AIX 5L Differences Guide Version 5.3 Edition

Front cover

AIX 5L Differences Guide

Version 5.3 Edition
AIX - The industrial strength UNIX
operating system
An expert’s guide to the new
release
AIX 5L Version 5.3
enhancements explained

Adrian Demeter
Shiv Dutta
Armin Röll
SeongLul Son

ibm.com/redbooks

International Technical Support Organization
December 2004

SG24-7463-00

Note: Before using this information and the product it supports, read the information in
“Notices” on page xvii.

First Edition (December 2004)
This edition applies to AIX 5L for POWER Version 5.3, program number 5765-G03.
© 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
Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Notices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
Trademarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviii
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
The team that wrote this redbook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
Become a published author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xx
Comments welcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxi
Chapter 1. Virtualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Advanced POWER Virtualization feature . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Introduction to the POWER Hypervisor. . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 POWER Hypervisor implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.2 POWER Hypervisor processor dispatch . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.3 POWER Hypervisor and virtual I/O . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3 Micro-Partitioning introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3.1 Shared processor partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.2 Shared pool overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.3 Dynamic partitioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3.4 Limitations and considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.4 Virtual Ethernet introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.4.1 Virtual LAN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.4.2 Virtual Ethernet connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.4.3 Benefits of Virtual Ethernet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.4.4 Dynamic partitioning for Virtual Ethernet devices . . . . . . . . . . . . . . . 38
1.5 Shared Ethernet Adapter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.5.1 Connecting a Virtual Ethernet to external networks . . . . . . . . . . . . . 40
1.5.2 Using link aggregation (EtherChannel) to external networks . . . . . . 43
1.6 Virtual SCSI introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.6.1 Partition access to virtual SCSI devices . . . . . . . . . . . . . . . . . . . . . . 47
1.7 Partition Load Manager introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

© Copyright IBM Corp. 2004. All rights reserved.

iii

1.7.1 Memory management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
1.7.2 Processor management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Chapter 2. Application development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.1 POSIX Realtime functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.1.1 Memlock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.1.2 Spin locks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.1.3 Clocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.1.4 Priority scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.1.5 Shared memory objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.1.6 Semaphores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.1.7 Timers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.1.8 Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.1.9 Thread options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.1.10 Message passing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.1.11 Advisory Info . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.2 Enhanced libc.a. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.3 New malloc() algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.4 Improvements to malloc subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.5 Block device mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.5.1 System calls and subroutines for block device mapping. . . . . . . . . . 75
2.6 Eclipse Runtime Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.7 Cryptographic sum command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.8 Perl 5.8.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.9 SVR4 linking affinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.9.1 dlsym() function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.9.2 New options of the ld command for SVR4 affinity . . . . . . . . . . . . . . . 79
2.10 Marking executable’s read/write sections . . . . . . . . . . . . . . . . . . . . . . . . 81
2.10.1 Updated flags for the ld command . . . . . . . . . . . . . . . . . . . . . . . . . 81
2.10.2 Updated flags for the dump command . . . . . . . . . . . . . . . . . . . . . . 81
2.10.3 Updated flags for the ldedit command . . . . . . . . . . . . . . . . . . . . . . 82
2.11 Thread-based credentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.12 Scalability enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.12.1 Kernel round robin locks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.12.2 Increased resource limit values in /etc/security/limits file . . . . . . . . 84
2.12.3 CPU time limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.12.4 New heap size and late staggering . . . . . . . . . . . . . . . . . . . . . . . . . 86
2.13 Increased inter-process communication limits . . . . . . . . . . . . . . . . . . . . 86
2.14 Thread support in gmon.out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.15 Enhanced DBX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.16 The tcpdump command. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.17 gprof thread support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

iv


2.18 Java 1.4.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.19 Java 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.20 Improved man command presentation . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Chapter 3. Storage management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.1 LVM enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.1.1 Performance improvement of LVM commands . . . . . . . . . . . . . . . . . 96
3.1.2 Removal of classical concurrent mode support . . . . . . . . . . . . . . . . 97
3.1.3 Scalable volume groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.1.4 Striped column support for logical volumes . . . . . . . . . . . . . . . . . . 106
3.1.5 Volume group pbuf pools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.1.6 Variable logical track group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.2 JFS2 enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.2.1 Disk quotas support for JFS2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.2.2 JFS2 file system shrink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.2.3 JFS2 logredo scalability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.2.4 JFS2 file system check scalability . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.2.5 JFS2 extended attributes Version 2 support . . . . . . . . . . . . . . . . . . 129
3.2.6 JFS2 ACL support for NFS V4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.2.7 ACL inheritance support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Chapter 4. Reliability, availability, and serviceability . . . . . . . . . . . . . . . . 141
4.1 Error log RAS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.2 Enhancements for a large number of devices . . . . . . . . . . . . . . . . . . . . . 142
4.3 Core file creation and compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.4 System dump enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.4.1 Dump information to TTY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.4.2 Verbose flag for the sysdumpdev command . . . . . . . . . . . . . . . . . . . 147
4.4.3 dmpfmt command changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.5 DVD support for system dumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.5.1 Copy system dumps to DVD media . . . . . . . . . . . . . . . . . . . . . . . . 150
4.6 snap command enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.7 Administrative control of the user trace buffers . . . . . . . . . . . . . . . . . . . . 154
4.8 Single thread trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Chapter 5. System management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.1 InfoCenter for AIX 5L Version 5.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
5.2 Multiple desktop selection from BOS menus. . . . . . . . . . . . . . . . . . . . . . 161
5.3 Erasing hard drive during BOS install . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.4 Service Update Management Assistant . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.4.1 Packaging and installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.4.2 Functional description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.4.3 Concepts and implementation specifics . . . . . . . . . . . . . . . . . . . . . 168
5.4.4 Command line interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

Contents

v

5.4.5 SMIT user interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.5 Long user and group name support . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.6 Dynamic reconfiguration usability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.7 Paging space garbage collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.7.1 Garbage collection on re-pagein . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.7.2 GC paging space scrubbing for in-memory frames. . . . . . . . . . . . . 184
5.7.3 VMM tuning parameters for PSGC . . . . . . . . . . . . . . . . . . . . . . . . . 185
5.8 Dynamic support for large page pools. . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.9 Interim Fix Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
5.10 List installed filesets by bundle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
5.11 Configuration file modification surveillance . . . . . . . . . . . . . . . . . . . . . . 192
5.12 DVD backup using the mkdvd command . . . . . . . . . . . . . . . . . . . . . . . 195
5.13 NIM security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
5.13.1 NIM service handler for client communication - NIMSH . . . . . . . . 196
5.13.2 NIM cryptographic authentication - OpenSSL . . . . . . . . . . . . . . . . 203
5.14 High Available NIM (HA NIM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
5.15 General NIM enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
5.15.1 Detailed output when creating a NIM lpp_source resource . . . . . 214
5.15.2 Creating a SPOT resource from a mksysb . . . . . . . . . . . . . . . . . . 214
5.15.3 Restore SPOT copy function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
5.15.4 Adjustments in NIM to process multiple CD media . . . . . . . . . . . . 215
5.15.5 NIM interface to change network attributes. . . . . . . . . . . . . . . . . . 216
5.15.6 VIPA and EtherChannel support in NIM adapters. . . . . . . . . . . . . 218
5.15.7 EZ NIM enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5.16 Alternate Disk Installation migration . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5.17 Enhancements to Alternate Disk Installation . . . . . . . . . . . . . . . . . . . . . 219
5.18 Advanced Accounting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
5.18.1 Data files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
5.18.2 Projects and policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
5.18.3 Transactional accounting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
5.18.4 Interval accounting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
5.18.5 Data aggregation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
5.18.6 Report and analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
5.18.7 Data file management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
5.18.8 Accounting records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
5.19 Date APIs past 2038 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
5.20 System V printer enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
5.21 Mozilla browser for AIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
5.21.1 Installing Mozilla for AIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
5.21.2 Mozilla as the browser for AIX Documentation Services . . . . . . . 249
5.21.3 Migrating Netscape Communicator Version 4 profile . . . . . . . . . . 250
5.22 Uniprocessor kernel support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
5.23 Enhancement of base commands and libraries . . . . . . . . . . . . . . . . . . 250

vi


Chapter 6. Performance monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
6.1 General information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
6.2 The procmon command - process monitoring tool . . . . . . . . . . . . . . . . . 254
6.3 Asynchronous I/O statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
6.3.1 Asynchronous I/O basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
6.3.2 AIO enhancement of the iostat command in Version 5.3 . . . . . . . . 258
6.4 Disk service and wait time monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6.4.1 dkstat and device support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6.4.2 avwait and avserv enhancements for the sar -d . . . . . . . . . . . . . . . 261
6.5 PMAPI M:N pthreads support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
6.6 Support for Micro-Partitioning and SMT . . . . . . . . . . . . . . . . . . . . . . . . . 263
6.6.1 perfstat library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
6.6.2 vmstat command enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . 263
6.6.3 iostat command enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
6.6.4 mpstat command - CPU performance monitoring . . . . . . . . . . . . . . 265
6.6.5 lparstat command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
6.6.6 sar command enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
6.6.7 topas command enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
6.7 Per file system I/O pacing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Chapter 7. Networking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
7.1 NFS Version 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
7.2 Multipath routing enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
7.3 PMTU enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
7.4 DHCPv6 support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
7.5 IPv6 functional updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
7.6 Interface layer support for hotplug network adapters . . . . . . . . . . . . . . . 304
7.7 Support for SLP client service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
7.8 Stream Control Transmission Protocol (SCTP). . . . . . . . . . . . . . . . . . . . 308
7.8.1 SCTP basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
7.8.2 SCTP API in AIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
7.8.3 SCTP management in AIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
7.9 Network RAS enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
7.10 Enable IP Security with intervening NAT devices . . . . . . . . . . . . . . . . . 313
7.11 AIX SNMP subagent enhancements. . . . . . . . . . . . . . . . . . . . . . . . . . . 316
7.11.1 MIB objects added in Version 5.3 . . . . . . . . . . . . . . . . . . . . . . . . . 316
7.11.2 The snmptrap command enhancement . . . . . . . . . . . . . . . . . . . . 318
7.12 SMBFS enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
7.13 Ported 64-bit DES NFS kernel extension . . . . . . . . . . . . . . . . . . . . . . . 319
7.14 BIND domain search improvement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
7.15 Network API support for ARP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
7.16 Withdraw older versions of TCP applications . . . . . . . . . . . . . . . . . . . . 323

Contents

vii

Chapter 8. Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
8.1 LDAP user authentication enhancements . . . . . . . . . . . . . . . . . . . . . . . . 326
8.2 Full PAM exploitation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
8.2.1 AIX authentication through PAM library . . . . . . . . . . . . . . . . . . . . . 330
8.2.2 Additional PAM modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
8.2.3 PAM application programming interface changes. . . . . . . . . . . . . . 339
8.3 Prioritizing LDAP servers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
8.4 Rsh login control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
8.5 chpasswd command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
8.6 Initial login license limit increased . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
8.7 NIS/NIS+ enhancements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
8.8 Support for passwd.adjunct NIS map . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
Chapter 9. Clustering technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
9.1 RSCT support for Micro-Partitioning technology . . . . . . . . . . . . . . . . . . . 344
9.1.1 Updated Host resource class - IBM.Host . . . . . . . . . . . . . . . . . . . . 344
9.1.2 Updated Processor resource class - IBM.Processors. . . . . . . . . . . 344
9.2 Least privilege resource manager . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
9.2.1 LPRM commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
9.2.2 LPRM security. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
9.3 Cluster Systems Management Version 1.4 for AIX . . . . . . . . . . . . . . . . . 349
9.3.1 CSM 1.4 new features. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
9.3.2 CSM migrations to Version 1.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
Chapter 10. National language support . . . . . . . . . . . . . . . . . . . . . . . . . . 353
10.1 Gujarati NLS enablement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
10.2 Tamil NLS enablement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
10.3 Kazakh NLS enablement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
10.4 Telegu NLS enablement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
10.5 Unicode extension. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
10.6 Update AGFA TrueType font rasterizer. . . . . . . . . . . . . . . . . . . . . . . . . 355
10.7 Unicode 4.0 support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
Appendix A. Web-based System Manager enhancements . . . . . . . . . . . 357
A.1 Java Web Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
A.2 Voicing support on Windows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
A.3 Keyboard accessibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
A.4 Inherit desktop theme on Windows remote client . . . . . . . . . . . . . . . . . . 360
Abbreviations and acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
Other publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371

viii


Online resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
How to get IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
Help from IBM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375

Contents

ix

x


Figures
1-1
1-2
1-3
1-4
1-5
1-6
1-7
1-8
1-9
1-10
1-11
1-12
1-13
1-14
1-15
1-16
1-17
1-18
1-19
1-20
1-21
1-22
1-23
1-24
1-25
1-26
1-27
3-1
3-2
3-3
3-4
3-5
3-6
3-7
3-8
3-9
3-10
3-11

Advanced POWER Virtualization feature. . . . . . . . . . . . . . . . . . . . . . . . . 2
HMC panel to enable the Virtualization Engine Technologies . . . . . . . . . 3
POWER Hypervisor functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Micro-Partitioning processor dispatch . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Processing units of capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Distribution of capacity entitlement on virtual processors . . . . . . . . . . . 20
Capped shared processor partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Uncapped shared processor partition . . . . . . . . . . . . . . . . . . . . . . . . . . 22
HMC panel Dynamic Logical Partitioning. . . . . . . . . . . . . . . . . . . . . . . . 24
Add Processor Resource panel on the HMC . . . . . . . . . . . . . . . . . . . . . 25
Changing the partition mode from Uncapped to Capped. . . . . . . . . . . . 26
Remove Processor Resource panel on the HMC . . . . . . . . . . . . . . . . . 28
Move Processor Resources panel on HMC . . . . . . . . . . . . . . . . . . . . . . 29
Example of a VLAN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
VLAN configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Logical view of an inter-partition VLAN . . . . . . . . . . . . . . . . . . . . . . . . . 37
Connection to external network using AIX routing . . . . . . . . . . . . . . . . . 40
Shared Ethernet Adapter configuration . . . . . . . . . . . . . . . . . . . . . . . . . 41
Multiple Shared Ethernet Adapter configuration . . . . . . . . . . . . . . . . . . 43
Link Aggregation (EtherChannel) pseudo device . . . . . . . . . . . . . . . . . 45
Virtual SCSI architecture overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Logical Remote Direct Memory Access . . . . . . . . . . . . . . . . . . . . . . . . . 49
Virtual SCSI device relationship on Virtual I/O Server . . . . . . . . . . . . . . 50
Virtual SCSI device relationship on AIX client partition . . . . . . . . . . . . . 50
Comparison of features of PLM and POWER Hypervisor . . . . . . . . . . . 52
Partition Load Manager overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
PLM resource distribution for partitions . . . . . . . . . . . . . . . . . . . . . . . . . 55
New SMIT menu: Add a Scalable Volume Group . . . . . . . . . . . . . . . . 105
New Web-based System Manager panel: New Volume Group . . . . . . 106
Logical volume with a stripe width of 3 and 2 striped columns . . . . . . 107
New SMIT menu: Manage Quotas for an Enhanced JFS . . . . . . . . . . 120
Change/Show Characteristics of an Enhanced Journaled File System 124
Shrink file system in command line . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
File system unchanged: Attempt to reduce the size . . . . . . . . . . . . . . 127
SMIT menu for converting EAv1 to EAv2 . . . . . . . . . . . . . . . . . . . . . . 132
Configuring ACL through Web-based System Manager . . . . . . . . . . . 135
File Access Control List dialog panel . . . . . . . . . . . . . . . . . . . . . . . . . . 135
File ACL dialog: Second panel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136


xi

3-12
3-13
4-1
4-2
4-3
5-1
5-2
5-3
5-4
5-5
5-6
5-7
5-8
5-9
5-10
5-11
5-12
5-13
5-14
5-15
5-16
6-1
6-2
6-3
6-4
6-5
6-6
6-7
6-8
6-9
7-1
7-2
7-3
7-4
7-5
A-1
A-2

xii

ACL edition dialog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Add ACE dialog: Access Mask tab . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Copy system dump to media boot time dialog . . . . . . . . . . . . . . . . . . . 150
smit cngtrace SMIT menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
smit trcstart SMIT menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
AIX Information Center on the documentation CD. . . . . . . . . . . . . . . . 160
Service Update Management Assistant: Control flow diagram . . . . . . 169
New SMIT menu: Service Update Management Assistant . . . . . . . . . 172
SMIT panel: Create a New SUMA Task . . . . . . . . . . . . . . . . . . . . . . . 174
Change/Show Characteristics of Operating System . . . . . . . . . . . . . . 176
smit nim_config_services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
smit nim_chmac . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
smit nim_ssl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
smit nim_config_services for SSL . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
High availability NIM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
smit nim_altmstr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
smit nim_mkaltmstr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
smit niminit_altmstr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
smit nim_mac_op_sync_hdr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
smit nim_mac_op_takeover_hdr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
smit nim_ch_if1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
procmon - new process monitoring tool . . . . . . . . . . . . . . . . . . . . . . . . 255
Global metrics for partition performance . . . . . . . . . . . . . . . . . . . . . . . 256
AIO flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
CPU statistics for POWER5 systems. . . . . . . . . . . . . . . . . . . . . . . . . . 266
Affinity domains and process dispatches . . . . . . . . . . . . . . . . . . . . . . . 272
mpstat -d command output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
topas default sample output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
topas logical partition sample output . . . . . . . . . . . . . . . . . . . . . . . . . . 281
smit crjfs2lvstd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
An example configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Add Static Route SMIT panel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Output of the pmtu display command . . . . . . . . . . . . . . . . . . . . . . . . . 299
The IP network protocol layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
NAT-enabled IP security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
Improved keyboard navigation in Web-based System Manager . . . . . 359
Native theme support in Web-based System Manager . . . . . . . . . . . . 360


Tables
1-1
1-2
1-3
1-4
1-5
2-1
3-1
3-2
3-3
3-4
3-5
3-6
3-7
3-8
5-1
5-2
5-3
5-4
5-5
5-6
5-7
5-8
5-9
5-10
5-11
5-12
5-13
6-1
7-1
7-2
7-3
7-4
7-5
7-6
7-7
7-8
7-9
7-10

Micro-Partitioning overview on p5 systems . . . . . . . . . . . . . . . . . . . . . . 15
Reasonable settings for shared processor partitions . . . . . . . . . . . . . . . 31
Interpartition VLAN communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
VLAN communication to external network . . . . . . . . . . . . . . . . . . . . . . . 36
Main differences between EC and LA aggregation . . . . . . . . . . . . . . . . 44
Resource limits in /etc/security/limits file . . . . . . . . . . . . . . . . . . . . . . . . 85
Configuration limits for volume groups. . . . . . . . . . . . . . . . . . . . . . . . . . 98
File system centric SMIT panels for JFS2 disk quota support . . . . . . . 119
Things to know about file system shrink . . . . . . . . . . . . . . . . . . . . . . . 128
Difference between EAv1 and EAv2 . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Compatibility of extended attributes . . . . . . . . . . . . . . . . . . . . . . . . . . 131
ACE type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
ACE access mask . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Keys for inheritance of NFS4 ACL . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
New and changed bosinst.data entries . . . . . . . . . . . . . . . . . . . . . . . . 164
SUMA SMIT menus and panels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
alt_disk_install command arguments . . . . . . . . . . . . . . . . . . . . . . . . . . 220
Data file management commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
User, group, and application rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Relative project classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Example of the No Accounting specification in a project list . . . . . . . . 228
Structure of a user policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Projects and policies commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Interval accounting commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
System-level data aggregation commands . . . . . . . . . . . . . . . . . . . . . 236
Project-level data aggregation commands. . . . . . . . . . . . . . . . . . . . . . 237
E-mail notification messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Sample CPU usage calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
Routing method for multipath routing . . . . . . . . . . . . . . . . . . . . . . . . . . 295
PMTU enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
Keywords in /etc/dhcpv6/dhcpsdv6.cnf . . . . . . . . . . . . . . . . . . . . . . . . 300
Keywords in /etc/dhcpv6/dhcpc6.cnf . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Key words in /etc/dhcprd.cnfi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Differences of network option ifsize . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Parameters for /etc/slp.conf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
TCP, UDP, and SCTP comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
New network options added in Version 5.3 . . . . . . . . . . . . . . . . . . . . . 313
New MIB resources implemented in Version 5.3 . . . . . . . . . . . . . . . . . 317


xiii

7-11
7-12
7-13
8-1

xiv

MIB resources implemented in version 5.2 . . . . . . . . . . . . . . . . . . . . . 318
Parameters for arpresolve_common . . . . . . . . . . . . . . . . . . . . . . . . . . 321
Parameters for arpupdate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
Difference between UNIX-type and LDAP-type authentication . . . . . . 326


Examples
2-1
3-1
3-2
3-3
3-4
3-5
3-6
3-7
4-1
4-2
4-3
4-4
4-5
5-1
5-2
5-3
5-4
5-5
5-6
6-1
6-2
6-3
6-4
6-5
6-6
6-7
7-1
7-2
7-3
7-4
7-5
7-6
7-7
7-8
7-9
7-10
7-11
7-12

Example of dump -ov command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
The value of LTG size is shown as MAX REQUEST . . . . . . . . . . . . . . 113
The output shows the value of LTG in use. . . . . . . . . . . . . . . . . . . . . . 113
Setting and reading extended attributes . . . . . . . . . . . . . . . . . . . . . . . 130
Configuring extended attributes with a file name . . . . . . . . . . . . . . . . . 130
Output of the acledit command before converting to NFS4 . . . . . . . . 132
Output of the acledit command after converting to NFS4 . . . . . . . . . . 133
Configuring ACL with inheritance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Managing core file creation settings . . . . . . . . . . . . . . . . . . . . . . . . . . 144
TTY output from system dump. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
sysdumpdev -vL output in case of I/O error during dump . . . . . . . . . . 147
sysdumpdev -vL output in case of power off during dump. . . . . . . . . . 148
snapsplit command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
New erasure menu. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
New disk selection menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Verify and change user name limit through the command line . . . . . . 177
Create a long user name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Cannot log in with longer user name because of the new lower limit . 177
lgpg command example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
iostat command AIO statistics in a Micro-Partitioning environment . . . 259
Output of the iostat -Aq command with AIO queue statistics . . . . . . . . 260
The iostat -AQ command output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
sar -d command output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
mpstat basic statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
mpstat -i output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
mpstat -s command output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
The nfsrgyd and gssd daemons. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
Setting up KDC server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
Add principal on KDC server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
Setting up the Kerberos client . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
Creating keytab file and setting up the gssd daemon for NFS Server . 293
Change nfs domain and set up map file . . . . . . . . . . . . . . . . . . . . . . . 293
Mount NFS4 file system with Kerberos on the NFS client . . . . . . . . . . 294
Mount failure without Kerberos configuration. . . . . . . . . . . . . . . . . . . . 294
Setting policy and weight for multipath routing . . . . . . . . . . . . . . . . . . 296
dhcpv6 server configuration file /etc/dhcpv6/dhcpsdv6.cnf . . . . . . . . . 299
dhcpv6 client configuration file: /etc/dhcpv6/dhcpc6.cnf . . . . . . . . . . . 302
dhcpv6 relay agent configuration file:/etc/dhcprd.cnf . . . . . . . . . . . . . . 302


xv

7-13
7-14
7-15
7-16
7-17
7-18
8-1
8-2
8-3

xvi

Network option ifsize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
sctpctrl command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
The output of nsdbg subcommand . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
/etc/isakmpd.conf used or NAT support . . . . . . . . . . . . . . . . . . . . . . . . 314
Filter rules added with ENABLE_IPSEC_NAT_TRAVERSAL . . . . . . . 315
Configuring SMB clients. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
New controls added in /etc/security/ldap/ldap.cfg . . . . . . . . . . . . . . . . 327
Converting AIX schema to RFC2307 . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Authentication through the passwd.adjunct NIS map file. . . . . . . . . . . 341


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.


xvii

Trademarks
The following terms are trademarks of the International Business Machines Corporation in the United States,
other countries, or both:
Eserver®
Eserver®
Redbooks (logo)
eServer™
ibm.com®
AFS®
AIX 5L™
AIX®
DB2®
developerWorks®
DFS™

™

Enterprise Storage Server®
Hypervisor™
HACMP™
i5/OS™
IBM®
Micro Channel®
Micro-Partitioning™
PowerPC®
POWER™
POWER2™
POWER4™

POWER5™
PTX®
pSeries®
QMF™
Redbooks™
RDN™
RS/6000®
SecureWay®
Versatile Storage Server™
ViaVoice®
Virtualization Engine™

The following terms are trademarks of other companies:
Microsoft, Windows, Windows NT, and the Windows logo are trademarks of Microsoft Corporation in the
United States, other countries, or both.
Java and all Java-based trademarks and logos are trademarks or registered trademarks of Sun
Microsystems, Inc. in the United States, other countries, or both.
UNIX is a registered trademark of The Open Group in the United States and other countries.
Other company, product, and service names may be trademarks or service marks of others.

xviii


Preface
This IBM® Redbook focuses on the differences introduced in AIX® 5L™ Version
5.3 when compared to AIX 5L Version 5.2. It is intended to help system
administrators, developers, and users understand these enhancements and
evaluate potential benefits in their own environments.
AIX 5L Version 5.3 introduces many new features, including NFS Version 4 and
Advanced Accounting, and exploits the advanced capabilities of POWER5™
equipped severs, such as Virtual SCSI, Virtual Ethernet, SMT,
Micro-Partitioning™ technology, and others. There are many other
enhancements available with AIX 5L Version 5.3, and you can explore them in
this redbook.
For customers who are not familiar with the enhancements of AIX 5L through
Version 5.2, a companion publication, AIX 5L Differences Guide Version 5.2
Edition, SG24-5765 is available.

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.
Adrian Demeter is a senior IT consultant and technical manager at IBM
Business Partner, GC System a.s. in Prague, Czech Republic. He holds the Ing.
academic degree in Electrical engineering and technical cybernetics from Czech
Technical University (CVUT), Prague. He has 20 years of experience in
computers and electronics, of which 12 years are in UNIX® and AIX. Adrian is
responsible for project design, implementation, and support of high-end
computer systems in Europe. He also teaches advanced IBM courses on AIX,
HACMP™, Cluster 1600, and others. He has written extensively on hardware
architectures, UNIX, communications, security, high availability, clustering, and
parallel computing.
Shiv Dutta works for IBM US in Austin, providing technical assistance to ISVs in
the areas of application benchmarking, porting, performance tuning and sizing
guides for pSeries®. He also provides the pSeries and AIX roadmaps for the ISV
community. Shiv has had extensive experience in application development,
database design, and system administration for RS/6000®, SP, and AIX. He has
been on the faculty of several colleges, universities, and training institutions
where he taught various C and UNIX courses. Shiv has authored a number of


xix

articles on IT topics for the IBM developerWorks® Web site. He has a Ph.D. in
Physics from Ohio University and Project Management Professional (PMP)
Certification from the Project Management Institute. He has a few published
patent disclosures to his credit.
Armin Röll works as a pSeries system engineer in Germany. He has nine years
of experience in pSeries and AIX pre-sales technical support and, as a
teamleader, he fosters the AIX skills community. He holds a degree in
experimental physics from the University of Hamburg, Germany. He co-authored
the AIX Version 4.3.3 and the AIX 5L Version 5.0 Differences Guide
Redbooks™.
SeongLul Son is an advisory education specialist working for IBM Korea. He
has eight years of experience in networking, e-learning, and operating systems
in various environments. His areas of expertise include AIX, TCP/IP networking,
and HACMP. He is an IBM certified Advanced Technical Expert - pSeries and AIX
5L and Cisco Certified Network Professionl (CCNP). He has written extensively
on system management, file system management, communcations, and security.
The team that authored this publication was managed by:
Scott Vetter
IBM Austin
Thanks to the following people for their contributions to this project:
Mathew Accapadi, David K Altobelli, Dwip Banerjee, Kavitha Baratakke, Anne Bohy, Joyce
Bolton, Paul S. Bostrom, Matthew T. Brandyberry, Larry Brenner, Luke Browning, Guoyou
Chen, Diane Chung, Mathew Cronk, Zhi-wei Dai, Medha Date, Christopher DeRobertis,
Saurabh Desai, Saravanan Devendran, Janet Ellsworth, John Emmons, Jianhua Feng,
Lilian Fernandes, Paul Finley, Greg Flaig, Arnold Flores, Eric P Fried, David Geise, Denise
Genty, Jeffrey George, Mark A. Grubbs, Julianne Haugh, David A. Hepkin, Mary Hoetzel,
Vinit Jain, Patrick Laffey, Joseph Lampit, Zachary Loafman, Yantian Lu, Sheena Madan,
Michael Mall, Richard Mankowski, Christopher McDonald, Laura L. McWilliams, Alex
Medvedev, Steven Molis, James Moody, Frank Nichols, Ketan Pancholi, Shiva Persaud,
Marcos Quezada, Dino Quintero, Meenatchi Ramalingam, Eduardo Reyes, Mark Rogers,
Adam Ruprecht, Scot Sakolish, Carolyn Scherrer, Michael Schmidt, Ravi Shankar, Johnny
Shieh, Luc Smolders, Andrew N Solomon, Michael Spornhauer, Marc Stephenson, Kent
Stuiber, Jean-philippe Sugarbroad, Robert Thompson, Duyen Tong, Marvin Toungate,
Kedron J. Touvell, Basu Vaidyanathan, Vasu Vallabhaneni, Marcos A Villarreal, Drew
Walters, Wayne Wheeler, Ann Wigginton, Andy Wong, Jason Wu, Wu Zheng

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

xx


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!
We want our Redbooks to be as helpful as possible. Send us your comments
about this or other Redbooks in one of the following ways:
Use the online Contact us review redbook form found at:
ibm.com/redbooks

Send your comments in an Internet note to:
redbook@us.ibm.com

Mail your comments to:
IBM Corporation, International Technical Support Organization
Dept. JN9B, Building 905 Internal Zip 9053D003
11501 Burnet Road
Austin, Texas 78758-3493

Preface

xxi

xxii


1

Chapter 1.

Virtualization
This chapter discusses the virtualization engine technologies that are now
integrated into the p5 servers, AIX 5L Version 5.3. It is also found in Chapter 3 of
Advanced POWER Virtualization on IBM Sserver p5 Servers: Introduction and
Basic Configuration, SG24-7940.
These technologies include:
POWER™ Hypervisor™
Micro-Partitioning
Virtual Ethernet
Shared Ethernet Adapter
Virtual SCSI
PLM


1

1.1 Advanced POWER Virtualization feature
This section describes the packaging and ordering details for the Advanced
POWER Virtualization feature available on p5 systems.
The Advanced POWER Virtualization feature is a combination of hardware
enablement that includes the following components that are available together as
a single priced feature:
Firmware enablement for Micro-Partitioning
Installation image for the Virtual I/O Server software, which supports:
– Shared Ethernet Adapter
– Virtual SCSI server
Partition Load Manager
Note that virtual Ethernet is available without this feature when the server is
attached to an HMC, as well as LPAR and dynamic LPAR. SMT is available on
the base hardware with no additional features required.
Figure 1-1 shows an overview of the different parts of the hardware order that
enables firmware and includes the software orders.

HW Order

KEY

CD in box

CD in box

Advanced POW ER Virtualization Feature

Micro-Partitioning
Key enables both Micro-partitioning and Virtual IO-Server

Virtual I/O-Server (VIO)
(I/O Appliance, VLAN & Virtual SCSI Server)
Virtual I/O-Server Software Maintenance

Partition Load Manger (PLM)
PLM Software Maintenance

HW FC
Firmware
SW

Figure 1-1 Advanced POWER Virtualization feature

2


When the hardware feature is specified with the initial system order, the firmware
is shipped activated to support Micro-Partitioning and the Virtual I/O Server. For
upgrade orders, IBM will ship a key to enable the firmware similar to the CUoD
key.
Figure 1-2 shows the HMC panel where you enable the Virtualization Engine™
Technologies.

