IBM - NFS in AIX

Front cover
Securing NFS in AIX

IX
An Introduction to NFS V4 in AIX 5L
Version 5.3
Set up and administer NFS V4 on your

systems
Command use changes and

other differences since V3
Learn to use NFS V4 in a

clustered environment
Chris Almond
Lutz Denefleh
Sridhar Murthy
Aniket Patel
John Trindle
ibm.com/redbooks
International Technical Support Organization
Securing NFS in AIX
November 2004
SG24-7204-00
Note: Before using this information and the product it supports, read the information in
“Notices” on page ix.
First Edition (November 2004)
This edition applies to NFS Volume 4 running on AIX 5L.
© Copyright International Business Machines Corporation 2004. 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 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Trademarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
The team that wrote this redbook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv
Become a published author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Comments welcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Part 1. NFS V4 fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 1. NFS Version 4 overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.1 What is NFS? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 NFS V2 and NFS V3 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 NFS V4 design motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Objectives of NFS V4 (RFC3530) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 AIX 5.3 specific implementation of NFS V4. . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5.1 Mandatory features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5.2 Optional features. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Planning and implementation considerations . . . . . . . . . . . . . . . . . . . . . . . 7
1.6.1 Pre-implementation design considerations . . . . . . . . . . . . . . . . . . . . . 7
1.7 Looking ahead to the rest of the book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Chapter 2. What’s new in NFS V4? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.1 How NFS works. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Protocols used by NFS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 UDP or TCP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.2 Remote Procedure Call (RPC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.3 eXternal Data Representation (XDR) . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 NFS daemons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1 The portmap daemon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.2 The rpc.mountd daemon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.3 The rpc.statd daemon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.4 The rpc.lockd daemon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.5 The nfsd daemon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.6 The block I/O daemon (biod). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 NFS V3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.5 The NFS Lock Manager protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.6 NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
© Copyright IBM Corp. 2004. All rights reserved. iii

2.6.1 Attribute classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.6.2 Username to UID mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6.3 Better namespace handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6.4 Built-in security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6.5 Client-side caching and delegation . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.6.6 Compound RPC procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.6.7 File locking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.6.8 Internationalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.6.9 Volatile file handles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.7 AIX 5L v5.3 implementation of NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.8 NFS V4 supported features in AIX 5.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.8.1 Mandatory feature support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.8.2 Other unsupported features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.8.3 Optional feature support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.8.4 NFS4 ACL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.8.5 AIXC ACLs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.8.6 External name space (exname) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.8.7 Protocol differences: server exporting and client mounting . . . . . . . 35
2.8.8 NFS files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.8.9 Restricting NFS port ranges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.8.10 Use of NFS_NOBODY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.9 NFS daemons, files, and commands: a quick reference. . . . . . . . . . . . . . 41
Chapter 3. Enhanced security in NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.1 General security concepts and terminology . . . . . . . . . . . . . . . . . . . . . . . 46
3.1.1 Broad security categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.1.2 Information security components . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.1.3 RPC security flavors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1.4 RPCSEC_GSS protection levels. . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1.5 RPCSEC_GSS protection mechanisms . . . . . . . . . . . . . . . . . . . . . . 47
3.1.6 Looking ahead to the rest of the chapter. . . . . . . . . . . . . . . . . . . . . . 48
3.2 NFS V4 user/group identification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.1 User identity management options . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.2 User/group identities and NFS V4. . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3 NFS V4 user authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3.1 AUTH_SYS user authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3.2 RPCSEC_GSS user authentication using Kerberos . . . . . . . . . . . . . 59
3.4 NFS V4 user authorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.4.1 Standard UNIX file permissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.4.2 AIXC ACLs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.4.3 NFS V4 ACLs: description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.4.4 NFS V4 ACLs: ACL evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.4.5 NFS V4 ACLs: administration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
iv Securing NFS in AIX

3.4.6 NFS V4 ACLs: permissions scenarios . . . . . . . . . . . . . . . . . . . . . . . 85
3.4.7 NFS V4 ACLs: NFS V3 clients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.5 NFS V4 host identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.5.1 Basic host identification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.5.2 Kerberos host identification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.6 NFS V4 host authentication. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.7 NFS V4 host authorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Part 2. Implementing NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Chapter 4. Planning for NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

4.1 Deployment of NFS V4 in general. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2 Mandatory requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.2.1 What is your name resolution type? . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.2.2 Choosing your NFS domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.3 Identification methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.3.1 Selecting the user/group repository . . . . . . . . . . . . . . . . . . . . . . . . 100
4.3.2 Other identification considerations . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.4 NFS Authentication methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.4.1 AUTH_SYS method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.4.2 Deploying Kerberos. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.4.3 Default types of encryption for KDC and security flavors . . . . . . . . 107
4.4.4 NFS client considerations when using Kerberos. . . . . . . . . . . . . . . 108
4.4.5 Deployment of LDAP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.5 Authorization methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.5.1 Choosing your user authorization method . . . . . . . . . . . . . . . . . . . 110
4.5.2 Other user authorization considerations . . . . . . . . . . . . . . . . . . . . . 111
4.6 Choosing the appropriate file system types . . . . . . . . . . . . . . . . . . . . . . 111
4.7 NFS protocols and namespace considerations. . . . . . . . . . . . . . . . . . . . 112
4.7.1 Pseudo-root FS - alias tree versus classic model . . . . . . . . . . . . . . 115
4.8 Sizing and capacity planning considerations. . . . . . . . . . . . . . . . . . . . . . 115
4.9 Migration considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Chapter 5. Sample implementation scenarios . . . . . . . . . . . . . . . . . . . . . 119

5.1 Setup of the sample environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.1.1 PATH variable for NAS deployment . . . . . . . . . . . . . . . . . . . . . . . . 121
5.1.2 syslogd settings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.2 Using NFS V4 as you did with NFS V3 . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.3 How to unmount an exported NFS V4 file system. . . . . . . . . . . . . . . . . . 123
5.4 Setting up the NFS domain name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.5 The pseudo-root FS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.5.1 Setting up the pseudo-root FS on an NFS V4 server . . . . . . . . . . . 125
5.5.2 Advantages of using the NFS V4 pseudo-root . . . . . . . . . . . . . . . . 128
5.5.3 Setting up the alias tree extension on an NFS V4 server . . . . . . . . 131
Contents v
5.6 Setting up the NAS with a legacy database . . . . . . . . . . . . . . . . . . . . . . 134
5.6.1 Setup of a KDC server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.6.2 Installing the IBM NAS file sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.6.3 Initial basic KDC functions test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
5.6.4 Create user principals on the KDC server. . . . . . . . . . . . . . . . . . . . 139
5.6.5 Create the NFS server principals on the KDC server . . . . . . . . . . . 141
5.7 Setting up an NFS V4 server with NAS on a different KDC server . . . . . 142
5.7.1 Create the NFS server keytab file entry . . . . . . . . . . . . . . . . . . . . . 142
5.7.2 Check the NFS V4 server before client access. . . . . . . . . . . . . . . . 143
5.7.3 Set up the NFS registry daemon. . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.7.4 Set up the gssd daemon on the NFS V4 server . . . . . . . . . . . . . . . 144
5.8 Setting up an NFS V4 client with NAS . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.8.1 General steps for all types of clients . . . . . . . . . . . . . . . . . . . . . . . . 145
5.8.2 Install the NAS client code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.8.3 Set up the NFS domain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.8.4 Set up the NFS domain-to-realm map . . . . . . . . . . . . . . . . . . . . . . 146
5.8.5 Full client installation steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.8.6 Slim client installation steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
5.8.7 Configuring RPCSEC_GSS on the clients . . . . . . . . . . . . . . . . . . . 154
5.9 Preparing the system for Tivoli Directory Server and Kerberos V5 . . . . . 155
5.9.1 Set up procedure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
5.9.2 Configure IBM Tivoli Directory Server . . . . . . . . . . . . . . . . . . . . . . . 160
5.9.3 Configure the KDC server with LDAP backend. . . . . . . . . . . . . . . . 165
5.9.4 Configure the NFS V4 client for integrated login services. . . . . . . . 170
5.10 Integrating NFS V4 with a Linux client . . . . . . . . . . . . . . . . . . . . . . . . . 176
5.10.1 NFS server and client setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
5.10.2 Read-only NFS V4 mount . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.10.3 Read/write NFS V4 mounts on Linux . . . . . . . . . . . . . . . . . . . . . . 185
5.10.4 Pseudo-file system in NFS V4 Linux client . . . . . . . . . . . . . . . . . . 187
5.11 Windows KDC and NFS V4 AIX 5.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
5.12 Setting up Kerberos cross-realm access. . . . . . . . . . . . . . . . . . . . . . . . 199
5.12.1 Add the krbtgt service principal to every KDC server . . . . . . . . . . 200
5.12.2 Kerberos configuration file changes on the KDC server, NFS V4 client
and server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
5.12.3 Add NFS domain-to-realm map on NFS V4 client and server . . . 204
5.12.4 Client access verification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
5.12.5 Client access mount using cross-realms. . . . . . . . . . . . . . . . . . . . 205
Chapter 6. Problem determination. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

6.1 Problem determination tools and techniques . . . . . . . . . . . . . . . . . . . . . 208
6.2 AIX problem determination tools and aids for NFS . . . . . . . . . . . . . . . . . 208
6.2.1 Enabling syslogd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
6.2.2 Using iptrace and ipreport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
vi Securing NFS in AIX

6.2.3 Using the fuser command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.2.4 Using the rpcinfo command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.2.5 Using the showmount command. . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.2.6 Using the nfs4cl command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.2.7 Using the nfsstat command. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.2.8 Using the errpt command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.3 IBM NAS problem determination tools . . . . . . . . . . . . . . . . . . . . . . . . . . 211
6.4 Tivoli Directory Server problem determination tools . . . . . . . . . . . . . . . . 212
6.5 Third-party problem determination tools . . . . . . . . . . . . . . . . . . . . . . . . . 212
6.5.1 Using the lsof command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
6.5.2 Using the Ethereal utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
6.6 General NFS V4 problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
6.6.1 Warning: EIM is not configured . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
6.6.2 Realm is already mapped to domain. . . . . . . . . . . . . . . . . . . . . . . . 214
6.7 Exporting file systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
6.7.1 Exportfs: cannot change the v4 root... . . . . . . . . . . . . . . . . . . . . . . 215
6.7.2 Exportfs: /<path>: Invalid argument . . . . . . . . . . . . . . . . . . . . . . . . 215
6.7.3 Exportfs: /var/<logfile>: Too many levels of symbolic links... . . . . . 217
6.8 Mount problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
6.8.1 General mount problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
6.8.2 Pseudo-root and nfs4cl problems . . . . . . . . . . . . . . . . . . . . . . . . . . 219
6.8.3 ‘vers’ mount option error: “...Program not registered” . . . . . . . . . . . 219
6.8.4 ‘vers’ mount option error: “...server <name> not responding” . . . . . 220
6.8.5 Mount command hangs - no system response . . . . . . . . . . . . . . . . 220
6.8.6 Mount with sec=krb5: “vmount: The file access permissions do not allow
the specified action” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
6.8.7 Mount with sec=krb5: “RPC: 1832-016 Unknown host...” . . . . . . . . 223
6.8.8 File and directory access: cd, ls, etc. return “permission denied” . . 226
6.8.9 File and directory access: file ownership is “nobody:nobody” . . . . . 229
6.8.10 NAS problem: kadmin: “Unable to initialize kadmin interface” . . . 231
6.9 GSS-API error codes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
6.9.1 Major GSS-API error codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
6.9.2 Kerberos v5 status codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
Part 3. Appendixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Appendix A. Kerberos. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Kerberos keys and initial setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Authenticating to the Kerberos server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Authenticating to an application server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Kerberos terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Where to find more information about Kerberos . . . . . . . . . . . . . . . . . . . . . . 252
Contents vii
IBM Redbooks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
Other IBM publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
Non-IBM publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Other information sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Appendix B. Sample scripts, files, and output . . . . . . . . . . . . . . . . . . . . . 255

Sample administrative scripts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
Change the pseudo-root FS sample script . . . . . . . . . . . . . . . . . . . . . . . . 256
Create a KDC server with NFS V4 server . . . . . . . . . . . . . . . . . . . . . . . . . 256
Create a full client with legacy KDC server backend . . . . . . . . . . . . . . . . 258
Create a Full Client with KDC and LDAP backend . . . . . . . . . . . . . . . . . . 259
Script to copy ACLs to an entire directory structure . . . . . . . . . . . . . . . . . 260
Windows command script to run ktpass . . . . . . . . . . . . . . . . . . . . . . . . . . 262
Script to gather additional information for local AIX software support. . . . 263
Sample client Kerberos configuration files . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
Kerberos configuration file /etc/krb5/krb5.conf with legacy backend . . . . 268
Kerberos configuration file /etc/krb5/krb5.conf with LDAP backend . . . . . 269
Kerberos configuration file /etc/krb5/krb5.conf with Windows Active Directory
backend . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
LDIF sample file for KDC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
Sample iptrace output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
Successful authentication during mount request . . . . . . . . . . . . . . . . . . . 271
Unsuccessful authentication during mount request . . . . . . . . . . . . . . . . . 281
Appendix C. AIX 5.3 NFS quick reference. . . . . . . . . . . . . . . . . . . . . . . . . 287

NFS configuration files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
NFS daemons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
NFS commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
Export options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
mount command options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
nfso command options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Abbreviations and acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301

IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Other publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Online resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
How to get IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Help from IBM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
viii Securing NFS in AIX

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 illustrates 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. 2004. All rights reserved. ix

Trademarks
The following terms are trademarks of the International Business Machines Corporation in the United States,
other countries, or both:
AFS® IBM® pSeries®

AIX® iSeries™ Redbooks™
AIX 5L™ Lotus® Redbooks (logo) ™
Domino® MQSeries® RS/6000®
DB2® Notes® Tivoli®
DFS™ OS/2® WebSphere®
DYNIX/ptx® POWERparallel®
The following terms are trademarks of other companies:
Microsoft, Windows, and the Windows logo are trademarks of Microsoft Corporation in the United States,
other countries, or both.
UNIX is a registered trademark of The Open Group in the United States and other countries.
Linux is a trademark of Linus Torvalds in the United States, other countries, or both.
Other company, product, and service names may be trademarks or service marks of others.
x Securing NFS in AIX

Preface
Network File System Version 4 (NFS V4) is the latest defined client-to-server
protocol for NFS. A significant upgrade of NFS V3, it was defined under the IETF
framework by many contributors. NFS V4 introduces a major changes to the way
NFS has been implemented and used up until now, including stronger security,
wide area network sharing, and broader platform adaptability.
This IBM® Redbook is intended to provide a broad understanding of NFS V4 and

specific AIX® NFS V4 implementation details. It discusses considerations for
deployment of NFS V4, with a focus on exploiting the stronger security features
of the new protocol.
In the initial implementation of NFS V4 in AIX 5.3, the most important functional
differences are related to security. Chapter 3 and parts of the planning and
implementation chapters in Part 2 cover this topic in detail.
Part 1 NFS V4 Fundamentals

Chapter 1. NFS Version 4 overview
This chapter provides a high-level discussion defining NFS, the evolution
of the protocol, design motivations for NFS V4, features implemented in
AIX 5.3, and an overview of planning and implementation design
decisions.
Read this chapter for an introduction to Version 4 protocol design
requirements. It includes an important section that introduces the planning
and implementation design decisions necessary for successful
deployment.
Chapter 2. What’s new in NFS V4?
This chapter provides a more detailed analysis of the differences between
NFS V4 and prior versions. Name space handling and psuedo file systems
are discussed in detail. It also provides reference information
demonstrating differences in supporting files and daemons.
Read the first part of the chapter for more detailed information on the NFS
protocol design and evolution from Version 2 to Version 4. In later parts of
the chapter, learn specific technical differences in the implementation of
NFS V4.
Chapter 3. Enhanced security in NFS V4
This chapter provides a deep dive into the enhanced security features
available in NFS V4.
© Copyright IBM Corp. 2004. All rights reserved. xi

Read this chapter for details on integration options to support
identification, authentication, and authorization of users and hosts in an
NFS V4 environment. Very important differences between standard
UNIX® permissions and the new NFS V4 access control lists (ACLs) are
described and demonstrated with sample scenarios. Discussion of ACL
management and ACL inheritance strategies is also included.
Part 2 Implementing NFS V4

Chapter 4. Planning for NFS V4
This chapter analyzes the deployment planning challenges for NFS V4
and provides some decision support tools for the system architect
responsible for this task.
Read this chapter to learn methods for planning a deployment. There are
useful tools (decision tree flow charts, sample before and after migration
scenarios) that should help the system designer to understand how to
map the NFS V4 migration challenges to their own environment.
Chapter 5. Sample implementation scenarios
This chapter documents actual sample installation, implementation, and
deployment tasks carried out to support the various NFS V4 configurations
tested as part of the development process for this book.
Refer to this chapter for how-to examples showing NFS V4 setup. See the
beginning of the chapter for a more detailed list of documented
configurations.
Chapter 6. Problem determination
This chapter begins by describing some best-practice setup and
configuration methods for capturing information in case you need to debug
error conditions when deploying NFS V4. Also included are some actual
error conditions and the information-gathering methods that led to
solutions for those conditions.
Read the beginning of this chapter for useful tips on configuring NSFv4
test systems. The latter part of the chapter could help you find solutions to
error conditions you may encounter.
Part 3 Appendices
Appendix A. Kerberos
This appendix provides detailed backgound information on the inner
workings of the Kerberos authentication system.
xii Securing NFS in AIX

Appendix B. Sample scripts, files, and output
This appendix contains sample administration scripts for automating NFS
V4 setup and configuration, sample configuration files, and sample
IPTrace output streams.
Appendix C. AIX 5.3 NFS quick reference
This appendix provides a quick reference to NFS. It includes
comprehensive lists of NFS-related configuration files, system daemons,
commands, exporting options, mount options, and nfso options.
The team that wrote this redbook

This redbook was produced by a team of specialists from around the world
working at the International Technical Support Organization, Austin Center.
Chris Almond is an ITSO Project Leader and IT Architect based in Austin,

Texas. He has a total of 14 years of IT industry experience, including the past
five with IBM. His experience includes UNIX/Linux® systems engineering,
network engineering, Lotus® Domino®–based content management solutions,
and WebSphere® Portal–based solution design.
Lutz Denefleh is a team leader at the PLM Technical Support Eastern Region
Support Center in Mainz, Germany. He holds a Graduate Engineer degree in
Fluid Dynamics. He has 16 years of experience in the Information Technology
field and has worked at IBM for 15 years. His areas of expertise include solution
implementations on RS/6000®, such as CATIA, Lotus Notes®/Domino, Tivoli®,
DCE/DFS™, and Internet technologies. He is now responsible for the IBM
internal infrastructure used by the PLM World Wide Technical Support.
Sridhar Murthy works for IBM as an IT Specialist in Dallas, Texas. He has 15

years of experience in the IT industry. He holds a Master’s degree in Computer
Science and a Bachelors’ degree in Electrical Engineering. He is also an AIX
Certified Advanced Technical Expert. His areas of expertise include solution
implementations using Distributed Computing Environment (DCE), Distributed
File System (DFS), Domino Servers, Tivoli Netview, and Network Authentication
Service on AIX. He is currently evaluating methods for designing an architecture
that will support migration of a 40,000+ user DEC/DFS environment to NFS V4.
Aniket Patel is a Team Leader at the IBM UK UNIX Support Centre. He has
eight years of experience in UNIX and has worked at IBM for six years. Aniket
graduated from Kingston University, UK, with a Bachelor of Science (Honours
Degree) in Computer Science. His areas of expertise include UNIX Support on
AIX, DYNIX/ptx®, and Linux, NFS, TCP/IP, SNA, X.25, DCE, Sendmail,
MQSeries®, and the Microsoft® Windows® operating systems. Prior to recently
Preface xiii
taking on the role of Team Leader, Aniket was responsible for a team of Network
Specialists within the IBM UK UNIX Support Centre.
John Trindle works for Boeing in Seattle, Washington, as a systems design and
integration specialist. He has worked with NFS on various UNIX systems
(including SunOS/Solaris, DEC Ultrix, HP-UX, and most recently AIX) for 18
years. He holds a Bachelor of Science degree in General Engineering from the
United Stated Military Academy, West Point, NY. His current areas of expertise
include the AIX operating system and storage integration with AIX. He also
designs and oversees several hierarchical storage management implementations
using IBM Tivoli Storage Manager for Space Management.
Acknowledgements
The following people provided significant support for this project:
Carl Burnett, IBM AIX Kernel Architecture Team: for his thoughtful insights and
overall guidance in helping us develop a content strategy for this book.
Ufuk Celikka, IBM AIX Security Development Team: for technical support during
NFS V4 testing.
Bill McAllister, Team Leader IBM UK UNIX Support Centre: for his support in
reviewing draft content and providing feedback during the project.
Brian L. McCorkle, IBM AIX NFS Development Team: for his continued patience
and assistance while we tried to understand the implementation of NFS V4 on
AIX. Brian’s support enabled the team to meet its content goals for the book.
Dave Sheffield, Team Leader IBM AIX NFS Development Team: for facilitating
access to Development Team resources whenever we needed it.
Steve Sipocz, IBM BCS Systems Integration Services: for technical support and
review of our integration scenarios.
Betsy Thaggard
ITSO Editor, Austin Center
And other members of the AIX NFS Development Team and the AIX NAS
Development Team, for fheir support on various technical issues.
xiv Securing NFS in AIX

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
Comments welcome
Your comments are important to us because 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 e-mail to:
[email protected]
򐂰 Mail your comments to:
IBM Corporation, International Technical Support Organization
Dept. JN9B Building 905 3D004
11501 Burnet Road
Austin, Texas 78758-3493
Preface xv
xvi Securing NFS in AIX
Part 1
Part 1 NFS V4
fundamentals
© Copyright IBM Corp. 2004. All rights reserved. 1

2 Securing NFS in AIX
1
Chapter 1. NFS Version 4 overview

The purpose of this chapter is to give a general overview of the Network File
System (NFS) protocol, from NFS V1 to NFS V4.
We give you an introduction to the NFS protocol, beginning with a look at how
NFS works. We then move on to a brief history of NFS followed by NFS V4. We
also introduce you to some of the AIX 5.3 implementation specifics of NFS V4.
The following topics are discussed in this chapter:

򐂰 What is NFS?
򐂰 NFS V2 and NFS V3 history
򐂰 NFS V4 design motivations
򐂰 Objectives of NFS V4 (RFC3530)
򐂰 AIX 5.3 specific implementation of NFS V4
򐂰 Planning and implementation considerations

1.1 What is NFS?
In many environments, we would like to share files and programs among
workstations in a local area network (LAN). Doing so requires programs that let
us share these files, create new ones, carry out file locking, and manage
ownership correctly.
Over the past few years there have been developments in multiple types of
network capable file systems. These include the Apollo Domain, the Andrew File
System (AFS®), the AT&T Remote File System (RFS), IBM DFS, and Sun
Microsystems’ Network File System (NFS). Each of these has had beneficial
features and limiting drawbacks.
Of all the network file systems, NFS is probably the most widely used. NFS is
available on almost all versions of UNIX, as well as on Apple Macintosh systems,
MS-DOS, Windows, OS/2®, and VMS. It has continued to mature, and the latest
revision, Version 4, will help to advance and expand its reach.
So, what is NFS? NFS is a distributed file system that enables users to access
files and directories on remote servers as if they were local. For example, the
user can use operating system commands to create, remove, read, write, and set
file attributes for remote files and directories. NFS is independent of machine
types, operating systems, and network architectures through the use of Remote
Procedure Calls (RPCs).
NFS operates on a client/server basis. An NFS server has files on a local disk,
which are accessed through NFS on a client machine. To handle these
operations, NFS consists of:
򐂰 Networking protocols
򐂰 Client and server daemons
򐂰 Kernel extensions
The kernel extensions are outside the scope of this book, but the protocols,
daemons, planning, and implementation of NFS V4 will be discussed.
1.2 NFS V2 and NFS V3 History

NFS was first introduced by Sun Microsystems in the early 1980s.
NFS V1 was Sun’s prototype version and was never released for public use.
NFS V2 was released in 1985 with the SunOS V2 operating system. Many UNIX
vendors licensed this version of NFS from Sun. NFS V2 suffered many
undocumented and subtle changes throughout its 10-year life. Some vendors

allowed NFS V2 to read or write more than 4 K bytes at a time; others increased
the number of groups provided as part of the RPC authentication from 8 to 16.
These minor changes created occasional incompatibilities between different NFS
implementations; however, the protocol continued to provide an exceptional
degree of compatibility between systems made by different vendors.
The NFS V3 specification was developed during July 1992. Working code for
NFS V3 was introduced by some vendors in 1995 and was made widely available
in 1996. Version 3 incorporated many performance improvements over Version 2
but did not significantly change the way that NFS worked or the security model
used by the network file system.
1.3 NFS V4 design motivations

Increasingly, businesses have needed to secure and protect data. Earlier
versions of NFS had weaknesses that kept them from meeting these needs. The
following list is an example of areas that NFS V2 and NFS V3 have failed to
address:
򐂰 Strong authentication to prevent malicious users from masquerading as a
valid user of the system
򐂰 Fine-grained access control to make sure only the right people have access
to sensitive data
򐂰 Encrypting data traffic to protect it from unauthorized disclosure as it travels
over the network
򐂰 Uniquely identifying users in a large organization
򐂰 Good system and file I/O performance, including access from remote
locations
򐂰 Being able to access shared data from many different platforms
򐂰 The use of an open systems design
1.4 Objectives of NFS V4 (RFC3530)

NFS V4 was originally defined under RFC3010 in December 2000. In April 2003
RFC3010 was superseded by RFC3530 to address some of the issues related to
ease of implementation or clarification. The NFS V4 implementation on AIX 5.3 is
based on RFC3530.
Chapter 1. NFS Version 4 overview 5

The main objectives of the Version 4 protocol focus on addressing the
shortcomings of previous versions as listed in 1.3, “NFS V4 design motivations”
on page 5, while achieving the following overall goals:
򐂰 Improved performance over the network
򐂰 Enhanced by adding security schemes built into the protocol
򐂰 Integrated locking support
򐂰 Cross-platform interoperability, including the Microsoft Windows environment
򐂰 Protocol extension support in a manner that does not invalidate backward
compatibility
򐂰 Movement toward an Open Standard, managed by the IETF, whereas
previous versions of NFS were proprietary
For more detailed information about the new features of the protocol, there are
many references available. Refer to “Related publications” on page 301. One
particularly useful reference for a more detailed discussion of Version 4 protocol
functional design goals is a white paper titled The NFS Version 4 Protocol by
Brian Pawlowski, et al., which can be found at:
http://www.nluug.nl/events/sane2000/papers/pawlowski.pdf
1.5 AIX 5.3 specific implementation of NFS V4

AIX 5.3 supports the latest NFS protocol, Version 4 (NFS V4). AIX also provides
support for NFS Version 3 (NFS V3) and NFS Version 2 (NFS V2) client/server. It
therefore provides backward compatibility with the existing install base of NFS
clients and servers in a heterogeneous environment.
Vendor-specific implementation of RFC3530 into their product depends on

several factors:
򐂰 Mandatory versus optional features described in the RFC
򐂰 Customer-specific requirements
򐂰 Derived business needs
The following features of the protocol are supported in AIX 5.3.
1.5.1 Mandatory features

The RFC has defined that the mandatory features are the minimal set of file or
file system attributes that must be provided by the server and must be properly

represented by the server. All implementations of NFS V4, including AIX 5.3,
must support these features to be compliant:
򐂰 GSS-API (RFC2078) with Kerberos V5 mechanism
򐂰 Identity mapping
򐂰 Pseudo-file system model
1.5.2 Optional features

The RFC specification also defines optional features. The term optional is used
here, meaning that the RFC gives the vendor the option to implement the feature
or not. AIX 5.3 implements the following optional features:
򐂰 NFS V4 Access Control Lists (ACL)
򐂰 Delegation
򐂰 Mapping of principals and file ownership attributes from one NFS V4 domain
into another
򐂰 Data replication and migration
1.6 Planning and implementation considerations

Planning and implementing NFS V4 is more complex than for previous NFS
versions. Because NFS V4 introduces new advanced capabilities, advanced
planning will make deployment less painful and ongoing management easier.
Therefore, it is essential that you choose the elements of your infrastructure
carefully before beginning your implementation. To help with the decisions you
need to make, we have dedicated a chapter to planning and one for
implementation considerations. We also discuss in detail what we see as
common migration paths. These are discussed in Chapter 5, “Sample
implementation scenarios” on page 119.
1.6.1 Pre-implementation design considerations

1. How will you manage user identities? The options to choose from include:
a. LDAP [RFC2307]
b. NIS
c. Standard UNIX password and group files
2. Will you need to accommodate user identities from other organizations?
3. How will you authenticate users? Options here include:
a. Kerberos

b. Standard UNIX password authentication
4. How will you control access to directories and files? Options here include:
a. Standard UNIX permissions
b. NFS V4 access control lists (ACLs)
• If you choose to use ACLs, how would you structure your directories to
maximize use of the inheritance features?
5. How will you structure your shared file system name space? Think about how
you want to present your NFS exported file systems to the clients. Additional
considerations include:
• Do you want to make the server location transparent to the users? For
example, will you need to move subdirectories between servers without
changing where they appear in the client directory tree?
• Do you have a large, flat directory structure (like home directories)
where using distributed structures would help?
• Do you have special NFS export restrictions that dictate directory
structure?
• Do your file and directory access control methods work better with a
particular directory structure? For example, do you want to structure
access controls grouped by organization or by role?
6. Do you require high availability? Services that might need to be designed for
high availability include:
a. Directory/Identity services
b. Authentication services
c. File services (the actual NFS servers themselves)
7. What is your network architecture? Will you need to place servers near users
in separate geographical locations?
8. How dispersed is your data access?
1.7 Looking ahead to the rest of the book

The questions highlighted in the pre-implementation considerations (outlined in
the previous section) are difficult to answer until a significant amount of planning,
research, and preparation is completed. It is the goal of this book to help prepare
the reader to make these design decisions.

The next two chapters provide technical background on the NFS V4 protocol as
compared to Version 3 (Chapter 2), then a detailed discussion of the new file
system security control features that Version 4 provides (Chapter 3).
These chapters provide technical background to prepare the reader for planning
an NFS V4 deployment. Before you can proceed with any implementation, you
have to plan it, so Chapter 4, “Planning for NFS V4” on page 93 is the key
chapter in this book. The chapter walks you through a design decision process
for planning your NFS V4 infrastructure.
After your planning is complete, we look at some scenarios and how to

implement them. The scenarios demonstrate how you could get to the final step
of having a running NFS V4 infrastructure. We consider migration paths from a
previous version of AIX and therefore an older version of NFS to AIX 5.3 and
NFS V4. We also consider migration paths from an older version of AIX running
NFS V3 to AIX 5.3 running NFS V3. In total, we look at three different migration
paths.

2
Chapter 2. What’s new in NFS V4?

This chapter reveals the differences between NFS V3 and NFS V4, but first we
look at NFS in general and how it has evolved over the years. The aim is to give
those who are new to NFS a basic grounding in the protocol. Those who already
know NFS can skip directly to 2.6, “NFS V4” on page 20.
Continuing from the NFS design discussion, we talk about the differences
between the AIX v5.3 NFS V4 implementation and what the NFS V4 RFC
RFC3530 describes. The reader is advised to refer to the NFS V4 RFC for a
detailed explanation of the protocol. Detailed scenarios and implementation
options will be discussed in the planning and implementation chapters.
The following topics are covered:

򐂰 How NFS works
򐂰 Protocols used by NFS
򐂰 NFS daemons
򐂰 NFS V3
򐂰 The NFS Lock Manager protocol
򐂰 NFS V4
򐂰 AIX5L v5.3 implementation of NFS V4
򐂰 NFS V4 support on AIX 5.3
򐂰 NFS files and commands: a quick reference

2.1 How NFS works
A server has files on a local disk, irrespective of file system type, that are
accessed through NFS on a client machine. NFS is an application layer protocol
that uses other underlying protocols defined in the TCP/IP model. NFS is built on
top of the TCP/IP protocol stack over the Remote Procedure Call (RPC) protocol.
To handle this operation, NFS consists of:

򐂰 Networking protocols
򐂰 Client and server daemons
Figure 2-1 How NFS works
2.2 Protocols used by NFS

The NFS-specific protocols are Sun RPC and External Data Representation
(XDR) protocol.
The Open Systems Interconnection (OSI) Reference Model is traditionally used

as a general purpose reference for describing and comparing protocols. We
assume that you are familiar with the OSI model. Figure 2-2 on page 13 attempts
to show the components of the NFS protocol in relation to the OSI Reference
Model. The mapping is only approximate and is intended as a conceptual guide.

Figure 2-2 Representing NFS with the seven-layer OSI model
2.2.1 UDP or TCP

Both NFS V2 and NFS V3 operate over the User Datagram Protocol (UDP) and
the Transmission Control Protocol (TCP). NFS V3 uses TCP by default. NFS V4
only uses TCP.
TCP is stateful, and NFS is stateless. A stateless server treats each request as
an independent transaction, unrelated to any previous request. This simplifies
the server design because it does not need to allocate storage to deal with
conversations in progress or worry about freeing it if a client dies in
mid-transaction. A disadvantage is that it may be necessary to include more
information in each request, and this extra information will have to be interpreted
by the server each time. UDP neither guarantees delivery nor does it require a
connection. As a result, it is lightweight and efficient, but all error processing and
retransmission must be taken care of by the application program.
A stateful server enables a host to send data in a continuous stream to another

host. The transport service guarantees that all data will be delivered to the other
end in the same order as sent and without duplication. Communication proceeds
through three well-defined phases: connection establishment, data transfer, and
connection release. TCP is a good example of this.
So, the statement that NFS is stateless would seem to be a contradiction and you
would think it impossible for NFS to work over TCP. However, NFS and TCP
Chapter 2. What’s new in NFS V4? 13

operate in a layer called Remote Procedure Call (RPC), and this hides the state
issues of TCP from NFS.
2.2.2 Remote Procedure Call (RPC)

RPC is a library of procedures. These procedures enable one process (the client
process) to direct another process (the server process) to execute procedure
calls as though the client process had executed the calls in its own address
space. Because the client and the server are two separate processes, they are
not required to be on the same physical system, although they can be. The RPC
call used is based on the file system action taken by the user. For example, when
issuing an ls -la command on an NFS mounted directory, the long listing will be
completed using RPCs NFSPROC3_READDIR or NFSPROC3_READDIRPLUS.
The server in turn will send the output from the command through RPC back to
the client. To the user, this transaction is totally transparent.
Note: The RPC described here is the Sun RPC and is not to be confused with
the RPC used by products such as DCE. RPC services use TCP or UDP to
transport data packets. NFS V4 only uses TCP as a transport protocol.
2.2.3 eXternal Data Representation (XDR)

Because the server and client processes can reside on two different physical
systems, which may have completely different architectures, RPC must address
the possibility that the two systems may not represent data in the same manner.
Therefore, RPC uses data types defined by the XDR protocol. XDR is the
specification for a standard representation of various data types. By using a
standard data type representation, data can be interpreted correctly, even if the
source of the data is a machine with a completely different architecture. XDR is
used when the vnode points out that the accessed file or directory is not local,
but resides on a remote system. A conversion of data into XDR format is needed
before sending the data. Conversely, when it receives data, it converts it from
XDR format into its own specific data type representation.
2.3 NFS daemons

A daemon is a process that runs continuously in the background and provides
services to clients. Some of the daemons that enable NFS to work are:
򐂰 portmap
򐂰 nfsd
򐂰 rpc.mountd
򐂰 rpc.statd

򐂰 rpc.lockd
򐂰 biod
The rpc.mountd and nfsd daemons run only on the server, and the rpc.lockd and
rpc.statd daemons run on the server and the client. The biod daemon runs only
on the client.
Figure 2-3 NFS client/server model with associated daemons
2.3.1 The portmap daemon

The portmap daemon converts RPC program numbers into Internet port
numbers. When an RPC server starts up, it registers with the portmap daemon.
The server tells the daemon which port number it is listening on and which RPC
program numbers it serves. By this process, the portmap daemon knows the
location of every registered port used by RPC servers on the host, and which
programs are available on each of these ports. When mounting, the mount
request starts with an RPC call named GETPORT that calls the portmap, which
in turn informs the client of the port number that the called RPC server listens to.
After this, the port number is used as reference for further communication. This is
why the NFS daemons must be registered with the portmap daemon.
A client will only consult the portmap daemon once for each program the client
tries to call. The portmap daemon tells the client which port to send the call to.
The client stores this information for future reference. As standard RPC servers
are normally started by the inetd daemon, the portmap daemon must be started
before the inetd daemon is invoked.

2.3.2 The rpc.mountd daemon
The rpc.mountd daemon handles the actual mount service, which is needed
when a client sends a mount request with an RPC procedure named
MOUNTPROC3_MNT to the server. In addition, it provides a list of currently
mounted file systems and the clients on which they are mounted.
2.3.3 The rpc.statd daemon

The rpc.statd daemon interacts with the rpc.lockd daemon to provide crash and
recovery functions for the locking services on NFS.
Important: The rpc.statd daemon should always be started before the lockd
daemon.
2.3.4 The rpc.lockd daemon

The rpc.lockd daemon processes lock requests that are either sent locally by the
kernel or remotely by another lock daemon. The rpc.lockd daemon forwards lock
requests for remote data to the server site lock daemon through the RPC
package. The rpc.lockd daemon then asks the rpc.statd (status monitor) daemon
for monitor service. The reply to the lock request is not sent to the kernel until
both the statd daemon and the server site lockd daemon reply.
2.3.5 The nfsd daemon

The nfsd daemon runs on a server and manages certain aspects of the server
configuration. The actual processing of client requests is handled by a
multi-threaded kernel process rather than the nfsd daemon itself. Each daemon
handles one request at a time. This means that on the server side, the receipt of
any one NFS protocol request from a client requires the dedicated attention of an
nfsd daemon until that request is satisfied, and the results of the request
processing are sent back to the client. The nfsd daemons are the active agents
providing NFS services.
The NFS daemons are inactive if there are no NFS requests to handle. When the
NFS server receives RPC calls on the nfsd daemon’s receive socket, the
daemon is awakened to pick up the packet off the socket and invoke the
requested operations.

2.3.6 The block I/O daemon (biod)
The block I/O daemon (biod) runs on all NFS client systems. When a user on a
client wants to read or write to a file on a server, the biod daemon sends this
request to the server. For each read or write request, one biod is requested. The
biod daemon is activated during system startup and runs continuously. The
number of biods are limited on a per-mount-point basis.
The actual work of reading and writing to a file on a server is handled by a

multi-threaded kernel process with one thread assigned for each read or write
request. The number of these threads changes dynamically depending on
demand, but the biod daemon controls the maximum number of these threads
per mount.
2.4 NFS V3
Even though NFS V2 (discussed briefly in the introduction) was widely accepted,
the protocol was not without its problems. This led to the introduction of NFS V3
RFC1813. The Version 2 protocol had the following major problems:
򐂰 A file size limit of 4 GB. Only files of up to 4 GB in size could be accessed. As
computer environments began to grow and the need for larger data
repositories became apparent, this limitation became a problem.
򐂰 Writes had to be synchronous. This led to write-intensive applications
suffering performance problems. There were quite a few workarounds for this
problem, but most violated the NFS V2 standard.
The main goals for the NFS V3 design were to solve these two problems. In
1994, Brian Pawlowski published a paper (referenced at end of this section) that
provided an overview of the process the designers of the Version 3 protocol went
through. This paper identified the following major areas where the protocol was
enhanced:
򐂰 The 4 GB file restriction. All arguments within the protocol (such as file sizes)
were changed from 32-bit to 64-bit.
򐂰 The write model was changed to introduce a write/commit phase. This
enabled asynchronous writes.
򐂰 A new ACCESS procedure was introduced. This resolved
permission-checking problems when mapping the ID of the superuser. This
procedure is also important for correct behavior when ACLs exist on files.
򐂰 The 8 Kb write per procedure call limit was relaxed.
򐂰 The READDIRPLUS procedure was introduced. This returned both a file
handle and attributes. This returns both directory entries and per-entry

attributes. READDIRPLUS is intended to reduce subsequent NFS LOOKUP
calls for each directory entry.
򐂰 The file handle size was changed to a variable size, up to a maximum of 64
bytes. NFS V2 had a fixed file handle size of 32 bytes.
򐂰 The CREATE procedure was modified. This allowed for exclusive file creates.
򐂰 Filenames and pathnames were limited to 255 and 1024 characters
respectively in NFS V2. This was changed in NFS V3 by variable length
strings agreed on between the client and server.
򐂰 NFS V3 tightened the errors that could be returned from the server. All error
values are iterated, and no errors outside the list are permitted.
򐂰 The blocksize field was removed. The blocks field was changed to used and
recorded the total number of bytes used by the file.
򐂰 The NFS3ERR_JUKEBOX error type was introduced. This allowed for
situations when a request is made to the server to read a file migrated to tape.
Obviously, the time to read the data back from the tape will be considerable.
This new error informs the client that the operation is in progress and the call
should be retried.
Forcing writes from asynchronous to synchronous mode will affect performance.

With NFS V3, the client can send a number of asynchronous WRITE requests
that it can then commit to disk on the server at a later date by issuing a COMMIT
request. After the server has received the COMMIT request, it cannot return until
all data has been flushed to disk. The COMMIT request is similar to calling
fsync(). The major difference is that the COMMIT request does not necessarily
cover the data for the whole file. However, it does allow the client to flush all the
data when a file is closed or to break up a large synchronous write request into
smaller writes. These are performed asynchronously, but followed by a COMMIT
request. This is an important enhancement, as it enables the file system on the
server to coalesce a number of write requests into a single large write, making it
more efficient. As the client is required to keep a copy of all data to be written to
the file until a COMMIT is issued, asynchronous writes should not affect the
crash/recovery properties of NFS.
The READDIRPLUS procedure, even though it is efficient, also presents

problems. The procedure was introduced to minimize the number of
over-the-wire LOOKUP requests when a READDIR procedure had been invoked.
An example of this would be an ls -F request on a directory.
The implementation of READDIRPLUS is significantly more expensive than

READDIR. The procedure should only be performed when first accessing the
directory in order to populate the name cache, depending on the underlying
operating system. It should then only be performed again in cases where the
cache was invalidated for the directory due to a directory modification.

Many of the goals of NFS V3 were to improve performance. A number of different
performance-related tests showed that NFS V3 did indeed meet its objectives
very well, as documented in Pawlowski’s 1994 white paper NFS Version 3 Design
and Implementation, by Brian Pawlowski et al, found at:
http://citeseer.ist.psu.edu/pawlowski94nfs.html
2.5 The NFS Lock Manager protocol

File locking was omitted during the design of the NFS protocol. The main reason
is that to support record locking, state would have to be maintained on the server.
This would dramatically increase the complexity of NFS implementations.
File locking was not something that could be overlooked easily and was therefore
implemented in SunOS as the Network Lock Manager (NLM). NLM went
through various iterations, with Version 3 being most widely used with NFS V2.
With the introduction of NFS V3, the definition of NLM (Version 4) was included in
the NFS specifications, but was still left as a separate protocol. The NLM also
relied on the Network Status Monitor protocol. This was required so the clients
and servers could be notified of a crash so that a lock state could be recovered.
Crash/recovery involves coordination between both the clients and server:

Server crash When locks are handed to clients, the server maintains a
list of clients and the locks that they own. If the server
were then to crash, these locks would then be lost. When
the server reboots, a status monitor runs and sends a
message to all known clients. The lock manager on each
client is notified and given an opportunity to reclaim all
locks it owns for the files on the server. There is a fixed
amount of time in which the clients can respond.
Client crash If the client crashes, any locks that the client holds on the
server must be cleaned up. When the client recovers, it
sends a message to the server to clean up its locks.
Through the use of a client state number, which is
incremented on reboots, the server can detect that the
client has been rebooted and removes any locks that
were held by the client before it crashed and rebooted.
As stated earlier, the NLM was not widely adopted. The NFS protocol in
Version 4 has been extended to include a file locking mechanism. You can find
more information in 2.6.7, “File locking” on page 29.

2.6 NFS V4
NFS V4 introduces many substantial changes to the protocol. For example, the
design of NFS predates the widespread adoption of the World Wide Web. It was
originally designed for use in local area networks. With the advent of
internetworking, there has been a greater need to use distributed file systems,
such as NFS, in wide area networks. When NFS is used for distributed file
systems across wide area networks, the security capabilities in pre-Version 4
implementations become apparent.
Important: Addressing existing security limitations in NFS is the most

significant area of change introduced by Version 4. This redbook devotes an
entire chapter to this topic: Chapter 3, “Enhanced security in NFS V4” on
page 45.
NFS V4 goes a long way to overcome the shortcomings of NFS V2 and V3, and
adds additional features left out in the NFS V3 implementation. NFS V4 is
described in detail in RFC3530. The main changes that are discussed in this
section are:
򐂰 Attribute classes
򐂰 User name to UID mapping
򐂰 Better namespace handling (pseudo-file systems)
򐂰 Built-in security
򐂰 Client-side caching and delegation
򐂰 Compound procedures
򐂰 File locking
򐂰 Internationalization
򐂰 Volatile file handles
2.6.1 Attribute classes

The set of attributes that were passed over-the-wire with earlier versions of NFS
were very UNIX-oriented. This meant that the information returned by the server
was sufficient to respond to a stat() call on the client. This made it difficult for
non-UNIX systems to understand the protocol properly. To address this issue,
NFS V4 introduces a new set of file attributes in three different classes:
򐂰 Mandatory
򐂰 Recommended
򐂰 Named

Mandatory attributes
Mandatory attributes are the minimal set of file or file system attributes that must
be provided by the server and must be properly represented by the server. They
must also be supported by every implementation. The mandatory set of
attributes contain information such as the file type and size, information about the
file handle expiration times, whether hard links and symbolic links are supported,
and whether the file has named data streams and attributes. These are
summarized in Table 2-1.
Table 2-1 Mandatory attributes and their description

Mandatory attribute Description
Object type The type of object (file, directory, symbolic link).
File handle expiration Specifies file handle expiration behavior on the

client.
Change indicator The value created by the server that the client can
use to determine if file data, directory contents, or
attributes of a given object have been modified. The
server may return the object’s modify time for this
attribute’s value, but only if the file system object
cannot be updated more frequently than the
resolution of the modify time.
fsid A unique file system identifier for the file system

holding a given object. Contains major and minor
components, each of which are unsigned 64-bit
integers.
Lease duration The time frame in which the leases at the server side
end, in seconds.
Size The size of a given object, in bytes.
UNIX LINK Support Determines whether UNIX hard links are supported.
UNIX SYMLINK support Determines whether UNIX symbolic links are

supported.
Recommended attributes
The recommended attributes contain information such as the type of ACLs that
the file system supports, the ACLs themselves, information about the owner and
group, access timestamps, and quota attributes. They also contain information
about the file system such as free space, total number of files, files available for
use, and the file system limits such as maximum filename length and maximum
number of links. These are summarized in Table 2-2 on page 22.

Table 2-2 Recommended attributes and their descriptions
Recommended attributes Description
ACL The Access Control List (ACL) for a given object.
Archive bit Checks to see whether the file has been

archived since the last time it was modified.
Case insensitive Checks whether file names on a given file

system are case-insensitive.
Case preservation Checks whether the file name case in a given file
system is preserved.
Change owner restricted If this attribute is set to TRUE, the server will
reject any request to change the owner or group
associated with a given file if the caller is not a
privileged user (for example, the root user in
UNIX operating systems).
No file name truncation beyond Checks to ensure that either an error is returned
maximum or the name is truncated when the maximum file
name size supported for a given object is
exceeded,
File handle An opaque data structure provided by the server

to the client in response to a lookup request.
File ID This attribute is a number uniquely identifying a

file within a given file system.
Hidden Checks to see whether a file is considered

hidden with respect to the WIN32 API.
Maximum file size The maximum file size supported for a file
system for a given object.
Maximum number of links The maximum number of links for a given object.
Maximum file name size The maximum file name size supported for a
given object.
Maximum read size The maximum read size supported for a given
object.
Maximum write size The maximum write size supported for a given
object. It should be supported if the file is
writable. Lack of this attribute can lead to the
client either wasting bandwidth or receiving poor
performance.
MIME type The MIME body type for a given object.

Recommended attributes Description
UNIX mode bits The UNIX-style permission bits for a given

object.
Owner string The string name of the owner for a given object.
Group string The string name for the group ownership of a

given object.
Modify time The time of the last modification of a given

object.
Create time The creation time of a given object. This attribute

does NOT have any relation to the traditional
UNIX file attribute ctime or change time.
Access time The time of last access for a given object.
Space available to user The disk space, in bytes, available to a user on

the file system containing a given object.
File system free space The free disk space, in bytes, on the file system
containing a given object.
File system total space The total disk space, in bytes, on the file system
containing a given object.
Space used by object The number of file system bytes used by a given
object.
Named attributes
Named attributes provide a mechanism to associate additional properties with a
filesystem object (such as file or directory). A named attribute is an uninterpreted
opaque stream of bytes with a name. Applications can use named attributes to
place auxiliary application specific data on files. Multiple named attributes can
exist on an object. The OPENATTR procedure is used to access named
attributes for an object. NFS V4 organizes named attributes as a directory of
attribute names. The READDIR and LOOKUP operations are used to obtain the
attribute names. The READ, WRITE, and CREATE operations are then used to
operate on the individual named attributes.
A client would access named attributes in the following way:

򐂰 The OPENATTR procedure sets the current filehandle to the named file
attribute directory for the file object.
򐂰 The READDIR and READDIRPLUS procedures retrieve the file handles for
the various named attributes associated with the original file system object.

Note: Named attributes are an optional protocol feature. Both the server and
client must support named attributes for them to be used.
2.6.2 Username to UID mapping

With the Version 3 protocol and the commonly used AUTH_SYS RPC flavor, NFS
requests contain a set of user credentials with a user ID (UID) and a list of group
IDs (GIDs) to which the UID belongs. On the NFS server, these credentials are
used to perform the permission checks that are part of UNIX file system access
control (for example, verifying write permission to remove a file, or execute
permission to search directories.) Additionally, filesystem object user and group
ownership information is transferred as numeric integer values. Using integers to
represent users and groups requires every client and server that would connect
to each other to agree on user and group ID assignments. This UID to name
mapping occurs only on the client.
NFS V4 departs from this model of identity representation. When transferring

ownership information, it represents users and groups as strings in the following
form:
user@nfs_domain
or
group@nfs_domain
Additionally, it is expected that the stronger RPCSEC-GSS RPC security flavor

will be largely deployed with NFS V4. With RPCSEC-GSS, an opaque security
token flows on the wire. From it, the receiver (NFS server) derives a string
representation (principal) of the accessing identity. The server transforms the
principal string into its native credentials. On UNIX systems, this results in a UID
and corresponding list of member GIDs.
String-based identities require NFS clients and servers to translate between the
protocol string attributes and the internal formats (UID and GID) that are used
within the operating system.
Representing users and groups as strings provides added flexibility and removes
the requirement that all systems utilize the same numeric ID space. It is expected
that all systems share a common view of the user and group name space within
a given NFS domain. The use of strings also opens the potential capability for
interdomain NFS sharing when the capability exists to map an identity from a
foreign domain into the receiver’s local domain.

