
<Key>Control_L: Reset()\n\ (Insert this line)
:<Key>: Default()
x3270.keymap.pc84: \
.....

72 EFS Systems on a Linux Base: Getting Started

 You need to use :w! to save the file in vi, because it is a read-only file. Make these changes
carefully; there is one altered line and three new lines. Be careful not to make any other
changes unless you understand the format and functions of this file.

We changed the large Enter key on the PC keyboard to perform the 3270 New Line function,
the left Cntrl key to perform the 3270 Reset function, and the right Cntrl key to perform the
3270 Enter function. This is the most familiar keyboard arrangement for most 3270 users.

7.7 S/390 identification

You may need to supply the machine type and serial number of your emulated S/390. This is
often needed for S/390 software licenses or registration on various Web sites. With z/OS, the
easiest way to obtain this is with the d m=cpu console command:
d m=CPU
PROCESSOR STATUS
ID CPU SERIAL
0 + 00F0971245

CPC.SI = 1245.L02.FSI.FREM.000000000000F097

You can obtain the same basic information with a command from the flexes prompt:
flexes> display cpuid
CPUID=0a00f097 12450000

Both these displays indicate that the machine type is 1245 and the serial number is 00F097.
(If a machine type-model number is needed, the model can be taken from the field containing
the L02 in the d m=cpu display.) IBM has reserved machine types 1245, 1246, and 1247 for
FLEX-ES emulated processors. Type 1247 is normally used for FLEX-ES systems obtained
through the IBM S/390 Partners in Development program. The serial number is assigned
through the FLEX-ES license code file and should be unique for each license.

The machine type and serial number are available through the S/390 instruction STIDP.

7.8 z/OS TCP/IP profile

This document is not intended to cover normal S/390 systems programming tasks. However,
the subject of the profile for z/OS TCP/IP arises so often that we include a typical listing for a
small EFS system. This is based on the profile for the AD CD-ROM 1.4s release, with all the
comment lines removed. Your needs may be different; this is offered only as a working
example.
ARPAGE 20
GLOBALCONFIG NOTCPIPSTATISTICS
IPCONFIG DATAGRAMFWD VARSUBNETTING SYSPLEXROUTING
SOMAXCONN 10
TCPCONFIG TCPSENDBFRSIZE 16K TCPRCVBUFRSIZE 16K SENDGARBAGE FALSE
TCPCONFIG RESTRICTLOWPORTS
UDPCONFIG RESTRICTLOWPORTS
DEVICE LCS1 LCS E20 AUTORESTART
LINK ETH1 ETHERNET 0 LCS1
HOME
192.168.0.121 ETH1
BEGINRoutes
; Destination Subnet Mask First Hop Link Packet Size
ROUTE 192.168.0.0 255.255.255.0 = ETH1 MTU 1500

Chapter 7. Basic discussions 73

 ROUTE DEFAULT 192.168.0.1 ETH1 MTU DEFAULTSIZE
ENDRoutes
AUTOLOG 5
FTPD JOBNAME FTPD1 ; FTP Server
PORTMAP JOBNAME PORTMAP1 ; USS Portmap Server (SUN 4.0)
ENDAUTOLOG
PORT
7 UDP MISCSERV ; Miscellaneous Server - echo
7 TCP MISCSERV ; Miscellaneous Server - echo
9 UDP MISCSERV ; Miscellaneous Server - discard
9 TCP MISCSERV ; Miscellaneous Server - discard
19 UDP MISCSERV ; Miscellaneous Server - chargen
19 TCP MISCSERV ; Miscellaneous Server - chargen
20 TCP * NOAUTOLOG ; FTP Server
; 20 TCP * NOAUTOLOG SAF FTPDATA ; FTP Server
21 TCP FTPD1 ; FTP Server
; 21 TCP FTPD2 BIND FEC0:0:0:1:0009:0067:0115:0066 ; FTP IPv6
23 TCP INTCLIEN ; Telnet 3270 Server
; 23 TCP INETD1 BIND 9.67.113.3 ; z/OS UNIX Telnet server
25 TCP SMTP ; SMTP Server
53 TCP NAMED ; Domain Name Server
53 UDP NAMED ; Domain Name Server
111 TCP PORTMAP ; Portmap Server (SUN 3.9)
111 UDP PORTMAP ; Portmap Server (SUN 3.9)
; 111 TCP PORTMAP1 ; Unix Portmap Server (SUN 4.0)
; 111 UDP PORTMAP1 ; Unix Portmap Server (SUN 4.0)
123 UDP SNTPD ; Simple Network Time Protocol Se
135 UDP LLBD ; NCS Location Broker
161 UDP OSNMPD ; SNMP Agent
162 UDP SNMPQE ; SNMP Query Engine
389 TCP LDAPSRV ; LDAP Server
443 TCP HTTPS ; http protocol over TLS/SSL
443 UDP HTTPS ; http protocol over TLS/SSL
512 TCP RXSERVE ; Remote Execution Server
514 TCP RXSERVE ; Remote Execution Server
; 512 TCP * SAF OREXECD ; z/OS UNIX Remote Execution Serv
; 514 TCP * SAF ORSHELLD ; z/OS UNIX Remote Shell Server
515 TCP LPSERVE ; LPD Server
520 UDP OROUTED ; OROUTED Server
580 UDP NCPROUT ; NCPROUTE Server
750 TCP MVSKERB ; Kerberos
750 UDP MVSKERB ; Kerberos
751 TCP ADM@SRV ; Kerberos Admin Server
751 UDP ADM@SRV ; Kerberos Admin Server
1933 TCP ILMTSRVR ; IBM LM MT Agent
1934 TCP ILMTSRVR ; IBM LM Appl Agent
3000 TCP CICSTCP ; CICS Socket
3389 TCP MSYSLDAP ; LDAP Server for Msys
SACONFIG ENABLED COMMUNITY public AGENT 161
TelnetParms
Port 23
TELNETDEVICE 3278-3-E NSX32703 ; 32x80
TELNETDEVICE 3279-3-E NSX32703 ; 32x80
TELNETDEVICE 3278-4-E NSX32704 ; 48x80
TELNETDEVICE 3279-4-E NSX32704 ; 48x80
TELNETDEVICE 3278-5-E NSX32705 ; 132x27
TELNETDEVICE 3279-5-E NSX32705 ; 132x27
LUSESSIONPEND
MSG07
CodePage ISO8859-1 IBM-1047 ; Linemode ASCII, EBCDIC code pa

74 EFS Systems on a Linux Base: Getting Started

 Inactive 0 ; Let connections stay around
PrtInactive 0 ; Let connections stay around
TimeMark 600
ScanInterval 120
; SMFinit std
; SMFterm std
WLMClusterName
TN3270E
EndWLMClusterName
EndTelnetParms
BeginVTAM
Port 23 ; 992
DEFAULTLUS
SC0TCP01..SC0TCP30
ENDDEFAULTLUS
LINEMODEAPPL TSO ; Send all line-mode terminals to TSO
ALLOWAPPL TSO* DISCONNECTABLE ; Allow all users access to TSO
USSTCP USSN
ALLOWAPPL * ; Allow all applications
EndVTAM
START LCS1

We changed only a few lines in the profile distributed with the AD system and these are
shown in bold above. Our corresponding resolver data (in TCPIP.TCPIP.DATA) is:
TCPIPJOBNAME TCPIP
HOSTNAME NF
DOMAINORIGIN ITSO.IBM.COM
DATASETPREFIX TCPIP
NSINTERADDR xxx.xxx.xxx.xxx (find your own name server!)
RESOLVEVIA UDP
RESOLVERTIMEOUT 10
RESOLVERUDPRETRIES 1
ALWAYSWTO NO

7.9 Initial debugging

We present here a few common debugging steps that may be useful if you have a problem
starting FLEX-ES. We must note that FLEX-ES typically does not produce error messages
oriented to a S/390 person. An error message indicates that an error occurred, but you may
need to ignore the specific text of the message and attempt to deduce the original cause of
the problem.