Figure 1-2 HMC panel to enable the Virtualization Engine Technologies

Virtual I/O Server and Partition Load Manager are licensed software components
of the Advanced POWER Virtualization feature. They contain one charge unit per
installed processor, including software maintenance. The initial software license
charge for Virtual I/O Server and PLM is included in the price of the Advanced
POWER Virtualization feature. The related hardware features that include Virtual
I/O Server and PLM are:
9111-520 Feature 7940
9113-550 Feature 7941
9117-570 Feature 7942

Chapter 1. Virtualization

3

For each Virtual I/O Server V1.1 license ordered, an order for either the one-year
(5771-VIO) or three-year (5773-VIO) Software Maintenance (SWMA) is also
submitted.
The processor-based license enables you to install multiple Virtual I/O Server
partitions on a single server to provide redundancy and to spread the Virtual I/O
Server workload across multiple partitions.
The Virtual I/O Server resides in a POWER5 partition as a single function
appliance that is created using the Hardware Management Console (HMC). The
Virtual I/O Server installation media ships with the POWER5 system that is
configured with the Advanced POWER Virtualization feature and supports
network install (NIMOL from HMC) or CD installation. It supports AIX 5L
Version 5.3, SUSE LINUX Enterprise Server 9 for POWER, and Red Hat
Enterprise Linux AS for POWER Version 3 as Virtual I/O clients.
The Virtual I/O Server provides the Virtual SCSI server and Shared Ethernet
Adapter virtual I/O function to client partitions (Linux or AIX). This POWER5
partition is not intended to run applications or for general user login.
For each Partition Load Manager V1.1 (5765-G31) license ordered, an order for
either the one-year (5771-PLM) or three-year (5773-PLM) Software Maintenance
(SWMA) must also be submitted. The Software Maintenance for the Partition
Load Manager is priced on a per processor basis, by processor group.
The Partition Load Manager for AIX 5L helps customers to maximize the
utilization of processor and memory resources on pSeries and p5 servers that
support dynamic logical partitioning. Within the constraints of a user-defined
policy, resources are automatically moved to partitions with a high demand, from
partitions with a lower demand. Resources that would otherwise go unused can
now be more fully utilized.
Partition Load Manager supports management of any dynamic LPAR running
AIX 5L Version 5.2 with the 5200-04 Recommended Maintenance Package or
AIX 5L Version 5.3 or later. The Partition Load Manager server can run on any
system with the appropriate level of IBM AIX.
The summary of features is as follows:
Automated processor and memory reconfiguration
Real-time partition configuration and load statistics
Support for dedicated and shared processor partition groups
Support for manual provisioning of resources
Graphical user interface (Web-based System Manager)

4


1.2 Introduction to the POWER Hypervisor
The POWER Hypervisor is an essential element of the IBM Virtualization Engine
system technologies implemented in the POWER5 processor based Sserver
family of products. Combined with features designed into the POWER5
processor, the POWER Hypervisor delivers functions that enable other system
technologies including Micro-Partitioning, virtualized processors, IEEE VLAN
compatible virtual switch, virtual SCSI adapters, and virtual consoles.
The POWER Hypervisor is a component of system firmware that will always be
installed and activated, regardless of system configuration. It operates as a
hidden partition, with no processor resources assigned to it.
Newly architected hypervisor calls (hcalls) provide a means for the operating
system to communicate with the POWER Hypervisor, allowing more efficient use
of physical processor capacity.
The POWER Hypervisor is a key component of the functions shown in
Figure 1-3. It performs the following tasks:
Provides an abstraction layer between the physical hardware resources and
the logical partitions using them
Enforces partition integrity by providing a security layer between logical
partitions
Controls the dispatch of virtual processors to physical processors
Saves and restores all processor state information during logical processor
context switch
Controls hardware I/O interrupts management facilities for logical partitions

Figure 1-3 POWER Hypervisor functions

The POWER Hypervisor, acting as the abstraction layer between the system
hardware and the logical partitions, allows multiple operating systems to run on
POWER5 technology with little or no modifications.


5

1.2.1 POWER Hypervisor implementation
The POWER4™ processor introduced support for logical partitioning with a new
privileged processor state called hypervisor mode. It is accessed using
hypervisor call functions, which are generated by the operating system kernel
running in a partition. Hypervisor mode allows for a secure mode of operation
that is required for various system functions where logical partition integrity and
security are required. The hypervisor validates that the partition has ownership of
the resources it is attempting to access, such as processor, memory, and I/O,
then completes the function. This mechanism allows for complete isolation of
partition resources.
In the POWER5 processor, further design enhancements are introduced that
enable the sharing of processors by multiple partitions. The hypervisor
decrementer (HDECR) is a new hardware facility in the POWER5 design that is
programmed to provide the POWER Hypervisor with a timed interrupt
independent of partition activity. HDECR interrupts are routed directly to the
POWER Hypervisor, and use only POWER Hypervisor resources to capture
state information from the partition. The HDECR is used for fine-grained
dispatching of multiple partitions on shared processors. It also provides a means
for the POWER Hypervisor to dispatch physical processor resources for its own
execution.
With the addition of shared partitions and SMT, a mechanism was required to
track physical processor resource utilization at a processor thread level. System
architecture for POWER5 introduces a new register called the Processor
Utilization Resource Register (PURR) to accomplish this. It provides the partition
with an accurate cycle count to measure activity during timeslices dispatched on
a physical processor. The PURR is a POWER Hypervisor resource, assigned
one per processor thread, that is incremented at a fixed rate whenever the thread
running on a virtual processor is dispatched on a physical processor.

1.2.2 POWER Hypervisor processor dispatch
Multiple logical partitions configured to run with a pool of shared physical
processors require a robust mechanism to guarantee the distribution of available
processing cycles. The POWER Hypervisor manages this task in the POWER5
processor based system.
Shared processor partition resource definition is discussed in detail in 1.3,
“Micro-Partitioning introduction” on page 14. This section introduces the
concepts of how the POWER Hypervisor dispatches work from multiple partitions
across physical processors.

6


Each shared processor partition is configured with a specific processor
entitlement, based on a quantity of processing units, which is referred to as the
partition’s entitled capacity. The entitled capacity, along with a defined number of
virtual processors, defines the physical processor resource that will be allotted to
the partition. The POWER Hypervisor uses the POWER5 HDECR, which is
programmed to generate an interrupt every 10 ms, as a timing mechanism for
controlling the dispatch of physical processors to system partitions. Each virtual
processor is guaranteed to get its entitled share of processor cycles during each
10 ms dispatch window.

Primary POWER Hypervisor hcalls
The primary POWER Hypervisor hcalls used by the operating system in the
dispatch of a virtual processor are:
cede

The cede call is used when a virtual processor or SMT thread
becomes idle, allowing the POWER Hypervisor to dispatch
other work.

confer

The confer call is used to grant the remaining cycles in a
dispatch interval to another virtual processor in the partition. It
can be used when one virtual processor cannot make forward
progress because it is waiting on an event to complete on
another virtual processor, such as a lock miss.

prod

The prod call is used to activate a virtual processor that has
ceded or conferred processor cycles.

A virtual processor will always be in one of four logical states. These states are:
running

Currently dispatched on a physical processor

runnable

Ready to run, waiting for dispatch

not-runnable

Has ceded or conferred its cycles

expired

Consumed its full entitled cycles for the current dispatch
window

Monitoring POWER Hypervisor hcalls
In AIX 5L Version 5.3, the lparstat command using the -h and -H flags will
display hypervisor statistical data about many hcalls, including cede, confer, and
prod. Using the -h flag adds summary hypervisor statistics to the default
lparstat output.


7

The following shows an example of this command, collecting statistics for one
five-second interval:
# lparstat -h 5 1
System configuration: type=Shared mode=Uncapped smt=On lcpu=4 mem=512 ent=0.50
%user %sys %wait %idle physc %entc lbusy
app vcsw phint %hypv hcalls
----- ---- ----- ----- ----- ----- -------- ---- ----- ----- -----0.0
0.5
0.0
99.4 0.00 1.0
0.0
- 1524
0
0.5
1542

Using the -H flag displays detailed hypervisor information, including statistics for
many hcall functions. The command shows the following for each of these hcalls:
Number of calls

Number of hypervisor calls made

Total Time Spent

Percentage of total time spent in this type of call

Hypervisor Time Spent Percentage of hypervisor time spent in this type of call
Average Call Time

Average call time for this type of call in nanoseconds

Maximum Call Time

Maximum call time for this type of call in nanoseconds

# lparstat -H 5 1
Detailed information on Hypervisor Calls
Hypervisor
Call
remove
read
nclear_mod
page_init
clear_ref
protect
put_tce
xirr
eoi
ipi
cppr
asr
others
enter
cede
migrate_dma
put_rtce
confer
prod

8

Number of
Calls

%Total Time
Spent

%Hypervisor
Time Spent

Avg Call
Time(ns)

Max Call
Time(ns)

2
21
0
3
0
0
14
10
10
0
0
0
0
5
1595
0
0
0
27

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.5
0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.0
0.0
0.0
0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
99.6
0.0
0.0
0.0
0.1

417
148
1
774
1
1
544
536
526
1
1
1
1
397
7918
1
1
1
672

550
294
0
2280
0
0
908
758
695
0
0
0
0
685
1343207
0
0
0
922


get_ppp
set_ppp
purr
pic
bulk_remove
send_crq
copy_rdma
get_tce
send_logical_lan
add_logicl_lan_buf

1
0
0
1
0
0
0
0
0
0

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

1502
1
1
309
1
1
1
1
1
1

2579
0
0
410
0
0
0
0
0
0

The basic concept of this dispatch mechanism is illustrated in Figure 1-4. In this
figure there are three logical partitions defined, sharing the processor cycles
from two physical processors, spanning two 10 ms hypervisor dispatch intervals.
Logical partition 1 is defined with an entitlement capacity of 0.8 processing units,
with two virtual processors. This allows the partition 80% of one physical
processor for each 10 ms dispatch window for the shared processor pool. For
each dispatch window, the workload is shown to use 40% of each physical
processor during each dispatch interval. It is possible for a virtual processor to be
dispatched more than one time during a dispatch interval. Note that in the first
dispatch interval, the workload executing on virtual processor 1 is not a
continuous utilization of physical processor resource. This can happen if the
operating system confers cycles, and is reactivated by a prod hcall.

Figure 1-4 Micro-Partitioning processor dispatch


9

Logical partition 2 is configured with one virtual processor and a capacity of 0.2
processing units, entitling it to 20% usage of a physical processor during each
dispatch interval. In this example, a worst case dispatch latency is shown for this
virtual processor, where the 2 ms are used in the beginning of dispatch interval 1
and the last 2 ms of dispatch interval 2, leaving 16 ms between processor
allocation.
Logical partition 3 contains 3 virtual processors, with an entitled capacity of 0.6
processing units. Each of the partition’s three virtual processors consumes 20%
of a physical processor in each dispatch interval, but in the case of virtual
processor 0 and 2, the physical processor they run on changes between dispatch
intervals. The POWER Hypervisor attempts to maintain physical processor
affinity when dispatching virtual processors. It will always first try to dispatch the
virtual processor on the same physical processor as it last ran on, and depending
on resource utilization will broaden its search out to the other processor on the
POWER5 chip, then to another chip on the same MCM, then to a chip on another
MCM.

Processor dispatch communication
The dispatch mechanism utilizes hcalls to communicate between the operating
system and the POWER Hypervisor. Implementing hcalls in the operating
system is desirable for performance reasons since they minimize the idle time of
a physical processor. Operating systems must be designed to use this new call
to exploit the full potential of POWER5.
When a virtual processor is active on a physical processor and the operating
system detects an inability to utilize processor cycles, it may cede or confer its
cycles back to the POWER Hypervisor, enabling it to schedule another virtual
processor on the physical processor for the remainder of the dispatch cycle.
Reasons for a cede or confer may include the virtual processor running out of
work and becoming idle, entering a spin loop to wait for a resource to free, or
waiting for a long latency access to complete. There is no concept of credit for
cycles that are ceded or conferred. Entitled cycles not used during a dispatch
interval are lost.
A virtual processor that has ceded cycles back to the POWER Hypervisor can be
reactivated using a prod hcall. If the operating system running on another virtual
processor within the logical partition detects that work is available for one of its
idle processors, it can use the prod hcall to signal the POWER Hypervisor to
make the prodded virtual processor runnable again. Once dispatched, this virtual
processor would resume execution at the return from the cede hcall.

10


In IBM AIX 5L Version 5.3, the mpstat command using the -d flag displays
detailed affinity and migration statistics for AIX threads and dispatching statistics
for logical processors.
# mpstat -d
System configuration: lcpu=4 ent=0.5
cpu
cs
0 68598
1
291
2 54514
3
751
ALL 124154

ics bound
38824
0
244
0
30174
1
624
0
69866
1

rq push S3pull S3grd
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0

S0rd
95.6
90.9
94.0
91.3
94.8

S1rd
0.0
7.4
0.1
2.9
0.1

S2rd
0.0
0.0
0.0
0.0
0.0

S3rd
4.4
1.7
6.0
5.8
5.1

S4rd
0.0
0.0
0.0
0.0
0.0

S5rd
ilcs
vlcs
0.0 174110 237393
0.0
1092 237759
0.0
2756 71779
0.0
1192 72971
0.0 89575 309951

The POWER Hypervisor dispatch affinity domains are defined as follows, and
statistics for virtual processor dispatch across these domains is given by the
mpstat command.
S0

The process redispatch occurs within the same logical processor. This
happens in the case of SMT-enabled systems.

S1

The process redispatch occurs within the same physical processor,
among different logical processors. This involves sharing of the L1, L2,
and L3 cache.

S2

The process redispatch occurs within the same processor chip, but
among different physical processors. This involves sharing of the L2 and
L3 cache.

S3

The process redispatch occurs within the same MCM module, but among
different processor chips.

S4

The process redispatch occurs within the same CEC plane, but among
different MCM modules. This involves access to the main memory or
L3-to-L3 transfer.

S5

The process redispatch occurs outside of the CEC plane.

As previously stated, the POWER Hypervisor will always first try to dispatch the
virtual processor on the same physical processor that it last ran on, and
depending on resource utilization will broaden its search out to the other
processor on the POWER5 chip, then to another chip on the same MCM, then to
a chip on another MCM.
Systems using DCMs share similar boundaries between processor cards, or
SMP Flex cables, as MCMs do between MCMs.


11

1.2.3 POWER Hypervisor and virtual I/O
Detailed discussion of POWER5 Sserver virtual I/O implementation is found in
various sections of this publication. This section introduces POWER Hypervisor
involvement in the virtual I/O functions.
With the introduction of Micro-Partitioning, the ability to dedicate physical
hardware adapter slots to each partition becomes impractical. Virtualization of
I/O devices allows many partitions to communicate with each other, and access
networks and storage devices external to the server, without dedicating physical
I/O resources to an individual partition. Many of the I/O virtualization capabilities
introduced with the POWER5 processor based Sserver products are
accomplished by functions designed into the POWER Hypervisor.
The POWER Hypervisor does not own any physical I/O devices, and it does not
provide virtual interfaces to them. All physical I/O devices in the system are
owned by logical partitions. Virtual I/O devices are owned by the Virtual I/O
Server, which provides access to the real hardware that the virtual device is
based on.
The POWER Hypervisor implements the following operations required by system
partitions to support virtual I/O:
Provides control and configuration structures for virtual adapter images
required by the logical partitions
Allows partitions controlled and secure access to physical I/O adapters in a
different partition
Along with these operations, the POWER Hypervisor allows for the virtualization
of I/O interrupts. To maintain partition isolation, the POWER Hypervisor controls
the hardware interrupt management facilities. Each logical partition is provided
controlled access to the interrupt management facilities using hcalls. Virtual I/O
adapters and real I/O adapters use the same set of hcall interfaces.

I/O types supported
Three types of virtual I/O adapters are supported by the POWER Hypervisor:
SCSI
Ethernet
Serial console (virtual TTY)
Virtual I/O adapters are defined by system administrators during logical partition
definition. Configuration information for the virtual adapters is presented to the
partition operating system by the system firmware.

12


Virtual SCSI support
The p5 server uses SCSI as the mechanism for virtual storage devices. This is
accomplished using two paired adapters: a virtual SCSI server adapter and a
virtual SCSI client adapter. These adapters are used to transfer SCSI commands
between partitions. To the operating system, the virtual SCSI client adapter is no
different than a typical SCSI adapter. The SCSI server, or target, adapter is
responsible for executing any SCSI command it receives. It is owned by the
Virtual I/O server partition. How this SCSI command is executed is dependent on
the hosting partition operating system. See 1.6, “Virtual SCSI introduction” on
page 46 for a detailed discussion of virtual SCSI.
Virtual SCSI operation is dependant on two system functions implemented in the
POWER Hypervisor.
Message queuing function that enables the passing of messages between
partitions, with an interrupt mechanism to confirm receipt of the message
DMA function between partitions that provides control structures for secure
transfers between logical partition memory spaces

Virtual Ethernet switch support
The POWER Hypervisor provides a Virtual Ethernet switch function that allows
partitions on the same server a means for fast and secure communication. It is
based on the IEEE 802.1Q VLAN standard. No physical I/O adapter is required
when creating a VLAN connection between partitions, and no access to an
outside network is required. Each partition operating system sees the VLAN
switch as an Ethernet adapter, without the physical link properties and
asynchronous data transmit operations. See 1.4, “Virtual Ethernet introduction”
on page 32 for a detailed discussion of Virtual Ethernet.
Virtual Ethernet transmit operations are performed synchronously and are
complete as soon as the hypervisor call to send a frame returns. Receive
operations are performed using a pool of buffers provided to the POWER
Hypervisor for receiving frames. As incoming frames are received by the
adapter, the POWER Hypervisor sends a virtual interrupt back to the device
driver.
Each Virtual Ethernet adapter can also be configured as a trunk adapter. This
enables layer-2 bridging to a physical Ethernet adapter to be performed,
extending the Virtual Ethernet network outside the server to a physical Ethernet
network. If a partition sends an Ethernet frame to a MAC address that is
unknown by the POWER Hypervisor virtual switch, it will be forwarded to any
trunk adapter defined for the same VLAN ID.


13

Virtual TTY console support
Each partition needs to have access to a system console. Tasks such as
operating system install, network setup, and some problem analysis activities
require a dedicated system console. The POWER Hypervisor provides virtual
console using a virtual TTY or serial adapter and a set of hypervisor calls to
operate on them.
Depending on the system configuration, the operating system console can be
provided by the Hardware Management Console (HMC) virtual TTY, or from a
terminal emulator connected to physical serial ports on the system’s service
processor.

1.3 Micro-Partitioning introduction
The concept of Micro-Partitioning allows the resource definition of a partition to
allocate fractions of processors to the partition.
On POWER4 systems, all partitions are considered dedicated, in that the
processors assigned to a partition can only be in whole multiples and only
used by that partition.
On POWER5 systems, you can choose between dedicated processor
partitions and shared processor partitions using Micro-Partitioning.
The benefit of Micro-Partitioning is that it allows increased overall utilization of
system resources by automatically applying only the required amount of
processor resource needed by each partition. Sales material tends to discuss
Micro-Partitioning in terms of fractional processor units; however, resource can
also be defined as increments greater than a single processor.
The POWER Hypervisor continually adjusts the amount of processor capacity
allocated to each shared processor partition and any excess capacity
unallocated based on current partition profiles within a shared pool. Tuning
parameters allow the administrator extensive control over the amount of
processor resources that each partition can use.
This section discusses the following topics about Micro-Partitioning:
Shared processor partitions
Shared pool overview
CUoD
Dynamic LPAR
Limitations of Micro-Partitioning

14


1.3.1 Shared processor partitions
The virtualization of processors enables the creation of a partitioning model
which is fundamentally different from the POWER4 systems, where whole
processors are assigned to partitions and are owned by them. In the new model,
physical processors are abstracted into virtual processors that are then assigned
to partitions, but the underlying physical processors are shared by these
partitions.
Virtual processor abstraction is implemented in the hardware and microcode.
From an operating system perspective, a virtual processor is indistinguishable
from a physical processor. The key benefit of implementing partitioning in the
hardware allows any operating system to run on POWER5 technology with little
or no changes. Optionally, for optimal performance, the operating system can be
enhanced to exploit shared processor pools more in-depth, for instance, by
voluntarily relinquishing CPU cycles to the hardware when they are not needed.
AIX 5L Version 5.3 is the first version of AIX 5L that includes such
enhancements.
Micro-Partitioning allows for multiple partitions to share one physical processor.
Partitions using Micro-Partitioning technology are referred to as shared
processor partitions.
A partition may be defined with a processor capacity as small as 10 processor
units. This represents 1/10 of a physical processor. Each processor can be
shared by up to 10 shared processor partitions. The shared processor partitions
are dispatched and time-sliced on the physical processors under control of the
POWER Hypervisor.
Micro-Partitioning is supported across the entire POWER5 product line from the
entry to the high-end systems. Table 1-1 shows the maximum number of logical
partitions and shared processor partitions supported on the different models.
Table 1-1 Micro-Partitioning overview on p5 systems
p5 servers

Model 520

Model 550

Model 570

Processors

2

4

16

Dedicated processor partitions

2

4

16

Shared processor partitions

20

40

160

It is important to point out that the maximums stated are supported by the
hardware, but the practical limits based on production workload demands may
be significantly lower.


15

Shared processor partitions still need dedicated memory, but the partitions’ I/O
requirements can be supported through Virtual Ethernet and Virtual SCSI.
Utilizing all virtualization features, support for up to 254 shared processor
partitions is planned.
The shared processor partitions are created and managed by the HMC. When
you start creating a partition you have to choose between a shared processor
partition and a dedicated processor partition.
When setting up a partition you have to define the resources that belong to the
partition, such as memory and I/O resources. For processor shared partitions
you have to configure these additional options:
Minimum, desired, and maximum processing units of capacity
The processing sharing mode, either capped or uncapped
If the partition is uncapped, specify its variable capacity weight.
Minimum, desired, and maximum virtual processors
These settings are the topic of the following sections.

Processing units of capacity
Processing capacity can be configured in fractions of 1/100 of a processor. The
minimum amount of processing capacity which has to be assigned to a partition
is 1/10 of a processor.
On the HMC, processing capacity is specified in terms of processing units. The
minimum capacity of 1/10 of a processor is specified as 0.1 processing units. To
assign a processing capacity representing 75% of a processor, 0.75 processing
units are specified on the HMC.
On a system with two processors a maximum of 2.0 processing units can be
assigned to a partition. Processing units specified on the HMC are used to
quantify the minimum, desired, and maximum amount of processing capacity for
a partition.
Once a partition is activated, processing capacity is usually referred to as
capacity entitlement or entitled capacity.
Figure 1-5 on page 17 shows a graphical view of the definitions of processor
capacity.

16


Minimum Requirement
0.1 Processing Units


Processing Capacity
1 Physical Processor

Figure 1-5 Processing units of capacity

Capped and uncapped mode
The next step in defining a shared processor partition is to specify whether the
partition is running in a capped or uncapped mode.
Capped mode

The processor unit never exceeds the assigned
processing capacity.

Uncapped mode

The processing capacity may be exceeded when the
shared processing pools have available resource.

When a partition is run in an uncapped mode, you must specify the uncapped
weight of that partition.
If multiple uncapped logical partitions require idle processing units, the managed
system distributes idle processing units to the logical partitions in proportion to
each logical partition's uncapped weight. The higher the uncapped weight of a
logical partition, the more processing units the logical partition gets.
The uncapped weight must be a whole number from 0 to 255. The default
uncapped weight for uncapped logical partitions is 128. A partition's share is
computed by dividing its variable capacity weight by the sum of the variable
capacity weights for all uncapped partitions. If you set the uncapped weight at 0,
the managed system treats the logical partition as a capped logical partition. A
logical partition with an uncapped weight of 0 cannot use more processing units
than those that are committed to the logical partition.


17

Virtual processors
Virtual processors are the whole number of concurrent operations that the
operating system can use. The processing power can be conceptualized as
being spread equally across these virtual processors. Selecting the optimal
number of virtual processors depends on the workload in the partition. Some
partitions benefit from greater concurrence, while other partitions require greater
power.
By default, the number of processing units that you specify is rounded up to the
minimum number of virtual processors needed to satisfy the assigned number of
processing units. The default settings maintain a balance of virtual processors to
processor units. For example:
If you specify 0.50 processing units, one virtual processor will be assigned.
If you specify 2.25 processing units, three virtual processors will be assigned.
You also can use the Advanced tab in your partitions profile to change the default
configuration and to assign more virtual processors.
At the time of publication, the maximum number of virtual processors per
partition is 64.
A logical partition in the shared processing pool will have at least as many virtual
processors as its assigned processing capacity. By making the number of virtual
processors too small, you limit the processing capacity of an uncapped partition.
If you have a partition with 0.50 processing units and 1 virtual processor, the
partition cannot exceed 1.00 processing units because it can only run one job at
a time, which cannot exceed 1.00 processing units. However, if the same
partition with 0.50 processing units was assigned two virtual processors and
processing resources were available, the partition could use an additional 1.50
processing units.

Dedicated processors
Dedicated processors are whole processors that are assigned to a single
partition. If you choose to assign dedicated processors to a logical partition, you
must assign at least one processor to that partition.
You cannot mix shared processors and dedicated processors in one partition.
By default, a powered-off logical partition using dedicated processors will have
its processors available to the shared processing pool. When the processors are
in the shared processing pool, an uncapped partition that needs more processing
power can use the idle processing resources. However, when you power on the
dedicated partition while the uncapped partition is using the processors, the
activated partition will regain all of its processing resources. If you want to

18


prevent dedicated processors from being used in the shared processing pool,
you can disable this function using the logical partition profile properties panels
on the Hardware Management Console.
Note: You cannot disable the “Allow idle processor to be shared” function
when you create a partition. You need to open the properties for the created
partition and change it on the Processor tab.

1.3.2 Shared pool overview
The POWER Hypervisor schedules shared processor partitions from a set of
physical processors that is called the shared processor pool. By definition, these
processors are not associated with dedicated partitions.
In shared partitions there is no fixed relationship between virtual processors and
physical processors. The POWER Hypervisor can use any physical processor in
the shared processor pool when it schedules the virtual processor. By default, it
attempts to use the same physical processor, but this cannot always be
guaranteed. The POWER Hypervisor uses the concept of a home node for virtual
processors, enabling it to select the best available physical processor from a
memory affinity perspective for the virtual processor that is to be scheduled.
Affinity scheduling is designed to preserve the content of memory caches, so that
the working data set of a job can be read or written in the shortest time period
possible. Affinity is actively managed by the POWER Hypervisor since each
partition has a completely different context. Currently, there is one shared
processor pool, so all virtual processors are implicitly associated with the same
pool.
Figure 1-6 on page 20 shows the relationship between two partitions using a
shared processor pool of a single physical CPU. One partition has two virtual
processors and the other a single one. The figure also shows how the capacity
entitlement is evenly divided over the number of virtual processors.
When you set up a partition profile, you set up the desired, minimum, and
maximum values you want for the profile. When a partition is started, the system
chooses the partition's entitled processor capacity from this specified capacity
range. The value that is chosen represents a commitment of capacity that is
reserved for the partition. This capacity cannot be used to start another shared
partition, otherwise capacity could be overcommitted.


19

LPAR 2 Capacity Entitlement 40

LPAR 1 Capacity Entitlement 50
Virtual
Processor 2

Virtual
Processor 1
25

Virtual
Processor 1
40

25


1 Physical Processor

Figure 1-6 Distribution of capacity entitlement on virtual processors

When starting a partition, preference is given to the desired value, but this value
cannot always be used because there may not be enough unassigned capacity
in the system. In that case, a different value is chosen, which must be greater
than or equal to the minimum capacity attribute. Otherwise, the partition cannot
be started.
The entitled processor capacity is distributed to the partitions in the sequence the
partitions are started. For example, consider a shared pool that has 2.0
processing units available.
Partitions 1, 2, and 3 are activated in sequence:
Partition 1 activated
Min. = 1.0, max = 2.0, desired = 1.5
Allocated capacity entitlement: 1.5
Min. = 1.0, max = 2.0, desired = 1.0
Partition 2 does not start because the minimum capacity is not met.
Min. = 0.1, max = 1.0, desired = 0.8
Allocated capacity entitlement: 0.5
The maximum value is only used as an upper limit for dynamic operations.

20


Figure 1-7 shows the usage of a capped partition of the shared processor pool.
Partitions using the capped mode are not able to assign more processing
capacity from the shared processor pool than the capacity entitlement will allow.

Pool Idle Capacity Available

Maximum Processor Capacity

Processor
Capacity
Utilization

Entitled Processor Capacity

Ceded
LPAR Capacity Capacity
Utilization
Minimum Processor Capacity

Utilized Capacity

Time

Figure 1-7 Capped shared processor partitions

Figure 1-8 on page 22 shows the usage of the shared processor pool by an
uncapped partition. The uncapped partition is able to assign idle processing
capacity if it needs more than the entitled capacity.


21

Pool Idle Capacity Available

Maximum Processor Capacity

Processor
Capacity
Utilization

Entitled Processor Capacity

Ceded Capacity
Minimum Processor Capacity

Utilized Capacity

Time
Figure 1-8 Uncapped shared processor partition

1.3.3 Dynamic partitioning
Dynamic partitioning was introduced with AIX 5L Version 5.2. An AIX 5L Version
5.2 based dynamic partition can consist of the following resource elements:
A dedicated processor
256 MB memory region
I/O adapter slot
Multiple resources can be placed under the exclusive control of a given logical
partition. Dynamic LPAR extends these capabilities by allowing this fine-grained
resource allocation to occur not only when activating a logical partition, but also
while the partitions are running. Individual processors, memory regions, and I/O
adapter slots can be released into a free pool, acquired from that free pool, or
moved directly from one partition to another.
On POWER5 with AIX 5L Version 5.3, a partition can consist of dedicated
processors, or virtual processors with a specific capacity entitlement running in
capped or uncapped mode, dedicated memory region, and virtual or physical I/O
adapter slots.

22


For dedicated and shared processor partitions it is possible to dynamically:
Add, move, or remove memory in a granularity of 16 MB regions
Add, move, or remove physical I/O adapter slots
Add or remove virtual I/O adapter slots
For a dedicated processor partition it is only possible to dynamically add, move,
or remove whole processors. When you dynamically remove a processor from a
dedicated partition on a system that uses shared processor partitions it is then
assigned to the shared processor pool.
For shared processor partitions it is also possible to dynamically:
Remove, move, or add entitled shared processor capacity
Change between capped and uncapped processing
Change the weight of an uncapped partition
Add and remove virtual processors
Figure 1-9 on page 24 shows the panel for dynamic reconfiguration of the
processor resources on the HMC. Here you can choose to add, remove, or move
your resources. Select the partition that you want to change dynamically and
press the right mouse button. Then choose Dynamic Logical Partitioning from
the menu, select Processor Resources, and choose the action you want to
perform.


23

Figure 1-9 HMC panel Dynamic Logical Partitioning

When you select Processor Resources → Add, the panel shown in Figure 1-10
on page 25 is displayed. Here you can specify the processing units and the
number of virtual processors you want to add to the selected partition. The limits
for adding processing units and virtual processors are the maximum values
defined in the partition profile. This panel also allows you to add variable weight
when the partition runs in uncapped mode.

24


Figure 1-10 Add Processor Resource panel on the HMC

Additionally, the Add Processor Resource panel allows you to dynamically
change the partition mode from uncapped to capped or vice versa. To show the
actual status of the partition, use the lparstat -i command from the AIX
command line interface of the partition, as shown in the following:
# lparstat -i
Node Name
Partition Name
Partition Number
Type
Mode
Entitled Capacity
Partition Group-ID
Shared Pool ID
Online Virtual CPUs
Maximum Virtual CPUs
Minimum Virtual CPUs
Online Memory
Maximum Memory
Minimum Memory

:
:
:
:
:
:
:
:
:
:
:
:
:
:

applsrv
Apps_Server
4
Shared-SMT
Uncapped
0.30
32772
0
2
10
1
512 MB
1024 MB
128 MB


25

Variable Capacity Weight
Minimum Capacity
Maximum Capacity
Capacity Increment
Maximum Dispatch Latency
Maximum Physical CPUs in system
Active Physical CPUs in system
Active CPUs in Pool
Unallocated Capacity
Physical CPU Percentage
Unallocated Weight

:
:
:
:
:
:
:
:
:
:
:

128
0.20
1.00
0.01
16999999
2
2
0.00
15.00%
0

Figure 1-11 shows the way to change the mode of the partition from uncapped to
capped mode. Un-check the Uncapped check box and click OK.