Important: For UID translation, NFS V4 requires that user and group names
are consistent between the client and server to avoid potential access errors
to secured data. However, the numeric UID space between systems does not
have to match. We look at how this works in practice in 3.2.2, “User/group
identities and NFS V4” on page 50.
2.6.3 Better namespace handling

NFS V2 and NFS V3 servers export a set of independent parts of their overall
namespace and do not allow NFS clients to cross mountpoints on the server.
This is because NFS expects all lookups to stay within a single file system. In
NFS V4, the server provides a single root file handle through which clients can
obtain file handles for any of the accessible exports.
NFS V4 no longer has a separate mount protocol. Instead of exporting a number

of distinct exports, an NFS V4 client sees the NFS V4 server’s exports as existing
inside a single file tree, called the nfsv4 pseudo-file system. The pseudo-file
system tree constructed by the server creates a single logical view of all the
different exported file systems (Figure 2-4).
Figure 2-4 Pseudo-file systems: server view

In Figure 2-4 on page 25, we would like to export the following directories:
򐂰 /exports/home/sally
򐂰 /exports/home/bob
򐂰 /exports/home/mary
򐂰 /exports/project/proja
򐂰 /exports/project/projb
On the server, you must:

1. Set the pseudo-root node; in our case, we set this to /exports
2. Add the directories to export to the /etc/exports file:
/exports/home/sally -vers=4,ro
/exports/home/bob -vers=4,ro
/exports/home/mary -vers=4,ro
/exports/project -vers=4,ro
/exports/project/projA -vers=4,ro
/exports/project/projB -vers=4,ro
3. Run:
exportfs -va
4. On the client, mount the root export:
mount -o vers=4 <nfsv4_svr_name>:/ /<local_mount_point>
The contents of the newly mounted file system appear in Example 2-1 and
Figure 2-5 on page 27.
Example 2-1 Client view of the pseudo-file system

# ls -al /nfs/*
/nfs/home:
total 26
drwxr-xr-x 4 root system 5 Jul 28 11:26 .
drwxr-xr-x 3 root system 4 Jul 28 11:26 ..
drwxr-xr-x 2 root system 512 Jul 28 11:24 bob
drwxr-xr-x 2 root system 512 Jul 28 11:25 mary
drwxr-xr-x 2 root system 512 Jul 28 11:24 sally
/nfs/project:
total 18
drwxr-xr-x 3 root system 4 Jul 28 11:26 .
drwxr-xr-x 3 root system 4 Jul 28 11:26 ..
drwxr-xr-x 2 root system 512 Jul 28 10:40 proja
drwxr-xr-x 2 root system 512 Jul 28 10:40 projb

Figure 2-5 Pseudo-file systems: client view
The server has provided the client with a single view of the exported file systems.
In NFS V4, a server’s named space is a single hierarchy. In the example above,
the export list hierarchy is not connected. When a server chooses to export a
disconnected portion of its name space, the server creates a pseudo-file system
to bridge the unexported portions of the name space, allowing a client to reach
the export points from the single common root. A pseudo-file system is a
structure containing only directories, created by the server having a unique fsid,
that enables a client to browse the hierarchy of the exported file systems.
The client view of the pseudo-file system is limited to paths that lead to the
exported file systems. Because /home/joe and /dept are not exported in the
example, they do not appear on the client during browsing operations
(Figure 2-5).
2.6.4 Built-in security

NFS has always relied on client-side authentication to provide security. This has
generally not been a problem because NFS has largely been used within private
networks. One of the objectives of the Version 4 protocol is to enable increased
use of NFS to wide area networks.
The basic NFS security mechanisms are extended in NFS V4 through the
mandated support of the RPCSEC_GSS RPC security flavor. RPCSEC_GSS is
implemented at the RPC layer. It is capable of supporting different security
mechanisms. Examples include Kerberos Version 5, and public-key-based
mechanisms such as SPKM. NFS V4 requires that RPCSEC-GSS be provided

as an available RPC security flavor. It mandates that the Kerberos Version 5,
SPKM, and LIPKEY security mechanisms be supported for full protocol
compliance. It still allows the support and use of other RPC security flavors such
as AUTH_SYS. A key weakness of the AUTH_SYS flavor has always been the
ease with which hackers can forge and impersonate credentials.
RPCSEC_GSS is different from AUTH_SYS in two ways:

򐂰 RPCSEC_GSS goes beyond authentication to perform integrity checksums
and encryption of the entire body of the RPC request and response.
򐂰 As RCPSEC_GSS encapsulates the GSS-API messaging tokens, it acts as a
transport for mechanism-specific tokens for security flavors such as Kerberos.
Adding new security mechanisms does not require rewriting significant
portions of NFS or any other ONCRPC-based application.
All versions of NFS are capable of using RPCSEC_GSS. The difference is that
while an implementation can claim to conform to NFS V2 and NFS V3 without
implementing support for RPCSEC_GSS, a conforming implementation (one that
claims to be based on RFC3530) of NFS V4 must implement security based on
Kerberos Version 5 (as done in AIX 5.3) and LIPKEY. Kerberos V5 (KRB5) and
LIPKEY are GSS-API conforming mechanisms.
Kerberos divides user communities into realms. Each realm has an administrator
responsible for maintaining a database of principals or users, one master Key
Distribution Center (KDC), and one or more slave KDCs that give user tickets to
access services on specific hosts in a realm. Users in one realm can access
services in another realm via trust relationships between the KDCs. Kerberos is a
very good choice for enterprise work groups operating within an intranet. It
provides centralized control, as well as single sign-on to the network. We will
discuss Kerberos further in upcoming chapters.
The Low Infrastructure Public Key (LIPKEY) system provides an SSL-like

model and is equivalent security for use on the Internet. LIPKEY is a GSS-API
security mechanism that uses a symmetric key cipher and server-side public key
certificates. LIPKEY will not be discussed in any more detail as the NFS V4
implementation in the initial release of AIX 5L™ v5.3 does not support it.
2.6.5 Client-side caching and delegation

Most NFS client implementations do caching of both data and attributes to
improve performance and reduce network traffic. Some vendors also support a
technology known as cachefs that allows extended caching of file and directory
content data to persistent storage (disk) on the client system. This further
increases the amount of data a client can cache.

With caching, some amount of server interaction is still required to maintain the
required semantics of the NFS protocol. Clients must check with servers and
open time to validate and flush cached information as appropriate. In addition,
the client periodically polls the server while files are in use. Depending on the
application environment, the network traffic associated with client cache
maintenance can be modest. In less reliable or slower networks, this traffic can
represent a performance restriction.
NFS V4 provides an optional protocol mechanism called delegation that can

improve the caching of NFS. With delegations, the open time network traffic can
be avoided as well as the periodic checks to servers. The reduction in network
traffic can help increase the performance and scale of an NFS environment.
NFS V4, like its predecessors, has a weak cache consistency model. Clients are
not guaranteed to see the most recent changes to data at a server. Delegations
are optional and are granted at the NFS server’s discretion. Without a delegation,
the NFS V4 client operates similar to previous versions of NFS.
2.6.6 Compound RPC procedures

Many file-related operations over NFS require a large number of RPC calls. In a
local area network this is not an issue; however, when operating in a wide area
network, the effect on performance can be much more noticeable. NFS V4
introduces a compound RPC model where it is possible to combine multiple
protocol operations into a single RPC request. This creates potential to more
efficiently utilize network resources by reducing the total number of RPC
transactions for a given workload.
2.6.7 File locking

We have already talked about the fact that file locking is not an integrated part of
NFS V2 and NFS V3. NFS V4 provides both UNIX-level file locking functions and
Windows-based share locking functions. This gives NFS V4 better
interoperability in heterogeneous environments. File locking in its simplest form is
the ability to block I/O operations by other applications on a file that contains a
record lock. Methods for managing state ID facilitates this operation.
A stateid is a unique 64-bit object that defines the locking state of a specific file.
When a client requests a lock, it presents a clientid and unique-per-client lock
owner identification to identify the lock owner.
When a client first contacts a server, it presents an opaque structure identifying

itself to the server. The opaque structure uniquely identifies a particular client.
After the server receives the client’s identifying data, it returns a 64-bit clientid.
The clientid is unique and will not conflict with those previously generated.

2.6.8 Internationalization
Previous versions of the NFS protocol handled file names as an opaque byte
stream. They were limited to a 7-bit US ASCII representation, but were
commonly encoded in 8-bit ISO-Latin-1. There was no way to specify the type of
encoding in the XDR. This limited the use of NFS in environments where there
may be mixed character sets. In order to provide better support for
internationalization, filesystem object names (such as files and directories) are
encoded as UTF-8 in NFS V4.
2.6.9 Volatile file handles

In previous versions of NFS, a file handle is a per-server unique identifier for a
file system object that is opaque to the client. It is defined as having a value that
is fixed for the lifetime of a file system object to which it refers. For example, a file
handle on UNIX contains the inode number and the generation count. As the
inodes are freed and reallocated, the generation count of the inode is
incremented when reused. This ensures that a client file handle that refers to the
old file cannot now be referred to the new file, even though the inode number
stays the same.
NFS V4 divides file handles into two types: persistent and volatile. Persistent file
handles describe the traditional file handle. Volatile file handles is a concept
introduced in NFS V4, in which a client must cache the mapping between path
name and file handle, and regenerate the file handle upon expiration.
2.7 AIX 5L v5.3 implementation of NFS V4

In this section, we discuss the parts of the NFS V4 protocol specification that IBM
has chosen to implement. AIX 5.3 is the first version of AIX to introduce support
for NFS V4. It is very important to note that it continues to support NFS V2 and
NFS V3. In fact, in AIX 5.3, Version 3 is still the default NFS protocol version that
is used in server exports and client mounts. This decision was primarily made to
allow for an easier migration to AIX 5.3 from previous versions. As you already
know, NFS V4 introduces some major changes, so it would not make sense to
make Version 4 the default version. The vers option may be used with mounts
and exports to specify NFS Version 4.
Note: In AIX 5.3, the default for exports is still NFS V3, not Version 4. You
must explicitly declare an export for NFS V4 using the vers option. See
Example 2-1 on page 26.

The initial AIX support places an emphasis on security, with support of the
optional NFS V4 ACL model when using the AIX Enhanced Journaled File
System. Support for managing access from foreign NFS V4 domains is also
included. NFS V4 provides the RPCSEC_GSS RPC authentication flavor
supporting the Kerberos 5 security mechanism with AIX 5.3. RPCSEC_GSS can
also be used with the NFS V3 protocol.
Note: The AIX Enhanced Journaled File System is a JFS2 file system with the
Extended Attributes Version 2 capability enabled to use NFS V4 ACLs.
When you first look at NFS on AIX V5.3, you will not see anything different. NFS
on AIX V5.3 supports NFS V2, NFS V3, and NFS V4. For this reason, all of the
daemons that you would see in previous versions of AIX are included in AIX 5.3.
One side effect of incorporating NFS V4 into AIX 5.3 is that NFS V3 security
controls can been extended by using some of the security features that are
required to create a V4 conforming implementation.
2.8 NFS V4 supported features in AIX 5.3

This section offers detailed descriptions of various features of NFS V4 that are
implemented in AIX 5L Version 5.3.
2.8.1 Mandatory feature support

The mandatory features of the NFS V4 protocol are supported as described in
RFC3530, with the following exceptions:
򐂰 The UTF-8 requirements are not fully supported.
Specifically, the transmission of file names and file system strings such as
symbolic link contents and directory entry names are not guaranteed to be in
UTF-8 format. Transmission of NFS attribute strings, such as owner and
owner group, are always in UTF-8 format. The NFS server and client do
perform UTF-8 validation on incoming string data as defined in RFC3530.
This checking may be administratively disabled using the nfso command.
Disabling UTF-8 checking may be necessary to use NFS V4 in environments
with non-UTF-8 configurations and data.
򐂰 The LIPKEY and SPKM-3 security mechanisms are not supported with
RPCSEC_GSS authentication.
Note: Kerberos V5 is the only RPCSEC_GSS security mechanism that is

supported in the implementation of NFS V4 in the initial release of AIX 5.3.

2.8.2 Other unsupported features
򐂰 Diskless client, Network Installation Management (NIM), and UDP are not
supported over NFS V4.
We have mentioned before that exporting individual files is not supported on
NFS V4, hence the lack of support for Diskless clients and NIM, which require
individual files to be exported.
򐂰 Delegation is not supported in the initial AIX V5.3 implementation.
2.8.3 Optional feature support

The following optional features of NFS V4 are supported:
򐂰 NFS V4 ACLs are supported by both the client and server.
The NFS client supports management of NFS V4 ACLs using the acledit,
aclget, and aclput utilities. The NFS server is capable of storing and
retrieving NFS V4 ACLs in underlying file systems that support the NFS V4
ACL model. We offer more information in the next section.
򐂰 Support is provided to map principals and file ownership attributes from one
NFS V4 domain into another.
This support is primarily intended for use at AIX NFS servers. It requires
deployment of Enterprise Identity Mapping. The NFS mappings are managed
using the chnfsim command.
򐂰 The AIX NFS V4 implementation supports two ACL types in underlying file
systems: NFS V4 and AIXC.
Both of these ACL types are described below and in more detail in 3.4.2,
“AIXC ACLs” on page 63 and 3.4.3, “NFS V4 ACLs: description” on page 65.
2.8.4 NFS4 ACL

The NFS4 ACL is the ACL defined by the NFS V4 protocol. It is
platform-independent, so it can be supported by other vendors’ clients or servers.
NFS V4 clients and servers are not required to support NFS4 ACL.
On an AIX server, if an underlying physical file system instance supports NFS4

ACLs, then the AIX NFS V4 server supports NFS4 ACL for that file system
instance. The NFS client supports management of NFS V4 ACLs using the
acledit, aclget, and aclput utilities. The NFS server is capable of storing and
retrieving NFS V4 ACLs in underlying file systems that support the NFS V4 ACL
model.

Most physical file system types on AIX do not support NFS4 ACL. These file
system types include, but are not limited to: CFS, UDF, JFS, and JFS2 with
extended attribute Version 1.
Note: Only JFS2 with extended attribute Version 2 (J2) supports NFS V4 ACLs.
See Chapter 3, “Enhanced security in NFS V4” on page 45 for more information
about NFS V4 ACL support.
2.8.5 AIXC ACLs

The AIXC ACL is an AIX-proprietary ACL. It is not defined by the NFS V4
protocol, and it is understood only by AIX servers and clients.
On an NFS V4 server, AIXC ACL is supported when the underlying file system
instance supports AIXC ACL. All instances of JFS and JFS2 support the AIXC
ACL.
An NFS V4 or NFS V3 client has a mount option that enables or disables support
for AIXC ACL. The default is to not support AIXC ACL. A user on an NFS V4
client file system can read and write AIXC ACL when both the client and the
server are running AIX, the underlying physical file system instance on the server
supports AIXC ACL, and the AIX client mounts the file system instance with
AIXC ACL enabled. AIXC ACL support in NFS V4 is similar to the AIXC ACL
support in AIX NFS V2 and NFS V3 implementations.
All instances of a JFS2 file system with extended attribute Version 2 support both
AIXC ACL and NFS V4 ACL. A file in this type of file system may have mode bits
only (no ACL), an NFS4 ACL, or an AIXC ACL. But it cannot have NFS4 ACL and
AIXC ACL at the same time.
The aclgettypes command can be used to determine the ACL types that can be
read and written on a file system instance. This command may return different
output when it runs against a physical file system on an NFS V4 server locally
than when it runs against the same file system on an NFS V4 client. For
example, an NFS V4 file system instance on an NFS V4 server may support NFS
V4 ACL and AIXC ACL, but the client is only configured to send and receive NFS
V4 ACL (mounted with the -noacl option). In this case, when aclgettypes is
executed from an NFS V4 client file system, only NFS V4 is returned. Also, if a
user on the client requests an AIXC ACL, an error is returned.
The authoritative source for access checking lies in the underlying file system
exported by the NFS server. The file system takes into consideration the file’s
access controls (ACLs or permission bits), the caller’s credentials, and other local
system restrictions that might apply.

Important: With AIXC or NFS V4 ACLs implemented in an exported file
system, applications and users should not assume that examination of UNIX
mode bits or ACLs alone can be used to conclusively predict access.
The aclget, aclput, and acledit commands can be used on the client to
manipulate either NFS4 or AIXC ACLs. For more information, see Access
Control Lists in AIX 5L Version 5.3 Security Guide, SC23-4907.
2.8.6 External name space (exname)

External name space (exname) is not part of the NFS V4 RFC. This is an option
specific to AIX implementation. The exname option extends the pseudo-file
system concept. The external name in your /etc/exports file must begin with the
nfsroot name. But an exname export does not have to correspond to the server’s
root. Figure 2-6 explains how this works.
Figure 2-6 Representation of how the server builds the exname pseudo FS view

We want to render the view in such a way that the client sees what is represented
by the pseudo-root part of Figure 2-6 on page 34. So, on the server, we want to
export the following file systems:
/local/home
/local/dept
/local/trans
/usr/codeshare/ThirdPartyProgs
We also want to make sure that we do not expose our server’s file system tree to
the client. How can we achieve this?
Attention: We must make sure that we set the pseudo-root on the server to
/exports. When the server renders the pseudo FS view for the client, the
directory or file under the /exports directory will be hidden. So, if you want to
have directories and files user /exports available to the clients, you should
either move them to another directory and export or choose a different
directory to be the anchor for the pseudo-root.
Using the exname option, we create the /etc/exports file:

/local/trans -vers=4,rw,exname=/exports/trans
/local/dept -vers=4,rw,exname=/exports/dept
/local/home -vers=4,rw,exname=/exports/home
/usr/codeshare/ThirdPartyProgs -vers=4,ro,exname=/exports/ThirdPartyProgs
We can see from this example that the exname option does not have the full path
to the individual exports under /exports. In fact, you could specify the full path for
all other exports. However, by following our example, the client is never shown
the server’s directory tree. This hides visibility of the actual server file system
layout from the client.
If you intend to export a large number of directories under /local allowing the local
component to be seen under /exports, then the exname option could also be
used as shown:
/local/trans -vers=4,rw,exname=/exports/local/trans
/local/dept -vers=4,rw,exname=/exports/local/dept
/local/home -vers=4,rw,exname=/exports/local/home
-vers=4,ro,exname=/exports/local/ThirdPartyProgs
2.8.7 Protocol differences: server exporting and client mounting

For a full description of allowable export restrictions and export semantics, see
the exportfs command description in AIX 5L Version 5.3 Commands Reference,

Volume 2, SC23-4889, and the /etc/exports file description in AIX 5L Version
5.3 Files Reference, SC23-4895.
Here are some differences between NFS V2, NFS V3, and NFS V4 in how
mounts are handled. In NFS V2 and NFS V3, the server exported the directories
that it wanted to make available for mounting. The NFS V2 or NFS V3 client then
had to explicitly mount each export to which it wanted access.
With NFS V4, the server still specifies export controls for each server directory or
file system to be exported for NFS access. From these export controls, the server
renders a single directory tree of all the exported directories. This tree, a
pseudo-file system, starts at the NFS V4 server’s pseudo-root. The NFS V4
pseudo-file system model enables an NFS V4 client, depending on its
implementation, to perform a single mount of the server’s pseudo-root in order to
access all of the server’s exported data. The AIX NFS V4 client supports this
feature. The actual content seen by the client is dependent on the server’s export
controls. (See 2.6.3, “Better namespace handling” on page 25.)
NFS V4 does not allow file-to-file mounting.
When a directory is exported by the server, that directory is only available to

clients using the NFS V2 or NFS V3 protocol by default. To allow access by NFS
V4, the export options must include the vers option. For details, see the exportfs
command description in AIX 5L Version 5.3 Commands Reference, Volume 2,
SC23-4889.
Client communication with the rpc.mountd daemon does not occur with NFS V4
mount processing. Operations in the core NFS V4 protocol are used to service
client side mount operations. The NFS V4 server implementation does utilize the
rpc.mountd daemon as part of handling NFS V4 access.
2.8.8 NFS files

This section includes information on the configuration files related to and used by
NFS and the new Version 4 protocol.
򐂰 “/etc/exports file” on page 37
򐂰 “/etc/xtab file” on page 38
򐂰 “/etc/nfs/hostkey file” on page 38
򐂰 “/etc/nfs/local_domain file” on page 38
򐂰 “/etc/nfs/realm.map file” on page 39
򐂰 “/etc/nfs/princmap” on page 39
򐂰 “/etc/nfs/security_default” on page 40

/etc/exports file
The /etc/exports file indicates all directories that a server exports to its clients.
Each line in the file specifies a single directory. The server automatically exports
the listed directories each time the NFS server is started. These exported
directories can then be mounted by clients. The syntax of a line in the
/etc/exports file is:
directory -option[,option]
Important: NFS V4 does not support file exporting. If you need to export a
specific file, export it as Version 2 or 3 (using the vers=2 or vers=3 options).
The directory is the full path name of the directory. Options can designate a
simple flag such as ro (read only) or a list of host names. See the specific
documentation of the /etc/exports file and the exportfs command for a complete
list of options and their descriptions.
Note: The /etc/rc.nfs script will not start the nfsd daemons or the rpc.mountd
daemon if the /etc/exports file does not exist.
Here are entries from a sample /etc/exports file:

/exports/project/proja -ro,access=nfs401:nfs402:nfs403:nfs404
/exports/home -root=nfs404,access=nfs404
/var/tmp
/exports/dept/sales -vers=4,sec=krb5,access=allsales,root=allsales2
The first entry in this example specifies that the /exports/project/proja

directory can be mounted by the systems named nfs401, nfs402, nfs403, and
nfs404. These systems can read data and run programs from the directory, but
cannot write to the directory.
The second entry in this example specifies that the /exports/home directory can
be mounted by the system nfs404 for read/write access. Additionally, nfs404 may
access data on the server as the root user.
The third entry in this example specifies that any client can mount the /var/tmp
directory (with read/write access).
Attention: You will notice there is no access list specified for the /var/tmp
entry. This means that all clients can mount this directory, and this is very
insecure. However, you now have the ability to use the sec=krb5 option, and
this would take away the worry of openly exporting the directory.

The fourth entry allows access to the /export/dept/sales directory only to
clients in the allsales netgroup using NFS V4 protocol and accessing the
directory using Kerberos 5 (krb5) authentication. Root access is allowed only
from allsales2.
Tip: Note the introduction of two new options: vers and sec.
/etc/xtab file
The /etc/xtab file has a format similar to the /etc/exports file and lists the currently
exported directories. Whenever the exportfs command is run, the /etc/xtab file
changes. This enables you to export a directory temporarily without having to
change the /etc/exports file. If the temporarily exported directory is unexported,
the directory is removed from the /etc/xtab file.
Important: The /etc/xtab file is updated automatically. You should not edit this
file manually.
/etc/nfs/hostkey file
This file is used by the NFS server to specify the Kerberos host principal and the
location of the keytab file. For instructions on how to configure and administer
this file, see the nfshostkey command description in AIX 5L Version 5.3
Commands Reference, Volume 4, SC23-4891, or the command’s man pages.
/etc/nfs/local_domain file
This file contains the local NFS domain of the system. It is implied that systems
that share the same NFS local domain also share the same user and group
registries. The chnfsdom command is used to administer this file:
# chnfsdom itsc.austin.ibm.com
#
# chnfsdom
Current local domain: itsc.austin.ibm.com
#
Important: You should not edit this file manually. Always use the chnfsdom
command. The command automatically tells the nfsrgyd daemon that you
have changed the local NFS domain. Also, a given NFS server or client
machine can only belong to one NFS V4 domain.
For further information about the chnfsdom command, see the AIX 5L Version 5.3
Commands Reference, Volume 1, SC23-4888, and the command’s man pages.

/etc/nfs/realm.map file
This file is used by the NFS registry daemon to map incoming Kerberos
principals of the form name@kerberos-realm to the form name@nfs-domain. It then
can resolve the name@nfs-domain to a local UNIX credential. This file provides
a simple way to map Kerberos principals into the server’s user registry. It is
appropriate when clients in different Kerberos realms will be accessing the
server, but the user namespace is global. For all Kerberos realms the server
supports, the file contains lines in the following format:
realm1 nfs-domain
realm2 nfs-domain
If the Kerberos realm name is always the same as the server’s NFS domain, this
file is not needed.
Important:
򐂰 Multiple Kerberos realms can map to a single nfs-domain. The previous
example demonstrates two realms mapping to one domain. However, a
single realm cannot map to multiple nfs domains.
򐂰 We recommended that you do not edit the /etc/nfs/realm.map file manually.
Always use the chnfsrtd command.
When the foreign identity mapping features of AIX NFS V4 support are used to
facilitate inter-nfs-domain access, the mapping rules managed by the chnfsim
utility allow mapping of realms into domains. In this case, chnfsim should be
used instead of chnfsrtd.
See the chnfsrtd command description in AIX 5L Version 5.3 Commands

Reference, Volume 1, SC23-4888, or the command’s man pages for more details.
/etc/nfs/princmap
This file maps host names to Kerberos principals when the principal is not the
fully qualified domain name of the server. It consists of any number of lines of the
following format:
<host part of principal> <alias1> <alias2> ...
To add, edit, or remove entries from this file, use the nfshostmap command. For
more information, see the nfshostmap command description in AIX 5L Version
5.3 Commands Reference, Volume 4, SC23-4891, or the command’s man(1)
pages.

/etc/nfs/security_default
The /etc/nfs/security_default file contains the list of security flavors that may be
used by the NFS client, in the order in which they should be used. Use the
chnfssec command to manage this file.
1. To create a new /etc/nfs/security_default file and tell the NFS client to first use
krb5, then krb5i, and then sys security:
# chnfssec -a krb5,krb5i,sys
#
# cat /etc/nfs/security_default
krb5
krb5i
sys
#
2. To add a security flavor:
# chnfssec -a krb5,krb5i,krb5p,sys
#
krb5
krb5i
krb5p
sys
Attention: If you add a security flavor to /etc/nfs/security_default, you

must specify all of the existing methods with the new methods when
running the chnfssec command. In this example, we add krb5p to
/etc/nfs/security_default. As you can see, we have already used all the
flavors in the file. Failure to do this (chnfssec -a krb5p) will result in the
existing file being overwritten with the flavor specified in the command.
3. To remove a security flavor:

# chnfssec -r krb5p
#
krb5
krb5i
sys
#
See the chnfssec command description in AIX 5L Version 5.3 Commands

Reference, Volume 1, SC23-4888, or man pages for more information.

2.8.9 Restricting NFS port ranges
The NFS_PORT_RANGE environment variable may be used to limit the source
port of network calls that the client makes to the server. This is very useful in a
firewalled environment. Prior to the introduction of this feature, you had no control
over what source port AIX would use, making firewall maintenance a difficult
task. If used, this environment variable should be added to the /etc/environment
file. The syntax of the environment variable is:
NFS_PORT_RANGE=udp[<starting_port>-<ending_port>]:tcp[<starting_port>-\
<ending_port>]
In the following example, UDP packets sent by the client will have a source port
in the range 3000 to 4000, and TCP connections will have a source port in the
range 5000 to 6000. This means that you can now restrict traffic on all other
higher range ports apart from the ones listed above.
NFS_PORT_RANGE=udp[3000-4000]:tcp[5000-6000]
2.8.10 Use of NFS_NOBODY

By default, and unauthenticated user is shown as user nobody belonging to group
nobody. This can be changed by setting the NFS_NOBODY environment variable.
This enables the user to be shown as whatever you set the NFS_NOBODY
variable to.
2.9 NFS daemons, files, and commands: a quick reference

Table 2-3 lists the NFS daemons and their subsystem names. The following
daemons are new to AIX and were specifically added to support NFS V4:
򐂰 /usr/sbin/gssd
򐂰 /usr/sbin/nfsrgyd
Table 2-3 List of NFS daemons

NFS daemons Description
/usr/sbin/gssd Services kernel requests for GSS operations.
/usr/sbin/nfsrgyd Provides a names translation service for NFS servers and

clients.
/usr/sbin/rpc.lockd Processes lock requests through the RPC package.
/usr/sbin/rpc.statd Provides crash-and-recovery functions for the locking

services on NFS.

NFS daemons Description
/usr/sbin/biod Sends the client’s read and write requests to the server. The
biod daemon is SRC-controlled.
/usr/sbin/rpc.mountd Answers requests from clients for file system mounts. The
mountd daemon is SRC-controlled.
/usr/sbin/nfsd Starts the daemons that handle a client’s request for file
system operations. nfsd is SRC-controlled.
/usr/sbin/portmap Maps RPC program numbers to Internet port numbers.

portmap is inetd-controlled.
/usr/sbin/rpc.rstatd Returns performance statistics obtained from the kernel.
/usr/sbin/rpc.pcnfsd Handles service requests from PC-NFS clients.
Table 2-4 gives an overview of all of the NFS files in AIX 5.3. The following files
are new to AIX 5.3 and were introduced specifically for NFS V4:
򐂰 /etc/nfs/hostkey
򐂰 /etc/nfs/local_domain
򐂰 /etc/nfs/realm.map
򐂰 /etc/nfs/princmap
򐂰 /etc/nfs/security_default
The /etc/filesystems file also has new options for NFS V4.
Table 2-4 List of NFS files and those new to AIX 5.3
Files Description
/etc/nfs/hostkey Used by the NFS server to specify the Kerberos host

principal and the location of the keytab file.
/etc/nfs/local_domain Contains the local NFS domain of the system.
/etc/nfs/realm.map Used by the NFS registry daemon to map incoming

Kerberos principals of the form name@kerberos-realm to the
form name@nfs-domain.
/etc/nfs/princmap Maps host names to Kerberos principals when the principal

is not the fully qualified domain name of the server.
/etc/nfs/security_default Contains the list of security flavors that may be used by the
NFS client, in the order in which they should be used.
/etc/filesystems Lists all file systems that can potentially be mounted and
their mounting configuration - persistent mounts.

Files Description
/etc/bootparms Lists the servers that diskless clients can use for booting
from.
/etc/exports Lists directories that can be exported to NFS clients.
/etc/networks Contains information about networks on the Internet

network.
/etc/pcnfsd.conf Options for the rpc.pcnfsd daemon.
/etc/rpc Contains database information for Remote Procedure Call

(RPC) programs.
/etc/xtab Lists directories that are currently exported.
Table 2-5 gives a list of all NFS commands in AIX 5.3. The following commands
are new in AIX 5.3 and were specifically included for NFS V4:
򐂰 /usr/sbin/chnfsdom
򐂰 /usr/sbin/chnfsrtd
򐂰 /usr/sbin/chnfssec
򐂰 /usr/sbin/nfshostkey
򐂰 /usr/sbin/nfshostmap
򐂰 /usr/sbin/nfs4cl
򐂰 /usr/sbin/chnfsim (delivered in the bos.cim.rte fileset)
Table 2-5 List of NFS commands and those new to AIX 5.3
Commands Description
/usr/sbin/chnfsdom Changes the local NFS domain.
/usr/sbin/chnfsrtd Changes NFS realm mappings.
/usr/sbin/chnfssec Manages the /etc/nfs/security_default file.
/usr/sbin/nfshostkey Configures the host key for an nfs server.
/usr/sbin/nfshostmap Manages mapping of host names to Kerberos principles.
/usr/sbin/nfs4cl Displays or modifies current NFS V4 statistics and

properties.
/usr/sbin/chnfsim Changes NFS foreign identity mappings.
/usr/sbin/chnfs Starts a specified number of biod and nfsd daemons.
/usr/sbin/mknfs Configures the system to run NFS and starts NFS

daemons.

Commands Description
/usr/sbin/nfso Configures NFS network options.
/usr/sbin/automount Mounts an NFS automatically on access.
/usr/sbin/chnfsexp Changes the attributes of an NFS-exported directory.
/usr/sbin/chnfsmnt Changes the attributes of an NFS-mounted directory.
/usr/sbin/exportfs Exports and unexports directories to NFS clients.
/usr/sbin/lsnfsexp Displays the characteristics of directories that are exported

with NFS.
/usr/sbin/lsnfsmnt Displays the characteristics of mounted NFS systems.
/usr/sbin/mknfsexp Exports a directory.
/usr/sbin/mknfsmnt Mounts a directory using NFS.
/usr/sbin/rmnfs Changes the configuration to stop the NFS daemons.
/usr/sbin/rpcinfo Reports the status of RPC servers.
/usr/sbin/rmnfsexp Removes NFS-exported directories from a server’s list of

exports.
/usr/sbin/rmnfsmnt Removes NFS-mounted file systems from a client’s list of

mounts.
/usr/sbin/nfsstat Displays information about a machine’s ability to receive

calls.
/usr/sbin/rpcgen Generates C code to implement an RPC protocol (delivered

by bos.net.nfs.server fileset).

3
Chapter 3. Enhanced security in NFS V4

This chapter describes the enhanced security options that are available in
NFS V4 as released with the AIXL Version 5.3 operating system.
It starts by introducing general terminology, and then goes on to describe specific

features.
The chapter includes the following sections:

򐂰 General security concepts and terminology
򐂰 NFS V4 user/group identification
򐂰 NFS V4 user authentication
򐂰 NFS V4 user authorization
򐂰 NFS V4 host identification
򐂰 NFS V4 host authentication
򐂰 NFS V4 host authorization

3.1 General security concepts and terminology
This section lays a foundation for the rest of the chapter by introducing
terminology that will that will be used throughout the chapter.
3.1.1 Broad security categories

The topic of security can be very broad and far-reaching. A discussion of security
measures might include items in the following categories:
Physical security Measures taken to control physical access to a facility or
resource. Padlocks, fences, guards, and dogs are
examples of physical security measures.
Personnel security Measures taken to help ensure that the people who are
granted access to secured resources are reliable and are
not likely to compromise the security of those resources.
Security clearances and photo ID badges are examples of
personnel security measures.
Information security Measures taken to protect important information from
unauthorized disclosure, tampering, or destruction.
Passwords, encryption, and file access permissions are
examples of information security measures.
We will be talking in this chapter primarily about what we call information

security, because NFS V4 is about sharing information. Organizations will not be
able to properly protect their information resources without also implementing
physical and personnel security measures, but we will not address those
measures in this document.
3.1.2 Information security components

In an NFS V4 context, information security falls into the following areas:
Identification Uniquely establishes the identity of information system
users, hosts, and services. Answers questions such as,
“Who are the users or hosts that are trying to access
shared data on my server?”
Authentication Confirms the identity of a system user, host, or service.
Answers questions such as, “Is this user really who he or
she claims to be?”

Authorization Controls what shared information each system user or
other entity can access. Answers questions such as,
“Does this user have the right to access a shared data
object on my server?
3.1.3 RPC security flavors

NFS V4 uses Sun’s Remote Procedure Call (RPC) protocol to communicate over
the network between the client and the server. The IBM implementation enables
you to use three different RPC security flavors:
򐂰 Basic UNIX security (AUTH_SYS, aka AUTH_UNIX)
򐂰 Diffie-Hellman security (AUTH_DH, aka AUTH_DES)
򐂰 RPCSEC_GSS security as defined in RFC2203
3.1.4 RPCSEC_GSS protection levels

When using RPCSEC_GSS security with RPCs, there are three levels of
protection that can be applied to the RPCs as they are transmitted over the
network between server and client:
Authentication Validates the identity of RPC sender
Integrity Validates that the contents of the RPC were not changed
during transmission (also includes authentication)
Privacy Prevents unauthorized viewing of data while it is in transit
between client and server (also includes authentication
and integrity)
Keep in mind that each increasing level of protection comes with a performance
penalty. Choose the minimum level that meets your data protection requirements.
3.1.5 RPCSEC_GSS protection mechanisms

The NFS V4 standard (RFC 3530) requires that NFS implementations support
three different RPCSEC_GSS mechanisms:
򐂰 Kerberos V5 (RFC1964)
򐂰 SPKM-3/LIPKEY (RFC2847)
򐂰 SPKM-3 on its own (RFC2847/RFC2025), for situations where the initiator
(the client) is anonymous or has its own certificate.
In AIX 5.3, IBM has implemented Kerberos V5, but not SPKM/LIPKEY at this
point in time.
Chapter 3. Enhanced security in NFS V4 47

IBM offers several forms of encryption with its Kerberos V5 implementation,
among which are single DES and triple DES.
Note: Triple DES encryption gives the best protection, but you may need to
use single DES to get better performance and interoperability.
3.1.6 Looking ahead to the rest of the chapter

The rest of this chapter compares different options for identification,
authentication, and authorization, which are different for hosts and users. We
discuss users first, and then hosts.
We use various Kerberos-related terms throughout this chapter. See Appendix A,

“Kerberos” on page 243 for a description of Kerberos and its related terms.
3.2 NFS V4 user/group identification

In this section, we discuss three different methods for managing user and group
identities: via the standard UNIX /etc/passwd and /etc/group files, via Sun
Microsystems’ Network Information Services (NIS), and via the Lightweight
Directory Access Protocol (LDAP) schema defined by RFC2307.
We also discuss how NFS V4 determines a user’s identity under AUTH_SYS

authentication and RPCSEC_GSS (Kerberos) authentication.
3.2.1 User identity management options

UNIX implementations typically use a 32-bit integer to identify users and groups.
These integers are referred to as UIDs for users and GIDs for groups. User and
group ownership for system processes and file system objects are maintained in
UID/GID form. NFS V2 and NFS V3 also use these UIDs and GIDs to identify
users and groups.
People do not usually work directly with the numeric IDs; they work with text user
and group names that are normally easier to associate with an actual individual
or group. When presenting information about process and file ownership, the
system translates the numeric IDs into the names. The relationship between the
names and the IDs is maintained in a user registry, which can be standard UNIX,
NIS, or LDAP.

Standard UNIX user registry
In standard UNIX, user-name-to-ID mappings are contained in the /etc/passwd
file, and group-name-to-ID mappings are contained in the /etc/group file.
All but the smallest organizations will want to use a shared user registry, rather
than maintaining separate /etc/passwd and /etc/group files on all hosts. Here is
one reason why. All clients using data stored on an NFS server directory should
use the same identifier to represent the same user or group. This is necessary to
maintain consistent file ownership. For example, in an NFS V2 and NFS V3
environment, if UID 100 is Joe on one NFS client and Mary on another client,
NFS files created by Joe from the first client will show as being owned by Mary
on the 2nd client, and vice versa. To avoid this, the separate /etc/passwd and
/etc/group files will need to be kept in sync on all NFS clients that access data on
a common NFS server. This can be a very expensive and error-prone task.
NIS user registry

Sun Microsystems developed Network Information Services (NIS) to help
centrally manage UID/GID- to-name mappings. With NIS, the user and group
identity information is maintained in one place and is distributed to all of the
machines participating in an NIS domain. Having all of an organization’s NFS
servers and clients participate in an NIS domain helps keep identity mappings
consistent.
NIS has been widely used in conjunction with NFS, and it can work well for
medium-sized organizations. It becomes problematic, however, when a large
organization requires multiple NIS domains, or when two different organizations
merge. Consolidating two different NIS domains can be a very difficult task.
More information about NIS can be found on Sun Microsystems’ Web site:
http://www.sun.com
LDAP user registry

Directory services based on the Lightweight Directory Access Protocol (LDAP)
can also be used to maintain user and group identities. LDAP-based directories
are typically more scalable and secure than NIS. RFC2307 describes an LDAP
schema that provides NIS-like functionality.
IBM Tivoli Directory Server is the IBM implementation of LDAP-based directory

services. As of Version 5.1, the Tivoli Directory Server supports RFC2307 user
and group identification. More information is available about the Tivoli Directory
Server on the IBM Web site at:
http://www.ibm.com/software/tivoli/products/directory-server

3.2.2 User/group identities and NFS V4
NFS V4 handles user and group identities very differently from previous versions.
NFS V2 and NFS V3 pass UIDs and GIDs between the client and server,
whereas NFS V4 passes string names in the form user@nfs_domain or
group@nfs_domain. We will describe identity mapping under three conditions:
򐂰 Single NFS domain using AUTH_SYS security
򐂰 Single NFS domain using Kerberos security
򐂰 Multiple NFS domains
We will use two sample operations in each case:

򐂰 Creating a file
򐂰 Requesting file ownership
Notes:
1. When we say “NFS” throughout the rest of this section, we mean NFS V4.
2. The examples in this section are not intended to depict everything that
goes on in an NFS transaction. They are simplified to convey pertinent
concepts.

Single NFS domain using AUTH_SYS security
Figure 3-1 shows how user information is passed from the NFS client to the
server when creating a file.
Figure 3-1 Creating a file under AUTH_SYS security
The ownership of the requesting user process is in the form of UID and GID. The
NFS client translates the UID into a user name string and translates the GID into
a group name string. These strings are then sent across to the server. The server
checks to make sure that the NFS domains in the request match its own NFS
domain. (If they didn’t match, extra steps would be required. See the multiple
domain discussion below.) The server then translates the strings back to
UID/GID using its user registry, which may not be the same as the client’s, and
stores the UID/GID as ownership attributes of the file being created.

Important: For ownership to be transferred properly between client and
server, the server must be able to convert the user and group names into IDs.
If the server cannot do that, it will make the file owned by the user nobody, or
the group nobody, or both. This is a key difference between NFS V4 and earlier
versions. Prior to NFS V4, the server just took the provided UID/GID
information and placed it in the file attributes. It did not have to do any
translation.
Figure 3-2 shows how the user information is passed from the NFS server to the
client when the client has requested ownership information for an existing file.
Figure 3-2 Getting file ownership under AUTH_SYS security
The ownership information that is stored with the file is in the form of UID and
GID. The NFS server translates the UID into a user name string and the GID into

a group name string. It then checks to make sure that its NFS domain is the
same as the requester’s NFS domain. In this case they are the same, so the
server can go ahead and send the owner and group strings across to the client.
The client translates the strings back to UID/GID using its user registry, which
may not be the same as the server’s, and reports that information back to the
requesting process.
The ls -n output would appear similar to this:

-rw-r--r-- 1 200 20 255 Aug 11 11:05 myfile
If an ls -l command was issued, the client would take the UID/GID information
and translate it back to names, resulting in output similar to:
-rw-r--r-- 1 sally hr 255 Aug 11 11:05 myfile
Single NFS domain using Kerberos security

Under Kerberos security, the NFS server does not take the client user’s identity
from the user and group strings in the NFS request. Instead, it determines the
user’s identity from the Kerberos principal that is part of the RPCSEC_GSS call.
The principal identity is in the form user@KERBEROS_REALM. (The realm
name is typically represented in all capital letters.)
Figure 3-3 on page 54 shows how an NFS server determines ownership when an
NFS client requests to create a file in a directory that is under Kerberos security.

Figure 3-3 Creating a file under RPCSEC_GSS (Kerberos) security
The server takes the requesting principal’s realm and maps it to an NFS domain
via the contents of the /etc/nfs/realm.map file. It then checks to see whether the
resulting NFS domain matches its own NFS domain. In this case, the domains
match, so the server then looks up the user name in its user registry to determine
the UID. The server also gets the user’s GID from its user registry, not from the
NFS request. It then places the UID and GID in the newly created file’s attributes.

Attention: Determining the group ownership under Kerberos is an important
difference from how NFS operates under AUTH_SYS security. With
AUTH_SYS, the server gets the group information from the client via the NFS
request. With Kerberos, the server only gets the user identity from the client. It
gets the group information from the server’s user registry.
Figure 3-3 illustrates this. Although the user sally’s primary group is eng on the
client, the created file is owned by the group hr because it is sally’s primary
group on the server. Using ls -l on the file on both the server and the client will
show that it is owned by the ‘hr’ group.
Figure 3-4 on page 56 shows what happens when an NFS client requests an
NFS server for ownership information for a file that is in a Kerberos-protected
directory.

Figure 3-4 Getting file ownership under RPCSEC_GSS (Kerberos) security
The process is the same as the AUTH_SYS process in Figure 3-2 on page 52,
except (again) that the client user’s identity is determined from the Kerberos
principal accompanying the request. The NFS server translates the principal’s
realm to an NFS domain to check to make sure that the requester’s domain is the

same as the server’s domain. Note that the file ownership information is still
communicated through the NFS protocol.
Multiple NFS domains

If the client’s NFS domain does not match the server’s, name mapping must be
made between the domains. User and group names in the client’s domain must
be mapped to user and group names in server’s domain. This can be done via
Enterprise Identity Mapping (EIM) that is available for free with AIX. (It is
available on the base AIX CD, in the fileset bos.eim.rte.) See the IBM Redbook
Windows-based Single Signon and the EIM Framework on the IBM ^
iSeries Server, SG24-6975, for an overview of EIM and a description of how to
configure it for use with the IBM Network Authentication Service. Although the
book is written for iSeries™ servers, the concepts also apply to pSeries® servers
and AIX.
If EIM is not in place, the NFS server will view the client identities as foreign, and
it will map them to the user/group nobody.
Figure 3-5 on page 58 shows the process of creating a file under AUTH_SYS
when EIM is used.

Figure 3-5 Creating a file under AUTH_SYS security: multiple domains
Note the extra EIM step after the server determines that the client’s NFS domain
does not match its own. The server requests from EIM which nfsdom1 user
matches sally@nfsdom2, and which nfsdom1 group matches hr@nfsdom2. The
figure shows the user and group names being the same between the two
domains, but it is possible for them to be different. The user name
sally@nfsdom2 might have mapped to the name sally2@nfsdom1, or it might
even have mapped to mary@nfsdom1.
The other scenarios presented earlier in this section all also require the
additional EIM step if the client’s NFS domain does not equal the server’s.

3.3 NFS V4 user authentication
By default, NFS uses the AUTH_SYS method to authenticate user identities.
Alternatively, you can use RPCSEC_GSS (Kerberos) to authenticate users.
3.3.1 AUTH_SYS user authentication

Under the AUTH_SYS security flavor, the user is authenticated at the client,
usually via a logon name and password. The NFS server trusts the user and
group identities presented by its clients. If someone gains administrative control
of an NFS client, or has control of a machine pretending to be a valid NFS client
via IP address spoofing, it is easy to masquerade as any valid NFS user.
For example, if tight physical security cannot be maintained on a network, it

would be easy for someone to bring in a Linux laptop, set the IP address to be
the same as a valid NFS client, and connect the laptop to the network using the
valid client’s connection. When connected, the person can set up any user
account desired on the laptop and gain access to any of the exported data on an
NFS server.
Attention: Because of this vulnerability, you should not use AUTH_SYS user
authentication if controlling access to your data is important.
3.3.2 RPCSEC_GSS user authentication using Kerberos

This section first gives a little background about Kerberos in general, and then it
describes how Kerberos is used in NFS.
What is Kerberos?
The AIX 5L Version 5.3 Security Guide describes Kerberos as a network
authentication service that provides a means of verifying the identities of
principals (users and hosts) on physically insecure networks. Kerberos provides
mutual authentication, data integrity, and privacy under the realistic assumption
that network traffic is vulnerable to capture, examination, and substitution.
Kerberos tickets are credentials that verify a principal’s identity. There are two
types of tickets: ticket-granting and service. The ticket-granting ticket is for the
initial identity request. When logging into a host system, a user needs something
that verifies his or her identity, such as a password or a token. This password or
token is used to obtain a ticket-granting ticket, which can then be used to request
service tickets for specific services. This two-ticket method is the called the
Kerberos third-party authentication model.

The trusted third-party or intermediary in Kerberos is called the Key Distribution
Center (KDC). The KDC issues all of the Kerberos tickets to the clients. Tickets
have a predetermined lifetime and have to be renewed periodically.
The Kerberos database keeps a record of every principal; the record contains the
name, private key, expiration date of the principal, and some administrative
information about each principal. This database is maintained on the master
KDC, and it can be replicated to one or more replica KDCs.
See Appendix A, “Kerberos” on page 243 for more information about what
Kerberos is and how it works.
How does NFS use Kerberos?

When an NFS client and server are using Kerberos 5 (krb5) authentication, the
client and server must establish a security context for NFS requests. The security
context is a data structure that indicates that the client and server have
completed a mutual authentication procedure. The context also contains the
encryption keys that will be used for protecting exchanged data, if such
protection has been requested. The security context has a lifetime and may need
to be refreshed by the client from time to time. The client and server each
maintain a cache of security contexts, one context per user/host combination.
A walkthrough of the authentication process helps explain how it works.

Figure 3-6 on page 61 illustrates the steps involved in authenticating an NFS
user request.

Figure 3-6 RPCSEC_GSS authentication flow
Here is a description of the authentication steps:

1. An authenticated user process on the NFS client attempts an operation on a
Kerberos-protected NFS object.
2. The client NFS has no current context for the user/server combination. It calls
out to the gssd to obtain an RPC security context with the target service (the
server NFS). It gives the gssd the name of the service (machine principal of
the NFS server) and the user’s identity.
3. If the user has valid Kerberos credentials, the gssd then contacts the KDC to
initiate context establishment to the target service.
4. If the service is known, the KDC returns an opaque token to the client gssd.
This token will authenticate the user to the target.
5. The gssd returns the token to the client NFS.
6. The client NFS then sends the token to the server NFS in a NULL RPC call to
request context establishment.