Starting msgmgr
One of the more common problems appears with messages similar to these:
FSIRA430 Aborting resadm -- unable to start resource manager; reason:
FSIRG840 Aborting -- unable to open “/dev/msgmgr/msg00000” (errno = 19)

Errno 19 indicates that the msgmgr loadable module is not loaded. You can use the Linux
command to list the Linux modules that are loaded:
$ /sbin/lsmod

The msgmgr module is found in /usr/flexes/modules. It is normally loaded by the S90FLEXES

startup script that was added to Linux by the FLEX-ES installation process. If the lsmod listing
does not include msgmgr, try this:
# cd /usr/flexes/modules

Chapter 7. Basic discussions 75

 # ls (verify that msgmgr.o exists)
# /sbin/insmod msgmgr.o (try to force loading)

Log file /var/log/messages should contain messages whenever msgmgr is loaded or

unloaded; you can look for error messages there:
# grep msgmgr /var/log/messages | more (look for error messages)

The msgmgr module is installed by the second rpm involved when installing FLEX-ES.
Among other actions, this installation consists of a compilation (gcc) and a link (ld). These
must complete successfully. If you did not observe this carefully during installation, you may
be able to remove msgmgr (rpm -e msgmgr) and install it again. The most common reason for
errors (with the gcc or ld steps) is that you did not include Development Tools and Kernel
Development selections when you installed Linux. You need these to install the compiler and
linker, and to install the kernel source. (If you selected the Everything option for Red Hat
installation, these are included.)

If no errors occur during the gcc and ld operations but you have this error when attempting to
use resadm to start the resource manager:
򐂰 The Linux kernel source tree in your Linux system may not be suitable for FLEX-ES. The
Red Hat releases we recommend have an appropriate kernel.
򐂰 If you rebuild the kernel you should use a kernel source tree provided by FSI.
򐂰 Perhaps you downloaded Red Hat updates that included a new kernel. If so, this kernel no
longer matches the source tree in your system. (You may be able to boot Linux with the
original kernel, installing and running FLEX-ES from there.)

The right kernel

You must run msgmgr with the same kernel you used when you installed it. You may have
built a new kernel. (This will become more common in the future because newer hardware
may require that you rebuild your kernel.) If you build and use a new kernel, you must
reinstall msgmgr to match that kernel.

Kernel or compiler mismatch

You may see messages about a mismatch between the compiler and modules. The
messages usually appear when you start msgmgr, and might be found in /var/log/messages.
The problem is that your kernel does not match the kernel source used to install msgmgr.
The most common reasons are:
򐂰 You downloaded a kernel (probably as part of a package of updates). Your kernel no
longer matches your kernel source and msgmgr does not work in this situation. You may
be able to boot with your original kernel and bypass this problem.
򐂰 You built a new kernel, but you forgot to select it when you last booted. (After appropriate
testing, you should consider making your new kernel the default kernel by changing the
default value in /boot/grub/grub.conf.)

In practice, you can probably use your msgmgr module, regardless of these messages,
provided you had a clean compile and link when you installed msgmgr. To try this, you can
force the loading of msgmgr:
# cd /usr/flexes/modules
# insmod -f msgmgr.o

CU will not start

If the resource manager starts without errors but you receive error messages about control
units when you start flexes (usually via a shell script), you should consider the following:

76 EFS Systems on a Linux Base: Getting Started

򐂰 The resource names in the FLEX-ES system definition must match the names in the
resource definitions. Check carefully for typos. We frequently see, for example, CU3390
when you meant to type CU3990, or similar errors.
򐂰 Carefully check the Linux file names in the resource definitions. The files must exist and
be in the right format. For example, if the DEVICE statement in the resource definition
says 3390-3, it must point to a Linux file that emulates a 3390-3. If the file emulates a
3390-2, FLEX-ES does not start.
򐂰 Check the ownership of all the files used by the resource definitions. These must be
owned by userid flexes and groupid flexes. You can check ownership with a command
such as ls -al /s390. You can change ownership (working as root) with a command such
as chown flexes:flexes /s390/WORK02.
򐂰 If you used our naming conventions, verify that directory /s390 is owned by flexes.

x3270 problems
The following error message means that you are starting the wrong copy of x3270:
Error: app-defaults version mismatch: want 3.2.18, got 3.1.1.0

See 4.1.3, “Location of x3270 program” on page 41, for more information.

Other errors
Do your rescf and syscf files match your current definitions (in your definition source file)?
Changing your definition source file of files has no effect until you recompile them and use
resadm to start (or restart) the resource manager with the newly compiled files. If you are not
certain, you can simply compile your definition file again. In our examples, this is:
$ cd /usr/flexes/rundir
$ cfcomp defR14A (using your correct source file name)

Be certain you have a clean compile. This is indicated by the messages:

Configuration processing *SUCCEEDED* with no errors
Data Space Manager Terminated

Are you using the correct files? The resadm command uses the rescf file and the flexes
command uses the syscf file. It is easy to reverse the two file suffixes, and the resulting error
message does not identify the exact problem.

Once the resource manager (resadm), the emulator (flexes), and the command line interface
(flexescli) are started, you may find the following CLI commands useful:
flexes> d devstate 700 (display state of emulated device)
flexes> d ckdcachestats A80 (verify I/O occurs to a disk)
flexes> d psw

If your emulated 3270 terminal operation is erratic, be certain you know where the 3270 Enter
key is located, and where the 3270 New Line key is located. You may be creating a 3270
Enter operation when you intended to use the New Line key. See 7.6, “x3270 client” on
page 71, for suggestions in this area.

Be certain that your disk model numbers are consistent. For example, if you created a 3390-2
(using ckdfmt or ckdconvaws) but you specify a 3390-3 in your resource definitions, the
system may not start.

FLEX-ES trace files

FLEX-ES writes trace files that are intended only for use by FSI. However, these sometimes
contain clues that may be useful. Consider the following:

Chapter 7. Basic discussions 77

 $ cd /var/adm/flexes (All the trace files begin here)
$ ls -al (Always use the -al switches)
drwxrwxr-x 2 flexes flexes 4096 June 12 12.02 CU3990
drwxrwxr-x 2 flexes flexes 4096 June 12 12.02 CU3991
drwxrwxr-x 2 flexes flexes 4096 June 12 12.02 S14B
(and other files, including .flexeslicense)
$ cd S14B (Move to one of the directories)
$ ls -al
-rwxrwxr-x 2 flexes flexes 514 June 12 12.02 devtr.1115.0
-rwxrwxr-x 2 flexes flexes 96 June 12 12.02 .stderr.0.1.1483
(and other files, such as “.” and “..”)
$ dtprint devtr.1115.0 (Display the trace)
$ cat .stderr.0.1.1483 (If the length is not zero)

There are several important points in this example:

򐂰 The FLEX-ES trace files (or directories) are in /var/adm/flexes.
򐂰 /var/adm/flexes usually contains directories with the names of your system (S14B) or your
resources (CU3990). The exact names depend on the names in your FLEX-ES
definitions.
򐂰 Always use the ls -al command when looking in these directories. Many file names
begin with a period and are not shown by a simple ls command.
򐂰 Many files (usually .stderr files) may have zero length. You can ignore these.
򐂰 You can cat any .stderr files with length greater than zero.
򐂰 You must use the dtprint command to view device trace (devtr) files.

If you attempted to start FLEX-ES many times, there may be many of these trace files and
directories. You can delete them all (but be careful not to delete the .flexeslicense file). The
time and date stamps can help you determine which are the most recent files.

Dongle problems
We found that the FSI dongle works only in the lower USB connector of a ThinkPad docking
station.4

There have been rare cases of FSI dongle failures. These have fallen into two categories:
򐂰 A partial failure, which is associated with the message “Cannot write to dongle.”
򐂰 Complete failure in which a Linux cat /proc/bus/usb/devices command does not display
the presence of the FSI dongle. An additional check is to connect the dongle to a Microsoft
Windows system. If Windows detects the dongle and offers to install a driver for it, this
means the dongle is probably working. (There is no Windows driver for the FSI dongle, so
you cannot go too far down this path.)

7.10 VTAM APAR

If you use XCA or OSA with SNA over Ethernet and you are using a multiprocessor system
then you should investigate APAR OA03594. This problem and fix is not unique to FLEX-ES,
but can occur on any system with more than one CPU and a fast XCA or OSA-1 function.

4 We used a T23 ThinkPad. The situation may be different for other hardware combinations.

78 EFS Systems on a Linux Base: Getting Started

7.11 Security
FLEX-ES is a layer of software that resides and operates between a S/390 operating system
and an underlying Linux system. All the security features and functions that come with a
z/OS system work as on any other S/390 platform. However, it is possible for a Linux user
with sufficient privilege to gain access to the contents of an emulated DASD or main storage
associated with an emulated CPU, and so forth.

An EFS owner must plan and manage traditional Linux security functions for the underlying
Linux system, as well as traditional S/390 security management. If the EFS platform is used
only for S/390 operation, this can be fairly simple.

A unique concern involves the z/OS master console(s). These can be implemented through
the Terminal Solicitor. This is convenient, but offers an opportunity for an unwanted person to
connect as a master console. We suggest that you have a master console on the server
display and direct any other master consoles5 to specific IP addresses and not through the
Terminal Solicitor. This is done by specifying an IP address in the FLEX-ES resource file,
instead of a terminal name. This means, of course, that the client systems connecting to the
emulated 3270 interfaces must have static IP addresses.