Figure 1-11 Changing the partition mode from Uncapped to Capped

26


To verify this dynamic action, use the lparstat -i command on the selected
partition again. The partition mode has changed from uncapped to capped.
# lparstat -i
Node Name
Partition Name
Partition Number
Type
Mode
Entitled Capacity
Partition Group-ID
Shared Pool ID
Online Virtual CPUs
Online Memory
Maximum Memory
Minimum Memory
Minimum Capacity
Maximum Capacity
Capacity Increment
Maximum Dispatch Latency
Active CPUs in Pool
Unallocated Weight

:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:

applsrv
Apps_Server
4
Shared-SMT
Capped
0.30
32772
0
2
10
1
512 MB
1024 MB
128 MB
128
0.20
1.00
0.01
16999999
2
2
0.00
15.00%
0

Figure 1-12 shows the Remove Processor Resources panel that allows you to
dynamically remove processing units and virtual processors. The limits for the
removal of processing units and virtual processors are the minimum values
defined in the partition profile.
This panel also allows you to remove variable weight when the partition runs in
uncapped mode.


27

Figure 1-12 Remove Processor Resource panel on the HMC

It is also possible to change the partition mode from capped to uncapped and
vice versa from the Remove Processor Resource panel.

28


When moving processing units you have to select the partition you want the
processing units removed from and choose the Move Processor Resource Panel
shown in Figure 1-13.

Figure 1-13 Move Processor Resources panel on HMC

In the Processing units field, select the amount of processor capacity you want to
remove from the selected partition and move to the partition you select from the
drop-down menu under Logical Partition. In this example, we want to move 0.1
processing units from the Web_Server partition to the Apps_Server partition.
You can also choose to move virtual processors to adjust the number of virtual
processors of your partition. This does not actually move the virtual processor
but removes and adds the defined number of virtual processors to the chosen
partitions.


29

1.3.4 Limitations and considerations
The following limitations must be considered when implementing shared
processor partitions:
The limitation for a shared processor partition is 0.1 processing units of a
physical processor, so the number of shared processor partitions you can
create for a system depends mostly on the number of processors in a system.
The maximum number of partitions planned is 254.
The maximum number of virtual processors in a partition is 64.
A mix of dedicated and shared processors within the same partition is not
supported.
If you dynamically remove a virtual processor you cannot specify a particular
virtual CPU to be removed. The operating system will choose the virtual CPU
to be removed.
Shared processors may render AIX affinity management useless. AIX will
continue to utilize affinity domain information as provided by firmware to build
associations of virtual processors to memory, and will continue to show
preference to redispatching a thread to the virtual CPU that it last ran on.
There is overhead associated with the maintenance of online virtual processors,
so you should carefully consider their capacity requirements before choosing
values for these attributes.
Virtual processors have dispatch latency since they are scheduled. When a
virtual processor is made runnable, it is placed on a run queue by the POWER
Hypervisor, where it waits until it is dispatched. The time between these two
events is referred to as dispatch latency.
The dispatch latency of a virtual processor depends on the partition entitlement
and the number of virtual processors that are online in the partition. The capacity
entitlement is equally divided among these online virtual processors, so the
number of online virtual processors impacts the length of each virtual processor's
dispatch. The smaller the dispatch cycle, the greater the dispatch latency.
At the time of writing, the worst case virtual processor dispatch latency is 18
milliseconds since the minimum dispatch cycle that is supported at the virtual
processor level is one millisecond. This latency is based on the minimum
partition entitlement of 1/10 of a physical processor and the 10 millisecond
rotation period of the Hypervisor's dispatch wheel. It can be easily visualized by
imagining that a virtual processor is scheduled in the first and last portions of two
10 millisecond intervals. In general, if these latencies are too great, then clients
may increase entitlement, minimize the number of online virtual processors
without reducing entitlement, or use dedicated processor partitions.

30


In general, the value of the minimum, desired, and maximum virtual processor
attributes should parallel those of the minimum, desired, and maximum capacity
attributes in some fashion. A special allowance should be made for uncapped
partitions, since they are allowed to consume more than their entitlement.
If the partition is uncapped, then the administrator may want to define the desired
and maximum virtual processor attributes x% above the corresponding
entitlement attributes. The exact percentage is installation-specific, but 25 to 50
percent is a reasonable number.
Table 1-2 shows several reasonable settings of number of virtual processors,
processing units, and the capped and uncapped mode.
Table 1-2 Reasonable settings for shared processor partitions
Min
VPsa

Desired
VPs

Max VPs

Min PUb

Desired
PU

Max. PU

Capped

1

2

4

0.1

2.0

4.0

Y

1

3 or 4

6 or 8

0.1

2.0

4.0

N

2

2

6

2.0

2.0

6.0

Y

2

3 or 4

8 or 10

2.0

2.0

6.0

N

a - Virtual processors
b - Processing units

Operating systems and applications running in shared partitions need not be
aware that they are sharing processors. However, overall system performance
can be significantly improved by minor operating system changes. AIX 5L
Version 5.3 provides support for optimizing overall system performance of
shared processor partitions.
In a shared partition there is not a fixed relationship between the virtual
processor and the physical processor. The POWER Hypervisor will try to use a
physical processor with the same memory affinity as the virtual processor, but it
is not guaranteed. Virtual processors have the concept of a home physical
processor. If it can't find a physical processor with the same memory affinity, then
it gradually broadens its search to include processors with weaker memory
affinity, until it finds one that it can use. As a consequence, memory affinity is
expected to be weaker in shared processor partitions.
Workload variability is also expected to be increased in shared partitions
because there are latencies associated with the scheduling of virtual processors
and interrupts. SMT may also increase variability, since it adds another level of


31

resource sharing, which could lead to a situation where one thread interferes with
the forward progress of its sibling.
Therefore, if an application is cache-sensitive or cannot tolerate variability, then it
should be deployed in a dedicated partition with SMT disabled. In dedicated
partitions, the entire processor is assigned to a partition. Processors are not
shared with other partitions, and they are not scheduled by the POWER
Hypervisor. Dedicated partitions must be explicitly created by the system
administrator using the Hardware Management Console.
Processor and memory affinity data is only provided in dedicated partitions. In a
shared processor partition, all processors are considered to have the same
affinity. Affinity information is provided through RSET APIs, which contain
discovery and bind services.

1.4 Virtual Ethernet introduction
Virtual Ethernet enables inter-partition communication without the need for
physical network adapters assigned to each partition. Virtual Ethernet allows the
administrator to define in-memory point-to-point connections between partitions.
These connections exhibit characteristics similar to physical high-bandwidth
Ethernet connections and support multiple protocols (IPv4, IPv6, ICMP). Virtual
Ethernet requires a p5 system with either AIX 5L Version 5.3 or the appropriate
level of Linux and an HMC to define the Virtual Ethernet devices. Virtual Ethernet
does not require the purchase of any additional features or software such as the
Advanced POWER Virtualization Feature.
The concepts of implementing Virtual Ethernet on p5 systems are discussed in
the following sections:
Virtual LAN
Virtual Ethernet connections
Benefits of Virtual Ethernet
Limitations

1.4.1 Virtual LAN
This section discusses the concepts of Virtual LAN (VLAN) technology with
specific reference to its implementation within AIX.

Virtual LAN overview
Virtual LAN is a technology used for establishing virtual network segments on top
of physical switch devices. If configured appropriately, a VLAN definition can

32


straddle multiple switches. Typically, a VLAN is a broadcast domain that enables
all nodes in the VLAN to communicate with each other without any L3 routing or
inter-VLAN bridging. In Figure 1-14, two VLANs (VLAN 1 and 2) are defined on
three switches (Switch A, B, and C). Although nodes C-1 and C-2 are physically
connected to the same switch C, traffic between two nodes can be blocked. To
enable communication between VLAN 1 and 2, L3 routing or inter-VLAN bridging
should be established between them; this is typically provided by an L3 device.

Figure 1-14 Example of a VLAN

The use of VLAN provides increased LAN security and flexible network
deployment over traditional network devices.


33

AIX virtual LAN support
Some of the technologies for implementing VLANs include:
Port-based VLAN
Layer 2 VLAN
Policy-based VLAN
IEEE 802.1Q VLAN
VLAN support in AIX is based on the IEEE 802.1Q VLAN implementation. The
IEEE 802.1Q VLAN is achieved by adding a VLAN ID tag to an Ethernet frame,
and the Ethernet switches restricting the frames to ports that are authorized to
receive frames with that VLAN ID. Switches also restrict broadcasts to the logical
network by ensuring that a broadcast packet is delivered to all ports which are
configured to receive frames with the VLAN ID that the broadcast frame was
tagged with.
A port on a VLAN-capable switch has a default PVID (Port virtual LAN ID) that
indicates the default VLAN the port belongs to. The switch adds the PVID tag to
untagged packets that are received by that port. In addition to a PVID, a port may
belong to additional VLANs and have those VLAN IDs assigned to it that indicate
the additional VLANs the port belongs to.
A port will only accept untagged packets or packets with a VLAN ID (PVID or
additional VIDs) tag of the VLANs the port belongs to. A port configured in the
untagged mode is only allowed to have a PVID and will receive untagged
packets or packets tagged with the PVID. The untagged port feature helps
systems that do not understand VLAN tagging communicate with other systems
using standard Ethernet.
Each VLAN ID is associated with a separate Ethernet interface to the upper
layers (for example IP) and creates unique logical Ethernet adapter instances
per VLAN (for example ent1 or ent2).
You can configure multiple VLAN logical devices on a single system. Each VLAN
logical device constitutes an additional Ethernet adapter instance. These logical
devices can be used to configure the same Ethernet IP interfaces as are used
with physical Ethernet adapters.

VLAN communication by example
This section discusses how VLAN communication between partitions and with
external networks works in more detail, using the sample configuration in
Figure 1-15 on page 35. The configuration is using four client partitions (Partition
1 through Partition 4) and one Virtual I/O Server. Each of the client partitions is
defined with one Virtual Ethernet adapter. The Virtual I/O Server has a Shared

34


Ethernet Adapter which bridges traffic to the external network. The Shared
Ethernet Adapter is discussed in more detail in 1.5, “Shared Ethernet Adapter”
on page 39.

Figure 1-15 VLAN configuration

Interpartition communication
Partition 2 and Partition 4 are using the PVID (Port virtual LAN ID) only. This
means that:
Only packets for the VLAN specified as PVID are received.
Packets sent have a VLAN tag added for the VLAN specified as PVID by the
Virtual Ethernet adapter.
In addition to the PVID, the Virtual Ethernet adapters in Partition 1 and Partition 3
are also configured for VLAN 10 using a specific network interface (en1) created
through smitty vlan. This means that:
Packets sent through network interfaces en1 are added a tag for VLAN 10 by
the network interface in AIX.
Only packets for VLAN 10 are received by the network interfaces en1.
Packets sent through en0 are automatically tagged for the VLAN specified as
PVID.
Only packets for the VLAN specified as PVID are received by the network
interfaces en0.


35

Table 1-3 lists which client partitions can communicate which each other through
what network interfaces.
Table 1-3 Interpartition VLAN communication
VLAN

Partition / Network interface

1

Partition 1 / en0
Partition 2 / en0

2

Partition 3 / en0
Partition 4 / en0

10

Partition 1 / en1
Partition 3 / en1

Communication with external networks
The Shared Ethernet Adapter is configured with PVID 1 and VLAN 10. This
means that untagged packets that are received by the Shared Ethernet Adapter
are tagged for VLAN 1. Handling of outgoing traffic depends on the VLAN tag of
the outgoing packets.
Packets tagged with the VLAN which matches the PVID of the Shared
Ethernet Adapter are untagged before being sent out to the external network.
Packets tagged with a VLAN other than the PVID of the Shared Ethernet
Adapter are sent out with the VLAN tag unmodified.
In our example, Partition 1 and Partition 2 have access to the external network
through network interface en0 using VLAN 1. Since these packets are using the
PVID, the Shared Ethernet Adapter will remove the VLAN tags before sending
the packets to the external network.
Partition 1 and Partition 3 have access to the external network using network
interface en1 and VLAN 10. These packets are sent out by the Shared Ethernet
Adapter with the VLAN tag. Therefore, only VLAN-capable destination devices
will be able to receive the packets. Table 1-4 lists this relationship.
Table 1-4 VLAN communication to external network
VLAN
1

Partition 1 / en0
Partition 2 / en0

10

36

Partition / Network interface

Partition 1 / en1
Partition 3 / en1


1.4.2 Virtual Ethernet connections
Virtual Ethernet connections supported in POWER5 systems use VLAN
technology to ensure that the partitions can only access data directed to them.
The POWER Hypervisor provides a Virtual Ethernet switch function based on the
IEEE 802.1Q VLAN standard that allows partition communication within the
same server. The connections are based on an implementation internal to the
hypervisor that moves data between partitions. This section describes the
various elements of a Virtual Ethernet and implications relevant to different types
of workloads. Figure 1-16 is an example of an inter-partition VLAN.

Figure 1-16 Logical view of an inter-partition VLAN

Virtual Ethernet adapter concepts
Partitions that communicate through a Virtual Ethernet channel will need to have
an additional in-memory channel. This requires the creation of an in-memory
channel between partitions on the HMC. The kernel creates a virtual device for
each memory channel indicated by the firmware. The AIX configuration manager
creates the device special files. A unique Media Access Control (MAC) address
is also generated when the Virtual Ethernet device is created. A prefix value can
be assigned for the system so that the generated MAC addresses in a system


37

consist of a common system prefix, plus an algorithmically-generated unique part
per adapter.
The Virtual Ethernet can also be used as a bootable device to allow such tasks
as operating system installations to be performed using NIM.

Performance considerations
The transmission speed of Virtual Ethernet adapters is in the range of 1-3
Gigabits per second, depending on the transmission (MTU) size. A partition can
support up to 256 Virtual Ethernet adapters with each Virtual Ethernet capable of
being associated with up to 21 VLANs (20 VID and 1 PVID).
The Virtual Ethernet connections generally take up more processor time than a
local adapter to move a packet (DMA versus copy). For shared processor
partitions, performance will be gated by the partition definitions (for example,
entitled capacity and number of processors). Small partitions communicating
with each other will experience more packet latency due to partition context
switching. In general, high bandwidth applications should not be deployed in
small shared processor partitions. For dedicated partitions, throughput should be
comparable to a 1 Gigabit Ethernet for small packets providing much better
performance than 1 Gigabit Ethernet for large packets. For large packets, the
Virtual Ethernet communication is copy bandwidth limited.

1.4.3 Benefits of Virtual Ethernet
Because the number of partitions possible on many systems is greater than the
number of I/O slots, Virtual Ethernet is a convenient and cost saving option to
enable partitions within a single system to communicate with one another
through a VLAN. The VLAN creates logical Ethernet connections between one or
more partitions and is designed to help prevent a failed or malfunctioning
operating system from being able to impact the communication between two
functioning operating systems. The Virtual Ethernet connections may also be
bridged to an external network to permit partitions without physical network
adapters to communicate outside of the server.

1.4.4 Dynamic partitioning for Virtual Ethernet devices
Virtual Ethernet resources can be assigned and removed dynamically. On the
HMC, Virtual Ethernet target and server adapters can be assigned and removed
from a partition using dynamic logical partitioning. The mapping between
physical and virtual resources on the Virtual I/O Server can also be done
dynamically.

38


The following are limitations that must be considered when implementing a
Virtual Ethernet:
A maximum of up to 256 Virtual Ethernet adapters are permitted per partition.
Virtual Ethernet can be used in both shared and dedicated processor
partitions provided the partition is running AIX 5L Version 5.3 or Linux with the
2.6 kernel or a kernel that supports virtualization.
A mixture of Virtual Ethernet connections, real network adapters, or both are
permitted within a partition.
Virtual Ethernet can only connect partitions within a single system.
Virtual Ethernet requires a POWER5 system and an HMC to define the Virtual
Ethernet adapters.
Virtual Ethernet connections from AIX or Linux partitions to an i5/OS™
partition may work; however, at the time of writing, these capabilities were
unsupported.
Virtual Ethernet uses the system processors for all communication functions
instead of offloading that load to processors on network adapter cards. As a
result, there is an increase in system processor load generated by the use of
Virtual Ethernet.

1.5 Shared Ethernet Adapter
A Shared Ethernet Adapter can be used to connect a physical Ethernet to the
Virtual Ethernet. It also provides the possibility for several client partitions to
share one physical adapter.
The following sections discuss the various aspects of Shared Ethernet Adapters
such as:
Connecting Virtual Ethernet to external networks
Ethernet adapter sharing
Benefits of Shared Ethernet Adapters
Using Link Aggregation (EtherChannel) for external network interface
Limitations and considerations


39

1.5.1 Connecting a Virtual Ethernet to external networks
There are two ways you can connect the Virtual Ethernet that enables the
communication between logical partitions on the same server to an external
network.

Routing
By enabling the AIX routing capabilities (ipforwarding network option) one
partition with a physical Ethernet adapter connected to an external network can
act as router. Figure 1-17 shows a sample configuration. In this type of
configuration the partition that routes the traffic to the external work does not
necessarily have to be the Virtual I/O Server as in the pictured example. It could
be any partition with a connection to the outside world. The client partitions would
have their default route set to the partition which routes traffic to the external
network.

Figure 1-17 Connection to external network using AIX routing

Shared Ethernet Adapter
Using a Shared Ethernet Adapter (SEA) you can connect internal and external
VLANs using one physical adapter. The Shared Ethernet Adapter hosted in the
Virtual I/O Server acts as a layer 2 switch between the internal and external
network.
Shared Ethernet Adapter is a new service that acts as a layer 2 network bridge to
securely transport network traffic from a Virtual Ethernet to a real network
adapter. The Shared Ethernet Adapter service runs in the Virtual I/O Server. It
cannot be run in a general purpose AIX partition.

40


Shared Ethernet Adapter requires the POWER Hypervisor component of
POWER5 systems and therefore cannot be used on POWER4 systems. It also
cannot be used with AIX 5L Version 5.2 because the device drivers for Virtual
Ethernet are only available for AIX 5L Version 5.3 and Linux. Thus there is no
way to connect an AIX 5L Version 5.2 system to a Shared Ethernet Adapter.
The Shared Ethernet Adapter allows partitions to communicate outside the
system without having to dedicate a physical I/O slot and a physical network
adapter to a client partition. The Shared Ethernet Adapter has the following
characteristics:
Virtual Ethernet MAC addresses are visible to outside systems.
Broadcast/multicast is supported.
ARP and NDP can work across a shared Ethernet.
In order to bridge network traffic between the Virtual Ethernet and external
networks, the Virtual I/O Server has to be configured with at least one physical
Ethernet adapter. One Shared Ethernet Adapter can be shared by multiple
VLANs, and multiple subnets can connect using a single adapter on the Virtual
I/O Server. Figure 1-18 shows a configuration example. A Shared Ethernet
Adapter can include up to 16 Virtual Ethernet adapters that share the physical
access.

Figure 1-18 Shared Ethernet Adapter configuration

A Virtual Ethernet adapter connected to the Shared Ethernet Adapter must have
the trunk flag set. Once an Ethernet frame is sent from the Virtual Ethernet
adapter on a client partition to the POWER Hypervisor, the POWER Hypervisor
searches for the destination MAC address within the VLAN. If no such MAC


41

address exists within the VLAN, it forwards the frame to the trunk Virtual Ethernet
adapter that is defined on the same VLAN. The trunk Virtual Ethernet adapter
enables a layer 2 bridge to a physical adapter.
The shared Ethernet directs packets based on the VLAN ID tags. It learns this
information based on observing the packets originating from the virtual adapters.
One of the virtual adapters in the Shared Ethernet adapter is designated as the
default PVID adapter. Ethernet frames without any VLAN ID tags are directed to
this adapter and assigned the default PVID.
When the shared Ethernet receives IP (or IPv6) packets that are larger than the
MTU of the adapter that the packet is forwarded through, either IP fragmentation
is performed and the fragments forwarded or an ICMP packet too big message is
returned to the source when the packet cannot be fragmented.
Theoretically, one adapter can act as the only contact with external networks for
all client partitions. Depending on the number of client partitions and the network
load they produce, performance can become a critical issue. Because the
Shared Ethernet Adapter is dependant on Virtual I/O, it consumes processor
time for all communications. A significant amount of CPU load can be generated
by the use of Virtual Ethernet and Shared Ethernet Adapter.
There are several different ways to configure physical and virtual Ethernet
adapters into Shared Ethernet Adapters to maximize throughput.
Using Link Aggregation (EtherChannel), several physical network adapters
can be aggregated. See 1.5.2, “Using link aggregation (EtherChannel) to
external networks” on page 43 for more details.
Using several Shared Ethernet Adapters provides more queues and more
performance. An example for this configuration is shown in Figure 1-19 on
page 43.
Other aspects which have to be taken into consideration are availability and the
possibility to connect to different networks.

42


Figure 1-19 Multiple Shared Ethernet Adapter configuration

1.5.2 Using link aggregation (EtherChannel) to external networks
Link aggregation is network port aggregation technology that allows several
Ethernet adapters to be aggregated together to form a single pseudo-Ethernet
device. This technology can be used to overcome the bandwidth limitation of a
single network adapter and avoid bottlenecks when sharing one network adapter
among many client partitions.
For example, ent0 and ent1 can be aggregated to ent3. Interface en3 would then
be configured with an IP address. The system considers these aggregated
adapters as one adapter. Therefore, IP is configured as on any other Ethernet
adapter. In addition, all adapters in the link aggregation are given the same
hardware (MAC) address, so they are treated by remote systems as though they
were one adapter. The main benefit of link aggregation is that the network
bandwidth of all of the adapters is in a single network presence. If an adapter
fails, the packets are automatically sent on the next available adapter without
disruption to existing user connections. The adapter is automatically returned to
service on the link aggregation when it recovers.
You can use EtherChannel (EC) or IEEE 802.3ad Link Aggregation (LA) to
aggregate network adapters. While EC is an AIX-specific implementation of
adapter aggregation, LA follows the IEEE 802.3ad standard. Table 1-5 on
page 44 shows the main differences between EC and LA.


43

Table 1-5 Main differences between EC and LA aggregation
EtherChannel

IEEE 802.3ad Link Aggregation

Requires switch configuration.

Little, if any, configuration of switch
required to form aggregation. Some initial
setup of the switch may be required.

Supports different packet distribution
modes.

Supports only standard distribution mode.

The main benefit of using LA is, that if the switch supports the Link Aggregation
Control Protocol (LACP) no special configuration of the switch ports is required.
The benefit of EC is the support of different packet distribution modes. This
means it is possible to influence the load balancing of the aggregated adapters.
In the remainder of this document, we use Link Aggregation where possible
since that is considered a more universally understood term.
Note: Only outgoing packets are subject to the following discussion; incoming
packets are distributed by the Ethernet switch.
Standard distribution mode selects the adapter for the outgoing packets by
algorithm. The adapter selection algorithm uses the last byte of the destination IP
address (for TCP/IP traffic) or MAC address (for ARP and other non-IP traffic).
Therefore all packets to a specific IP address will always go through the same
adapter. There are other adapter selection algorithms based on source,
destination, or a combination of source and destination ports available. EC
provides one further distribution mode called round robin. This mode will rotate
through the adapters, giving each adapter one packet before repeating. The
packets may be sent out in a slightly different order than they were given to the
EC. It will make the best use of its bandwidth, but consider that it also introduces
the potential for out-of-order packets at the receiving system. This risk is
particularly high when there are few, long-lived, streaming TCP connections.
When there are many such connections between a host pair, packets from
different connections could be intermingled, thereby decreasing the chance of
packets for the same connection arriving out-of-order.
To avoid the loss of network connectivity by switch failure, EC and LA can
provide a backup adapter. The backup adapter should be connected to a
different switch than the adapter of the aggregation. Now in case of switch failure
the traffic can be moved with no disruption of user connections to the backup
adapter.
Figure 1-20 on page 45 shows the aggregation of three plus one adapters to a
single pseudo-Ethernet device including a backup feature.

44


Figure 1-20 Link Aggregation (EtherChannel) pseudo device

You must consider the following limitations when implementing Shared Ethernet
Adapters in the Virtual I/O Server:
Because Shared Ethernet Adapter depends on Virtual Ethernet, which uses
the system processors for all communications functions, a significant amount
of system processor load can be generated by the use of Virtual Ethernet and
Shared Ethernet Adapter.
One of the virtual adapters in the Shared Ethernet Adapter on the Virtual I/O
Server must be defined as the default adapter with a default PVID. This virtual
adapter is designated as the PVID adapter and Ethernet frames without any
VLAN ID tags are assigned the default PVID and directed to this adapter.
Up to 16 Virtual Ethernet adapters with 21 VLANs (20 VID and 1 PVID) on
each can be shared on a single physical network adapter. There is no limit on
the number of partitions that can attach to a VLAN, so the theoretical limit is
very high. In practice, the amount of network traffic will limit the number of
clients that can be served through a single adapter.
For performance and latency related information refer to Advanced POWER
Virtualization on IBM Sserver p5 Servers Architecture and Performance
Considerations, SG24-5768.


45

1.6 Virtual SCSI introduction
Available at the time of writing is the first support of Virtual I/O, which pertains to
a virtualized implementation of the SCSI protocol.
Virtual SCSI requires POWER5 hardware with the Advanced POWER
Virtualization feature activated. It provides Virtual SCSI support for AIX 5L
Version 5.3 and Linux.
The driving forces behind virtual I/O are:
The advanced technological capabilities of today’s hardware and operating
systems like POWER5 and IBM AIX 5L Version 5.3.
The value proposition enabling on demand computing and server
consolidation. Virtual I/O also provides a more economical I/O model by using
physical resources more efficiently through sharing.
At the time of writing, the virtualization features of the POWER5 platform support
up to 254 partitions, while the server hardware only provides up to 160 I/O slots
per machine. With each partition typically requiring one I/O slot for disk
attachment and another one for network attachment, this puts a constraint on the
number of partitions. To overcome these physical limitations, I/O resources have
to be shared. Virtual SCSI provides the means to do this for SCSI storage
devices.
Furthermore, virtual I/O allows attachment of previously unsupported storage
solutions. As long as the Virtual I/O Server supports the attachment of a storage
resource, any client partition can access this storage by using Virtual SCSI
adapters.
For example, if there is no native support for EMC storage devices on Linux,
running Linux in a logical partition of a POWER5 server makes this possible.
A Linux client partition can access the EMC storage through a Virtual SCSI
adapter. Requests from the virtual adapters are mapped to the physical
resources in the Virtual I/O Server. Driver support for the physical resources is
therefore only needed in the Virtual I/O Server.

46


Note: You will see different terms in this publication that refer to the various
components involved with virtual SCSI. These terms vary depending on the
context. With SCSI, usually the terms initiator and target are used, so you
may see terms such as virtual SCSI initiator and virtual SCSI target. On the
HMC, the terms virtual SCSI server adapter and virtual SCSI client adapter
are used. Basically they refer to the same thing. When describing the
client/server relationship between the partitions involved in virtual SCSI, the
terms hosting partition (meaning the Virtual I/O Server) and hosted partition
(meaning the client partition) are used.
The terms Virtual I/O Server partition and Virtual I/O Server both refer to the
Virtual I/O Server. The terms are used interchangeably in this section.

1.6.1 Partition access to virtual SCSI devices
The following sections describe the virtual SCSI architecture and the protocols
used.

Virtual SCSI client and server architecture overview
Virtual SCSI is based on a client/server relationship. The Virtual I/O Server owns
the physical resources and acts as the server or, in SCSI terms, target device.
The logical partitions access the virtual SCSI resources provided by the Virtual
I/O Server as clients.
The virtual I/O adapters are configured using an HMC. The provisioning of virtual
disk resources is provided by the Virtual I/O Server.
Often the Virtual I/O Server is also referred to as the hosting partition and the
client partitions as hosted partitions.
Physical disks owned by the Virtual I/O Server can either be exported and
assigned to a client partition whole, or can be partitioned into several logical
volumes. The logical volumes can then be assigned to different partitions.
Therefore, Virtual SCSI enables sharing of adapters as well as disk devices.
To make a physical or a logical volume available to a client partition it is assigned
to a virtual SCSI server adapter in the Virtual I/O Server.
The client partition accesses its assigned disks through a virtual SCSI client
adapter. The virtual SCSI client adapter sees standard SCSI devices and LUNs
through this virtual adapter. The commands in the following example show how
the disks appear on an AIX client partition.
# lsdev -Cc disk -s vscsi
hdisk2 Available Virtual SCSI Disk Drive


47

# lscfg -vpl hdisk2
hdisk2 111.520.10DDEDC-V3-C5-T1-L810000000000 Virtual SCSI Disk Drive

Figure 1-21 shows an example where one physical disk is partitioned into two
logical volumes inside the Virtual I/O Server. Each of the two client partitions is
assigned one logical volume which it accesses through a virtual I/O adapter
(vSCSI Client Adapter). Inside the partition the disk is seen as normal hdisk.

Figure 1-21 Virtual SCSI architecture overview

SCSI Remote Direct Memory Access
The SCSI family of standards provides many different transport protocols that
define the rules for exchanging information between SCSI initiators and targets.
Virtual SCSI uses the SCSI RDMA Protocol (SRP) which defines the rules for
exchanging SCSI information in an environment where the SCSI initiators and
targets have the ability to directly transfer information between their respective
address spaces.
SCSI requests and responses are sent using the virtual SCSI adapters that
communicate through the POWER Hypervisor.
The actual data transfer, however, is done directly between a data buffer in the
client partition and the physical adapter in the Virtual I/O Server using the Logical
Remote Direct Memory Access (LRDMA) protocol.
Figure 1-22 shows how the data transfer using LRDMA appears.

48


Figure 1-22 Logical Remote Direct Memory Access

AIX device configuration for virtual SCSI
The virtual I/O adapters are connected to a virtual host bridge which AIX treats
much like a PCI host bridge. It is represented in the ODM as a bus device whose
parent is sysplanar0. The virtual I/O adapters are represented as adapter
devices with the virtual host bridge as their parent.
On the Virtual I/O Server, each logical volume or physical volume that is
exported to a client partition is represented by a virtual target device that is a
child of a Virtual SCSI server adapter.
On the client partition, the exported disks are visible as normal hdisks, but they
are defined in subclass vscsi. They have a virtual SCSI client adapter as parent.
Figure 1-23 and Figure 1-24 on page 50 show the relationship of the devices
used by AIX for virtual SCSI and their physical counterparts.


49

Figure 1-23 Virtual SCSI device relationship on Virtual I/O Server

Figure 1-24 Virtual SCSI device relationship on AIX client partition

50


Dynamic partitioning for virtual SCSI devices
Virtual SCSI resources can be assigned and removed dynamically. On the HMC,
Virtual SCSI target and server adapters can be assigned and removed from a
partition using dynamic logical partitioning.
The mapping between physical and virtual resources on the Virtual I/O Server
can also be done dynamically.

The following areas should be considered when implementing virtual SCSI:
At the time of writing virtual SCSI supports Fibre Channel, parallel SCSI, and
SCSI RAID devices. Other protocols such as SSA or tape and CD-ROM
devices are not supported.
Virtual SCSI itself does not have any limitations in terms of number of
supported devices or adapters. However, the Virtual I/O Server supports a
maximum of 65535 virtual I/O slots. A maximum of 256 virtual I/O slots can be
assigned to a single partition.
Every I/O slot needs some resources to be instantiated. Therefore, the size of
the Virtual I/O Server puts a limit to the number of virtual adapters that can be
configured. For details see Advanced POWER Virtualization on IBM
Sserver p5 Servers Architecture and Performance Considerations,
SG24-5768.
The SCSI protocol defines mandatory and optional commands. While virtual
SCSI supports all the mandatory commands, not all optional commands are
supported.
There are performance implications when using virtual SCSI devices. It is
important to understand that, due to the overhead associated with POWER
Hypervisor calls, virtual SCSI will use additional CPU cycles when processing
I/O requests. When putting heavy I/O load on virtual SCSI devices, this
means you will use considerably more CPU cycles. Provided that there is
sufficient CPU processing capacity available the performance of virtual SCSI
should be comparable to dedicated I/O devices.
Suitable applications for virtual SCSI include boot disks for the operating
system or Web servers which will typically cache a lot of data. When
designing a virtual I/O configuration, performance is an important aspect
which should be given careful consideration. For a more in-depth discussion
of performance issues see Advanced POWER Virtualization on IBM
Sserver p5 Servers Architecture and Performance Considerations,
SG24-5768.