7. The server NFS passes the context token to the server gssd in a request to
accept the context.
8. The server gssd passes the token to the KDC to verify the requesting user’s
identity.
9. If the KDC accepts the token, it returns another token to the server gssd and
a context handle. This token will authenticate the server’s identity to the
requesting client.
10.The server gssd returns this token to the server NFS. The server NFS now
has established context.
11.The server NFS returns the accept token to the client in response to the NULL
RPC call.
12.The client passes the accept token to the client gssd in a second call to
initiate the context.
13.The client gssd calls the KDC to verify the server’s token.
14.If the KDC accepts the token, it returns a context handle to the client gssd.
15.The client gssd returns the context handle to the client NFS. The client now
has established context.
16.The original NFS operation can now proceed under the established context.
17.The NFS server responds to the original operation.
18.Results of the original operation are returned to the user process.
3.4 NFS V4 user authorization

User authorization in an NFS context means controlling user access to
directories and files in the exported file systems. This section describes two ways
to control this access: via standard UNIX file permissions and via NFS V4
Access Control Lists (ACLs).
Note: Although you can also use AIX ACLs (known in AIX 5.3 as the AIXC
ACL type), they are only supported on AIX systems, and they have not been
widely adopted. We do not discuss them in detail in this document. For more
about AIXC ACLs, see the AIX 5L Version 5.3 Security Guide, SC23-4907.
ACLs provide for more granular access control than standard UNIX file
permissions. One of the main differences is in group access. Whereas standard
UNIX permissions only provide access control for one group (the group that owns
the file), ACLs enable different access permissions to be specified for multiple
groups. ACLs also allow access permissions to be specified at a user level, but

controlling access on a per-user basis is usually not practical for organizations of
any size.
3.4.1 Standard UNIX file permissions

The standard UNIX file permission model consists of granting read, write, and
execute access to three categories of users:
User The user who owns the file
Group The group that owns the file
Other Everyone else
This model frequently is inadequate when applied to a real-world organization

structure. Individuals often operate in multiple roles, or different individuals
operating in the same role may have attributes that require different data access.
For example, access to data is sometimes restricted based on company

affiliation. Two people working on an engineering project might be from different
companies. Although they both can have access to much of the engineering
data, some of that data may be restricted to employees of one company and not
the other. This sort of granular access control is very difficult, if not impossible, to
implement using standard UNIX file permissions.
3.4.2 AIXC ACLs

An AIXC ACL has two parts:
򐂰 Base permissions, which map directly to the standard user/group/other UNIX
file permissions
򐂰 Extended permissions, which enable you to control access to other specific
users and groups
Here is an example of an AIXC ACL (in aclget format) that has extended
permissions disabled:
*
* ACL_type AIXC
*
attributes: SGID
base permissions
owner(root): rwx
group(system): r-x
others: r-x
extended permissions
disabled

Here is an example of an AIXC ACL that has extended permissions enabled:
*
* ACL_type AIXC
*
attributes: SGID
base permissions
owner(root): rwx
group(system): r-x
others: r-x
enabled
permit rw- u:sally
deny rwx g:sales
AIXC ACLs from an NFS client’s viewpoint

AIXC ACLs are supported on both NFS V3 and NFS V4 AIX clients. To be able to
see and modify AIXC ACLs from an NFS client, you must mount your file
systems with the acl option (noacl is the default).
There is one aspect in which an NFS V4 client handles AIXC ACLs differently
from an NFS V3 client. An NFS V4 client depicts an AIXC ACL as an NFS V4
ACL if the AIXC ACL has extended permissions disabled, or if extended
permissions are enabled but there are no entries in the extended permissions
list. If the AIXC ACL has extended permissions enabled and there are entries in
the extended permissions list, the NFS V4 client shows it as an AIXC ACL. An
NFS V3 client always shows an AIXC ACL as an AIXC ACL.
For example, assume a file has the following AIXC ACL on the server:
*
* ACL_type AIXC
*
attributes:
base permissions
owner(root): rw-
group(system): r--
others: r--
disabled
Then the NFS V4 client will show the ACL as:

*
* ACL_type NFS4
*
*
* Owner: root
* Group: system

*
s:(OWNER@): a rwpRWaAdcCs
s:(OWNER@): d xo
s:(GROUP@): a rRadcs
s:(GROUP@): d wpWxACo
s:(EVERYONE@): a rRadcs
s:(EVERYONE@): d wpWxACo
The AIX 5L Version 5.3 Security Guide, SC23-4907, has more about AIXC ACLs.
3.4.3 NFS V4 ACLs: description

NFS V4 ACLs are similar to Windows NTFS ACLs, but they are not identical. The
developers of the NFS V4 standard chose the Windows ACLs model over POSIX
ACLs because the Windows ACL model is both richer and more widely deployed.
Although many UNIX vendors implemented ACLs based on the POSIX Draft
ACL specification, those implementations tended to be proprietary, and the
POSIX specification was never standardized.1
Note: In order to use NFS V4 ACLs, the server file system must support them.
As of this writing, AIX5L Version 5.3 only supports NFS V4 ACLs in two file
system types: Enhanced Journaled File System (JFS2) with the extended
attribute format set to Version 2 (EAv2), and General Parallel File System
(GPFS). For more information about NFS V4 ACL support, see the AIX 5L
Version 5.3 Security Guide, SC23-4907, and the AIX 5L Differences Guide
Version 5.3 Edition, SG24-7463.
NFS V4 ACL format

According to the NFS V4 protocol specification, an NFS V4 ACL is an array of
Access Control Entries (ACEs) that have four elements: a type, a set of flags, an
access bit mask, and an identity. As implemented in the AIX JFS2 EAv2 file
system, each ACL can be a maximum of 64 KB.
The textual representation of an NFS V4 ACL consists of a list of Access Control

Entries (ACEs), one per line. Each ACE has four elements in the following format:
IDENTITY ACE_TYPE ACE_MASK ACE_FLAGS
IDENTITY has the format:

IDENTITY_type:(IDENTITY_name or IDENTITY_ID or IDENTITY_who):
Table 3-1 on page 66 lists possible values for IDENTITY_type.

1
Pawlowski, Brian, Spencer Shepler, Carl Beame, Brent Callaghan, Michael Eisler, David Noveck,
David Robinson, Robert Thurlow. The NFS Version 4 Protocol, SANE Conference, NL, 2000.
http://www.nluug.nl/events/sane2000/papers/pawlowski.pdf.

Table 3-1 ACE IDENTITY_type values
IDENTITY_type Description
u user (IDENTITY_name or IDENTITY_ID)
g group (IDENTITY_name or IDENTITY_ID)
s special who string (IDENTITY_who)
IDENTITY_name is the user or group name.
IDENTITY_ID is the user or group numeric ID.
IDENTITY_who is a special who string that should be understood universally

rather than in the context of a particular NFS domain. The string can be one of
those shown in Table 3-2.
Table 3-2 ACE special who strings supported by AIX

IDENTITY_who Description
OWNER@ The owner of the file
GROUP@ The group associated with the file
EVERYONE@ The world
The additional special who strings specified in RFC 3530 (shown in Table 3-3)
are not currently supported in AIX.
Table 3-3 ACE special who strings not supported in AIX

IDENTITY_who Description
ANONYMOUS@ Accessed without any authentication
AUTHENTICATED@ Any authenticated user (opposite of ANONYMOUS)
INTERACTIVE@ Accessed from an interactive session
NETWORK@ Accessed via the network
DIALUP@ Accessed via a dial-up connection
BATCH@ Accessed from a batch job
SERVICE@ Accessed from a system service
ACE_TYPE is a single character. Table 3-4 on page 67 shows possible values.

Table 3-4 ACE_TYPE values
ACE_TYPE Description
a Allow access
d Deny access
l Generate a system alarm when an access is attempted

(currently not supported in AIX)
u Generate an audit log entry when an access is attempted

(currently not supported in AIX)
ACE_MASK is a set of permission flags (permission bits) that can be combined

without any separator. Table 3-5 shows possible flags. Note that the flag values
are case-sensitive.
Table 3-5 ACE_MASK values

ACE_MASK RFC 3530 name Description
r READ_DATA or Permission to read the data of the file or

LIST_DIRECTORY list the contents of the directory
w WRITE_DATA or Permission to modify the file’s data or add

ADD_FILE a new file to the directory
p APPEND_DATA or Permission to append data to the file or

ADD_SUBDIRECTORY add a new subdirectory to the directory
R READ_NAMED_ATTRS Permission to read the named attributes

of the file or directory
W WRITE_NAMED_ATTRS Permission to write the named attributes

of the file or directory
x EXECUTE Permission to execute the file or traverse

the directory
D DELETE_CHILD Permission to delete files or

subdirectories from within the directory
a READ_ATTRIBUTES Permission to read basic attributes

(non-ACLs) of the file or directory
A WRITE_ATTRIBUTES Permission to change basic attributes

(non-ACLs) of the file or directory
d DELETE Permission to delete the file or directory
c READ_ACL Permission to read the ACL of the file or

directory

ACE_MASK RFC 3530 name Description
C WRITE_ACL Permission to change the ACL of the file

or directory
o WRITE_OWNER Permission to change the owner of the file

or directory
s SYNCHRONIZE Permission to access file locally at the

server with synchronous reads and writes
ACE_FLAGS (optional) is a combination of one or more of the following two-letter

flags without any separator. Four of the currently defined flags have to do with
ACL inheritance (Table 3-6), and the other two have to do with auditing
(Table 3-7). The inheritance flags only have meaning when applied to a directory.
Auditing flags only have meaning when used with the audit or alarm ACE types.
Table 3-6 ACE_FLAG inheritance-related values

ACE_FLAG RFC 3530 name Description
fi FILE_INHERIT Indicates that this ACE should be added to each

newly created non-directory file.
di DIRECTORY_ Indicates that this ACE should be added to each

INHERIT newly created subdirectory.
oi INHERIT_ONLY Indicates that this ACE does not apply to the

current directory; it is only to be added to newly
created files/subdirectories as specified by the
above two flags.
ni NO_ Indicates that this ACE should be added to newly

PROPAGATE_ created files/subdirectories immediately under
INHERIT the directory, but subdirectories should not pass
it on to their children.
Table 3-7 ACE_FLAG auditing-related values

ACE_FLAG RFC 3530 name Description
sf SUCCESSFUL_ Generate audit or alarm when an access attempt

ACCESS_ succeeds.
ACE_FLAG
ff FAILED_ Generate audit or alarm when an access attempt

ACCESS_ fails.
ACE_FLAG

NFS V4 ACL permission restrictions
Some permission bits are interrelated and must be used together under the
following circumstances:
򐂰 The WRITE_DATA (w) and APPEND_DATA (p) bits must be specified
together in a file’s ACE, or in a directory’s ACE that has the FILE_INHERIT
flag set.
Special user permissions

In NFS V4, two classes of users have special access to files:
򐂰 The UNIX super user (UID=0) is allowed all access permissions, regardless
of the ACE permission bit settings. (The only exception to this is execute
permission.) This special access applies to processes running on the NFS
server and processes running on NFS clients that have been given root
access via the exportfs command.
򐂰 The owner of a file always has the permissions READ_ACL, WRITE_ACL,
READ_ATTRIBUTES, and WRITE_ATTRIBUTES, regardless of the actual
settings in the ACL.
3.4.4 NFS V4 ACLs: ACL evaluation

In order to properly use NFS V4 ACLs, it is important to understand how they are
evaluated when determining whether an access request will be granted or
denied.
Per the RFC 3530 NFS V4 standard and the AIX 5L Version 5.3 Security Guide,
an AIX NFS V4 server evaluates the ACL list from the top down, applying the
following rules:
򐂰 Only ACEs that have a who that matches the requester are considered. The
credentials of the requester are not checked while processing the ACE with
special who EVERYONE@.
򐂰 Each ACE is processed until all of the bits of the requester’s access have
been allowed or at least one of the requested bits not previously allowed has
been denied.
򐂰 When a permission bit has been allowed, it is no longer considered in the
processing of later ACEs.
򐂰 If a deny ACE_TYPE is encountered where the ACE_MASK has bits in
common with not-yet-allowed bits in the request, access is denied, and the
remaining ACEs are not processed.
򐂰 If the entire ACL has been processed and some of the requested access bits
still have not been allowed, access is denied.

NFS V4 ACL evaluation examples
The following examples help illustrate ACL evaluation. For more examples, see
the AIX 5L Version 5.3 Security Guide.
Given the following ACL on a file:

*
* ACL_type NFS4
*
*
* Owner: sally
* Group: staff
*
g:sales: d wp
s:(OWNER@): a rRWDaAdcCs
s:(OWNER@): d wpo
s:(GROUP@): a rwpRxadcs
s:(GROUP@): d WDACo
s:(EVERYONE@): a rwpRxadcs
s:(EVERYONE@): d WDACo
If the user sally requests READ_DATA (r) and WRITE_DATA (w) access, the ACL
evaluation will proceed as follows:
򐂰 The s:(OWNER@):a... ACE is processed because sally owns the file.
– READ_DATA is allowed because that bit is set in the ACE_MASK.
– WRITE_DATA is not yet allowed because it is not set in the ACE_MASK.
򐂰 The s:(OWNER@):d... ACE is processed because sally owns the file.
– WRITE_DATA is denied because that bit is set in the ACE_MASK, and
WRITE_DATA has not yet been allowed by a previous ACE.
򐂰 No further ACEs are processed, and the requested access is denied.
Notes:
1. In the previous example, even though the GROUP@ and EVERYONE@
ACEs allow WRITE_DATA access, Sally is denied WRITE_DATA access
because it is specifically denied by the owner ACE.
2. The ACE order is important. If the group allow ACE had appeared in the list
before the owner deny ACE, then Sally would be allowed write access to
the file.

If the user sally, who is a member of the group staff, requests READ_DATA (r)
and EXECUTE (x) access, the ACL evaluation will proceed as follows:
򐂰 The s:(OWNER@):a... ACE is processed because sally owns the file.
– READ_DATA is allowed because that bit is set in the ACE_MASK.
– EXECUTE is not yet allowed because it is not set in the ACE_MASK.
򐂰 The s:(OWNER@):d... ACE is processed because sally owns the file.
– EXECUTE is not yet denied because it is not set in the ACE_MASK.
򐂰 The s:(GROUP@):a... ACE is processed because sally is a member of the
group staff, which owns the file.
– EXECUTE is allowed because that bit is set in the ACE_MASK.
򐂰 All requested permission bits have now been allowed. No further ACLs are
processed, and the requested access is granted.
If the user joe, who is a member of the group sales, requests READ_DATA (r)
and WRITE_DATA (w) access, the ACL evaluation will proceed as follows:
򐂰 The g:sales:d... ACE is processed because joe is a member of the group
sales.
– READ_DATA is not denied because it is not set in the ACE_MASK.
– WRITE_DATA is denied because that bit is set in the ACE_MASK, and
WRITE_DATA has not yet been allowed by a previous ACE.
򐂰 No further ACEs are processed, and the requested access is denied.
If joe requests just READ_DATA (r) access, the ACL evaluation will proceed as
follows:
򐂰 The g:sales:d... ACE is processed because joe is a member of the group
sales.
– READ_DATA is not denied because it is not set in the ACE_MASK.
򐂰 The s:(EVERYONE@):a... ACE is processed.
– READ_DATA is allowed because it is set in the ACE_MASK.
򐂰 All requested permission bits have now been allowed. No further ACLs are
processed, and the requested access is granted.
Relationship between NFS V4 ACLs and UNIX permissions

Those familiar with standard UNIX read (r), write (w), and execute (x) permission
bits may want to know how these bits correlate with the bits in the ACE_MASK.
As you can probably tell from the ACE_MASK definitions above, the r, w, and x
permission bits basically provide the same access as the standard UNIX
permission bits.

For example, as with UNIX permissions, the w permission bit in a directory’s ACL
has to do with creating, deleting, and renaming files and subdirectories within
that directory, rather than changing the contents of those files and subdirectories.
The difference comes into play when mapping the rwx bits to user (owner),
group, and other. This mapping is unspecified in RFC 3530. Here is an example
of the mapping we observe in AIX.
Consider a file with the following ACL:

*
* ACL_type NFS4
*
*
* Owner: sally
* Group: staff
*
s:(OWNER@): a cCs
s:(OWNER@): d o
s:(GROUP@): a rRxadcs
s:(GROUP@): d wpWDACo
s:(EVERYONE@): a rwpRxadcs
s:(EVERYONE@): d WDACo
Applying ls -l to the file shows:

-rwxr-xrwx 1 sally staff 0 Jul 30 11:20 testfile
Initially, you might think that the bits will map straight from OWNER@ to user,
GROUP@ to group, and EVERYONE@ to other. As you can see from the
example, this is not the case. Here you see that the user permissions show as
rwx, where the s:OWNER@:a ACE has none of those bits set. Furthermore, even
though the ls -l output makes it look like the user sally has write access to the
file, she actually does not. Evaluating the ACEs from top down, write access is
denied by the s:GROUP@:d entry.
Based on this, we must draw the conclusion that you cannot use the standard
UNIX permissions bits to reliably predict access when using NFS V4 ACLs.
3.4.5 NFS V4 ACLs: administration

In AIX, NFS V4 ACLs (and AIXC ACLs) can be administered from either the NFS
server or an NFS V4 client via the command line or via the AIX Web-based
System Manager (WSM).

Manipulating ACLs via the command line
ACLs can be administered using the following commands:
aclget Writes the textual representation of an ACL to standard
output or to a named file.
aclput Replaces the contents of an ACL from a textual
representation provided either from standard input or from
a named file.
acledit Retrieves the ACL’s textual representation into a text
editor (specified by the EDITOR environment variable)
and then replaces the ACL from the modified text.
aclconvert Converts an ACL’s format from either AIXC to NFS4 or
from NFS4 to AIXC. The translation is not necessarily
straightforward. Use this with caution, and make sure that
the end result is what you intended.
aclgettypes Returns a list of the ACL types supported by the file
system that contains a given file or directory.
Note: The aclconvert and aclgettypes commands are new in AIX 5L V5.3.
Manipulating ACLs via WSM

You can also use the AIX Web-based System Manager (WSM) to manipulate
ACLs. This may be easier for novice users to grasp, because it is GUI-based. We
think, however, that WSM will be useful for only the most basic operations and
that experienced administrators will mainly use the command line utilities.
Here is an example of how to use WSM to change an ACL:

1. Start up the WSM console and double-click the File Systems icon. Then
double-click the Overview and Tasks icon. You will see a window that looks
like Figure 3-7 on page 74.

Figure 3-7 WSM File Systems → Overview and Tasks screen
2. Choose Access Control List either from the main window or from the
Filesystems menu. This opens a window like Figure 3-8.
Figure 3-8 WSM ACL file or directory name prompt
3. Either: Type in the full path name of the file or directory whose ACL you would
like to modify, or click Browse to choose the file or directory from the GUI.
After you have entered the name, either type Enter or click Next.

4. This opens a window similar to Figure 3-9. Make sure that Edit ACL is
selected and click Next.
Figure 3-9 WSM ACL operation selection

5. The window shown in Figure 3-10 opens. Select the ACE you would like to
change and click Edit.
Figure 3-10 WSM ACL edit screen

6. The ensuing window has two tabs: General and Access Mask. Under the
General tab, you can set the user type and identity, the ACE type, and the
ACE flags (inheritance, audit and alarm). If you select the Access Mask tab,
you will see a window like Figure 3-11.
Figure 3-11 WSM ACE mask screen
7. Make sure that the selected access mask entries are the ones you want and
click OK. You will be returned to the ACL edit screen (Figure 3-10 on
page 76).
8. Repeat steps 5 and 6 for other ACEs that you want to change. After you have
finished with all of the ACEs that you want to change, click OK on the ACL
edit screen. You will see a pop-up window that indicates the status of the
operation. When you are done viewing the status, click Close on that window.
This concludes the WSM example.
Use of the chmod command

When working with files and directories that have NFS V4 ACLs, chmod can only
be used to set UNIX permission bits that are outside of what is stored in the ACL.
These are the setuid bit, the setgid bit, and the sticky bit (Table 3-8 on page 78).

Table 3-8 Chmod operations compatible with NFS V4 ACLs
Chmod command Description
chmod u+s Set or unset the setuid bit.

chmod u-s
chmod g+s Set or unset the setgid bit.

chmod g-s
chmod +t Set or unset the sticky bit.

chmod -t
Important: Using the chmod command to manipulate the rwx permission bits,
either in octal form (for example, 755) or in symbolic form (u+x) replaces the
NFS V4 ACL with an AIXC ACL, wiping out the original permissions that were
on the file or directory.
Never use the octal form of the chmod command if you are using NFS V4
ACLs. Even if you think that you are leaving the rwx bits alone, using the octal
form will replace the NFS V4 ACL with an AIXC ACL.
Note: If you use chmod to manipulate rwx permission bits on an NFS client and
then (again on the client) run aclget on the file, the ACL will still appear to be
an NFS V4 ACL. However, it will be an AIXC ACL on the NFS server. The NFS
protocol translates AIXC ACLs that have extended permissions disabled to
look like NFS V4 ACLs at the client.
ACL inheritance and umask

Does the UNIX umask have any impact on inherited ACL settings when creating
a new file or directory? The answer is no. The umask has no effect on inherited
ACL settings when using NFS V4 ACLs.
For example, if a directory has the following ACL, a file created in that directory
will have the same ACL, even if the umask is set to 777.
*
* ACL_type NFS4
*
*
* Owner: root
* Group: system
*
s:(OWNER@): a rwpRWxDaAdcCs fidi
s:(OWNER@): d o fidi
s:(GROUP@): a rwpRxadcs fidi

s:(GROUP@): d WDACo fidi
s:(EVERYONE@): a rwpRxadcs fidi
s:(EVERYONE@): d WDACo fidi
ACL inheritance and move vs. copy

Table 3-9 describes the impact that the UNIX mv, cp, and cp -p commands have
on a file’s ACL (assuming that the destination file system also supports NFS V4
ACLs).
Table 3-9 UNIX commands and impact on ACLs

Command Resulting file ACL
mv The file retains the same ACL that it had in the original location if the
source and target file systems are the same. If not, then ACL
assignment occurs as per cp below.
cp The file inherits its ACL from the directory where it is being placed, just
as if it were a newly created file.
cp -p The file retains the same ACL it had in the original location.
Directory structure and ACLs

There are many different ways you can choose to organize your data into a
directory structure and implement ACLs to control access to that data. For
example, you might choose from two different methods to control read access to
data:
1. Controlling access at the directory (container) level by maintaining uniform
access permissions on files and subdirectories within a directory
or
2. Leaving access wide open at the directory level and setting unique access
restrictions at each individual file.
Choosing a method depends on your requirements. Each method has its own
characteristics, some of which are as follows.
Characteristics of method #1:

򐂰 It is easier for most people to keep track of permissions when files with like
permissions are grouped together.
򐂰 Permissions can be easily changed via a bulk replacement of the ACLs. (See
“Maintaining an existing directory structure” on page 82.)
򐂰 If a subset of files in a directory need to have their permissions changed, the
files must be moved to a different directory. Directory location changes can be

disruptive to operations. (for example, symbolic links and other path name
references might have to be updated.)
򐂰 A file could inadvertently inherit incorrect permissions if it is placed in the
wrong directory.
򐂰 If a directory has less restrictive access permissions than a parent directory,
an NFS client could possibly mount that directory on a path that has more
open access. A user whose access would normally be blocked by the parent
directory might then be able to access the directory via the mounted path.
Characteristics of method #2:

򐂰 Every file’s permissions can be tailored to its unique access requirements.
򐂰 A file does not have to be moved when its permissions need to be different
than the files around it.
򐂰 It is more difficult to keep track of the different file permissions.
򐂰 It is easy to mistakenly overwrite a file’s permissions via a bulk update, and it
is relatively complicated to detect a mistake and then to restore the correct
permissions after a mistake has happened. (You would have to know via
some external source what the original permissions were.)
򐂰 To prevent inadvertent access, each file would have to start out with the most
restrictive set of permissions (via inheritance), requiring manual intervention
by the user to share that file with others.
Maximizing the benefits of ACL inheritance

If you choose to implement method 1 on page 79, you should organize your
directory structure to maximize the use of ACL inheritance. To do this, carefully
plan out your directory structure so that files and subdirectories with the same
access requirements are collocated under a single parent directory.
This example helps illustrate this:
An organization has three departments: engineering (eng), sales, and human

resources (hr). People from each of the departments are working on two different
projects: projA and projB. For business reasons, the two projects must be entirely
separate, and people working on one project must not be able to access the data
belonging to the other project.
Each department has its own directory for data, and each department creates
separate project directories under its directory. The resulting directory structure
is depicted in Figure 3-12 on page 81.

Figure 3-12 Directory structure that makes poor use of ACL inheritance
This structure has the project directories replicated under each department. If a
permissions change has to be made to one of the projects, those changes must
be made in three different places.
A directory structure that better lends itself to managing the project permissions
is depicted in Figure 3-13.
Figure 3-13 Directory structure that makes better use of ACL inheritance
This structure has a separate projects directory where the permissions for each
special project can be managed in one place. The departments still put their own
non-project-related data under the dept directory.

Maintaining an existing directory structure
Inheritance takes care of setting permissions for newly created files and
directories, but it does not affect permissions for existing files and directories. No
matter how carefully you plan ahead, you will eventually need to change
permissions on an existing directory structure and all of the files it contains.
There are two possible ways to make a large-scale permissions change:

1. Make the change to one file or directory and then propagate the change to
other files and directories by copying the whole ACL from the file or directory
that you already changed.
2. Incrementally change the ACL for every file and directory.
The first method is simpler to implement, but all files and directories being
changed will take on exactly the same permissions, eradicating any variation that
may have existed. This may be a good thing or a bad thing, depending on your
structure. The second method is much more complex to implement, but it does
allow for other differences to exist in the ACLs. This is another example where
carefully planning your directory structure around permissions requirements can
make administration easier.
The rest of this section illustrates possible ways to implement the ACL
propagation method #1 above.
It is possible to propagate an ACL to an entire directory structure using a

combination of aclget and aclput as follows:
aclget dirname | aclput -R dirname
You can use a different source and directory name, or you can specify the same
directory name for both source and destination to copy a directory’s ACL to all of
its descendants (including itself).
Caution: Only use aclput -R on a directory structure that has a uniform

permissions structure. The command will make a wholesale replacement of all
existing ACLs at and below the specified directory. Any variations in ACLs that
previously existed will be lost.
Using the aclget | aclput combination is convenient, but there are drawbacks:
򐂰 If you mistype the name of the source directory, you will completely wipe out
the permissions in the destination directory. This can be remedied quickly by
reissuing the command with the correct source name, but meanwhile you will
have blocked access to any user or application that tries to access the data.
򐂰 The aclput -R command stops at the first error it encounters, leaving the rest
of the files untouched.

The sample script in Example 3-1 addresses these issues. It does not attempt to
run aclput if either the source or destination does not exist, and it runs aclput on
each individual file and directory so that all possible ACL changes will be made.
Example 3-1 Sample script for copying an ACL (with recursive option)
#!/usr/bin/ksh
#
# copy_acl.sh
#
# Copy the ACL for the given source file/directory to other files/directories
#
# Name of this script

scrname=${0##*/}
#
# Functions
#
function usage {
echo "Usage: $scrname [-R] <source> <dest>"
echo " where"
echo " -R indicates a recursive copy"
echo " (copy ACL to all files and directories below and including"
echo " the destination.)"
echo " <source> = the name of the file or directory to copy the ACL from"
echo " <dest> = the name of the file or directory to copy the ACL to"
exit 1
}
if [[ $# -eq 0 ]]
then
usage
fi
#
# Process input parameters
#
if [[ "$1" = "-R" ]]; then

SETSUBTREE="true"
shift
else
SETSUBTREE="false"
fi
if [[ -n "$1" ]]; then

SRC_NAME="$1"
else
usage
fi
if [[ -n "$2" ]]; then

DEST_NAME="$2"
else
usage
fi
#
# Initialize other variables
#
NBERR=0
TMP_ACLFILE="/tmp/.AIXACL_$$"
if [[ -e "${SRC_NAME}" ]]; then

aclget -o "${TMP_ACLFILE}" "${SRC_NAME}"
NBERR=$?
else
echo "Source \"${SRC_NAME}\" does not exist"
NBERR=1
fi
if [[ "${NBERR}" -eq 0 ]]; then

if [[ -e "${DEST_NAME}" ]]; then
if [[ -d "${DEST_NAME}" && "${SETSUBTREE}" = "true" ]]; then
find "${DEST_NAME}" -print | while read NAME
do
aclput -i "${TMP_ACLFILE}" "${NAME}"
(( NBERR += $? ))
ls -dl "${NAME}"
done
else
aclput -i "${TMP_ACLFILE}" "${DEST_NAME}"
(( NBERR += $? ))
ls -dl "${DEST_NAME}"
fi
else
echo "Destination \"${DEST_NAME}\" does not exist"
NBERR=1
fi
fi
rm -f "${TMP_ACLFILE}"
exit ${NBERR}

3.4.6 NFS V4 ACLs: permissions scenarios
The following scenarios help illustrate how NFS V4 ACLs can be applied.
1. Restricting home directory access to just the associated user, and not
allowing users to change permissions and open up their home directories to
others.
2. Making sure that a particular group is denied access to a set of data.
ACL scenario 1: home directories

Users’ home directories can be a collector for all types of data. Users often place
data in their home directory while working on it. The data may have come from a
directory with strict access controls, and the home directory’s permissions should
not allow wider access to that data. One way to manage this is to lock down each
home directory so that only its associated user can access it.
This is difficult to do with standard UNIX permissions. There are two basic
options:
򐂰 Make the user the owner of the directory and allow only owner access.
Because the user owns the directory, he or she can change its permissions,
which we do not want to allow.
򐂰 Create a group for each user, where the user is the only member; make the
home directory owned by root and the user’s group; and allow only owner and
group access.
The user cannot change the directory permissions, but this option requires
maintaining a whole set of groups, one for each user.
This is easier to do with NFS V4 ACLs. Make root the owner of the directory and
add a user ACE to allow the user access to the directory. This is how that ACL
would look:
*
* ACL_type NFS4
*
*
* Owner: root
* Group: system
*
u:sally(sally@nfsdom1): a rwpRWxDaAdcs
u:sally(sally@nfsdom1): d Co
s:(GROUP@): d rwpRWxDaAdcCos fidi
s:(EVERYONE@): d rwpRWxDaAdcCos fidi

(Note that the user ACEs do not have to be inherited because files that are
created below the directory will be owned by the user.)
The user can open up permissions for files and subdirectories that he or she
creates in the directory because the user owns them, but the home directory
itself will still block access to those files.
Note: It might be possible to NFS mount a lower-level directory that has more
open permissions and gain access to those files, but normally the mount
operation is under system administrator control. If mounts are managed
correctly, users will not be able to get directly at lower directories underneath
the home directory.
ACL scenario 2: block a group’s access

If you have two subcontractors working on a project, and you want to make sure
that the subcontractors are not able to access each other’s data, you can do the
following:
Create a group for each subcontractor; put each subcontractor’s data in a

separate directory structure; and put an ACE at the top of the ACL that denies
access to the other subcontractor. (It is important that the ACE be at the top of
the list to prevent other ACEs from allowing access before the subcontractor’s
access is blocked.)
If the groups are company1 and company2, the ACL on company1’s data would
look like this:
*
* ACL_type NFS4
*
*
* Owner: root
* Group: system
*
g:company2(company2@nfsdom1): d rwpRWxDaAdcCos fidi
s:(GROUP@): a rRxadcs fidi
s:(GROUP@): d wpWDACo fidi
s:(EVERYONE@): a rRxadcs fidi
s:(EVERYONE@): d wpWDACo fidi
No matter what the rest of the ACEs are, company2 will be denied access to
company1’s data.

3.4.7 NFS V4 ACLs: NFS V3 clients
It is possible to make an NFS V3 mount of a file system that contains NFS V4
ACLs. This combination exhibits the following behavior:
򐂰 The NFS server will still grant or deny data access based on the NFS V4
ACLs.
򐂰 The NFS client will not be able to view or manipulate the ACLs directly. For
example, an aclget command on the client returns the error:
aclget: The system call does not exist on this system.
򐂰 If the mount is made with the acl option (noacl is the default), the NFS client
will be able to manipulate AIXC ACLs, but not NFS V4 ACLs.
The bottom line: You may have NFS V3 clients mount file systems that use NFS
V4 ACLs. ACL inheritance and evaluation will work normally on the server.
However, do not attempt to manipulate access permissions directly from the NFS
V3 client. Any permissions change at the NFS V3 client will overwrite the NFS V4
ACL with an AIXC ACL.
Unfortunately, there is no good way to block a user on the NFS V3 client from
running chmod or aclput on files or directories that he or she owns. You will have
to publish policy and rely on well-behaved users. (You could completely disable
the chmod and aclput commands on the client, but that would also disable them
for other client file systems where using those commands is perfectly valid.)
Also keep in mind when using NFS V3 clients that the UIDs and GIDs have to
match between server and client.
3.5 NFS V4 host identification

We discuss two forms of host identification: basic identification using IP address
and host name, and Kerberos identification by machine principal.
3.5.1 Basic host identification

An NFS V4 server identifies client hosts by the IP address given in the RPC
packets. The NFS server turns this IP address into a host name by way of the
host resolver, which can get its information from the Domain Name Service
(DNS), Network Information Services (NIS), or the local /etc/hosts file.

3.5.2 Kerberos host identification
Kerberos authentication uses a unique identifier called a machine principal to
identify hosts. The machine principal is established when configuring a host into
a Kerberos realm. The machine principal name is the fully qualified host name
prefixed with host/ (for example, host/nfs402.itsc.austin.ibm.com).
Another way that Kerberos indirectly identifies a host is through the NFS service
principal. (This is the identification of the NFS service running on the host.) The
service principal name is the fully qualified host name prefixed with nfs/ (as in
nfs/nfs402.itsc.austin.ibm.com). NFS clients using Kerberos authentication
identify NFS servers with this service principal.
3.6 NFS V4 host authentication

NFS servers always identify client hosts by IP addresses and host names,
regardless of the authentication method used. The value added with NFS V4 is
that when Kerberos authentication is the only allowed security method for an
exported directory (see the host authorization section next), the NFS client
session must be properly authenticated before gaining access to any of the data
in that directory.
Think of NFS V4 authentication of clients being mostly at the user level rather
than at the host level. (See 3.3.2, “RPCSEC_GSS user authentication using
Kerberos” on page 59.)
Kerberos does authenticate NFS server identities to the clients via the NFS
service principal. (See the NFS service principal discussion in the previous
section.)
3.7 NFS V4 host authorization

Host authorization in an NFS context means controlling which NFS client hosts
can mount exported directories from the NFS server. This is accomplished in AIX
with a combination of the /etc/exports file and the exportfs command.
Exporting directories from an NFS server is still fundamentally the same in NFS
V4 as it was in NFS V3. The main difference is that NFS V4 has added the new
security-related options shown in Table 3-10 on page 89.

Table 3-10 New /etc/exports security options for NFS V4
Option Description
vers Controls which version NFS mounts are allowed. Possible values are 2, 3,
and 4. Versions 2 and 3 cannot be enforced separately. Specifying Version 2
or 3 allows access by clients using either NFS protocol versions 2 or 3.
Version 4 can be specified independently and must be specified to allow
access by clients using Version 4 protocol. The default is 2 and 3.
sec Controls which security methods are allowed.

Possible values are:
sys (UNIX authentication)
dh (DES authentication)
krb5 (Kerberos, authentication only)
krb5i (Kerberos, authentication, and integrity)
krb5p (Kerberos, authentication, integrity, and privacy)
none (Allow mount requests to proceed with anonymous credentials if the
mount request uses an authentication flavor not specified in the export.
Otherwise a weak auth error is returned. By default, all flavors are allowed.)
In the absence of any sec option, sys (UNIX authentication) is assumed.
The sec option is unique in that it can appear more than once in the exports
definition for a directory. This allows different ro, rw, root, and access options to
be specified for the different security options. For example, hosts using the sys
security method might only be allowed read access, while hosts using the krb5
security method might be allowed read and write access.
Note: You cannot specify the same security option in more than one sec=
stanza in a single exports definition.
Here is a sample /etc/exports line with the new NFS V4 security options:
/exports -vers=3:4,sec=krb5:krb5i:krb5p,rw,sec=sys:none,ro
For more about exporting directories, see the following publications:

򐂰 AIX 5L Version 5.3 System Management Guide: Communications and
Networks, SC23-4909
򐂰 The exportfs command in AIX 5L Version 5.3 Commands Reference
򐂰 The /etc/exports file in AIX 5L Version 5.3 Files Reference, SC23-4895

Part 2
Part 2 Implementing
ing
NFS V4
In this part we introduce you to the planning and implementation methodologies
that were used while preparing this book.

4
Chapter 4. Planning for NFS V4

The most crucial part of any new software and hardware implementation project
is the planning phase. If we get this part of the project right, then the design and
implementation phases become easier. Therefore, in this chapter we discuss
and provide detailed information about planning a reliable infrastructure for NFS
V4 deployment. The underlying concepts are complex, but it is easy to plan,
design, and implement when the concepts are well-understood.
Considerations that must be taken into account when beginning any planning
phase include:
򐂰 Currently available and deployed hardware, software, and applications
Consider the following questions:
– Do you have an inventory of your currents assets and their usage?
– Do you have a logical overview of your infrastructure?
򐂰 The future IT strategies of your company
– Do you want to have centralized user management?
򐂰 Business-driven design and implementation issues
– Do you have the need to exchange data with other customers or
departments?
– Do you have other (third-party) applications to be considered?

Important: Careful planning for your NFS V4 deployment is crucial if you want
to properly use the enhanced security features provided by Kerberos
Authentication as well as NFS V4 ACLs.
We also look at the necessary planning requirements if you want to use NFS V4
without some of the enhanced features, simply as a replacement to your current
NFS V3 environment.
This chapter discusses the following:

򐂰 Deployment of NFS V4 in general
򐂰 Mandatory requirements
򐂰 Identification methods
򐂰 NFS Authentication methods
򐂰 Authorization methods
򐂰 Choosing the appropriate file system types
򐂰 NFS protocols and namespace considerations
򐂰 Sizing and capacity planning considerations
򐂰 Migration considerations

4.1 Deployment of NFS V4 in general
Deployment of NFS V4 should be planned in such a way that a smooth
implementation and integration into the existing infrastructure can be achieved.
Normally, this is done by:
1. Evaluating the basic conditions for deployment.
2. Defining the architectural design and logical layout.
3. Integrating the design in a sample environment.
4. Defining how the rollout will take place into your infrastructure.
Important: With NFS V4, the following two general design constraints should
be taken into account while planning the deployment. These two topics reach
a level of primary importance because of functional changes introduced by the
NFS V4 standard:
򐂰 NFS domain and Identity Mapping
򐂰 Authentication
We have attempted to develop a logical approach to the planning process for

implementing NFS V4. Figure 4-1 on page 96 shows a planning flow chart that
we developed and used to build our test environment and to evaluate different
migration scenarios. We recommend that you use this chart as a template when
approaching your infrastructure planning.
For this redbook project, we did not have an existing environment to migrate
from, so we planned and built one from scratch using the flow chart. However,
most of you will be looking at this from an existing systems migration point of
view. Some considerations in using the flow chart:
򐂰 The flow chart is designed to be modular. So, based on what you have
already implemented in your existing environment, you can skip to the next
step on the flow chart and continue with the planning.
򐂰 It was impractical to cover all possible permutations when considering a
migration path, so in 4.9, “Migration considerations” on page 116, we walk
you through some of the scenarios we consider to be most common.
򐂰 Certain parts of the flow chart may not easily plug into your environment, as
they are new concepts introduced by NFS V4. On the whole, the flow chart
should serve as a good building block for your planning considerations.
Chapter 4. Planning for NFS V4 95

Figure 4-1 Flow chart: implementation process decisions

4.2 Mandatory requirements
The “What is your name resolution type?” and “Choose your NFS Domain”
questions in the flow chart (Figure 4-1 on page 96) are two areas that dictate
how you plan and implement your design.
4.2.1 What is your name resolution type?

Whatever name resolution option you choose (/etc/hosts or DNS), it must be in
100% working order. Name resolution plays a major part throughout the planning
and implementation phase.
NFS V4–related name resolution

NFS V4 does not support referencing hosts by IP address. If applying NFS
export restrictions based on client host names is desired, proper setup of host
name resolution is essential for correct operation. Therefore, you should observe
the following guidelines to avoid most name-resolution-related problems:
򐂰 Ensure valid forward and reverse name mappings for all systems on your
network. The machines that will participate in your NFS domain or Kerberos
realm must have resolvable name entries, forward and reverse. This means
that if you are using DNS, an entry for all participating machines must exist
that maps the host name to an IP address, and a reverse entry must exist for
that IP address mapping it back to the original host name. If you are using
/etc/hosts, the same applies, but you also have to ensure that all participating
machines have the same common information within /etc/hosts file.
򐂰 Ensure fully qualified domain name (FQDN) to host name mapping. What a
system considers its host name is also important. When NFS V4 or Kerberos
creates the name of the host principal, it uses the output from the
gethostname() call to resolve the so-called canonical name (CNAME) and
generates the principal name. If this host principal does not match the host
principal that is held in the KDC’s database, then the application will fail,
saying it cannot find the principal.
For example, if a host resolves the CNAME bigserv for its host name
bigserver, yet its host principal is host/[email protected], it will
attempt to find a host principal of host/[email protected] and fail.
򐂰 Ensure that both the localhost and loopback interface names (by default, IP
address 127.0.0.1) can be resolved in any case.
You may take RFC1537, Common DNS Data File Configuration Errors, and
RFC2181, Clarifications to the DNS Specification into account while planning
your name resolution infrastructure.

Tip: Using the /etc/hosts file as the only form of name resolution on all
systems will work, but using a DNS server provides a more reliable and easily
managed solution to your name resolution. This gives you a single point to
manage your name resolution rather than having to add or change entries on
each host in your network.
Kerberos-related name resolution

Traditional UNIX Kerberos V5 implementations use a flat file (/etc/krb5/krb5.conf
or /etc/krb5.conf) for hostname-to-Kerberos realm mapping. This is similar to the
/etc/hosts file being used for the name-to-ipaddress mapping and vice versa. The
Kerberos configuration file contains two major pieces of information.
򐂰 The DNS domain to Kerberos realm mappings
򐂰 A list of KDCs for each Kerberos realm
Obviously, this method does not scale, so just as DNS now serves the purpose
of the old /etc/hosts file, DNS can also be used to provide Kerberos
configuration.
Kerberos can use DNS as a service location protocol by using the DNS SRV
record as defined in RFC2052. In addition, Kerberos can use a TXT record to
locate the appropriate realm for a given host or domain name. These DNS
entries are not required to run a Kerberos realm, but they do eliminate the need
for manual configuration of clients. With these DNS records, Kerberos clients can
find the appropriate KDCs without the use of a configuration file.
Windows will establish the necessary SRV records automatically when an Active
Directory domain is created. Those using UNIX for their KDCs can create these
DNS entries manually in their zone files as a convenience to the DNS clients.
4.2.2 Choosing your NFS domain

From a traditional point of view, sharing data between servers and clients is done
by the use of NFS. All features of NFS V3 are fully supported by the NFS
implementation in AIX 5.3.
If you do not wish to take advantage of the security enhancements offered by

NFS V4, very little has to be installed or implemented. The changes that do have
to be made are as follows:
򐂰 NFS V4 requires that the NFS domain be set on all servers and clients.
򐂰 NFS V4 changes the way users and groups are looked at. Previous versions
of NFS used UIDs and GIDs. NFS V4 now changes this to user@domain and
group@domain.

If you have not already done so, we recommend that you take the time to read
through Chapter 2, “What’s new in NFS V4?” on page 11 and Chapter 3,
“Enhanced security in NFS V4” on page 45 for a better understanding of what
has changed in NFS V4 and the new security enhancements offered. The topics
that are discussed from here on require you to have a basic understanding of the
new concepts being brought in by NFS V4. The new protocol requires a major
change in your mind set; you can no longer think along the lines of NFS V2 and
NFS V3. We cannot stress this enough!
The NFS domain name is:

򐂰 By the standard defined by the RFC, bound to the DNS domain name.
򐂰 Case insensitive, upper-case characters will be treated as lower case during
runtime.
򐂰 Upper-case characters will be converted automatically to lower case when
using the preferred method to change or set the NFS domain: the chnfsdom
<NFS Domain Name> command.
We recommend that you do not edit the /etc/nfs/local_domain file manually
but if you do so and use upper-case characters, the chnfsdom command will
show upper-case characters.
Important: The AIX implementation does not require the NFS domain to
match your DNS domain. You can call your NFS domain pretty much anything
you like, but keeping a relationship to your DNS domain simplifies managing
the environment. It also helps make sure that the name is unique.
For simplified management, we logically linked our NFS domain to the DNS
domain. In our sample environment our DNS domain was:
itsc.austin.ibm.com
Therefore our NFS domain became:

itsc.austin.ibm.com
4.3 Identification methods

For a fully supported login process in an AIX environment, you need both user
identification and user authentication. In this section, we describe some details
you should take into account while planning user identification in your
environment. See 3.2, “NFS V4 user/group identification” on page 48 for general
information on user identification.

4.3.1 Selecting the user/group repository
User identification comprises all the necessary information about what user IDs
exist and what the attributes are for these user IDs. This information must be
consistent so that you can accommodate the following considerations:
򐂰 Management of identities
How do you intend to manage creation, modification, and deletion of users
and groups?
򐂰 Flexibility of the end user
Will end users each use only one system, or will they also log on to other
systems using the same user name and password combination?
򐂰 Scalability / availability
How many NFS clients will you have? What happens if you have to integrate
new systems into your environment? How much server downtime can you
tolerate?
򐂰 Scope of data sharing
How widely will end users share data: only within each department, or
throughout the entire organization?
򐂰 Access control
Access control decisions are based on user and group identity. To properly
control access to a resource, you must make sure that all accessing entities
are uniquely identified. With NFS V4, the identifiers are strings:
user@nfs_domain or group@nfs_domain. If you also need to support NFS
V3 clients in your environment, you will also need to make sure that the
numeric UIDs and GIDs are unique.
User and group identities are maintained in some kind of repository. This
repository typically is implemented through one of the following methods:
򐂰 Standard UNIX /etc/passwd and /etc/group files
򐂰 Network Information Services (NIS)
򐂰 Lightweight Directory Access Protocol (LDAP)
Use the following guidelines when choosing which method to use for your
user/group repository.
Using /etc/passwd and /etc/group might be desirable if each user uses only one
system. You will need to somehow maintain a central clearing house for user
identities to make sure that you choose names and IDs that are unique.
If users log into multiple systems using the same user name and password, you
will need to replicate your /etc/passwd and /etc/group information to all

participating systems. This becomes very difficult to manage when the number of
systems is more than just a few. In this situation, you will probably want to use a
centrally managed repository such as NIS or LDAP.
If you are using Kerberos authentication with NFS, we recommend that you
choose LDAP, using the schema defined in RFC2307 as your user/group
repository. LDAP supports Kerberos integrated logon, but NIS does not.
Note: The RFC2307 schema enables NIS maps to be imported into an LDAP
directory. If your existing infrastructure uses NIS to manage user and group
information, you may want to consider migrating to LDAP. After the NIS maps
have been migrated, AIX 5L and other RFC2307 compliant platforms can use
LDAP instead of NIS to access this information.
For further information on how this can be achieved we refer you to the
following Technote: AIX - Migrating NIS Maps into LDAP, TIPS-0123, at:
http://www.redbooks.ibm.com/abstracts/tips0123.html?Open
You can also use a combination of methods for your user/group repository. For
example, if your security policy does not allow the same password to be used on
multiple systems, you could maintain the password locally and still use a central
repository such as NIS or LDAP for the rest of the user and group information.
4.3.2 Other identification considerations

If your organizational needs require that you operate across different NFS
domains for data access, you must use identity mapping to map users and
groups from one NFS domain to the other. Enterprise Identity Mapping (EIM)
from IBM provides a way to manage and perform this cross-domain mapping.
Note: Use of EIM requires LDAP.
For more information about EIM, refer to the AIX 5L version 5.3 Security Guide
and the IBM Redbook Windows-based Single Signon and the EIM Framework on
the IBM ^ iSeries Server, SG24-6975.

4.4 NFS Authentication methods
Before planning the NFS V4 authentication method used in your environment, we
suggest reading 3.3, “NFS V4 user authentication” on page 59. With NFS V4,
you have the following choice:
򐂰 AUTH_SYS
򐂰 Kerberos
4.4.1 AUTH_SYS method

By default, NFS uses the AUTH_SYS method to authenticate user identities.
Under the AUTH_SYS security flavor, the user is authenticated at the client,
usually via a logon name and password. The NFS server trusts the user and
group identities presented by its clients.
For example, if tight physical security cannot be maintained on a network, it

would be easy for someone to bring in a Linux laptop, set the IP address to be
the same as a valid NFS client, and connect the laptop to the network using the
valid client’s connection. When connected, the person can set up any user
account desired on the laptop. That way, someone who knows what they are
doing can gain access to any of the exported data on an NFS server.
If someone gains administrative control of an NFS client, or has control of a

machine pretending to be a valid NFS client via IP address spoofing, it is easy to
masquerade as any valid NFS user.
Because of this vulnerability, you should not use AUTH_SYS user authentication
if controlling access to your data is important.
4.4.2 Deploying Kerberos

To take advantage of features offered by NFS V4, we operate under a Kerberos
authentication environment. If your organization has a Kerberos infrastructure in
place, you can use the existing KDC. This can be done only if the installed
version is compatible with the software requirements for NFS V4. IBM provides a
detailed migration path for customers who have previous versions of Network
Authentication Services (NAS). If you plan to set up a new Kerberos
infrastructure to support NFS V4, we provide some guidelines in this section.
See Appendix A, “Kerberos” on page 243 for more information. Also, detailed
information can be found in IBM Network Authentication Service Version 1.4 for
AIX, Linux, and Solaris Administrator’s and User’s Guide. This document is
made available with fileset krb5.doc.en_US on the AIX 5.3 Expansion CD.