The FLEX-ES resource manager uses TCP/IP port 555 to talk with other instances of
FLEX-ES resource managers. There may be a potential for problems if someone hacks this
port. We are not aware of any instances of this, but it is a potential concern if you use an
open network to link multiple FLEX-ES systems at this level.

7.12 RAS discussion

Is a ThinkPad EFS or xSeries EFS system as reliable as a “real” S/390? No. A reliability
discussion can quickly become extended and open-ended and we do not attempt to explore
all the potential discussion points. We can summarize our views this way:
򐂰 An xSeries EFS or a ThinkPad EFS system is not a S/390 (or a zSeries) system in terms
of reliability. It does not have dual instruction units in each processor, or an Application
Preservation function, for example.
򐂰 A production EFS system is based on a high-quality server and can be expected to exhibit
much better reliability than a typical desktop PC. Contributing factors are:
– ECC memory (or other types of correcting memory)
This is extremely important in preventing the random lockups and crashes that many
PCs experience.
– Use of a RAID adapter
– ECC memory in the RAID adapter
– Battery operated cache in the RAID adapter
– Multiple, redundant power supplies
– Avoidance of random applications (of uncertain trustworthiness) on the server
That is, games are not downloaded from random sources on the Web, we hope. The
system is used only for S/390 emulation.
– Thorough physical inspection of the hardware by the Business Partner involved
While this sounds trivial, it contributes to stability.

5 We use the term “master console” loosely here to mean any z/OS operator console.

Chapter 7. Basic discussions 79

 – Using a UPS system “in front” of the server
– Better cooling
This is a significant factor for a system that is intended to run continuously.
򐂰 A ThinkPad EFS is based on a quality mobile computer, but this does not compete with an
xSeries EFS server in many RAS characteristics.
– The ThinkPad does not have RAID protection for its disk(s). With a single disk drive,
or, at the most, two drives, there is no practical method to add RAID protection.
– Multiple power supplies are not available.
– ECC memory is not used, for a number of reasons.
– Interfaces (such as SCSI or S/390 channels) are not typically available to ThinkPad
EFS users, making routine system backup functions more difficult.
– Cooling effectiveness, especially when used without a docking station, can be greatly
influenced by the exact location and positioning of the ThinkPad.

If you must have S/390 RAS, you need to buy a S/390. We believe an xSeries EFS system
provides more than adequate reliability within its place in the hierarchy of servers. A ThinkPad
EFS system is not normally intended as a production system and should not be compared
with the RAS characteristics of production systems.

7.13 Useful Linux commands

Not all EFS owners have Linux or UNIX backgrounds, and a few brief notes may help these
users.

Common commands
New users may find Linux commands a little confusing. The following is intended as a brief
reminder of the syntax for a number of commands often used while working with Linux in an
EFS environment.
chown flexes /s390/volA Make userid “flexes” owner of this file
chgrp flexes /s390/volA Make group “flexes” the group owner of this file
chown aaa:bbb /home/myfile Make aaa the owner userid and bbb the owner groupid
chmod 600 /s390/volA Allow only file owner to read & write this file

cat filename List all of a file on the terminal

more filename List a file, page by page (Spacebar to advance page)
(Ctrl-c to exit from the program)
less filename A newer version of more

cp /s390/volA /s391/volX Copy first file to second file

find / -name unzip Start search in root and find file named “unzip”

updatedb Update the database used by the locate command (root)

locate unzip Locate all files with “unzip” in their name
file /usr/bin/unzip Describe nature of named file

mkdir /home/ogden/source Create a new directory

ls -al List current directory, in some detail
ls -lrt List current directory, most recently updated files last
mv oldfile newfile Rename or move a file
rm filename Delete a file
head filename List first 10 lines of file
tail filename List last 10 lines of file

80 EFS Systems on a Linux Base: Getting Started

 tar -cvxf mytar.tar /flexes/rundir Create tar file from files in /flexes/rundir
tar -tvf mytar.tar List files contained in “my.tar”
tar -xvf mytar.tar Extract all the files in the tar file
tar -zxvf mytar.tar.gz Extract from compressed tar file

unzip /dev/cdrom/os390/os39ra.zip -d/s390 Unzip a file on CD-ROM

mdir a: Display DOS diskette

mcopy /flexes/x/y a:x.txt Copy Linux file to a DOS diskette file
mdel a:afile.txt Delete a file on diskette
mformat a: Format a DOS diskette

shutdown -h now Shut down the Linux system

last Who has logged into Linux?

whoami Displays your userid

ifconfig Display current LAN adapter status

netstat -r Display current routing state
route add default gw 9.12.17.150 eth0 Create default gateway
route add -net 10.10.0.0 255.255.255.0 eth0 Add network route

top Monitor activity. “q” to quit

vmstat 10 5 5 activity reports. 10 seconds between reports.

mount /dev/cdrom /mnt/cdrom Mount a CD-ROM

mount /dev/fd0 /mnt/floppy Mount a diskette (but not a DOS diskette)
mount /dev/hdc2 /s391 Mount file system on second HDD at mount point /s391
mount Display the mount table
umount /dev/cdrom Unlock CD drive so you can remove CD

Multiple consoles, sessions, screens

Linux offers multiple consoles on the PC display. The keys Ctrl-Alt-Fn are used to select a
console. Ctrl-Alt-F1 through Ctrl-Alt-F6 produce traditional command lines, and Ctrl-Alt-F7
produces the X windows GUI desktop. Each of these consoles is a separate Linux session,
requiring a separate login. You can log in with different userids in each session, or use the
same userid in several sessions. You can switch between sessions at any time, using
Ctrl-Alt-Fn keys.

You can also have multiple command line windows open on the GUI desktop. These do not
require separate logins.

Text editors
Traditional UNIX users and many Linux users use vi as their primary text editor. Traditional
mainframe users often avoid vi, if possible. Gnome offers gedit. This is much like a simple PC
text editor and very unlike vi. It is suitable for almost all the text editing we required while
installing and customizing FLEX-ES. The Red Hat Linux distribution includes a number of
editors, and you can select the ones you like best.

Chapter 7. Basic discussions 81

82 EFS Systems on a Linux Base: Getting Started
 8

Chapter 8. FAQs
Q: I sometimes need to connect a USB device to my ThinkPad, but the FSI dongle is using
the USB port. Can I disconnect it? When?
A: You can disconnect it when FLEX-ES is not running. Linux appears to sense the insertion
or removal of the dongle and reacts appropriately. You may also consider the use of a USB
hub to attach multiple USB devices, including the FSI dongle.

Q: Can I disconnect the dongle after FLEX-ES is started?

A: We tried this. FLEX-ES continued to run for about a minute and then terminated. It
appears that the dongle is sensed at irregular intervals.

Q: If I install a new release of FLEX-ES (perhaps by downloading it), do I need a new license
key? A new dongle?
A: You probably need a new license key file, but can use your existing dongle. A license key
file may be used with the FLEX-ES release that corresponds to the license key level or with
any earlier release.

Q: How do I display my FLEX-ES release number?

A: Here are three different ways:
# rpm -qa flexes (from a Linux command prompt)
flexes> d release (from a flexes> prompt)
# resadm -V (if the resource manager is running)

Q: Does each S/390 user need a Linux userid?

A: No. A typical z/OS TSO user, for example, never “sees” Linux.

Q: Must I use vi to administer the Linux system?

A: No. Linux has a number of other editors, some of which are very “PC-like.” However, you
may find it easier to use vi to edit files whose names begin with a period.

Q: Why do I need to define emulated control units and channels? I did not need to do this with
emulated I/O on P/390 and MP3000 systems. For these machines, I can simply define
emulated devices.
A: There are two major reasons. The first is that working through (emulated) channels and
devices provides a more accurate emulation of S/390 hardware. In most cases, this has little
practical effect. The second reason involves emulating multiple S/390 instances, and sharing

© Copyright IBM Corp. 2003. All rights reserved. 83

 emulated I/O—for example, shared direct access storage device (DASD). This is a more
complex situation that does not exist on a P/390 and is not permitted (via LPARs) on an
MP3000. The emulated control units provide the logical interlock and queuing points for
sharing devices according to S/390 architectural specifications.

Q: How real is the channel emulation? Can I make my system faster by defining lots of
emulated channels and spreading out my I/O devices?
A: As best we could tell, the emulated channels are not very real in this sense. Many system
definitions we examined define only two or three channels of each channel type.

Q: I understand that a LAN MAC address is used to control FLEX-ES licenses in some cases.
Can I use this method instead of a dongle?
A: FLEX-ES licenses for systems based on Linux normally require the use of a dongle.