51

1.7 Partition Load Manager introduction
The Partition Load Manager (PLM) software is part of the Advanced POWER
Virtualization feature and helps customers maximize the utilization of processor
and memory resources of DLPAR-capable logical partitions running AIX 5L on
pSeries servers.
The Partition Load Manager is a resource manager that assigns and moves
resources based on defined policies and utilization of the resources. PLM
manages memory, both dedicated processors and partitions using
Micro-Partitioning technology, to readjust the resources. This adds additional
flexibility on top of the micro-partitions flexibility added by the POWER
Hypervisor.
PLM, however, has no knowledge about the importance of a workload running in
the partitions and cannot readjust priority based on the changes in types of
workloads. PLM does not manage Linux and i5/OS partitions. Figure 1-25 shows
a comparison of features between PLM and the POWER Hypervisor.
PLM Differentiation

Capability

PLM

OS Support

POWER4 PLM automates DLPAR adjustment for P4 install base

X

POWER5

X

AIX 5.2 PLM runs on AIX 5.2 on P4 and P5 systems (through PRPQ)

HW Support

X

AIX 5.3

X

P5
PHYP

X

X
X

pLinux
Dedicated PLM runs on AIX 5.2 and/or P4 systems

X

Capped shared

X

Uncapped shared

X

Virtual Processor
Management

Virtual processor minimization for efficiency

X

Virtual processor adjustment for physical processor growth

X

Physical Memory
Management

Share-based

X

Minimum and maximum entitlements

X

Entitlement-based

X

Physical Processor
Management

Management Policy

X

X

Goal-based
Application/middleware instrumentation required

Management Domains

Multiple management domains on a single CEC
Cross platform (CEC)

X

Simple administration
Administration

Centralized LPAR monitoring

(PLM command provides usage stats)

TOD-driven policy adjustment (PLM command supports new policy load based as TOD)

Figure 1-25 Comparison of features of PLM and POWER Hypervisor

52

X


X
X

X

Partition Load Manager is set up in a partition or on another system running AIX
5L Version 5.2 ML4 or AIX 5L Version 5.3. Linux or i5OS support for PLM and
the clients is not available. You can have other installed applications on the
partition or system running the Partition Load Manager as well. A single instance
of the Partition Load Manager can only manage a single server.
You can use the command line interface to configure Partition Load Manager, or
the Web-based System Manager for graphical setup.
Partition Load Manager uses a client/server model to report and manage
resource utilization. The clients (managed partitions) notify the PLM server when
resources are either under- or over-utilized. Upon notification of one of these
events, the PLM server makes resource allocation decisions based on a policy
file defined by the administrator.
Partition Load Manager uses the Resource Monitoring and Control (RMC)
subsystem for network communication, which provides a robust and stable
framework for monitoring and managing resources. Communication with the
HMC to gather system information and execute commands PLM requires a
configured SSH connection.
Figure 1-26 shows an overview of the components of Partition Load Manager.

Figure 1-26 Partition Load Manager overview


53

The policy file defines managed partitions, their entitlements and their
thresholds, and organizes the partitions into groups. Every node managed by
PLM must be defined in the policy file along with several associated attribute
values:
Optional maximum, minimum, and guaranteed resource values
The relative priority or weight of the partition
Upper and lower load thresholds for resource event notification
For each resource (processor and memory), the administrator specifies an upper
and a lower threshold for which a resource event should be generated. You can
also choose to manage only one resource.
Partitions that have reached an upper threshold become resource requesters.
Partitions that have reached a lower threshold become resource donors. When a
request for a resource is received, it is honored by taking resources from one of
three sources when the requester has not reached its maximum value:
A pool of free, unallocated resources
A resource donor
A lower priority partition with excess resources over entitled amount
As long as there are resources available in the free pool, they will be given to the
requester. If there are no resources in the free pool, the list of resource donors is
checked. If there is a resource donor, the resource is moved from the donor to
the requester. The amount of resource moved is the minimum of the delta values
for the two partitions, as specified by the policy. If there are no resource donors,
the list of excess users is checked.
When determining if resources can be taken from an excess user, the weight of
the partition is determined to define the priority. Higher priority partitions can take
resources from lower priority partitions. A partition's priority is defined as the ratio
of its excess to its weight, where excess is expressed with the formula (current
amount - desired amount) and weight is the policy-defined weight. A lower value
for this ratio represents a higher priority. Figure 1-27 on page 55 shows an
overview of the process for partitions.

54


Figure 1-27 PLM resource distribution for partitions

In Figure 1-27, all partitions are capped partitions. LPAR3 is under heavy load
and over its high CPU average threshold value for becoming a requestor. There
are no free resources in the free pool and no donor partitions available. PLM now
checks the excess list to find a partition having resources allocated over its
guaranteed value and with a lower priority. Calculating the priority, LPAR1 has
the highest ratio number and therefore the lowest priority. PLM deallocates
resources from LPAR1 and allocates them to LPAR3.
If the request for a resource cannot be honored, it is queued and re-evaluated
when resources become available. A partition cannot fall below its minimum or
rise above its maximum definition for each resource.
The policy file, once loaded, is static, and has no knowledge of the nature of the
workload on the managed partitions. A partition's priority does not change upon
the arrival of high priority work. The priority of partitions can only be changed by
some action, external to PLM, loading a new policy.
Partition Load Manager handles memory and both types of processor partitions:
dedicated and shared processor partitions. All the partitions in a group must be
of the same processor type.


55

1.7.1 Memory management
PLM manages memory by moving Logical Memory Blocks (LMBs) across
partitions. To determine when there is demand for memory, PLM uses two
metrics:
Utilization percentage (ratio of memory in use to available)
The page replacement rate
For workloads that result in significant file caching, the memory utilization on AIX
may never fall below the specified lower threshold. With this type of workload, a
partition may never become a memory donor, even if the memory is not currently
being used.
In the absence of memory donors, PLM can only take memory from excess
users. Since the presence of memory donors cannot be guaranteed, and is
unlikely with some workloads, memory management with PLM may only be
effective if there are excess users present. One way to ensure the presence of
excess users is to assign each managed partition a low guaranteed value, such
that it will always have more than its guaranteed amount. With this sort of policy,
PLM will always be able to redistribute memory to partitions based on their
demand and priority.

1.7.2 Processor management
For dedicated processor partitions, PLM moves physical processors, one at a
time, from partitions that are not utilizing them to partitions that have demand for
them. This enables dedicated processor partitions running AIX 5L Version 5.2
and AIX 5L Version 5.3 to better utilize their resources. If one partition needs
more processor capacity, PLM automatically moves processors from a partition
that has idle capacity.
For shared processor partitions, PLM manages the entitled capacity and the
number of virtual processors (VPs) for capped or uncapped partitions. When a
partition has requested more processor capacity, PLM will increase the entitled
capacity for the requesting partition if additional processor capacity is available.
For uncapped partitions, PLM can increase the number of virtual processors to
increase the partition's potential to consume processor resources under high
load conditions. Conversely, PLM will also decrease entitled capacity and the
number of virtual processors under low-load conditions, to more efficiently utilize
the underlying physical processors.

56


With the goal of maximizing a partition's and the system's ability to consume
available processor resources, the administrator now has two choices:
1. Configure partitions that have high workload peaks as uncapped partitions
with a large number of virtual processors. This has the advantage of allowing
these partitions to consume more processor resource when it is needed and
available, with very low latency and no dynamic reconfiguration. For example,
consider a 16-way system utilizing two highly loaded partitions configured
with eight virtual processors each, in which case, all physical processors
could have been fully utilized. The disadvantage of this approach is that when
these partitions are consuming at or below their desired capacity there is an
overhead associated with the large number of virtual processors defined.
2. Use PLM to vary the capacity and number of virtual processors for the
partitions. This has the advantages of allowing partitions to consume all of the
available processor resource on demand, and it maintains a more optimal
number of VPs. The disadvantage to this approach is that since PLM
performs dynamic reconfiguration operations to shift capacity to and from
partitions, there is a much higher latency for the reallocation of resources.
Though this approach offers the potential to more fully utilize the available
resource in some cases, it significantly increases the latency for redistribution
of available capacity under a dynamic workload, since dynamic
reconfiguration operations are required.

You must consider the following limitations when managing your system with the
Partition Load Manager:
The Partition Load Manager can be used in partitions running AIX 5L Version
5.2 ML4 or AIX 5L Version 5.3. Linux or i5OS support is not available.
A single instance of the Partition Load Manager can only manage a single
server. However, multiple instances of the Partition Load Manager can be run
on a single system, each managing a different server.
The Partition Load Manager cannot move I/O resources between partitions.
Only processor and memory resources can be managed by Partition Load
Manager.
Partition Load Manager requires HMC Release 3 Version 2.6 or newer on an
HMC and an IBM Sserver p5 system.


57

58


2

Chapter 2.

Application development
AIX 5L provides several enhancements that assist you in developing your own
software. Among the enhancements described in this chapter are:
POSIX Realtime compliant features
Enhancements to memory allocation
Marking of executable read/write sections
Java™ enhancements


59

2.1 POSIX Realtime functions
One standard that has been included in AIX 5L Version 5.3 is the POSIX
REALTIME extension. The options available within this standard are described in
the section on System Interfaces, Issue 6, 2003 IEEE and the Open Group book,
available on the following Web site:
http://www.opengroup.org

AIX 5L Version 5.3 provides system interfaces for the following options:
Memlock
Spin locks
Clocks
Priority scheduling
Shared memory objects
Semaphores
Timers
Barriers
Thread options
Message passing
Advisory info
These options are discussed in detail in the sections that follow.

2.1.1 Memlock
The Memlock option uses POSIX MEMLOCK and POSIX MEMLOCK RANGE
interfaces to provide a POSIX-compliant implementation for the memory locking
and unlocking function. These interfaces allow repeated locking and unlocking of
memory within the lifetime of a process to support applications that undergo
major mode changes where, in one mode, locking is required, but in another it is
not. The locking function guarantees the residence in memory of all, or portions
of, a process address space including the text, data and stack areas, shared
libraries and shared memory areas.
Memory locking is used to eliminate the indeterminacy introduced by paging and
swapping.

60


AIX 5L Version 5.3 provides the following interfaces for memlock:
mlock()

Causes the whole pages containing any part of the address
space of the process starting at a specifiable address and
continuing for a specifiable number of bytes to be
memory-resident until unlocked or until the process exits or
executes another process image.

munlock()

Unlocks the whole pages containing any part of the address
space of the process starting at a specifiable address and
continuing for a specifiable number of bytes, regardless of how
many times mlock() has been called by the process for any of the
pages in the specified range.
If any of the pages in the range specified in a call to the
munlock() function are also mapped into the address spaces of
other processes, any locks established on those pages by
another process are unaffected by the call of this process to the
munlock() function. If any of the pages in the range specified by a
call to the munlock() function are also mapped into other portions
of the address space of the calling process outside the range
specified, any locks established on those pages through other
mappings are also unaffected by this call.

mlockall()

Causes all of the pages mapped by the address space of a
process to be memory-resident until unlocked or until the
process exits or executes another process image.

munlockall()

Unlocks all currently mapped pages of the address space of the
process. Any pages that become mapped into the address space
of the process after a call to the munlockall function are not
locked, unless there is an intervening call to the mlockall
function. If pages mapped into the address space of the process
are also mapped into the address spaces of other processes and
are locked by those processes, the locks established by the
other processes are unaffected by a call to the munlockall
function.

Note: Stacking calls to mlock() and mlockall() in a single process has the
same effect as a single call to one of these interfaces.
Stacking calls to munlock() and munlockall() in a single process has the same
effect as a single call to one of these interfaces.
These functions are available by default. They use the standard C library (libc.a).
There are no tunables to turn them on or off. However, the calling process must
have the root user authority to use these functions.

Chapter 2. Application development

61

In general, plock() and mlock() calls should not be mixed. In certain situations
they can run into conflict with each other. For instance:
It is forbidden to call the POSIX memory locking interfaces while a flag is set
in the U_lock field of the user structure, due to a call to plock(*LOCK).
It is forbidden to call plock() interface while a memory region is still locked due
to a call to mlockall() or mlock().

2.1.2 Spin locks
Spin locks are synchronization objects used to allow multiple threads to serialize
their access to shared data. They represent an extremely low-level
synchronization mechanism suitable primarily for use on shared memory
multi-processors. When a caller requests a spin lock that is already held, it
typically spins in a loop testing whether the lock has become available. Such
spinning wastes processor cycles. Spin locks provide an efficient locking
mechanism for short-duration locks.
Locking mechanisms typically must trade off processor resources consumed
while setting up to block the thread and the processor resources consumed by
the thread while it is blocked. Spin locks require very little resources to set up the
blocking of a thread compared to mutexes.
AIX 5L Version 5.3 provides the following interfaces for a POSIX compliant
implementation of POSIX SPIN LOCKS:
pthread_spin_destroy()

Destroy a spin lock object.
The pthread_spin_destroy function destroys the spin
lock and releases any resources used by the lock.

pthread_spin_init()

Initialize a spin lock object.
The pthread_spin_init function allocates any resources
required to use the spin lock and initializes the lock to
an unlocked state.

pthread_spin_lock()

Lock a spin lock object.

pthread_spin_trylock()

The pthread_spin_lock function locks the spin lock.
The calling thread shall acquire the lock if it is not held
by another thread. Otherwise, the thread spins (that is,
does not return from the pthread_spin_lock call) until
the lock becomes available. The pthread_spin_trylock
function locks the spin lock if it is not held by any
thread. Otherwise, the function fails.

62


pthread_spin_unlock()

The pthread_spin_unlock function releases the spin
lock referenced by the lock parameter which was
locked using the pthread_spin_lock function or the
pthread_spin_trylock function.

The result of referring to copies of a spin lock object in calls to
pthread_spin_destroy function, pthread_spin_lock function, pthread_spin_trylock
function, or the pthread_spin_unlock function is undefined.
You can use the dbx program to debug spinlocks. To do so, set the
AIXTHREAD_SPINLOCKS environment variable to ON.

2.1.3 Clocks
The IEEE standard defines functions for obtaining system time. Implicit behind
these functions is a mechanism for measuring the passage of time. The
specification makes this mechanism explicit and calls it a clock. The
CLOCK_REALTIME clock required by POSIX is a higher resolution version of
the clock that maintains system time. This is a system-wide clock in the sense
that it is visible to all processes.
The interface was extended to define additional clocks. This was done because
many realtime platforms support multiple clocks. However, such clocks do not
necessarily represent hardware devices nor are they necessarily system-wide.
POSIX_MONOTONIC_CLOCK, POSIX_CLOCK_SELECTION,
POSIX_CPUTIME, and POSIX_THREAD_CPUTIME provide a POSIX-compliant
implementation of the CLOCKS option as part of the POSIX Realtime support.
POSIX_MONOTONIC_CLOCK provides a clock whose value cannot be set
using clock_settime() and that cannot have backward clock jumps.
POSIX_CLOCK_SELECTION provides the following three interfaces:
clock_nanosleep()

Suspend current thread from execution until the time
specified has elapsed.

pthread_condattr_getclock() Get the clock selection condition variable attribute.
pthread_condattr_setclock() Set the clock selection condition variable attribute.
The clocks supported by these interfaces are:
The realtime clock, CLOCK_REALTIME
The monotonic clock, CLOCK_MONOTONIC


63

CPU-time clocks are not supported by these interfaces.
POSIX_CPUTIME provides access to a process CPU-time clock that can be
invoked to measure execution time of a process using the clock_getcpuclockid()
interface.
POSIX_THREAD_CPUTIME provides support for a thread CPU-time clock that
measures execution time of threads with the pthread_getcpuclockid() interface. It
is a thread-wide clock rather than a process- or system-wide clock.

2.1.4 Priority scheduling
Realtime applications require that process scheduling be fast and deterministic:
the highest-priority process is always run first and, among processes of equal
priority, the process that has been runnable for the longest time is executed first.
POSIX PRIORITY SCHEDULING option provides a POSIX-compliant
implementation of POSIX process scheduling extensions.
POSIX PRIORITY SCHEDULING provides the following interfaces:
sched_setparam()

Sets the scheduling parameters of a process.

sched_getparam()

Returns the scheduling parameters of a process.

sched_setscheduler()

Sets the scheduling policy and parameters of a
process.

sched_getscheduler()

Returns the scheduling policy of a process.

sched_get_priority_min()

Returns the minimum priority value for a scheduling
policy.

sched_get_priority_max()

Returns the maximum priority value for a scheduling
policy.

sched_rr_get_interval()

Returns the execution time limit of a process.

sched_yield()

Makes the calling thread yield the processor.

It should be noted that AIX 5L Version 5.2 already conforms to POSIX Realtime
Process Scheduling option to a large extent. Version 5.3 fills in the gaps.

2.1.5 Shared memory objects
Shared memory is used to share data among several processes. For virtual
memory systems, a shared memory interface is required to prevent processes
from accessing each other’s data. However, in unprotected systems such as are
found in embedded controllers, a shared memory interface is needed to provide
a portable mechanism to allocate a region of memory to be shared and then to

64


communicate the address of that region to other processes. Therefore, to
support realtime applications, a shared memory interface must be able to
allocate and name an object to be mapped into memory for potential sharing and
make the memory object available within the address space of a process. The
interface must also provide a mechanism for deleting the name of the sharable
object that was previously created.
POSIX SHARED MEMORY OBJECTS option fulfils these requirements by
means of the following interfaces:
shm_open()

Establish a connection between a shared memory object
and a file descriptor. If it does not already exist, first
create the object.

shm_unlink()

Remove a shared memory object.

2.1.6 Semaphores
A realtime system requires a robust synchronization mechanism between the
processes within the same overall application. Such synchronization has to allow
more than one scheduled process mutually-exclusive access to the same
resource. POSIX SEMAPHORES option provides a POSIX-compliant
implementation for the POSIX semaphores which offer such a synchronization
mechanism for realtime applications. The POSIX standard for realtime
inter-process communication describes two types of semaphores: named
semaphores and unnamed semaphores.
POSIX SEMAPHORES provides the following interfaces:
sem_close()

Close a named semaphore.

sem_destroy()

Destroy an unnamed semaphore.

sem_getvalue()

Get the value of a semaphore.

sem_init()

Initialize an unnamed semaphore.

sem_open()

Establish a connection between a named semaphore and
a process.

sem_post()

Unlock a semaphore.

sem_trywait()

Lock the semaphore only if the semaphore is currently not
locked.

sem_unlink()

Remove a named semaphore.

sem_wait()

Lock the semaphore by performing a semaphore lock
operation on that semaphore.

sem_timedwait()

Lock a semaphore within a specified timeout. If the
semaphore cannot be locked without waiting for another


65

process or thread to unlock the semaphore by performing
a sem_post() function, this wait shall be terminated when
the specified timeout expires.

2.1.7 Timers
Two types of timers are required for a system to support realtime applications.
One-shot

A one-shot timer is a timer that is armed with an initial
expiration time, either relative to the current time or at an
absolute time. The timer expires once and then is
disarmed.

Periodic

A periodic timer is a timer that is armed with an initial
expiration time, either relative or absolute, and a
repetition interval. When the initial expiration occurs, the
timer is reloaded with the repetition interval and continues
counting.

POSIX TIMERS option provides a POSIX-compliant implementation for the
timers extension. This option relies for timer management not only on the real
time clock, but also on the other clocks defined in other POSIX Extensions.
POSIX TIMERS provides the following interfaces:
clock_getres()
clock_gettime()

Returns the time as measured by a specific clock.

clock_settime()

Sets the time of a specific clock.

nanosleep()

Sleeps for an amount of time as measured by the real
time clock as a relative interval.

timer_create()

Creates a per-process timer using a specific clock as the
timing base.

timer_delete()

Deletes a per-process timer previously created by
timer_create.

timer_getoverrun()

Returns the timer overrun count for a timer previously
created by timer_create.

timer_gettime()

Returns the amount of time until a timer previously
created by timer_create expires, as well as the reload
value of the timer.

timer_settime()

66

Returns the resolution of a specific clock.

Sets the time until next expiration of a timer and its reload
value.


2.1.8 Barriers
POSIX BARRIERS provide a POSIX-compliant implementation for the POSIX
barriers option. They are used for pthreads synchronization purposes and are
typically used in parallel DO/FOR loops to ensure that all threads have reached a
particular stage in a parallel computation before allowing any to proceed to the
next stage. Highly efficient implementation of barriers is possible on machines
which support a Fetch and Add operation. Although they can be implemented
with mutexes and condition variables, such implementations of barriers are
significantly less efficient than is possible.
POSIX BARRIERS provide the following interfaces:
pthread_barrier_init()

Initialize a barrier object.

pthread_barrier_destroy()

Destroy a barrier object.

pthread_barrierattr_init()

Initialize the barrier attributes object.

pthread_barrierattr_destroy()

Destroy the barrier attributes object.

pthread_barrierattr_setpshared() Set the process-shared attribute of the barrier
attributes object.
pthread_barrierattr_getpshared() Get the process-shared attribute of the barrier
attributes object.
pthread_barrier_wait()

Synchronize at a barrier.

The symbol _POSIX_BARRIERS is defined to 200112L in <unistd.h>.
Barriers can be debugged by setting environment variable
AIXTHREAD_BARRIER_DEBUG to ON.

2.1.9 Thread options
POSIX THREAD PRIORITY SCHEDULING, POSIX THREAD PRIO PROTECT,
and POSIX THREAD PRIO INHERIT options provide a POSIX implementation of
these options as part of POSIX Realtime support. This option also provides
additional thread interfaces from the POSIX TIMEOUTS option.
POSIX THREAD PRIORITY SCHEDULING provides the following interfaces:
pthread_attr_getinheritsched() Get the inheritsched attribute.
pthread_attr_setinheritsched() Set the inheritsched attribute.
pthread_attr_getschedpolicy() Get the schedpolicy attribute.
pthread_attr_setschedpolicy()

Set the schedpolicy attribute.

pthread_attr_getscope()

Get the contentionscope attribute.


67

pthread_attr_setscope()

Set the contentionscope attribute.

pthread_getschedparam()

Get the current schedparam attributes of a
thread.

pthread_setschedparam()

Set the current schedparam attributes of a
thread.

pthread_setschedprio()

Set the scheduling priority for the thread.

POSIX THREAD PRIO PROTECT (and INHERIT) provide the following
interfaces:
pthread_mutex_getprioceiling()

Get the priority ceiling of a mutex.

pthread_mutex_setprioceiling()

Set the priority ceiling of a mutex.

pthread_mutexattr_getprioceiling() Get the prioceiling attribute of the mutex
attributes object.
pthread_mutexattr_setprioceiling() Set the prioceiling attribute of the mutex
attributes object.
pthread_mutexattr_getprotocol()

Get the protocol attribute of the mutex
attributes object.

pthread_mutexattr_setprotocol()

Set the protocol attribute of the mutex
attributes object.

POSIX TIMEOUTS provide the following interfaces:
pthread_mutex_timedlock()

Locks a mutex, with a timed wait if the mutex
is already locked.

pthread_rwlock_timedrdlock()

Locks a read-write lock for reading.

pthread_rwlock_timedwrlock()

Locks a read-write lock for writing.

2.1.10 Message passing
Many applications, including both realtime and database applications, require a
means of passing data between processes of the same application, executing on
one or more processors. Most conventional interfaces for inter-process
communication are insufficient for such data passing for realtime applications.
POSIX MESSAGE PASSING option provides interfaces that enable such data
passing. It allows processes to communicate through system-wide queues.
POSIX MESSAGE PASSING provides the following interfaces:
mq_close()
mq_getattr()

68

Close a message queue.
Get message queue attributes.


mq_notify()

Register the calling process to be notified of message
arrival at an empty message queue.

mq_open()

Establish a connection between a process and a
message queue.

mq_receive()

Receive the oldest of the highest priority messages from
the message queue.

mq_send()

Send a message to a message queue.

mq_setattr()

Set message queue attributes.

mq_timedreceive()

Receive a message from a message queue within a
specified timeout.

mq_timedsend()

Send a message to a message queue within a specified
timeout.

mq_unlink()

Remove a message queue.

2.1.11 Advisory Info
The purpose of the Advisory Info option is to provide a method for an application
to advise the OS on its (the application's) future behavior with respect to a given
file. In addition, it provides a way for the application to pre-allocate certain
resources from the OS as well as specify the alignment of memory allocated
from the system.
It provides the following interfaces:
posix_fadvise()

Advises the system on the expected future behavior of the
application with regards to a given file. The system may
take this advice into account when performing operations
on file data specified by this function.

posix_fallocate()

Reserves storage space on the file system media for a
given File Descriptor.

posix_memalign()

Allocates memory with specified alignment.

posix_madvise()

Advises the system on the expected future behavior of the
application with regard to a given range of memory. The
system may take this advice into account when
performing operations on the data in memory specified by
this function.

The posix_fallocate() and posix_memalign() functions actually define behaviors
while the advise functions simply provide the OS the opportunity to optimize
future behaviors.


69

Advice must be one of the following values specified in fcntl.h:
POSIX_FADV_NORMAL
POSIX_MADV_NORMAL
POSIX_FADV_SEQUENTIAL
POSIX_MADV_SEQUENTIAL
POSIX_FADV_WILLNEED
POSIX_MADV_WILLNEED
POSIX_FADV_RANDOM
POSIX_MADV_RANDOM
POSIX_FADV_DONTNEED
POSIX_MADV_DONTNEED
POSIX_FADV_NOREUSE
POSIX_FADV_NORMAL and POSIX_MADV_NORMAL mean the application
has no advice to give on its behavior with respect to the specified data. It is the
default characteristic if no advice is given for an open file.
The POSIX_FADV_SEQUENTIAL and POSIX_MADV_SEQUENTIAL advice
tells the OS to expect serial access. Typically the system will prefetch the next
several serial accesses in order to overlap I/O. It may also free previously
accessed serial data if memory is tight. If the application is not doing serial
access it can use POSIX_FADV_WILLNEED and POSIX_MADV_WILLNEED to
accomplish I/O overlap, as required. When the application advises
POSIX_FADV_RANDOM or POSIX_MADV_RANDOM behavior, the OS usually
tries to fetch a minimum amount of data with each request and it does not expect
much locality. POSIX_FADV_DONTNEED and POSIX_MADV_DONTNEED
allow the system to free up caching resources since the data will not be required
in the near future.
POSIX_FADV_NOREUSE tells the system that caching the specified data is not
optimal because the application expects to access the specified data once and
then not reuse it thereafter.

2.2 Enhanced libc.a
AIX 5L Version 5.3 has implemented a number of new APIs for the libc.a library.
A number of existing functions in this library have been modified as well. This has
been done to improve system performance, reliability, and servicability for the
identification and authentication services provided by this library.

70


The following list describes new and modified APIs. Functions ending in the letter
x do everything that their old counterparts did and more, as described here.
authenticatex()

Introduces the ability to communicate authentication
method information between library routines.

chpassx()

Introduces the ability to communicate authentication
information about the mechanisms which were actually
used during the authentication process.

loginrestrictionsx()

Introduces the ability to examine all methods used in an
authentication decision and query all of those methods to
verify that all of the methods grant the user access.

passwdexpiredx()

Introduces the ability to examine all methods used in an
authentication decision and query all of those methods to
determine which passwords have expired, and then
update those passwords using chpassx().

getgroupattrs()

Allows an application to request one or more group
attributes with a single library call.

getuserattrs()

Allows an application to request one or more user

putgroupattrs()

Allows an application to modify (update) one or more
group attributes with a single library call.

putuserattrs()

Allows an application to modify (update) one or more user

getgroupattr()

This function has been modified to make a single call to
getgroupattrs() using the supplied name, attribute, and
type information.

getuserattr()

getuserattrs() using the supplied name, attribute, and type
information.

putgroupattr()

putgroupattrs() using the supplied name, attribute, and
type information.

putuserattr()

putuserattrs() using the supplied name, attribute, and type
information.

usersec.h

This header file has been updated to include function
prototypes for each of the new functions listed above.


71

2.3 New malloc() algorithm
A new malloc allocator called Watson Malloc has been implemented in AIX 5L
Version 5.3. Watson Malloc provides improvement over the default allocator in
the areas of memory fragmentation and speed performance in massively
multi-threaded applications. In addition, two new options have been added for
the default allocator, a thread cache front end and a new buckets-based front
end.
The current default allocator is the Yorktown allocator. With respect to speed, the
Yorktown allocator does not efficiently handle repeated small to medium size
requests. This deficiency was previously addressed by adding the Malloc
Buckets algorithm. Malloc Buckets, however, provides no way to consolidate
freed memory into the heap. The new bucket allocator will allow freed memory to
be reclaimed. Through the use of the new bucket allocator, the Watson Allocator
handles small requests quickly and with comparatively little wasted memory. The
Watson Allocator also performs quicker than the default allocator, and with less
internal fragmentation than the old bucket allocator.
The Watson Allocator can be configured in three distinct ways to try and identify
which sections reveal the largest gains. It can be enabled with caching
mechanisms (a per thread cache and an adaptive heap cache) and with the new
bucket allocator. It can be enabled without the caching mechanisms and with the
new bucket allocator. It can also be enabled without the caching mechanisms
and without the new bucket allocator. The new bucket allocator and thread cache
have been adapted to work with the Yorktown allocator.
These enhancements lead to increased performance and reduced resource
overhead for large multithreaded applications. However, any of these
configurations can be disabled if they do not prove to be beneficial in regard to
speed or memory usage.
Only one MALLOCTYPE can be specified at a time. The MALLOCTYPE is
specified prior to process startup and cannot be changed for a running process
as in the following:
MALLOCTYPE=watson

2.4 Improvements to malloc subsystem
Currently user actions required to enable the malloc environment are quite
confusing. They are complicated by the fact that some of the environment
variables that the malloc subsystem supports often have conflicting purposes.
This situation has been simplified in AIX 5L Version 5.3 by reducing the number
of environment variables to three and redefining the attributes they can assume.

72


They are MALLOCTYPE, MALLOCOPTIONS, and MALLOCDEBUG.
MALLOCOPTIONS is a new environment variable that has been added to take
care of all current and future options to the MALLOCTYPE allocators. It
supplants the MALLOCBUCKETS, MALLOCMULTIHEAP, and
MALLOCDISCLAIM environment variables which have been deprecated since
they really are just algorithmic options which conceptually should not require a
separate environment variable.
The following are the definitions of these three environment variables and some
of the attributes they can assume:
MALLOCTYPE

Used to select allocator type. Attributes include: 3.1,
default allocator, Watson, user.

MALLOCOPTIONS

Allocator options. Attributes include: buckets, disclaim,
threadcache, multiheap.

MALLOCDEBUG

debugging options. Attributes include: catch_overflow,
report_allocations, Malloc Log, Malloc Trace,
validate_ptrs.

AIX 5L Version 5.3 has also implemented the following enhancements:
Threadcache option which reduces contention for heap lock is now available
for applications using the default (Yorktown) as well as the Watson Allocator.
Malloc Log and Malloc Trace functions have been enhanced. Malloc Log
offers an extended log and provides for automatic logging of allocations and
metadata. Malloc Trace now logs more data including heap and thread IDs.
The output debug option
Currently, all malloc debugging options which produce printed output send
their output to stderr. The output option provides a mechanism to instruct the
subsystem to send printed output to a different file stream as in the following:
$ MALLOCDEBUG=output:/dev/null
$ MALLOCDEBUG=output:stdout

The continue debug option
Currently, many malloc debugging options call the abort() procedure when
they encounter an error. It is often the case, however, that a developer may
wish to debug other classes of errors first and would prefer that less serious
errors do not produce fatal flaws. With the continue option enabled,
synchronous errors do not call the abort() function. Error messages are still
logged to the appropriate channels as in the following:
$ MALLOCDEBUG=continue


73

Watson allocator now supports the following debugging features:
catch_overflow This is a debugging option that attempts to catch writes and
reads past the end of a buffer. This is done by allocating an
extra guard page and positioning the allocation in such a
manner that the end of the allocation is flush against the
guard page. The guard page is then protected so that any
access will cause a segfault.
Malloc Log

This is a debugging option that creates a live database of
active malloc allocations that can be accessed through the
use of a debugger.