The need for Kerberos
Kerberos has grown out of the need for a secure protocol that enables users to
identify themselves to applications and services. Normally, information is
transmitted across the network in an insecure form, which is usually just plain
readable text. Kerberos removes many of the risks by encrypting passwords,
reducing or eliminating the use of passwords, and providing optional methods to
validate data integrity and encrypt data.
The main components of the Kerberos protocol implementation are:

򐂰 Key Distribution Center (KDC)
򐂰 Principals and realms
򐂰 Tickets
򐂰 Services
In addition to data encryption and integrity checking, you can use NFS V4 with
Kerberos authentication to help control which clients can access exported
directories on an NFS server. We demonstrate this in Figure 4-2.
Figure 4-2 NFS V4 Kerberos authentication on system level

In Figure 4-2 on page 103, clientB cannot mount the exported file system
because the client is not:
򐂰 In the Kerberos realm
򐂰 Authenticated to Kerberos
This authentication checking was not possible with NFS V3, and to achieve a
similar control over mounts it is very common to use the exportfs option
-access=Client[:Client] within the file /etc/exports. Sometimes the list of clients
becomes very large and the server exports list must be changed every time a
client is added, deleted, or renamed.
With NFS V4 and Kerberos authentication, with the option sec=krb5, the client is
required to be authenticated before contacting the NFS V4 server. This means
that even if clientA resides in the same Kerberos realm as the NFS V4 Server, a
mount request will fail if clientA itself or the user is not authenticated.
KDC considerations
Authentication requests to the KDC can be easily handled with today’s available
processors, therefore a single or dual processor machine should suffice for
thousands of clients.
Note: This applies only to UNIX-based systems running a KDC. Windows

domain controllers function as much more than just a Kerberos KDC and may
have a higher server-to-client ratio than a dedicated UNIX KDC.
The KDC server will be one of the most important servers in your network, so:
򐂰 The system should be running 24x7.
򐂰 Planning for disaster recovery should be taken into account. Your KDC
database should be replicated if possible.
򐂰 The KDC server should ideally not be used for any other purpose because if
the KDC is compromised, all Kerberos principals are compromised.
System time synchronization

Kerberos requires that the system time be reasonably close throughout the
network (within five minutes by default). Before beginning with the
implementation, you should set up a mechanism to automatically synchronize
time throughout the network between the servers and the clients. The following
are possible methods for synchronizing time:
򐂰 AIX timed daemon.
򐂰 NTP (Network Time Protocol, official home page at: http://www.ntp.org/)

򐂰 In a small business network, it should suffice to use the AIX setclock
command, scheduled through a crontab entry, to synchronize the system time
between the client and KDC server.
򐂰 On Windows OSs, use of NTP on the server and client is also supported.
Note: The default value for maximum clock skew is 300 seconds (five
minutes). For security reasons, we recommend that you not change this value.
Multi-homed servers
Multi-homed servers use more than one physical network interface with different
IP addresses to serve clients. There are several reasons why you would want to
do this, including:
򐂰 Subnetworks
򐂰 Load balancing
The KDC server can only be bound to one IP address. This is defined during the
setup of the KDC by providing the system host name.
Therefore, mapping between the client and server network interfaces as well as
the KDC bound network interface has to be planned and managed.
Network infrastructure
You should take into consideration not only how many authentication clients you
will be serving, but also where these clients are located. While the bandwidth
requirements for Kerberos authentication are miniscule, the important metric for
Kerberos performance is the network latency between clients and the Kerberos
KDC. Each authentication exchange requires time for at least one full round trip
between client and KDC. Users’ authentication requests will become noticeably
slow if this latency is long, for reasons such as:
򐂰 Going through a satellite uplink
򐂰 Traversing across DSL connected backbone
Consequently, you should position your KDCs so that they are as close to the
clients’ network as possible. To support geographically dispersed networks with
possibly different types of connections, Kerberos implementations such as IBM
Network Authentication Service are capable of using the replication mechanism
to set up a KDC server and propagate the KDC database as needed.
Principals and realms

To understand how principal names are generated in Kerberos, we must first
understand Kerberos realms. A Kerberos realm is often referred to as an
administrative domain. A realm consists of members, which can be users,

servers, services, or network resources that are registered within a KDC’s
database. Each of these members has a unique identifier that is called a
principal. The Kerberos realm is made up of the KDC and all of its principals.
Realm naming considerations

Although the realm can be any ASCII string, conventionally the realm name is
the same as the domain name, in uppercase letters. If multiple realms are
needed, use descriptive names that end with the domain name, such as
CARPROJECT.AUSTIN.IBM.COM or POWERTRAIN.AUSTIN.IBM.COM.
Important: The normal conventions of naming Kerberos realms is to create

them using upper case.
In our environment we created two realms:

REALM1.IBM.COM
and
REALM2.IBM.COM
Principal naming considerations

By default, every entity contained within a Kerberos installation, including
individual users, computers, and services running on servers, has a principal
associated with it. The principal is a unique identifier to which KDC assigns
tickets. There are three parts of a principal name: the primary, the instance
(optional) and the realm. The primary component of the principal name is
separated from the instance by a forward slash ( / ) The realm is separated from
the rest of the principal name by the at sign ( @ ):
򐂰 In the case of a host principal, the user name is host/<hostname>@REALM
򐂰 For an NFS V4 service principal, it is nfs/<hostname>@REALM
򐂰 For a user, it is <username>@REALM
Because the principal is a unique identifier, you should plan a strategy for
principal naming throughout your Kerberos environment. This is especially true
for the NFS service principal names, because NFS V4 must have proper name
resolution in place to compile the ticket request. See “NFS V4–related name
resolution” on page 97.
Tip: Always use the FQDN for the NFS service principal.
We decided that all NFS V4 Server and NFS V4 full clients in our environment
would have the following service principal name:
nfs/<hostname>.itsc.austin.ibm.com@REALM

User principal names can follow your current naming convention for user logon
names. As of today, you should restrict principal names to the printable portion of
the POSIX portable character set, which is equivalent to 7-bit ASCII.
4.4.3 Default types of encryption for KDC and security flavors

Several types of encryption are available for use with Network Authentication
Service; by default, Triple-DES is used. Multiple encryption types can be used
within a Kerberos realm. The NFS client and server will negotiate which of the
supported encryption types to use.
NAS 1.4 supports the following types of encryption:

򐂰 aes128–cts, aes256–cts, des–cbc–crc, des–cbc–md4, des–cbc–md5,
des3–cbc–sha1, arcfour–hmac, arcfour–hmac–exp
However, the NFS V4 implementation on AIX 5.3 supports only the following
types of encryption:
򐂰 des–cbc–crc, des–cbc–md4, des–cbc–md5, des3–cbc–sha1
For the best performance and interoperability, we recommend that you use
Single-DES as the standard encryption type on your installation. This may
depend, of course, on your overall security strategy. If protecting packet privacy
is critical, you might need to stay with Triple-DES encryption.
In addition to the chosen standard Kerberos encryption, you can select between
the following security flavors, which are options used during exporting file
systems and depend on your security needs:
򐂰 krb5
Use Kerberos V5 protocol to authenticate users before granting access to the
shared file system.
Note: We only used krb5, and we recommend that you start with this.
򐂰 krb5i
Use Kerberos V5 authentication with integrity checking (checksums) to verify
that the data has not been tampered with in transmission.
򐂰 krb5p
Use Kerberos V5 authentication, integrity checksums, and privacy protection
(encryption) on the shared file data. This provides the most secure file
sharing, as all traffic is encrypted.
Note that each increasing level of protection carries with it a performance

penalty.

4.4.4 NFS client considerations when using Kerberos
This section discusses considerations you should take into account while
planning and implementing your NFS V4 clients.
NFS V4 slim client versus full client

NFS V4 introduces the concept of a full client and a slim client similar to what is
already used in the Distributed Computing Environment (DCE Version 3). This
concept does not necessarily apply to other applications that use Kerberos
authentication.
By default, every entity contained within a Kerberos installation (including

individual users, computers, and services running on computers) has a principal
associated with it. In the case of a host principal (machine principal), the principal
name is host/<hostname>@REALM; for an NFS service principal, it is
nfs/<hostname>@REALM.
In the AIX 5.3’s NFS V4 implementation, two types of Kerberos clients are
defined: slim clients and full clients.
򐂰 An NFS client without a dedicated NFS service principal is called a slim client.
򐂰 An NFS client with a dedicated NFS service principal in the form
nfs/<hostname>@REALM is called a full client. The full client provides
stronger security. However, it is the stronger Kerberos-based RPC security
that this type of client provides that requires more administrative overhead:
You should run the config.krb5 command on each client with the Kerberos
Administrative ID.
If you trust all systems connected to your internal network and authentication at a
user level fulfills your needs, then using slim clients would suffice. The following
reasons are why you would want to deploy slim clients in your infrastructure:
򐂰 Pre-installed new systems
򐂰 Mass rollout of new systems
򐂰 Unprompted upgrade of clients
򐂰 Share all the same Kerberos realm as well as NFS domain
In addition, there are several choices for installing the slim clients, and these are
based on the way your systems are installed:
򐂰 Complete new installation (scratch installation)
򐂰 Cloning by use of a full system backup image

Attention: As this book is being written, a slim client is not capable of
automatically mounting file systems that require RPCSEC_GSS authorization
only. To enable a slim client to automatically mount file systems at boot time,
the server exports definition must include AUTH_SYS in the first sec= stanza.
Kerberos integrated logon

User authentication at a system level is a process where a user claims to have a
certain identity and the system has to check whether this is true. For this
process, the system requires a unique piece of information about this user
(usually a password). When users authenticate, the system challenges them by
requesting that they type in their password. The user’s response is then
compared to the stored unique piece of information, and the request is accepted
or denied depending on the outcome of this comparison.
AIX 5L supports load modules that are responsible for identification, for
authentication, or both. You can use either one load module, which supports
both, or you can specify one load module that is responsible for the identification
part, and another that is responsible for authentication. Such a combination of
two modules is called a compound module.
We recommend that you use a compound module comprised of NAS 1.4 for
authentication and LDAP RFC2307 for identification. With NAS and LDAP in
place and KRB5LDAP as the authentication module on the clients, you do not
have to deal with creating users on every client.
For more about configuring the system to use these authentication modules, see
5.9.4, “Configure the NFS V4 client for integrated login services” on page 170.
For more details about using AIX directory integration to support user
authentication and identification, refer to the Security chapter in the redbook AIX
5L Version 5.3 Differences Guide, SG24-7463.
4.4.5 Deployment of LDAP

LDAP can be used as a backend system to provide additional user identification
information that is not covered by the KDC environment. Because the NFS V4
implementation of RPCSEC_GSS is based on the MIT Kerberos Standard
RFC1510, all compliant third-party LDAP servers can be substituted for IBM
Tivoli Directory Server.

4.5 Authorization methods
Authorization is used to control access to either client hosts or client users.
There is only one way to control access to client hosts: via the /etc/exports file
and the exportfs command. This is described in 3.7, “NFS V4 host
authorization” on page 88.
You have three options when it comes to controlling client user access to files
and directories:
򐂰 Standard UNIX permissions
򐂰 AIXC ACLs
򐂰 NFS V4 ACLs
4.5.1 Choosing your user authorization method

When deciding which user authorization method to implement, think about your
requirements for controlling access to data. Do you have simple or complex data
access requirements?
Standard UNIX permissions enable you to control access to only three identities:
the owning user, the owning group, and everyone else. If that is not sufficient to
meet your access control requirements, then you should choose one of the ACL
options. For example, if you have data where you need one group to have write
access, one or more other groups to have read-only access, and everyone else
to have no access at all, you will not be able to accomplish this using standard
UNIX permissions.
If standard UNIX permissions do not meet your requirements, you can then
choose to use AIXC ACLs or NFS V4 ACLs. If you choose NFS V4 ACLs, then
make sure that you choose file system types on your server that support this.
See 4.6, “Choosing the appropriate file system types” on page 111 for more
information.
You should not use AIXC ACLs if your requirements include one of the following:
򐂰 You have non-AIX NFS clients that must be able to manipulate ACLs for data
on your NFS server. AIXC ACLs are only supported in AIX.
򐂰 You require finer granularity access control than AIXC ACLs support. For
example, AIXC ACLs do not provide a way for you to set up a directory where
users can create files but not delete them after they are created.

Note: The choice of AIXC ACLs over NFS V4 ACLs should make sense in
environments that already have experience with AIX ACL models and are
running predominantly AIX platforms. Otherwise, use of NFS V4 ACLs is
recommended as the default choice due to its compliance with open
standards and to minimize potential interoperability issues with non-AIX
platforms.
More details about permissions and ACLs can be found in 3.4, “NFS V4 user
authorization” on page 62.
4.5.2 Other user authorization considerations

Maintaining access control lists is more complicated than maintaining standard
UNIX permissions. You must understand how the ACLs are evaluated before you
can correctly construct an ACL. (See 3.4.4, “NFS V4 ACLs: ACL evaluation” on
page 69 for information about how NFS V4 ACLs are evaluated.)
Depending on how computer-savvy your users are, you will need to either
educate them on how to properly maintain their own ACLs or set up ACL
inheritance so that they do not have to worry about changing permissions. (See
“Directory structure and ACLs” on page 79 for more about setting up ACL
inheritance.)
What if you do not want to allow end users to change permissions, thus keeping
that task under the control of designated system or data administrators? The
system by default does not accommodate this. Files created by a user are owned
by that user, and a file’s owner can always change its permissions. You would
need to devise some way to make sure that all files are owned by an
administrator account. One way is to run a periodic cron job that changes
ownership of all files to that account. This leaves a window of time when a user
can change permissions on a newly created file, but this can be remedied by the
job that changes ownership. It can also make sure that the permissions are set
as they should be after it changes ownership.
You can see from this discussion that you may need to implement additional
administrative controls on top of those that are provided by default.
4.6 Choosing the appropriate file system types

AIX supports several different file system types, including:
򐂰 Journaled File System (JFS)
򐂰 Enhanced Journaled File System (JFS2)

򐂰 JFS2 with extended attribute format Version 2 (EAv2)
򐂰 General Parallel File System (GPFS)
The file system you can choose will be dictated by the authorization method you
chose above. If you want to use NFS V4 ACLs, then your choice should be either
JFS2 EAv2 or GPFS. This does not mean that your system cannot use the other
file system types. The restriction only applies to the file systems where you want
to use NFS V4 ACLs.
4.7 NFS protocols and namespace considerations

At this decision point, we are in the final planning stages. We now need to make
decisions about whether we will be running in a mixed environment (NFS V3 and
NFS V4) or creating a pure NFS V4 environment.
Namespace is a new concept to NFS V4. How do you go about implementing

something like this in a production environment? We begin by considering what it
means. We can then look at some ways that it can fit into your environment.
From an administrative point of view, a single namespace reduces the number of

mounts a client has to make per server down to one. With previous versions of
NFS, you had to mount all the required NFS exports from a server. A single
namespace allows the server to render a single view (pseudo-FS) for the client to
mount, hence the single mount. What does this mean to you? To gain the optimal
benefits of this feature you have to consider your server’s NFS exports layout.
Figure 4-3 on page 113 shows one way the single namespace could be
implemented and how it could benefit you, using one NFS server and five NFS
clients.

Figure 4-3 One NFS server serving five NFS clients
You want to export five file systems from your server. If you use NFS V3, then
your server’s exports file will look something like Example 4-1.
Example 4-1 /etc/exports file on NFS V3 server

/home -rw
/usr/codeshare/ThirdPartyProgs -rw
/usr/local -ro
/var/db2/v81/DB -rw
/exports/scratch -rw
The NFS V3 client would then have to mount each export individually, as shown
in Example 4-2.
Example 4-2 The mount commands you would need to run on each NFS V3 client
mount serverY:/home /mount1
mount serverY:/usr/codeshare/ThirdPartyProgs /mount2
mount serverY:/usr/local /mount3

mount serverY:/var/db2/v81/DB /mount4
mount serverY:/exports/scratch /mount5
The NFS V3 server exported five file systems and the client had to mount five file
systems to five different mount points.
Now we look at how a single namespace would simplify the export and mount
process. We continue to use the directory structure defined in Example 4-1 on
page 113 to carry this out. The server’s /etc/exports file will look something like
Example 4-3.
Example 4-3 Rendering a single view for the NFS V4 client on the NFS V4 server
/exports -nfsroot
/home -vers=4,rw,exname=/exports/home
/usr/codeshare/ThirdPartyProgs -vers=4,rw,exname=/exports/ThirdPartyProgs
/usr/local -vers=4,ro,exname=/exports/local
/var/db2/v81/DB -vers=4,rw,exname=/exports/DB
/exports/scratch -vers=4,rw,exname=/exports/scratch
What have we done to the /etc/exports file to make the single namespace
concept into a practical implementation?
1. We need to set the pseudo-root on the server: This acts as the glue for all
other file systems. We used the /exports directory.
2. We then export all the other file systems, as we did with the NFS V3 exports,
but we add two new options:
vers=4 This tells the NFS server that the export is of type NFS V4.
exname This is the AIX implementation extension of the single
namespace concept. We have taken the file systems that we
want to export and glued them to /exports (defined in the first
line of the /etc/exports file).
If you look at the exname option (for all exports), we have also chosen to not
show the parent directories of all the exported directories. This enables us to
keep the server’s directory structure private and unexposed to the clients.
We will look at this from a practical point after we have shown what the client
must do to see the logical view. The command that the client has to run to mount
the exported file systems is:
# mount -o vers=4 serverY:/ /nfs
The mount command tells serverY that the client is requesting a view of its single
namespace, and after the server allows the client access to the view, the client
will mount it onto the /nfs mount point locally.

This is how the view is represented to the client:
# ls /nfs
DB ThirdPartyProgs home local scratch
We can see that all the file systems we exported on the server are represented
under the client’s mount point (/nfs). If we now decide that we want to move the
/exports/scratch file system, on the server, to a different location, say /scratch,
and we also want to move /home to /users/home, we only have to make the
appropriate changes to the server’s /etc/exports file, and the client’s mount
command will remain the same.
Several other considerations must be taken in to account when allowing

concurrent access to the same data with NFS V2, NFS V3, and NFS V4. NFS V3
access may receive errors due to the NFS V4 granted state. Also, NFS V3
performance may be affected when data is also exported for NFS V4 access (the
use of vers=3:4 in the /etc/exports file).
4.7.1 Pseudo-root FS - alias tree versus classic model

Implementation of pseudo-root FS or the AIX-specific alias tree implementation
must be taken into account before deploying NFS V4. It introduces several
changes compared to the classic model as with NFS V3. There is no need to
migrate off the classic model when using NFS V4. Nevertheless, the alias tree
model seems to be the best from our point of view. See 2.8.6, “External name
space (exname)” on page 34.
While the pseudo-root FS - alias tree model delivers a good solution to build a
Global Namespace, you have to think about the current logical as well as
physical layout of your File systems on the NFS V4 Server. Nevertheless, you
can adopt the pseudo-root FS - alias tree model to run with your current physical
and logical layout.
Important: Either all Version 4 exports specify an external name, or none

specify an external name.
4.8 Sizing and capacity planning considerations

Capacity planning of your infrastructure environment is crucial to your overall
operation. It is not the purpose of this book to discuss this topic. For further
reading see Sizing and Capacity Planning, SG24-7071.

4.9 Migration considerations
As already mentioned, we assume that, in most situations, a version of NFS is
already deployed in your environment and therefore migration has to be taken
into account. In this book, we cannot cover all possible migration paths in terms
of how to migrate your current installation onto AIX 5.3. We have designed logical
migration paths that can be used for migration planning, as shown in Figure 4-4
on page 117.
All logical migration paths have a normal NFS V3 deployment on AIX as the entry
point, and end with deployment of Kerberos-based authentication. In addition,
the identification method can be chosen between classic (using local
/etc/passwd) or centralized, by the use of IBM Tivoli Directory Server or any
other backend system.
Choosing the final Namespace model in terms of using, for example,

pseudo-root FS goes with migration planning, which we do not cover in this book.
This migration will affect your general physical as well as logical file system and
mount structure of your server and client system but can be seen as independent
to your security model on which we are focused.
Figure 4-4 on page 117 shows three possible logical migration paths indicated by
the different numbers and arrows:
Dark arrows Probable migration paths expected in production
environments
Dashed arrows Alternative paths for crossover
Number 1 target Migration to NFS V4 without enhanced security
Number 2 target Migration to NFS V4 with enhanced security using new
authentication infrastructure
Number 3 target Migration to NFS V4 with enhanced security by extending
the available identification infrastructure

Figure 4-4 Possible NFS V4 migration scenarios
What does Can coexist at the top of Figure 4-4 mean?

򐂰 Server file systems can be exported to support NFS V4 and NFS V3 at the
same time using the exportfs option -vers=3:4. This enables the client to
mount the file system using either NFS V3 or NFS V4, and enables you to use
clients that do not support NFS V4 to access the server file systems.
򐂰 With the AIX 5.3 NFS implementation, you can also use enhanced security
(RPCSEC_GSS API) with NFS V3, but only if all servers and clients are
running AIX 5.3.
At this point we discuss only the migration paths marked by the dark arrows.
1. In this case, migration to NFS V4 has been considered using a two-step
approach by migrating all server systems first to AIX 5.3 without having NFS
V4 deployed. After migrating all client systems to AIX 5.3, general
deployment of NFS V4 can take place.

2. If you are able to migrate all of your systems to AIX 5.3, you may plan to
integrate NFS V4 with enhanced security in one step. This implies paying
particular attention when planning your authentication infrastructure.
3. If you already have centralized identification, migration using a two-step
approach is reasonable. The first step introduces authentication services into
your existing identification infrastructure, and step 2 migrates the existing
NFS V3 environment into NFS V4.

5
Chapter 5. Sample implementation

scenarios
This chapter provides practical information about installation, administration, and
deployment of NFS V4 implementation on AIX 5.3. You should get an idea of how
NFS V4 can be implemented in your environment through the step-by-step
scenarios that we used based on our own redbook development systems layout.
The following actions are demonstrated and discussed:

򐂰 General setup
What system level configurations are needed: PATH variables, Application
Event logging and so forth.
򐂰 Deploying NFS V4 in an classic NFS V3 manner without additional security
enabled.
We provide a sample that shows deployment of NFS V3 on AIX 5.3.
򐂰 Setting up a pseudo-root file system and a global filespace
We show how to set up and deploy the pseudo-root FS as well as the
advantages of using the pseudo-root FS - alias tree model to provide a Global
Filespace to the client.

򐂰 NAS Service environment to support the additional NFS V4 security features
This includes the required steps to set up a minimal NAS environment to
deploy Kerberos on the server and client. We also show the steps required to
integrate NFS V4 security flavors.
򐂰 Additional LDAP backend integration into KDC with integrated logon
This section guides you through the necessary steps to set up the
environment to deploy integrated logon on your client system while running a
KDC server with an LDAP backend for identification and authentication.

5.1 Setup of the sample environment
This section describes the setup used on our systems for compiling the samples.
The setup may not adapt to your needs if you do not use NAS, but we believe
that it is useful to provide the information here.
Important: It is important that you do not confuse the term NAS with Network
Attached Storage. In this context NAS stands for Network Authentication
Service.
5.1.1 PATH variable for NAS deployment

After installing the AIX 5.3 Base Operating System, the PATH environment
variable must be adjusted to make the required commands available to the end
user. In addition, correlations with other installed products such as DCE and
Tivoli Netview should be considered while setting the PATH variable systemwide.
The Java14.sdk file set is automatically installed as a co-requisite of

sysmgt.websm.rte 5.3.0.0. You can verify this by running the installp -ugp
Java14.sdk command. The standard PATH variable in /etc/environment is set to:
PATH=/usr/bin:/etc:/usr/sbin:/usr/ucb:/usr/bin/X11:/sbin:/usr/java14/jre/bi
n:/usr/java14/bin:/usr/java131/jre/bin:/usr/java131/bin
We found that the NAS user commands, such as kinit, are not found in the
PATH. Further, kinit is also installed with Java14.sdk.
Example 5-1 Sample output of different kinit versions with different PATH settings
# type kinit
kinit is /usr/java14/jre/bin/kinit
#
# type kinit
kinit is /usr/krb5/bin/kinit
#
Therefore, we changed the PATH variable in /etc/environment to:

PATH=/usr/bin:/etc:/usr/sbin:/usr/ucb:/usr/bin/X11:/sbin:/usr/krb5/bin/:/us
r/java14/jre/bin:/usr/java14/bin:/usr/java131/jre/bin:/usr/java131/bin
We distributed this file to all of our AIX NFS V4 servers and clients.
5.1.2 syslogd settings

By default, all NFS V4 related debug output is written to the syslog.
Chapter 5. Sample implementation scenarios 121

Tip: We recommend that you always enable syslog logging when carrying out
your NFS setup. It will be beneficial in obtaining detailed messages for the
errors you see when running NFS commands. You can disable the logging
when you are satisfied that your environment is working as you expect it to.
To capture errors logged by the NFS daemons, we made the following changes
to our systems:
1. A local file system was created and mounted to the /var/nfsv4log directory. (It
is always better to have a separate file system for log daemons, so that your
root file system is safeguarded from being filled up.)
2. The following stanza was added to the /etc/syslog.conf file:
*.debug /var/nfs4log/syslog.out rotate time 1d archive /var/nfs4log/archive/
3. We then refreshed the syslogd using the refresh -s syslog command to
activate these changes immediately on the running system.
Using the setting for syslogd as given in step 2 may log more information as
needed while running a system in production mode. Thus you can set back the
log level to ERROR in the /etc/syslog.conf file to limit the amount of data written.
The new stanza will now show:
*.error /var/nfs4log/syslog.out rotate time 1d archive /var/nfs4log/archive/
The changes will take affect only after you refresh the syslog daemon.
5.2 Using NFS V4 as you did with NFS V3

NFS V4 still supports the legacy file system export and mount model because it
is common in NFS V3. There are no changes required to your server /etc/exports
file or to the client /etc/filesystems file or automount map. This is true only if you
are not planning to use the new features available with NFS V4.
Example 5-2 shows the server view of NFS V3 exported file systems.
Example 5-2 Server output of NFS V3 exported file systems on AIX 5.3
# pg /etc/exports
/exports -rw
/exports/home -rw
/exports/project/projA -ro
/exports/project/projB -ro
/usr/ldap -vers=3,ro
#
#exportfs -va

Exported /exports
Exported /exports/home
Exported /exports/project/projA
Exported /exports/project/projB
Exported /usr/ldap
#
Example 5-3 shows the client view for the NFS V3 mounted file systems from
server nfs404.
Example 5-3 Client output for NFS V3 mounted file system on AIX 5.3
# showmount -e nfs404
export list for nfs404:
/exports (everyone)
/exports/home (everyone)
/exports/project/projA (everyone)
/exports/project/projB (everyone)
/usr/ldap (everyone)
# mount nfs404:/exports /nfs
# mount nfs404:/exports/home /nfs/home
# mount nfs404:/exports/project/projA /nfs/project/projA
# mount nfs404:/exports/project/projB /nfs/project/projB
# cd /nfs/project/projB
# df .
Filesystem 512-blocks Free %Used Iused %Iused Mounted on
nfs404:/exports/project/projB 65536 64856 2% 5 1%
/nfs/project/projB
5.3 How to unmount an exported NFS V4 file system

With NFS V3 you could unmount any physical file systems NFS exported, on the
NFS server, without first unexporting it. With NFS V4 this is no longer possible.
As long as the file systems are exported they are locked within the server code
and cannot be unmounted. To do so, you have to unexport the file system first.
Example 5-4 Unmount of exported file systems on NFS V4 server

# exportfs
/exports -vers=4,rw
/exports/home -vers=4,rw
/usr/ldap -vers=3,ro
#
# umount /exports/project/projB
umount: 0506-349 Cannot unmount /dev/fslv01: The requested resource is busy.

#
# exportfs -u /exports/project/projB
# umount /exports/project/projB
5.4 Setting up the NFS domain name

As described in 4.2.2, “Choosing your NFS domain” on page 98, it is mandatory
to have the NFS domain name set before you can use NFS V4. The current
setting can be checked using chnfsdom or smitty chnfsdom commands.
If the NFS domain is not set, the output looks similar to Example 5-5.
Example 5-5 chnfsdom sample output

# chnfsdom
Current local domain: N/A
In our sample environment, we already set the NFS domain so that the output
looks like Example 5-6.
Example 5-6 Setting the NFS domain on the server

# chnfsdom
Current local domain:
nfs.ibm.com
Note: As this book is being written, changing the NFS domain does not
recycle or start nfsrgyd. You have to manually start or recycle the daemon.
We started the nfsrgyd by using the command startsrc -s nfsrgyd.
Example 5-7 startup of the nfsrgyd daemon

# startsrc -s nfsrgyd
0513-059 The nfsrgyd Subsystem has been started. Subsystem PID is 14496.
5.5 The pseudo-root FS

The mechanism and meaning of the NFS V4 pseudo-root has been described in
2.6.3, “Better namespace handling” on page 25. This chapter shows how to set
up the pseudo-root and point out the advantages compared to the classic NFS
V3 export model.

5.5.1 Setting up the pseudo-root FS on an NFS V4 server
The pseudo-root FS setup on an NFS V4 server can be achieved easily. This
requires that the pseudo-root directory be available and that all exported file
systems are locally mounted within this directory.
Note: There are several ways to change the pseudo-root on the server. We
recommend using the -nfsroot option within the /etc/exports file; however, all
NFS exported server file systems have to be unexported or this will not work.
By default, the root directory / is defined and exported as the pseudo-root FS

from the server. This can be seen when the nfsd -getnodes command is
executed on the running system. Example 5-8 shows the output.
Example 5-8 Output of command nfsd -getnodes

$nfsd -getnodes
#root:public
/:/
$
Using the -nfsroot option in file /etc/exports

Perform these steps on the server to set the pseudo-root to /exports and export it
to the clients:
1. Check the current settings for pseudo-root by running the nfsd -getnodes
command.
2. Check that no NFS file systems are exported by using the exportfs
command. If there are exported file systems, run the exportfs -ua command
to unexport them.
3. Change the /etc/exports file so that the option /exports -nfsroot is listed in the
first place.
Example 5-9 shows all commands and system responses when changing the
pseudo-root FS using -nfsroot option.
Example 5-9 Changing the pseudo-root on an NFS V4 server using -nfsroot

# nfsd -getnodes
#root:public
/:/
#
# exportfs -ua
# exportfs
exportfs: nothing exported
# pg /etc/exports

#
/exports -nfsroot
#
# exportfs -va
exported /local/trans
exported /local/dept
exported /local/home
exported /usr/codeshare/ThirdPartyProgs
# nfsd -getnodes
#root:public
/exports:/exports
#
#ps -ef |grep nfsd
root 270550 159906 0 17:12:13 - 0:00 /usr/sbin/nfsd -root / 3891
Important: Although the nfsd process still shows the option -root as /
the actual pseudo-root FS has been changed to /exports. This has been
verified by running the nfsd -getnodes command.
Changing the nfsd daemon

Perform these steps on the server to set the pseudo-root to /exports and export it
to the clients:
1. Check the current settings for pseudo-root by running the nfsd -getnodes
command.
2. Check that no NFS file systems are exported by using the exportfs
command. If there are exported file systems, run the exportfs -ua command
to unexport them.
3. Stop all NFS server processes by running /etc/nfs.clean
4. Change the pseudo-root with chnfs -r /exports or smitty chrootfh
The chnfs command:
a. Changes the entries within the AIX Subsystem Resource for nfsd. This
can be verified using the lssrc -Ss nfsd command.
b. Starts the nfsd in command line mode outside the control of the
Subsystem Resource Controller.
We have two options to proceed:
– Reboot the system so that all changes take effect.

– Stop all NFS processes using /etc/nfs.clean and restart NFS using
/etc/rc.nfs.
We decided to use the second option, and we continue with Step 6.
5. Verify the change using the nfsd -getnodes command.
6. Stop all NFS processes on the system using /etc/nfs.clean
7. Start all NFS server processes with /etc/rc.nfs
Example 5-10 shows all commands and system responses when changing the
pseudo-root FS using chnfs.
Example 5-10 Changing the pseudo-root on an NFS V4 server using chnfs

# nfsd -getnodes
#root:public
/:/
#
# exportfs
/exports -vers=4,rw
/exports/project/projA -vers=4,rw
/exports/project/projB -vers=4,rw
#
# exportfs -ua
#
# exportfs
exportfs: 1831-182 nothing exported
#
# /etc/nfs.clean
nfs_clean: Stopping NFS/NIS Daemons
0513-044 The nfsd Subsystem was requested to stop.
0513-044 The biod Subsystem was requested to stop.
0513-044 The rpc.lockd Subsystem was requested to stop.
0513-044 The rpc.statd Subsystem was requested to stop.
0513-044 The rpc.mountd Subsystem was requested to stop.
0513-085 The ypserv Subsystem is not on file.
0513-004 The Subsystem or Group, ypbind, is currently inoperative.
0513-085 The yppasswdd Subsystem is not on file.
0513-085 The ypupdated Subsystem is not on file.
#
# chnfs -r /exports
0513-077 Subsystem has been changed.
#
# nfsd -getnodes
#root:public
/exports:/exports
#
# /etc/nfs.clean
Stopping NFS/NIS Daemons

0513-004 The Subsystem or Group, nfsd, is currently inoperative.
0513-004 The Subsystem or Group, biod, is currently inoperative.
0513-004 The Subsystem or Group, rpc.lockd, is currently inoperative.
0513-004 The Subsystem or Group, rpc.statd, is currently inoperative.
0513-004 The Subsystem or Group, rpc.mountd, is currently inoperative.
#
# /etc/rc.nfs
Starting NFS services:
0513-059 The biod Subsystem has been started. Subsystem PID is 266398.
0513-029 The nfsrgyd Subsystem is already active.
Multiple instances are not supported.
0513-059 The nfsd Subsystem has been started. Subsystem PID is 303296.
0513-059 The rpc.mountd Subsystem has been started. Subsystem PID is 258216.
0513-059 The rpc.statd Subsystem has been started. Subsystem PID is 262314.
0513-059 The rpc.lockd Subsystem has been started. Subsystem PID is 274644.
0513-029 The gssd Subsystem is already active.
Completed NFS services.
#
# ps -ef |grep nfs
root 286874 204944 0 09:48:14 - 0:00 /usr/sbin/nfsrgyd
root 303296 204944 0 09:53:06 - 0:00 /usr/sbin/nfsd -root /exports
3891
root 352494 344286 0 09:53:25 pts/0 0:00 grep nfs
#
# lssrc -g nfs
Subsystem Group PID Status
biod nfs 274482 active
nfsd nfs 266406 active
rpc.mountd nfs 249996 active
nfsrgyd nfs 172234 active
rpc.lockd nfs 254144 active
rpc.statd nfs 245884 active
gssd nfs inoperative
#
Note: The gssd daemon is not running because we are not using
RPCSEC_GSS at this point in our sample environment.
A simple shell script called SetNewRootFS.ksh to achieve the above described

changes is available in “Change the pseudo-root FS sample script” on page 256.
5.5.2 Advantages of using the NFS V4 pseudo-root

As already described in 2.6.3, “Better namespace handling” on page 25, there
are several advantages to using NFS V4 with pseudo-root. This section shows

examples of these advantages. We assume that the NFS V4 server pseudo-root
FS has already been set to /exports.
On the server, we created three file systems (Example 5-11), which were
mounted under the /exports directory using following subdirectories:
򐂰 /exports/home
򐂰 /exports/project/projA
򐂰 /exports/project/projB
Example 5-11 Sample df output from the NFS V4 server

/dev/fslv00 65536 64832 2% 5 1% /exports/project/projA
/dev/fslv01 65536 64864 2% 4 1% /exports/project/projB
/dev/fslv02 65536 64864 2% 4 1% /exports/home
To be able to export these file systems to the clients, we created the /etc/exports
file on the server using smitty mknfsexp with the entries in Example 5-12.
Example 5-12 Output of file /etc/exports for pseudo-root

# pg /etc/exports
/exports -vers=4,rw
/exports/home -vers=4,rw
/exports/project/projB -vers=4,ro
In Example 5-12, the entries are similar to the /etc/exports file used with NFS V3
except for the fact that option vers=4 indicates that the export is of type NFS V4
only. In addition, you still have to export every local file system so that the NFS
server is capable of internally rendering and crossing the file system borders.
The main changes are seen on the client side. They are:
1. Only a single mount command is required enable retrieval and changing of
directory to all available file systems.
2. With NFS V3, the same functionality is not available, so to mount all file
systems, a mount command for each exported file system is required.
In our example, only the following command is needed on the client to make the
pseudo-root FS and all exported file systems beneath available to the client:
mount -o vers=4 nfs404:/ /nfs
On traversal of the mount, for example, from /nfs to /nfs/project/projA, no

subsequent mounts are required as NFS V4 will handle this by use of the fsid().

Note: This works only with NFS V4 exported file systems. If you use NFS V3
by default on the client, the mount command will fail with this message:
mount nfs404:/ /nfs
mount: 1831-011 access denied for nfs404:/
mount: 1831-008 giving up on:
nfs404:/
The file access permissions do not allow the specified action.
Example 5-13 shows the output of an NFS V4 client accessing the pseudo-root
FS.
Example 5-13 Client output while accessing a pseudo-root FS

# mount -o vers=4 nfs404:/ /nfs
# cd /nfs
# ls
dept home project tt
# cd project/projA
# ls -l
total 0
drwxr-xr-x 2 nobody system 256 Jul 26 17:43 Targets
drwxr-xr-x 2 root system 256 Jul 26 17:42 lost+found
#
# mount
node mounted mounted over vfs date options
-------- --------------- --------------- ------ ---------------------
nfs404 / /nfs nfs4 Jul 29 14:56 vers=4
#
# pwd
/nfs/project/projA
#
# df .
[NFSv4] 65536 64832 2% 5 1% .
#
# nfs4cl showfs
Server Remote Path fsid Local Path
-------- --------------- --------------- ---------------
nfs404 10:13 /nfs/projB
nfs404 10:12 /nfs/projA
nfs404 10:17 /nfs/home
nfs404 / 10:4 /nfs
You may have recognized that the command df is not useful in this context. That
is why the nfs4cl command is recommended.

5.5.3 Setting up the alias tree extension on an NFS V4 server
With reference to the description in 4.7, “NFS protocols and namespace
considerations” on page 112, the alias tree model seems to be the best model for
using pseudo-root FS to create a global NFS namespace. The following steps
show how to set up the alias tree model on one of our sample servers.
Note: The described pseudo-root FS setup cannot coexist with the alias tree
model. You have to choose between the two models.
We created three new file systems on the server that does not share the same
root mount path:
򐂰 /local/trans
򐂰 /local/trans1
򐂰 /usr/codeshare/ThirdPartyProgs
Example 5-14 Sample df output from NFS V4 server for alias tree model
/dev/exname_lv 65536 63376 4% 20 1% /local/trans
/dev/exname2_lv 65536 63392 4% 18 1% /local/trans1
/dev/exname3_lv 65536 63392 4% 18 1%
The second step is to generate a new /etc/exports file and include the exname
option as shown in Example 5-15.
Example 5-15 Sample file /etc/exports using the exname option

/local -vers=4,rw,exname=/exports/local
/local/trans -vers=4,rw,exname=/exports/local/trans
/local1/trans1 -vers=4,rw,exname=/exports/local1/trans1
-vers=4,ro,exname=/exports/usr/codeshare/ThirdPartyProgs
Finally, we export the new file systems to the clients using the exportfs -va
command as shown in Example 5-16.
Example 5-16 Exporting file systems using exportfs -va

root@nfs402 #exportfs -va
exportfs: 1831-187 re-exported /local
exportfs: 1831-187 re-exported /local/trans
exportfs: 1831-187 re-exported /local1/trans1
exportfs: 1831-187 re-exported /usr/codeshare/ThirdPartyProgs

You can use this command to mount the pseudo-root FS on the client:
mount -o vers=4,sec=krb5 nfs402:/ /nfs
The global namespace is available using the pseudo-root FS - alias tree model.
In Example 5-17 we show the access to the alias tree on a NFS V4 client.
Example 5-17 Client view of the alias tree

# mount
-------- --------------- --------------- ------ ------------ ---------------
nfs402 / /nfs nfs4 Aug 02 17:26 vers=4,
# find /nfs -print

/nfs
/nfs/usr
/nfs/usr/codeshare
/nfs/usr/codeshare/ThirdPartyProgs
/nfs/local1
/nfs/local1/trans1
/nfs/local1/trans1/trash
/nfs/local
/nfs/local/trans
/nfs/local/trans/var
/nfs/local/trans/trash
/nfs/local/trans/translog
/nfs/local/trans1
# nfs4cl showfs
#
-------- --------------- --------------- ---------------
nfs402 10:15 /nfs/trans
nfs402 10:16 /nfs/trans1
nfs402 10:17 /nfs/ThirdPartyProgs
nfs402 / 10:4 /nfs
nfs403 /exports/acltest 10:13 /mntjt
# pwd
/nfs/local1/trans1
# df .
[NFSv4] 65536 63392 4% 18 1% .
#

A more realistic example was discussed in 2.6.3, “Better namespace handling”
on page 25 and will be covered in this chapter. On the server that does not share
the same root mount path, we created four new file systems called:
򐂰 /local/trans
򐂰 /local/home
򐂰 /local/dept
򐂰 /usr/codeshare/ThirdPartyProgs
The intention is to map these directories on the client in a global namespace

called /nfs with the following directory structure:
򐂰 /nfs/home
򐂰 /nfs/home
򐂰 /nfs/dept
򐂰 /nfs/ThirdPartyProgs
To do so we created the following /etc/exports file on the server and included the
exname option shown in Example 5-18.
Example 5-18 Sample alias tree file /etc/exports file on the server
To mount the pseudo-root FS on the client, the following command is used:

The Global Namespace is made available using the pseudo-root FS - alias tree
model.
Example 5-19 shows the access to the alias tree on an NFS V4 client.
Example 5-19 Extended client view of the pseudo-root FS - alias tree model
#mount -o sec=krb5,vers=4 nfs404:/ /nfs
#
#cd /nfs
#
# find . -print
.
./ThirdPartyProgs
./ThirdPartyProgs/bin
./ThirdPartyProgs/src
./ThirdPartyProgs/contrib
./home
./home/sally

./home/bob
./home/mary
./home/joe
./dept
./dept/eng
./dept/hr
./dept/sales
./trans
./trans/lost+found
#
# cd /nfs/home
#
# ls -ltr
total 40
drwxrwx--- 2 root system 512 Aug 12 18:41 lost+found
drwxr-sr-x 2 sally staff 512 Aug 12 18:50 sally
drwxr-sr-x 2 bob staff 512 Aug 12 18:50 bob
drwxr-sr-x 2 mary staff 512 Aug 12 18:50 mary
drwxr-sr-x 2 joe staff 512 Aug 12 18:50 joe
# mount
-------- --------------- --------------- ------ ------------ ---------------
/dev/hd4 / jfs Aug 10 14:59 rw,log=/dev/hd8
/dev/hd2 /usr jfs Aug 10 14:59 rw,log=/dev/hd8
/dev/hd9var /var jfs Aug 10 14:59 rw,log=/dev/hd8
/dev/hd3 /tmp jfs Aug 10 14:59 rw,log=/dev/hd8
/dev/hd1 /home jfs Aug 10 14:59 rw,log=/dev/hd8
/proc /proc procfs Aug 10 14:59 rw
/dev/hd10opt /opt jfs Aug 10 14:59 rw,log=/dev/hd8
/dev/log_lv /var/nfs4log jfs Aug 10 14:59 rw,log=/dev/hd8
nfs404 / /nfs nfs4 Aug 10 15:07 vers=4,sec=krb5
#
# nfs4cl showfs

-------- --------------- --------------- ---------------
nfs404 10:12 /nfs/trans
nfs404 10:18 /nfs/dept
nfs404 10:17 /nfs/home
nfs404 10:19 /nfs/ThirdPartyProgs
nfs404 / 10:4 /nfs
5.6 Setting up the NAS with a legacy database

In this section we describe the sample setup of IBM NAS Version 1.4 on AIX with
a KDC legacy database. The identification of the user will be done using

standard AIX local mechanisms: by using the /etc/passwd file entries. This
means that we do not use a centralized user identification registry such as LDAP
or NIS.
The following definitions are used in the example:

NFS and KDC server host name = nfs403, OS: AIX 5.3
NFS client host name = nfs406, OS: AIX 5.3
NFS Domain Name itsc.austin.ibm.com
Realm Name REALM1.IBM.COM
Figure 5-1 gives a better view of the setup.
Figure 5-1 Sample AIX environment with AIX KDC server
5.6.1 Setup of a KDC server

In this section, we describe the minimal setup of the KDC server using IBM NAS
V1.4 so that NFS V4 can use RPCSEC_GSS for authentication. Further
information about IBM NAS and the KDC infrastructure can be found in IBM
Network Authentication Service Version 1.4 for AIX, Linux, and Solaris
Administrator’s and User’s Guide, which is delivered with the krb5.doc.en_US file
set.
5.6.2 Installing the IBM NAS file sets

The IBM NAS Version1.4 file sets are delivered with the AIX 5.3 Expansion CD.
They can be installed using smit or the installp command. We used the
command line interface for installation. The command we used was:
installp -aqXgd . krb5.server modcrypt.base

Set up the KDC server
To configure NAS to use the legacy database, which is stored on the local file
system, we used the following command:
mkkrb5srv -r REALM2.IBM.COM -d itsc.austin.ibm.com -s \
nfs403.itsc.austin.ibm.com
While running this command, the system asks for a Master Database password
and a password for the administrative principal called admin. Record the name
and chosen password in an secure place as these principals are essential for
your NAS environment.
Example 5-20 shows the output from the mkkrb5srv command with the options
shown above.
Example 5-20 Output of command mkkrb5srv on the KDC server

# mkkrb5srv -r REALM2.IBM.COM -d itsc.austin.ibm.com \
-s nfs403.itsc.austin.ibm.com
Fileset Level State Description

----------------------------------------------------------------------------
Path: /usr/lib/objrepos
krb5.server.rte 1.4.0.0 COMMITTED Network Authentication Service
Server
Path: /etc/objrepos
Server
The -s option is not supported.
The administration server will be the local host.
Initializing configuration...
Creating /etc/krb5/krb5_cfg_type...
Creating /etc/krb5/krb5.conf...
Creating /var/krb5/krb5kdc/kdc.conf...
Creating database files...
Initializing database '/var/krb5/krb5kdc/principal' for realm 'REALM2.IBM.COM'
master key name 'K/[email protected]'
You are prompted for the database Master Password.
It is important that you DO NOT FORGET this password.
Enter database Master Password:
Re-enter database Master Password to verify:
WARNING: no policy specified for admin/[email protected];
defaulting to no policy. Note that policy may be overridden by
ACL restrictions.
Enter password for principal "admin/[email protected]":
Re-enter password for principal "admin/[email protected]":
Principal "admin/[email protected]" created.
Creating keytable...

Creating /var/krb5/krb5kdc/kadm5.acl...
Starting krb5kdc...
krb5kdc was started successfully.
Starting kadmind...
kadmind was started successfully.
The command completed successfully.
Restarting kadmind and krb5kdc
The same can be achieved by running the NAS command:

config.krb5 -S -d itsc.austin.ibm.com -r REALM2.IBM.COM
In addition, the two lines shown in Example 5-21 are added automatically to the
/etc/inittab file so that the KDC server is automatically started after system
reboot.
Example 5-21 /etc/inittab entries for NAS

# lsitab krb5kdc
krb5kdc:2:once:/usr/krb5/sbin/krb5kdc
#
# lsitab kadm
kadm:2:once:/usr/krb5/sbin/kadmind
#
Attention: If you intend to install the KDC on an NFS V4 server, be aware that
after activating the KDC server your NFS V4 server will not work properly until
you have completed all further steps.
After successful installation and configuration of the KDC, on the server, the files
shown in Figure 5-1 are installed in the /etc/krb5 directory.
Table 5-1 Additional files installed at KDC configuration

krb5.conf Contains general information for clients and servers. It must reside
on each system that contains the administration server, a KDC, or
a client. If two or more Network Authentication Service servers or
clients reside on the same system, they must share the same
krb5.conf file.
krb5_cfg_type Determines the configuration type of the machine (master, slave, or

client). This file must reside on the system that contains the
administration server.
Example 5-22 on page 138 shows the contents of the /etc/krb5/krb5.conf file,
and Example 5-23 on page 138 shows the contents of the
/etc/krb5/krb5_cfg_type file.

Example 5-22 Sample /etc/krb5/krb5.conf file on the KDC server
# cat krb5.conf
[libdefaults]
default_realm = REALM2.IBM.COM
default_keytab_name = FILE:/etc/krb5/krb5.keytab
default_tkt_enctypes = des3-cbc-sha1 arcfour-hmac aes256-cts
des-cbc-md5 des-cbc-crc
default_tgs_enctypes = des3-cbc-sha1 arcfour-hmac aes256-cts
[realms]
REALM2.IBM.COM = {
kdc = nfs404.itsc.austin.ibm.com:88
admin_server = nfs404.itsc.austin.ibm.com:749
default_domain = ibm.com
}
[domain_realm]
.ibm.com = REALM2.IBM.COM
nfs404.itsc.austin.ibm.com = REALM2.IBM.COM
[logging]
kdc = FILE:/var/krb5/log/krb5kdc.log
admin_server = FILE:/var/krb5/log/kadmin.log
default = FILE:/var/krb5/log/krb5lib.log
#
Note: The default_tkt_enctypes and default_tgs_enctypes can be reduced to

show only des-cbc-crc and des-cbc-md5, but keep in mind that this may only
be correct with NFS V4.
Example 5-23 Sample contents of file /etc/krb5/krb5_cfg_type on the KDC server

# cat /etc/krb5/krb5_cfg_type
master
5.6.3 Initial basic KDC functions test

Before continuing, we test the initial setup of the KDC server by using the
following command sequence to verify that all required processes for the KDC
server have started:
ps -ef |grep krb |grep -v grep
This verifies the logon principal as admin:

kinit admin/[email protected]

This verifies that a ticket has been granted to the admin principal:
klist
Example 5-24 shows the output of the above mentioned commands.
Example 5-24 Basic verification of KDC server

# ps -ef |grep krb |grep -v grep
root 299160 1 0 18:20:51 - 0:00 /usr/krb5/sbin/krb5kdc
root 315554 1 0 18:20:51 - 0:00 /usr/krb5/sbin/kadmind
#
# kinit admin/[email protected]
Password for admin/[email protected]:
#
# klist
Ticket cache: FILE:/var/krb5/security/creds/krb5cc_0
Default principal: admin/[email protected]
Valid starting Expires Service principal

07/28/04 14:16:18 07/29/04 14:16:15 krbtgt/[email protected]
5.6.4 Create user principals on the KDC server

In this section, we create Kerberos user principals that match the existing UNIX
user names. The principal name will be mapped to the user name by NFS to
determine the UNIX credential associated with the principal. In general,
principals can be added on the KDC server using this command:
kadmin.local -> add_principal (or addprinc)
Note: kadmin.local can only be run on the master KDC, whereas kadmin can
be run on any machine that is part of the Kerberos realm. We use both
variations of the command in examples in this chapter.
In Example 5-25, the principal named sally is added to the KDC database via
command line. This requires that you have already authenticated to Kerberos
using an administrative principal, for example admin/admin.
Note: We assume that the local AIX user sally already exists on the KDC
server. In addition, this user name must also exist on all NFS V4 clients.
Example 5-25 Adding the principal named sally

# kadmin.local
kadmin.local: addprinc -e des-cbc-crc:normal sally
WARNING: no policy specified for [email protected];

ACL restrictions.
Enter password for principal "[email protected]":
Re-enter password for principal "[email protected]":
Principal "[email protected]" created.
#
To verify the newly created principal, we use the kadmin.local command as

shown in Example 5-26. The kadmin.local interface also gives the options
listprincs, get_principal, and getprinc to achieve the result.
Example 5-26 Output of kadmin.local getprinc command

kadmin.local: getprinc sally
Principal: [email protected]
Expiration date: [never]
Last password change: Wed Jul 28 14:58:25 CDT 2004
Password expiration date: [none]
Maximum ticket life: 1 day 00:00:00
Maximum renewable life: 7 days 00:00:00
Last modified: Wed Jul 28 14:58:25 CDT 2004 (admin/[email protected])
Last successful authentication: [never]
Last failed authentication: [never]
Failed password attempts: 0
Number of keys: 1
Key: vno 1, DES cbc mode with CRC-32,
no salt
Attributes:
REQUIRES_PRE_AUTH
Policy: [none]
Example 5-26 shows that the number of keys equals 1, and the standard
encryption for this principal is set to Single-DES. You could have principals with
different number of keys and standard encryption types within the KDC.
In the next step we verify that the user can be authenticated using the created
principle sally as shown in Example 5-27.
Example 5-27 Verify the principal sally by use of kinit

# kinit [email protected]
Password for [email protected]:
#
# klist
Default principal: [email protected]

Renew until 07/29/04 15:17:22
This procedure has to be executed for every AIX user in your environment. For a
large number of users you may want to create a script to automate this process.
Example 5-28 should give you an idea of how to achieve this.
Example 5-28 User principal creation shell script

#/bin/ksh
NAME=joe
NPASSWD=”new01new”
/usr/krb5/sbin/kadmin.local <<EOF
add_principal -e des-cbc-crc:normal -pw ${NPASSWD} $NAME
EOF
Note: This script will run only if you are already authenticated as admin. Using
this flag in a shell script can be dangerous if unauthorized users gain read
access to this script. The script is only provided to give you an idea of how this
can be achieved.
You also have the option to run the utility mkseckrb5, delivered with the AIX Base
Operating System bos.rte.security file set, to import all existing users from the
local system into the KDC database. For further details, see the commands
reference chapter of the AIX 5.3 online documentation.
5.6.5 Create the NFS server principals on the KDC server

For each NFS server in your KDC environment, you must define a principal of
type nfs/<full_qualified_hostname>@REALM, for example:
nfs/[email protected]
The setup for this principal is slightly different from the user principal setup, as we
use a random password instead of one that is user-defined.
Example 5-29 Creation of NFS server principal nfs/nfs@REALM

# /usr/krb5/sbin/kadmin
kadmin.local: add_principal -e des-cbc-crc:normal -randkey
nfs/nfs404.itsc.austin.ibm.com
WARNING: no policy specified for nfs/[email protected];
ACL restrictions.
Principal "nfs/[email protected]" created.

kadmin.local: getprinc nfs/nfs404.itsc.austin.ibm.com
Principal: nfs/[email protected]
Last password change: Wed Jul 28 16:21:44 CDT 2004
Last modified: Wed Jul 28 16:21:44 CDT 2004 (admin/[email protected])
Number of keys: 1
Key: vno 2, DES cbc mode with CRC-32,
no salt
Attributes:
Policy: [none]
5.7 Setting up an NFS V4 server with NAS on a different

KDC server
In this section, we demonstrate the steps that are required to set up an NFS V4
server in your NAS environment by using the KDC server created in “Installing
the IBM NAS file sets” on page 135 to act as a NFS V4 server as well.
5.7.1 Create the NFS server keytab file entry

As described previously, the principal for the NFS server has already been
created as:
The next step is to set up a keytab entry on the NFS V4 server using the kadmin
Example 5-30 Sample ktadd to create an NFS server keytab file

kadmin: ktadd nfs/nfs404.itsc.austin.ibm.com
Entry for principal nfs/nfs404.itsc.austin.ibm.com with kvno 3, encryption type
Triple DES cbc mode with HMAC/sha1 added to keytab
WRFILE:/etc/krb5/krb5.keytab.
ArcFour with HMAC/md5 added to keytab WRFILE:/etc/krb5/krb5.keytab.