Q: Can I run a ThinkPad in 24x7 operation? (24 hours/day, 7 days/week)

A: There is no rule against it, and some ThinkPads are left running in docking stations.
However, in our opinion, this is not a reasonable solution for an EFS system. The
fully-loaded ThinkPad used for EFS develops considerable heat (on the bottom) and may
require additional air flow. In addition, a system intended for 24x7 S/390 operation should
have RAID disks.

Q: Can I use token-ring LAN adapters? You mention only Ethernet.

A: Yes, token ring can be used. We usually discuss Ethernet because it is what we used, and
the majority of EFS users have Ethernet LANs.

Q: How many Ethernet adapters should I have?

A: One should be sufficient. You can share this adapter (using multiple IP addresses)
between Linux and S/390 operating systems.

Q: Can I use the FSI channel adapters or ICA adapter with a ThinkPad? I could put them in a
docking station.
A: No. This is not supported.

Q: Can I have “odd size” 3390 drives, such as a 100 cylinder 3390 on a P/390?
A: No. Only standard size drives are emulated at this time.

Q: An emulated 3390-3 was 3339 cylinders on my P/390. Why is it 3343 cylinders under
FLEX-ES?
A: A standard 3390-3 has 3339 cylinders of directly usable space. The additional cylinders
include the “CE cylinder” (for testing) and alternate tracks. P/390 emulation did not include
these additional cylinders, while FLEX-ES emulation does include them. In normal use, you
do not see any difference between the two methods.

Q: Are 4mm tapes compatible between my MP3000 and my xSeries EFS system?
A: Yes, they should be, for those EFS systems that support 4mm drives.

Q: How many emulated S/390 devices can I have?

A: There is a FLEX-ES maximum of 2048 emulated devices per S/390 instance. If you
emulated two S/390s, each has this maximum number of devices.

Q: I have a fixed amount of real memory in my system. You say that I should not overcommit
this for S/390 emulation. Should I make my emulated S/390 main storage as large as possible
(without triggering Linux swapping) or should I define some expanded storage at the expense
of main storage?
A: Provided that you do not exceed the maximum of approximately 2 GB permitted for main
storage, we think that server memory is better used as S/390 main storage than as S/390
expanded storage. However, you might have specific applications that require expanded

84 EFS Systems on a Linux Base: Getting Started

storage, so we cannot provide an absolute answer. You face the same quandary when
dividing up real memory on an MP3000, for example, between main storage, expanded
storage, and a disk cache. If you want to define the largest possible S/390, you should read
Chapter 2 in S/390 PartnerWorld for Developers, ITSO/EFS Project EFS Systems on a Linux
Base: Additional Topics, SG24-7008, for a discussion of various tradeoffs.

Q: Can I write FakeTape files from my Linux applications?

A: Yes. The format is well documented. APIs are provided if you work under Linux.

Q: How many x3270 sessions or TN3270 sessions can I have on the Linux desktop?
A: There is no particular limit. Reasonable screen usage may indicate a practical maximum of
something like three or four on a single panel.

Q: Should I use Terminal Solicitor sessions (via Linux TCP/IP) or direct z/OS TCP/IP
connections for my TSO and CICS users?
A: Assuming z/OS TCP/IP is configured and working, it does not appear to make much
difference. z/OS TCP/IP connects TN3270 sessions to VTAM and whether VTAM is using a
local, non-SNA 3270 (as emulated by the Terminal Solicitor) or working through z/OS TCP/IP
has no observable effect at the end-user level. Systems programmers may prefer to work
through the Terminal Solicitor because it is operational even if z/OS TCP/IP is not working.
You can use odd-sized 3270 sessions through z/OS TCP/IP and through the Terminal
Solicitor (starting with FLEX-ES version 6.2.10).

Q: When I connect via z/OS TCP/IP, I do not get the terminal logo I always receive if I connect
via the Terminal Solicitor. Why not?
A: This may be due to your z/OS TCP/IP setup. The z/OS TCP/IP PROFILE needs to point to
a USSTAB module that generates the logo. Recent AD releases have this set up.

Q: Can I connect to z/OS with an ASCII Telnet session? I want to use vi.
A: Yes, although this is not really an EFS question. Recent AD releases have port 1023 (of
z/OS TCP/IP) configured for ASCII Telnet sessions. If you are not using an AD release, you
need to configure z/OS TCP/IP.

Q: I have the impression there are frequent FLEX-ES releases. Do I receive these
automatically? When should I upgrade? Where is there a concise list of what is new in each
release?
A: In general, releases are obtained through the FSI Business Partner from which you
purchased the system. We suggest you should upgrade only when a later release contains a
feature you need or when your FSI Business Partner recommends an upgrade. Remember
that you need a new license key file for a new release.

Q: Where can I order a printed copy of the FLEX-ES manuals?

A: Consult your FSI Business Partner for this. FSI does not routinely produce printed
manuals.

Q: Can I use the system for other purposes when FLEX-ES is not running? Can I run Linux
server functions on it when FLEX-ES is running?
A: When FLEX-ES is not running, you can run anything you want on your system—taking
care not to disturb the FLEX-ES customization. When FLEX-ES is running, you could still run
other Linux workload. However, we generally recommend against this. The external
workload may force Linux swapping and this can have negative effects on FLEX-ES operation
and performance. Having gone to considerable effort to obtain a small S/390 (in the form of
your EFS system), we suggest you use it only as a S/390. If you need to also run a Linux
server (at the same time), we suggest that you obtain another machine.

Chapter 8. FAQs 85
 Q: Should I purchase an Uninterruptible Power Supply (UPS) for use with my EFS machine?
A: The ThinkPad battery provides, in effect, a UPS function. You might consider an external
surge protector. For an xSeries EFS system, we believe a UPS is a good idea. (You may need
to configure the BIOS options correctly for a ThinkPad in a docking station to continue
operation (on the battery) if the AC power fails.)

Q: How do I order an EFS system? What if I am not a PWD member?

A: You do not need to be a PWD member to order an EFS system (although you do not have
access to AD CD-ROM systems or some of the other S/390 operating system packages
mentioned in this document). The FSI Web site has a contact point for locating their Business
Partners.

Q: Where can I find documentation about the various FLEX-ES traces?

A: There are several FLEX-ES traces:
– The processor event trace is always active and uses a permanently defined circular
buffer. It is intended only for FLEX-ES debugging and is not documented.
– The processor instruction trace uses the tracesize buffer you define in your system
definition. It is intended for FLEX-ES debugging but might be useful to more
sophisticated customers. It is minimally documented at present, but FSI intends to
provide better documentation.
– Device traces, which are usually obtained with the devtrsnap command. These are not
documented at present, but a user familiar with channel programming techniques
might find these traces useful.

The traces are not well documented at this time.

Q: Are any modifications, PTFs, or SPEs required for z/OS on EFS?

A: No.

Q: Are printer FCBs emulated?

A: Yes. FLEX-ES printer emulation supports the Load Forms Control Buffer channel
command. The default FCB provides 66 lines at 6 lines/inch, with a “channel 1 punch” at line
3 and a “channel 12 punch” at line 64.

Q: Is a default 1403 carriage tape emulated?

A: Yes. It is set for 66 lines at 6 lines/inch, and has a channel 1 “punch” at line 3. There is no
way to alter these settings.

Q: Are token ring and Ethernet equally supported?

A: Yes, but remember that a Linux patch may be needed to use token ring.

Q: Can I manage an ESCON director from an EFS system?

A: At the time of writing, the FSI ESCON adapters were not yet available and we were unable
to address this question.

Q: I am interested in using FLEX-ES remote resources for disks and tapes. Do I need a
FLEX-ES license just to run the resource manager on remote systems?
A: Yes, you need a FLEX-ES license for every system running the resource manager.

Q: If I obtain a faster server, do I need a new license? Do I need to notify someone?

A: PWD members should ask their Business Partner if a new license key file is needed. You
may need to manage any potential S/390 software fee changes related to the increased MIPS
of your new machine. This is obviously a complex area, and these processes may change in
the future. For example, your basic FLEX-ES license may contain a speed limit based on your
S/390 software license.

86 EFS Systems on a Linux Base: Getting Started

Q: What is the difference between swapping and paging?
A: None, in these discussions.

Q: Various people mention different MIPS ranges for EFS systems. What is correct?
A: An emulated S/390 has a much wider range of MIPS than a “real” hardware S/390. Any
MIPS measurement is very strongly dependent on the instruction mix and the addressing mix
involved, as well as the emulation techniques used. We could probably contrive code to
produce anything from below one MIPS to 60+ MIPS. We try to use conservative numbers in
this series of redbooks, while marketing representatives tend to use optimistic numbers. An
exact MIPS number simply does not exist.