DBX malloc command
Malloc debugging features have been integrated into DBX. This would allow a
developer to query the current state of the malloc subsystem without resorting
to the creation of complex, unwieldy scripts requiring internal knowledge of
the malloc subsystem. An application developer will be able to get information
on such items as specific malloc allocations, organization of the heaps, and
the free space tree. These features are supported in both live processes and
core file dbx sessions. DBX malloc command works in tandem with Malloc
Log and allows key-based searches for such items as heap, address, size,
and PID.
The following are some examples:
(dbx) malloc prints out options, heap statistics, brk statistics
(dbx) malloc freespace prints out all free nodes in the process’s heaps

2.5 Block device mapping
A device is treated like an ordinary file in many respects, but previous AIX
releases never had the ability to map a device. AIX 5L Version 5.3 extends the
mapping capability of the mmap subroutine to block devices.
A block device is a special file that provides access to a device driver that
presents a block interface. A block interface to a device driver requires data
access in blocks of a fixed size. The interface is typically used for data storage
devices.
AIX 5L Version 5.3 utilizes block device mapping to improve performance in the
file system maintenance commands. The Version 5.3 JFS2 fsck command in
particular makes use of this new capability to reduce the access time to file
system meta data. Also, the control of the memory used for buffering is greatly
simplified by mapping the device that the file system resides on.

74


The user of block device mapping should be aware of the following
characteristics and restrictions:
All operations valid on memory resulting from mmap() of a file will be valid on
memory resulting from mmap() of a block device.
The mapped device is accessed using the device files in /dev. The mapping
function is limited to devices opened using the block special file.
The mmap subroutine allows mapping of a device by only one process at a
time.
The mapped device data cache will not be coherent with respect to read and
write system calls to the device. Nor will it be coherent with respect to the file
system cache when accessing a device that contains a mounted file system.

2.5.1 System calls and subroutines for block device mapping
The new block device mapping capability has the following impact on system
calls and subroutines:

open()
A program that wants to do device mapping will open the special file in the
normal way. No special flags or arguments are needed. All existing interfaces
and interactions with the file descriptor returned by open will continue to operate
as before.

mmap()
The mapping function is initiated by a call to mmap specifying the file descriptor
for the device. The mmap call is the same as for a regular file mapping. The
resultant mapping obeys the semantics of regular file mapping. If the file
descriptor refers to a device that has already been mapped, the mmap call will
fail and set errno to EAGAIN.

msync()
The data that is written to the mapped device segment can be flushed to disk
using the msync call. The flags and arguments used for regular files also apply to
devices.

disclaim()
The disclaim subroutine can be used to release pages in the mapped segment.


75

2.6 Eclipse Runtime Environment
Eclipse Runtime Environment is shipped with AIX 5L Version 5.3. This provides a
platform on which Eclipse-based tools, including the new procmon performance
monitoring tool, will run. AIX 5L does not support Eclipse as a development
platform.

2.7 Cryptographic sum command
Users often use the sum command to generate a checksum to verify the integrity
of a file. However, it is possible that two distinct files will generate the same
checksum. A cryptographic sum command, csum, has been implemented in AIX
5L Version 5.3 that offers a more reliable tool to verify file integrity. This
command allows users to generate message-digests using the AIX
Cryptographic Library. A cryptographic checksum is considered secure because
it is computationally infeasible to construct data to generate a known checksum,
and vice-versa.
The csum command allows users the option to select the algorithm they prefer,
including both MD5 and SHA-1, which are considered secure. It is estimated that
the order of 264 operations would be required to derive two different files which
generate the same MD5 message-digest and that the order of 2128 operations
would be needed to derive a file that would generate a specified MD5
message-digest.
csum will aid in improving the AIX e-fix upgrade process by offering users a
mechanism to verify that a file has not been tampered with or corrupted during
download.
The csum command is installed as part of the bos.rte.commands fileset.

2.8 Perl 5.8.2
The latest version of Perl that ships with AIX 5L is Perl 5.8.0, which was first
shipped on AIX 5.2.0. A number of reported problems in Perl 5.8.0 have been
fixed by the Perl community as well as the AIX Perl team. Perl 5.8.2, the latest
version of the code, is shipped with AIX 5L Version 5.3, as can be shown with the
following command:
# perl -v
This is perl, v5.8.2 built for aix-thread-multi
Copyright 1987-2003, Larry Wall

76


Perl may be copied only under the terms of either the Artistic License or the
GNU General Public License, which may be found in the Perl 5 source kit.
Complete documentation for Perl, including FAQ lists, should be found on this
system using `man perl' or `perldoc perl'. If you have access to the Internet,
point your browser at http://www.perl.com/, the Perl Home Page.

You can run the perldoc command to view the delta in changes between two
Perl versions. The first command line in the following displays the delta in
changes between Perl version 5.8.0 and 5.8.1 and the second one does the
same between Perl version 5.8.1 and 5.8.2:
$ perldoc perl581delta
$ perldoc perldelta

The Perl environment is packaged and shipped in two filesets: perl.rte and
perl.man.en_US. The perl.rte fileset is on the first AIX CD and is automatically
installed when installing AIX 5L Version 5.3 or when migrating from a previous
level of AIX.
The man pages fileset (perl.man.en_US) is shipped on the second CD and is not
automatically installed.
Useful information is documented in the /usr/lpp/perl.rte/README.perl.aix file.
Perl installs in /usr/opt/perl5. Links for the Perl executables in /usr/opt/perl5/bin to
/usr/bin have been added.
The 64-bit and 32-bit versions are packaged together, with the 32-bit version
being the default version. Both versions reside under the /usr/opt/perl5 directory.
Both versions are Perl thread capable, built using the newer ithreads (interpreter
threads) and also have built-in support for PerlIO.
This version of Perl also supports loading of large files into memory using the
maxdata linker option passed in at compile time.
The 64-bit version was built using both optimum 64-bit support along with 64-bit
integer support.
Also of interest is the libperl.a library located in the following places under
/usr/opt/perl5:
./lib/5.8.2/aix-thread-multi/CORE/libperl.a
./lib64/5.8.2/aix-thread-multi-64all/CORE/libperl.a
Switching to use the 64-bit version only requires redirecting the Perl symbolic
links in /usr/bin to point to the 64-bit versions of the same command in
/usr/opt/perl5/bin.


77

The /usr/opt/perl5 directory includes scripts you can use to change the links for
you.
To switch from 32-bit to 64-bit Perl, use the following script:
/usr/opt/perl5/link_perl_64

To switch from 64-bit to 32-bit Perl, use the following script:
/usr/opt/perl5/link_perl_32

Note: You must run these scripts as root for them to work.

2.9 SVR4 linking affinity
Nowadays, there are two major object file formats used by the operating
systems:
XCOFF The COFF enhanced by IBM and others. The original COFF (Common
Object FIle Format) was the base of SVR3 and BSD 4.2 systems.
ELF

Executable and Linking Format that was developed by AT&T and is a
base for SVR4 UNIX.

The modifications done to the static and dynamic linker are to make linking on
AIX look and behave more like that on the SVR4.

2.9.1 dlsym() function
The dlsym() function accepts, as a first argument, either the value returned from
the call to dlopen(), or one of the special handles RTLD_DEFAULT,
RTLD_NEXT, or RTLD_SELF. It also accepts, as a second argument, a special
symbol named RTLD_ENTRY. These special handles and special symbol name
are defined in the <dlfcn.h> header file as follows:
#define
#define
#define
#define

RTLD_DEFAULT
RTLD_NEXT
RTLD_SELF
RTLD_ENTRY

(-1)
(-2)
(-3)
((const char *)(-1)

In the case of the special handle RTLD_DEFAULT, dlsym() searches for the
named symbol starting with the first object loaded and proceeding through the list
of initial loaded objects, and any global objects obtained with dlopen(), until a
match is found. This search follows the default model employed to relocate all
objects within the process.

78


In the case of the special handle RTLD_NEXT, dlsym() searches for the named
symbol in the objects that were loaded following the object from which the
dlsym() call is being made.
In the case of the special handle RTLD_SELF, dlsym() searches for the named
symbol in the objects that were loaded starting with the object from which the
dlsym() call is being made.
In the case of the special symbol name RTLD_ENTRY, dlsym() returns the
module's entry point. The entry point is a value of the module's loader section
symbol marked as entry point (bit LDR_ENTRY is set in l_smtype member of the
LDSYM structure - see the <loader.h> header file).
In the case of RTLD_DEFAULT, RTLD_NEXT, and RTLD_SELF, if the objects
being searched have been loaded from dlopen() calls, dlsym() searches the
object only if the caller is part of the same dependency hierarchy, or if the object
was given global search access, that is, it has been opened in dlopen()
RTLD_GLOBAL mode by dlopen().

2.9.2 New options of the ld command for SVR4 affinity
The static linker utility, the ld command, is enhanced by the following new
options:
-b svr4

Sets the SVR4 affinity of the linker. The option changes
the meaning of some other options on the command line
and the standard behavior of the linker.

-V

Prints the version string of the static linker to stderr.

The behavior of the following options were modified and are valid only when
-b svr4 is specified:
-dn

Specifies that ld uses static linking. This option is
equivalent to -b nso option.

-dy

Specifies that ld uses dynamic linking. Dynamic linking is
a default. This option is equivalent to -b so option.

-R path

A colon-separated list of directories used to specify a
runtime library search path. If present and not NULL, it is
recorded in the output file’s loader section. This is used
when linking an executable with shared libraries. Multiple
instances of this option are concatenated together. Run
time linker uses this path to locate shared libraries at
runtime.

-z defs

Forces fatal error if any undefined symbols remain at the
end of the link. This is the default when an executable is


79

built. It is also useful when building a shared library to
assure that the object is self-contained, that is, all its
symbolic references are resolved internally.
-z nodefs

Allows undefined symbols. This is the default when a
shared library is built. When used with executable, the
behavior of references to such undefined symbols is
unspecified. It is equivalent to -b erok option.

-z multidefs

Allows multiple symbol definitions. By default, multiple
symbol definitions that occur between relocatable objects
(.o files) will result in a fatal error condition. This option
suppresses the error condition and allows the first symbol
definition to be taken.

-z text

In dynamic mode only, it forces a fatal error if any
relocations against non-writable sections remain.

-z nowarntext

In dynamic mode only, it allows relocations against all
mappable sections, including non-writable ones. This is
the default when building a shared library.

-z warntext

In dynamic mode only, it warns if any relocations against
the .text section remain. This is the default when building
an executable.

When the -b svr4 flag is set, the following behavior is changed to assure better
compatibility:
The -b E or -b export options take precedence over -b expall or -b expfull.
This means that only symbols specified by -b E or -b export options are
exported.
There is no need to specify the -b noentry option when building a shared
library. It is the default.
When producing modules with the -b rtl option specified, multiple definitions
are not reported when the symbol is defined in different modules. They are
reported only when they occur in the same module.
The library search path specified with the -L option is not included in the
runtime library search path recorded in the file’s loader section. Only
/lib:/usr/lib path is recorded in the file. The path specified with -b libpath option
is recorded in the loader section. It is similar to the -R option, however, unlike
the -R option, in case of multiple -b libpath options, only the last path is
recorded. Multiple paths are not concatenated.
No matter if the -b svr4 flag is set, during symbol resolution if a symbol is found to
belong to an object file within an archive library, all global symbols from this
object file are exported.

80


2.10 Marking executable’s read/write sections
The new F_NONEXEC bit can be set in the f_flags field in the XCOFF header as
follows:
#define F_NONEXEC 0x8000

When this bit is set, the read/write sections (.data, .bss) of the executable,
shared library, or object file are not executable. If the bit is not set, all read/write
sections are executable. You can set or un-set the bit using the ld or ldedit
commands. Explanation of the commands follows.

2.10.1 Updated flags for the ld command
The ld command is updated by new flags that enable you to set the
F_NONEXEC flag as follows:
-b rwexec

Marks the read/write sections in the file as executable. It
does not set the F_NONEXEC bit in the XCOFF file
header. This is the default.

-b norwexec

This is a complimentary option to -b rwexec. It marks all
read/write sections in the file as non-executable and
indicates to the loader to make the stack non-executable.
This option sets the F_NONEXEC bit in the XCOFF file
header.

2.10.2 Updated flags for the dump command
The dump -ov binary utility command additionally displays the value of the
F_NONEXEC flag as RWNONEXEC. When the flag is set, you will get the
following format as highlighted in Example 2-1.
Example 2-1 Example of dump -ov command
# dump -ov a.out
a.out:
***Object Module Header***
# Sections
Symbol Ptr
# Symbols
Opt Hdr Len
4
0x0000190c
236
72
Flags=( EXEC DYNLOAD RWNONEXEC )
Timestamp = "Jul 16 13:35:57 2004"
Magic = 0x1df (32-bit XCOFF)

Tsize
0x00000acc

***Optional Header***
Dsize
Bsize
Tstart
0x00000368 0x00001004 0x10000128

Flags
0x9002

Dstart
0x20000bf4


81

SNloader
0x0004

SNentry
0x0002

SNtext
0x0001

SNtoc
0x0002

TXTalign
0x0005

DATAalign TOC
vstamp
0x0003
0x20000ed8 0x0001

maxSTACK
0x00000000

maxDATA
SNbss
0x00000000 0x0003

magic
0x010b

SNdata
0x0002
entry
0x20000ec4
modtype
1L

2.10.3 Updated flags for the ldedit command
If the executable is linked with the RWNONEXEC flag set in the f_flags field of
the XCOFF header, then it will have its writable and mappable sections
executable. This is not always desirable. The ldedit command allows you to set
or reset this bit in the file without need to re-link the file.
Use the ldedit command to clear the F_NONEXEC flag as follows:
ldedit -b rwexec

Use the ldedit command to set the F_NONEXEC flag as follows:
ldedit -b norwexec

2.11 Thread-based credentials
GRID applications need the operating system to support thread-based
credentials. AIX 5L Version 5.3 introduced a pthread-based API to set and
destroy thread-based credential information, such as UID and GID.
Per-thread credentials is the property by which different threads in a process own
different credentials and the access rights to a resource are checked at the
thread-level, ignoring any process-level credentials. This is a very useful concept
because it allows a single process to handle requests requiring different
credentials by creating threads with the appropriate credentials for each request.
Traditionally on systems that do not support per-thread credentials, this is
accomplished by having the main process run as root and forking another
process. There are several disadvantages with this approach, most notably the
fact that creating a process and acquiring a reduced or different set of credentials
involves many system calls and is usually an expensive operation.
AIX 5L Version 5.3 enables you to create threads with a reduced set of
credentials. The pthread_create_withcred_np() function allows you to create a
thread and switch the base credentials for the thread. The call of the thread

82


creation by itself is very similar to the pthread_create() function. Comparing to
pthread_create(), the pthread_create_withcred_np() additionally accepts the
structure that defines the credentials. The syntax is the following:
#include <pthread.h>
#include <sys/cred.h>
int pthread_create_withcred_np(pthread_t *thread, const pthread_attr_t *,
void *(*start_subroutine)(void), void *arg,
struct __pthrdscreds *credp)

The struct __pthrdscreds *credp input to the function is a structure where you set
the credentials. The remaining inputs of the system call are the same as for the
pthread_create() system call. The following credentials can be modified:
Effective, real, and saved user IDs
Effective, real, and saved group IDs
Supplementary group IDs
The following program excerpt provides an example. The program creates a
thread that runs our cred_thread() function with credentials of the user with
UID=203 and group with GID=1.
# more pt.c
#include <sys/types.h>
#include <pthread.h>
cred_thread()
{
... write your thread ...
}
int main()
{
pthread_t
thread;
struct __pthrdscreds thrcrs;
int rc=0;
...
thrcrs.__pc_flags=PTHRDSCREDS_INHERIT_UIDS | PTHRDSCREDS_INHERIT_GIDS;
thrcrs.__pc_cred.crx_ruid=203;
thrcrs.__pc_cred.crx_uid=203;
thrcrs.__pc_cred.crx_suid=203;
thrcrs.__pc_cred.crx_luid=203;
thrcrs.__pc_cred.crx_acctid=203;
thrcrs.__pc_cred.crx_gid=1;
thrcrs.__pc_cred.crx_rgid=1;
thrcrs.__pc_cred.crx_sgid=1;


83

rc=pthread_create_withcred_np(&thread, NULL, &cred_thread, NULL, &thrcrs);
...
}

When creating threads with credentials of other users, only processes entitled to
set credentials will succeed. Root users can allow the applications in the context
of another identity to use this feature. This is enabled through capabilities of the
system infrastructure. Typically a relevant application, like GRID, will pose a
question to the root user to provide this authority to take advantage of this
feature. It will also require root user to set any other relevant parameters such as
noswitchuid and noswitchgid. By defining the noswitchuid and noswitchgid
system parameters the administrator can set the lowest user ID and group ID
that a process with credentials capability is allowed to switch to.

2.12 Scalability enhancements
This section describes the improvements made to the operating system to help
achieve better scalability.

2.12.1 Kernel round robin locks
The kernel round robin locks (krlock) were introduced in AIX 5L. Version 5.3
makes performance optimizations to them to enable systems to scale over 48
CPUs. The new krlock reduces cache bounces and contentions while spinning
on the lock.
When there is a lock miss during a simple_lock() function call, krlock processing
takes place. The simple_lock() function is not changed and current applications
using simple_lock() on Version 5.2 are compatible with Version 5.3.

2.12.2 Increased resource limit values in /etc/security/limits file
The restrictions for resource limits specified in the /etc/security/limits file have
been removed along with the corresponding checks in the mkuser and chuser
commands. Now all resource limits can have a maximum value of 2147483647
or 2^31-1. When combined with the assumed block size of 512 bytes, this results
in a maximum of 2**40-1.
Table 2-1 provides a list of the old and new resource limit maximums allowed in
the /etc/security/limits file in AIX 5L Version 5.3.

84


Table 2-1 Resource limits in /etc/security/limits file
Limit

Old value

New value

CPU time

2.147.483.647 (2^31-1)

2.147.483.647 (2^31-1)

File size

4.194.303

2.147.483.647 (2^31-1)

Stack size

523.264

2.147.483.647 (2^31-1)

Memory size

2.147.483.647 (2^31-1)

2.147.483.647 (2^31-1)

Open files

2.147.483.647 (2^31-1)

2.147.483.647 (2^31-1)

Data size

4.194.303

2.147.483.647 (2^31-1)

Core size

2.147.483.647 (2^31-1)

2.147.483.647 (2^31-1)

2.12.3 CPU time limits
In prior releases, a program could avoid a cputime limit set with the ulimit
command by simply blocking, ignoring, or handling the SIGXCPU signal, that
says you're out of time to the process exceeding the limit. Now, however, such a
program can get away with doing this to get around a SOFT limit, but not a
HARD limit. The ability to handle an out of time condition is useful if a program
wants to format its own debugging information in response to the SIGXCPU.
The SIGXCPU signal is generated not only when the soft limit is reached, but
periodically afterwards (the old system did this, too) until the hard limit is
reached. At this time, however, the ability to block, ignore, or handle the signal is
overridden by the kernel, so the SIGXCPU will inescapably terminate the
process.
Note: The process that reaches the cpu_hard limit is killed by the SIGXCPU
signal and not with the SIGKILL signal.
This might create a binary compatibility problem, since a program that did not get
killed will now get killed! The system administrator has an easy way to control
this. If the old behavior is preferred, the administrator can set just a soft limit, and
leave the hard limit at infinity. To make the limit absolute, the administrator can
set the hard and soft limits to the same value. And to allow a program a grace
period in which to clean up, the administrator can set the hard limit a little bit
higher than the soft limit.


85

2.12.4 New heap size and late staggering
The vmgetinfo() function is enhanced by adding the following options that can be
passed to the function:
VM_STAGGER_DATA
VM_NEW_HEAP_SIZE
The use of these options with the vmgetinfo() function may improve the
performance, for example the TPCC numbers for databases, when using large
pages.
The VM_STAGGER_DATA option staggers the calling process's current sbreak
value by a cumulative per-MCM stagger value. This stagger value must be set
through the vmo command option data_stagger_interval. The out and arg
arguments should be NULL and 0, respectively. An example follows:
#include <sys/vminfo.h>
...
int rc=0;
rc=vmgetinfo(NULL, VM_STAGGER_DATA, NULL);

The VM_NEW_HEAP_SIZE option sets a new preferred page size for future
sbreak allocations for the calling process's private data heap. This page size
setting is advisory. The out parameter should be a pointer to a psize_t structure
that contains the preferred page size, in bytes, to use to back any future sbreak
allocations by the calling process. Presently, only 16 M (0x1000000) and 4 K
(0x1000) are supported. The arg parameter should be that of the sizeof(psize_t).
An example follows:
#include <sys/vminfo.h>
...
int rc=0;
psize_t buffer=0x1000000;
// Set to 16M value for LGPG
rc=vmgetinfo(&buffer, VM_STAGGER_DATA, sizeof(psize_t));

2.13 Increased inter-process communication limits
In AIX, upper limits are defined for the Inter-Process Communication (IPC)
mechanisms, which are not configurable. The individual IPC data structures are
allocated and deallocated as needed, so memory requirements depend on the
current system usage of IPC mechanisms.
AIX 5L releases prior to Version 5.3 defined the maximum number of semaphore
IDs, shared memory segment IDs, and message queue IDs to be 131072
(128 K) for the 64-bit kernel.

86


To cope with anticipated future scalability demands, AIX 5L Version 5.3
preventively increases the maximum number of data structures for each of the
IPC identifier types to 1048576 (1024 K).
This enhancement changes the scalability of the subroutines which are used to
get the IPC data structure IDs:
semget()

Returns the semaphore identifier

shmget()

Returns the shared memory identifier

msgget()

Returns the message queue identifier

The maximum numbers of IPC data structure IDs for all three subroutines are as
follows:
4096 for operating system releases before AIX 4.3.2 (32-bit kernel).
131072 for releases AIX 4.3.2 through AIX 5L Version 5.2 (32-bit and 64-bit
kernel).
1048576 for release AIX 5L Version 5.3 and later (64-bit kernel only).

2.14 Thread support in gmon.out
When applications consisting of multiple steps in which different executables, all
built with -p or -pg flags to generate profiling information, are invoked in a
sequence, each executable causes the previous gmon.out file to be overwritten.
It is often difficult to work through the nested scripts and source code to insert
commands or subroutine calls to rename the gmon.out files so that they are not
overwritten.
In AIX 5L Version 5.3, the gmon.out file has been made thread-safe, so that each
thread from a multi-threaded application has its data in it. The gmon.out file thus
created contains one histogram per thread and arc count per thread in the call
graph.
The current design of gprof provides process-level interpretation of gmon.out.
With a thread-level gmon.out gprof enables the user to analyze complex
multi-threaded applications as well as retain profiling information for a large
number of programs all executing from within the same directory.

2.15 Enhanced DBX
Major enhancements have been made to DBX in AIX 5L Version 5.3. The
following list identifies some of these enhancements:


87

Users now have additional control over the way DBX uses debug events.
Debug events are the user-defined breakpoints, tracepoints, and watchpoints
that control the execution of the debugged process. Should a user need to
disable the effects of a debug event, DBX currently requires that the event be
completely removed. In a scenario where an event must be repeatedly
removed and added, a user must manually reenter the event each time it is to
be enabled. This inconvenience has been removed with the addition of two
new DBX subcommands: disable and enable. These subcommands allow
the DBX user to temporarily disable debug events without permanently
removing them.
The function of the map subcommand has been extended. The new function
adds to DBX the ability to resolve addresses and certain types of symbols to
their respective loaded modules. The extensions to the map subcommand also
include facilities for requesting loaded module information for single modules,
ranges of modules, and with various levels of detail. In addition, a new
internal variable $mapformat has been introduced which allows user
customization of the default verbosity level for the map subcommand.
The stop and trace subcommands have been extended to allow stops and
traces that depend on changes in loaded module information. The new debug
event types allow user notification upon load or unload of specific modules, or
upon load or unload of all modules. The new function also enables DBX to
report the affected modules for each load change.
Currently, the DBX use subcommand sets the list of directories to be
searched when the DBX debug program looks for source files. These source
files, when located, are used by the DBX list, edit and file subcommands.
The “@” (at-sign) is a special symbol that, when included in the list of
directories specified with use, directs DBX to search the directory specified in
the object file (the file's location at compile-time) for the source file in
question.
This method of providing source code locations to DBX is cumbersome. If
source files reside in nested directory structures and the structures have been
relocated (for instance, when debugging remotely with the source mounted),
it would be more favorable to simply tell DBX the details of the relocation
rather than have to communicate to DBX each and every directory to search.
The function of the use subcommand has been extended by allowing the user
to specify, along with the search directories, mappings in the form of
[<oldpath>=<newpath>]. These mappings are then used in conjunction with
the special @ directory. Path replacements are performed in all cases where
the beginning of the full-path name as recorded in the object file matches
<oldpath>.

88


When using DBX to debug a corefile, it is important that dependent modules
that were loaded in the process at core creation time are available and
unchanged at debug time. DBX relies on the information in these dependent
modules to provide correct definitions and locations of text and data values.
Currently, if any dependent module referenced by the corefile is not available
at debug time, DBX displays the fatal error “cannot open <dependent module
path>” and exits. This behavior prevents any debugging of the corefile from
occurring until all of the dependent modules are located. Frequently,
however, this restriction is impractical. A missing dependent module may not
ultimately contribute any meaningful information to the debug session.
Furthermore, if a dependent module has changed since core creation time,
the definitions and locations of values in the process can be misrepresented
to the user by DBX. This usually has the effect of confusing the user and
leads them to believe that the debugger is functioning incorrectly. In the
extreme case, such as when the incorrect version of the libpthreads.a library
is used during debugging, DBX cannot report pthread information to the user.
Since DBX does not check for mismatched libraries, and thus does not warn
the user, the user has no idea why DBX may be acting erratically.
DBX now allows the debug session to continue even if any number of
dependent modules referenced by the corefile are unreadable. A notification
message is displayed by DBX during initialization for each missing dependent
module. The message includes the name of the missing module.
DBX also sends notification messages upon initialization and use of any
dependent module referenced in the corefile that is determined to be different
than at core file creation. These messages include the name of the
mismatched module.
Several new subcommands have been added to DBX. These include proc,
kthread, and fd. The purpose of these subcommands is to display data
contained within the data structures used by the kernel to keep track of
processes and threads. This is beneficial because it makes information about
the debug available that was previously very difficult for users to obtain. Much
of the data within the user process or file descriptor structures might have
been accessible with various DBX subcommands, but with these new
interfaces, all of the data is obtainable from one interface.
It is expected that these new subcommands will be particularly useful for
debugging core files. Previously, a wealth of the information locked in the
corefile would be very difficult to extract. The new subcommands expose this
data in a highly legible format.
To facilitate debugging pthreaded code, the scope of DBX function has been
enhanced by providing several new subcommands that reveal information
about pthread objects and display them in a readable, user-friendly format.
These include handler, onceblock, and resource subcommands.


89

DBX has been enhanced by adding a subcommand, corefile, that will
display information about a debuggee <corefile>. Without arguments, the
corefile subcommand will print basic information about the <corefile>. With
arguments, more detailed information can be requested. Some function may
be limited in cases where FULLCORE was not enabled at dump-time. In
these cases, a warning message is printed, as appropriate.
The corefile subcommand displays information from the header of a
corefile, including the executable name, corefile format versioning
information, flags indicating which data is available, the signal that caused the
crash, and the execution mode of the process that dumped core.
DBX can now list spinlocks by number, tid, or current state (similar to mutex).
It can also list all barriers, just one, or those with/without waiters (similar to
condition)

2.16 The tcpdump command
The tcpdump command has been updated to Version 3.8 for AIX 5L Version 5.3.
As a consequence of this upgrade, iptrace and ipreport were also changed to
use the new upgraded libcap library (version 0.8) for packet capture and dump
reading.
The tcpdump command prior to AIX 5L Version 5.3 displayed packet timestamps
down to 1 ns (10-9s). The open source tcpdump command displays timestamps at
10-6s, as BPF (Berkeley Packet Filter) supplies 10-6s accuracy on most
platforms. The new tcpdump command will have 10-6s time stamp resolution.
A total of 87 protocol printers have been included to facilitate printing when using
the tcpdump command.
Support for the latest SP switch has been added.
A number of new flags have been added to the tcpdump command. Some of the
most important ones are:
-D
-r

Does not require root privileges

-X

Dumps packets in hex and ASCIII

-vvv

More verbose than -v and -vv flags

-E

90

Lists supported link types

Decrypts ipsec ESP packets


2.17 gprof thread support
AIX 5L Version 5.3 provides support for threads in gprof. This enhancement is
aimed at adding new function to gprof so that it can interpret additional
information available for thread-level profiled programs. The additional
information would include per thread function call count, per thread arc execution
count, and amount of time spent executing a function per thread. The current
design of gprof provides process-level interpretation of gmon.out. The
enhancement to gprof is to interpret call graph and histogram information on a
per thread basis. The enhancement aims at enabling the user to analyze
complex multi-threaded applications as well as retain profiling information about
a large number of programs all executing from within the same directory.
The new mode is controlled with a new environment variable, GPROF, with the
following syntax:
GPROF=[profile:{process|thread}][,][scale:<scaling_factor>][,][file:{one|multi|
multithread}]

Where:
profile indicates whether thread or process level profiling is to be done.
scaling_factor represents the granularity of the profiling data collected.
file indicates whether a single or multiple gmon.out files should be
generated:
– multi: creates a file for each process (for each fork or exec)
•

gmon.out.<progname>.<pid>

– multithread: creates a file for each pthread
•

gmon.out.<progname>.<pid>.Pthread<pid>

•

Can be used to look at one pthread at a time with gprof or xprofiler

The default values are:
– profile: process
– scale: 2 for process level and 8 for thread level
Thread level profiling consumes considerably more memory
– file: one
The following new flags have been added to optionally separate output into
multiple files:
-g filename
– Writes the call graph information to the specified output filename.


91

– Suppresses the profile information unless -p is used.
-p filename
– Writes flat profile information to the specified output filename.
– Suppresses the call graph information unless -g is used.
-i filename
– Writes the routine index table to the specified output filename.
– If this flag is not used, the index table goes either:
•

At the end of the standard output

•

At the bottom of the filename(s) specified with -p and -g

Format of data itself is unchanged. It is presented in multiple sets. The first set
has cumulative data followed by one set of data per pthread.

2.18 Java 1.4.2
Both 32-bit and 64-bit versions of Java 1.4.1 are shipped with AIX 5L Version
5.3. However Java 1.4.2. is supported on AIX 5L Version 5.3 (and 5.1 and 5.2)
and is available from:
http://www.ibm.com/developerworks/java/jdk/aix/service.html

2.19 Java 3D
Java 3D version 1.3.1 is part of Java 1.4.2 which is supported on AIX 5L Version
5.3 and is available from:
http://www.ibm.com/developerworks/java/jdk/aix/service.html

2.20 Improved man command presentation
AIX 5L Version 5.3 includes the following improvements to the man command
presentation of man pages that are shipped on the AIX base media and the
documentation CD:
Sections are now indented. For example, the text of the Purpose section is
indented from the heading Purpose, which is lined up in the first column. The
same is true of all the sections.
The description of each flag is indented so it is clear where the description of
the flag begins and ends.

92


There is now visual differentiation between command names, parameters, file
names, and normal descriptive text. This makes the descriptive text easy to
read.
Multi-column tables are now neatly formatted.


93

94


3

Chapter 3.

Storage management
AIX 5L introduces several new features for the current and emerging storage
requirements. These enhancements include:
LVM enhancements
– Performance improvement of LVM commands
– Removal of classical concurrent mode support
– Scalable volume groups
– Striped column support for logical volumes
– Volume group pbuf pools
– Variable logical track group
JFS2 enhancements
– Disk quotas support for JFS2
– JFS2 file system shrink
– JFS2 extended attributes Version 2 support
– JFS2 ACL support for NFS V4
– ACL inheritance support
– JFS2 logredo scalability
– JFS2 file system check scalability


95

3.1 LVM enhancements
The following sections discuss the LVM enhancements in detail.

3.1.1 Performance improvement of LVM commands
A large number of changes and enhancements have been implemented in AIX
5L Version 5.3 to reduce the command execution time of several Logical Volume
Manager (LVM) high-level commands. All volume group types will benefit from
the improvements made to the following commands:
extendvg
importvg
mkvg
varyonvg
chlvcopy
mklvcopy
lslv
lspv
The scalability enhancements which were made to the LVM of AIX over the past
years required a constant effort to adjust the performance of user level
commands and library functions to the growing number and size of the utilized
resources. The prospect to configure and use a scalable volume group in AIX 5L
Version 5.3 which holds up to 1024 hard disk drives initiated several new code
changes and additions.
LVM metadata, which is critical for system integrity and operation, needs to be
stored in the volume group descriptor area (VGDA) and the volume group status
area (VGSA) of every single disk which belongs to a configured volume group.
AIX 5L Version 5.3 reduces the number of access incidences to VGDAs and
VGSAs that are necessary for LVM operations on volume groups and logical
volumes. Required read or write operations on metadata are now executed in
parallel by an automatically scaled number of threads. Many other improvements
focus on the efficiency of metadata access and optimize the effectiveness of their
usage. Specific focus was given to the following commands: extendvg, importvg,
mkvg, varyonvg, chlvcopy, mklvcopy, and lslv.
To speed up the creation of logical partition copies, the mklvcopy script was
rewritten in the C programing language. The new mklvcopy executables utilize
the optimized metadata access routines that were mentioned previous but also
exploit a new generic hashtable function integrated in the previous LVM library.