AES-256 CTS mode with 96-bit SHA-1 HMAC added to keytab
DES cbc mode with RSA-MD5 added to keytab WRFILE:/etc/krb5/krb5.keytab.
This command created the /etc/krb5/krb5.keytab file on the local system.
5.7.2 Check the NFS V4 server before client access

To be sure that the setup on the NFS V4 server is correct, we check the following:
򐂰 Are the keytab entries valid? Use the ktutil command.
Example 5-31 Check of valid keytab on the NFS V4 server

# /usr/krb5/sbin/ktutil
ktutil: read_kt /etc/krb5/krb5.keytab
ktutil: l
slot KVNO Principal
------ ------ ------------------------------------------------------
1 3 nfs/[email protected]
򐂰 Can the server principal get valid tickets? Use the kinit command.
Example 5-32 kinit using the server keytab file

# kinit -kt /etc/krb5/krb5.keytab\ nfs/nfs404.itsc.austin.ibm.com
#
# klist
Default principal: nfs/[email protected]

#
5.7.3 Set up the NFS registry daemon

The NFS domain name has to be set. The current setting can be checked using
chnfsdom or smitty chnfsdom commands.

If the NFS domain is not set, the output appears as:
# chnfsdom
In our sample environment, we already set the NFS domain so that the output
appears as:
# chnfsdom
Current local domain: nfs.ibm.com
Note: Changing the NFS domain does not recycle or start the nfsrgyd. You
have to start and recycle the daemon manually.
We started the nfsrgyd by using the startsrc -s nfsrgyd command:

# startsrc -s nfsrgyd
5.7.4 Set up the gssd daemon on the NFS V4 server

To enable NFS V4 using RPSEC-GSS, we have to create the map file between
the server’s keytab file and the NFS server principal. This is done using the
nfshostkey command as shown in Example 5-33.
Example 5-33 Creating the server’s hostkey map file

# nfshostkey -p nfs/nfs404.itsc.austin.ibm.com -f /etc/krb5/krb5.keytab
#
# nfshostkey -l
/etc/krb5/krb5.keytab
#
To set up the gssd daemon to start up automatically, run the chnfs -S -B or

smitty addsecurity command. You have to repeat this setup for each NFS V4
server in your infrastructure.
The next step is to create the file /etc/nfs/realm.map. Use the chnfsrtd
command. In our sample environment, the realm is REALM2.IBM.COM and the
NFS domain is itsc.austin.ibm.com.
We used the command chnfsrtd -a REALM2.IBM.COM itsc.austin.ibm.com or

smit chnfsrtd.
Example 5-34 Setting the realm.map file on the NFS V4 server

# cat /etc/nfs/realm.map
realm2.ibm.com itsc.austin.ibm.com

# chnfsrtd
Note: The realm entry in the /etc/nfs/realm.map file is not case-sensitive.
5.8 Setting up an NFS V4 client with NAS

In this section, we describe the steps that are required to set up a client system
to use the created NAS and NFS V4 environment. We describe the basic
required steps as well as the differences between a full client and slim client.
5.8.1 General steps for all types of clients

In general, the steps required to set up an NFS V4 client with NAS support are:
1. Install the NAS file set from the AIX 5.3 Expansion CD.
2. Set the NFS domain using smit chnfsdom or chnfsdom command.
3. Set the NFS domain-to-realm mapping using chnfsrtd.
4. Set up the NAS environment:
– Full client: Create the NFS host principal and keytab file.
– Slim client: Retrieve the required configuration files from a master.
For details about slim and full clients, see 4.4.4, “NFS client considerations
when using Kerberos” on page 108 in this book.
5. Configure the gssd daemon after the Kerberos installation and verification
have been completed.
5.8.2 Install the NAS client code

Next, we install IBM NAS Version 1.4 from the AIX 5.3 Expansion CD. Install the
following file sets using the smit or installp commands:
krb5.client Delivers all required client commands and libraries.
modcrypt.base Delivers the required kernel extension for using NFS V4.
Example 5-35 shows the installation process and the file sets that are installed
as part of the above.
Example 5-35 NAS client file set installation

# installp -aqXgYd . krb5.client modcrypt.base
+-----------------------------------------------------------------------------+

Installation Summary
--------------------
Name Level Part Event Result
-------------------------------------------------------------------------------
modcrypt.base.lib 5.3.0.0 USR APPLY SUCCESS
modcrypt.base.includes 5.3.0.0 USR APPLY SUCCESS
krb5.client.rte 1.4.0.0 USR APPLY SUCCESS
krb5.client.samples 1.4.0.0 USR APPLY SUCCESS
krb5.client.rte 1.4.0.0 ROOT APPLY SUCCESS
5.8.3 Set up the NFS domain

In Example 5-36, we set the NFS domain to itsc.austin.ibm.com by using the
chnfsdom command.
Example 5-36 Client NFS domain setup

# chnfsdom
#
#
# chnfsdom
#
5.8.4 Set up the NFS domain-to-realm map

The AIX 5.3 implementation of NFS V4 requires that you have a cross-relation
definition between the NFS domain and the Kerberos realm that is used. This is
done by generating a realm map file by using the smit or chnfsrtd commands.
Example 5-37 Client NFS domain-to-realm mapping

# chnfsrtd -a REALM1.IBM.COM itsc.austin.ibm.com
#
#chnfsrtd
#
At this point, you should start the necessary NFS processes by running the
/etc/rc.nfs script. Before completing the client setup by enabling RPSSEC_GSS,
we have to set up the NAS environment by using a full or slim NFS V4 Kerberos
client, which we describe in the next two sections.

5.8.5 Full client installation steps
After installing the NAS client file sets, we configure the Kerberos client using the
mkkrb5clnt command as shown in Example 5-38.
Example 5-38 mkkrb5clnt output

# mkkrb5clnt -d itsc.austin.ibm.com -r REALM1.IBM.COM -c\
nfs407.itsc.austin.ibm.com -s nfs407.itsc.austin.ibm.com
#
#
The same result can be achieved by using the config.krb5 command with
options -C -d itsc.austin.ibm.com -r REALM1.IBM.COM -c
nfs407.itsc.austin.ibm.com -s nfs407.itsc.austin.ibm.com
The NFS server (or an NFS V4 Full Kerberos client), using RPCSEC_GSS
security, must be able to acquire credentials for its service principal
nfs/<host_name>@REALM to authenticate requests. This type of principal must
be created using the kadmin command on either a server or the client during
installation. We create this principal on the client. For this operation, you should
know the Kerberos administrative principal (default is admin/admin@REALM)
and the password for the administrative principal.
In addition, this operation verifies that the client can communicate with the KDC
server.
Example 5-39 kinit to admin/admin

# kinit admin/admin
#
# klist

#
The next step is to add the NFS service principal using the kadmin command.
Example 5-40 on page 148 shows that we used the kadmin: listprincs
commands to verify that no service principal of type

nfs/nfs403.itsc.austin.ibm.com has been registered for the host
nfs403.itsc.austin.ibm.com. If this is the case, you can omit the principal creation
and continue with creation of the keytab file.
Example 5-40 kadmin to add the client service principal

Authenticating as principal admin/[email protected] with password.
kadmin: listprincs
[email protected]
[email protected]
[email protected]
[email protected]
host/[email protected]
root/[email protected]
admin/[email protected]
kadmin:
kadmin: add_principal -e des-cbc-crc:normal -randkey
ACL restrictions.
kadmin:
Now we create the keytab file so that the machine can request a valid ticket after
reboot. This enables the machine to mount NFS V4 directories with the security
flavor sec=krb5.
Example 5-41 Creating the client keytab file

Triple DES cbc mode with HMAC/sha1 added to keytab
ArcFour with HMAC/md5 added to keytab WRFILE:/etc/krb5/krb5.keytab.
AES-256 CTS mode with 96-bit SHA-1 HMAC added to keytab
DES cbc mode with RSA-MD5 added to keytab WRFILE:/etc/krb5/krb5.keytab.
kadmin:
This step created the local keytab file krb5.keytab in the /etc/krb5 directory. If this
keytab file is generated on another machine, such as a Windows KDC server,

you can transfer it using binary ftp to the correct location on the client and import
the file using the ktutil command.
We verify the created keytab file using the kinit command (Example 5-42).
Example 5-42 kinit using the client keytab file

# kinit -kt /etc/krb5/krb5.keytab nfs/nfs403.itsc.austin.ibm.com
#
# klist

#
To enable NFS V4 to use RPSEC-GSS, we create a map between the client

keytab file and the NFS service principal. This is done using the nfshostkey
command (Example 5-43).
Example 5-43 hostkey map creation on the client

#
# nfshostkey -l
#
We again verify that the nfshostkey map file can be used to get a valid Kerberos
ticket.
Example 5-44 Verification of the client nfs hostkey

# kdestroy
#
# kinit -kt `tail -n 1 /etc/nfs/hostkey` `head -n 1 /etc/nfs/hostkey`
#
# klist

#
The successful verification completes the full client installation.

5.8.6 Slim client installation steps
As described in “NFS V4 slim client versus full client” on page 108, a slim client
does not have a service principal registered within the KDC server. In addition,
you cannot create a keytab file, so no Kerberos ticket can be issued automatically
on reboot.
Restriction: While writing this book, we learned that a slim client is not
capable of automatically mounting file systems that require RPCSEC_GSS
authorization only. In this case, the security flavor on the server must be at
least sec=sys:krb5(x).
Another recommended approach is to utilize the NFS V4 pseudo-FS model

and use a single client mount of the server’s exported pseudo-root. The
pseudo-root typically allows non-RPCSEC-GSS access.
The following are reasons why you would want to deploy slim clients in your
infrastructure:
򐂰 Pre-installed new systems
򐂰 Mass rollout of new systems
򐂰 Unprompted upgrade of clients
In addition, there are several choices for how to install the slim clients, based on
the way your systems are installed:
򐂰 Complete new installation (scratch installation)
򐂰 Cloning by use of a full system backup image
In the following section, we show how a slim client can be installed:

򐂰 To be used as a master image for cloning purposes
򐂰 Using the NAS and NFS V4 configuration files from the master machine
The provided examples show the configuration of a slim client. They can be used
if the requirement is met and all targeted systems are to be in the same NFS
domain and Kerberos realm. Otherwise, the slim client will not function properly
and you will have to manually change the Kerberos configuration, the NFS
domain configuration, or both.
Configure a slim client for cloning

We decided to use the following method for cloning:
1. Install one client with all required file sets to use NAS as described in “Full
client installation steps” on page 147.

2. Configure the Kerberos client using the mkkrb5clnt command, but do not
create a service principal or a keytab file.
3. Set up the NFS domain and enable RPCSEC_GSS.
4. Create all required /etc/filesystem entries.
5. Create a full system backup image to be used by NIM.
Example 5-45 Slim client configuration for cloning

# lslpp -L krb5*
Fileset Level State Type Description (Uninstaller)
----------------------------------------------------------------------------
krb5.client.rte 1.4.0.0 C F Network Authentication
Service
Client
krb5.client.samples 1.4.0.0 C F Network Authentication
Service
Samples
#
# mkkrb5clnt -d itsc.austin.ibm.com -r REALM1.IBM.COM -c\
#
# chnfsdom
#
# chnfsdom nfs.ibm.com
# chnfsdom
#
# chnfs -S -B
0513-059 The gssd Subsystem has been started. Subsystem PID is 14812.
#
# /etc/rc.nfs
0513-059 The rpc.lockd Subsystem has been started. Subsystem PID is 14522.
0513-029 The gssd Subsystem is already active.
#
# klist
Unable to get cache name (ticket cache: /var/krb5/security/creds/krb5cc_0).
Status 0x96c73ac3 - No credentials cache found.

#
# ls /etc/krb5
krb5.conf krb5_cfg_type
#
# ls /etc/nfs
local_domain
#
Configure a slim client on a preinstalled AIX 5.3 system

To be able to configure an already installed AIX 5.3 system as slim client, we
used the following method:
1. Create a tar image of the basic NAS and NFS configuration files on the
master system as described above (Example 5-46).
This is the minimal list of files that are needed on a slim client. If your
environment requires the need to have, for example, Enterprise Identity
Mapping, there will be more files in the /etc/nfs directory.
Example 5-46 Creating a tar image of the basic NAS and NFS configuration files
# ls /etc/nfs/* > /tmp/SlimClientInList
#
# ls /etc/krb5/* >> /tmp/SlimClientInList
#
# tar -cvf /tmp/SlimClientImage.tar -L /tmp/SlimClientInList
a /etc/nfs/local_domain 1 blocks.
a /etc/nfs/realm.map 1 blocks.
a /etc/krb5/krb5.conf 2 blocks.
a /etc/krb5/krb5_cfg_type 1 blocks.
#
2. Copy the tar image onto the target system.

3. Install the contents of the tar image on the system (Example 5-47).
Example 5-47 Installing the contents of the tar image onto the target system
# tar -xvf /tmp/SlimClientImage.tar
x /etc/nfs/local_domain, 12 bytes, 1 media blocks.
x /etc/nfs/realm.map, 68 bytes, 1 media blocks.
x /etc/krb5/krb5.conf, 864 bytes, 2 media blocks.
x /etc/krb5/krb5_cfg_type, 7 bytes, 1 media blocks.
#
# kinit sally
#
# klist


#
Verification of the slim client

On the server, validate the NFS pseudo-root. Include, at the very least, the
security flavor sec=sys to enable mount during startup on the slim client. The
following file systems have been exported on the server:
򐂰 /exports -vers=4,sec=sys:krb5,rw
򐂰 /exports/home -vers=4,sec=sys:krb5,rw
򐂰 /exports/project/projA -vers=4,sec=sys:krb5,ro
򐂰 /exports/project/projB -vers=4,sec=sys:krb5,ro
The slim client will not be able to mount a file system using security flavor
sec=krb5 without having valid Kerberos tickets. To achieve automatic mount of
the pseudo-root FS during startup, the security flavor sec=sys on the server and
the client for the NFS pseudo-root FS has to be used.
We used smitty mknfsmnt to create NFS pseudo-root FS mount on the client as

shown in Example 5-48.
Example 5-48 Slim client /etc/filesystems NFS mount

# tail /etc/filesystems
/nfs:
dev = "/"
vfs = nfs
nodename = nfs404
mount = true
options = bg,hard,intr,vers=4,sec=sys:krb5
account = false
The slim client is now rebooted and with UID root, we access the NFS remote
mounted directory.
Example 5-49 Slim client access after reboot

# uptime
12:22PM up 8 mins, 1 user, load average: 0.02, 0.26, 0.19
#
#klist
#

# id
uid=0(root) gid=0(system)
groups=2(bin),3(sys),7(security),8(cron),10(audit),11(lp)
#
#mount
-------- --------------- --------------- ------ ------------ ---------------
nfs404 / /nfs nfs4 Aug 12 12:14
rw,bg,hard,intr,vers=4,sec=sys:krb5
#
# cd /nfs
#
# ls
dept home local project tmp
#
This system is now ready to be used as the master full system backup for
prompted installation of the clients.
5.8.7 Configuring RPCSEC_GSS on the clients

If you want to use Kerberos authentication on the client, you must enable NFS V4
enhanced security on the client. You can enable enhanced security using smit
Configure NFS on This System → Enable RPCSEC_GSS or by using the
chnfs -S -B command:
# chnfs -S -B
0513-059 The gssd Subsystem has been started. Subsystem PID is 22158.
#
At this point, the client setup is finished and can mount and access directories
with enhanced security.
Example 5-50 Verify the client using mount command

# mount -o vers=4,sec=krb5 nfs404:/ /nfs
#
# nfs4cl showfs /nfs

-------- --------------- --------------- ---------------
nfs404 / 10:4 /nfs
Current Server: nfs404:/nfs
options :
rw,intr,rsize=32768,wsize=32768,timeo=300,retrans=5,biods=0,numclust=2,maxgroup
s=0,acregmin=3,acregmax=60,acdirmin=30,acdirmax=60,sec=krb5
#
The client is now ready for deploying NFS V4.
5.9 Preparing the system for Tivoli Directory Server and

Kerberos V5
In this section, we lead you through the mandatory preparation steps on your AIX
5.3 system before you can deploy Kerberos V5 with the IBM Tivoli Directory
Server LDAP backend.
Attention: As we write this book, there is a requirement for the system to be

running in 64-bit kernel mode. This is required for the db2_08_01.ldap file set,
which gives you the LDAP DB2® backend. This may change in future releases
of the file set.
The following definitions are used in the example:

NFS server hostname = nfs404, OS = AIX 5.3
NFS client hostname = nfs405, OS: AIX 5.3
KDC and LDAP server hostname = nfs407, OS: AIX 5.3
Figure 5-2 on page 156 gives a better view of the setup.

Figure 5-2 Sample AIX environment with an AIX KDC and LDAP backend
5.9.1 Set up procedure

The following procedure needs to be carried to ensure that your system is
running a 64-bit kernel.
1. Check to see if your system is capable of running in 64-bit mode:
# bootinfo -y
64
#
If the result is 32, then your system is a 32-bit machine and you will not be able to
proceed any further. If the result is as shown in the previous example, then
continue with the rest of the steps.
2. We confirm what mode the system is running in:
# bootinfo -K
32
#

As can be seen from the output, the system is currently running in 32-bit
mode. Before we can progress with the install, we have to configure it to run in
64-bit mode.
3. To change the system to run in 64-bit mode:
# cd /
#
# ln -sf /usr/lib/boot/unix_64 /unix
#
# ls -al /unix
lrwxrwxrwx 1 root system 21 Aug 04 15:30 /unix -> /usr/lib/boot/unix_64
#
# bosboot -ad /dev/ipldevice
bosboot: Boot image is 22469 512 byte blocks.

#
4. We now make sure that aio is enabled at system start.
# lsattr -El aio0
autoconfig defined STATE to be configured at system restart True
fastpath enable State of fast path True
kprocprio 39 Server PRIORITY True
maxreqs 4096 Maximum number of REQUESTS True
maxservers 10 MAXIMUM number of servers per cpu True
minservers 1 MINIMUM number of servers True
As the example shows, autoconfig is in a defined state for aio, and we need
to change this to an available state. Use the following command and change
STATE to be configured at system restart to available:
# smitty chgaio
5. We now reboot the system to allow for the 64-bit kernel.
6. After the system has rebooted, verify that it is running in 64-bit mode:
# bootinfo -K
64
#
7. We also confirm that aio is now in available state:
# lsattr -El aio0
autoconfig available STATE to be configured at system restart True
fastpath enable State of fast path True
kprocprio 39 Server PRIORITY True
maxreqs 4096 Maximum number of REQUESTS True
maxservers 10 MAXIMUM number of servers per cpu True
minservers 1 MINIMUM number of servers True
#

8. Next, we need to create a 350 MB file system so that we can install the DB2
binaries. (We want to keep the root file system as small as possible.) The new
file system should be created to mount on to /usr/opt/db2_08_01. You will
need to change the owner and group to bin respectively.
9. We also need to create a 200 MB file system to mount to /home/ldapdb2.
10.Create a group called dbsysadm and add the root user to this group.
11.Create a user called ldapdb2. This user must have these characteristics:
a. Primary group: dbsysadm
b. Password REGISTRY: files
c. HOME directory: /home/ldapdb2
d. Change the owner and group of /home/ldapdb2 to ldapdb2:dbsysadm
Important: Make sure that user ldapdb2 has a password assigned and can
log on without any challenges before proceeding.
12.Add the following entries to /etc/services:

DB2_ldapdb2 60000/tcp
DB2_ldapdb2_1 60001/tcp
DB2_ldapdb2_2 60002/tcp
DB2_ldapdb2_END 60003/tcp
13.The following file sets must be installed for IBM Directory Server. They can be
found on the AIX Base Installation media:
– ldap.client.adt
– ldap.client.rte
– ldap.html.en_US.config
– ldap.html.en_US.man
– ldap.msg.en_US
– ldap.server.cfg
– ldap.server.com
– ldap.server.java
– ldap.server.rte
– ldap.webdadmin
– ldap.client.rte
– ldap.server.cfg
– ldap.server.com
– db2_08_01.ldap
To install the file sets, this command was used:
# installp -aXYgd /dev/cd0 ldap.server.rte

At the end of the install, the screen in Example 5-51 should confirm a
successful install.
Example 5-51 Successful install of ldap.rte.server

+-----------------------------------------------------------------------------+
Summaries:
+-----------------------------------------------------------------------------+
--------------------
-------------------------------------------------------------------------------
ldap.client.rte 5.2.0.0 USR APPLY SUCCESS
ldap.client.adt 5.2.0.0 USR APPLY SUCCESS
ldap.client.rte 5.2.0.0 ROOT APPLY SUCCESS
ldap.server.java 5.2.0.0 USR APPLY SUCCESS
db2_08_01.client 8.1.1.16 USR APPLY SUCCESS
db2_08_01.cnvucs 8.1.1.16 USR APPLY SUCCESS
db2_08_01.conv 8.1.1.16 USR APPLY SUCCESS
db2_08_01.db2.rte 8.1.1.16 USR APPLY SUCCESS
db2_08_01.db2.samples 8.1.1.16 USR APPLY SUCCESS
db2_08_01.essg 8.1.1.16 USR APPLY SUCCESS
db2_08_01.icuc 8.1.1.16 USR APPLY SUCCESS
db2_08_01.icut 8.1.1.16 USR APPLY SUCCESS
db2_08_01.jdbc 8.1.1.16 USR APPLY SUCCESS
db2_08_01.jhlp.en_US.iso885 8.1.1.16 USR APPLY SUCCESS
db2_08_01.cj 8.1.1.16 USR APPLY SUCCESS
db2_08_01.ldap 8.1.1.16 USR APPLY SUCCESS
db2_08_01.msg.en_US.iso8859 8.1.1.16 USR APPLY SUCCESS
db2_08_01.pext 8.1.1.16 USR APPLY SUCCESS
db2_08_01.repl 8.1.1.16 USR APPLY SUCCESS
db2_08_01.sqlproc 8.1.1.16 USR APPLY SUCCESS
ldap.server.rte 5.2.0.0 USR APPLY SUCCESS
ldap.server.com 5.2.0.0 USR APPLY SUCCESS
ldap.server.cfg 5.2.0.0 USR APPLY SUCCESS
ldap.server.com 5.2.0.0 ROOT APPLY SUCCESS
ldap.server.cfg 5.2.0.0 ROOT APPLY SUCCESS
db2_08_01.conn 8.1.1.16 USR APPLY SUCCESS
db2_08_01.cs.rte 8.1.1.16 USR APPLY SUCCESS
db2_08_01.das 8.1.1.16 USR APPLY SUCCESS
db2_08_01.db2.engn 8.1.1.16 USR APPLY SUCCESS
#
14.Install the Kerberos V5 file sets as described in 5.6.2, “Installing the IBM NAS
file sets” on page 135.

Note: If you need SSL capabilities, install the AIX Certificate and SSL base file
sets (gskta for 64-bit kernel, gsksa for 32-bit kernel) and the crypto file sets
from the AIX 5.3 Expansion Pack. Both gskta and gsksa are required. gsksa is
used by the LDAP client apps such as ldapsearch.
We have now successfully installed all of the software that is required to support
KDC and RFC2307 with IBM Directory Server.
5.9.2 Configure IBM Tivoli Directory Server

After installing the required programs, we now proceed with the initial
configuration:
1. Ensure that ibmslapd is not running; otherwise the command in the next step
would fail with the following message:
08/13/04 15:55:15 Non-SSL port initialized to 389.
08/13/04 15:55:16 Attempt to bind failed; errno 67 (Address already in
use).
08/13/04 15:55:16 SocketInit failed for port 389.
2. We start with configuring the directory.
Example 5-52 Configuration of the Directory Server

# mksecldap -s -a cn=admin -p succ3ss -S rfc2307aix -u ALL
Filesystem size changed to 212992
ldapdb2's New password:
Enter the new password again:
You have chosen the following actions:
Administrator DN 'cn=admin' and password will be set.
Setting administrator DN 'cn=admin' and password.

Set administrator DN 'cn=admin' and password.
IBM Tivoli Directory Server Configuration complete.
You have chosen the following actions:
Database 'ldapdb2' will be configured in instance 'ldapdb2'.
Configuring IBM Tivoli Directory Server Database.

Creating instance: 'ldapdb2'.
Created instance: 'ldapdb2'.
Cataloging instance node: 'ldapdb2'.
Cataloged instance node: 'ldapdb2'.

Starting database manager for instance: 'ldapdb2'.
Started database manager for instance: 'ldapdb2'.
Creating database: 'ldapdb2'.
Created database: 'ldapdb2'.
Updating the database: 'ldapdb2'
Updated the database: 'ldapdb2'
Updating the database manager: 'ldapdb2'
Updated the database manager: 'ldapdb2'
Enabling multi-page file allocation: 'ldapdb2'
Enabled multi-page file allocation: 'ldapdb2'
Configuring database: 'ldapdb2'
Configured database: 'ldapdb2'
Adding local loop back to database: 'ldapdb2'.
Added local loop back to database: 'ldapdb2'.
Stopping database manager for instance: 'ldapdb2'.
Stopped database manager for instance: 'ldapdb2'.
Starting database manager for instance: 'ldapdb2'.
Started database manager for instance: 'ldapdb2'.
Configured IBM Tivoli Directory Server Database.
IBM Tivoli Directory Server Configuration complete.

Server starting.
Plugin of type EXTENDEDOP is successfully loaded from libevent.a.
Plugin of type EXTENDEDOP is successfully loaded from libtranext.a.
Plugin of type EXTENDEDOP is successfully loaded from libldaprepl.a.
Plugin of type PREOPERATION is successfully loaded from libDSP.a.
Plugin of type PREOPERATION is successfully loaded from libDigest.a.
Plugin of type AUDIT is successfully loaded from /lib/libldapaudit.a.
Plugin of type AUDIT is successfully loaded from
/usr/ccs/lib/libsecldapaudit64.a(shr.o).
Plugin of type DATABASE is successfully loaded from /lib/libback-rdbm.a.
Plugin of type REPLICATION is successfully loaded from /lib/libldaprepl.a.
Plugin of type EXTENDEDOP is successfully loaded from /lib/libback-rdbm.a.
Plugin of type DATABASE is successfully loaded from /lib/libback-config.a.
Plugin of type EXTENDEDOP is successfully loaded from libloga.a.
Non-SSL port initialized to 389.
Migrating users and groups to LDAP server.
#
3. Although the DB2 database is now configured and running, we need to set
the database to autostart after reboot of the system.

Example 5-53 Change autostart setting for DB2 instance ldapdb2
# su - ldapdb2
$
$ db2iauto -on ldapdb2
$
$ db2set
DB2COMM=TCPIP
DB2AUTOSTART=YES
$
4. We now tune the DB2 database.
Example 5-54 Tuning of the DB2 database

# su - ldapdb2
$
$ db2 update db cfg for ldapdb2 using DBHEAP 20000
DB20000I The UPDATE DATABASE CONFIGURATION command completed successfully.
$
$ db2 update db cfg for ldapdb2 using SORTHEAP 5000
$
$ db2 update db cfg for ldapdb2 using APPLHEAPSZ 10000
$
$ exit
#
5. We now stop the IBM Tivoli Directory Server to facilitate the addition of the
container object for Kerberos:
# ibmdirctl -D cn=admin -w succ3ss stop
Stop operation succeeded
#
6. To be able to provide an LDAP backend to the KDC, run this command:
# ldapcfg -q -s "o=IBM,c=US"
#
We do this so that we do not have to use a legacy or file-based backend. For
a detailed description of why this is done, refer to the NAS Admin Guide IBM
Network Authentication Service Version 1.4 for AIX, Linux, and Solaris
Administrator’s and User’s Guide delivered with the krb5.doc.en_US file set.
7. We now start the IBM Directory Server to enable us to add the Kerberos
schema:
# ibmdirctl -D cn=admin -w succ3ss start
Start operation succeeded
#

8. We add the schema:
# ldapmodify -h localhost -D cn=admin -w succ3ss -f\
/usr/krb5/ldif/IBM.KRB.schema.ldif -v -c
#
9. We now create a schema for the KDC realm. Create
/usr/ldap/etc/realm_add_ibm.ldif. (The NAS Admin Guide is used as
reference for creating the schema file.) Add the following lines to the new file.
Example 5-55 Sample LDIF file for the KDC realm

# The suffix "ou=, o=, c=" should be defined before attempting to
# load this data. Or change the suffix to be an already defined object.
# Change all references of YOURHOSTNAME.DOMAINNAME to be your realm name
#
# version: 1
dn: o=IBM, c=US

objectclass: top
objectclass: organization
o: IBM
dn: krbrealmName-V2=REALM1.IBM.COM, o=IBM, c=US

objectclass: KrbRealm-V2
objectclass: KrbRealmExt
krbrealmName-V2: REALM1.IBM.COM
krbprincSubtree: krbrealmName-V2=REALM1.IBM.COM, o=IBM, c=US
krbDeleteType: 3
dn: cn=principal, krbrealmName-V2=REALM1.IBM.COM, o=IBM, c=US

objectclass: container
cn: principal
dn: cn=policy, krbrealmName-V2=REALM1.IBM.COM, o=IBM, c=US

cn: policy
The sample file contents are shown in “LDIF sample file for KDC” on
page 270.
10.We now modify the dn: o=IBM, c=US container by adding the schema.
Example 5-56 Adding the modified realm schema into LDAP

# ldapmodify -a -h localhost -D cn=admin -w succ3ss -f\
/usr/ldap/etc/realm_add_ibm.ldif -v -c
ldap_init(localhost, 389)
add objectclass:

BINARY (3 bytes) top
BINARY (12 bytes) organization
add o:
BINARY (4 bytes) IBM
adding new entry o=IBM, c=US
add objectclass:
BINARY (11 bytes) KrbRealm-V2
BINARY (11 bytes) KrbRealmExt
add krbrealmName-V2:
BINARY (15 bytes) REALM1.IBM.COM
add krbprincSubtree:
BINARY (45 bytes) krbrealmName-V2=REALM1.IBM.COM, o=IBM, c=US
add krbDeleteType:
BINARY (1 bytes) 3
adding new entry krbrealmName-V2=REALM1.IBM.COM, o=IBM, c=US
add objectclass:
BINARY (9 bytes) container
add cn:
BINARY (9 bytes) principal
adding new entry cn=principal, krbrealmName-V2=REALM1.IBM.COM, o=IBM, c=US
add objectclass:
BINARY (9 bytes) container
add cn:
BINARY (6 bytes) policy
adding new entry cn=policy, krbrealmName-V2=REALM1.IBM.COM, o=IBM, c=U
#
11.At this point we can verify the IBM Directory using a simple ldap query that
shows all available container names in the newly created LDAP directory.
Example 5-57 Verify the LDAP naming contexts

# ldapsearch -b "" -s base "objectclass=*" namingcontexts
namingcontexts=CN=SCHEMA
namingcontexts=CN=LOCALHOST
namingcontexts=CN=AIXDATA
namingcontexts=CN=PWDPOLICY
namingcontexts=CN=IBMPOLICIES
namingcontexts=O=IBM,C=US
#

5.9.3 Configure the KDC server with LDAP backend
1. Make sure that the NAS V1.4 server file sets are installed on the system. The
next step is to configure the KDC server.
Example 5-58 Creating the KDC server using mkkrb5srv

# mkkrb5srv -r REALM1.IBM.COM -s nfs407.itsc.austin.ibm.com \
-d itsc.austin.ibm.com -a admin/admin -l nfs407.itsc.austin.ibm.com \
-u “cn=admin” -p succ3ss
Fileset Level State Description

----------------------------------------------------------------------------
Path: /usr/lib/objrepos
Server
Path: /etc/objrepos
Server
The -s option is not supported.
The administration server will be the local host.
Creating /var/krb5/krb5kdc/kdc.conf...
Creating database files...
Initializing database 'LDAP' for realm 'REALM1.IBM.COM'
master key name 'K/[email protected]'
Attempting to bind to one or more LDAP servers. This may take a while...
You are prompted for the database Master Password.
Enter database Master Password:
Re-enter database Master Password to verify:
WARNING: no policy specified for admin/[email protected];
ACL restrictions.
Enter password for principal "admin/[email protected]":
Re-enter password for principal "admin/[email protected]":
Principal "admin/[email protected]" created.
Creating keytable...
Creating /var/krb5/krb5kdc/kadm5.acl...
Starting krb5kdc...
krb5kdc was started successfully.
Starting kadmind...
kadmind was started successfully.

Restarting kadmind and krb5kdc
#
2. Check that the KDC server process krb5kdc and kadmind has been started.
Example 5-59 Verify the KDC server processes

# ps -ef |grep krb5 |grep -v grep
root 290896 1 0 13:57:30 - 0:00 /usr/krb5/sbin/krb5kdc
root 512076 1 0 13:57:30 - 0:00 /usr/krb5/sbin/kadmind
#
3. Now we log on to the KDC to check that a ticket can be issued.
Example 5-60 Verification of the KDC administrator

# kinit admin/admin
#
# klist

#
4. To be able to add, modify, and delete users and groups using the
mkgroup/mkuser -R KRB5LDAP commands, we have to add the following
authentication grammar to the /usr/lib/security/methods.cfg file and append
the following lines.
Example 5-61 Authentication grammar in file /usr/lib/security/methods.cfg

LDAP:
program = /usr/lib/security/LDAP
program_64 =/usr/lib/security/LDAP64
KRB5:
program = /usr/lib/security/KRB5
KRB5LDAP:
options = db=LDAP,auth=KRB5
5. In addition, the local LDAP security client daemon must be running on the
system. This can be accomplished with the following command.

Example 5-62 Enable the local security client daemon
# mksecldap -c -h nfs407.itsc.austin.ibm.com -a cn=admin -p succ3ss
#
6. To verify the LDAP security client daemon, you can use the ls-secldapclntd
command.
Example 5-63 Output of command ls-secldapclntd

# /usr/sbin/ls-secldapclntd
ldapservers=nfs407.itsc.austin.ibm.com
ldapport=389
ldapversion=3
userbasedn=ou=aixuser,cn=aixsecdb,cn=aixdata
groupbasedn=ou=aixgroup,cn=aixsecdb,cn=aixdata
idbasedn=cn=aixid,ou=system,cn=aixsecdb,cn=aixdata
usercachesize=1000
usercacheused=0
groupcachesize=100
groupcacheused=0
cachetimeout=300
heartbeatT=300
numberofthread=10
alwaysmaster=no
authtype=UNIX_AUTH
searchmode=ALL
defaultentrylocation=LDAP
ldaptimeout=60
userobjectclass=account,posixaccount,shadowaccount,aixauxaccount
groupobjectclass=posixgroup,aixauxgroup
#
At this point, the KDC and IBM Directory Server initial setup is complete.
In order to create users in the Kerberos realms and store the user/group
identification information in the RFC2307, we run the following commands:
1. We added the root user to the realm to enable the AIX mkuser command to
access the LDAP backend database.
Example 5-64 Creation of the root principal

# kinit admin/admin
#
# klist

#
# /usr/krb5/sbin/kadmin.local
kadmin.local:
kadmin.local: add_principal root/nfs407.itsc.austin.ibm.com
WARNING: no policy specified for
root/[email protected];
ACL restrictions.
Enter password for principal "root/[email protected]":
Re-enter password for principal
"root/[email protected]":
Principal "root/[email protected]" created.
kadmin.local: exit
#
# kdestroy
#
2. The next step is to verify the newly created principal. In addition, we need to
have valid tickets to perform all further user creation steps (shown in
Example 5-65).
Example 5-65 Verify the newly created principle

# kinit root/nfs407.itsc.austin.ibm.com
Password for root/[email protected]:
#
# /usr/krb5/bin/klist
Default principal: root/[email protected]:

08/06/04 10:39:45 08/06/04 13:39:45
root/[email protected]
#
3. Now we can create the users and groups as in Example 5-66.
Example 5-66 Creation of user and group in the KDC with LDAP backend
# mkgroup -R KRB5LDAP -a id=1400 eng
#
# mkuser -R KRB5LDAP id='6023' pgrp='staff' groups='eng' \ home='/home/sally'
shell='/bin/ksh' gecos='NFS V4 KDC Test user sally' sally
#
# passwd -R KRB5LDAP sally

sally's Old password:
sally's New password:
Enter the new password again:
#
4. Verify that the user sally is created correctly by executing the lsuser sally
command.
Example 5-67 Output of command lsuser for user sally

# lsuser -R KRB5LDAP sally
sally id=6023 pgrp=staff groups=staff,eng home=/home/sally shell=/bin/ksh
gecos=NFSv4 KDC Test user sally login=true su=true rlogin=true telnet=true
daemon=true admin=false sugroups=ALL admgroups= tpath=nosak ttys=ALL expires=0
auth1=SYSTEM auth2=NONE umask=22 registry=KRB5LDAP SYSTEM=KRB5LDAP OR compat
logintimes= loginretries=0 pwdwarntime=0 account_locked=false minage=0 maxage=0
maxexpired=-1 minalpha=0 minother=0 mindiff=0 maxrepeats=8 minlen=0
histexpire=0 histsize=0 pwdchecks= dictionlist= fsize=2097151 cpu=-1
data=262144 stack=65536 core=2097151 rss=65536 nofiles=2000 time_last_login=0
time_last_unsuccessful_login=0 unsuccessful_login_count=0 roles=
[email protected] [email protected]
krb5_realm=REALM1.IBM.COM maxage=0 expires=0 krb5_last_pwd_change=1091805538
admchk=false krb5_attributes=requires_preauth
krb5_mod_name=root/[email protected]
krb5_mod_date=1091805538 krb5_kvno=2 krb5_mkvno=0
krb5_max_renewable_life=604800 time_last_login=0 time_last_unsuccessful_login=0
unsuccessful_login_count=0 krb5_names=sally:nfs407.itsc.austin.ibm.com
#
5. To verify that the user ID is valid, try to log on as sally and display the
Kerberos ticket details.
Example 5-68 Integrated logon with user sally

# telnet nfs407
Trying...
Connected to nfs407.itsc.austin.ibm.com.
Escape character is '^]'.
telnet (nfs407)
AIX Version 5
(C) Copyrights by IBM and by others 1982, 2004.
login: sally
sally's Password:
$
$ /usr/krb5/bin/klist
Ticket cache: FILE:/var/krb5/security/creds/[email protected]_6023

08/06/04 10:39:45 08/06/04 13:39:45 kadmin/[email protected]
$
The user sally can log on to any machine in the Kerberos realm
REALM1.IBM.COM without being defined as a local user (if the client has been
set up with the previously mentioned authentication grammar). The user’s
identification will be stored in LDAP RFC2307.
Also, because user sally has a Kerberos ticket granted automatically during the
AIX logon process, she can have access to all NFS V4 mounted file systems with
sec=krb5 option if provided ACLs are set for access.
All further steps to set up an NFS V4 server were described in previous sections.
5.9.4 Configure the NFS V4 client for integrated login services

We installed AIX 5.3 using the base install media. We now install the following
additional file sets from the AIX 5.3 Expansion Pack:
򐂰 krb5.client.rte
򐂰 modcrypt.base
Example 5-69 Output of NAS file set installation verification

#lslpp -l | grep -i krb5
krb5.client.rte 1.4.0.0 COMMITTED Network Authentication Service
krb5.client.samples 1.4.0.0 COMMITTED Network Authentication Service
#
#lslpp -l |grep modcrypt
modcrypt.base.includes 5.3.0.0 COMMITTED Cryptographic Library Include
modcrypt.base.lib 5.3.0.0 COMMITTED Cryptographic Library
(libmodcrypt.a)
#
1. The plan is to use AIX integrated login within KDC and IBM Directory Server
as the LDAP backend, so we install the LDAP client software from the AIX
5.3. Base media.
Example 5-70 Output of LDAP client file set installation

# installp -agXd . ldap.client
--------------------
-------------------------------------------------------------------------------
ldap.client.rte 5.2.0.0 USR APPLY SUCCESS

ldap.client.adt 5.2.0.0 USR APPLY SUCCESS
ldap.client.rte 5.2.0.0 ROOT APPLY SUCCESS
X11.adt.lib 5.3.0.0 USR APPLY SUCCESS
#
2. Check that the system uses the correct PATH to locate the NAS binaries.
Example 5-71 Output of command type kinit

# type kinit
kinit is /usr/krb5/bin/kinit
#
3. Now we can configure this system into our NAS infrastructure as a client
system using the mkkrb5clnt command, delivered by AIX BOS file set
bos.rte.security. (nfs407.itsc.austin.ibm.com is the KDC serving the Kerberos
realm REALM1.IBM.COM and also the IBM Tivoli Directory Server.)
Example 5-72 Output of Kerberos client configuration using mkkrb5clnt

# mkkrb5clnt -c nfs407.itsc.austin.ibm.com -r REALM1.IBM.COM -s
nfs407.itsc.austin.ibm.com -d itsc.austin.ibm.com -l nfs407.itsc.austin.ibm.com
-i files -A -K -T
Configuring fully integrated login
Authenticating as principal admin/admin with existing credentials.
host/[email protected];
ACL restrictions.
Principal "host/[email protected]" created.
Administration credentials NOT DESTROYED.

Making root a Kerberos administrator
Authenticating as principal admin/admin with existing credentials.
root/[email protected];
ACL restrictions.
Enter password for principal "root/[email protected]":
Re-enter password for principal
"root/[email protected]":
Principal "root/[email protected]" created.
Administration credentials NOT DESTROYED.

Configuring Kerberos as the default authentication scheme
Cleaning administrator credentials and exiting.
#
4. We verify that the newly created principal can be used to authenticate with
Kerberos.
Example 5-73 Principal root verification

# kinit root/nfs405.itsc.austin.ibm.com
Password for root/[email protected]:
root@nfs405 [ppc] klist
Default principal: root/[email protected]

#
Note: This principal cannot be used for NFS V4 authentication. You still
must create the service principal of type nfs/<FQDN>@REALM.
If using the config.krb5 command, delivered with IBM NAS V1.4, the
configuration will be valid, but the principal nfs/<FQDN>@REALM will not be
created.
5. The next step is to add an NFS service (machine) principal and create a
keytab file for the NFS V4 client nfs402.itsc.austin.ibm.com. This is done by
carrying out the following process.
Example 5-74 Adding the NFS service principal

# kadmin -p admin/admin
Authenticating as principal admin/admin with password.
kadmin:
kadmin: add_principal -randkey -e des-cbc-crc:normal \
ACL restrictions.
kadmin:
DES cbc mode with CRC-32 added to keytab

kadmin:
kadmin: quit
#
6. Verify that the machine principal can log on using the keytab file that we just
generated.
Example 5-75 Verification of the NFS service principal

# kinit -t /etc/krb5/krb5.keytab nfs/nfs402.itsc.austin.ibm.com
# klist
#
7. To create the hostkey map file, change the NFS domain and enable
RPCSEC_GSS by executing the commands shown in Example 5-76.
Example 5-76 Enable NFS V4 on the system nfs402

# mkdir -p /etc/nfs
#
#
#
# /usr/sbin/chnfsrtd -a 'REALM1.IBM.COM' 'nfs.ibm.com'
#
# chnfs -S -B
#
# /etc/rc.nfs
#
8. The next step is to change the authentication grammar on

nfs402.itsc.austin.ibm.com so that all users, except the root user, use
integrated logon. We decided to change the Password Registry for the root
user to files and the authentication grammar to compat:
# chuser registry=files root
#
# chuser SYSTEM="compat" root
#
9. We also make the following changes to the default stanza in the
/etc/security/user file:
SYSTEM = "compat"

Change the line to:
SYSTEM = "KRB5LDAP OR compat"
We also need to add the following line to the default stanza in
/etc/security/user:
registry = KRB5LDAP
Both changes can be achieved by running the following command.
Example 5-77 Change default logon using chsec command

#chsec -f /etc/security/user -s default -a registry=KRB5LDAP
#
#chsec -f /etc/security/user -s default -a "SYSTEM=\"KRB5LDAP OR compat\""
#
10.Edit the /usr/lib/security/methods.cfg file and add the following lines.
Example 5-78 Add KRB5LDAP stanza to file /usr/lib/security/methods.cfg

LDAP:
program = /usr/lib/security/LDAP
program_64 =/usr/lib/security/LDAP64
KRB5:
program = /usr/lib/security/KRB5
KRB5LDAP:
options = db=LDAP,auth=KRB5
11.Communication to the IBM Directory Server from the client must be enabled:
#mksecldap -c -h nfs407.itsc.austin.ibm.com -a cn=admin -p succ3ss
#
With these settings in place, any principal that is defined in the realm
REALM1.IBM.COM can log on to nfs402.itsc.austin.ibm.com. Before trying to log
on to the system, we have to verify that the local LDAP security client daemon is
running. Otherwise, the logon attempt would fail as the AIX login process will not
be able to contact the LDAP server.
Example 5-79 Verification that secldapclntd is running

# /usr/sbin/ls-secldapclntd
ldapservers=nfs407.itsc.austin.ibm.com
ldapport=389
ldapversion=3
userbasedn=ou=People,cn=aixdata
groupbasedn=ou=Groups,cn=aixdata
idbasedn=cn=aixid,ou=System,cn=aixdata
usercachesize=1000

usercacheused=0
groupcachesize=100
groupcacheused=0
cachetimeout=300
heartbeatT=300
numberofthread=10
alwaysmaster=no
authtype=UNIX_AUTH
searchmode=ALL
defaultentrylocation=LDAP
ldaptimeout=60
userobjectclass=account,posixaccount,shadowaccount,aixauxaccount
groupobjectclass=posixgroup,aixauxgroup
#
The local LDAP security client daemon is running and we can try the integrated
logon for user sally.
Example 5-80 Sample integrated logon output

# telnet loopback
AIX Version 5
login: sally
sally's Password:
*******************************************************************************
* *
* *
* Welcome to AIX Version 5.3! *
* *
* *
* Please see the README file in /usr/lpp/bos for information pertinent to *
* this release of the AIX Operating System. *
* *
* *
*******************************************************************************
$
$ id
uid=6023(sally) gid=1(staff) groups=1400(eng)
$
$ lsuser -a registry sally
sally registry=KRB5LDAP
$
$ klist
Ticket cache: FILE:/var/krb5/security/creds/[email protected]_6023

08/09/04 18:35:34 08/09/04 21:35:34 kadmin/[email protected]
$
The user sally can log on to any machine in the Kerberos realm
REALM1.IBM.COM without being defined as a local user if the client has been
set up with the previously mentioned authentication grammar. The user’s
identification will be stored in LDAP RFC2307.
Also, because user sally has a Kerberos ticket granted automatically during the
AIX logon process, she can have access to all NFS V4 mounted file systems with
sec=krb5 option if ACLs are set for access.
5.10 Integrating NFS V4 with a Linux client

NFS V4 is shipped with the Linux kernel Version 2.6.5. For the purposes of our
example, we use Fedora Core 2 with kernel Version 2.6.5.1.358. This is the
current downloadable version of Fedora Core at the time of writing this book.
Note: As we write this book, the unmodified Fedora Core 2 Linux does not
contain a complete and working RPCSEC_GSS (Kerberos 5, LIPKEY,
SPKM-3) implementation. This will probably happen in Fedora Core 3 Linux.
Therefore we are unable to test this functionality in this book. You are welcome
to attempt to patch the 2.6.5.x kernel to a later kernel manually. This should
give you the missing functionality.
Linux was installed from the install media with the NFS services chosen at install
time. In this section, we look at setting up some basic NFS V4 scenarios using
the AUTH_SYS security mechanism.
The following definitions will be used in the example:

NFS client hostname = nfs408, OS: Fedora Core 2 with v2.6.5.1.358
kernel.

Figure 5-3 Sample environment
There are many configuration files that make up a Linux NFS client and you
should make sure that they exist before you proceed:
򐂰 /etc/fstab
򐂰 /etc/auto.master
򐂰 /etc/idmapd.conf
򐂰 /etc/gssapi_mech.conf
򐂰 /etc/init.d/portmap
򐂰 /etc/init.d/rpcidmapd
򐂰 /etc/init.d/rpcgssd (required on the client when RPCSEC_GSS is used)
5.10.1 NFS server and client setup

The ID mapper daemon is required on the client. It maps NFS V4
username@domain user strings back and forth into numerous UIDs and GIDs.
The client and server domains must match. This is set in the /etc/idmapd.conf
file.
1. The domain on our server is itsc.austin.ibm.com, so we modify
/etc/idmapd.conf to reflect this. Figure 5-4 on page 178 shows this.