Q: How is floating point performance?

A: It is acceptable for casual use, but you would probably not select this platform for major
floating point applications. Binary (IEEE) floating point has considerably better performance
than hex floating point.

Q: There are always new Linux functions available, as well as kernel fixes and upgrades.
Should we install these? Do we need to reinstall any part of FLEX-ES afterward?
A: If you have a stable system, we suggest you do not apply Linux changes. You should
regard your EFS system as a S/390 and not as a Linux machine. If you rebuild the Linux
kernel, you need to reinstall the FLEX-ES msgmgr package and you need to make a special
effort to make the USB dongle interface work. You must use a kernel that matches the kernel
source tree present when msgmgr was installed.

Q: Can I use DHCP to obtain a LAN IP address?

A: You can do this for Linux TCP/IP, but not for z/OS TCP/IP. Your EFS system is essentially a
S/390 server, not a client, and you may need a stable IP address for it.

Q: How difficult is it to change the IP address of my system?

A: You are likely to have two IP addresses: one for Linux and one for z/OS. To change the
z/OS address you must edit the PROFILE data set and then restart TCP/IP. If you are sharing
an Ethernet adapter with Linux, you also need to change the IP address in the FLEX-ES
resource stanza that defines the emulated 3172. To change the Linux IP address, use:
(gnome): Programs --> System --> Network Configuration. Select the Interfaces tab.
Deactivate the appropriate interface. Edit the interface to change the IP
address. Activate the interface. Save. Quit.

Q: As a developer, I sometimes need a real hardcopy listing. You provide very little
information about setting up printers. Why?
A: See the appropriate chapter in S/390 PartnerWorld for Developers, ITSO/EFS Project EFS
Systems on a Linux Base: Additional Topics, SG24-7008, for more information about printing.

Q: How long does the ThinkPad battery last?

A: With our fully loaded T20 system (512 MB, two disk drives, heavy computational load) the
battery usually lasted about 45 minutes. Newer ThinkPads use less power and have better
batteries, so the duration may be longer.

Q: Is SNA available over Ethernet?

A: Yes.

Q: Can I use a locally administered MAC address for Ethernet?

A: We have never done this, although it appears that some Ethernet drivers may support this
option.

Q: How many Ethernet adapters do I need if I want to use SNA and TCP/IP under z/OS?
A: One.

Chapter 8. FAQs 87
 Q: You did not say much about the ICA adapter. Can I emulate EP lines (or PEP lines)?
A: We did not say much about this FSI adapter because z/OS does not support ICA devices.
No, you cannot use it for EP (or EP under PEP) lines. This mode is not emulated.

Q: My xSeries has four processors, with one licensed for S/390 emulation. Can I also license
more processors for S/390 emulation? Will this make a faster system? How much faster?
A: Yes, you can license more processors for S/390 emulation. You cannot dedicate all
processors for this purpose, because this would leave no processor for other required Linux
functions and auxiliary FLEX-ES processes.

Remember that enabling more processors for S/390 work changes the MIPS rating of your
machine and may affect the license costs for much of your S/390 software.

Q: Why is a dedicated processor faster? What makes it different?

A: A complete context switch for a processor creates additional overhead for the “hardware”
of the system. The processor cache contents (L1 and L2, typically) become useless and
many cache misses occur for a while. Likewise, the virtual address translation mechanism
must be refreshed with new address ranges. A processor dedicated to S/390 emulation has
fewer context switches and avoids this overhead. We have no specific measurements on
current hardware as to how much performance is gained by dedicated processors.

Q: Does all the FLEX-ES work occur only on the licensed number of processors? Does it ever
use the additional processor or processors?
A: All the S/390 instruction emulation is done only on the licensed number of Pentium
processors. Some of the I/O and control processes run as normal Linux processes and are
dispatched on the “other” Pentium processors.

Q: I have an older S/390 and am interested in changing to an xSeries EFS machine. I

currently use seven channels on my machine. Do I need this many channels on the xSeries
EFS system?
A: Probably not, but this takes careful analysis. In this redbook, we assume that your DASD
will be emulated on internal server disks. (Indeed, nothing else was supported at the time of
writing. However, FSI plans to support external CKD DASD attachment through both parallel
and ESCON channels.) Even if external DASD can be connected, the internal emulated
DASD is probably faster and larger. This topic (channel analysis) is highly specific to each
situation and we suggest you work with a knowledgeable Business Partner who is trained on
xSeries EFS configurations.

Important: Remember that no channels, parallel or ESCON, are available for Linux-based
FLEX-ES systems at the time of writing. The comments here are based on experience with
UNIX-based FLEX-ES systems and are likely to apply to Linux-based systems if they
become available.

Q: Can an ESCON adapter (on an EFS system) go through an ESCON director? How is it
configured for director port addresses?
A: The FSI ESCON adapter was not available at the time of writing, but we expect that it will
be usable through ESCON directors.

Q: How many parallel channel adapters can I have?

A: The limitation is usually the number of PCI slots available in your machine. A secondary
limitation may be with cable connectors. There is no reasonable limit from the software
viewpoint.

88 EFS Systems on a Linux Base: Getting Started

Q: I am interested in a substantial number of parallel channels, say fifteen, on my xSeries
EFS system. How are cables handled? Is this a problem?
A: This would require five FSI parallel channel adapter cards, with three channels per card.
FSI provides cables as part of the adapter package. These cables have small connectors on
one end (for the adapter card) and standard bus and tag connectors on the other end. There
are two basic issues: “finger space” at the back of the server becomes very tight when several
of these adapters are used, and strain relief for the cables is extremely important. You must
plan your cable routing and physical support. FSI can provide a strain relief structure or you
could construct your own. In any event, this number of cables cannot simply hang from the
back of the server. You should talk with your Business Partner about your plans.

Q: How much electrical power does your xSeries EFS system use? How much does it cost to
run all the time?
A: This varies with system configurations, of course. We informally measured the current
used by our Netfinity 5100 system. Once started, it averages about 2.0 amps (at 120 volts)
when the disks are active and the flat panel console display is on. When the system is idle
and the console is off, it uses about 1.5 amps. Assuming a power factor near 1, this is a range
of 180 to 240 watts. At typical U.S. electricity prices, it costs about $.60 (60 cents) per day to
run the system. You should provide power connections that meet the published requirements
for your server, and you should realize that startup currents are higher than the average
running currents mentioned here. Also, remember that a CRT display (instead of the flat panel
we used) substantially increases the power required.

Q: Why do you make the root file system so large? Why not make separate file systems for
/usr and /tmp?
A: We like to have ample work space. Again, the goal is to minimize activity in /s390 that
might lead to disk fragmentation. You can create separate file systems for /usr and /tmp (and
anything else you want) while installing Linux. If the sole purpose of the system is to run
S/390 emulation, there is no particular need for separate file systems (partitions) for various
traditional parts of Linux.

However, we note that you might improve long-term stability by making a separate file system
for /tmp and /var. If some logging function (assuming it writes to /tmp) fills up /tmp, the system
may crash. It is less likely to crash if /tmp is a separate file system. We have not had any such
problems, but it could be a consideration for a production operation.

Q: Can I upload from diskettes using the IND$FILE function?

A: Yes. We used PCOM to log onto TSO (from another PC). We then used PCOM’s SEND
and RECEIVE commands at the PC. If you upload/download using x3270, you must ensure
that the Linux file system involved is mounted. If it is on a diskette, you can use:
mount -t msdos /dev/fd0 /mnt/floppy

Q: Your examples always assign a Linux file name that is the same as the volser of the S/390
volume in that file. I used a different convention, but now I forget which volser is stored in
which Linux file. How can I easily check the logical contents of a file?
A: Assume you have an emulated 3390 volume in /my/linux/files/AAA and you want to
determine the volser. Use the ckddump command:
$ ckddump /my/linux/files/AAA 0 0 | more

Then scroll to the third record. This is a standard VOL1 label and the volser starts in byte five.
The ckddump command produces a storage dump type of listing (in hex and characters) and
includes all the CKD fields of the volume. A Ctrl-c keyboard function terminates ckddump
output. If you want to see only the volser, use:
$ ckddump /my/linux/files/AAA 0 0 | grep VOL1

Chapter 8. FAQs 89
 Q: You said that Parallel Sysplex is not supported. Can I run basic sysplex, using shared
DASD for coupling data sets?
A: We have had reports of a user or users running and testing base sysplex on FLEX-ES
6.1.15 (or later) using emulated CTC connections between the emulated S/390 instances, but
it can be very complex and is not supported for production use.

Q: Can I run z/OS under z/VM? Can I emulate sysplex functions under z/VM?
A: Yes and yes (but only for basic susplex).