96


The default algorithm for the importvg command was further enhanced to reduce
the execution time while maintaining a maximum of integrity protection. With AIX
5L Version 5.3, it is no longer the default to scan every disk of a system for an
import operation. In Version 5.3 the importvg command uses the redefinevg
command to get all physical volume identifiers (PVIDs) by reading the VGDA of
the key device that is related to the volume group in question and after that only
the initial LVM records for those physical volumes are examined. The default
method of previous AIX releases used to read the LVM record of every disk in
the system trying to match the disks that are listed in the VGDA. Beginning with
AIX 5L Version 5.3, this method will be an error path to try other disks in the
system if needed.
The lspv command, when used without any flags or arguments, applies
additional hashing logic while assembling a complete list of all physical disks of a
given RS/6000, pSeries, or p5 system.

3.1.2 Removal of classical concurrent mode support
In conjunction with the importvg command redesign focused on execution time
improvement, the support for classical concurrent mode volume groups has been
removed. When trying to import a classical concurrent mode volume group, an
error message will inform the system administrator to convert the volume group
to an enhanced concurrent mode capable volume group.
The classical concurrent mode volume groups are only supported on Serial
DASD and SSA disks in conjunction with the 32-bit kernel. Beginning with AIX 5L
Version 5.1 the enhanced concurrent mode volume group was introduced to
extend the concurrent mode support to all other disk types. The enhanced
concurrent volume groups use AIX group services. AIX 5L Version 5.2 is no
longer allowed to create classical concurrent mode volume groups, and finally
Version 5.3 removes the support for that volume groups type entirely.

3.1.3 Scalable volume groups
Originally AIX allowed you to configure volume groups (VGs) with a maximum of
32 physical volumes (PVs), no more than 1016 physical partitions (PPs) per disk,
and an upper limit of 256 logical volumes (LVs) per VG. This VG type is
commonly referred to as the standard volume group. To handle the increase in
hard disk drive capacity over time, AIX Version 4.3.1 implemented a new volume
group factor, which can be specified by the -t flag of the mkvg command, that
allows you to increase the maximum number of PPs per disk proportional to the
given integer multiplier value. However, the maximum number of PVs is
adversely affected by this multiplier and decreases proportionally to the specified
factor. AIX Version 4.3.2 expanded the LVM scalability by introducing big volume

Chapter 3. Storage management

97

groups. A big VG can have up to 128 physical volumes and a maximum of 512
logical volumes defined with it.
AIX 5L Version 5.3 takes the LVM scalability to the next level and offers a new
scalable volume group (scalable VG) type. A scalable VG can accommodate a
maximum of 1024 PVs and raises the limit for the number of LVs to 4096. The
maximum number of PPs is no longer defined on a per disk basis, but applies to
the entire VG. This opens up the prospect of configuring VGs with a relatively
small number of disks, but fine-grained storage allocation options through a large
number of PPs which are small in size. The scalable VG can hold up to 2097152
(2048 K) PPs. Optimally, the size of a physical partition can also be configured
for a scalable VG. As with the older VG types the size is specified in units of
megabytes and the size variable must be equal to a power of 2. The range of PP
sizes starts at 1 (1 MB) and goes up to 131072 (128 GB), which is more than two
orders of magnitude above the 1024 (1 GB) maximum for AIX 5L Version 5.2.
(The new maximum PP size provides an architectural support for 256 petabyte
disks.) Table 3-1 shows the variation of configuration limits with the different VG
types. Note that the maximum number of user definable LVs is given by the
maximum number of LVs per VG minus 1 because one LV is reserved for system
use. Consequently, System administrators can configure 255 LVs in normal VGs,
511 in big VGs, and 4095 in scalable VGs.
Table 3-1 Configuration limits for volume groups
VG type
Normal VG
Big VG
Scalable VG

Maximum
PVs

Maximum
LVs

Maximum
PPs per VG

Maximum
PP size

32

256

32512
(1016 * 32)

1 GB

128

512

130048
(1016 * 128)

1 GB

1024

4096

2097152

128 GB

Metadata structure for scalable VGs
In AIX the first 128 sectors of a PV are reserved to store various types of
information which is indispensable for proper device integration and usage. The
physical sector number (PSN) layout for hard disk drives is defined by the
/usr/include/sys/hd_psn.h header file. AIX 5L Version 5.3 adds the new global
variable PSN_SVG_MWC_REC that defines the PSN of the first MWC cache
record for scalable VGs to start with sector 100.
There are no other changes to the PSN layout so you still find the LVM record to
start at sector 7 on the disk. However, the LVM record for a scalable VG differs
substantial in format and content from the LVM record for a standard or big VG.
All LVM record structures are defined in the /usr/include/lvmrec.h file.

98


Like in previous AIX releases the VGSA and the VGDA areas follow after the first
128 reserved sectors, but the structure of both entities has changed
considerably.
The VGSA for scalable VGs consists of three areas: PV missing area (PVMA),
MWC dirty bit area (MWC_DBA), and PP status area (PPSA). The overall size
reserved for the VGSA is independent of the configuration parameters of the
scalable VG and stays constant. However, the size of the contained PPSA
changes proportional to the configured maximum number of PPs:
PV missing areas

PVMA tracks if any of the disks are missing.

MWC dirty bit area

MWC_DBA holds the status for each LV if passive mirror
write consistence is used.

PP status area

PPSA logs any stale PPs.

The new type VGDA area starts at sector number 650 and consists of 5 different
areas. Again, while the overall size reserved for the VGDA stays constant, the
sizes of the contained areas are variable in proportion to the configured
maximum number of PPs and LVs for the scalable VG:
VG information area Holds VG-related metadata such as the current number of
LVs, the allowed maximum number of LVs, the number of
PVs, the allowed maximum number of PVs, and other
information of significance for a VG.
PV information area Keeps metadata related to PVs, such as the state of
every PV which belongs to a scalable VG.
LVCB information area
Maintains a record of logical volume control block (LVCB)
metadata for each LV defined on a scalable VG but which
is only related to the file system space.
LV entry area

Stores the LVM-related metadata such as strictness,
inter- and intra-policy for each LV in a scalable VG.

PP entry area

Records information pertinent to physical partitions, such
as their state.


99

Note: The LVCB contains metadata about a logical volume. For standard VGs
the LVCB resides in the first block of the user data within the LV. Big VGs keep
additional LVCB information in the ondisk VGDA. The LVCB structure on the
first LV user block and the LVCB structure within the VGDA are similar but not
identical. (If a big VG was created with the -T O option of the mkvg command,
no LVCB will occupy the first block of the LV.) With scalable VGs, logical
volume control information is no longer stored on the first user block of any
LV. All relevant logical volume control information is kept in the VGDA as part
of the LVCB information area and the LV entry area. Therefore, no
precautions have to be met when using raw logical volumes because there is
no longer a need to preserve the information held by the first 512 bytes of the
logical device.

Command interface changes
The scalable volume group implementation in AIX 5L Version 5.3 provides
configuration flexibility with respect to the number of physical and logical
volumes that can be accommodated by a given instance of the new volume
group type. The configuration options allow any scalable VG to contain 32, 64,
128, 256, 512, 768, or 1024 disks and 256, 512, 1024, 2048, or 4096 LVs. The
maximum values of 1024 PVs and 4096 LVs do not need to be configured at the
time of VG creation to account for future growth. The initial settings can always
be increased at a later date as required.
The following user commands were changed in support for the new scalable
volume group type: chvg, lsvg, and mkvg.

mkvg command changes
The enhanced mkvg command implementation provides four new command
flags, -S, -v, -P, and -I and three flags,-t, -L, and -s that are modified to adjust for
the option to create a scalable VG. The following example shows the syntax
statement of the mkvg command manual page and describes the new and
modified flags in more detail.
...
mkvg Command
Purpose
Creates a volume group.
Syntax
mkvg [ -d MaximumPhysicalVolumes ] [ -B ] [ -t factor ] [ -S [ -v
LogicalVolumes ] [ -P Partitions ] ] [ -C ] [ -G ] [ -f ] [ -i ] [ -I ][ -c ]

100


[ -x ] [ -L LTGSize ] [ -m MaxPvSize ] [ -n ] [ -s Size ] [ -V MajorNumber ] [
-y VolumeGroup ] PhysicalVolume ...
...

-S

The -S flag creates a scalable-type volume group. By default, this
volume group can accommodate up to 1024 physical volumes, 256
logical volumes, and 32768 physical partitions. To increase the number
of logical volumes, use the -v option. To increase the number of physical
partitions, use the -P option.

Note: Increasing the maximum number of logical volumes and the number of
physical partitions beyond the default values for a scalable volume group can
significantly increase the size of the VGDA proportionately. These values
should only be increased as needed because they cannot be decreased.
Meanwhile, as the VGDA space increases, all VGDA update operations
(creating a logical volume, changing a logical volume, adding a physical
volume, and so on) can take longer to run.
-v

The -v flag defines the number of logical volumes that can be created.
The corresponding value is specified by the logical volumes variable
which directly follows the flag. Valid values are 256, 512, 1024, 2048 and
4096. The default is 256. The chvg command can be used to increase
the number of logical volumes up to the maximum of 4096. This option is
only valid with the -S option.

-P

The -P flag defines the total number of partitions in a scalable volume
group. The corresponding value is specified by the partition variable that
directly follows the flag. The partitions variable is represented in units of
1024 partitions. Valid values are 32, 64, 128, 256, 512 768, 1024, and
2048. The default is 32 K (32768 partitions). The chvg command can be
used to increase the number of partitions up to the maximum of 2048 K
(2097152 partitions). This option is only valid with the -S option.

-I

The -I flag creates a volume group that can be imported to AIX 5L
Version 5.1 and 5.2. The logical track group (LTG) size will behave as if
the volume group had been created prior to AIX 5L Version 5.3. If the
logical volumes are later created with a stripe width that is larger than the
supported stripe size on AIX 5L Version 5.1 or 5.2, then attempting to
import the volume group back to AIX 5L Version 5.1 or 5.2 is not
supported.

-t

The -t flag changes the limit of the number of physical partitions per
physical volume for standard and big volume groups. The flag is directly
followed by the factor value. The concept to configure the number of PPs


101

per disk does not ably to scalable volume groups and consequently the -t
option is ignored with the -S option.
-L

For volume groups created on AIX 5L Version 5.3 without the -I flag, the
-L flag is ignored. When the volume group is varied on the logical track
group size will be set to the common max transfer size of the disks.
For volume groups created on Version 5.3 with the -I flag or for volume
groups created prior to Version 5.3, the -L flag is immediately followed by
the logical track group size value which is expressed in units of kilobytes,
and which must be 128, 256, 512, or 1024. In addition, the LTG value
should be less than or equal to the maximum transfer size of all disks in
the volume group. The default LTG size is 128 kilobytes.

-s

Sets the number of megabytes in each physical partition. The -s flag is
immediately followed by the size variable. The size is expressed in units
of megabytes from 1 (1 MB) through 131072 (128 GB) for scalable VGs.
The maximum supported PP size for standard and big VGs stays
unchanged at 1024 MB. The size variable must be equal to a power of 2
(for example 1, 2, 4, 8, and so on). The default value for 32 and 128 PV
volume groups will be the lowest value to remain within the limitation of
1016 physical partitions per PV. The default value for scalable volume
groups will be the lowest value to accommodate 2040 physical partitions
per PV.

chvg command changes
The enhanced chvg command implementation provides four new command
flags, -G, -P, -v, and -I; and two flags, -L, and -t that were modified to adjust for
the option to change a scalable VG. The new flags allow you to convert standard
and big VGs to scalable VGs and to increase the maximum number of physical
partitions and logical volumes for existing scalable VGs if desired.
The conversion of a previously configured standard or big VG to a scalable VG
type requires the volume group to be varied off (with the varyoffvg command).
This is different compared to the conversion of a standard VG to a big VG, which
can be carried out without taking the source VG offline. The chvg command
returns the following error message if you make an attempt to convert to a
scalable VG while the source VG is still in an active state:
# chvg -G testvg
0516-1707 chvg: The volume group must be varied off during conversion to
scalable volume group format.
0516-732 chvg: Unable to change volume group testvg.

This behavior accounts for the substantial transformations which are necessary
to convert the source VG’s metadata to the new scalable VG metadata format.

102


The new -P and -v flags of the chvg command provide the analogous
configuration options with the same set and range of values as the -P and -v
flags of the mkvg command. The -t flag of the chvg command will be ignored for
scalable VGs, as is the case with the mkvg command. The -L flag has no effect on
volume groups created on AIX 5L Version 5.3 but allows the LTG size to change
on standard and big VGs which were created on previous AIX releases. The -L
flag combined with a flag value of 0 can be used to enable the variable LTG
option on VGs which were created prior to Version 5.3. Refer to 3.1.6, “Variable
logical track group” on page 111 for more information about this particular option.
The new -G and -I flag of the chvg command offer the following functions:
Changes the volume group to scalable VG format. The conversion of
a standard or big VG to a scalable VG type requires the volume
group to be varied off (with the varyoffvg command).

-G

Note:
The -G flag cannot be used if there are any stale physical partitions.
Once the volume group is converted, it cannot be imported into AIX 5L
Version 5.2 or previous versions.
The -G flag cannot be used if the volume group is varied on in concurrent
mode.
There must be enough free partitions available on each physical volume for
the VGDA expansion for this operation to be successful.
Since the VGDA resides on the edge of the disk and it requires contiguous
space for expansion, the free partitions are required on the edge of the
disk. If those partitions are allocated for user usage, they will be migrated
to other free partitions on the same disk. The rest of the physical partitions
will be renumbered to reflect the loss of the partitions for VGDA usage.
This will change the mappings of the logical to physical partitions in all the
PVs of this VG. If you have saved the mappings of the LVs for a potential
recovery operation, you should generate the maps again after the
completion of the conversion operation. Also, if the backup of the VG is
taken with the map option and you plan to restore using those maps, the
restore operation may fail since the partition number may no longer exist
(due to reduction). It is recommended that backup is taken before the
conversion, and right after the conversion if the map option is utilized.
Because the VGDA space has been increased substantially, every VGDA
update operation (creating a logical volume, changing a logical volume,
adding a physical volume, and so on) may take considerably longer to run.


103

-I

The -I flag modifies the volume group so that it can be imported to
AIX 5L Version 5.1 and 5.2. The LTG size will behave as if the
volume group had been created prior to AIX 5L Version 5.3. This
operation might fail if the volume group contains striped logical
volumes whose stripe size is larger than the supported stripe size on
AIX 5L Version 5.1 or 5.2. If logical volumes are later created with a
stripe width that is larger than the supported stripe width on AIX 5L
Version 5.1 or 5.2, then attempting to import the volume group back
to AIX 5L Version 5.1 or 5.2 is not supported. The -I flag modifies the
specified standard or big VG to no longer allow the LVM to configure
the optimum LTG size dynamically during the varyon process. The -I
flag will not work for scalable VGs.

lsvg command changes
The lsvg command output on AIX 5L Version 5.3 has changed to show the
maximum physical partitions per volume group for all volume group types. The
corresponding number is listed as the value for the MAX PPs per VG field. The
lsvg output for scalable VGs will not display the maximum number of PPs per PV
because for the scalable VG type the limit for PPs is defined per VG and not per
PV.

SMIT and graphical user interface changes
The new scalable volume group is fully supported by the System Management
Interface Tool (SMIT) and the Web-based System Manager graphical user
interface.
Existing SMIT panels which are related to VG management tasks are changed
and many new panels are added to account for the new scalable VG type. For
example, you can use the new SMIT fast path _mksvg to directly access the Add
a Scalable Volume Group SMIT menu. Figure 3-1 on page 105 depicts the
appearance of the new menu when invoked by the smitty _mksvg command.

104


Figure 3-1 New SMIT menu: Add a Scalable Volume Group

AIX 5L Version 5.3 implements changes to the Volume Group container of the
Web-based System Manager to account for the new VG type.
Like in the previous AIX release, the wizard that supports the creation of new
VGs is dedicated to the standard VG type, but with Version 5.3 additional usage
information regarding the available VG types is provided and you are referred to
the advanced method for big VG or scalable VG creation.
The new task guide of the advanced VG creation method now offers the
additional option to create a scalable VG. You can activate the task guide by
selecting Volumes → New → Volume Group (Advanced Method)... in the
menu bar of the Volume Group container. Figure 3-2 on page 106 shows the new
inserted top-level panel that allows you to choose the type of VG you want to
create. Other Web-based System Manager dialogs impacted by the introduction
of the scalable VG type are related to the Volume Group Properties panel.


105

Figure 3-2 New Web-based System Manager panel: New Volume Group

3.1.4 Striped column support for logical volumes
Redundant Array of Independent Disks (RAID) is a term used to describe the
technique of improving data availability and data I/O rates through the use of
arrays of disks and various data-striping methodologies. RAID algorithms can be
implemented as part of the operating system's file system software, or as part of
a disk device driver.
AIX LVM supports the following RAID options:
RAID 0

Striping

RAID 1

Mirroring

RAID 10 or 0+1

Mirroring and striping

AIX 5L Version 5.3 further enhances the LVM RAID 0 and the LVM RAID 10
implementation and provides striped columns support for logical volumes. This
new feature allows to extend a striped logical volume even if one of the physical
volumes in the disk array became full.
In previous AIX releases you could enlarge the size of a striped logical volume
with the extendlv command as long as enough physical partitions were available
within the group of disks which define the RAID disk array. Rebuilding the entire
LV was the only way to expand a striped logical volume beyond the hard limits
imposed by the disk capacities. This work-around required to backup and delete

106


the striped LV and then to recreate the LV with a larger stripe width followed by a
restore operation of the LV data.
To overcome the disadvantages of this time-consuming procedure, AIX 5L
Version 5.3 introduces the concept of striped columns for LVs.
Prior to Version 5.3 the stripe width of a striped LV was determined at the time of
LV creation by either of the following two methods:
Direct specification of all physical volume names
Specification of the maximum number of physical volumes allocated to the
striped logical volume
The maximum number of PVs allocated to the striped LV is determined by the
upper bound value specified with the -u flag of the mklv command. If you use the
-u flag and also directly specify the PV names, the later takes precedence and
the -u flag will be ignored. It was not permitted to configure a striped logical
volume with an upper bound larger than the stripe width.
In AIX 5L Version 5.3 the upper bound can be a multiple of the stripe width. One
set of disks, as determined by the stripe width, can be considered as one striped
column. Note that the upper bound value is not related to the number of mirror
copies in case you are using a RAID 10 configuration. Figure 3-3 illustrates the
new concept of striped columns for a logical volume with stripe width 3 and an
upper bound of 6.

Figure 3-3 Logical volume with a stripe width of 3 and 2 striped columns


107

If you use the extendlv command to extend a striped logical volume beyond the
physical limits of the first striped column, an entire new set of disks will be used
to fulfill the allocation request for additional logical partitions. If you further
expand the LV more striped columns may get added as required and as long as
you stay within the upper bound limit. The chlv -u command allows you to
increase the upper bound limit to provide additional headroom for striped LV
expansion. The -u flag of the enhanced extendlv command can be used to raise
the upper bound and extend the LV all in one operation.

Command interface changes
The following commands are affected by the introduction of the new striped
column feature in AIX 5L Version 5.3:
mklv
chlv
extendlv
mklvcopy
The mklv command supports the striped columns concept for logical volumes
with a new -C flag and changed characteristics for the existing -u flag. The
following paragraph shows the usage information of the mkvg command as it is
returned when entered without any flag or parameter:
# mklv
0516-606 mklv: Volume group name not entered.
Usage: mklv [-a IntraPolicy] [-b BadBlocks] [-c Copies]
[-d Schedule] [-e InterPolicy] [-i] [-L Label] [-m MapFile]
[-r Relocate] [-s Strict][-t Type] [-u UpperBound]
[-v Verify] [-x MaxLPs] [-y LVname] [-S StripeSize] [-Y Prefix]
[-o Overlapping IO] [-C StripeWidth] [-T IOoption] VGname NumberOfLPs
[PVname...]
Makes a logical volume.

The new -C flag directly defines the intended stripe width for the LV. The -u flag
continues to be used to specify the maximum number of physical volumes for
new allocation, but beginning with AIX 5L Version 5.3 the given upper bound
value can be larger than the stripe width. If the striped LV has to be extended to
the next column a full new stripe width of disks will be used; for that reason, the
upper bound value must be specified in multiples of the stripe width. The
following list describes the new -C command flag and the modified -u flag in more
detail.
-C Stripe_width

108

Sets the Stripe width of the logical volume. If the
Stripe_width variable value is not entered it is assumed to


be the UpperBound variable value or the total number of
disks specified on the command line.
-u UpperBound

Sets the maximum number of physical volumes for new
allocation. The value of the UpperBound variable should
be between one and the total number of physical
volumes. When using super strictness, the upper bound
indicates the maximum number of physical volumes
allowed for each mirror copy. When using striped logical
volumes, the UpperBound value must be a multiple of the
Stripe_width value. If upper bound is not specified, it is
assumed to be Stripe_width for striped logical volumes.

As mentioned previously the -u flag of the chlv, extendlv, and mklvcopy
commands will now allow you to change the upper bound to be a multiple of the
stripe width. The extendlv -u command can be used to change the upper bound
and to extend the LV in a single operation.

3.1.5 Volume group pbuf pools
The Logical Volume Manager (LVM) uses a construct named pbuf to control a
pending disk I/O. Pbufs are pinned memory buffers and one pbuf always is used
for each individual I/O request, regardless of the amount of data that is supposed
to be transferred. AIX creates extra pbufs when a new physical volume is added
to a volume group.
In previous AIX releases, the pbuf pool was a system-wide resource, but starting
with AIX 5L Version 5.3 the LVM assigns and manages one pbuf pool per volume
group. This enhancement supports advanced scalability and performance for
systems with a large number of VGs and applies to all VG types. As a
consequence of the new pbuf pool implementation, AIX displays and manages
additional LVM statistics and tuning parameters. AIX 5L Version 5.3 introduces
the lvmo command to provide support for the new pbuf pool-related
administrative tasks.
The lvmo command can be used to display pbuf and blocked I/O statistics and
the settings for pbuf tunables regardless of whether the scope of the entity is
system-wide or VG-specific. However, the lvmo command only allows the
settings to change for the LVM pbuf tunables that are dedicated to specific VGs.
The sole pbuf tunable with system-wide scope continues to be managed by the
ioo command. Also, the system wide number of I/Os that were blocked due to
lack of free pbufs can still be displayed by the vmstat -v command like in AIX
releases prior to Version 5.3.


109

Use lvmo -? to display the command usage statement:
# lvmo -?
lvmo: Not a recognized flag: ?
lvmo -v VgName -o Tunable [ =NewValue ]

The lvmo command displays and tunes the following items for the VG specified
by the -v flag:
pv_pbuf_count

Number of pbufs added when the specified VG is
extended by one PV. This parameter is tunable with the
lvmo command and takes effect immediately at run time.
The pv_pbuf_count default values are 256 for the 32-bit
kernel and 512 for the 64-bit kernel. The lvmo command
performs no range checking.

total_vg_pbufs

Number of pbufs currently available for the specified VG.
This parameter is tunable with the lvmo command and
takes effect after the VG has been varied off and varied
on again.

max_vg_pbuf_count Maximum number of pbufs that can be allocated for the
specified VG. This parameter is tunable with the lvmo
command and takes effect after the VG has been varied
off and varied on again.
pervg_blocked_io_count Number of I/Os that were blocked due to lack of free
pbufs for the specified VG. This parameter provides
statistics information about congestion for the VG pbuf
pool and can serve as an indicator for required
adjustments to the pv_pbuf_count and total_vg_pbufs
tunables. This counter will be reset whenever the VG is
varied on.

Note: The pbuf pools are not necessarily incorrectly sized just because LVM
uses up all of its pbufs and as a result I/Os are blocked due to the lack of free
pbufs. As a rule of thumb you can say that having an LVM pbuf pool resource
which is larger than 2 or 3 times the queue depth of the underlying devices
may provide little if any increase in overall throughput. A pbuf resource that is
too large could adversely affect the performance of other VG on the same
adapters.

The lvmo command displays the following system-wide items:
global_pbuf_count

110

Minimum number of pbufs that are added when a PV is
added to any VG. This system-wide parameter is tunable


with the ioo command but not with the lvmo command.
Changes to the parameter take effect for any VG that
has been varied off and varied on again after the new
value was configured through the ioo command. Note
that the global_pbuf_count parameter label of the lvmo
command maps to the pv_min_pbuf parameter label of
the ioo command. The global_pbuf_count default values
are the same as for the pv_pbuf_count parameter: 256
for the 32-bit kernel and 512 for the 64-bit kernel.
global_blocked_io_count System-wide number of I/Os that were blocked due
to lack of free pbufs. This parameter provides
accumulated statistics on congestion for the combined
pbuf pool resources on your system and can serve as an
indicator for required adjustments to the
global_pbuf_count, pv_pbuf_count, and total_vg_pbufs
tunables. The global_blocked_io_count parameter value
is also displayed by the vmstat -v command. The
counter is reset at system startup.
If both the VG-specific pv_pbuf_count and the system-wide global_pbuf_count
are configured, the larger value takes precedence over the smaller.
Use the lvmo -a command to display the current value of the pbufs statistics and
tuning parameters. The following shows an example output for the rootvg of a
given system:
# lvmo -a
vgname = rootvg
pv_pbuf_count = 512
total_vg_pbufs = 512
max_vg_pbuf_count = 16384
pervg_blocked_io_count = 1
global_pbuf_count = 512
global_blocked_io_count = 1

3.1.6 Variable logical track group
When the Logical Volume Manager receives a request for an I/O it breaks the I/O
down into what is called logical track group (LTG) sizes before it passes the
request down to the device driver of the underlying disks. The LTG is the
maximum transfer size of a logical volume and is common to all the logical
volumes in the volume group since it is a volume group attribute.
AIX 5L Version 5.2 accepted LTG values of 128 KB, 256 KB, 512 KB, and
1024 KB. However, many disks now support transfer sizes larger than 1 MB. To


111

take advantage of these larger transfer sizes and get better disk I/O performance
AIX 5L now accepts values of 128 KB, 256 KB, 512 KB, 1 MB, 2 MB, 4 MB,
8 MB, and 16 MB for the LTG size.
In contrast to previous releases AIX 5L Version 5.3 also allows the stripe size of
a Logical Volume to be larger than the LTG size in use and expands the range of
valid stripe sizes significantly. Version 5.3 adds support for 2 MB, 4 MB, 8 MB,
16 MB, 32 MB, 64 MB, and 128 MB stripe sizes to complement the 4 KB, 8 KB,
16 KB, 32 KB, 64 KB, 128 KB, 256 KB, 512 KB, and 1 MB stripe size options of
the former release.
In AIX 5L Version 5.2 the LTG size was set by the -L flag on the chvg or mkvg
command. In AIX 5L Version 5.3 it is set by varyonvg using the flag -M. The LTG
size thus created is called the variable logical track group size.
The following command sets the LTG size of the tmpvg volume group at 512 KB:
# varyonvg -M512K tmpvg

The LTG size is specified either in K or M units implying KB or MB respectively.
When the LTG size is set using the -M flag, the varyonvg and extendvg
commands may fail if an underlying disk has a maximum transfer size that is
smaller than the LTG size.
If the -M flag is not used, the varyonvg command will select the optimal LTG size
automatically. This optimal size is the largest common maximum transfer size
among the underlying disks. If an extendvg or reducevg command is
subsequently executed and the optimal LTG size changes, the LVM driver will
correct the LTG size on the fly. That is, it will hold the I/O, modify the LTG size,
and then resume.
The following command will enable variable LTG for a tmpvg volume group
created prior to AIX 5L Version 5.3 at the next varyonvg and will set the logical
track group size to the common max transfer size of the disks:
# chvg -L 0 tmpvg

If this command is not executed a volume group created prior to AIX 5L Version
5.3 will have the old LTG size in Version 5.3 and this LTG size will behave the
same way it did in the prior release.
You can specify 128, 256, 512, or 1024 instead of 0 on the command line to
indicate 128 KB, 256 KB, 512 KB, or 1024 KB LTG size, respectively. The value
should be less than or equal to the maximum transfer size of all disks in the
volume group. The default size is 128 kilobytes.

112


By default, volume groups in AIX 5L Version 5.3 are created with variable logical
track group size. If you want to import it to a previous release of AIX, you first
need to disable the variable LTG using the -I option for mkvg or chvg (then do a
varyoffvg followed by exportvg), otherwise the importvg command on the
previous release will fail.
The -L flag on the mkvg or chvg command is ignored for volume groups created in
AIX 5L Version 5.3. A message will be printed that says varyonvg will determine
LTG size based on the maximum transfer size of the disks at varyon time.
To obtain what the maximum supported LTG size of your hard disk is, you can
use the lquerypv command with the -M flag. The output gives the LTG size in
KB, as can be seen from the following lines:
# /usr/sbin/lquerypv -M hdisk0
256

The lspv command will display the same value as MAX REQUEST, as shown in
Example 3-1.
Example 3-1 The value of LTG size is shown as MAX REQUEST
# lspv hdisk0
PHYSICAL VOLUME:
hdisk0
VOLUME GROUP:
PV IDENTIFIER:
000bc6fdbff92812 VG IDENTIFIER
000bc6fd00004c00000000fda
469279d
PV STATE:
active
STALE PARTITIONS:
0
ALLOCATABLE:
PP SIZE:
16 megabyte(s)
LOGICAL VOLUMES:
TOTAL PPs:
542 (8672 megabytes)
VG DESCRIPTORS:
FREE PPs:
HOT SPARE:
USED PPs:
MAX REQUEST:
FREE DISTRIBUTION: 108..76..30..108..109
USED DISTRIBUTION: 01..32..78..00..00

rootvg

yes
9
2
no
256 kilobytes

You can list the value of the LTG in use with the lsvg command shown in
Example 3-2.
Example 3-2 The output shows the value of LTG in use.
# lsvg rootvg
VOLUME GROUP:
rootvg
000bc6fd00004c00000000fda469279d
VG STATE:
active
VG PERMISSION:
read/write
megabytes)
MAX LVs:
256
megabytes)

VG IDENTIFIER:
PP SIZE:
TOTAL PPs:

16 megabyte(s)
542 (8672

FREE PPs:

431 (6896


113

LVs:
megabytes)
OPEN LVs:
TOTAL PVs:
STALE PVs:
ACTIVE PVs:
MAX PPs per VG:
MAX PPs per PV:
LTG size (Dynamic):
HOT SPARE:

9

USED PPs:

111 (1776

8
1
0
1
32512
1016
256 kilobyte(s)
no

QUORUM:
VG DESCRIPTORS:
STALE PPs:
AUTO ON:

2
2
0
yes

MAX PVs:
AUTO SYNC:
BB POLICY:

32
no
relocatable

Note: The LTG size for a volume group created in AIX 5L Version 5.3 will be
displayed as Dynamic in the lsvg command output as shown in Example 3-2.

3.2 JFS2 enhancements
The following sections describe JFS2 enhancements found in AIX 5L Version
5.3.

3.2.1 Disk quotas support for JFS2
AIX 5L introduced the Enhanced Journaled File System (JFS2), which offers
unprecedented scalability and performance characteristics. AIX 5L Version 5.3
further expands the JFS2 function and provides a mechanism to control usage of
persistent storage by implementing disk usage quotas.
Persistent storage limits, from here on referred to as quotas, may be set for
individual users or groups on a per file system basis. The quota system will issue
a warning to the user when a particular quota is exceeded, but allow some extra
space for current work. Remaining over quota beyond a specified grace period
will result in further allocation attempts being denied until the total usage is
reduced below the user’s or group’s quota.
To reduce quota information overhead and system administration time the JFS2
quota implementation introduces the concept of Limits Classes, which is
explained in more detail in “Concepts and implementation specifics” on
page 115. Also, in contrast to the JFS disk quota system, AIX 5L Version 5.3
offers extensive configuration support through a comprehensive set of new
System Management Interface (SMIT) menus.