Figure 5-4 /etc/idmapd.conf showing the nfsdomain
2. Our NFS server has the directories exported as shown in Figure 5-5. You can
edit /etc/exports using either a text editor or smitty mknfsexp to create your
exports list.
Tip: Manually editing /etc/exports is easier if you need to export more than
one file system. Run exportfs -va after applying your changes to the
exports file. With smitty, add one file system at a time.
Figure 5-5 Samples exported directories on the NFS server

Name resolution between the server and client must work. This is an NFS V4
requirement.
Figure 5-6 NFS server name resolution of NFS client
Figure 5-7 NFS client name resolution of NFS server

3. Make sure that the /etc/gssapi_mech.conf file exists. It should have been
installed by default.
Figure 5-8 /etc/gssapi_mech.conf on NFS client
4. We use /nfs as our NFS mount point.
Figure 5-9 NFS mount point on the NFS client

5. Ensure that all NFS-related scripts start and stop automatically. The
chkconfig utility can be used for this.
Figure 5-10 chkconfig to make sure all services start and stop automatically
6. Ensure that all of the right daemons are restarted or stopped on the NFS
client.
Figure 5-11 Confirm the correct daemons are restarted or stopped on the NFS client

Figure 5-12 tells what NFS-related daemons are running on the client, and what
UDP and TCP ports they are listening on.
Figure 5-12 Checking NFS daemons and what ports they are listening on
5.10.2 Read-only NFS V4 mount

We first look at carrying out a read-only NFS V4 mount.
1. On the server, modify /etc/exports and add a file system to export in read-only
mode.
Figure 5-13 /etc/exports on the AIX NFS server

2. Export the directory specified on /etc/exports.
Figure 5-14 Exporting the directory on the NFS server
3. Mount the exported directory on the NFS client. We first mount it manually
and then show how the /etc/fstab file can be modified to allow mounts via
directory name.
Figure 5-15 Manually mounting an NFS V4 file system in read-only mode

4. Look at the file system that we have just mounted. If you try to write to the
directories, you will get an error. (Remember, we mounted the directory in
read-only mode.)
Figure 5-16 Checking to see whether read-only mode works
5. Unmount the NFS file system that was mounted in the previous steps.
Figure 5-17 Unmounting the NFS V4 file system

6. Append the following line to /etc/fstab:
nfs403:/ /nfs nfs4 ro,hard,intr,proto=tcp,port=2049,noauto 0 0
7. Mount the NFS V4 file system using the information in /etc/fstab.
Figure 5-18 NFS mount using entry in /etc/fstab
As you can see, Linux can mount AIX NFS V4 exports without any problems. We
now unmount the file system as done in step 5 on page 184 to continue with the
next section.
5.10.3 Read/write NFS V4 mounts on Linux

We can now look at the most common type of NFS mount: the read/write mount.
For this example, we add another export to our server. We create the directory
with the sticky bit set. This enables any remote user to read and write to the
directory. If the root user on the NFS V4 client writes to the directory, the
ownership will be changed to nfsnobody. Root squashing is turned on by default
(which means that the root user on the NFS client does not have root privileges
on the NFS server).
The NFS server already has a user called sally, with UID 401, created on it. We
create the same user and UID on our NFS V4 client. For this example, we are not
using LDAP. If you use LDAP, you can use users created there for testing.
1. Create user sally on the NFS V4 client.
Figure 5-19 Adding a user on the Linux client to match the NFS server

2. We modify /etc/exports on the server to export user sally’s home directory
and export the file system.
Figure 5-20 Adding and exporting user sally’s home directory on the NFS V4 server
3.Next, we modify /etc/fstab on the client so that it looks like this:

nfs403:/ /nfs nfs4 rw,hard,intr,proto=tcp,port=2049,noauto 0 0
4. We then mount the file system.
Figure 5-21 Mounting file system on NFS client

5. User sally should now be able to write to /nfs/exports/home/sally.
Figure 5-22 Testing write as user sally
6. We can try writing to the directory owned by sally as root.
Figure 5-23 Test on user sally’s home directory as the root user
5.10.4 Pseudo-file system in NFS V4 Linux client

Remember, the pseudo-file system functionality in NFS V4 enables the server to
present a single, seamless view of all exported file systems to an NFS V4 client.
Preparing the AIX NFS V4 server for pseudo-FS
Important: Ensure that the NFS V4 client has no directories mounted from the
NFS V4 server before proceeding. Use the umount command to unmount it.

To set the NFS root on the server:
1. Unexport all file systems.
Example 5-81 Unexporting all file systems on the NFS server

# exportfs -vu
/exports/project/proja -vers=4,ro
/exports/home/sally -vers=4,rw
/exports/acltest -vers=4:3,rw,root=nfs401:nfs406
/exports/acltest-jfs -vers=4:3,rw,root=nfs406
#
2.Stop NFS.
Example 5-82 Shutting down the NFS daemons cleanly

# /etc/nfs.clean
nfs_clean: Stopping NFS/NIS Daemons
0513-044 The nfsd Subsystem was requested to stop.
0513-044 The biod Subsystem was requested to stop.
0513-006 The rpc.lockd was requested to stop
0513-044 The rpc.statd Subsystem was requested to stop.
0513-044 The rpc.mountd Subsystem was requested to stop.
#
3. Stop the nfsrgyd subsystem.
Example 5-83 Stopping the registry daemons

# stopsrc -s nfsrgyd
0513-044 The nfsrgyd Subsystem was requested to stop.
#
4. Change the NFS root to /exports and confirm the change.
Example 5-84 Setting the root node on the NFS server

# /usr/sbin/chnfs -r '/exports' '-B'
#
# nfsd -getnodes
#root:public
#

5. Restart the NFS daemons.
Example 5-85 Restarting the NFS daemons

# /etc/rc.nfs
0513-059 The nfsd Subsystem has been started. Subsystem PID is 315644.
0513-059 The rpc.mountd Subsystem has been started. Subsystem PID is 319678.
0513-029 The rpc.lockd Subsystem has been started. Subsystem PID 266398
#
6. Export all of the file systems.
Example 5-86 Exporting all file systems on the NFS server

# exportfs -va
exportfs: 1831-187 re-exported /exports/project/proja
exportfs: 1831-187 re-exported /exports/home/sally
#
We are now ready to move on to the client.
Mounting the pseudo-file system on the NFS V4 Linux client

The process of mounting the pseudo-file system on the client is the same as in
the previous examples. The entry in /etc/fstab will remain the same.
1. Mount the pseudo-file system on the client.
Figure 5-24 Mounting the pseudo-file system on the NFS V4 client

2. We can now traverse the mount point to see what we have underneath.
Figure 5-25 Client view of the pseudo-root
The output shows that, although the server is exporting /exports/home/sally and
/export/project/proja, the client only sees these as /nfs/home/sally and
/nfs/project/proja. This is as a direct result of setting the nfs root on the server to
/exports.
5.11 Windows KDC and NFS V4 AIX 5.3

In this section, we set up a basic Active Directory and using the built-in KDC on
the AIX 5.3 systems with the NFS V4 RPCSEC_GSS security mechanism.
We installed Windows Server 2003 Standard Edition, which includes a KDC

server built in with the Active Directory. The Windows server was installed from
the install media with no additional options. We also installed the Windows
Server 2003 Resource Kit and the Windows Server 2003 Support Tools.

KDC server hostname = nfs409.kdc.austin.ibm.com, OS: Windows
2003 Server Standard Edition

Figure 5-26 Sample AIX environment with Windows 2003 Active Directory
1. We installed Active Directory through the Windows Server configuration

wizard on the Windows Server 2003 Standard Server.
Note: The Kerberos realm name is derived from the Active Directory
domain name. The realm name is the domain name converted to
uppercase (For the example.com domain name, the realm name is
EXAMPLE.COM.) In our case this is KDC.ITSC.AUSTIN.IBM.COM on
server nfs409.kdc.austin.ibm.com.

The following figures show the chosen setup.
Figure 5-27 DNS Active Directory registration in Windows 2003 Server
Figure 5-28 Active Directory Installation Wizard summary

Figure 5-29 Active Directory Installation Wizard confirmation
2. Download and install the public Windows MIT Kerberos V5 utilities, which are
available at:
http://web.mit.edu/kerberos/www/
3. We created a sample user called sally within Windows Active directory. This
is achieved by carrying out the following steps:
a. Click Start.
b. Click All Programs.
c. Click Administrative Tools.
d. Click Active Directory Users and Computers.

4. We used the same method to create the user for the NFS V4 server (nfs405)
and client (nfs403). See the following figures for details.
Figure 5-30 User creation main panel
Figure 5-31 User creation password panel

5. Next we use the Leash Kerberos Ticket Manager (leash32.exe) delivered with
the MIT Kerberos V5 utilities to get a ticket for user sally.
Figure 5-32 Kerberos ticket verification on Windows Server 2003
Use the following guidelines to configure the AIX NFS V4 server hosts.
Services running on UNIX or Linux systems can be configured with service

instance accounts in the Active Directory. This allows full interoperability.
Kerberos clients and servers on UNIX systems can authenticate using the
Windows Server 2003 Kerberos server, and Windows 2000 Professional-based
clients can authenticate to Kerberos services that support GSS-API.
Unlike Kerberos principal names, Windows Server 2003 account names are not
multipart. Because of this, it is not possible to directly create an account of the
name nfs/[email protected]. Such a principal instance is created
through the service principal name mappings.
Create a service instance account in the Active Directory:

1. Use the Active Directory Management tool to create a user account for the
NFS V4 server; for example, create an account with the name nfs405.
2. Use the ktpass tool to set up an identity mapping for the user account. Use
this command (from the Windows COMMAND prompt):
ktpass -princ service-instance@REALM -mapuser account-name -pass
password -out UNIXmachine.keytab
The format of the Kerberos service-instance name is: nfs/host@realm_name,
for example:
ktpass.exe -princ nfs/[email protected] -mapuser nfs405
-pass nfs405 -out c:\misc\nfs405.keytab

In this case, an account is created with the name sample, and a service
principal name mapping is added for sample/UNIX_SRV.example.com. This is
the purpose of using ktpass with the -princ and -mapuser switches.
Figure 5-33 Sample output of ktpass tool
3. Verify the created SPN (service principal name) using the setspn command
on Windows. The SPN must show the NFS service principal; otherwise, the
systems will be not be able to authenticate with the KDC.
Figure 5-34 Output of setpsn tool

Note: On Windows, you cannot map multiple service instances to the
same user account. Therefore we recommend that you do not add another
instance with the same name.
4. Create the /etc/krb5/krb5.conf file on the AIX 5.3 server.
Attention: Be sure to have set all host names using the Fully Qualified
Domain Name as used within your Active Directory; otherwise, the gssd
daemon will not be able to authenticate with the Windows KDC. In our
example, all systems used <hostname>.kdc.austin.ibm.com.
Example 5-87 Sample Kerberos configuration file for the Windows Active Directory
[libdefaults]
default_realm = KDC.ITSC.AUSTIN.IBM.COM
default_tkt_enctypes = des-cbc-md5 des-cbc-crc
default_tgs_enctypes = des-cbc-md5 des-cbc-crc
[realms]
KDC.ITSC.AUSTIN.IBM.COM = {
kdc = nfs409.kdc.itsc.austin.ibm.com:88
admin_server = nfs409.kdc.itsc.austin.ibm.com:749
default_domain = kdc.itsc.austin.ibm.com
}
[domain_realm]
.kdc.itsc.austin.ibm.com = KDC.ITSC.AUSTIN.IBM.COM
nfs409.kdc.itsc.austin.ibm.com = KDC.ITSC.AUSTIN.IBM.COM
[logging]
Note: The default encryption type entries, default_txx_enctype, are optional.

However, if the Kerberos client receives an encryption type error, set the
default encryption type to either des-cbc-md5 or des-cbc-crc.
5. Merge the keytab file with the /etc/krb5/krb5.keytab file on the AIX System by
copying the keytab file to the AIX system (for example, using binary FTP),
then entering this command in the directory that it has been copied to:
/usr/krb5/sbin/ktutil

Then, at the ktutil command prompt, enter this command (where
UNIX-SRV.keytab is the name of the keytab file you created in step 2 on
page 195):
rkt UNIX_SRV.keytab
Exit kutil by entering quit at the kutil prompt.
Example 5-88 Adding the Windows keytab file to the AIX System
# /usr/krb5/sbin/ktutil
ktutil: rkt nfs403.keytab
ktutil: l
slot KVNO Principal
------ ------ ------------------------------------------------------
ktutil: wkt /etc/krb5/krb5.keytab
ktutil: quit
#
6. Verify that the system can get tickets.
Example 5-89 Verification of the keytab file

# kinit -kt /etc/krb5/krb5.keytab nfs/[email protected]
#
# klist

08/04/04 17:05:20 08/05/04 03:05:20
krbtgt/[email protected]
Renew until 08/05/04 17:05:20
#
7. Set up your NFS V4 AIX server and client as already described.
Example 5-90 Configuration of the NFS V4 Windows KDC client

# chnfsdom nfs.ibm.com
#
#chnfsdom
#
#nfshostkey -p nfs/[email protected] -f /etc/krb5/krb5.keytab
#
#nfshostkey -l
#

#chnfsrtd -a KDC.ITSC.AUSTIN.IBM.COM nfs.ibm.com
#
#chnfsrtd
kdc.itsc.austin.ibm.com nfs.ibm.com
#
#chnfs -S -B
#
#/etc/rc.nfs
8. Now, we can mount the exported file system.
Example 5-91 Sample mount output

# mount -o vers=4,sec=krb5 nfs403:/exports/home /nfs
#
# kinit sally
#
# cd /nfs
#
# ls
bob joe lost+found mary sally
#
For further details about configuring AIX Integrated Login using Kerberos and
Windows Active Directory, see the AIX 5.3 online documentation chapter about
security and authenticating to AIX using Kerberos.
5.12 Setting up Kerberos cross-realm access

If you have several realms and you want to enable access throughout them, then
you should activate cross-realm access on your server and client systems.
Note: Setting up cross-realm access is necessary to support systems

belonging to multiple realms sharing access to secure data. Multiple realms
can reside within a single NFS V4 domain to share user and group definitions.
If the realms reside in different NFS V4 domains, then use of foreign domain
identity mappings is likely required to provide access to secured data.

KDC server hostname = nfs407, OS = AIX 5.3

KDC server hostname = nfs403, OS = AIX 5.3
Figure 5-35 gives a better view of the setup.
Figure 5-35 Sample cross realm layout
The goal of this sample is to show that a principal with tickets in

REALM1.IBM.COM is capable of accessing NFS data exported from a server in
REALM2.IBM.COM.
5.12.1 Add the krbtgt service principal to every KDC server

For a KDC in one realm to authenticate Kerberos users in a different realm, it
must share a key with the KDC in the other realm. In both databases, there must
be krbtgt service principals for realms.These principals should all have:
򐂰 The same passwords

򐂰 Key version numbers
򐂰 Encryption types
For example, if the administrators of REALM1.IBM.COM and REALM2.IBM.COM

want to enable authentication across the realms, they would run the following
commands on the KDCs in both realms:
kadmin: addprinc -requires_preauth krbtgt/[email protected]
and
Even if most principals in a realm are generally created with the requires_preauth
flag enabled, this flag is not desirable on cross-realm authentication keys
because doing so makes it impossible to disable preauthentication on a
service-by-service basis. Disabling it as in the example above is recommended.
It is also very important that these principals have good passwords. MIT
recommends that the TGT principal passwords be at least 26 characters of
random ASCII text.
Example 5-92 shows the command and output as performed on our KDC server
in realm REALM1.IBM.COM.
Example 5-92 kadmin output on nfs407 KDC server

WARNING: no policy specified for krbtgt/[email protected];
ACL restrictions.
Enter password for principal "krbtgt/[email protected]":
Re-enter password for principal "krbtgt/[email protected]":
Principal "krbtgt/[email protected]" created.
ACL restrictions.
kadmin:
kadmin: getprinc krbtgt/[email protected]
Principal: krbtgt/[email protected]
Last password change: Thu Aug 19 18:14:06 CDT 2004
Last modified: Thu Aug 19 18:14:07 CDT 2004 (admin/[email protected])

Number of keys: 4
Key: vno 1, Triple DES cbc mode with HMAC/sha1,
no salt
Key: vno 1, ArcFour with HMAC/md5,
no salt
Key: vno 1, AES-256 CTS mode with 96-bit SHA-1 HMAC,
no salt
Key: vno 1, DES cbc mode with RSA-MD5,
no salt
Attributes:
Policy: [none]
The same has to be done on the KDC server in the other realm.
Example 5-93 kadmin output on nfs403 KDC server

# /usr/krb5/sbin/kadmin -p admin/admin
ACL restrictions.
ACL restrictions.
kadmin: getprinc krbtgt/[email protected]

Principal: krbtgt/[email protected]
Last password change: Thu Aug 19 18:18:10 2004
Last modified: Thu Aug 19 18:18:10 2004 (admin/[email protected])

Number of keys: 4
Key: vno 1, Triple DES cbc mode with HMAC/sha1,
no salt
Key: vno 1, ArcFour with HMAC/md5,
no salt
Key: vno 1, AES-256 CTS mode with 96-bit SHA-1 HMAC,
no salt
Key: vno 1, DES cbc mode with RSA-MD5,
no salt
Attributes:
Policy: [none]
kadmin:
5.12.2 Kerberos configuration file changes on the KDC server,

NFS V4 client and server
1. We have to add the new realm to /etc/krb5/krb5.conf using vi to modify the
two stanzas [realms] and [domain_realm]. Example 5-94 shows the changed
contents of the realms stanza.
Example 5-94 New realms stanza entries

[realms]
REALM1.IBM.COM = {
default_domain = itsc.austin.ibm.com
}
REALM2.IBM.COM = {
}
2. Now we change the domain_realm stanza as shown in Example 5-95.
Example 5-95 Correct domain_realm stanza for cross-realm access

[domain_realm]

Important: Do not try to map one DNS domain into a second Kerberos realm
map. Example 5-96 shows an invalid entry in the krb5.conf file.
Example 5-96 Incorrect domain_realm stanza for cross-realm access

[domain_realm]
.itsc.austin.ibm.com = REALM1.IBM.COM
3. Now we restart the gssd daemon using stopsrc -g gssd and startsrc -g
gssd.
Note: Due to internal kernel caching, it may take up to two minutes until
gssd recognizes the change.
5.12.3 Add NFS domain-to-realm map on NFS V4 client and server

The NFS domain is the same, so we do not need identity mapping.
To enable the nfsrgyd daemon to map identification requests, we have to add an

additional NFS domain-to-realm map into file /etc/nfs/realm.map; otherwise, the
NFS user nobody will be mapped.
Example 5-97 Cross-realm NFS domain-to-realm map file on nfs406 client

# chnfsrtd -a realm2.ibm.com itsc.austin.ibm.com
#
# chnfsrtd
# pg /etc/nfs/realm.map
Note: You have to recycle the nfsrgyd daemon for the new mapping to be
loaded.
5.12.4 Client access verification

Example 5-98 on page 205 shows verification from the client to a principal in the
other realm using the kinit command.

Example 5-98 kinit into the new added realm
# klist

#
# kinit [email protected]
# klist

Renew until 08/20/04 18:34:55
5.12.5 Client access mount using cross-realms

At this stage the Kerberos environment modification has been validated. But our
target is to access the server in REALM2.IBM.COM without having to use kinit
into the realm. That is shown in the next sample.
Example 5-99 Client mount to a server within realm2

# kdestroy
# klist
# mount
-------- --------------- --------------- ------ ------------ ---------------
/dev/home_lv /exports/home jfs Aug 19 18:29 rw,log=/dev/hd8
/dev/exname_lv /local/trans jfs Aug 19 18:29 rw,log=/dev/hd8
/dev/projA_lv /exports/project/projA jfs2 Aug 19 18:29
rw,log=/dev/loglv00

/dev/projB_lv /exports/project/projB jfs2 Aug 19 18:29
rw,log=/dev/loglv00
#
# kinit sally
# mount
-------- --------------- --------------- ------ ------------ ---------------
/dev/home_lv /exports/home jfs Aug 19 18:29 rw,log=/dev/hd8
/dev/exname_lv /local/trans jfs Aug 19 18:29 rw,log=/dev/hd8
/dev/projA_lv /exports/project/projA jfs2 Aug 19 18:29
rw,log=/dev/loglv00
/dev/projB_lv /exports/project/projB jfs2 Aug 19 18:29
rw,log=/dev/loglv00
#
# cd /nfs/exports2/home
# ls -ltr
total 48
drwxr-sr-x 2 sally staff 512 Jul 30 19:42 sally
drwxr-sr-x 2 bob staff 512 Jul 30 19:42 bob
drwxr-sr-x 2 mary staff 512 Jul 30 19:42 mary
drwxr-sr-x 2 joe staff 512 Jul 30 19:42 joe
drwxrwx--- 2 root system 512 Aug 16 11:09 lost+found
#

6
Chapter 6. Problem determination

This chapter provides guidance in carrying out problem determination with
NFS V4. It is beyond to the scope of this book to cover the possible problems, but
we will guide you through the thought process you should consider when trying to
debug NFS V4 problems.
We start by giving a brief introduction to the tools and aids available with AIX and
other third-party tools we have found useful. We then guide you through issues
that we encountered while writing the book and how we resolved them.
The sections covered in this chapter are as follows:

򐂰 Problem determination tools and techniques
򐂰 AIX problem determination tools and aids for NFS
򐂰 IBM NAS problem determination tools and aids
򐂰 IBM Tivoli Directory Server problem determination tools and aids
򐂰 Third-party problem determination tools and aids
򐂰 File system exports problem
򐂰 Mount problems
򐂰 Major GSS-API error codes
򐂰 Kerberos V5 status codes

6.1 Problem determination tools and techniques
Before we can carry out any kind of problem determination, we need to be
familiar with the tools that are available with AIX and provided by third parties.
Third-party tools such as Ethereal (free at http://www.ethereal.com) can be
very useful when you need to examine a tcpdump or an iptrace. Of course, AIX
provides all of the necessary tools to convert an iptrace to something readable to
the human eye, but it is not always easy to look through a large trace.
You may find additional useful information on this topic in Introduction to the IBM
Problem Determination Tools, SG24-6296.
6.2 AIX problem determination tools and aids for NFS

We begin by looking at the tools provided by AIX. For more about using these
tools, refer to the relevant man pages and AIX system reference documentation.
6.2.1 Enabling syslogd

NFS uses the syslog to write its error and debug information. Before you carry
out any problem determination, we recommend that you turn syslog logging on.
The following steps take you through the process of enabling syslog logging (if it
is not already enabled on your system).
Tip: We created a separate file system for the syslog output and mounted it to
/var/logs/. It is always good to do this, as logging daemons are notorious at
filling up file systems. Allowing the syslog daemon to write to the root file
system, then having it subsequently fill up, can lead to service interruption.
1. Edit the /etc/syslog.conf file and add an entry:

*.debug /var/logs/syslog/syslog.out rotate time 1d archive \
/var/logs/syslog/archive/
We used extra parameters in this example to enable us to automatically rotate
the log on a daily basis.
2. We now create the /var/logs/syslog and /var/logs/syslog/archive directories.
We also create the /var/logs/syslog/syslog.out log file (Example 6-1).
Example 6-1 Creating the necessary directories and log file

# mkdir /var/logs/syslog
#
# mkdir /var/logs/syslog/archive

#
# ls -ld /var/logs/syslog
drwxr-sr-x 3 root sys 512 Aug 10 18:59 /var/logs/syslog
#
# ls -ld /var/logs/syslog/archive
drwxr-sr-x 2 root sys 512 Aug 10 18:59 /var/logs/syslog/archive
#
# touch /var/logs/syslog/syslog.out
#
# ls -al /var/logs/syslog/syslog.out
-rw-r--r-- 1 root sys 168 Aug 10 19:00 /var/logs/syslog/syslog.out
#
3. We tell the syslogd that we have made changes to its configuration file and
that it should reread it.
# refresh -s syslogd
0513-095 The request for subsystem refresh was completed successfully.
#
Example 6-21 on page 220 shows sample entries in the syslog output file.
4. When you have completed your problem determination steps, you may want
to disable syslog logging. Doing so is usually a personal preference. Some
systems administrators prefer to have all logging turned on while others
choose to do it on demand. To disable syslog logging, comment out the line
that was added to /etc/syslog.conf in step 1 on page 208 and refresh the
syslogd as shown in Example 6-2.
Example 6-2 Disabling syslogd logging

#*.debug /var/logs/syslog/syslog.out rotate time 1d archive \
#
# refresh -s syslogd
0513-095 The request for subsystem refresh was completed successfully.
#
Using the setting for syslogd shown in Example 6-2 may log more information
than you need while running your system in production mode. You can set
syslog to log only ERROR in the /etc/syslog.conf file. This limits the amount of
data written to the log file. This is the new line in the /etc/syslog.conf file:
#*.error /var/logs/syslog/syslog.out rotate time 1d archive \
Chapter 6. Problem determination 209

6.2.2 Using iptrace and ipreport
iptrace is very useful when trying to debug problems that are not obvious. It is not
the purpose of this book to show you how to interpret an iptrace, but to show you
how to use the available tools. Follow these steps to create an iptrace:
1. Run the following command to start an iptrace:
startsrc -s iptrace -a “-a -d [source host] -b [LogFile]”
2. Recreate the problem scenario.
3. Stop the iptrace by running:
stopsrc -s iptrace
We now decode the iptrace file into a readable format using the ipreport
command. You can also use Ethereal to decode the trace file (see 6.5.2, “Using
the Ethereal utility” on page 212). We recommend either of two ways to decode
the iptrace file:
򐂰 The first method is to decode everything:
ipreport -v [LogFile] > [OutputFile]
򐂰 The second method is to decode RPC packets only:
ipreport -nsrv [LogFile] > [OutputFile]
If you are trying to debug NFS V2 or NFS V3 problems, then the second method
would be the correct option to choose. For NFS V4, we recommend that you use
the first method. This enables you to decode the Kerberos packets as well as all
other relevant information.
6.2.3 Using the fuser command

The fuser command lists the process numbers of local processes that use the
local or remote files specified by the file parameter. For block special devices, the
command lists the processes that use any file on that device. It can be useful
when trying to determine mount or umount problems.
6.2.4 Using the rpcinfo command

The rpcinfo command reports the status of remote procedure call (RPC)
servers. It is very useful when you need to check whether a system is running the
correct NFS programs.

6.2.5 Using the showmount command
The showmount command displays a list of all clients that have remotely mounted
file systems. The showmount -a command cannot show any NFS V4 exported file
systems. We recommend using the nfs4cl command on the NFS V4 client.
6.2.6 Using the nfs4cl command

The nfs4cl command displays or modifies current NFS V4 statistics and
properties. See the showmount command in the previous section.
6.2.7 Using the nfsstat command

The nfsstat command displays statistical information about NFS and RPC calls.
6.2.8 Using the errpt command

The errpt command generates a report of logged errors from the error log. It is
useful when you are trying to determine why a daemon is core dumping or not
starting. For the most detailed output, run the command in the following way:
errpt -a | more
6.3 IBM NAS problem determination tools

By default, krb5kdc and kadmind log directly to a file specified in the [logging]
stanza of the krb5.conf file. (The default location of this file is /var/krb5/log/.)
Logging to syslog instead can give you more control over the amount of logged
information. The krb5kdc process logs an informational message for every ticket
it issues. On a busy KDC, this can cause the log file to grow rapidly. If you log to
syslog, you can log only the KDC errors. Table 6-1 shows the default files.
Table 6-1 IBM NAS Server log files

Filename Description
krb5kdc.log Log file of the KDC server process

krb5kdc
kadmin.log Log file of the administrative server

process kadmind

6.4 Tivoli Directory Server problem determination tools
When a problem occurs that appears to be related to the IBM Tivoli Directory
Server, you should first check the following files for error messages. By default,
Tivoli Directory Server uses the /var/ldap directory to generate and write its error
and debug information. The files shown in Example 6-2 are used by default.
Table 6-2 IBM Tivoli Directory Server log files

Filename Description
ibmslapd.log Log file of the Tivoli Directory Server main

server process
db2cli.log Log file for Tivoli Directory Server DB2

client interface
If you want to have more information logged by Tivoli Directory Server, you can
change the level of variable ibm-slapdSysLogLevel in the /etc/ibmslapd.conf file.
6.5 Third-party problem determination tools

Here we will discuss tools that complement available AIX tools.
Restriction: Third-party tools are not supported by IBM. If you choose to use
these tools, you do so at your own risk.
6.5.1 Using the lsof command

lsof stands for list open files. It lists information about files that are currently
open by processes. The main difference between lsof and fuser is that lsof
takes files, file systems, and PIDs as arguments, but fuser only accepts files and
file systems. If fuser does not give you any output, try using lsof.
lsof for AIX is available for download from the following sites:
http://aixpdslib.seas.ucla.edu
http://www.bullfreeware.com
6.5.2 Using the Ethereal utility

Ethereal is used by network professionals around the world for troubleshooting,
analysis, software and protocol development, and education. It has all of the
standard features you would expect in a protocol analyzer, and several features

not seen in any other product. Its open source license enables experts in the
networking community to add enhancements.
Ethereal can be installed on a Windows or Linux system. It is capable of reading

the iptrace output.
Figure 6-1 Sample screen shot of Ethereal
Ethereal is available for download from:

http://www.ethereal.com
6.6 General NFS V4 problems

The following messages may appear when using NFS V4.
6.6.1 Warning: EIM is not configured

After installing AIX 5.3 and activating an NFS V4 NFS domain the error entry
shown in Example 6-3 on page 214 is logged into syslogd output file on the
server and client side.

Example 6-3 EIM syslogd error entry
Aug 16 11:19:51 nfs403 syslog: nfsrgyd: dlopen(/usr/lib/libeim.a(shr.o))
failed: No such file or directory
Aug 16 11:19:51 nfs403 syslog: nfsrgyd: Warning: EIM is not configured
Problem determination steps

Not applicable
Solution
The message is telling you that Enterprise Identity Mapping (EIM) is not
configured. If you are not using EIM, you can safely ignore the message. If you
are using EIM, consult your EIM documentation for further assistance.
6.6.2 Realm is already mapped to domain

Trying to add a new NFS domain-to-realm map using chnfsrtd results in the
message in Example 6-4.
Example 6-4 chnfsrtd message for duplicate domain

# chnfsrtd -a realm2.ibm.com nfs1.itsc.austin.ibm.com
Realm (realm2.ibm.com) is already mapped to domain (itsc.austin.ibm.com)
#

Use the chnfsrtd command without any argument to see the current settings.
Example 6-5 Output of chnfsrtd

# chnfsrtd
realm2.ibm.com nfs.itsc.austin.ibm.com
#
Solution
This behavior is expected. You cannot map the same realm to a second NFS
domain.

6.7 Exporting file systems
The following error messages may appear when exporting file systems.
6.7.1 Exportfs: cannot change the v4 root...

We changed the /etc/exports file and added the following line to adopt the new
pseudo-root FS:
/exports -nfsroot
The full /etc/exports file looked as shown in Example 6-6.
Example 6-6 The complete /etc/exports file for pseudo-root FS modification

/exports -nfsroot
After this, the /etc/rc.nfs command always fails with the message shown in
Example 6-7.
Example 6-7 /etc/rc.nfs failure

# /etc/rc.nfs
exportfs: cannot change the V4 root to /exports

Not applicable.
Solution
Check that the /exports directory exists. After creating the directory using
command mkdir /exports everything worked fine.
6.7.2 Exportfs: /<path>: Invalid argument

After unexporting all file systems using the exportfs -ua command and
subsequent export using the exportfs -va command, we get following output:
Example 6-8 Invalid argument while running exportfs -va

# exportfs -ua
# exportfs

exportfs: nothing exported
#
# exportfs -va
exportfs: /exports2: Invalid argument
exportfs: /exports2/home: Invalid argument
exportfs: /exports2/project/projA: Invalid argument
exportfs: /exports2/project/projB: Invalid argument

Check whether the pseudo-root FS has been set using the nfsd -getnodes
command.
Example 6-9 Output of nfsd -getnodes command for invalid argument problem
# nfsd -getnodes
#root:public
/prootfs:/prootfs
#
Next step is to check whether the /etc/exports file contains the -nfsroot option.
Example 6-10 Check whether the -nfsroot option is in file /etc/exports

# pg /etc/exports
/exports2 -vers=4,sec=krb5:sys,rw
/exports2/home -vers=4,sec=krb5:sys,rw
/exports2/project/projA -vers=4,sec=krb5:sys,ro
/exports2/project/projB -vers=4,sec=krb5:sys,ro
Finally we verified that the nfsd daemon shows the -root / option.
Example 6-11 Check that nfsd is running with -root / option

# ps -ef |grep nfs |grep -v grep
root 286924 159906 0 17:53:33 - 0:00 /usr/sbin/nfsd -root / 3891
Solution
The problem is caused by the missing -nfsroot option within the /etc/exports file.
After adding the / -nfsroot option to the /etc/exports file as shown in the next
example, everything worked fine.
Example 6-12 Correct /etc/exports file with -nfsroot option set to /

# pg /etc/exports
/ -nfsroot
/exports2/home -vers=4,sec=krb5:sys,rw

/exports2/project/projA -vers=4,sec=krb5:sys,ro
/exports2/project/projB -vers=4,sec=krb5:sys,ro
#
# exportfs -va
exported /exports2
exported /exports2/home
exported /exports2/project/projA
exported /exports2/project/projB
#
# nfsd -getnodes
#root:public
/:/
6.7.3 Exportfs: /var/<logfile>: Too many levels of symbolic links...

We changed the /etc/exports file to include this line to adopt the exname option:
/var/nfs4log -vers=4,sec=krb5:sys,ro,exname=/exports/logs
The full /etc/exports file looked as shown in Example 6-13.
Example 6-13 The complete /etc/exports file after modification

/exports -vers=4,sec=krb5:krb5i:krb5p:sys,rw
/exports/home -vers=4,sec=krb5:sys,rw
/exports/project/projA -vers=4,sec=krb5:sys,ro
/exports/project/projB -vers=4,sec=krb5:sys,ro
/var/nfs4log -vers=4,sec=krb5:sys,ro,exname=/exports/logs
After this, the exportfs -va command always fails with the message shown in
Example 6-14.
Example 6-14 exportfs -va failure

exportfs: 1831-187 re-exported /exports
exportfs: 1831-187 re-exported /exports/home
exportfs: 1831-187 re-exported /exports/project/projA
exportfs: 1831-187 re-exported /exports/project/projB
exportfs: /var/nfs4log: There are too many levels of symbolic links to
translate a path name.

Check to see what the nfsroot is set to with the nfsd -getnodes command. The
exname must begin with the nfsroot.

Solution
The use of the exname option with the pseudo-file system was incorrect. If you
start using the exname option in the /etc/exports files you have to take the
following into account:
1. Either all Version 4 exports must specify an external name, or none of the
exports specify an external name.
2. The nfsroot, /exports, must not be exported.
After removing all non-exname entries from the /etc/exports file, the exportfs
command ran without any further problems.
6.8 Mount problems

The NFS client can experience various types of mount problems, and in this
section we look at the most common ones. Remember that if you are using NFS
V4 with Kerberos, problem determination should always start at the client side. In
this case using an iptrace should also be mandatory to break down the root
cause of your problem. See “Sample iptrace output” on page 271 for a sample
iptrace breakdown.
6.8.1 General mount problem

Problem
The mount on a client fails with access denied as shown in Example 6-15.
Example 6-15 Client mount returns access denied

# mount nfs403:/ /mnt
mount: 1831-011 access denied for nfs403:/
nfs403:/
The file access permissions do not allow the specified action.

Check which NFS version has been used on the server to export the file system.
In this case, the file system was exported using option:
Solution
Caused by use of the vers=4 option on the client; by default, NFS V3 is used.

6.8.2 Pseudo-root and nfs4cl problems
Problem
This command works fine on the client:
However, this command output seems to be incorrect:

nfs4cl showfs
Example 6-16 Incorrect output from the nfs4cl showfs command

-------- --------------- --------------- ---------------
nfs404 10:13 /nfs/projB
Example 6-17 shows the output we expect to see.
Example 6-17 Correct output from the nfs4cl showfs command

-------- --------------- --------------- ---------------
nfs404 10:13 /nfs/project/projB
We see in Example 6-16 that in the Local Path column, nfs4cl showfs reports
the mounted file system incorrectly. Example 6-17 shows the correct output.

Not applicable.
Solution
This is a problem with the nfs4cl command on AIX 5.3.
Note: As we wrote this book, this problem had been identified as a bug and
should be resolved in a future Maintenance Level for AIX 5.3.
6.8.3 ‘vers’ mount option error: “...Program not registered”

The mount command with option vers=4 reports the error in Example 6-18.
Example 6-18 Errors when the mount command is used

RPC: 1832-019 Program not registered
Verify the NFS local domain has been set, and the nfsrgyd process is running

Check that the NFS domain is set using the chnfsdom command, and that the
associated NFS registry daemon nfsrgyd is running.
Example 6-19 Confirming the NFS domain is set and nfsrgyd is running
# chnfsdom
#
# lssrc -s nfsrgyd
nfsrgyd nfs inoperative
#
Solution
Start the nfsrgyd using /etc/rc.nfs.
6.8.4 ‘vers’ mount option error: “...server <name> not responding”

If you try to mount with an unsupported NFS version number, the mount
command will hang with message shown in Example 6-20.
Example 6-20 Mount command used with an incorrect NFS version number
mount: 1831-010 server nfs402 not responding: RPC: Success
mount: retrying
Solution
Use a supported version number (Version 2, 3, or 4).
6.8.5 Mount command hangs - no system response

The mount command seems to hang without any message:

Check the syslog entries and see if there are recently logged errors:
Example 6-21 Syslog entries showing output from gssd

Jul 30 20:24:50 localhost unix: kgss_init_sec_context returned
GSS_S_UNAVAILABLE

Jul 30 20:25:20 localhost unix: kgss_init_sec_context returned
GSS_S_UNAVAILABLE
Solution
Restart the gssd subsystem or make sure it is running.
6.8.6 Mount with sec=krb5: “vmount: The file access permissions do

not allow the specified action”
Running the following mount command returns the error shown in Example 6-22:
Example 6-22 Error returned by the mount command

nfs404:/
vmount: The file access permissions do not allow the specified action.
This problem has multiple fail reasons even though this command works:
mount -o vers=4 nfs404:/ /nfs
Failure mode 1: problem determination steps

Check the syslog for errors similar to Example 6-23 and Example 6-24.
Example 6-23 Syslog entry on NFS V4 server if mounted from server

Jul 30 11:53:27 nfs404 unix: kgss_init_sec_context returned GSS_S_FAILURE
KRB5_FCC_NOFILE
Example 6-24 Syslog entry on NFS V4 server if mounted from client

Jul 29 17:17:52 nfs404 syslog: accept_context failed, major=d0000,
minor=96c73ae3
Jul 29 17:17:52 nfs404 unix: kgss_accept_sec_context failed: GSS_S_FAILURE
FFFFFFFF
Jul 30 11:13:48 nfs404 syslog: nfsrgyd: Unable to map realm
(itsc.austin.ibm.com) to a domain
Solution to failure mode 1

The /etc/nfs/realm.map file has been created incorrectly. Use the chnfsrtd
command to recreate the file. This is the incorrect syntax for /etc/nfs/realm.map:
realm.map realm2.ibm.com itsc.austin.ibm.com

This is the correct syntax:

Example 6-25 Errors from syslog output
Jul 30 16:02:45 nfs407 gssd[11968]: The user option is different from the
address family passed into the API.
Jul 30 16:02:45 nfs407 gssd[11968]: The user option is different from the
address family passed into the API.
Jul 30 16:02:45 nfs407 syslog: init_context failed, major=d0000, minor=96c73a07
Jul 30 16:02:45 nfs407 unix: kgss_init_sec_context returned GSS_S_FAILURE
KRB5KDC_ERR_S_PRINCIPAL_UNKNOWN
Solutions to failure mode 2

1.) Incorrect hostkey entry
There is an incorrect entry in the /etc/nfs/hostkey file for the NFS V4 server
name. Use nfshostkey -l to check this:
Example 6-26 Using the nfshostkey command

# nfshostkey -l
#
The correct entry should be:

Carry out the following steps to resolve the problem:
a. Run the following command:
nfshostkey -p nfs/nfs407.itsc.austin.ibm.com -f\ /etc/krb5/krb5.keytab
b. Stop and restart NFS.
2.) Incorrect nfshostmap entry

There is an incorrect entry in the /etc/nfs/princmap file. Use the nfshostmap -l
command to check this:
nfs407 nfs407.itsc.austin.ibm.com
This entry is incorrect. We do not have to map this host, as nfs407 does not have
a second interface. Carry out the following steps to resolve the problem:
1. Run the following command:
nfshostmap -d nfs407

2. Stop and restart the gssd subsystem:
stopsrc -s gssd
startsrc -s gssd
Restriction: As we write this book, we have found that the gssd takes
approximately two minutes from the stop-and-start operation to update all
local changes.
3.) Incorrect credentials

The credentials within the NAS credential cache file
/var/krb5/security/creds/krb5cc_0 are incorrect. Carry out the following steps to
resolve the problem:
1. Unset the KRB5CCNAME variable with the following command:
unset KRB5CCNAME
2. To check the credentials available to the gssd daemon, run:
klist
3. Stop and restart the gssd subsystem:
stopsrc -s gssd
startsrc -s gssd

The syslog file gives the entries shown in Example 6-27.
Example 6-27 KRB5_NO_TKT_IN_RLM in syslog file

Aug 16 11:14:50 nfs406 su: from root to sally at /dev/pts/0
Aug 16 11:14:50 nfs406 syslog: init_context failed, major=d0000, minor=96c73ab9
Aug 16 11:14:50 nfs406 unix: kgss_init_sec_context returned GSS_S_FAILURE
KRB5_NO_TKT_IN_RLM

The client does not belong to the same Kerberos realm as the NFS V4 server.
Thus you have to add the NFS domain to the /etc/realm.map file using the
chnfsrtd or smitty chnfsrtd commands. After the change, this error will not be
logged any more.
6.8.7 Mount with sec=krb5: “RPC: 1832-016 Unknown host...”

Running the following mount command returns the error shown in Example 6-28
on page 224, even though the NFS domain is set on the system:

Example 6-28 Unknown host error returned from the mount command
RPC: 1832-016 Unknown host
Verify the NFS local domain has been set, and the nfsrgyd process is running
#
#chnfsdom
#
# lssrc -s nfsrgyd
nfsrgyd nfs 16530 active
#

This message seems to indicate a local DNS/HOSTNAME resolution problem:
RPC: 1832-016 Unknown host
All other messages can be ignored at this time.
The local gssd must be able resolve the NFS V4 server as well as the KDC
server, so we check whether the server names can be found using the host
Example 6-29 NFS V4 host IP / DNS test

# host nfs403
nfs403.itsc.austin.ibm.com is 9.3.5.173
#
# ping -c 1 nfs403
PING nfs403.itsc.austin.ibm.com (9.3.5.173): 56 data bytes
64 bytes from 9.3.5.173: icmp_seq=0 ttl=255 time=0 ms
--- nfs403.itsc.austin.ibm.com ping statistics ---

1 packets transmitted, 1 packets received, 0% packet loss
round-trip min/avg/max = 0/0/0 ms
#
This test passed, so the NFS server can be resolved. Next, we test whether the
KDC server is known and resolvable (Example 6-30).
Example 6-30 Test to see whether the KDC server is known and accessible
# grep ":88" /etc/krb5/krb5.conf
#
# host nfs409.kdc.itsc.austin.ibm.com
nfs409.kdc.itsc.austin.ibm.com is 9.3.4.71

# host nfs409
# tn nfs409 88
Trying...
^C
# tn nfs409.kdc.itsc.austin.ibm.com 88
Trying...
Connected to nfs409.kdc.itsc.austin.ibm.com.
Escape character is '^T'.
^CConnection closed.
#
This test passed as well. We finally check the local host name and loopback
interface using the hostname and rcpinfo commands (Example 6-31).
Example 6-31 Checking the local hostname and loopback interface

# hostname
#
root@nfs405 [scripts] host nfs405
root@nfs405 [scripts] ping -c1 nfs405
PING nfs405.itsc.austin.ibm.com (9.3.5.175): 56 data bytes
64 bytes from 9.3.5.175: icmp_seq=0 ttl=255 time=0 ms
--- nfs405.itsc.austin.ibm.com ping statistics ---

1 packets transmitted, 1 packets received, 0% packet loss
round-trip min/avg/max = 0/0/0 ms
root@nfs405 [scripts]
#
root@nfs405 [scripts] rpcinfo -p
program vers proto port service
100000 4 udp 111 portmapper
100000 4 tcp 111 portmapper
100003 2 udp 2049 nfs
100003 3 udp 2049 nfs
100003 2 tcp 2049 nfs
100003 3 tcp 2049 nfs
100003 4 tcp 2049 nfs
200006 1 udp 2049
200006 4 udp 2049
200006 1 tcp 2049
200006 4 tcp 2049
400003 1 udp 33843

100005 1 tcp 32787 mountd
100005 1 udp 33847 mountd
100005 2 udp 33847 mountd
100005 3 udp 33847 mountd
400234 1 udp 33849
100024 1 tcp 32788 status
100024 1 udp 33851 status
100133 1 tcp 32788
100133 1 udp 33851
200001 1 tcp 32788
200001 1 udp 33851
200001 2 tcp 32788
200001 2 udp 33851
400005 1 udp 33848
100021 1 udp 33882 nlockmgr
100021 1 tcp 32789 nlockmgr
#
#root@nfs405 [scripts] host loopback
host: name loopback NOT FOUND
#
Solution
The system’s loopback interface name resolution failed (see RFC1537 for
details). This indicates an incorrect entry on the DNS server or /etc/hosts setup.
Correcting the system loopback entry resolves the problem and the host
command returns the correct entry, shown in Example 6-32.
Example 6-32 Correct entry returned by the host command

#root@nfs405 [scripts] host loopback
loopback is 127.0.0.1, Aliases: localhost
root@nfs405 [scripts] mount -o vers=4,sec=krb5 nfs402:/ /nfs
#
6.8.8 File and directory access: cd, ls, etc. return “permission denied”
A NFS client cannot access the mounted files and directories. Several
commands, such as cd or ls, return a permission-denied message
(Example 6-33 on page 227).

Example 6-33 Permission denied messages from several commands
# nfs4cl showfs

-------- --------------- --------------- ---------------
nfs403 /exports/home 10:12 /nfs
#
# cd /nfs
ksh: /nfs: permission denied
#
# ls -l /nfs
/nfs: No permission
#

1. Check the syslog entries on the client, as shown in Example 6-34.
Example 6-34 Syslog entries seen as a result of the permissions problem

KRB5_NO_TKT_IN_RLM
These errors indicate that no valid ticket has been issued to the user.
2. You can also check the ticket status by running the klist command.
Example 6-35 Checking the ticket status using klist

#date
Tue Aug 10 15:00:52 CDT 2004
#
#klist

08/02/04 11:14:19 08/03/04 11:14:11

Request a new ticket using the kinit command to resolve the problem
(Example 6-36 on page 228).

Example 6-36 Requesting a new ticket for the user using the kinit command
# kinit sally
#
# klist

08/06/04 09:27:06 08/06/04 19:27:13
krbtgt/[email protected]
Renew until 08/07/04 09:27:06
#
# cd /nfs
#
# ls
bob joe lost+found mary sally
#

Check the syslog log file and check whether you can see an entry similar to the
one in Example 6-37.
Example 6-37 Error showing the gssd being unable to access a valid ticket cache
KRB5_FCC_NOFILE
The error indicates that the gssd process is unable to access a valid ticket cache
file for the principal.
Check to see whether the KRB5CCNAME environment variable has been set.
Example 6-38 Has the KRB5CCNAME environment variable been set?

$ set |grep -i krb
AUTHSTATE=KRB5LDAP
KRB5CCNAME=FILE:/var/krb5/security/creds/[email protected]_6023
PATH=/usr/bin:/etc:/usr/sbin:/usr/ucb:/usr/bin/X11:/sbin:/usr/krb5/bin/:/usr/ja
va14/jre/bin:/usr/java14/bin:/usr/java131/jre/bin:/usr/java131/bin

Change the user environment by unsetting the KRB5CCNAME environment
variable and obtaining a new ticket, as shown in Example 6-39 on page 229.

Example 6-39 Unsetting KRB5CCNAME and obtaining a new ticket
$ unset KRB5CCNAME
$
$ klist
$
$ kinit sally
$
$ cd /nfs
$
6.8.9 File and directory access: file ownership is “nobody:nobody”

A newly created file has ownership nobody:nobody instead of the format
userame:groupname. The server and client are in the same NFS domain, and
the identity of the user is valid. Example 6-40 shows the problem.
Example 6-40 The nobody:nobody ownership issue

# tn nfs406
Trying...
Connected to nfs406.itsc.austin.ibm.com.
Escape character is '^T'.
AIX Version 5
login: sally
sally's Password:
$
$ id
uid=6023(sally) gid=1(staff) groups=1400(eng)
$
$ klist

08/11/04 11:06:34 08/12/04 11:06:23
$
$ mount
-------- --------------- --------------- ------ ------------ ---------------
$ cd /nfs/tmp

$
$ touch thisisatest
$
$ ls -l thisisatest
-rw-r--r-- 1 nobody nobody 0 Aug 11 10:54 thisisatest
$

On the NFS server, the syslog log file gives entries as shown in Example 6-41.
Example 6-41 NFS V4 registry daemon entries in the syslog log file
Aug 11 11:09:22 nfs404 syslog: nfsrgyd: Unable to map local user (sally) to a
foreign user
Aug 11 11:09:22 nfs404 syslog: nfsrgyd: Unable to map local group (staff) to a
foreign group
Solution
The realm to NFS domain mapping in the /etc/nfs/realm.map file is incorrect.
Carry out the following steps to resolve the problem:
1. Change the mapping using the chnfsrtd command to make the change from:
realm1.ibm.com nfstest.itsc.austin.ibm.com
to
2. Stop and start the nfsrgyd on NFS server nfs404:
stopsrc -s nfsrgyd
startsrc -s nfsrgyd
Example 6-42 shows the test we carried out to prove that the issue had been
resolved.
Example 6-42 The resolved issue after steps 1 and 2 have been carried out
$ cd /nfs
$
$ touch thisisatest2
$
$ ls -l thisisatest2
-rw-r--r-- 1 sally staff 0 Aug 11 11:08 thisisatest2
$

6.8.10 NAS problem: kadmin: “Unable to initialize kadmin interface”
After installing the NAS file sets and configuration of the NAS client, kadmin no
longer works and results in the error shown in Example 6-43.
Example 6-43 kadmin initialization error

Authenticating as principal root/[email protected] with password.
Unable to initialize kadmin interface.
Status 0x29c2500 - Operation failed for unspecified reason.