Q: Using the installation processes described in this redbook (and not using raw devices for
emulated volumes), it seems that the AD CD-ROMs install faster if the target files do not
already exist. That is, if I do not overwrite existing emulated volumes. Is this normal?
A: We have noticed the same effect. It might be faster to delete the existing volumes and
create new ones (using the unzip and ckdconvaws piped commands described in 5.1.2,
“Installation tasks” on page 48). New AD systems are not installed often enough to worry
much about the installation speed, so we usually overwrite old volumes (for example, rename
Z4RES1 to S4RES1 and overwrite it) and ignore any performance issues.

Q: Looking through the Red Hat list of services, I see something named FLEXES09 that is
enabled. The comments are about token ring MAC addresses. I do not have token ring. Do I
need this?
A: This is part of FLEX-ES support for SNA on token ring. It is a small component that must
run (if needed) during system startup. Leave it as it is.

90 EFS Systems on a Linux Base: Getting Started

 A

Appendix A. Sample FLEX-ES operation files

This appendix lists the FLEX-ES definitions and shell scripts we used when installing our EFS
systems and writing this redbook. These files were copied from our working systems and
represent the actual controls and scripts we used.

© Copyright IBM Corp. 2003. All rights reserved. 91

Definitions for the Dallas DEMOpkg
The following FLEX-ES definition file was used to load and IPL the first five volumes of the
Dallas z/OS 1.4 DEMOpkg. This is a subset of the FLEX-ES definition samples included on
the CDs. Also, the CDs contained separate files for the system definition and the resource
definitions. We combined these into a single definition file to be more consistent with our other
examples.
system DDS:
memsize(524288)
cachesize(4096)
instset(z)
tracesize(512)
cpu(0)
channel(0) local
channel(1) local
channel(2) local
channel(3) local
channel(4) local
cu devad(0x300,5) path(2) resource(RSC39902)
cu devad(0x460,4) path(3) resource(RSC31741)
cu devad(0x470,4) path(3) resource(RSC31742)
cu devad(0x5DE,2) path(4) resource(RSC3088)
end DDS

resources DDR:
RSC31741: cu 3174
interface local(1)
device(00) 3278 Tso1
device(01) 3278 Tso2
device(02) 3278 Tso3
device(03) 3278 MstrCon
end RSC31741

RSC31742: cu 3174
interface local(1)
device(00) 3278 Tso4
device(01) 3278 Tso5
device(02) 3278 Tso6
device(03) 3278 Tso7
end RSC31742

RSC39902: cu 3990
interface local(1)
device(00) 3390-3 /s390/dmtres
device(01) 3390-3 /s390/dmtcat
device(02) 3390-3 /s390/dmtos1
device(03) 3390-3 /s390/dmtos2
device(04) 3390-3 /s390/dmtos3
end RSC39902

RSC3088: cu 3172
options ‘ipaddress=192.168.0.111’
interface local(1)
device(00) 3172 eth0
device(01) 3172 OFFLINE
end RSC3088

end DDR

92 EFS Systems on a Linux Base: Getting Started

 The following shell script was used to start operation. It opens three 3270 sessions on the
Linux desktop:
flexes DDS.syscf
x3270 -model 4 -port tn3270 localhost:MstrCon &
x3270 -model 4 -port tn3270 localhost:Tso4 &
x3270 -model 4 -port tn3270 localhost:Tso5 &
flexescli localhost DDS

Minimal AD system definitions

Our system and resource definitions (both in the same file) used for the four volumes required
to IPL the z/OS Release 1.4s AD CD-ROM system were as follows:
system S14A:
memsize(524288)
cachesize(4096)
instset(z)
cpu(0)
channel(0) local
channel(1) local
channel(2) local
cu devad(0xA80,8) path(2) resource(CU3990)
cu devad(0x700,5) path(0) resource(CU3174)
cu devad(0xE20,2) path(1) resource(CU3172)
end S14A

resources R14A:
CU3990: cu 3990
interface local(1)
device(00) 3390-3 /s390/S4RES1
device(01) 3390-3 /s390/S4RES2
device(02) 3390-3 /s390/OS39M1
device(03) 3390-3 OFFLINE
device(04) 3390-3 OFFLINE
device(05) 3390-3 OFFLINE
device(06) 3390-3 OFFLINE
device(07) 3390-3 /s390/S4USS1
end CU3990

CU3174: cu 3174
interface local(1)
device(00) 3278 mstcon
device(01) 3278 L701
device(02) 3278 L702
device(03) 3278 L703
device(04) 3278 L704
end CU3174

CU3172: cu 3172
interface local(1)
options 'ipaddress=192.168.0.111,adapternumber=0'
device(00) 3172 eth0
device(01) 3172 OFFLINE
end CU3172

end R14A

Appendix A. Sample FLEX-ES operation files 93

 The file names shown, such as S4RES1, are typical for the AD 1.4s release. The previous
AD 1.4 release typically used names such as Z4RES1.

Some FLEX-ES users elect to define emulated 3270 devices as OFFLINE and then use
FLEX-ES mount commands (in a shell script) to define and name terminals. For our small
system, we found it easier to simply define a number of 3270 terminals (with names such as
L701) in our resources definition.

The S/390 volumes used for our initial IPL are normally mounted at addresses A80, A81,
A82, and A87 (for the four system volumes). The “gap” between A82 and A87 required the
definition of several OFFLINE 3390 volumes. (In this case, the device model (3390-1, -2, -3)
is not meaningful.)

It would be possible, of course, to mount these volumes at any 3390 addresses contained in
the z/OS IODF. We could have placed the A87 volume on A83 and avoided defining the
OFFLINE devices, for example. However, other AD CD-ROM documentation (including
several Redbooks) uses the addresses we selected and we want to preserve compatibility
with this existing documentation. This is the only reason for inserting OFFLINE devices to
manipulate the addresses.

The 3172 definition is not required for a minimal system. You could omit it and the
corresponding cu definition. However, many users want to establish their z/OS TCP/IP
connectivity, even when working with a minimal AD system and the 3172 is needed for this.

We strongly suggest that new EFS users study this example until every detail is understood. It
is a simple example, and contains nothing clever or unusual.

After using cfcomp to compile the above definitions, we used this shell script to start
operation. (A resadm -s R14A.rescf command must be issued first.)
flexes S14A.syscf
xmodmap -e 'keysym Alt_L = Alt_L Meta_L'
xset fp+ /usr/flexes/fonts
xset fp rehash
x3270 -model 3 -keymap pc -port tn3270 localhost:mstcon &
x3270 -model 3 -keymap pc -port tn3270 localhost:L701 &
flexescli localhost S14A

Full AD system definitions

The following definitions assume that all the AD CD-ROM volumes have been loaded. The
definitions also assume that three scratch volumes have been created (WORK01, WORK02,
WORK03) and that a tape drive may be used. (The tape drive could be a real SCSI drive or a
FakeTape drive.) More 3270 terminals are defined for use by remote users via the Terminal
Solicitor. Note that we used system name S14B and resource name R14B so we could have
the minimal and full definitions in the same directory.

The file names shown, such as S4RES1, are typical for the AD 1.4s release. The previous
AD 1.4 release typically used names such as Z4RES1. In these definitions we assume that
all the S/390 emulated volumes fit in one Linux file system (/s390). If this is not the case, you
need to modify the file names as needed.
system S14B:
memsize(524288)
cachesize(4096)
instset(z)
cpu(0)

94 EFS Systems on a Linux Base: Getting Started

channel(0) local
channel(1) local
channel(2) local
cu devad(0xA80,14) path(2) resource(CU3990)
cu devad(0xA90,3) path(2) resource(CU3991)
cu devad(0x700,20) path(0) resource(CU3174)
cu devad(0xE20,2) path(1) resource(CU3172)
cu devad(0x560,1) path(2) resource(CU3480)
end S14B

resources R14B:
CU3990: cu 3990
interface local(1)
device(00) 3390-3 /s390/S4RES1
device(01) 3390-3 /s390/S4RES2
device(02) 3390-3 /s390/OS39M1
device(03) 3390-3 /s390/S4DB21
device(04) 3390-2 /s390/S4CIC1
device(05) 3390-3 /s390/S4DIS1
device(06) 3390-3 /s390/S4DIS2
device(07) 3390-3 /s390/S4USS1
device(08) 3390-3 /s390/S4DIS3
device(09) 3390-2 /s390/S4IMS1
device(10) 3390-3 /s390/S4WAS1
device(11) 3390-3 /s390/S4WAS2
device(12) 3390-3 /s390/SARES1
device(13) 3390-3 /s390/S4DIS4
end CU3990

CU3991: cu 3990
interface local(1)
device(00) 3390-1 /s390/WORK01
device(01) 3390-1 /s390/WORK02
device(02) 3390-1 /s390/WORK03
end CU3991

CU3174: cu 3174
interface local(1)
device(00) 3278 mstcon
device(01) 3278 L701
device(02) 3278 L702
device(03) 3278 L703
device(04) 3278 L704
device(05) 3278 L705
device(06) 3278 L706
device(07) 3278 L707
device(08) 3278 L708
device(09) 3278 L709
device(10) 3278 L70A
device(11) 3278 L70B
device(12) 3278 L70C
device(13) 3278 L70D
device(14) 3278 L70E
device(15) 3278 L70F
device(16) 3278 L710
device(17) 3278 L711
device(18) 3278 L712
device(19) 3278 L713
end CU3174

Appendix A. Sample FLEX-ES operation files 95

 CU3172: cu 3172
interface local(1)
options 'ipaddress=192.168.0.111'
device(00) 3172 eth0
device(01) 3172 OFFLINE
end CU3172

CU3480: cu 3480
interface local(1)
device(00) 3480 OFFLINE
end CU3480

end R14B

There are many arbitrary choices in these definitions. You need to examine the definitions
and customize them to your purposes.