114


Packaging and installation
All executables that are required to use the AIX disk quota system for JFS and
JFS2 file system types are included in the bos.sysmgt.quota fileset. This fileset is
available by default after any AIX Base Operating System (BOS) installation.
The lslpp output given in the following example shows the file content of the
bos.sysmgt.quota fileset in AIX 5L Version 5.3:
# lslpp -f bos.sysmgt.quota
Fileset
File
---------------------------------------------------------------------------Path: /usr/lib/objrepos
bos.sysmgt.quota 5.3.0.0
/usr/sbin/repquota
/usr/sbin/edquota
/usr/sbin/quot
/usr/sbin/quotacheck
/usr/sbin/quotaoff -> /usr/sbin/quotaon
/usr/sbin/quotaon
/usr/sbin/j2edlimit
/usr/sbin/quota

To support the new and enhanced user-level commands, JFS2-specific
commands were added to the quotactl() subroutine which is included in the AIX
standard C Library (libc.a). The standard C library is installed by default on any
AIX system as part of the bos.rte.libc fileset.
Additional changes had to be made to the chfs command to enable quotas on
JFS2 file systems.

Concepts and implementation specifics
The disk quota system in AIX is based on the functional description documented
in UNIX System Manager's Manual (4.3 Berkeley Software Distribution): Disk
Quotas in a UNIX Environment. This document can be accessed through the
following URL:
http://www.openbsd.org

The JFS-specific implementation in previous AIX releases further extends the
function by the support of distinct quotas for users and groups and the optional
enforcement of hard quota limits. The JFS-specific disk quota system is
managed by command-line utilities.
The new JFS2-specific disk quota system implementation in AIX 5L Version 5.3
additionally introduces the concept of Limits Classes. Limits classes allow the


115

configuration of per file system default limits and offer comprehensive support
through dedicated SMIT panels.

Limits Classes
In most installations that configure disk usage quotas there is little variation in the
limits placed on individual users.
There may be a small number of conceptual classes where all members of one
given class have the same individual limits (note that this is different from group
quotas, where the total space used by everyone in that group is limited), but for
the most part the traditional JFS-specific implementation duplicates usage limits
information in every quota record.
Duplicating this information makes it unreasonable to implement any sort of
default limits scheme, especially one that allows such defaults to be changed. To
eliminate this problem the new JFS2-specific disk quota system introduces Limits
Classes which define a set of hard and soft disk block and file allocation limits,
and the grace periods before the soft limit becomes enforced as the hard limit.
System administration then becomes a process of defining Limits Classes (using
different sets for user and group limits) and assigning users or groups to a
particular class. Members of a Limits Class are bound by the limits defined by the
class; a user or group that is not a member of a specific class is automatically a
member of and bound by the limits of a per file system Default Class. The
Default Class is always present as soon as the JFS2 file system has been
enabled for quotas. The system administrator, however, can change the limit
settings for any Default Class. Deleting an assignment to a user-defined Limits
Class can be accomplished by assigning the user or group in question back to
the Default Class. Changing the limits for an entire class is simply a matter of
changing the limits in the Limits Class; there is no need to propagate those
changes to every member of the class as in the traditional JFS-specific quota
implementation.
To implement the Limits Class concept, the traditional limits information records
are replaced by two new record types: a Limits Record which contains the hard
and soft limits and grace period information, and a Usage Record which contains
the usage and warning information for a particular user or group.
Each Usage Record simply references a Limits Record, and users with common
limits defined all reference a single Limits Record, thereby reducing quotas
information overhead. This method also drastically reduces system
administration time. By predefining and designating a Limits Record to be the file
system default, the administrator does not need to create any records at all for
users who are subject to those default limits. If a user does not have a Usage

116


Record defined at either quotacheck or resource allocation time one is
automatically created which references the default limits.

AIX 5L Version 5.3 quota user commands
The section “Packaging and installation” on page 115 gives an overview of all
components which have to be present to configure the disk quota system for JFS
and JFS2 files systems. The following paragraphs provide an outline of the quota
user commands in AIX 5L Version 5.3.
The chfs command is used to enable quotas on JFS or JFS2 file systems. For
JFS file systems you are allowed to define alternate names and locations for the
quota files that are used to store the quota records. The default names are
quota.user and quota.group, and they are created in the root of the JFS file
system as soon as you run the quotacheck or the appropriate edquota command.
When a JFS2 file system is quota enabled by the use of the chfs command, the
default quota files are automatically created in the root of the JFS2 file system
and the system administrator cannot change their names and locations.
Note: Do not configure user-defined quota file names and locations for a
quota-enabled JFS file system that coincide with the default quota file names
and locations of a quota-enabled JFS2 file system. Also, be aware of the fact
that the quota files only get created if no existing instances can be found.
The behavior of the chfs command has also changed significantly with respect to
file system quota activation.
If you use chfs to configure a JFS file system for quotas only the file system
stanza in /etc/filesystems receives an additional entry for the quota attribute.
After that the quotacheck command has to be used to update the quota files with
the current disk usage information. The quotaon command must follow
immediately to initiate the concurrent tracking of disk usage and the enforcement
of the limits. Also, you have to apply the quotacheck command and the quotaon
command in sequence after a JFS quota enabled file system was mounted to
ensure the accuracy and currency of the disk usage information and the
enforcement of the limits.
If you use chfs to configure a JFS2 file system for quotas the quotacheck
command will be invoked by default, the quota system becomes automatically
active, and the defined quota limits are enforced. Likewise, if you mount a JFS2
quota-enabled file system, no quotacheck and no quotaon command is required
to activate the quota system in a current and consistent state.
In the presence of mounted quota-enabled JFS2 file systems, the disk usage
statistics are always maintained, the quota system is active, and the quota limits


117

will be applied. The quotaon and quotaoff commands are only required to assert
control over quota limit enforcement.
The edquota command edits user and group quotas that pertain to JFS file
systems only. If you use this command no quota-enabled JFS2 file system will be
listed in the temporary file that is opened by the edquota command and contains
the disk quota limit settings.
The new j2edlimit command manages quota Limits Classes for JFS2 file
systems and as such can be considered as the JFS2 counterpart of the edquota
command. Note the important distinction between these two commands: edquota
manages quota information from a user point of view, whereas j2edlimit
manages Limits Classes from a file system point of view.
The j2edlimit command enables the root user to perform the following tasks:
Edit quota Limits Classes
List quota Limits Classes
Set an Existing Limits Class as the Default Limits Class
Assign a user or group to a Limits Class
The j2edlimit command manages quota Limits Classes for one JFS2 file
system at a time. In case you need to duplicate the identical Group Limits Class
and User Limits Class definitions to another JFS2 file system you can use the
following procedure:
1. Use the cp command to copy the relevant quota files from the root of the
source to the root of the target JFS2 file system. If you merely control user
quota only, the quota.user file needs to be copied. An analogous statement
holds true for group quota control, which means that you have to copy the
quota.group file only. You will have to copy both quota files for a full quota
system implementation.
2. Run the chfs command against the target JFS2 file system. The chfs
command does not recreate or overwrite the existing copy of the quota files,
but will call quotacheck to update the quota files usage records while keeping
the Limits Class definitions.
The remaining quota-related executables of the bos.sysmgt.quota file set,
repquotas, quotacheck, quotaoff, quotaon, and quota examine the file system
type and either execute the current code for a JFS-type file system or execute a
JFS2-specific utility. Any JFS2-specific flags used when a JFS file system is
specified result in a warning message and are ignored; when specified with
commands that operate on multiple/all file systems, the extraneous flags are
ignored silently.

118


The changes and additions made to the user-level commands are enabled by
new JFS2-specific commands for the quotactl() subroutine. The quotactl()
subroutine manipulates disk quotas and is included in the AIX Standard C Library
(libc.a).

File system centric SMIT interfaces
Previous AIX releases use a user/group centric approach to manage quotas.
Selected file systems are enabled for quotas by the chfs command but the quota
limits are defined by the edquota command for individual users or groups.
Since the JFS2 quota support in AIX 5L Version 5.3 introduces the concept of
per file system Limits Classes, a file system centric approach to manage quotas
is potentially more intuitive.
In AIX 5L Version 5.3 the JFS2 quota support provides SMIT panels for the file
system centric approach which directly map to JFS2 quota management
commands. Table 3-2 provides an overview of the changes and additions to the
SMIT interface in support for the JFS2 disk quota system.
Table 3-2 File system centric SMIT panels for JFS2 disk quota support
SMIT panel title

Fast path

Type

Status

Enhanced Journaled File System

jfs2

Menu

modified

Add an Enhanced Journaled File System

crjfs2std

Dialog

modified

Add a JFS2 on a previously defined logical
volume

crjfs2lvstd

Dialog

modified

Change / Show Characteristics of a JFS2

chjfs2

Dialog

modified

Manage Quotas for an Enhanced Journaled
File System

j2fsquotas

Menu

added

Enable / Disable Quota Management

j2enablequotas

Dialog

added

Stop / Restart Quota Limits Enforcement

j2enforcequotas

Dialog

added

List Quota Usage

j2repquota

Dialog

added

Recalculate Current Disk Block and File
Usage Statistics

j2quotacheck

Dialog

added

Add a Limits Class

j2addlimit

Dialog

added

Change / Show Characteristics of a Limits
Class

j2changelimit

Dialog

added

Make a Limits Class the Default for a File
System

j2defaultlimit

Dialog

added


119

SMIT panel title

Fast path

Type

Status

Assign a User or Group to a Limits Class

j2assignlimit

Dialog

added

List Limits Classes for a File System

j2listlimits

Dialog

added

Remove a Limits Class

j2removelimit

Dialog

added

One SMIT menu and three dialogs have been modified to support quotas on
JFS2 file systems. But the new SMIT menu Manage Quotas for an Enhanced
Journaled File System can be considered as the most important addition to the
SMIT interface since it provides a single point of control to handle JFS2 quota
tasks. Figure 3-4 shows the actual appearance of this new SMIT menu.

Figure 3-4 New SMIT menu: Manage Quotas for an Enhanced JFS

Setting up the Disk Quota System for JFS2
Depending on the specifics of a given system environment you may encounter
the situation where it is required to configure the AIX disk quota system for a
mixture of JFS and JFS2 file systems.
Consider the following scenario: The International Technical Support
Organization (ITSO) is one of many departments of IBM that is concerned with
technical support. Within the ITSO a number of different job roles exist. One of
the ITSO employees holds the user ID aroell and works as an author on a given
project.

120


The following steps provides you with some insight in the required procedure to
establish a disk quota system on an IBM internal server for this particular
scenario:
1. Log in with root authority.
2. Enable user and group quotas on the JFS2 /home file system and the
mounted JFS /authors sample file system:
# chfs -a "quota = userquota,groupquota" /home
Initializing quota file /home/quota.user
Initializing quota file /home/quota.group
*** Checking user and group quotas for /dev/hd1 (/home)
root
fixed: inodes 0 -> 3
blocks 0 -> 64
bin
guest
aroell
blocks 0 -> 4
sdutta
blocks 0 -> 4
system
blocks 0 -> 64
staff
blocks 0 -> 8
bin
usr
# chfs -a "quota = userquota,groupquota" /authors

3. Use the edquota command to set the quota limits for the sample group ITSO
(edquota -g ITSO) and the sample user aroell (edquota -u aroell) on the
/authors JFS file system. The following example entries show the chosen
quota limits:
Quotas for group ITSO:
/authors: blocks in use: 86324, limits (soft = 5242880, hard = 10485760)
inodes in use: 167, limits (soft = 50000, hard = 100000)
Quotas for user aroell:
/authors: blocks in use: 86324, limits (soft = 51200, hard = 102400)
inodes in use: 167, limits (soft = 500, hard = 1000)

The block size is 1 KB and consequently the group ITSO has used about 84
MB of the maximum 5 GB of disk space. Of the maximum 50,000 files, users
belonging to the ITSO group created just 167. The ITSO allows buffers of 5
GB of disk space and 50,000 files that can be allocated to temporary storage.
The quotas for user aroell can be interpreted likewise.
4. Use the j2edlimit -g /home command to define a new Group Limits Class
for the /home JFS2 file system. Within this example the new Group Limits
Class defines the quota limits for all technical support departments within
IBM. The + character in the ID field indicates that a new Group Limit Class
shall be added. The related Group Limit Class ID will get automatically
assigned by the system when you write your changes to the j2edlimit
temporary file and leave the editing session. In our case the Group Limit


121

Class ID 1 is chosen. The following example entry lists the related quota
limits:
Group Limits Classes for file system /home
Block Limits units: g=gigabyte, m=megabyte, or optional k=kilobyte
Grace Period units: d=days, h=hours, m=minutes, or s=seconds
Limits Class ID 0 is the default class.
Prepend '-' to ID to delete Limits Class.
Use '+' for ID to add new Limits Class.

ID

Block Limits
soft
hard

File Limits
soft
hard

Grace Period
block
file

0
+

0
50g

0
50000

0
4d

0
1024g

0
100000

0
4d

5. Use the j2edlimit command to assign the ITSO sample group to the Group
Limits Class ID 1:
# j2edlimit -a 1 -g ITSO /home

Because of the scope of the Group Limits Class ID 1 definition the system
administrator could assign any other group of any given technical support
department as required. This eliminates the necessity to configure and
administer quota limits for each technical support department group
individually. This is only a valid statement as long as the defined limits apply
to every group in question.
6. Use the j2edlimit -u /home command to define a new User Limits Class for
the /home JFS2 file system. Within this example the new User Limits Class
defines the quota limits for all ITSO employees regardless of their specific job
role. The related User Limit Class ID will get automatically assigned by the
system when you write your changes to the j2edlimit temporary file and
leave the editing session. In our case the User Limit Class ID 1 is chosen.
The following example entry lists the related quota limits:
User Limits Classes for file system /home
Block Limits units: g=gigabyte, m=megabyte, or optional k=kilobyte
Grace Period units: d=days, h=hours, m=minutes, or s=seconds
Limits Class ID 0 is the default class.
Prepend '-' to ID to delete Limits Class.
Use '+' for ID to add new Limits Class.

ID

File Limits
soft
hard

Grace Period
block
file

0
+

122

Block Limits
soft
hard
0
51200

0
500

0
0

0
102400


0
1000

0
0

7. Use the j2edlimit command to assign the user aroell to the User Limits
Class ID 1:
# j2edlimit -a 1 -u aroell /home

Because of the scope of the User Limits Class ID 1 definition the system
administrator could assign any other user in the ITSO department as
required. This eliminates the necessity to configure and administer quota
limits for each ITSO employee individually. This is only a valid statement as
long as the defined limits apply to every ITSO system user in question.
8. Enable the quota system with the quotaon command for all file systems:
# quotaon -v -a

9. Use the quotacheck command to check the consistency of the quota files
against actual disk usage:
# quotacheck -v -a

10.Use the repquota command to view a summary of quotas and disk usage for
all your file systems:
# repquota -v -a

3.2.2 JFS2 file system shrink
Prior to Version 5.3 there is no way to shrink a file system dynamically while you
are using it, although you can easily extend as needed. The procedure to shrink
a file system was to create a new smaller version, copy the data, take the old
version offline, then delete the old version. In Version 5.3 file system shrink is
now available with JFS2.

Shrink a file system using SMIT
The easiest way to shrink a file system is through SMIT. Use SMIT chjfs2 fast
path and choose the file system you want to shrink; you will see the menu shown
in Figure 3-5 on page 124.


123

Figure 3-5 Change/Show Characteristics of an Enhanced Journaled File System

You can simply change the file system to your desired size in exactly the same
way you extend it, but with the smaller number of units as shown in Figure 3-6 on
page 126.

Shrink a file system using a command
User commands are available to shrink a file system. Use the chfs command
with the following options:
chfs [ -n NodeName ] [ -m NewMountPoint ] [ -u MountGroup ]
[ -A { yes | no } ] [ -p { ro | rw } ] [ -t { yes | no } ]
[ -a Attribute=Value ] [ -d Attribute ] FileSystem

The enhanced options of the chfs command are discussed in the following
sections.

-a size=NewSize
Specifies the size of the Enhanced Journaled File System in 512-byte blocks,
megabytes or gigabytes. If NewSize has the M suffix, it is interpreted to be in
Megabytes. If NewSize has a G suffix, it is interpreted to be in Gigabytes. If
NewSize begins with a +, it is interpreted as a request to increase the file system
size by the specified amount. If NewSize begins with a -, it is interpreted as a
request to reduce the file system size by the specified amount.

124


If the specified size does not begin with a + or -, but it is greater or smaller than
the file system current size, it is also a request to increase or reduce the file
system size.
The inline log size remains unchanged if this file system new size is the same as
the current file system size.
If the specified size is not evenly divisible by the physical partition size, it is
rounded up to the closest number that is evenly divisible. The volume group in
which the file system resides defines a maximum logical volume size and limits
the file system size. When a request to reduce the file system size is successful,
the logical volume should be equal to or smaller than the original LV size
depending on the requested file system size.

-a logsize=LogSize
Specifies the size for an inline log in MB. Starting with Version 5.3, a + or - can
be specified with LogSize. It is ignored if inLinelog not being used. It cannot be
greater than 10% the size of the file system for an extending request, or smaller
than 0.4% the size of the file system for a shrinking request.

Shrink an inline log
You can have an inline log in a Enhanced Journaled File system and possibly
extend or shrink it. Here are the general guidelines:
When new file system size is specified, but its inline log size is not specified,
the new log size will be adjusted to be extended or shrunk proportionally,
based on the specified extended/shrunk file system size. The log size
increased or reduced should not more than 40% of the file system size
increased or reduced.
When a new file system size is not specified and there is an inline log, if a
new logsize is specified the file system size might be changed to include the
new log size.

Shrink a file system to minimum size
There is no command to show exactly how much a file system can be shrunk
since the df command does not show the size of the metadata. In addition, the
freed space reported by the df command is not necessarily the space that can be
truncated by a shrink request due to file system fragmentation. A fragmented file
system may not be shrunk if it does not have enough free space for an object to
be moved out of the region to be truncated, and a shrink does not perform file
system defragmentation. In this case, the chfs command should fail with the
returned code 28. A way to work around this is shown in Figure 3-6 on page 126.
Enter the df command to get information about the file system and look at the
size of the Free field. That might be the maximum size (2324048 in this example)


125

you can shrink from the file system. Use the chfs -a command to shrink the file
system and then check the final size with the df command. Remember that a file
system can be extended or shrunk by a multiple of the physical partition size,
which is 16 MB in this example.

Figure 3-6 Shrink file system in command line

Note: Use the df command without -k, -m, -g, and the chfs command with a
unit of 512-byte blocks to shrink a file system to minimum size in the previous
example. Otherwise, the result can be different from what you expected.
Note that you cannot shrink a file system if the requested size is less than a
physical partition size. If you attempt to reduce the size with less than a physical
partition size, the request will be ignored, as shown in Figure 3-7 on page 127.

126


Figure 3-7 File system unchanged: Attempt to reduce the size

Three internal steps to shrink file system
There are three steps to shrink a file system, internally provided by the
command.
1. Validation to figure out if outside extents fit inside: The first step is to scan the
allocation map (bMap) for allocated space outside a fence and to compare it
to free space inside of the fence. If there is not enough free space, the
command will fail.
2. Shrink the file system: Extents outside the fence are relocated to free spaces
inside the fence.
3. Shrink LV: Finally, remove partitions in request.

Considerations
Table 3-3 provides a list of common questions and answers about this function.


127

Table 3-3 Things to know about file system shrink
When a file system is shrunk, will the logical volume that this file system
resides on be automatically shrunk?

Yes

When a logical volume is created, and before the file system is defined
for the logical volume, is there any way to shrink the logical volume only?

No

Can I shrink a standard JFS other than enhanced JFS (JFS2)?

No

Is it possible to shrink a an file system that has inline log in it?

Yes

Can I shrink an inline log while the size of the file system is unchanged?

Yes

During a shrink of the file system, are writes to the file system blocked?

Yes

Important: Shrinking a file system with snapshots is discouraged.

3.2.3 JFS2 logredo scalability
Logredo is the Journaled File System utility that applies all committed
transactions recorded in the journal log since the most recent sync point. Its goal
is to put the Journaled File System in the state it would have been in if all
transactions had been applied, in order, by the Journaled File System itself.
The job of the logredo is accomplished in one pass over the log, reading
backwards from log end to the first sync point encountered. This means that the
log entries are read and processed in Last-In-First-Out (LIFO) order. In other
words, the records logged latest in time are the first records processed.
AIX 5L Version 5.3 provides the following enhancements in the area of logredo to
improve performance and to support large numbers of file systems:
Support for minor numbers greater than 512 in a volume group
Support for copy-on-write and cached updates to reduce I/O activity
Support for shrink file system
Version 5.3 will allow a maximum of 4096 logical volumes in a volume group with
the scalable volume group option. In addition, JFS2 logredo now can take
advantage of copy-on-write and cached updates for performance reasons.
Shrink file systems capability is included in Version 5.3 and is supported by the
logredo utility.

128


Note: Logredo is called by the fsck and mount commands, and is not
recommended to be run as a stand-alone utility unless completely necessary,
and only by advanced users with a clear understanding of journaling in a file
system.

3.2.4 JFS2 file system check scalability
The fsck command in AIX checks file system consistency and interactively
repairs the file system if required. AIX 5L Version 5.3 enhanced the
implementation of the helper which specifically performs the file system check for
JFS2 file systems. The new code makes a better use of system resources and
includes algorithms that improve the scalability and performance.
The fsck command, in particular, must read a large amount of data from the file
system device. Access to this data and control of the memory used for buffering
was greatly simplified by mapping the device that the file system resides on. As
such the fsck command utilizes the new block device mapping capability
introduced in AIX 5L Version 5.3, which is described in more detail in 2.5, “Block
device mapping” on page 74.

3.2.5 JFS2 extended attributes Version 2 support
Extended attributes are an extension of the normal attributes of a file (such as
size and mode). They are (name, value) pairs associated with a file or directory.
The name of an attribute is a null-terminated string. The value is arbitrary data of
any length. There are two types of extended attribute: extended attribute version
1 (EAv1) and extended attribute version 2 (EAv2). Starting with AIX 5L Version
5.3, EAv2 with JFS2 is now available. Table 3-4 summarizes the difference
between EAv1 and EAv2. It should be noted that EAv2 is required to use an
NFS4 ACL.
Table 3-4 Difference between EAv1 and EAv2
EAv1

EAv2

No. of attributes

Max. 8 attributes

Architecturally unlimited

Size of attributes

4 KB per attribute

Max 16 TB per attribute

Name space

Only encoded name of
16 bits

String of max NAME_MAX length

User-defined
attributes support

System-defined
attributes only

Support both system-defined and
user-defined attributes


129

Setting extended attribute version 2
If you created a file named report1 and want to set attributes to the file such as
author, date, revision number, comments, and so on (SeonglulSon as author in
this example), you can accomplish this with the setea command to set the value
of an extended attribute and the getea command to read the value of an
extended attribute, as shown in Example 3-3.
#setea -n Name { -v Value | -d | -f EAFile } FileName ...
#getea [-n Name] [-e RegExp] [-s] FileName
Example 3-3 Setting and reading extended attributes
#setea -n Author -v SeonglulSon report1
#getea report1
EAName: Author
EAValue:
SeonglulSon

In addition, you can specify a file name instead of value name with the setea
command. Example 3-4 shows the command output.
Example 3-4 Configuring extended attributes with a file name
#getea report1
EAName: Author
EAValue:
SeonglulSon
#vi co-authors
Scott Vetter
Adrian Demeter
Armin Roell
Shiv Dutta
Seonglul Son
#setea -n Author -f co-authors report1
#getea report1
EAName: Author
EAValue:
Scott Vetter
Adrian Demeter
Armin Roell
Shiv Dutta
Seonglul Son

130


Compatibility
AIX 5L Version 5.3 continues to support EAv1 as the default format, and
provides an option to create a file system with EAv2 and a runtime command to
convert dynamically from EAv1 to EAv2 to create or access named attributes and
advanced ACL. However, once a file system is created with EAv2 or conversion
has been initiated, AIX 5L Version 5.2 cannot access the file system and
attempting to mount will result in an EFORMAT error. The compatibility is
summarized in Table 3-5.
Table 3-5 Compatibility of extended attributes
JFS2 file system with EAv1

JFS2 file system with EAv2

AIX 5L Version
5.2 or earlier

No change

Mount not allowed. Other file
system commands which
attempt to read the formatted file
system also not allowed.
(backbyinode, for example)

AIX 5L Version
5.3 or later

Migration to EAv2 is not required.
File system can be mounted with
no changes. Other file system
commands which attempt to read
the formatted file system are also
allowed.

New support

In order to change a JFS2 file system to EAv2, use the following command:
#chfs -a ea=v2 fs_name

3.2.6 JFS2 ACL support for NFS V4
Starting with AIX 5L Version 5.3, the Enhanced Journaled File system now
supports ACLs for NFS version 4. This allows you to establish fine-grained
access control for file system objects and support inheritance features. In order
to take advantage of ACL support for NFS V4, you need to create the JFS2 file
system with extended attribute format version 2 (EAv2). If you have file systems
that have already been created with extended attribute format version 1(EAv1),
you need to convert to EAv2 first, as shown in Figure 3-8 on page 132.


131

Figure 3-8 SMIT menu for converting EAv1 to EAv2

NFS4 ACLs using the command line
The JFS2 file system included with AIX supports two ACL types from AIX 5L
Version 5.3: AIXC and NFS4. The AIXC (AIX Classic) ACL type provides for the
ACL behavior as defined on previous releases of AIX. This ACL type consists of
the regular base mode bits and extended permissions. With extended
permissions, you can permit or deny file access to specific individuals or groups
without changing the base permissions. Example 3-5 shows the typical output of
acledit before converting AIXC ACL to NFS4.
Example 3-5 Output of the acledit command before converting to NFS4
#acledit /fs2
*
* ACL_type
AIXC
*
attributes:
base permissions
owner(root): r-x
group(system): r-x
others: r-x
extended permissions
disabled~

132


Now convert ACL with the aclconvert command. The usage of the aclconvert
command and the output of the acledit command after converting ACL to NFS4
are shown in Example 3-6.
Note: Before you convert AIXC to NFS4 or NFS4 to AIXC ACL, it is always
recommended to back up the original ACL or files. You can go back to AIXC
from NFS4, or NFS4 to AIXC ACL, but the resulting permissions are not
necessarily equivalent to the original ones.
Example 3-6 Output of the acledit command after converting to NFS4
# aclconvert -t NFS4 /fs2
# acledit /fs2
*
* ACL_type
NFS4
**
* Owner: root
* Group: system
*
s:(OWNER@):
d
wpDd
s:(OWNER@):
a
rRWxaAcCo
s:(GROUP@):
a
rx
s:(OWNER@):
d
rwpRWxDdos
s:(GROUP@):
d
rwpRWxDaAdcCos
s:(EVERYONE@): a
rRxac
s:(EVERYONE@): d
wpDd

If you use the acledit -v command instead of the acledit command with no
flags, you will be presented a lot of useful explanations of access control entry
(ACE) types, ACE access masks, and so on. Table 3-6 and Table 3-7 describe
what the keys mean.
Table 3-6 ACE type
Key

Description

a

ACE type allows the access described in the access mask

d

ACE type is denied the access described in the access mask

l

Audit type ACE

u

Alarm type ACE


133

Table 3-7 ACE access mask
Key

Description

r

Permission to read the data for the file or permission to list the contents
for a directory

w

Permission to modify the file's data or permission to add a new file to a
directory

p

Permission to append data to a file or permission to create a
subdirectory to a directory

R

Permission to read the named attributes of a file

W

Permission to write the named attributes of a file

x

Permission to execute a file

D

Permission to delete a file or directory within a directory

a

The ability to read basic attributes (non-ACLs) of a file (Read Base
Attributes)

A

Permission to change basic attributes (non-ACLs) of a file

d

Permission to delete the file

c

Permission to read the ACL

C

Permission to write the ACL

o

Permission to change the owner

NFS4 ACLs using Web-based System Manager
You can apply NFS4 ACLs using the Web-based System Manager. It may be
easier for you to understand since it is well organized with the GUI. Go to the
Web-based System Manager and choose File Systems → Overview and
tasks; you will find a new menu Access control list added in Version 5.3. It will
lead you to configure NFS4 ACL as shown in Figure 3-9 on page 135.

134


Figure 3-9 Configuring ACL through Web-based System Manager

You will be prompted to select a directory or file. This allows you to choose the
file whose ACL the selected operation will be performed on, as shown in
Figure 3-10.

Figure 3-10 File Access Control List dialog panel

After you choose a file or directory, the second File Access Control panel is
returned. Here you select the action you wish to perform, as shown in
Figure 3-11 on page 136.


135

Figure 3-11 File ACL dialog: Second panel

In this panel you can choose to edit ACL or to convert ACL to the selected ACL
type. Selecting Edit ACL will open the ACL edition dialog in the selected ACL
type shown in Figure 3-12 on page 137. The following columns are displayed in
this panel:
Identity
User type

The type of user which will use the ACE. It can be one of user,
group, or special. This is given by the prefix of the identity string.

ACE Type

The type of the ACE entry. It is one of the following: allow, deny,
alarm, or audit. This is given by the ACE type string.

Mask

136

Represents to whom this ACE applies. It is the rest of the string
following the prefix of the identity string.

Indicates the various access rights defined in the ACE,
expressed as a string of keys. Tips are added to those table cells
that display in clear the meaning of the flags.


Figure 3-12 ACL edition dialog

If you choose Edit and click the Access Mask tab, you will see the panel shown in
Figure 3-13 on page 138. These keys were defined in Table 3-7 on page 134.


137

Figure 3-13 Add ACE dialog: Access Mask tab

3.2.7 ACL inheritance support
AIX classical ACL (AIXC) does not support inheritance; only NFSv4 ACL
supports inheritance. When a new file or directory is created and the parent
directory has an NFSv4 ACL, then an ACL will be established for the new file or
directory. If an ACL is established for a child file or directory explicitly, it is
decoupled from the parent's ACL. Any changes to the parent's ACL will not affect
the child's ACL in this case. In an environment where users are not actively
managing ACLs, all the ACLs of a directory's child files will likely have the same
ACL. It makes sense to keep one copy of the ACL when the ACL data is
common.

Configuring ACL with inheritance
If the following conditions are met for a certain file system, inheritance feature
can be used:
1. It is a JFS2 file system with AIX 5L Version 5.3 or later.
2. Extended attribute version 2 (EAv2) is enabled with the JFS2 file system.

138


3. NFS4 ACL is applied.
When you configure ACL with the acledit command, there are fields to define
different inheritance flags such as fi and di. Table 3-8 describes the inheritance
flags.
Table 3-8 Keys for inheritance of NFS4 ACL
Key

Descriptions

fi

Can be placed on a directory and indicates that this ACE should be added to
each new non-directory file created.

di

Can be placed on a directory and indicates that this ACE should be added to
each new directory file created.

oi

Can be placed on a directory but does not apply to the directory, only to newly
created files/directories as specified by the previous two flags.

ni

For child only, no inherit for grandchild.

Suppose that you have a directory named dir1 and want to inherit ACL to the
sub-directories and files under the dir1 directory. Example 3-7 shows you how to
accomplish this.
First, you convert to NFS4 ACL for the dir1 directory and edit the ACL with the
acledit command. Modify ACL and add inheritance flags as needed. Then
create a new file named file1 and verify file1 with the aclget command to see if
the ACL of file1 is the same as that of dir1.
Example 3-7 Configuring ACL with inheritance
# aclconvert -tNFS4 dir1
# acledit dir1
*
* ACL_type
NFS4
*
*
* Owner: root
* Group: system
*
s:(OWNER@):
a
rwpRWxDaAdcCo
u:slson:
a
r

fidi
fidi

# touch dir1/file1
# aclget dir1/file1
*
* ACL_type
NFS4
*
*


139

* Owner: root
* Group: system
*
s:(OWNER@):
a
u:slson:
a

rwpRWxDaAdcCo
r

fidi
fidi

The inheritance as implemented in AIX 5L Version 5.3 is only valid for the newly
created files and directories under the parent directory, not for the existing files
and directories. if you want to inherit the ACL to all existing files and directories
as well, then use the aclget parent_directory | aclput -R directory
command.
Note: Before you convert AIXC to NFS4 or NFS4 to AIXC ACL, it is always
recommended to back up the original ACL or files. You can go back to AIXC
from NFS4 or NFS4 to AIXC ACL, but the resultant permissions are not
necessarily equivalent to the original ones.

140


4

Chapter 4.