See whether you can log on to Kerberos using the principal admin/admin.
Example 6-44 kinit verification with principal admin/admin
# kinit admin/admin
#
# klist

Solution
The problem is caused by the fact that principal root/admin is not known as an
administrative principal. So this message is normal and we recommend that you
use the kadmin command with the -p admin/admin option.
Example 6-45 kadmin verification with -p option

# /usr/krb5/sbin/kadmin -p admin/admin
kadmin:

6.9 GSS-API error codes
Each GSS-API function returns two status codes: a major status code and a
minor status code. Major status codes indicate behavior of the GSS-API itself.
For example, if an application attempts to transmit a message after a security

context has expired, GSS-API returns a major status code of
GSS_S_CONTEXT_EXPIRED. Major status codes are listed among the
GSS-API status codes.
Minor status codes are returned by the underlying security mechanisms that are
supported by a given implementation of GSS-API. Every GSS-API function takes
as the first argument a minor_status or minor_stat parameter. An application can
examine this parameter when the function returns, successfully or not, to see the
status that is returned by the underlying mechanism.
Example 6-46 GSS-API syslog entries

KRB5_NO_TKT_IN_RLM
In Example 6-46, the following important information is provided:

򐂰 The GSS subroutine called kgss_init_sec_context returned with GSS-API:
– Major status: GSS_S_FAILURE
– Minor status: KRB5_NO_TKT_IN_RLM
The following two sections provide a quick messages reference. A more detailed
list can be found in IBM Network Authentication Service Version 1.4 for AIX,
Linux, Solaris and Windows Application Development Reference, available in the
krb5.doc.en_US file set.
6.9.1 Major GSS-API error codes

The following table lists calling errors returned by GSS-API.
Table 6-3 GSS-API routine failures

Error code Description
GSS_S_BAD_MECH An unsupported mechanism was

requested.
GSS_S_BAD_NAME An invalid name was supplied.

Error code Description
GSS_S_BAD_NAMETYPE A supplied name was of an unsupported

type.
GSS_S_BAD_BINDINGS Incorrect channel bindings were supplied.
GSS_S_BAD_STATUS An invalid status code was supplied.
GSS_S_BAD_SIG A token had an invalid signature.
GSS_S_NO_CRED No credentials were supplied, or the

credentials were unavailable or
inaccessible.
GSS_S_NO_CONTEXT No context has been established.
GSS_S_DEFECTIVE_TOKEN A token was invalid.
GSS_S_DEFECTIVE_CREDENTIAL A credential was invalid.
GSS_S_CREDENTIALS_EXPIRED The referenced credentials have expired.
GSS_S_CONTEXT_EXPIRED The context has expired.
GSS_S_FAILURE Miscellaneous failure. The underlying

mechanism detected an error for which no
specific GSS–API status code is defined.
The mechanism-specific status code
(minor-status code) provides more details
about the error.
GSS_S_BAD_QOP The quality-of-protection requested could

not be provided.
GSS_S_UNAUTHORIZED The operation is forbidden by local

security policy.
GSS_S_UNAVAILABLE The operation or option is unavailable.
GSS_S_DUPLICATE_ELEMENT The requested credential element already

exists.
GSS_S_NAME_NOT_MN The provided name was not a mechanism

name (MN).

6.9.2 Kerberos v5 status codes
The following tables list the status messages that can be returned by Kerberos
V5 in the minor_status argument. For more about GSS-API status codes, see
GSS-API Status Codes.
Messages returned in Kerberos v5 for KRB5KDC status codes

Table 6-4 Kerberos V5 KRB5KDC status codes
Minor status code Description
KRB5KDC_ERR_BAD_PVNO Requested protocol version not

supported.
KRB5KDC_ERR_BADOPTION KDC cannot fulfill requested option.
KRB5KDC_ERR_C_OLD_MAST_KVNO Client’s key is encrypted in an old

master key.
KRB5KDC_ERR_C_PRINCIPAL_UNKNOWN Client not found in Kerberos

database.
KRB5KDC_ERR_CANNOT_POSTDATE Ticket is ineligible for postdating.
KRB5KDC_ERR_CLIENT_NOTYET Client not yet valid; try again later.
KRB5KDC_ERR_CLIENT_REVOKED Client’s credentials have been

revoked.
KRB5KDC_ERR_ETYPE_NOSUPP KDC has no support for encryption

type.
KRB5KDC_ERR_KEY_EXP Password has expired.
KRB5KDC_ERR_NAME_EXP Client’s entry in database has

expired.
KRB5KDC_ERR_NEVER_VALID Requested effective lifetime is

negative or too short.
KRB5KDC_ERR_NONE No error.
KRB5KDC_ERR_NULL_KEY Client or server has a null key.
KRB5KDC_ERR_PADATA_TYPE_NOSUPP KDC has no support for padata type.
KRB5KDC_ERR_POLICY KDC policy rejects request.
KRB5KDC_ERR_PREAUTH_FAILED Preauthentication failed.
KRB5KDC_ERR_PREAUTH_REQUIRED Additional preauthentication

required.

Minor status code Description
KRB5KDC_ERR_PRINCIPAL_NOT_UNIQUE Principal has multiple entries in

Kerberos database.
KRB5KDC_ERR_S_OLD_MAST_KVNO Server’s key is encrypted in an old

master key.
KRB5KDC_ERR_S_PRINCIPAL_UNKNOWN Server not found in Kerberos

database.
KRB5KDC_ERR_SERVER_NOMATCH Requested server and ticket do not

match.
KRB5KDC_ERR_SERVICE_EXP Server’s entry in database has

expired.
KRB5KDC_ERR_SERVICE_NOTYET Server not yet valid; try again later.
KRB5KDC_ERR_SERVICE_REVOKED. Credentials for server have been

revoked.
KRB5KDC_ERR_SUMTYPE_NOSUPP KDC has no support for checksum

type.
KRB5KDC_ERR_TGT_REVOKED TGT has been revoked.
KRB5KDC_ERR_TRTYPE_NOSUPP KDC has no support for transited

type.
Messages returned in Kerberos V5 for KRB5KRB status code

Table 6-5 Kerberos V5 KRB5KRB status codes
Minor status codes Description
KRB5KRB_AP_ERR_BAD_INTEGRITY Decrypt integrity check failed.
KRB5KRB_AP_ERR_BADADDR Incorrect net address.
KRB5KRB_AP_ERR_BADDIRECTION Incorrect message direction.
KRB5KRB_AP_ERR_BADKEYVER. Key version is not available.
KRB5KRB_AP_ERR_BADMATCH Ticket and authenticator do not

match.
KRB5KRB_AP_ERR_BADORDER Message out of order.
KRB5KRB_AP_ERR_BADSEQ Incorrect sequence number in

message.
KRB5KRB_AP_ERR_BADVERSION Protocol version mismatch.

Minor status codes Description
KRB5KRB_AP_ERR_ILL_CR_TKT Illegal cross-realm ticket.
KRB5KRB_AP_ERR_INAPP_CKSUM Inappropriate type of checksum in

message.
KRB5KRB_AP_ERR_METHOD Alternative authentication method

required.
KRB5KRB_AP_ERR_MODIFIED Message stream modified.
KRB5KRB_AP_ERR_MSG_TYPE Invalid message type.
KRB5KRB_AP_ERR_MUT_FAIL Mutual authentication failed.
KRB5KRB_AP_ERR_NOKEY Service key not available.
KRB5KRB_AP_ERR_NOT_US The ticket is not for us.
KRB5KRB_AP_ERR_REPEAT Request is a replay.
KRB5KRB_AP_ERR_SKEW Clock skew too great.
KRB5KRB_AP_ERR_TKT_EXPIRED Ticket expired.
KRB5KRB_AP_ERR_TKT_INVALID Ticket has invalid flag set.
KRB5KRB_AP_ERR_TKT_NYV Ticket not yet valid.
KRB5KRB_AP_ERR_V4_REPLY Initial ticket response appears to be

Version 4 error.
KRB5KRB_AP_WRONG_PRINC Wrong principal in request.
KRB5KRB_ERR_FIELD_TOOLONG Field is too long for this

implementation.
KRB5KRB_ERR_GENERIC Generic error.
Messages returned in Kerberos for KRB5 status codes

Table 6-6 Kerberos V5 general KRB5 status codes
Minor codes Description
KRB5_BAD_ENCTYPE Bad encryption type.
KRB5_BAD_KEYSIZE Key size is incompatible with

encryption type.
KRB5_BAD_MSIZE Message size is incompatible with

encryption type.

KRB5_BADMSGTYPE Invalid message type specified for

encoding.
KRB5_CC_BADNAME Credential cache name malformed.
KRB5_CC_END End of credential cache reached.
KRB5_CC_FORMAT Bad format in credentials cache.
KRB5_CC_IO Credentials cache I/O operation

failed.
KRB5_CC_NOMEM No more memory to allocate (in

credentials cache code).
KRB5_CC_NOTFOUND Matching credential not found.
KRB5_CC_TYPE_EXISTS Credentials cache type is already

registered.
KRB5_CC_UNKNOWN_TYPE Unknown credential cache type.
KRB5_CC_WRITE Error writing to credentials cache file.
KRB5_CCACHE_BADVNO Unsupported credentials cache

format version number.
KRB5_CONF_NOT_CONFIGURED Kerberos /etc/krb5/krb5.conf

configuration file not configured.
KRB5_CONFIG_BADFORMAT Improper format of Kerberos

/etc/krb5/krb5 configuration file.
KRB5_CONFIG_CANTOPEN Cannot open or find Kerberos

/etc/krb5/krb5 configuration file.
KRB5_CONFIG_NODEFREALM Configuration file /etc/krb5/krb5.conf

does not specify default realm.
KRB5_CONFIG_NOTENUFSPACE Insufficient space to return complete

information.
KRB5_ERR_BAD_HOSTNAME Host name cannot be canonicalized.
KRB5_ERR_HOST_REALM_UNKNOWN Cannot determine realm for host.
KRB5_FCC_INTERNAL Internal file credentials cache error.
KRB5_FCC_NOFILE No credentials cache file found.
KRB5_FCC_PERM Credentials cache file permissions

incorrect.

KRB5_FWD_BAD_PRINCIPAL Bad principal name while trying to

forward credentials.
KRB5_GET_IN_TKT_LOOP Looping detected in krb5_get_in_tkt.
KRB5_IN_TKT_REALM_MISMATCH Client/server realm mismatch in

initial ticket request.
KRB5_KDC_UNREACH Cannot contact any KDC for

requested realm.
KRB5_KDCREP_MODIFIED KDC reply did not match

expectations.
KRB5_KDCREP_SKEW Clock skew too great in KDC reply.
KRB5_KEYTAB_BADVNO Unsupported key table format

version number.
KRB5_KT_BADNAME Key table name malformed.
KRB5_KT_END End of key table reached.
KRB5_KT_IOERR Error writing to key table.
KRB5_KT_KVNONOTFOUND Key version number for principal in

key table is incorrect.
KRB5_KT_NAME_TOOLONG Keytab name too long.
KRB5_KT_NOTFOUND Key table entry not found.
KRB5_KT_NOWRITE Cannot write to specified key table.
KRB5_KT_TYPE_EXISTS Key table type is already registered.
KRB5_KT_UNKNOWN_TYPE Unknown key table type.
KRB5_MUTUAL_FAILED Mutual authentication failed.
KRB5_NO_2ND_TKT Request missing second ticket.
KRB5_NO_LOCALNAME No local name found for principal

name.
KRB5_NO_TKT_IN_RLM Cannot find ticket for requested

realm.
KRB5_NO_TKT_SUPPLIED Request did not supply a ticket.
KRB5_NOCREDS_SUPPLIED No credentials supplied to library

routine.

KRB5_PREAUTH_BAD_TYPE Unsupported preauthentication type.
KRB5_PREAUTH_FAILED Generic preauthentication failure.
KRB5_PREAUTH_NO_KEY Required preauthentication key not

supplied.
KRB5_PRINC_NOMATCH Requested principal and ticket do not

match.
KRB5_PROG_ATYPE_NOSUPP Program lacks support for address

type.
KRB5_PROG_ETYPE_NOSUPP Program lacks support for encryption

type.
KRB5_PROG_KEYTYPE_NOSUPP Program lacks support for key type.
KRB5_PROG_SUMTYPE_NOSUPP Program lacks support for checksum

type.
KRB5_RC_IO_EOF End-of-file on replay cache I/O.
KRB5_RC_IO_MALLOC No more memory to allocate (in

replay cache I/O code).
KRB5_RC_IO_PERM Permission denied in replay cache

code.
KRB5_RC_MALLOC No more memory to allocate (in

replay cache code).
KRB5_RC_PARSE Replay cache name parse/format

error.
KRB5_RC_REQUIRED Message replay detection requires

rcache parameter.
KRB5_RC_TYPE_EXISTS Replay cache type is already

registered.
KRB5_RC_TYPE_NOTFOUND Replay cache type is unknown.
KRB5_RCACHE_BADVNO Unsupported replay cache format

version number.
KRB5_REALM_CANT_RESOLVE Cannot resolve KDC for requested

realm.
KRB5_REALM_UNKNOWN Cannot find KDC for requested

realm.

KRB5_SAM_UNSUPPORTED Bad SAM flags in

obtain_sam_padata.
KRB5_SENDAUTH_BADAPPLVERS Bad application version was sent (by

sendauth).
KRB5_SENDAUTH_BADAUTHVERS Bad sendauth version was sent.
KRB5_SENDAUTH_BADRESPONSE Bad response (during sendauth

exchange).
KRB5_SENDAUTH_REJECTED Server rejected authentication

(during sendauth exchange).
KRB5_SERVICE_UNKNOWN Kerberos service unknown.
KRB5_SNAME_UNSUPP_NAMETYPE Conversion to service principal

undefined for name type.
KRB5_TKT_NOT_FORWARDABLE Requesting ticket cannot get

forwardable tickets.
KRB5_WRONG_ETYPE Requested encryption type not used

in message.
Messages returned in Kerberos v5 for KRB5DES status codes

Table 6-7 Kerberos V5 KRB5DES status codes
KRB5DES_BAD_KEYPAR DES key has bad parity.
KRB5DES_WEAK_KEY DES key is a weak key.

Part 3
Part 3 Appendixes
This part includes the following sections:
򐂰 Appendix A, “Kerberos” on page 243
򐂰 Appendix B, “Sample scripts, files, and output” on page 255
򐂰 Appendix C, “AIX 5.3 NFS quick reference” on page 287

A
Appendix A. Kerberos
This appendix describes the Kerberos authentication method. It also provides a
list of references that contain more in-depth information about Kerberos.
The appendix contains the following sections:

򐂰 Overview
򐂰 Kerberos keys and initial setup
򐂰 Authenticating to the Kerberos server
򐂰 Authenticating to an application server
򐂰 Kerberos terminology
򐂰 Where to find more information about Kerberos
The first four sections are reprinted from the IBM Redbook The RS/6000 SP
Inside Out, SG24-5374.

Overview
Kerberos: Also spelled Cerberus, the watchdog of Hades, whose duty was to
guard the entrance (against whom or what does not clearly appear)… It is
known to have had three heads.
- Ambrose Bierce, The Enlarged Devil's Dictionary
Kerberos is a trusted third-party authentication system for use on physically

insecure networks. It enables entities communicating over the network to prove
their identity to each other while preventing eavesdropping or replay attacks.
Figure 6-2 shows the three parties involved in the authentication. The Kerberos
system was designed and developed in the 1980s by the Massachusetts Institute
of Technology (MIT) as part of the Athena project. The current version of
Kerberos is Version 5, which is standardized in RFC 1510, The Kerberos
Network Authentication Service (V5).
Figure 6-2 Partners in a third-party authentication
Kerberos provides two services to Kerberos principals (users or services): an

authentication service (AS) and a ticket-granting service (TGS). Principals can
prove their identity to the AS with a single sign-on and get a ticket-granting ticket
(TGT) back from the AS. When one authenticated principal (the client) wants to

use the services of a second authenticated principal (the server), it can get a
service ticket for this service by presenting its TGT to the Kerberos TGS. The
service ticket is then sent from the client to the server, which can use it to verify
the client’s identity.
This section describes the protocol that Kerberos uses to provide these services,
independent of a specific implementation. A more detailed rationale for the
Kerberos design can be found in the MIT article Designing an Authentication
System: a Dialogue in Four Scenes, which is available from:
http://web.mit.edu/kerberos/www/dialogue.html
Kerberos keys and initial setup

To encrypt the messages that are sent over the network, Kerberos uses a
symmetric encryption method, normally the Data Encryption Standard (DES).
This means that the same key is used to encrypt and decrypt a message, and
consequently the two partners of a communication must share this key if they
want to use encryption. The key is called a secret key for obvious reasons: it
must be kept secret by the two parties; otherwise the encryption will not be
effective.
This approach must be distinguished from public key cryptography, which is an

asymmetric encryption method. There, two keys are used: a public key and a
private key. A message encrypted with one of the keys can only be decrypted by
the other key, not by the one that encrypted it. The public keys do not need to be
kept secret (hence the name “public”), and a private key is known only to its
owner. (It is not even to the communication partner as in the case of symmetric
cryptography.) This has the advantage that no key must be transferred between
the partners prior to the first use of encrypted messages.
With symmetric encryption, principals must provide a password to the Kerberos

server before they can use the Kerberos services. The Kerberos server then
encrypts it and stores the resulting key in its database. This key is the shared
information that the Kerberos server and the principal can use to encrypt and
decrypt the messages they send each other. Initially, two principals who want to
communicate with each other do not share a key, and so cannot encrypt their
messages. But since the Kerberos server knows the keys of all the principals, it is
called a trusted third party and can securely provide a common session key to
the two parties.
Obviously, the initial passwords have to be entered securely, if possible at the

console of the Kerberos server machine. They might also be generated by the
Kerberos server (especially if the principal is a host or service). In that case they
must be securely transferred to the principal that stores (or remembers) them.
Appendix A. Kerberos 245

Authenticating to the Kerberos server
If a principal (typically a user) wants to use Kerberos services (for example,
because it wants to use an application service that requires Kerberos
authentication), it first has to prove its identity to the Kerberos server. This is
done in the following way:
A command to sign on to the Kerberos system is issued on the application client,

typically kinit. This command sends an authentication request to the Kerberos
server, as shown in Figure A-1. This contains the type of service that is
requested (here, the client wants to get service from the ticket-granting service),
the client’s (principal’s) name, and the IP address of the client machine. This
request is sent in plain text. Note that the principal’s password is not sent in this
packet, so there is no security exposure in sending the request in plain text.
Figure A-1 Client's authentication request
The request is processed by the authentication server. Using the client’s name, it
looks up the corresponding key in the Kerberos database. It also generates a
random session key to be shared by the client and the TGS, which will be used to
encrypt all future communication of the client with the TGS. With this information,
the AS constructs the ticket-granting ticket for the client, which (as with all
Kerberos tickets) contains six parts:
1. The service for which the ticket is good (here, the TGS)
2. The client’s (principal’s) name
3. The client machine’s IP address
4. A timestamp showing when the ticket was issued
5. The ticket lifetime (configurable in K5)
6. The session key for client/TGS communications
This ticket is encrypted with the secret key of the TGS, so only the TGS can
decrypt it. Because the client needs to know the session key, the AS sends back
a reply that contains both the TGT and the session key, all of which is encrypted
by the client’s secret key. This is shown in Figure A-2 on page 247.

Figure A-2 Authentication server's reply: TGT
Now the sign-on command prompts the user for the password and generates a
DES key from the password using the same algorithm as the Kerberos server. It
then attempts to decrypt the reply message with that key. If this succeeds, the
password has matched the one used to create the user’s key in the Kerberos
database, and the user has authenticated herself. If the decryption fails, the
sign-on is rejected and the reply message is useless. Assuming success, the
client now has the encrypted TGT and the session key for use with the TGS, and
stores them both in a safe place. Note that the authentication has been done
locally on the client machine, and the password has not been transferred over the
network.
Authenticating to an application server

If the client now wants to access an application service that requires Kerberos
authentication, it must get a service ticket from the TGS. The TGT that was
obtained during the Kerberos sign-on can be used to authenticate the client to
the TGS; there is no need to type in a password each time a service ticket is
requested.

If the client sent only the (encrypted) TGT to the Kerberos TGS, this might be
captured and replayed by an intruder who has impersonated the client machine.
To protect the protocol against such attacks, the client also generates an
authenticator which consists of three parts:
1. The client’s (principal’s) name
2. The client machine’s IP address
3. A timestamp showing when the authenticator was created
The authenticator is encrypted with the session key that the client shares with the
TGS. The client then sends a request to the TGS consisting of the name of the
service for which a ticket is requested, the encrypted TGT, and the encrypted
authenticator. This is shown in Figure A-3.
Figure A-3 Client's service ticket request
The TGS can decrypt the TGT because it is encrypted with its own secret key. In
that ticket, it finds the session key to share with the client. It uses this session key
to decrypt the authenticator, and can then compare the client’s name and
address in the TGT and the authenticator.

If the timestamp that the TGS finds in the authenticator differs from the current
time by more than a prescribed difference (typically 5 minutes), a ticket replay
attack is assumed and the request is discarded.
If all checks pass, the TGS generates a service ticket for the service indicated in
the client’s request. The structure of this service ticket is identical to the TGT
described in “Authenticating to the Kerberos server” on page 246. The content
differs in the service field (which now indicates the application service rather than
the TGS), the timestamp, and the session key. The TGS generates a new,
random key that the client and application service will share to encrypt their
communications. One copy is put into the service ticket (for the server), and
another copy is added to the reply package for the client since the client cannot
decrypt the service ticket. The service ticket is encrypted with the secret key of
the service, and the whole package is encrypted with the session key that the
TGS and the client share. The resulting reply is shown in Figure A-4. Compare
this to Figure A-2 on page 247.
Figure A-4 Ticket-granting service's reply: service ticket
The client can decrypt this message using the session key it shares with the
TGS. It then stores the encrypted service ticket and the session key to share with
the application server, normally in the same ticket cache where it already has
stored the TGT and session key for the TGS.

To actually request the application service, the client sends a request to that
server that consists of the name of the requested service, the encrypted service
ticket, and a newly generated authenticator to protect this message against
replay attacks. The authenticator is encrypted with the session key that the client
and the service share. The resulting application service request is shown in
Figure A-5. Note the resemblance to the request for ticket-granting service in
Figure A-3 on page 248.
Figure A-5 Client's application service request
The application server decrypts the service ticket with its secret key, uses the
enclosed session key to decrypt the authenticator, and checks the user’s identity
and the authenticator’s timestamp. Again, this processing is the same as for the
TGS processing the service ticket request. If all checks pass, the server performs
the requested service on behalf of the user.
Authorization: Kerberos is only responsible for authenticating the two

partners. Any authorization mechanism must be enforced by the application
itself.

If the client requires mutual authentication (that is, the service has to prove its
identity to the client), the server could send back a message that is encrypted by
the session key it shares with the client, and application-dependent contents that
the client can verify. Because the service can only know the session key if it was
able to decrypt the service ticket, it must have known its secret key and so has
proven its identity.
Kerberos terminology
The following terms are used when discussing Kerberos authentication:
Realm A Kerberos domain that can consist of a number of
machines providing authentication services.
Principal A user or a service that uses authentication services and
is identified in the authentication database. For example,
[email protected], where root is the user
identity and admin the instance.
Instance In the case of a service instance, it represents the
occurrence of the server. Service example:
hardmon.sp21cw0, where hardmon represents the
service and sp21cw0 represents the node providing the
service.
In the case of a user, the instance represents the
Kerberos authority granted to the user. User example:
root.admin, where admin represents a Kerberos
authorization for administrative tasks in Kerberos.
Authentication database
A set of files containing the definitions of the Kerberos
authentication information. The authentication database is
maintained at the Kerberos server.
Ticket An encrypted message containing the identity of a user. A
ticket is passed from a client to a server as soon as a
Kerberos service is requested. Tickets have a
predetermined lifetime and have to be renewed
periodically.
Key An eight-byte form of a user or service password stored in
the authentication database. This password is associated
with a Kerberos user or service principal, not a user ID.

Ticket-granting ticket
A ticket that is generated by the Kerberos authentication
database as proof that Kerberos recognizes the user as
an authorized user.
KDC The trusted third-party or intermediary in Kerberos that
issues all Kerberos tickets to the clients.
Where to find more information about Kerberos

Here is a list of references where you can find more information about Kerberos
and related products.
IBM Redbooks
(Redbooks are available for purchase or download at
http://www.redbooks.ibm.com.)
AIX 5L Version 5.2 Security Supplement, SG24-6066

This book includes a chapter titled “Exploiting Network Authentication
Service.” The chapter discusses IBM Network Authentication Service Version
1.3 for AIX, which is the IBM AIX implementation of Kerberos.
RS/6000 SP System Management: Easy, Lean and Mean, GG24-2563

This book discusses Kerberos in the context of the IBM Scalable
POWERparallel® System. Although the book is becoming dated (published in
1995), it contains useful overview material, and it has a useful step-by-step
illustration of the ticket-granting process.
Other IBM publications

AIX 5L Version 5.3 Security Guide, SC23-4907
This manual has a chapter dedicated to Kerberos. The chapter describes how
to implement and troubleshoot Kerberos authentication in AIX.
IBM Network Authentication Service Version 1.4 for AIX, Linux, and Solaris
Administrator’s and User’s Guide
This manual describes how to plan for, install, configure, and administer the
IBM Network Authentication Service Version 1.4.

Non-IBM publications
Kerberos: The Definitive Guide, by Jason Garman. O’Reilly & Associates, Inc.,
2003. ISBN 0596004036
The author’s own words describing this book (quoted from the preface): “This
book is geared toward the system administrator who wants to establish a
single sign-on network using Kerberos. This book is also useful for anyone
interested in how Kerberos performs its magic: the first three chapters will be
most helpful to these people.”
Red Hat Enterprise Linux 3 Reference Guide, Red Hat, Inc., 2003
This book has a chapter dedicated to Kerberos, where it briefly describes
Kerberos, discusses advantages and disadvantages, explains how it works,
and defines terminology.
The book is available for download from http://www.redhat.com.
The following Requests for Comment (RFCs) specify standards for Kerberos
implementations. The RFCs are available in text format at
http://www.ietf.org/rfc.html. They are also available both in text and PDF
format at http://www.faqs.org/rfcs/index.html.
RFC 1508 Generic Security Service Application Program Interface
RFC 1510 The Kerberos Network Authentication Service (V5)
RFC 1964 The Kerberos Version 5 GSS-API Mechanism
Other information sources

http://web.mit.edu/kerberos/index.html
This Web site is the main page for the MIT Kerberos project. It includes
documentation and tutorials to provide a better understanding of the protocol
and downloadable source code for interested developers.
http://www.cmf.nrl.navy.mil/CCS/people/kenh/kerberos-faq.html
This is a comprehensive list of frequently asked questions compiled by Ken
Hornstein of the U.S. Naval Research Laboratory. It also contains links to
other useful sources of information.

B
Appendix B. Sample scripts, files, and

output
This appendix provides copies of sample scripts from throughout the book.
It contains the following sections:

򐂰 Sample administrative scripts
򐂰 Sample client Kerberos configuration files
򐂰 LDIF sample file for KDC
򐂰 Sample iptrace output

Sample administrative scripts
In this section you will find sample scripts used during our testing and
implementation scenarios. If you decide to use the scripts, you should use them
only as templates. We do not recommend that you use the scripts in the provided
form without first carefully reviewing the code and adding error checking and
logging where needed. As provided, there is no form of error checking in the
scripts and this may lead to unexpected results.
Change the pseudo-root FS sample script

The script in Example B-1 can be used to set the pseudo-root FS.
Example: B-1 Sample script to set/change the pseudo-root FS

#!/bin/ksh
NEWNFSROOT="/exports"
#
exportfs -ua
/etc/nfs.clean
#
chnfs -r ${NEWNFSROOT}
#
lssrc -Ss nfsd | grep ${NEWNFSROOT}
#
nfsd -getnodes
#
/etc/nfs.clean
#
/etc/rc.nfs
#
lssrc -g nfs
#
ps -ef | grep nfsd
#
exportfs
Create a KDC server with NFS V4 server

The script in Example B-2 can be used to create and configure your KDC server
on the same system as your NFS V4 server.
Example: B-2 Sample script to create a KDC server with legacy database
#!/bin/ksh
HOST=$(hostname)

IREALM=”REALM2.IBM.COM”
KDCSERV=”nfs403.itsc.austin.ibm.com”
DNSDOM=”itsc.austin.ibm.com”
NFSDOM=”itsc.austin.ibm.com”
#
HOST=”${HOST}.${DNSDOM}”
unset KRB5CCNAME
#
SECRET=”succ3ss”
#
#config.krb5 -S -d ${DNSDOM} -r ${IREALM}
mkkrb5srv -r ${IREALM} -d ${DNSDOM} -s ${HOST}
#
kinit admin/admin
klist
#
/mnt/scripts/create_initial_kdc_user.ksh
#
/usr/krb5/sbin/kadmin -p admin/admin -w ${SECRET}<< EOF
add_principal -e des-cbc-crc:normal -randkey nfs/${HOST}
EOF
#
ktadd nfs/${HOST}
EOF
#
nfshostkey -p nfs/${HOST} -f /etc/krb5/krb5.keytab
nfshostkey -l
#
chnfsdom ${NFSDOM}
chnfsdom
#
chnfsrtd -a ${IREALM} ${NFSDOM}
chnfsrtd
#
exportfs -ua
/etc/nfs.clean
chnfs -r /exports
/etc/nfs.clean
#
chnfs -S -B
#
cp /mnt/scripts/exports.v4krb5.fullclients /etc/exports
#
/etc/rc.nfs
exportfs -va
Appendix B. Sample scripts, files, and output 257

Create a full client with legacy KDC server backend
This script can be used to create a full client with a legacy KDC server backend.
Example: B-3 Sample script to create a full client with a legacy backend
#!/bin/ksh
HOST=$(hostname)
#
#
#
unset KRB5CCNAME
#
setclock ${KDCSERV}
#
#config.krb5 -C -d $DNSDOM -r $IREALM -c $KDCSERV -s $KDCSERV
mkkrb5clnt -c $KDCSERV -r $IREALM -s $KDCSERV -d $DNSDOM
#
kinit admin/admin
klist
#
EOF
#
ktadd nfs/${HOST}
EOF
#
nfshostkey -l
#
chnfsdom ${NFSDOM}
chnfsdom
#
chnfsrtd
#
mkdir /nfs
#
chnfs -S -B
#
/etc/rc.nfs

Create a Full Client with KDC and LDAP backend
The script in Example B-4 performs all operations to configure an NFS V4 full
Kerberos client with integrated login.
Example: B-4 Sample script to create a full client with KDC and LDAP backend
#!/bin/ksh
HOST=$(hostname)
LDAPSERV=”nfs407.itsc.austin.ibm.com”
#
#
ADMINDN=”cn=admin”
#
unset KRB5CCNAME
#
#synchronize the system time with the KDC Server
setclock ${KDCSERV}
#
mkkrb5clnt -c $KDCSERV -r $IREALM -s $KDCSERV -d $DNSDOM -l $LDAPSERV -i files
-A -K -T
#
mksecldap -c -h $LDAPSERV -a ${ADMINDN} -p ${SECRET}
#
/usr/sbin/ls-secldapclntd
#
echo “file /usr/lib/security/methods.cfg has to edited”
if grep -p ^KRB5LDAP /usr/lib/security/methods.cfg
then
echo “file /usr/lib/security/methods.cfg contains KRB5LDAP no changes will
occure”
else
echo “file /usr/lib/security/methods.cfg need to be changed”
cp /usr/lib/security/methods.cfg /usr/lib/security/methods.cfg.save
echo “\nKRB5LDAP:\n\toptions = db=LDAP,auth=KRB5” >>
/usr/lib/security/methods.cfg
fi
#
chsec -f /etc/security/user -s default -a registry=KRB5LDAP
chsec -f /etc/security/user -s default -a “SYSTEM=\”KRB5LDAP OR compat\””
#
chuser registry=files root
chuser SYSTEM=”compat” root

#
kinit admin/admin
klist
#
EOF
#
ktadd nfs/${HOST}
EOF
#
nfshostkey -l
#
kdestroy
kinit -kt /etc/krb5/krb5.keytab nfs/${HOST}
#
chnfsdom ${NFSDOM}
chnfsdom
#
chnfsrtd
#
mkdir /nfs
#
chnfs -S -B
#
/etc/rc.nfs
Script to copy ACLs to an entire directory structure

The script in Example B-5 copies an ACL from a source file or directory to a
destination file or directory. If the -r option is specified and the destination is a
directory, the ACL is also copied to all files or subdirectories underneath the
destination directory.
Example: B-5 Sample script for copying an ACL (with recursive option)
#!/usr/bin/ksh
#
# copy_acl.sh
#
# Copy the ACL for the given source file/directory to other files/directories
#
# Name of this script

scrname=${0##*/}

#
# Functions
#
function usage {
echo "Usage: $scrname [-r] <source> <dest>"
echo " where"
echo " -r indicates a recursive copy"
echo " (copy ACL to all files and directories below and including"
echo " the destination.)"
echo " <source> = the name of the file or directory to copy the ACL from"
echo " <dest> = the name of the file or directory to copy the ACL to"
exit 1
}
if [[ $# -eq 0 ]]
then
usage
fi
#
# Process input parameters
#
if [[ "$1" = "-r" ]]; then

SETSUBTREE="true"
shift
else
SETSUBTREE="false"
fi
if [[ -n "$1" ]]; then

SRC_NAME="$1"
else
usage
fi
if [[ -n "$2" ]]; then

DEST_NAME="$2"
else
usage
fi
#
# Initialize other variables
#

NBERR=0
TMP_ACLFILE="/tmp/.AIXACL_$$"
if [[ -e "${SRC_NAME}" ]]; then

aclget -o "${TMP_ACLFILE}" "${SRC_NAME}"
NBERR=$?
else
echo "Source \"${SRC_NAME}\" does not exist"
NBERR=1
fi
if [[ "${NBERR}" -eq 0 ]]; then

if [[ -e "${DEST_NAME}" ]]; then
if [[ -d "${DEST_NAME}" && "${SETSUBTREE}" = "true" ]]; then
find "${DEST_NAME}" -print | while read NAME
do
aclput -i "${TMP_ACLFILE}" "${NAME}"
(( NBERR += $? ))
ls -dl "${NAME}"
done
else
aclput -i "${TMP_ACLFILE}" "${DEST_NAME}"
(( NBERR += $? ))
ls -dl "${DEST_NAME}"
fi
else
echo "Destination \"${DEST_NAME}\" does not exist"
NBERR=1
fi
fi
rm -f "${TMP_ACLFILE}"
exit ${NBERR}
Windows command script to run ktpass

The Windows command script in Example B-6 was used to add the NFS service
principals for the created Windows Active Directory users.
Example: B-6 Sample ktpass_WinKDC command script

C:\PROGRA~1\SUPPOR~1\ktpass.exe -princ nfs/[email protected]
-mapuser nfs403 -pass nfs403 -out c:\misc\nfs403.keytab
C:\PROGRA~1\SUPPOR~1\ktpass.exe -princ nfs/[email protected]
-mapuser nfs405 -pass nfs405 -out c:\misc\nfs405.keytab

Script to gather additional information for local AIX software support
The script in Example B-7 gathers additional information that the AIX snap
command does not gather. It is not intended as a substitute for the snap
command but should be used as an addition to the snap command and provided
to your local AIX Support Center.
Example: B-7 nfs_pd.script to gather additional information for IBM AIX support
#! /bin/ksh
# ****************************README***********************************
# SPECIAL NOTICES
#
# Information in this document is correct to the best of our
# knowledge at the time of this writing.
# Please use this information with care. IBM will not be
# responsible for damages of any kind resulting from its use.
# The use of this information is the sole responsibility of
# the customer and depends on the customer's ability to eval-
# uate and integrate this information into the customer's
# operational environment.
#
# This script will gather information about your networking environment
# so that IBM may attempt to determine why your computer is experiencing
# problems. You will need to run this script as the root user. If you
# do not have enough space in the /tmp filesystem, you may need to increase
# the size of this filesystem.
# In order to run the script, you will need to give it execute permissions.
# In order to do this, make sure your working directory is the directory
# where the script is located. Then issue -
# chmod +x nfs.script
#
# To run the script from this directory, type -
# ./nfs_pd.script
# *********************************************************************
#
# In addition to this script, you may be asked to supply an iptrace with
# your testcase. If you are asked to do so, the syntax is as follows -
# startsrc -s iptrace -a "-a /tmp/iptrace.bin"
#
# When you wish to stop the trace, issue -
# stopsrc -s iptrace
# You can tar this file along with the rest of the test case and upload it
# following the instructions that the script outlines.
#
# ******************************************************************
clear
echo ""

echo "******************************WELCOME***************************"
echo ""
echo "************************EXECUTING NFS SCRIPT********************"
echo ""
echo "This script will generate a TESTCASE which should be sent to IBM"
echo "to be analyzed by AIX Software Support."
echo ""
echo "Please enter your 5 digit PMR number: "
read pmr
echo "Please enter you 3 digit branch number: "
read branch
echo "Please enter you 3 digit country code: "
read country
echo ""
echo "**************************Thank You*****************************"
echo ""
echo "**************Gathering TESTCASE...Please Standby...************"
echo ""
if [ -d /tmp/ibm ] ; then
echo ""
echo "Unable to create the /tmp/ibm directory because it already exists."
echo ""
echo "Exiting the script."
echo ""
exit
else
mkdir /tmp/ibm
fi
/usr/bin/date > /tmp/ibm/date.start

/usr/bin/uptime > /tmp/ibm/uptime.out
/usr/bin/uname -M > /tmp/ibm/uname-M.out
/usr/bin/uname -n > /tmp/ibm/uname-n.out
/usr/bin/hostname > /tmp/ibm/hostname.out
/usr/sbin/bootinfo -K > /tmp/ibm/bootinfo-K.out
echo "Gathering config files ..."

/usr/bin/cp /etc/inetd.conf /tmp/ibm/inetd.conf
/usr/bin/cp /etc/services /tmp/ibm/services
/usr/bin/cp /etc/inittab /tmp/ibm/inittab
/usr/bin/cp /etc/hosts /tmp/ibm/hosts
/usr/bin/cp /etc/filesystems /tmp/ibm/filesystems
if [ -f /etc/auto* ]; then
/usr/bin/cp /etc/auto* /tmp/ibm/
fi
if [ -f /etc/exports ] ; then

/usr/bin/cp /etc/exports /tmp/ibm/exports
fi
if [ -f /etc/xtab ] ; then
/usr/bin/cp /etc/xtab /tmp/ibm/xtab
fi
if [ -f /etc/rmtab ] ; then
/usr/bin/cp /etc/rmtab /tmp/ibm/rmtab
fi
if [ -f /etc/netsvc.conf ] ; then
/usr/bin/cp /etc/netsvc.conf /tmp/ibm/netsvc.conf
fi
if [ -f /etc/resolv.conf ] ; then
/usr/bin/cp /etc/resolv.conf /tmp/ibm/resolv.conf
fi
if [ -f /etc/irs.conf ] ; then
/usr/bin/cp /etc/irs.conf /tmp/ibm/irs.conf
fi
if [ -f /etc/rc.nfs ] ; then
/usr/bin/cp /etc/rc.nfs /tmp/ibm/rc.nfs
fi
if [ -f /etc/nfs/hostkey ] ; then
/usr/bin/cp /etc/nfs/hostkey /tmp/ibm/hostkey
fi
if [ -f /etc/nfs/local_domain ] ; then
/usr/bin/cp /etc/nfs/local_domain /tmp/ibm/local_domain
fi
if [ -f /etc/nfs/realm.map ] ; then
/usr/bin/cp /etc/nfs/realm.map /tmp/ibm/realm.map
fi
if [ -f /etc/nfs/princmap ] ; then
/usr/bin/cp /etc/nfs/princmap /tmp/ibm/princmap
fi
if [ -f /etc/nfs/security_default ] ; then
/usr/bin/cp /etc/nfs/security_default /tmp/ibm/security_default
fi
echo "Gathering TCP related command outputs ..."

/usr/bin/netstat -v > /tmp/ibm/netstat-v.out

/usr/bin/netstat -in > /tmp/ibm/netstat-in.out
/usr/bin/netstat -rn > /tmp/ibm/netstat-rn.out
/usr/bin/netstat -an > /tmp/ibm/netstat-an.out
/usr/bin/netstat -s > /tmp/ibm/netstat-s.out
/usr/bin/netstat -m > /tmp/ibm/netstat-m.out
echo "Gathering OS related command ouput ..."

/usr/sbin/instfix -i > /tmp/ibm/instfix.out 2>&1
grep AIX /tmp/ibm/instfix.out > /tmp/ibm/aix_ml
/usr/sbin/lsattr -El sys0 > /tmp/ibm/sys0.out
/usr/sbin/lsattr -El inet0 > /tmp/ibm/inet0.out
/usr/sbin/lsattr -El aio0 > /tmp/ibm/aio0.out
/usr/bin/lssrc -ls inetd > /tmp/ibm/inetd.out
/usr/bin/lssrc -a > /tmp/ibm/lssrc-a.out
/usr/bin/lssrc -g tcpip > /tmp/ibm/lssrc-tcp.status
/usr/bin/lssrc -g nfs > /tmp/ibm/lssrc-nfs.status
/usr/bin/lslpp -h > /tmp/ibm/lslpp-h.out
/usr/bin/lppchk -v > /tmp/ibm/lppchk-v.out
/usr/bin/lppchk -c bos.net.* > /tmp/ibm/lppchk-c.out
/usr/sbin/scls -l > /tmp/ibm/scls-l.out
/usr/bin/ls -l /etc/*.conf > /tmp/ibm/conf.out
/usr/bin/what /usr/lib/drivers/netinet > /tmp/ibm/what-netinet
/usr/bin/cp /etc/trcfmt /tmp/ibm/trcfmt
/usr/bin/trcnm -a > /tmp/ibm/namelist
/usr/bin/errpt -a > /tmp/ibm/errpt-a.out

/usr/bin/ps -ef > /tmp/ibm/ps-ef.out
/usr/bin/ps aux > /tmp/ibm/psaux.out
/usr/bin/env > /tmp/ibm/env.out
/usr/bin/env | grep -i krb > /tmp/ibm/env_krb.out
/usr/bin/df -k > /tmp/ibm/df-k.out
/usr/sbin/no -a > /tmp/ibm/no.out
/usr/sbin/nfso -a > /tmp/ibm/nfso.out
/usr/bin/rpcinfo -p > /tmp/ibm/rpcinfo.out
/usr/sbin/arp -an > /tmp/ibm/arp.out
/usr/sbin/lsdev -Cc adapter > /tmp/ibm/adapter.out
/usr/sbin/lsdev -Cc if > /tmp/ibm/lsdev-if.out
/usr/sbin/lscfg -v > /tmp/ibm/lscfg-v.out
echo "Gathering NFS related command output ..."

/usr/bin/showmount -e > /tmp/ibm/showmount-e.out
/usr/sbin/mount > /tmp/ibm/mount.out
/usr/sbin/nfsstat -m > /tmp/ibm/nfsstat-m.out
/usr/sbin/nfsstat -cr > /tmp/ibm/nfsstat-cr.out
/usr/sbin/nfs4cl showfs > /tmp/ibm/nfs4cl_showfs.out
/usr/sbin/nfs4cl showstat > /tmp/ibm/nfs4cl_showstat.out
/usr/sbin/chnfsdom > /tmp/ibm/chnfsdom.out
/usr/bin/netstat -s > /tmp/ibm/netstat-s.out2

/usr/bin/netstat -an > /tmp/ibm/netstat-an.out2
if [ -f `which lsof` ] ; then

lsof > /tmp/ibm/lsof.out
fi
ADAPTERLIST="ent tok atm mpc fddi escon cat"

for i in $ADAPTERLIST
do
for j in `/usr/sbin/lsdev -Cc adapter|grep "^$i"|awk '{print $1}'`
do
/usr/sbin/lsattr -El $j > /tmp/ibm/lsattr-El.$j.out
/usr/sbin/lscfg -v $j > /tmp/ibm/lscfg-v.$j.out
done
done
for i in `/usr/bin/netstat -in|grep -Ev "::1|Name|ink"|awk '{print $1}'`

do
/etc/ifconfig $i > /tmp/ibm/ifconfig.$i.out
/usr/sbin/lsattr -El $i > /tmp/ibm/lsattr-El.$i.out
done
/usr/bin/netstat -v > /tmp/ibm/netstat-v.out2

/usr/bin/netstat -in > /tmp/ibm/netstat-in.out2
/usr/bin/netstat -m > /tmp/ibm/netstat-m.out2
/usr/bin/netstat -v > /tmp/ibm/netstat-v.out3
/usr/bin/netstat -in > /tmp/ibm/netstat-in.out3
/usr/bin/netstat -m > /tmp/ibm/netstat-m.out3
/usr/sbin/nfs4cl showfs > /tmp/ibm/nfs4cl_showfs.out2
/usr/sbin/nfs4cl showstat > /tmp/ibm/nfs4cl_showstat.out2
/usr/bin/ps -ef > /tmp/ibm/ps-ef.out2

/usr/bin/ps aux > /tmp/ibm/psaux.out2
/usr/bin/date > /tmp/ibm/date.stop

cd /tmp/ibm
/usr/bin/tar -cf /tmp/$pmr.$branch.$country.tar *
cd /tmp
/usr/bin/compress $pmr.$branch.$country.tar
/usr/bin/rm -rf /tmp/ibm
clear
echo ""
echo "The script has completed. "
echo "There should be a "$pmr.$branch.$country".tar.Z file in your "

echo "/tmp directory. Please ftp this file to your local IBM Support "
echo "Centre - your local IBM Support representative should be able to "
echo "provide the relevant details."
echo ""
echo "**************************Thank You*****************************"
echo ""
echo "Please be sure to use the bin mode within FTP, otherwise "
echo "the tar file may be corrupted."
echo ""
echo "*************Thank You For Using IBM AIX Software Support*******"
echo ""
Sample client Kerberos configuration files

This section provides the Kerberos configuration files from our test environment.
Kerberos configuration file /etc/krb5/krb5.conf with legacy backend

This is the sample configuration file used on clients with a legacy KDC database.
Example: B-8 Sample /etc/krb5/krb5.conf file with a legacy KDC database

[libdefaults]
[realms]
REALM2.IBM.COM = {
default_domain = ibm.com
}
[domain_realm]
.ibm.com = REALM2.IBM.COM
[logging]

Kerberos configuration file /etc/krb5/krb5.conf with LDAP backend
Example B-9 shows the sample configuration file used on the clients with
Kerberos LDAP backend.
Example: B-9 Sample /etc/krb5/krb5.conf with KRB5 and LDAP backend

[libdefaults]
use_ldap_lookup = 1
ldap_server = nfs407.itsc.austin.ibm.com
[realms]
REALM1.IBM.COM = {
}
[domain_realm]
[logging]
Kerberos configuration file /etc/krb5/krb5.conf with Windows Active

Directory backend
Example B-10 shows the sample configuration file used on clients with a
Miscrosoft Windows Active Directory backend.
Example: B-10 Sample /etc/krb5/krb5.conf with Windows Active Directory backend

[libdefaults]
default_realm = KDC.ITSC.AUSTIN.IBM.COM
default_tkt_enctypes = des-cbc-md5 des-cbc-crc
default_tgs_enctypes = des-cbc-md5 des-cbc-crc

[realms]
KDC.ITSC.AUSTIN.IBM.COM = {
admin_server = nfs409.kdc.itsc.austin.ibm.com:749
default_domain = kdc.itsc.austin.ibm.com
}
[domain_realm]
.kdc.itsc.austin.ibm.com = KDC.ITSC.AUSTIN.IBM.COM
nfs409.kdc.itsc.austin.ibm.com = KDC.ITSC.AUSTIN.IBM.COM
[logging]
LDIF sample file for KDC

Example B-11 shows a sample LDIF file for the KDC LDAP container to work
with integrated logon into AIX. It is used in 5.9, “Preparing the system for Tivoli
Directory Server and Kerberos V5” on page 155.
Example: B-11 Sample realm_add_ibm.ldif file

# The suffix “ou=Austin, o=IBM, c=US” should be defined before attempting to
# load this data. Or change the suffix to be an already defined object.
# Change all references of YOURHOSTNAME.AUSTIN.IBM.COM to be your realm name
#
# version: 1
dn: o=IBM, c=US

objectclass: top
objectclass: organization
o: IBM
dn: krbrealmName-V2=REALM2.IBM.COM, o=IBM, c=US

objectclass: KrbRealm-V2
objectclass: KrbRealmExt
krbrealmName-V2: REALM2.IBM.COM
krbprincSubtree: krbrealmName-V2=REALM2.IBM.COM, o=IBM, c=US
krbDeleteType: 3
dn: cn=principal, krbrealmName-V2=REALM2.IBM.COM, o=IBM, c=US

cn: principal

dn: cn=policy, krbrealmName-V2=REALM2.IBM.COM, o=IBM, c=US
cn: policy
Sample iptrace output

In this section we provide some sample iptrace output taken while setting up and
debugging our test environment.
Successful authentication during mount request

Example B-12 shows the iptrace taken during a succesful authentication while
carrying out an NFS V4 mount. We outlined the three major packages, which are:
򐂰 Kerberos ticket request KRB5 TGS-REQ
򐂰 Valid Kerberos ticket reply KRB5 TGS-REP
򐂰 The NFS V4 compound call
Example: B-12 Sample iptrace output showing successful authentication during mount
No. Time Source Destination Protocol Info
167 20.373421 9.3.5.173 9.3.4.71 KRB5
TGS-REQ
Frame 167 (1311 bytes on wire, 1311 bytes captured)

Arrival Time: Aug 6, 2004 15:30:07.127965000
Time delta from previous packet: 0.000276000 seconds
Time since reference or first frame: 20.373421000 seconds
Frame Number: 167
Packet Length: 1311 bytes
Capture Length: 1311 bytes
Ethernet II, Src: 00:02:55:af:1c:8d, Dst: 00:09:12:48:3c:02
Destination: 00:09:12:48:3c:02 (Cisco_48:3c:02)
Source: 00:02:55:af:1c:8d (Ibm_af:1c:8d)
Type: IP (0x0800)
Internet Protocol, Src Addr: 9.3.5.173 (9.3.5.173), Dst Addr: 9.3.4.71
(9.3.4.71)
Version: 4
Header length: 20 bytes
Differentiated Services Field: 0x00 (DSCP 0x00: Default; ECN: 0x00)
0000 00.. = Differentiated Services Codepoint: Default (0x00)
.... ..0. = ECN-Capable Transport (ECT): 0
.... ...0 = ECN-CE: 0
Total Length: 1297
Identification: 0x0586 (1414)
Flags: 0x00

0... = Reserved bit: Not set
.0.. = Don’t fragment: Not set
..0. = More fragments: Not set
Fragment offset: 0
Time to live: 30
Protocol: UDP (0x11)
Header checksum: 0x765d (correct)
Source: 9.3.5.173 (9.3.5.173)
Destination: 9.3.4.71 (9.3.4.71)
User Datagram Protocol, Src Port: 32925 (32925), Dst Port: kerberos (88)
Source port: 32925 (32925)
Destination port: kerberos (88)
Length: 1277
Checksum: 0xc985 (correct)
Kerberos TGS-REQ
Pvno: 5
MSG Type: TGS-REQ (12)
padata: PA-TGS-REQ
Type: PA-TGS-REQ (1)
Value: 6E82045F3082045BA003020105A10302... AP-REQ
Pvno: 5
MSG Type: AP-REQ (14)
Padding: 0
APOptions: 00000000
.0.. .... .... .... .... .... .... .... = Use Session Key:
Do NOT use the session key to encrypt the ticket
..0. .... .... .... .... .... .... .... = Mutual required:
Mutual authentication is NOT required
Ticket
Tkt-vno: 5
Realm: KDC.ITSC.AUSTIN.IBM.COM
Server Name (Unknown): krbtgt KDC.ITSC.AUSTIN.IBM.COM
Name-type: Unknown (0)
Name: krbtgt
Name: KDC.ITSC.AUSTIN.IBM.COM
enc-part rc4-hmac
Encryption type: rc4-hmac (23)
Kvno: 2
enc-part: 68A3BA1DDDD3C5E05848A0D3BA2C1F05...
Authenticator des-cbc-md5
Encryption type: des-cbc-md5 (3)
Authenticator data: 7333893FE36467A6E3AB514BE8740E20...
KDC_REQ_BODY
Padding: 0
KDCOptions: 00800000 (Renewable)
.0.. .... .... .... .... .... .... .... = Forwardable: Do NOT use
forwardable tickets
..0. .... .... .... .... .... .... .... = Forwarded: This is NOT a
forwarded ticket

...0 .... .... .... .... .... .... .... = Proxyable: Do NOT use
proxiable tickets
.... 0... .... .... .... .... .... .... = Proxy: This ticket has
NOT been proxied
.... .0.. .... .... .... .... .... .... = Allow Postdate: We do NOT
allow the ticket to be postdated
.... ..0. .... .... .... .... .... .... = Postdated: This ticket is
NOT postdated
.... .... 1... .... .... .... .... .... = Renewable: This ticket is
RENEWABLE
.... .... ...0 .... .... .... .... .... = Opt HW Auth: False
.... .... .... ...0 .... .... .... .... = Canonicalize: This is NOT
a canonicalized ticket request
.... .... .... .... .... .... ..0. .... = Disable Transited Check:
Transited checking is NOT disabled
.... .... .... .... .... .... ...0 .... = Renewable OK: We do NOT
accept renewed tickets
.... .... .... .... .... .... .... 0... = Enc-Tkt-in-Skey: Do NOT
encrypt the tkt inside the skey
.... .... .... .... .... .... .... ..0. = Renew: This is NOT a
request to renew a ticket
.... .... .... .... .... .... .... ...0 = Validate: This is NOT a
request to validate a postdated ticket
Server Name (Principal): nfs nfs403
Name-type: Principal (1)
Name: nfs
Name: nfs403
till: 2004-08-07 06:31:00 (Z)
Nonce: 1091824267
Encryption Types: des-cbc-md5 des-cbc-crc
Encryption type: des-cbc-crc (1)

168 20.375701 9.3.4.71 9.3.5.173 KRB5
TGS-REP

Arrival Time: Aug 6, 2004 15:30:07.130245000
Frame Number: 168
Ethernet II, Src: 00:09:12:48:3c:02, Dst: 00:02:55:af:1c:8d
Destination: 00:02:55:af:1c:8d (Ibm_af:1c:8d)
Source: 00:09:12:48:3c:02 (Cisco_48:3c:02)
Type: IP (0x0800)

(9.3.5.173)
Version: 4
.... ...0 = ECN-CE: 0
Total Length: 1268
Identification: 0xb478 (46200)
Flags: 0x00
Fragment offset: 0
Time to live: 127
Header checksum: 0x6687 (correct)
Source: 9.3.4.71 (9.3.4.71)
Destination: 9.3.5.173 (9.3.5.173)
User Datagram Protocol, Src Port: kerberos (88), Dst Port: 32925 (32925)
Source port: kerberos (88)
Destination port: 32925 (32925)
Length: 1248
Checksum: 0x5cb7 (correct)
Kerberos TGS-REP
Pvno: 5
MSG Type: TGS-REP (13)
Client Realm: KDC.ITSC.AUSTIN.IBM.COM
Client Name (Principal): nfs nfs403
Name: nfs
Name: nfs403
Ticket
Tkt-vno: 5
Name: nfs
Name: nfs403
enc-part des-cbc-md5
Kvno: 5
enc-part: 2F97563D681D6BB1DC8B010025F82DB9...
enc-part: 76C193F0DB908DB403831A7B45167357...