This shell script is just like the first one, except that it uses system name S14B instead of
S14A.
flexes S14B.syscf
xmodmap -e 'keysym Alt_L = Alt_L Meta_L'
xset fp+ /usr/flexes/fonts
xset fp rehash
x3270 -model 3 -keymap pc -port tn3270 localhost:mstcon &
x3270 -model 3 -keymap pc -port tn3270 localhost:L701 &
flexescli localhost S14B

Customized Offerings Driver definitions

We used the following FLEX-ES definitions and shell script with the Customized Offerings
Driver:
system SCOD:
memsize(262144)
cachesize(2048)
instset(esa)
cpu(0)
channel(0) local
channel(1) local
channel(2) local
cu devad(0x320,2) path(2) resource(CU3990)
cu devad(0xF00,3) path(0) resource(CU3174)
cu devad(0x0A1,1) path(0) resource(CU3175)
cu devad(0x390,1) path(2) resource(CU3490)
end SCOD

resources RCOD:
CU3990: cu 3990
interface local(1)
device(00) 3390-3 /s390/TEMP01 #address 320
device(01) 3390-3 /s390/TEMP02 #address 321
end CU3990

CU3174: cu 3174
interface local(1)
device(00) 3278 LF00 #address F00
device(01) 3278 LF01
device(02) 3278 LF02

96 EFS Systems on a Linux Base: Getting Started

 end CU3174

CU3175: cu 3174
interface local(1)
device(00) 3178 mstcon #address 0A1
end CU3175

CU3490: cu 3490
interface local(1)
device(00) 3490-E /dev/sg1 #address 390
end CU3490

end RCOD

The following is the shell script we used to start the system:

flexes SCOD.syscf
xmodmap -e ‘keysym Alt_L = Alt_L Meta_L’
xset fp+ /usr/flexes/fonts
xset fp rehash
echo ‘Verify your master console is active’
x3270 -model 3 -keymap pc -port tn3270 localhost:LF00 &
x3270 -model 3 -keymap pc -port tn3270 localhost:mstcon &
flexescli localhost SCOD

z/VM OMA/2 definitions

The following provide a minimum system definition for the installation of z/VM from OMA/2
CDs:
system VMS:
memsize(524288)
cachesize(4096)
instset(z)
tracesize(512)
cpu(0)
channel(1) local
channel(2) local
channel(3) local
cu devad(0x400,2) path(1) resource(CU3990)
cu devad(0x600,3) path(2) resource(CU3174)
cu devad(0x500,1) path(3) resource(CU3422)
end VMS

resources VMR:
CU3174: cu 3174
interface local(1)
device(00) 3278 L600 #address 600
device(01) 3278 L601
device(02) 3278 L602
end CU3174

CU3990: cu 3990
interface local(1)
device(00) 3390-3 /s390/VM1000 #address 400
device(01) 3390-3 /s390/VM2000
end CU3990

Appendix A. Sample FLEX-ES operation files 97

 CU3422: cu 3422
interface local(1)
device(00) 3422 OFFLINE #address 500
end CU3422

end VMR

We used the following shell script with these definitions:

flexes VMS.syscf
x3270 -model 3 -keyboard pc -port tn3270 localhost:L600 &
x3270 -model 3 -keyboard pc -port tn3270 localhost:L601 &
flexescli localhost VMS

Definition statements
A system keyword is required in the system definition section. The name following the
keyword is the name of the syscf created when the definition is compiled. A colon must follow
the name.

memsize expresses kilobytes of storage and the largest permitted value is 2097152.

essize is for expanded storage and is expressed in megabytes. Furthermore, this parameter
must be a multiple of 16.

instset is either esa or z for the systems we discuss in this document. An esa parameter
causes 31-bit S/390 emulation and a z parameter enables 64-bit capable zSeries emulation.

cachesize is expressed as kilobytes of S/390 storage. We suggest using 4096 for this
parameter. This cache is used internally by FLEX-ES.

tracesize is 512 unless you are directed to a different size by FSI or your Business Partner.

cpu and channel numbers are arbitrary, but usually start with zero. You cannot have more
CPUs defined than you have licensed FLEX-ES processors.

The dedicated keyword on a cpu statement causes FLEX-ES to reserve an Intel processor
solely for S/390 emulation. This improves performance. You must have at least one Intel
processor that is not dedicated.

Additional information about FLEX-ES definitions can be found in S/390 PartnerWorld for
Developers, ITSO/EFS Project EFS Systems on a Linux Base: Additional Topics, SG24-7008.

98 EFS Systems on a Linux Base: Getting Started

Related publications

The publications listed in this section are considered particularly suitable for a more detailed
discussion of the topics covered in this redbook.

IBM Redbooks
For information about ordering these publications, see “How to get IBM Redbooks” on
page 100. Note that some of the documents referenced here may be available in softcopy
only.
򐂰 S/390 Partners in Development: OS/390 (and z/OS) New User’s Cookbook, SG24-6204
򐂰 S/390 PartnerWorld for Developers, ITSO/EFS Project EFS Systems on a Linux Base:
Additional Topics, SG24-7008

Other publications
These publication z/VM Installation Guide, GC24-5992, is also relevant as a further
information source. You may also want to refer to the following publications from Fundamental
Software, Inc.:
򐂰 FLEX-ES Technical FAQ, FSIMM040
򐂰 FLEX-ES Planning Guide, FSIMM100
򐂰 FLEX-ES Operator’s Guide, FSIMM200
򐂰 FLEX-ES CLI Language Reference, FSIMM210
򐂰 FLEX-ES System Programmer’s Guide, FSIMM300
򐂰 FLEX-ES Resource Language Reference, FSIMM310

Online resources
These Web sites and URLs are also relevant as further information sources:
򐂰 Fundamental Software, Inc.:
http://www.funsoft.com
򐂰 Sistina home page
http://www.sistina.com

© Copyright IBM Corp. 2003. All rights reserved. 99

How to get IBM Redbooks
You can search for, view, or download Redbooks, Redpapers, Hints and Tips, draft
publications and Additional materials, as well as order hardcopy Redbooks or CD-ROMs, at
this Web site:
ibm.com/redbooks

Help from IBM

IBM Support and downloads
ibm.com/support

IBM Global Services

ibm.com/services

100 EFS Systems on a Linux Base: Getting Started

Index
consoles, sessions, screens 81
Symbols control units and channels, which 83
.flexeslicense file 42 Coupling Facility processors 7
/etc/init.d/flexes file 65 cryptographic hardware 7
/hda terminology 25 CU will not start 76
/sda 25 Customized Offerings Driver definitions 96
/usr/flexes/bin 58, 72 Customized Offers Driver 50
/usr/flexes/rundir 59

D
Numerics Dallas DEMOpkg 52
1403 carriage tape 86 debugging steps 75
3339 cylinders 84 dedicate server processor 8
3390 drives, odd size 84 dedicated processor 88
3422 tape drive 54 desktop windows 58
4mm tapes, compatibility 84 Device Support Facilities (DSF) 50
64-bit operation 69 DFSMSdss 50
DHCP 87
A DHCP server 16
AD CD-ROM distributions 46 Disk Druid 24
Additional Topics 3 disk space 14
alternate boot 11, 20 size 5
APAR OA03594 78 diskette drive 18
apm=off parameter 29 T40 36
app-defaults version 41, 77 DLIBs 15
Architectural Level Set 3 (ALS3) 69 DMA for disk accesses 36
ASCII Telnet sessions 3, 9, 85 docking station 78
AWSCKD format 47 dongle 5, 42, 83–84
AWSTAPE 46 problems 78
dtprint command 78
dual boot 11, 20
B dual-boot system 5
battery life 87 Dual-port IBM 10/100 adapter 33
BIOS and RAID adapter updates 11, 21 dump/restore distribution 50
BIOS settings 17 dump/restore format 45
BIOS upgrades 21
boot diskette 30
Boot Loader Configuration 25 E
boot partition 15 ECC memory 79
Broadcom Tigon3 adapter 33 EFS systems 1
Business Partners, IBM 2 emulated S/390 devices, maximum 84
bzImage 33 emulated volumes 69
Enabled For S/390 1
EP lines 88
C ESCON channel adapters 5
CD-ROM drive 18, 25 ESCON director 86, 88
CD-ROM formats 47 Ethernet (1000 Mbit) 33
cfcomp command 58–59 Ethernet adapters
channel emulation 84 how many 84
ckdconvaws command 47–48, 77 number 87
ckddump command 89 Ethernet interface 3
ckdfmt command 70, 77 Ethernet port 29
CLI program 60 ext3 file systems 15
clock accuracy 28
COD system 50
command line interface (CLI) 9

© Copyright IBM Corp. 2003. All rights reserved. 101

F IND$FILE function 89
FakeTape 46, 85 instance
FCBs emulated 86 definition 3
fdisk 24 multiple 9
fdisk utility 15 Intel PRO/100 VE Network Connector 33
firewall security option 26 Intel PRO/1000 Gigabit Ethernet 33
FLEX-ES IOCDS 62
console 7 IODF 46
definition file 57 IODF requirements 62
formats 47 IP address 16
installation 39–40 changing 87
instance 3 ipl command 62
license 1 IPL z/OS 62
license key, installation 42
manuals 85 K
mount command 71 KDE 27
operation 68 kernel fixes and upgrades 87
product 1 kernel name 31
release number 83 kernel source tree 12, 31
releases 85 key mapping, x3270 72
remote resources 86
trace files 77
traces 86 L
upgrades 43 LAN adapter
flexes command 60, 77 new 30
flexes program 57 shared 63
FLEXES09 service 90 LAN adapters 9, 26
flexes-6.2-9.i386.rpm 40 LAN network 16
flexescli program 57 Large memory 33
floating point performance 87 license key 42, 86
FSI 1 license key file 83
FSI channel adapters 84 Linux commands 80
FSI dongle 83 Linux distribution 11
FSI patches 31 Linux file names 46
ftlib-6.2-9.i386.rpm 40 Linux file system layout 15
Fundamental Software, Incorporated 1 Linux for S/390 3
Linux kernel 30
Linux windows 68
G logical partitions 3
gedit 81 Logical Volume Manager 12, 23
Gnome 27 logo, USSTAB 85
gnome 28, 30 LPAR 3, 7
grub 20

M
H MAC address 84, 87
hard disk on ThinkPad 28 make clean 32
hda 24 make mrproper 32
hdc=ide-scsi 25, 28 make xconfig 32
hdparm command 36 Makefile 32
High Memory Support 33 memory
HIGHMEM I/O 33 server 9
HiperSockets 7 size 5
hwc command 71 MIPS 2, 87
MIPS range 2
I mismatch, kernel 76
IBM distribution formats 45 mount command 94
ICA (integrated communications adapter) 5 mount point /s390, ownership 41
ICA adapter 84, 88 MPTS adapter number 64
ICKDSF job 70 msgmgr, problems 75

102 EFS Systems on a Linux Base: Getting Started

msgmgr-6.2-9.i386.rpm 40 rpm -U option 44
multiple paths 7 RSU (service) tapes 56

N S
N3270 sessions, limit 85 S/390 adapter 4
Netfinity 3 S/390 central storage 84
new emulated DASD volume 69 S/390 CPU, definition 3
S/390 expanded storage 84
S/390 identification 73
O S/390 operating system 45
OMA/2 46 S/390 parallel channels 88
OMA/2 distribution is for z/VM 53 screen saver 30
Operating Systems Messages console 71 SCSI-connected tape drives 5
OS/390 3 security 79
OS/390 device configuration 47 SEND and RECEIVE commands 89
OSA-Express 7 ServeRAID Manager 21
ServeRAID-4Mx Ultra160 SCSI Controller 5
P ServerPacs 45
parallel channel adapter 5, 88–89 shell script 60
Parallel Sysplex 7 shutdown process 69
PartitionMagic 20 shutting down 64
PartnerWorld for Developers 2, 46 SMP kernel 29
PATH environment 58 SMP machine 13
PATH, default 41 SMP system 30
PC memory, large 30 SNA over Ethernet 78, 87
PCOM 4 source tree 31
PEP lines 88 SPECIFY SYSTEM PARAMETERS message 52
port 24 9, 61 staging 49
port replicators 18 Stand-alone z/OS restore 80
power consumption 89 Starter System 50
power supplies, redundant 79 startup process 68
PowerQuest 20 STIDP instruction 73
printers 87 stock Linux kernel 12
Processor type and features panel 33 STORAGEGROUP parameter 70
processor, definition 3 swap partition 15, 25
processors, use 88 sysplex
PTFs 86 basic 90
functions 90
system console 71
R system definitions 93
RAID configuration 11
steps 21
ram disk image 33 T
RAS discussion 79 T23 ThinkPads 33
raw device 15 T40 ThinkPads 13
Red Hat Linux 5, 12 tar file distribution 52
Red Hat Linux installation 23 tar format 40
Red Hat Linux updates 37 TCP/IP for z/OS 63
Redbooks Web site 100 TCP/IP port 555 79
Contact us x Telnet application 4
remote resources 86 terminal logo 85
resadm 57–58, 61, 68, 70 Terminal Solicitor 3, 7, 9, 61, 79, 85
resadm command 59, 62, 65, 77 terminology 3
resadm -r command 59 text editors 81
rescf command 65 ThinkPad docking station 18
resource definition 58, 68, 93 ThinkPad/EFS, name 3
resource manager 7 TN3270 client 9
Resource Measurement Facility (RMF) 63 TN3270 server 9
root file system 15, 89 TN3270 session 62
rpm commands 40, 43 TN3270e session 57

Index 103
token-ring LAN adapter 84
token-ring support 29
trace 86
file 77
I/O device 86
processor event 86
processor instruction 86
TSO and CICS users 85
TSO user 62

U
Ultrabay 28
Ultrabay usage 11, 17
Uninterruptable Power Supply (UPS) 86
UNIX System Services 4, 9
UnixWare 1
unzip program 48
up2date command 37
USB 2.0 13, 30, 33
USB 2.0 ports 13
USB 2.0 systems 36
USB device 83
USB hub 83
USB port 42
userids, Linux 83
USSTAB logo 62
USSTAB module 85

V
VATLST00 71
vi editor 83
VM/ESA 3
volser 89
VSE/ESA 3
VTAM APAR 78

X
X Customization 27
x3270 client 71
x3270 command 60, 72
x3270 keyboard modifications 44
x3270 problems 77
x3270 program 41
x3270 sessions 9, 85
xSeries EFS, name 3

Z
z/OS TCP/IP 9
z/OS TCP/IP profile 73
z/OS TCP/IP setup 85
z/OS UNIX System Services 3
z/VM 3, 46, 90

104 EFS Systems on a Linux Base: Getting Started

EFS Systems on a Linux Base: Getting Started
(0.2”spine)
0.17”<->0.473”
90<->249 pages
 Back cover ®

EFS Systems on a Linux

Base: Getting Started

Planning for ThinkPad This IBM Redbook describes the basic installation of FLEX-ES (a
product of Fundamental Software, Inc., Fremont, California) with
INTERNATIONAL
and Netfinity bases
several z/OS operating system packages. Much of the material is TECHNICAL
in cookbook format. More advanced techniques and options are SUPPORT
Installation,
customization, described in S/390 PartnerWorld for Developers, ITSO/EFS ORGANIZATION
operation Project EFS Systems on a Linux Base: Additional Topics,
SG24-7008. These books are intended primarily for members of
the IBM PartnerWorld for Development organization and for
Common techniques internal IBM users of the ITSO/EFS package.
BUILDING TECHNICAL
INFORMATION BASED ON
This second edition includes the latest updates to the first edition PRACTICAL EXPERIENCE
of this book. It also contains more information about using Red
Hat 9.0 with FLEX-ES Release 6.2.14. IBM Redbooks are developed
by the IBM International
Technical Support
Organization. Experts from
IBM, Customers and Partners
from around the world create
timely technical information
based on realistic scenarios.
Specific recommendations
are provided to help you
implement IT solutions more
effectively in your
environment.

For more information:

ibm.com/redbooks

SG24-7007-01 ISBN 0738499404