173 20.385891 9.3.5.173 9.3.5.173 NFS V4
NULL Call (Reply In 175)

Arrival Time: Aug 6, 2004 15:30:07.140435000
Frame Number: 173
Raw packet data
No link information available
(9.3.5.173)
Version: 4
.... ...0 = ECN-CE: 0
Total Length: 1288
Identification: 0x058c (1420)
Flags: 0x04 (Don’t Fragment)
.1.. = Don’t fragment: Set
Fragment offset: 0
Time to live: 60
Protocol: TCP (0x06)
Header checksum: 0x0000 (incorrect, should be 0x1705)
Source: 9.3.5.173 (9.3.5.173)
Destination: 9.3.5.173 (9.3.5.173)
Transmission Control Protocol, Src Port: 32775 (32775), Dst Port: 2049 (2049),
Seq: 1, Ack: 1, Len: 1236
Source port: 32775 (32775)
Sequence number: 1 (relative sequence number)
Next sequence number: 1237 (relative sequence number)
Acknowledgement number: 1 (relative ack number)
Flags: 0x0018 (PSH, ACK)
0... .... = Congestion Window Reduced (CWR): Not set
.0.. .... = ECN-Echo: Not set
..0. .... = Urgent: Not set
...1 .... = Acknowledgment: Set
.... 1... = Push: Set
.... .0.. = Reset: Not set
.... ..0. = Syn: Not set

.... ...0 = Fin: Not set
Window size: 262140
Checksum: 0xa5d8 (correct)
Options: (12 bytes)
NOP
NOP
Time stamp: tsval 1091824482, tsecr 1091824482
Remote Procedure Call, Type:Call XID:0x77709946
Fragment header: Last fragment, 1232 bytes
1... .... .... .... .... .... .... .... = Last Fragment: Yes
.000 0000 0000 0000 0000 0100 1101 0000 = Fragment Length: 1232
XID: 0x77709946 (2003867974)
Message Type: Call (0)
RPC Version: 2
Program: NFS (100003)
Program Version: 4
Procedure: NULL (0)
The reply to this request is in frame 175
Credentials
Flavor: RPCSEC_GSS (6)
Length: 20
GSS Version: 1
GSS Procedure: RPCSEC_GSS_INIT (1)
GSS Sequence Number: 1
GSS Service: rpcsec_gss_svc_none (1)
GSS Context: <EMPTY>
length: 0
contents: <EMPTY>
Verifier
Flavor: AUTH_NULL (0)
Length: 0
Network File System
Program Version: 4
V4 Procedure: NULL (0)
GSS Token
GSS Token Length: 1165
GSS-API
OID: 1.2.840.113554.1.2.2 (iso.2.840.113554.1.2.2) (KRB5 - Kerberos
5)
krb5_blob: 01006E82047830820474A003020105A1...
krb5_tok_id: KRB5_AP_REQ (0x0001)
Kerberos AP-REQ
Pvno: 5
Padding: 0
APOptions: 20000000 (Mutual required)
.0.. .... .... .... .... .... .... .... = Use Session
Key: Do NOT use the session key to encrypt the ticket

..1. .... .... .... .... .... .... .... = Mutual
required: MUTUAL authentication is REQUIRED
Ticket
Tkt-vno: 5
Name: nfs
Name: nfs403
Kvno: 5
enc-part: 2F97563D681D6BB1DC8B010025F82DB9...
Authenticator data: 5B51E74785A4CBB0369C571DED75B7EB...

175 20.431180 9.3.5.173 9.3.5.173 NFS V4
NULL Reply (Call In 173)

Arrival Time: Aug 6, 2004 15:30:07.185724000
Frame Number: 175
Raw packet data
(9.3.5.173)
Version: 4
.... ...0 = ECN-CE: 0
Total Length: 276
Identification: 0x0596 (1430)
Fragment offset: 0
Time to live: 60
Header checksum: 0x0000 (incorrect, should be 0x1aef)
Source: 9.3.5.173 (9.3.5.173)

Destination: 9.3.5.173 (9.3.5.173)
Seq: 1, Ack: 1237, Len: 224
Source port: 2049 (2049)
.... 1... = Push: Set
.... ..0. = Syn: Not set
.... ...0 = Fin: Not set
Window size: 262140
Checksum: 0x31fd (correct)
Options: (12 bytes)
NOP
NOP
Remote Procedure Call, Type:Reply XID:0x77709946
.000 0000 0000 0000 0000 0000 1101 1100 = Fragment Length: 220
XID: 0x77709946 (2003867974)
Message Type: Reply (1)
Program Version: 4
Procedure: NULL (0)
Reply State: accepted (0)
This is a reply to a request in frame 173
Time from request: 0.045289000 seconds
Verifier
GSS Token
GSS-API
OID: 1.2.840.113554.1.2.2 (iso.2.840.113554.1.2.2) (KRB5 -
Kerberos 5)
krb5_blob: 01010000FFFFFFFFE975FD4C7682CD7D...
krb5_tok_id: KRB5_GSS_GetMIC (0x0101)
krb5_sgn_alg: DES MAC MD5 (0x0000)
krb5_snd_seq: E975FD4C7682CD7D
krb5_sgn_cksum: 8E63BCC171B105D2
Accept State: RPC executed successfully (0)

Network File System
Program Version: 4
V4 Procedure: NULL (0)
GSS Context: <DATA>
length: 4
contents: <DATA>
GSS Major Status: 0
GSS Minor Status: 0
GSS Sequence Window: 48
GSS Token
GSS-API
OID: 1.2.840.113554.1.2.2 (iso.2.840.113554.1.2.2) (KRB5 - Kerberos
5)
krb5_blob: 02006F71306FA003020105A10302010F...
krb5_tok_id: KRB5_AP_REP (0x0002)
Kerberos AP-REP
Pvno: 5
MSG Type: AP-REP (15)
enc-part: A949C61E03CE98601CEC6B2A55D24725...

176 20.434340 9.3.5.173 9.3.5.173 NFS V4
COMPOUND Call (Reply In 177)

Arrival Time: Aug 6, 2004 15:30:07.188884000
Frame Number: 176
Raw packet data
(9.3.5.173)
Version: 4
.... ...0 = ECN-CE: 0
Total Length: 212
Identification: 0x059b (1435)

Fragment offset: 0
Time to live: 60
Header checksum: 0x0000 (incorrect, should be 0x1b2a)
Source: 9.3.5.173 (9.3.5.173)
Destination: 9.3.5.173 (9.3.5.173)
Seq: 1237, Ack: 225, Len: 160
Source port: 32775 (32775)
.... 1... = Push: Set
.... ..0. = Syn: Not set
.... ...0 = Fin: Not set
Window size: 262140
Checksum: 0x56db (correct)
Options: (12 bytes)
NOP
NOP
SEQ/ACK analysis
This is an ACK to the segment in frame: 175
The RTT to ACK the segment was: 0.003160000 seconds
Remote Procedure Call, Type:Call XID:0x7770994d
.000 0000 0000 0000 0000 0000 1001 1100 = Fragment Length: 156
XID: 0x7770994d (2003867981)
Message Type: Call (0)
RPC Version: 2
Program Version: 4
Procedure: COMPOUND (1)
The reply to this request is in frame 177
Credentials
Length: 24
GSS Version: 1
GSS Procedure: RPCSEC_GSS_DATA (0)

GSS Sequence Number: 2
GSS Service: rpcsec_gss_svc_none (1)
GSS Context: <DATA>
length: 4
contents: <DATA>
Verifier
GSS Token
GSS-API
OID: 1.2.840.113554.1.2.2 (iso.2.840.113554.1.2.2) (KRB5 -
Kerberos 5)
krb5_blob: 01010000FFFFFFFF28D134F10157FBBD...
krb5_tok_id: KRB5_GSS_GetMIC (0x0101)
krb5_sgn_alg: DES MAC MD5 (0x0000)
krb5_snd_seq: 28D134F10157FBBD
krb5_sgn_cksum: 9971CB9E15D6C1A5
Network File System
Program Version: 4
V4 Procedure: COMPOUND (1)
Tag: nfs4pathlookup
length: 14
contents: nfs4pathlookup
fill bytes: opaque data
minorversion: 0
Operations (count: 3)
Opcode: PUTROOTFH (24)
Opcode: LOOKUP (15)
Filename: exports
length: 7
contents: exports
fill bytes: opaque data
Opcode: GETFH (10)
Unsuccessful authentication during mount request

Example B-13 shows the iptrace output taken during an unsuccessful
authentication that was done while carrying out an NFS V4 mount. We outlined
the two major packages, which are:
򐂰 Kerberos ticket request KRB5 TGS-REQ
򐂰 Invalid Kerberos ticket reply KRB5 TGS-REP
Example: B-13 iptrace output showing unsuccessful authentication during mount

115 10.421160 9.3.5.175 9.3.4.71 KRB5
TGS-REQ

Arrival Time: Aug 6, 2004 15:34:37.198215000
Frame Number: 115
Ethernet II, Src: 00:02:55:af:12:7a, Dst: 00:09:12:48:3c:02
Destination: 00:09:12:48:3c:02 (Cisco_48:3c:02)
Source: 00:02:55:af:12:7a (Ibm_af:12:7a)
Type: IP (0x0800)
(9.3.4.71)
Version: 4
.... ...0 = ECN-CE: 0
Total Length: 1321
Identification: 0x8efa (36602)
Flags: 0x00
Fragment offset: 0
Time to live: 30
Header checksum: 0xecce (correct)
Source: 9.3.5.175 (9.3.5.175)
Destination: 9.3.4.71 (9.3.4.71)
User Datagram Protocol, Src Port: 33735 (33735), Dst Port: kerberos (88)
Source port: 33735 (33735)
Destination port: kerberos (88)
Length: 1301
Checksum: 0xe902 (correct)
Kerberos TGS-REQ
Pvno: 5
MSG Type: TGS-REQ (12)
padata: PA-TGS-REQ
Type: PA-TGS-REQ (1)
Value: 6E82045F3082045BA003020105A10302... AP-REQ
Pvno: 5
Padding: 0
APOptions: 00000000
.0.. .... .... .... .... .... .... .... = Use Session Key:
Do NOT use the session key to encrypt the ticket

..0. .... .... .... .... .... .... .... = Mutual required:
Mutual authentication is NOT required
Ticket
Tkt-vno: 5
Server Name (Unknown): krbtgt KDC.ITSC.AUSTIN.IBM.COM
Name-type: Unknown (0)
Name: krbtgt
Name: KDC.ITSC.AUSTIN.IBM.COM
enc-part rc4-hmac
Encryption type: rc4-hmac (23)
Kvno: 2
enc-part: 242AC4DAA22D420C592754948B7C2399...
Authenticator data: 0EDE8A3F1B5C370A818DC7CFEF1BC12B...
KDC_REQ_BODY
Padding: 0
KDCOptions: 00800000 (Renewable)
.0.. .... .... .... .... .... .... .... = Forwardable: Do NOT use
forwardable tickets
..0. .... .... .... .... .... .... .... = Forwarded: This is NOT a
forwarded ticket
...0 .... .... .... .... .... .... .... = Proxyable: Do NOT use
proxiable tickets
.... 0... .... .... .... .... .... .... = Proxy: This ticket has
NOT been proxied
.... .0.. .... .... .... .... .... .... = Allow Postdate: We do NOT
allow the ticket to be postdated
.... ..0. .... .... .... .... .... .... = Postdated: This ticket is
NOT postdated
.... .... 1... .... .... .... .... .... = Renewable: This ticket is
RENEWABLE
.... .... ...0 .... .... .... .... .... = Opt HW Auth: False
.... .... .... ...0 .... .... .... .... = Canonicalize: This is NOT
a canonicalized ticket request
.... .... .... .... .... .... ..0. .... = Disable Transited Check:
Transited checking is NOT disabled
.... .... .... .... .... .... ...0 .... = Renewable OK: We do NOT
accept renewed tickets
.... .... .... .... .... .... .... 0... = Enc-Tkt-in-Skey: Do NOT
encrypt the tkt inside the skey
.... .... .... .... .... .... .... ..0. = Renew: This is NOT a
request to renew a ticket
.... .... .... .... .... .... .... ...0 = Validate: This is NOT a
request to validate a postdated ticket
Server Name (Principal): nfs nfs403.kdc.itsc.austin.ibm.com

Name: nfs
Name: nfs403.kdc.itsc.austin.ibm.com
till: 2004-08-07 06:35:49 (Z)
Nonce: 1091824487
Encryption Types: des-cbc-md5 des-cbc-crc
Encryption type: des-cbc-crc (1)

116 10.422686 9.3.4.71 9.3.5.175 KRB5 KRB
Error: KRB5KDC_ERR_S_PRINCIPAL_UNKNOWN

Arrival Time: Aug 6, 2004 15:34:37.199741000
Frame Number: 116
Ethernet II, Src: 00:09:12:48:3c:02, Dst: 00:02:55:af:12:7a
Destination: 00:02:55:af:12:7a (Ibm_af:12:7a)
Source: 00:09:12:48:3c:02 (Cisco_48:3c:02)
Type: IP (0x0800)
(9.3.5.175)
Version: 4
.... ...0 = ECN-CE: 0
Total Length: 150
Identification: 0xb67d (46717)
Flags: 0x00
Fragment offset: 0
Time to live: 127
Header checksum: 0x68de (correct)
Source: 9.3.4.71 (9.3.4.71)
Destination: 9.3.5.175 (9.3.5.175)
User Datagram Protocol, Src Port: kerberos (88), Dst Port: 33735 (33735)
Source port: kerberos (88)
Length: 130
Checksum: 0x31df (correct)
Kerberos KRB-ERROR

Pvno: 5
MSG Type: KRB-ERROR (30)
stime: 2004-08-06 20:35:56 (Z)
susec: 211147
error_code: KRB5KDC_ERR_S_PRINCIPAL_UNKNOWN (7)
Server Name (Principal): nfs nfs403.kdc.itsc.austin.ibm.com
Name: nfs
Name: nfs403.kdc.itsc.austin.ibm.com

C
Appendix C. AIX 5.3 NFS quick reference

The purpose of this appendix is to provide a quick reference sheet for NFS on
AIX 5.3. It includes the following sections:
򐂰 NFS configuration files
򐂰 NFS daemons
򐂰 NFS commands
򐂰 Export options
򐂰 mount command options
򐂰 nfso command options

NFS configuration files
/etc/rc.nfs
Starts NFS on boot.
/etc/filesystems
Contains file systems to be mounted.
/etc/exports
Contains NFS export definitions.
/etc/xtab
Contains names of file systems currently exported.
/etc/rmtab
Contains names of machines and the file systems they
have mounted.
/etc/sm
Directory used by rpc.statd
/etc/sm.bak
Directory used by rpc.statd
/etc/state
File used by rpc.statd
/etc/nfs/realm.map
File used by the NFS registry daemon to map
incoming Kerberos principals.
/etc/nfs/local_domain
File containing the local NFS domain.
/etc/nfs/hostkey
File used by the NFS server to specify the Kerberos
host principal and location of the keytab file.
/etc/nfs/princmap
File maps host names to Kerberos principals when the
principal is not the fully qualified domain name of the
server.
/etc/nfs/security_default
File containing the list of security flavors that may be
used by the NFS client.

/etc/bootparms
File containing list of servers that diskless clients can
use to boot from.
/etc/networks
File containing information about networks.
/etc/pcnfsd.conf
File containing options for the rpc.pcnfsd daemon.
/etc/rpc
File containing database information for RPC
programs.
/etc/netgroup
File defining network-wide groups; used for checking
permissions when doing remote mounts, remote
logons, and remote shells.
Appendix C. AIX 5.3 NFS quick reference 289

NFS daemons
/usr/sbin/rpc.lockd
Processes lock requests through the RPC package.
/usr/sbin/rpc.statd
Provides crash-and-recovery functions for the NFS
locking services.
/usr/sbin/biod
Sends the client’s read and write requests to the
server; runs only on the client.
/usr/sbin/rpc.mountd
Answers requests from clients for file system mounts;
runs only on the server
/usr/sbin/nfsd
Starts the daemons that handle a client’s request for
file system operations; runs only on the server.
/usr/sbin/portmap
Maps RPC program numbers to Internet port numbers.
/usr/sbin/rpc.statd
Returns performance statistics obtained from the
kernel.
/usr/sbin/pcnfsd
Handles service requests from PC-NFS clients.
/usr/bin/gssd
New daemon for NFS V4 that services kernel requests
for GSS operations.
/usr/sbin/nfsrgyd
New daemon for NFS V4 that provides a name
translation service for NFS servers and clients.

NFS commands
/usr/sbin/mount
Shows what file systems are mounted on the machine
the command is run on, including name of the server,
and mount options.
/usr/bin/showmount -e [host]
Shows contents of /etc/xtab file on the [host].
/usr/bin/showmount -a [host]
Shows the contents of the /etc/rmtab on the [host].
/usr/sbin/exportfs -va
Exports all file systems defined in /etc/exports and
prints the name of each directory as it is exported.
/usr/sbin/exportfs -vua
Unexports all exported directories and prints the name
of each directory as it is unexported.
/usr/sbin/mknfs
Configures a system to run NFS and starts NFS
daemons.
/usr/sbin/nfso
Configures and lists NFS network options.
/usr/sbin/automount
Mounts an NFS automatically.
/usr/bin/chnfsexp
Changes the attributes of an NFS exported directory.
/usr/sbin/chnfsmnt
Changes the attributes of an NFS mounted directory.
/usr/sbin/lsnfsexp
Displays the characteristics of directories that are
exported with NFS.
/usr/sbin/lsnfsmnt
Displays the characteristics of mounted NFS.
/usr/sbin/mknfsexp
Exports a directory.

/usr/sbin/mknfsmnt
Mounts a directory using NFS.
/usr/sbin/rmnfs
Changes the configuration of NFS in system inittab
and stops the NFS daemons.
/usr/sbin/rmnfsexp
Removes NFS exported directories from a server’s list
of exports.
/usr/sbin/chnfsdom
Changes the local NFS domain.
/usr/sbin/nfs4cl
Displays or modifies current NFS V4 statistics and
properties.
/usr/sbin/nfshostkey
Configures the host key for an NFS server.
/usr/sbin/chnfsim
Changes the NFS foreign identity mappings.

Export options
-o options
-rw
All clients have read-write permission (default).
-ro
All clients have read-only permission.
-rw=Client [:Client]
Exports the directory with read-write permission to the
specified clients. Exports the directory read-only to
clients not in the list.
-access=Client[:Client,...]
Gives mount access to each client listed. A client can
be either a host name or a net group name.
-root=Client[:Client]
Allows root access from the specified clients.
-vers=version_number [:version_number]
Specifies which versions of NFS are allowed to access
the exported directory. Valid versions are 2, 3, and 4.
-exname=external-name
Exports the directory by the specified external name.
The external name must begin with the nfsroot name.
-sec=flavor[:flavor...]
Used to specify a list of security methods that may be
used to access files under the exported directory.
Allowable flavor values are: sys, dh, none, krb5, krb5i
and krb5p.
-nfsroot
Sets the nfsroot to a specified directory. For example,
/exports -nfsroot.

mount command options
-o options
ro
Specifies that the mounted file is read-only.
rw
Specifies that the mounted file is read/write accessible
(default).
fg
Attempts mount in foreground if first attempt is
unsuccessful (default).
bg
Attempts mount in background if first attempt is
unsuccessful.
hard
Retries a request until server responds (default).
soft
Returns an error if the server does not respond.
intr
Allows keyboard interrupts on hard mounts.
nointr
Specifies no keyboard interrupts allowed on hard
mounts.
acl
Requests using the Access Control List RPC program
for this NFS mount.
sec=[flavor1:...:flavorn]
Specifies a list of security methods that may be used
to access files under the mount point. Allowable
security flavors are: sys, dh, krb5, krb5i and krb5p
vers=Version
Specifies NFS version. Options are 2, 3, and 4.
vers=4 is only applicable to AIX 5.3.

nfso command options
Use the nfso command to configure Network File System tuning parameters.
The nfso command sets or displays current or next boot values for Network File
System tuning parameters. This command can also make permanent changes or
defer changes until the next reboot. Whether the command sets or displays a
parameter is determined by the accompanying flag. The -o flag performs both
actions. It can either display the value of a parameter or set a new value for a
parameter.
Important: Extreme care must be taken before values are changed using the
nsfo command. An incorrect change can render the system unusable.
NFS V4 introduces the following new tunable parameters:

utf8
This option enables NFS V4 to perform UTF8
checking.
A value of 1 turns on UTF-8 checking of file names.
A value of 0 turns it off.
utf8_validation
Enables checking of file names for the NFS V 4 client
and server to ensure that they conform to the UTF-8
specification.
A value of 1 turns on UTF-8 checking of file names.
A value of 0 turns it off.
nfs_v4_pdts
Sets the number of tables for memory pools used by
the biods for NFS V4 mounts.
You should use the vmstat -v command to look for
non-zero values in the client file system I/Os blocked
with no fsbuf field.
Increase the number until the blocked I/O count is no
longer incremented during workload. The number
might need to be increased in conjunction with
nfs_v4_vm_bufs.
nfs_v4_vm_bufs
Sets the number of initial free memory buffers used for
each NFS V4 paging device table (pdt) created after
the first table. The very first pdt has a set value of 256,

512, 640, or 1000, depending on system memory. This
initial value is also the default value of each newly
created pdt.
You should use the vmstat -v command to look for
non-zero values in the client file system I/Os blocked
with no fsbuf field.
The nfs_v4_vm_bufs option must be set prior to
nfs_v4_pdts.
Useful examples how the nfso command can be used:

1. To print, in colon-delimited format, a list of all tunable parameters and their
current values, run:
nfso -a -c
2. To list the current and reboot value, range, unit, type, and dependencies of all
tunable parameters managed by the nfso command, run:
nfso -L
3. To list the reboot values for all Network File System tuning parameters, run:
nfso -r -a
4. To set a tunable parameter, for example utf8, to a value of 1, run:
nfso -o utf8=1
5. To set a tunable parameter, for example nfs_v4_pdts, to its default value of 1
at the next reboot,run:
nfso -r -d nfs_v4_pdts
The following references are also useful:
“Network File System (NFS) Overview for System Management” and “TCP/IP
Overview for System Management,” AIX 5L Version 5.3 System User’s Guide:
Communications and Networks, SC23-4909
“Monitoring and Tuning NFS Use,” AIX 5L Version 5.3 Performance Management
Guide, SC23-4905
To view these documents, choose System management guides in the left

navigation bar at:
http://publib.boulder.ibm.com/infocenter/pseries/index.jsp?topic=/com.ibm.aix.d
oc/cmds/aixcmds6/ypserv.htm

Abbreviations and acronyms
ACE Access Control Entry LDAP Lightweight Directory Access
ACL access control list Protocol
AFS Andrew file system LDIF LDAP Data Interchange

Format
AIO asynchronous input/output
LIPKEY Low Infrastructure Public Key
ASCII American Standard Code for
Information Interchange NAS Network Authentication
Service
DCE Distributed Computing
Environment NFS Network File System
DES Data Encryption Standard NFS V2 NFS Version 2
DFS Distributed File System NFS V3 NFS Version 3
DNS Domain Name Service NFS V4 NFS Version 4
EIM Enterprise Identity Mapping NIM Network Installation

Management
FS file system
NIS Network Information Service
GID group ID
NLM Network Lock Manager
GPFS General Parallel File System
NTP Network Time Protocol
GSS-API Generic Security Service
Application Programming OSI open systems interconnection
Interface POSIX Portable Operating System
GUI graphical user interface Interface
IBM International Business RFC Request For Comments

Machines Corporation RFS (AT&T) Remote File System
IEEE Institute of Electrical and RPC remote procedure call
Electronics Engineers SMIT System Management
IETF Internet Engineering Task Interface Tool
Force SPKM Simple Public-Key
IP Internet Protocol Mechanism
ISO International Organization for SSL Secure Sockets Layer
Standardization TCP Transmission Control Protocol
ITSO International Technical UDP User Datagram Protocol
Support Organization
UID User ID
JFS journaled file system
UTF Unicode Transformation
JFS2 enhanced JFS Format
KDC key distribution center XDR External Data Representation
KRB5 Kerberos version 5

Glossary
ACE. (Access Control Entry) One of the entries in GSS-API. Generic Security Services Application
an Access Control List (ACL). Programming Interface. A generic API for doing
client-server authentication.
ACL. (Access Control List) A list of permission
entries that control user and group access to an NFS mount. The operation that maps an exported
object. directory from an NFS server into a client’s directory
structure, making it appear as if the served directory
Authentication. Security method used to confirm is local on the client.
the identity of a system user, host, or service.
NFS server. A host that makes available for NFS
Authorization. Security method used to control access one or more of its directories.
what shared information each system user or client
machine can access. NFS. (Network File System) A protocol for sharing
files over a computer network.
GID. (Group Identifier) Number used to identify a
UNIX group. Opaque. Used in conjunction with bytes, data
structures, tokens, and so on, to represent a
Identification. Security method used to uniquely collection of bytes whose internal structure is
establish the identity of information system users, unknown at the time but will be interpreted later on
hosts, and services. in a processing sequence.
Integrated login. System login configured to Pseudo-file system. The portion of an NFS
obtain user authentication and identification server’s name space that is exported to NFS clients.
(optional) from an external source, such as
Kerberos/LDAP. Pseudo-root. The top level of an NFS server’s
pseudo-file system. By default, it corresponds to the
KDC. (Key Distribution Center) The trusted root directory in the server’s file tree, but it can be
third-party or intermediary in Kerberos that issues all specified to be a lower-level directory, such as
the Kerberos tickets to the clients. /exports.
Kerberos realm. Comprises a set of managed RPC. (Remote Procedure Call) A protocol whereby
hosts that share the same Kerberos database. one computer process (the client process) can direct
another process (the server process) to run a
NFS client. A host that accesses, via a mount, one procedure, appearing as if the client process had run
or more directories from an NFS server. the procedure in its own address space. The client
and server processes are typically on two separate
NFS domain. A name that identifies an operating computers, although they can both be on the same
context for NFS servers and clients. It is implied that computer.
systems that share the same NFS domain also
share the same user and group registries. RPCSEC_GSS. A security flavor that provides
authentication, integrity, and privacy protection for
NFS export. The operation that makes a directory remote procedure calls.
on an NFS server available for access by NFS
clients.

UID. (User Identifier) Number used to identify a
UNIX user.
WSM. (Web-based System Manager) A

Web-based interface for administering an AIX
system.

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 on ordering these publications, see “How to get IBM Redbooks”
on page 302. Note that some of the documents referenced here may be available
in softcopy only.
򐂰 AIX - Migrating NIS Maps into LDAP, TIPS0123
򐂰 AIX 4.3 Elements of Security, SG24-5962
򐂰 AIX 5L Version 5.2 Security Supplement, SG24-6066
򐂰 AIX 5L Differences Guide Version 5.3 Edition, SG24-7463
򐂰 Exploiting RS/6000 SP Security: Keeping It Safe, SG24-5521
򐂰 IBM IBM Eserver Certification Study Guide - AIX 5L Communications,
SG24-6186
򐂰 IBM IBM Eserver pSeries Sizing and Capacity Planning: A Practical Guide,
SG24-7071
򐂰 Introduction to the IBM Problem Determination Tools, SG24-6296
򐂰 RS/6000 SP System Management: Easy, Lean and Mean, GG24-2563
򐂰 Using LDAP for Directory Integration, SG24-6163
򐂰 Windows-based Single Signon and the EIM Framework on the IBM
IBM Eserver iSeries Server, SG24-6975
Other publications
These publications are also relevant as further information sources:
򐂰 AIX 5L Version 5.3 Commands Reference, Volume 1, SC23-4888
򐂰 AIX 5L Version 5.3 Files Reference, SC23-4895

򐂰 AIX 5L Version 5.3 Performance Management Guide, SC23-4905
򐂰 AIX 5L Version 5.3 Security Guide, SC23-4907
򐂰 AIX 5L Version 5.3 System Management Guide: Communications and
Networks, SC23-4909
򐂰 Garman, Kerberos: The Definitive Guide, O'Reilly, 2003, ISBN 0596004036
򐂰 Stern, et al., Managing NFS and NIS, 2nd Edition, O'Reilly, 2001, ISBN
1565925106
Online resources
These Web sites and URLs are also relevant as further information sources:
򐂰 NFS Version 4 Open Source Reference Implementation
http://www.citi.umich.edu/projects/nfsv4/linux/
򐂰 The NFS Version 4 Protocol, Brian Pawlowski, et al
http://www.nluug.nl/events/sane2000/papers/pawlowski.pdf
򐂰 NFS Version 3 Design and Implementation, Brian Pawlowski, et al
http://citeseer.ist.psu.edu/pawlowski94nfs.html
򐂰 IETF RFC page
http://www.ietf.org/rfc.html
򐂰 Other RFC references for NFS version 4
http://www.nfsv4.org/nfsv4techinfo.html
򐂰 NFS V4 Working Group
http://www.nfsv4.org/
򐂰 Open Systems Interconnection (OSI) Reference Model
http://ourworld.compuserve.com/homepages/timothydevans/osi.htm
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 hard-copy 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
Related publications 303

Index
Symbols A
aclconvert command 73
/etc/auto.master 177
acledit command 32, 73
/etc/bootparms 289
aclget command 32, 73, 82
/etc/exports 37
aclgettypes command 33, 73
/etc/exports file 88
aclput command 32, 73, 82
/etc/filesystems 288
ACLs
/etc/fstab (Linux) 177
AIXC ACLs
/etc/group file 49
AIXC ACL and NFS V4 client 64
/etc/gssapi_mech.conf (Linux) 177
commands 73
/etc/ibmslapd.conf 212
NFS V4 ACLs
/etc/idmapd.conf (Linux) 177
access evaluation 69
/etc/init.d/portmap (Linux) 177
administration 72
/etc/init.d/rpcgssd (Linux) 177
chmod command and 77
/etc/init.d/rpcidmapd (Linux) 177
directory structure and 79
/etc/inittab 137
file system support 65
/etc/krb5/krb5.conf 268
format 65
/etc/krb5/krb5.keytab 143
inheritance 68
/etc/netgroup 289
inheritance and move vs. copy 79
/etc/networks 289
inheritance and umask 78
/etc/nfs.clean 126
inheritance, maximizing benefits of 80
/etc/nfs/hostkey 38, 288
maintaining existing ACLs 82
/etc/nfs/local_domain 38, 288
NFS V3 clients and 87
/etc/nfs/princmap 288
permission bits 67
/etc/nfs/princmap file 39
permission restrictions 69
/etc/nfs/realm.map 39, 288
permissions scenarios 85
/etc/nfs/realm.map file 54
special user permissions 69
/etc/nfs/security_default 40, 288
UNIX permissions and 71
/etc/passwd file 49
types
/etc/pcnfsd.conf 289
AIXC ACL 33, 62–63
/etc/rc.nfs 288
NFS V4 ACL 32, 65
/etc/rmtab 288
add_principal command 139
/etc/rpc 289
addprinc command 139
/etc/security/user 173
administration
/etc/services 158
sample scripts 256
/etc/sm 288
aio 157
/etc/sm.bak 288
alias tree
/etc/state 288
configuration 131
/etc/syslog.conf 208
setting up alias tree extension in NFS V4 131
/etc/xtab 38
attributes
/usr/lib/security/methods.cfg 166
mandatory attributes 21
/var/krb5/log/krb5kdc.log 211
named attributes 23
/var/ldap/db2cli.log 212
recommended attributes 21
/var/ldap/ibmslapd.log 212
authentication 46

AUTH_SYS 47, 51, 59 chnfsim 32, 43, 292
example unsuccessful 281 chnfsmnt 44, 291
host 88 chnfsrtd 39, 43, 146
Kerberos 53, 59 chnfssec 40, 43
user 59 chnfssim 39
authentication grammar 166, 173 chsec 174
authorization 47 chuser 173
access control lists 62 config.krb5 137
exports 88 db2iauto 162
host 88 db2set 162
standard UNIX file permissions 63 errpt 211
user 62 exportfs 44, 88, 291
automount command 44, 291 options 293
find 132
fsuser 210
B get_principals 140
bootinfo command 156
getprincs 140
bosboot command 157
ibmdirctl 162
id 154
C installp 135
chkconfig (Linux) command 181 ipreport 210
chnfs command 43, 127 iptrace 210
chnfsdom command 38, 43, 292 kadmin.local 139
chnfsexp command 44, 291 kdestroy 149
chnfsim command 32, 43, 292 kinit 138
chnfsmnt command 44, 291 klist 139
chnfsrtd command 39, 43, 146 ktpass.exe (Windows) 195
chnfssec command 40, 43 ktutil 143
chnfssim command 39 ldapcfg 162
chsec command 174 ldapmodify 163
chuser command 173 ldapsearch 164
commands lsattr 157
/etc/nfs.clean 126 lsitab 137
/etc/rc.nfs 127 lsnfsexp 44, 291
aclconvert 73 lsnfsmnt 44, 291
acledit 32, 73 lsof 212
aclget 32, 73, 82 ls-secldapclntd 167
aclgettypes 33, 73 lssrc 128
aclput 32, 73, 82 lsuser 169
add_principal 139 mkgroup 166
addprinc 139 mkkrb5srv 136
automount 44, 291 mknfs 43, 291
bootinfo 156 mknfsexp 44, 291
bosboot 157 mknfsmnt 44, 292
chkconfig (Linux) 181 mkuser 166
chmod 77 mount 129, 291
chnfs 43, 127 options 294
chnfsdom 38, 43, 292 nfs4cl 43, 132, 211, 292
chnfsexp 44, 291 nfshostkey 38, 43, 144, 292

nfshostmap 39, 43 file sets
nfso 44, 291 bos.eim.rte 57
options 294–295 db2_08_01.ldap 158
nfsstat 44, 211 krb5.client 145
ps 166 ldap.client.adt 158
refresh 209 ldap.client.rte 158
rmnfs 44, 292 ldap.html.en_US.config 158
rmnfsexp 44, 292 ldap.html.en_US.man 158
rmnfsmnt 44 ldap.msg.en_US 158
rpcgen 44 ldap.server.cfg 158
rpcinfo 44, 210 ldap.server.com 158
setclock 105 ldap.server.java 158
showmount 123, 211, 291 ldap.server.rte 158
snap 263 ldap.webdadmin 158
startsrc 124 modcrypt.base 145
stopsrc 188 files
umount 123 /etc/bootparms 43, 289
uptime 153 /etc/environment 41
config.krb5 command 137 /etc/exports 43, 88, 288
/etc/filesystems 42, 288
/etc/fstab (Linux) 177
D /etc/group 49
db2iauto command 162
/etc/gssapi_mech.conf (Linux) 177
db2set command 162
/etc/ibmslapd.conf 212
diskless client 32
/etc/idmapd.conf (Linux) 177
domain name 124
/etc/inittab 137
/etc/krb5/krb5.conf 137, 268–269
E /etc/krb5/krb5.keytab 143
EIM /etc/krb5/krb5_cfg_type 137
bos.eim.rte file set 57 /etc/netgroup 289
Enterprise Identity Mapping /etc/networks 43, 289
See EIM /etc/nfs/hostkey 38, 42, 288
errpt command 211 /etc/nfs/local_domain 38, 42, 288
Ethereal 208, 210, 212–213 /etc/nfs/princmap 39, 42, 288
exportfs command 44, 88, 291 /etc/nfs/realm.map 39, 42, 54, 288
options 293 /etc/nfs/security_default 40, 42, 288
exports file 88 /etc/passwd 49
External Data Representation /etc/pcnfsd.conf 43, 289
See XDR /etc/rc.nfs 288
external name space (exname) 34 /etc/rmtab 288
/etc/rpc 43, 289
/etc/security/user 173
F
file handle 30 /etc/services 158
persistent 30 /etc/sm 288
volatile 30 /etc/sm.bak 288
file locking /etc/state 288
clientid 29 /etc/syslog.conf 208
stateid 29 /etc/xtab 38, 43, 288
/usr/lib/security/methods.cfg 166
Index 307
/var/krb5/log/krb5kdc.log 211 configuration file changes on KDC server 203
/var/ldap/db2cli.log 212 cross-realm access 199
find command 132 database 60
full client 147 description 59
fuser command 210 host identification 88
machine principal 88
mapping realm to NFS domain 54
G MIT Athena project 244
get_principals command 140
need for 103
GID 48
NFS client considerations 108
group file 49
NFS use of 60
GSS-API 28
other information sources 252
error codes 232
principal 106
gssd daemon 144
realms 105
service principal 88
I status codes 234
IBM Tivoli Directory Server 49, 155, 160 symmetric encryption 245
configuration 160 third-party authentication described 244
ibmdirctl command 162 ticket lifetime 60
id command 154 ticket-granting service 246
identification 46 tickets 59
host 87 user authentication 101
user/group 48 user principal 53
identity management Kerberos definitions
LDAP user registry 49 authentication database 251
NIS user registry 49 instance 251
UNIX user registry 49 KDC 252
identity mapping 50 key 251
AUTH_SYS 51 principal 251
Kerberos 53 realm 251
multiple NFS domains 57 ticket 251
installing the IBM NAS file sets 135 ticket-granting ticket 252
installp command 135 Kerberos overview 243
ipreport command 210 Kerberos V5 (KRB5) 28
iptrace command 210 KerberosV5 (KRB5)
sample output 271 Key Distribution Center (KDC) 28
Key Distribution Center
See KDC
K
kadmin.local command 139 keytab
KDC create NFS server entry 142
adding krbtgt service principal 200 krb5.keytab file 143
creating server principals 141 kinit command 138, 246
creating user principals 139 klist command 139
encryption types 107 krbtgt service principal 200
testing basic functions 135, 138 ktpass.exe (Windows) command 195
with LDAP backend 165 ktutil command 143
kdestroy command 149
Kerberos

L /usr/sbin/portmap 290
LDAP user registry 49 /usr/sbin/rpc.lockd 290
ldapcfg command 162 /usr/sbin/rpc.mountd 290
ldapmodify command 163 /usr/sbin/rpc.statd 290
ldapsearch command 164 biod 17, 42
LDIF gssd 41, 61
sample file 270 NFS Lock Manager (NLM) 19
Linux client nfsd 16
pseudo-file system in NFS V4 187 nfsrgyd 41
Read/Write NFS V4 mounts 185 portmap 15, 42
Read-only NFS V4 mount 182 rpc.lockd 16, 41
lsattr command 157 rpc.mountd 16, 42
lsitab command 137 rpc.pcnfsd 42
lsnfsexp command 44, 291 rpc.rstatd 42
lsnfsmnt command 44, 291 rpc.statd 16, 41
lsof command 212 NFS domain
choosing your NFS domain 98
configuration 146
M multiple NFS domains 57
mknfs command 43, 291
single NFS domain 51
mknfsexp command 44, 291
NFS domain name 124
mknfsmnt command 44, 292
NFS registry daemon 143
mount command 291
NFS V4
options 294
ACLs 65
multi-homed server 105
AIXC ACL and NFS V4 client 64
alias tree extension 131
N authentication methods 102
namespace 25 client with integrated login services 170
pseudo-file system 27 exports options 89
NAS full client 108
configuration 134 general deployment strategy 95
installing IBM NAS file sets 135 host authentication 88
path variable 121 host authorization 88
set up with legacy database 134 host identification 87
Network File System identity mapping 24
See NFS implementation 119
Network Installation Management integrating with Linux client 176
See NIM Linux client 170
Network Time Protocol mandatory attributes 21
See NTP named attributes 23
NFS 4 nobody, mapping unknown users to 57
definition 3 pseudo root FS 124
NFS V3 (NFS Version 3) 17 recommended attributes 21
NFS daemons 14 security context 60
/usr/bin/gssd 290 setting up client with NAS 145
/usr/sbin/biod 290 slim client 108
/usr/sbin/nfsd 290 slim client versus full client 108
/usr/sbin/nfsrgyd 290 syslogd settings 121
/usr/sbin/pcnfsd 290 unknown user 57
Index 309
user authentication 59 LIPKEY 47
user authorization 62 protection levels
user/group identification 50 authentication 47
with Windows KDC 190 integrity 47
NFS V4 (NFS version 4) 20 privacy 47
NFS_NOBODY 41 security context 60
NFS_PORT_RANGE 41
nfs4cl command 43, 211, 292
nfshostkey command 38, 43, 292
S
scripts
nfshostmap command 39, 43
change pseudo-root FS 256
nfso command 44, 291
copy ACLs 260
options 294–295
create full client (LDAP) 259
nfsstat command 44, 211
create full client (legacy db) 258
NIM 32
create KDC server 256
NIS user registry 49
nfs_pd script 263
NTP
sec= exports option 89
security
P AUTH_SYS 51
passwd file 49 flavors 107
princmap file 39 Kerberos 28, 53
pseudo-file system 27 krb5 107
advantages 128 krb5i 107
configuration 125 krb5p 107
pseudo root FS 124 LIPKEY 28
RPC flavors 47
security categories 46
R security components
realm.map file 54
authentication 46
Redbooks Web site 302
authorization 47
contact us xv
identification 46
refresh command 209
service ticket 59
Remote Procedure Call
showmount command 211, 291
See RPC)
slim client 150, 152
RFC
slim client for cloning 150
RFC2078 7
slim client installation steps 150
RFC2307 7, 170
slim client verification 153
RFC3010 5
snap command 263
RFC3530 5
stateful 13
rmnfs command 44, 292
stateless 13
rmnfsexp command 44, 292
Symbols
RPC
/etc/exports 288
security flavors 47
/etc/krb5/krb5.conf 269
rpcgen command 44
/etc/syslog.conf 122
rpcinfo command 44, 210
/etc/xtab 288
RPCSEC_GSS 47
/var/krb5/log/kadmin.log 211
authentication flow 61
NFS_NOBODY 41
configuration 154
NFS_PORT_RANGE 41
configuring on clients 154
syslogd 121
Kerberos 47

T
ticket-granting ticket 59
time synchronization 104
NTP 104
setclock 105
timed 104
troubleshooting 207
EIM not configured 213
exporting file systems 215
GSS-API error codes 232
IBM Tivoli Directory Server 212
Kerberos status codes 234
mount problems 218
NAS 211
realm is already mapped to domain 214
tools 208
using errpt command 211
using Ethereal 212
using fuser command 210
using iptrace and ipreport 210
using lsof command 212
using nfs4cl command 211
using nfsstat command 211
using rpcinfo command 210
using showmount command 210
using syslogd 208
U
UID 48
UNIX user registry 49
unmounting exported NFS V4 file system 123
user registries
NIS 49
UTF-8 30–31
V
vers=exports option 89
W
Web-based Systems Manager 73
ACL administration 73
Windows KDC 190
X
XDR
Index 311
Securing NFS in AIX An Introduction to NFS V4 in AIX 5L
(0.5” spine)
0.475”<->0.875”
250 <-> 459 pages
Back cover ®
Securing NFS in AIX

An Introduction to NFS V4 in AIX 5L
Version 5.3
NFS Version 4 (NFS V4) is the latest defined client-to-server

protocol for NFS. A significant upgrade from NFS V3, it was INTERNATIONAL
defined under the IETF framework by many contributors. NFS TECHNICAL
V4 introduces a major changes to the way NFS has been SUPPORT
implemented and used up until now, including stronger ORGANIZATION
security, wide area network sharing, and broader platform
adaptability.
This IBM Redbook is intended to provide a broad BUILDING TECHNICAL

understanding of NFS V4 and specific AIX NFS V4 INFORMATION BASED ON
PRACTICAL EXPERIENCE
implementation details. It discusses considerations for
deployment of NFS V4, with a focus on exploiting the stronger
security features of the new protocol. IBM Redbooks are developed by
the IBM International Technical
In the initial implementation of NFS V4 in AIX 5.3, the most Support Organization. Experts
important functional differences are related to security. from IBM, Customers and
Partners from around the world
Chapter 3 and parts of the planning and implementation create timely technical
chapters in Part 2 cover this topic in detail. 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
SG24-7204-00 ISBN 0738490806

IBM - NFS in AIX

Uploaded by

Copyright:

Available Formats

IBM - NFS in AIX

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

IBM - NFS in AIX

Uploaded by

Copyright:

Available Formats

Front cover

Securing NFS in AIX

Set up and administer NFS V4 on your

Command use changes and

Learn to use NFS V4 in a

Securing NFS in AIX

First Edition (November 2004)

This edition applies to NFS Volume 4 running on AIX 5L.

© Copyright International Business Machines Corporation 2004. All rights reserved.

Part 1. NFS V4 fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Chapter 1. NFS Version 4 overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Chapter 2. What’s new in NFS V4? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

© Copyright IBM Corp. 2004. All rights reserved. iii

Chapter 3. Enhanced security in NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . 45

iv Securing NFS in AIX

Part 2. Implementing NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

Chapter 4. Planning for NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

Chapter 5. Sample implementation scenarios . . . . . . . . . . . . . . . . . . . . . 119

Chapter 6. Problem determination. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

vi Securing NFS in AIX

Part 3. Appendixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241

Appendix A. Kerberos. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

Appendix B. Sample scripts, files, and output . . . . . . . . . . . . . . . . . . . . . 255

Appendix C. AIX 5.3 NFS quick reference. . . . . . . . . . . . . . . . . . . . . . . . . 287

Abbreviations and acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297

Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301

viii Securing NFS in AIX

© Copyright IBM Corp. 2004. All rights reserved. ix

AFS® IBM® pSeries®

The following terms are trademarks of other companies:

x Securing NFS in AIX

This IBM® Redbook is intended to provide a broad understanding of NFS V4 and

Part 1 NFS V4 Fundamentals

© Copyright IBM Corp. 2004. All rights reserved. xi

Part 2 Implementing NFS V4

xii Securing NFS in AIX

The team that wrote this redbook

Chris Almond is an ITSO Project Leader and IT Architect based in Austin,

Sridhar Murthy works for IBM as an IT Specialist in Dallas, Texas. He has 15

xiv Securing NFS in AIX

© Copyright IBM Corp. 2004. All rights reserved. 1

Chapter 1. NFS Version 4 overview

The following topics are discussed in this chapter:

© Copyright IBM Corp. 2004. All rights reserved. 3

1.2 NFS V2 and NFS V3 History

4 Securing NFS in AIX

1.3 NFS V4 design motivations

1.4 Objectives of NFS V4 (RFC3530)

Chapter 1. NFS Version 4 overview 5

1.5 AIX 5.3 specific implementation of NFS V4

Vendor-specific implementation of RFC3530 into their product depends on

The following features of the protocol are supported in AIX 5.3.

1.5.1 Mandatory features

6 Securing NFS in AIX

1.5.2 Optional features

1.6 Planning and implementation considerations

1.6.1 Pre-implementation design considerations

Chapter 1. NFS Version 4 overview 7

1.7 Looking ahead to the rest of the book