Reliability, availability, and
serviceability
This chapter includes descriptions of the following enhancements for AIX 5L:
Error logging, core files, and system dumps
– Error log RAS
– Enhancements for a large number of devices
– Core file creation and compression
– System dump enhancements
– DVD support for system dumps
– snap command enhancements
Trace enhancements
– Administrative control of the user trace buffers
– Single thread trace


141

4.1 Error log RAS
Under very rare circumstances, such as powering off the system exactly while
the errdemon is writing into the error log, the error log may become corrupted. In
AIX 5L Version 5.3 there are minor modifications made to the errdemon to
improve its robustness and to recover the error log file at its start.
When the errdemon starts, it checks for error log consistency. First it makes a
backup copy of the existing error log file to /tmp/errlog.save and then it corrects
the error log file, while preserving consistent error log entries.
The difference from the previous versions of AIX is that the errdemon used to
reset the log file if it was corrupted, instead of repairing it.
If the error log copy to the /tmp/errlog.save fails, the CORRUPT_LOG and the
LOG_COPY_FAILED or LOG_COPY_IO entry is written to the error log. These
entries will be logged using the errlog() function. They will be logged along with
other entries waiting in the buffer. Note that, in many corruption scenarios, the
system can continue to write entries to the error log. The main reason the copy
would fail is insufficient space in /tmp.
If the error log was copied and repaired, the CORRUPT_LOG error entry is
written to the error log.

4.2 Enhancements for a large number of devices
For each device configured, an entry is made in the /dev directory. On systems
with many devices, it may be possible for the system to run out of space in the
root file system or to run out of inodes. Prior to Version 5.3, you would not know
the cause of errors related to configuring a large number of devices since that
error would not be reported. Starting with Version 5.3, the cfgmgr command now
reports this error to the user. If an error is detected, the cfgmgr command
determines if the problem is lack of space or inodes and sets a flag. If this occurs
during runtime, the cfgmgr command displays an error message and exits. If this
scenario occurs during boot, AIX makes an entry in the boot log to record the
event and continues. Before the cfgmgr command exits, if the flag is set, it
displays the error message.

Error message support
If you have an error message related to configuring a large number of devices, it
appears similar to the following:
cfgmgr: 0514-624 Unable to define or configure one or more devices
because the filesystem is full or is out of inodes

142


If a system runs into this problem when a user runs the cfgmgr command at the
command line, the user will see the error message on the console. If the system
runs into this problem during boot, the user can view the boot log with the alog
-ot boot command.

Configuring a large number of devices
The number of devices that AIX can support varies from system to system,
depending on several important factors. The following factors have an impact on
the file systems that support the devices:
Configuring a large number of devices requires storage of more information in
the ODM device-configuration database. It can also require more device
special files. As a result, more space and more inodes are required of the file
system.
Some devices require more space than others in the ODM
device-configuration database. The number of special files or inodes used
also varies from device to device. As a result, the amount of space and
inodes required of the file system depends on the types of devices on the
system.
Multipath I/O (MPIO) devices require more space than non-MPIO devices
because information is stored in the ODM for the device itself as well as for
each path to the device. As a rough guideline, assume that each path takes
up the space of one-fifth of a device. For example, an MPIO device with five
paths will have the space equivalent to two non-MPIO devices.
AIX includes both logical devices and physical devices in the ODM
device-configuration database. Logical devices include volume groups,
logical volumes, network interfaces, and so on. In some cases, the
relationship between logical and physical devices can greatly affect the total
number of devices supported. For example, if you define a volume group with
two logical volumes for each physical disk that is attached to a system, this
will result in four AIX devices for each disk. On the other hand, if you define a
volume group with six logical volumes for each physical disk, there will be
eight AIX devices for each disk. Therefore, only half as many disks could be
attached.
Changing device attributes from their default settings results in a larger ODM
device-configuration database and could lead to fewer devices that can be
supported.
More devices means more real memory is required.
Two file systems are used by AIX to support devices:
The RAM file system is used during boot in an environment that has no
paging space and no disk file systems mounted. The size of the RAM file
system is 25 percent of the system memory size up to a maximum of 128 MB.

Chapter 4. Reliability, availability, and serviceability

143

One inode is allocated for every KB in the RAM file system. If the system
memory size is 512 MB or larger, then the RAM file system will be at its
maximum size of 128 MB with 131072 inodes. If either the amount of RAM file
system space or number of inodes needed to support the attached devices
exceeds what has been allocated to the RAM disk, the system might not boot.
If this is the case, you must remove some of the devices.
The space and inodes of the root file system (rootvg) on the disk can be
increased as long as there are unallocated partitions in the rootvg. With the
maximum RAM file system size, it is likely that up to 25,000 AIX devices could
be configured. These numbers include both physical and logical devices.
Depending on the various factors mentioned in this section, your system
might be able to configure more or fewer devices than this number.

4.3 Core file creation and compression
Applications running on the AIX system may produce quite large core files.
The changes to the core file creation introduced in Version 5.3 are the following:
Core file compression
Enabling core file destination directory and core file naming
To check the core file creation settings or change them you can use two new
commands:
lscore

Checks the current or default settings

chcore

Changes the core file creation settings

Example 4-1 shows how to use the commands to check and change the core file
creation settings. The discussion that follows shows the steps you probably will
use in your environment.
Example 4-1 Managing core file creation settings
# lscore
compression: off
path specification: off
corefile location: not set
naming specification: off
# mkdir /tmp/cores
# chcore -c on -p on -l /tmp/cores -n on
<logout and login>
# lscore
compression: on

144


path specification: on
corefile location: /tmp/cores
naming specification: on
# sleep 1000 &
[1]
360620
# kill -11 %1
#
[1] + Memory fault(coredump)
sleep 1000 &
# ls -l /tmp/cores
total 16
-rw-r--r-1 root
system
5418 Jun 22 14:44 core.360620.22194424.Z
#

First we check the settings using the lscore command. In the example, we start
from scratch and the displayed settings are the default ones.
Log in as the user you want to change the settings for. Use the mkdir
/tmp/corefiles command to create the target directory for the core files and
check that the directory is writable to the user. Use the chcore -c on -p on -l
/tmp/corefiles -n on command to change the core file creation settings. The
basic flags have the following meanings:
-c on/off/default

Turns the compression on, off, or returns to default.

-p on/off/default

Turns the use of destination directory on, off, or returns to
default.

-l directory

Specifies the target directory where core files are created.
The directory must exist before you run the command and
must be specified using fully qualified name.

-n on/off/default

This flag sets the core file naming as if set by the
CORE_NAMING environment variable.

After you make changes to the core file creation settings, users need to log out
and log in again in order to apply the change.
The easiest way to test the new settings is to start any process, such as sleep
1000 & in the background and kill this process by the kill -11 %1 command
using signal SIGSEGV. The core file is created in the defined directory and its
name is created from the process ID and the timestamp: core.PID.timestamp.Z.
The compressed core file is compressed by the LZW (Lempel-Zev-Welch)
algorithm and can be decompressed using the uncompress command if needed.


145

Note: If you uncompress the compressed core file and compress it back using
the compress command, the file size will differ. This is because the in kernel
core file compression stores some information in raw format. This is a correct
behavior.
With the chcore command you can specify the user you want to change. The
command can be run by any user, but only root can change settings for other
users. After the settings are changed, an entry to /etc/security/user is added:
root:
...
core_compress = on
core_path = on
core_naming = on
core_pathname = /tmp/cores

If the -R load_module is used the core settings are stored in an external security
repository, like LDAP, DCE, or GSA, instead of the default /etc/security/user file.
The load_module has to be supplied by the external program.
If the core is generated by the gencore command, the core file naming and path
may be omitted depending on the file name specified to the command.
Specifying the full path name will omit the core file naming and path settings. For
example, the gencore PID /path/filename command will omit the core file
naming and path, but the gencore PID filename command will observe the core
file naming and path settings.

4.4 System dump enhancements
System dumps are enhanced in AIX 5L Version 5.3 to provide more information
about the status of the dumps. These enhancements include:
Dump information to TTY
Verbose flag to the sysdumdev command
The dumpfmt command formatting

4.4.1 Dump information to TTY
During the creation of the system dump, additional information is displayed on
the TTY about the progress of the system dump. Example 4-2 on page 147
shows a sample output to the TTY console.

146


Example 4-2 TTY output from system dump
# sysdumpstart -p
Preparing for AIX System Dump . . .
Dump Started .. Please wait for completion message
AIX Dump .. 23330816 bytes written - time elapsed is 47 secs
Dump Complete .. type=4, status=0x0, dump size:23356416 bytes
Rebooting . . .

At the time of writing, the kernel debugger and the 32-bit kernel must be enabled
to see this function, and we tested the function only on the S1 port. However, this
limitation may change in the future.

4.4.2 Verbose flag for the sysdumpdev command
Following a system crash there exist scenarios where a system dump may crash
or fail without one byte of data written out to the dump device, for example power
off or disk errors. For cases where a failed dump does not include the dump
minimal table, it is useful to save some trace back information in the NVRAM.
Starting with Version 5.3, the dump procedure is enhanced to use the NVRAM to
store minimal dump information.
If the dump fails, use the sysdumpdev -vL command to check the reason for the
failure. Example 4-3 and Example 4-4 on page 148 show samples of failed
dumps.
Example 4-3 sysdumpdev -vL output in case of I/O error during dump
# sysdumpdev -Lv
0453-039
Device name:
/dev/hd6
Major device number: 10
Minor device number: 2
Size:
262144 bytes
Uncompressed Size:
4491225 bytes
Date/Time:
Fri Jun 25 16:04:19 CDT 2004
Dump status:
-4
I/O error
Dump copy filename: /var/adm/ras/vmcore.2.Z
End Date/Time:
Fri Jun 25 16:04:25 CDT 2004
Dump Duration:
6s
Traceback:
iar:000d88bc
dmpnow+28
addr:000d88c8
dmpnow+34
addr:0037e21c
dmpioctl+23c
addr:002cde68
rdevioctl+c8
addr:003c0524
spec_ioctl+68


147

addr:002e032c
Dump History:
Cdt:
Entry:
Places:
Leds Line 1:
Leds Line 2:

vnop_ioctl+24
ldr
ldr32l
jwrite
0C2 0C1
CDT: thrd
DUMP TABLE
CDT: ldr
5
5
5
CDT: thrd

3

In Example 4-3 we initiated a system dump using the sysdumpstart -p command
and while the system dump was writing to the disk we pulled out the disk. The
sysdumpdev -Lv command after a failed system dump shows us information that
the I/O operation had failed and provides trace back information for analysis.
Example 4-4 sysdumpdev -vL output in case of power off during dump
root@Server4: sysdumpdev -Lv
0453-039
Device name:
/dev/hd6
Major device number: 10
Minor device number: 2
Size:
0 bytes
Uncompressed Size:
17717 bytes
Date/Time:
Fri Jun 25 16:16:38 CDT 2004
Dump status:
-3
dump crashed or did not start
End Date/Time:
Fri Jun 25 16:16:38 CDT 2004
Dump Duration:
0ns
Traceback:
iar:000d88bc
dmpnow+28
addr:000d88c8
dmpnow+34
addr:0037e21c
dmpioctl+23c
addr:002cde68
rdevioctl+c8
addr:003c0524
spec_ioctl+68
addr:002e032c
vnop_ioctl+24
Dump History:
Cdt:
dmp_minimal
Entry:
dbgtbls
Places:
jwrite
jwrite
jwrite
Leds Line 1:
0C2

148


Leds Line 2:
DUMP
DUMP
CDT:
DUMP
CDT:

ON CPU 1
TABLE
0
dmp_minimal
TABLE
1
proc

In Example 4-4 we initiated the system dump and pressed the power button while
the dump was running. In this case the sysdumpdev shows different dump status
and dump history.
Tip: During the dump the LCD display shows progress codes and information.
If you could not see the LCD you can get the codes from the sysdumpdev -Lv
command output. See “Leds Line 1:” and “Leds Line 2:” sections of
Example 4-3 or Example 4-4.
In the unlikely event a dump crashes, a second dump traceback is displayed
instead of the dump history.

4.4.3 dmpfmt command changes
The dmpfmt -c command is extended to verify that certain dump components are
present in the dump in addition to the dump integrity test. The dmpfmt -c
command is executed by the system at boot time and its output is added to the
DUMP_STATS error log entry.

4.5 DVD support for system dumps
This enhancement adds the possibility to:
Copy system dumps to DVD media
Use DVD as a primary or secondary dump device
Enhance the snap command to handle system dumps on DVD
In order to support this function the target DVD device should be DVD-RAM or
writable DVD. Remember to insert an empty writable DVD in the drive when
using the sysdumpdev or snap commands, or when you require the dump to be
copied to the DVD at boot time after a crash. If the DVD media is not present the
commands will give error messages or will not recognize the device as suitable
for system dump copy.


149

4.5.1 Copy system dumps to DVD media
Figure 4-1 shows the dialog you get at boot time. This dialog prompts you to
select the target device to be used to receive the dump when there is insufficient
space in the copy directory, usually in the /var file system.

Figure 4-1 Copy system dump to media boot time dialog

Select the DVD device as the target device and the system will copy the dump to
the DVD using the pax command format.
After the system dump is successfully copied to the DVD media, you can check
or restore the dump from the system prompt using the pax command as follows:
# pax -v -f /dev/rcd0
PAX format archive
-rwx--x--- 0 root

system

22884864 Jun 28 17:29 dump_file.Z

Setting the dump device to DVD
The sysdumpdev command as modified in AIX 5L Version 5.3 accepts DVD
devices as the primary or secondary dump device. The following example shows
how to set the primary dump device to /dev/cd0 and the secondary to /dev/cd1:
sysdumpdev -p /dev/cd0 -s /dev/cd1

If the system dump destination is a DVD, then the system dump should not
exceed one DVD media size.

150


Using DVD for the snap command
The snap command is enhanced in that it enables the DVD device to be selected
as source for the dump and that its -o flag can accept the DVD device as an
output device. Use the snap command to generate its output to DVD as follows:
# snap -a -o /dev/cd0
...
The previous dump to paging space was not copied to disk at reboot.
You were given the opportunity to copy the dump to external media.
Would you like to restore the dump? (y/n) y
Please enter external device(default /dev/rmt0): /dev/cd0
Setting input device to /dev/cd0... done.
Please insert the media containing the dump into /dev/cd0 and press Enter
done.
...
Please remove dump media and replace with snap output media.
Press enter to continue
...
Copying information to /dev/cd0... Please wait... done.
...

The snap uses the pax format to copy the data to the DVD. You can check the
DVD content using the pax -v -f /dev/cd0 command.

4.6 snap command enhancements
The snap command already exists in the previous versions of AIX, but Version
5.3 extends its function in using external scripts, enabling the snap command to
split up the output pax format file into smaller pieces or extending the collected
data.

Extending snap to run external scripts
The scripts can either be parameters to the snap command, or in a case that the
all option is specified, the scripts are expected to be in the
/usr/lib/ras/snapscripts directory. The snap command can call the scripts in three
different way:
Specifying the script name in the /usr/lib/ras/snapscripts directory that the
snap will call.
Specifying the all keyword to the snap command. The snap command then
calls all scripts in the /usr/lib/ras/snapscripts directory.


151

Specifying the file:filename keyword with filename reference to the file
containing the list of external scripts. The file must contain one script name
per line.
First, prepare a sample script in the /usr/lib/ras/snapscripts directory and name it,
for example, test_snap_script.ksh. We created just a short script that writes the
environment to the log file.
# cat /usr/lib/ras/snapscripts/test_snap_script.ksh
#!/usr/bin/ksh
case $PASSNO in
"1") echo "4096" > $SCRIPTSIZE
;;
"2") set > $SCRIPTLOG
;;
*)
echo "ERROR IN THE ENVIRONMENT"
esac
exit 0

Then we ran the snap command and checked the output of the script. The output
is placed in the /tmp/ibmsupt/scriptname directory.
# snap test_snap_script.ksh
********Checking and initializing directory structure
Creating /tmp/ibmsupt directory tree... done.
Creating /tmp/ibmsupt/test_snap_script.ksh directory tree... done.
Creating /tmp/ibmsupt/testcase directory tree... done.
Creating /tmp/ibmsupt/other directory tree... done.
********Finished setting up directory /tmp/ibmsupt
Checking Space requirement for test_snap_script.ksh
Checking for enough free space in filesystem... done.
Gathering test_snap_script.ksh data
# ls -al /tmp/ibmsupt/test_snap_script.ksh
total 16
-rw------1 root
system
0
test_snap_script.ksh.err
-rw------1 root
system
900
test_snap_script.ksh.log
-rw------1 root
system
0
test_snap_script.ksh.out
-rw------1 root
system
5
test_snap_script.ksh.size

152


Jun 24 16:24
Jun 24 16:24
Jun 24 16:24
Jun 24 16:24

The system environment that is passed to the script from the snap command
contains the following variables:
SNAPDIR

The destination directory for all snap command output.

PASSNO

Pass number of the script. The snap command calls the
third party script twice and tells the script the phase (1 or
2) it is calling the script.

SCRIPTSIZE

The path to a file where the script is expected to write the
estimated size during the first pass.

SCRIPTLOG

The path to an output log file to which the script is
expected to write.

Note: The snap command calls the third party script in two phases. In the first
phase the script is expected to prepare for the second run and calculate the
estimated size of the output data. The second phase should gather the
necessary data and store it in the $SNAPDIR directory.

The snapsplit command
The snapsplit command is introduced in Version 5.3. The command splits the
snap.pax.Z file into smaller files or joins together split snap files. The command
expects to be run from the /tmp/ibmsupt directory, where the snap.pax.Z file
resides. In Example 4-5 we show the use of the command to split the snap into 4
MB files, and the command to rejoin the split files.
Example 4-5 snapsplit command
# snapsplit -s 4
was split successfully ...
#ls -al
...
-rw------1 root
system
-rw-r--r-1 root
system
snap.sq1test1.062504143158.pax.Zaa
-rw-r--r-1 root
system
snap.sq1test1.062504143158.pax.Zab
...
# snapsplit -u -T 062504143158
Restoring . . .
was successfully restored ...
# ls -al
-rw------1 root
system
-rw-r--r-1 root
system
snap.sq1test1.062504143158.pax.Z
-rw-r--r-1 root
system
snap.sq1test1.062504143158.pax.Zaa

7179388 Jun 25 13:03 snap.pax.Z
4194304 Jun 25 14:31
2985084 Jun 25 14:31

7179388 Jun 25 13:03 snap.pax.Z
7179388 Jun 25 14:39
4194304 Jun 25 14:31


153

-rw-r--r-1 root
system
2985084 Jun 25 14:31
snap.sq1test1.062504143158.pax.Zab

The snapsplit command created the snap.hostname.timestamp.pax.Zxx files of
size 4 MB or less. The snapsplit -u command rejoined the files to
snap.hostname.timestamp.pax.Z file. You can take the timestamp for the -T flag
from the name of the split files.
The -T or -h flags available with the snapsplit command enable you to handle
snap dumps from different systems taken at different times. The -f flag enables
you to handle renamed snap files.

Splitting the snap output file with the snap command
The size of the snap output file can be a problem. There is a new flag, -O
megabytes, introduced in Version 5.3 that enables a split of the snap output file.
The snap command calls the snapsplit command. You can use the flag as
follows to split the large snap output into smaller 4 MB files:
# snap -a -c -O 4

Additional data collected by snap
In Version 5.3, the snap command collects additional information about the
system, or the output is reorganized. For example, lslpp -La, sysdumpdev -Lv,
lsdev -Ccscsi is collected to lsdev.scsi; lsdev -Ccproc is collected to the
processor.log; bindprocessor -q is collected to the processor.log; and lsvpd is
collected to lsvpd.out.

4.7 Administrative control of the user trace buffers
In previous versions of AIX, the trace buffer size for a regular user is restricted to
a maximum of 1 MB. This was done to prevent a malicious user using up all the
memory to be as pinned and consequently cause failures on the system. Also, in
the early days of computing, this restriction was necessary because memory was
at a premium.
This size threshold is working out to be a limitation for some applications, such
as MQ Series, to monitor their messaging system, and consequently affects their
serviceability.
Version 5.3 allows the users belonging to the system group to set the buffer
threshold for the user trace buffers. The threshold is set for the system. To do
this, the trcctl command is introduced and a menu is added to SMIT. The
settings are stored in the SWservAt ODM class.

154


The following example shows how to use the trcctl command. First we check
the current settings:
# trcctl -l
Default Buffer Size: 262144
Default Log File Size: 2621440
Default Log File: /var/adm/ras/trcfile
Non-Root User Buffer Size Maximum: 1048576

We change the log file size to 40 MB using the -L flag, the maximum buffer size
that a non-root user can specify to 20 MB using the -N flag, the path and file
name of the trace file to /tmp/tracefile using the -o flag and the default trace
buffer size that the trace command uses to 1 MB with the -T flag.
# trcctl -L 40M -N 20M -o /tmp/tracefile -T 1M

We check the new settings using the -l flag.
# trcctl -l
Default Log File: /tmp/tracefile

Finally, we show how to return the settings back to default using the -r flag. The
defaults differ if you use 32-bit kernel and 64-bit kernel. Our example shows the
case of the 64-bit kernel.
# trcctl -r
# trcctl -l
Default Log File: /var/adm/ras/trcfile

To ease the use of the trcctl command the SMIT is updated with a new menu
Manage Trace under the smit trace fastpath, followed by two sub menus that
change the settings or restore the default. Figure 4-2 on page 156 shows the
smit cngtrace menu.


155

Figure 4-2 smit cngtrace SMIT menu

4.8 Single thread trace
In versions prior to Version 5.3, the AIX system trace traced the entire system.
The trace command in Version 5.3 is enhanced with new flags. These flags
enable the trace to run only for specified processes, threads, or programs. This
will save space in the trace file and also helps to focus on just the area you are
interested in.
The new flags introduced to the trace command have the following meanings:
-A

Sets process trace for the listed processes following the flag.

-t

Sets threads to be traced.

-I

Specifies that interrupt-level events are to be traced.

-P

Specifies if the trace is propagated to child processes or threads.

-x

Specifies the trace is to be run for the specified program. After the
program ends, the trace ends.

-X

Specifies the trace is to be run for the specified program. After the
program ends, the trace continues to run.

To trace one specific command, its child processes and threads, start the trace
command as follows:
# trace -J fact -x /ad01/fop_test -P p -a
# trcrpt -O exec=y -O pid=y -O tid=y -O svc=y -O timestamp=1
...

156


In this example the trace is started for the /ad01/fop_test program and the file
activities are traced. The trace stops immediately after the program ends. Then
the trace report can be generated using the trcrpt command.
If you started the trace with the -X flag instead of the -x flag, the trace does not
stop when the traced program ends. Do not forget to stop the trace after you are
done.
The following is another example where we start a trace that traces two
processes, inetd and sendmail:
# ps -ef | grep inetd
root 131266 225456 0
Jun 22
- 0:00 /usr/sbin/inetd
# ps -ef | grep sendmail
root 184494 225456 0
Jun 22
- 0:00 sendmail: accepting
connections
# trace -J fact -A '131266 184494' -P p -a
# trcstop
# trcrpt -O exec=y -O pid=y -O tid=y -O svc=y -O timestamp=1
...

In this example we check for the PID of the two processes we intend to trace and
start the trace for those two PIDs. In this case, the trace is not stopped
automatically when the process stops and we need to use the trcstop command
to stop it. Use the trcrpt command to generate the trace report.
When you want to generate thread-specific traces using the -t flag, you need to
supply the TID additionally to the -t flag. You can get the TID from, for example,
the ps -efm -o THREAD command.
Using the -I flag you can add additional information to the trace containing the
interrupts.
In addition to the new flags of the trace command, the smit trcstart menu is
updated by items reflecting the new flags. See Figure 4-3 on page 158 for a
sample smit trcstart menu.


157

Figure 4-3 smit trcstart SMIT menu

158


5

Chapter 5.

System management
AIX 5L provides many enhancements in the area of system management and
utilities. This chapter discusses these enhancements. Topics include:
InfoCenter for AIX 5L Version 5.3
Multiple desktop selection from BOS menus
Erasing hard drive during BOS install
Service Update Management Assistant
Long user and group name support
Dynamic reconfiguration usability
Paging space garbage collection
Dynamic support for large page pools
Interim Fix Management
List installed filesets by bundle
Configuration file modification surveillance
DVD backup using the mkdvd command
NIM security
High Available NIM (HA NIM)
General NIM enhancements


159

5.1 InfoCenter for AIX 5L Version 5.3
Starting with AIX 5L Version 5.3, IBM Sserver pSeries and AIX documentation
will be available in one of two information centers: the IBM Sserver pSeries and
AIX Information Center on the Web, and the AIX Information Center on the
documentation CD. Figure 5-1 shows the AIX Information Center on the
documentation CD.

Figure 5-1 AIX Information Center on the documentation CD

The AIX and pSeries Information Center is more than a portal to documentation.
From this Web site, you can access the following tools and resources:
http://publib16.boulder.ibm.com/pseries/en_US/infocenter/base/

A message database that shows what error messages mean and, in many
cases, how you can recover. This database also provides information for LED
codes and error identifiers.
How-to tips with step-by-step instructions for completing system administrator
and user tasks.
FAQs for quick answers to common questions.
The entire AIX software documentation library for Version 5.1, Version 5.2,
and Version 5.3. Each publication is available in PDF format, and abstracts
are provided for books for Version 5.2 and Version 5.3.

160


Centralized information previously located throughout the library and easier
access to information about some new AIX functions:
– A new selection in the navigation bar centralizes all partitioning
information, including planning, installation, and implementation
information for partitioned-system operations.
– A new Advanced Accounting publication is available. Advanced
Accounting offers increased flexibility, allowing users to customize it to
meet their needs. Advanced Accounting allows for the collection of data
that is not necessarily associated with a user, but may be associated with
specific, user-defined projects. Interval Accounting is now possible, so that
users can collect accounting data from long-running jobs while the job is
running.
– A new Partition Load Manager for AIX Guide and Reference provides
experienced system administrators with information about how to perform
such tasks as installing, configuring, and managing Partition Load
Manager for AIX. This guide also provides the administrator with reference
information about commands and files that are used to run and manage
Partition Load Manager for AIX.
– Links to the entire pSeries and p5 hardware documentation library.
– A resources page that links users to other IBM and non-IBM Web sites
proven useful to system administrators, application developers, and users.
– Links to related documentation from IBM, including white papers, IBM
Redbooks, and technical reports on topics such as RS/6000, SP, and
HACMP for AIX. Release Notes and readme files are also available
through the information center.
– Several new videos are available for customer-installable features and
customer-replaceable parts.
– A new application, the AIX Information Center, will be available for
installation beginning with AIX 5L Version 5.3. The information center will
provide navigation and search capabilities for all installed AIX 5L Version
5.3 publications. The information center will be included on the AIX
Documentation CD. It can be installed and used on a local system or
installed on a documentation server for intranet use. The information
center is powered by Eclipse technology.

5.2 Multiple desktop selection from BOS menus
Prior to Version 5.3, you can choose only one of three desktops during BOS
install (CDE, GNOME, or KDE). If you need any other desktops, you can add
them after rebooting the system. Starting with Version 5.3, multiple desktop

Chapter 5. System management

161

selection from the BOS install menus is available. The available choices are
CDE, GNOME, and KDE.

Changes to the BOS menus
The Install Options menu remains the same. The Install More Software menu is
enhanced with two additional prompts:
If CDE is the desktop selected from the Install Options menu, prompts would
appear similar to the following:
4. GNOME Desktop (Toolbox for Linux Application)........No
5. KDE Desktop (Toolbox for Linux Application)..........No

If GNOME is the selected desktop:
4. CDE Desktop (Volume 2)..........No
5. KDE Desktop (Toolbox for Linux Application)..........No

If KDE is the selected desktop:
4. CDE Desktop (Volume 2)..........No
5. GNOME Desktop (Toolbox for Linux Application)........No

Each prompt to install an additional desktop defaults to No. If no desktop is
chosen (Desktop = "NONE" option on the Install Options menu), the prompts for
an additional desktop are not displayed. If the INSTALL_TYPE = CC_EVAL or
the CONSOLE is not /dev/lft0, the prompts for an additional desktop are not
displayed.

Login screens
The existing implementations of the three desktops (CDE, GNOME, or KDE)
each provide an interface to allow selection of another desktop manager.
From the GNOME login screen, you can select Session and access a menu
of possible sessions. CDE and KDE are displayed on the session menu if
they are available.
From the KDE login screen, you can select CDE or GNOME as well as KDE.
From the CDE login screen, you can select the option button and then select
Session to get a menu of session types that includes GNOME and KDE if
they are installed.

5.3 Erasing hard drive during BOS install
AIX 5L Version 5.3 now provides hard disk erasure of the installation drives at
install time through the BOS install menus and through the bosinst variables.

162


When you choose to erase hard drives, the erasure process will be internally
performed by the diag command.

Changes to BOS menus
The following changes were made to the BOS install menus.
For the installation case
If you are doing an overwrite install and you select the disks to install to, then
you will see a new option 55 More Disk Options which will take you to the
Erasure Options for Disks menu. It will in turn continue on from there
normally.
For the erase and not install case (the maintenance case)
The maintenance menu will have option 4 Erase Disks, which will take the
user to the Select Disk(s) That You Want to Erase menu. Continuation from
there will take the user to the Erasure Options for Disks menu. It will be the
last menu, and as such will then run the erase and exit.

New erasure menu
The new erasure menu is shown in Example 5-1.
Example 5-1 New erasure menu
Erasure Options for Disks
Select the number of times the disk(s) will be erased, and select the
corresponding pattern to use for each disk erasure. If the number of patterns
to write is 0 then no disk erasure will occur. This will be a time consuming
process. Either type 0 and press Enter to continue with the current settings,
or type the number of the setting you want to change and press Enter.
1
2
3
4
5
6
7
8
9

Number of patterns to write...............
Pattern #1................................
Pattern #2................................
Pattern #3................................
Pattern #4................................
Pattern #5................................
Pattern #6................................
Pattern #7................................
Pattern #8................................

0
00
ff
a5
5a
00
ff
a5
5a

>>> 0 Continue with choices indicated above
88 Help ?
99 Previous Menu
>>> Choice[0]:


163

New disk selection menu
Example 5-2 shows a new disk selection menu very similar to the one where
users select the disks to install to. This is used when the user is selecting to
erase the drives but not to install them.
Example 5-2 New disk selection menu
Select Disk(s) That You Want to Erase
Type one or more numbers for the disk(s) to be used for erasure and press
Enter. To cancel a choice, type the corresponding number and press Enter. At
least one disk must be selected. The current choices are indicated by >>>.
Name
hdisk0
hdisk1

>>> 1
2
>>> 0

Location Code
10-80-00-0,0
10-80-00-0,4

Size
4303
4303

VG Status
rootvg
none

Bootable
Yes
Yes

Maps
No
No

Continue with choices indicated above

66

Devices not known to Base Operating System Installation

77

Display More Disk Information

88

Help ?

99

Previous Menu

>>> Choice[0]:

New and changed bosinst.data entries
No changes are needed to the target disk stanza. These stanzas will be used to
also designate the drive to erase. There is a way to designate a set to install and
a different set to erase. Table 5-1 lists the new and changed bosinst entries that
will be under the control_flow stanza.
Table 5-1 New and changed bosinst.data entries
Stanza
ERASE_PATTERNS

164

Description
Specifies the patterns to write to the chosen hard drives. The
value for this field is a comma separated list of the patterns to
use for each erasure of the drives. A valid pattern is a
hexadecimal value from 0 to ffffffff. The number of patterns
specified must be equal to or greater than the number of
iterations specified in ERASE_ITERATIONS. If
ERASE_ITERATIONS is 0 then this field is ignored. Ex: If
ERASE_ITERATIONS = 3 then a valid entry for this field could
be ERASE_PATTERNS = 00, ff, 0a0a0a0a.


Stanza

Description

ERASE_ITERATIONS

Specifies the number of times to erase the chosen hard drives
before the installation occurs. This field is only valid when the
INSTALL_METHOD field is set to overwrite or erase_only.
The choices for this field are the numbers from 0 to 8. If the
field is set to 0 then no erasure of the hard drives will occur.
The default is 0.

INSTALL_METHOD

A new value besides overwrite, migrate, and preserve will be
added. It will be erase_only, which will only erase the drives
and not do an installation.

5.4 Service Update Management Assistant
AIX 5L development is highly focused on implementing tools and functions which
help to fulfill IBM’s Autonomic Computing strategy. If you apply the Autonomic
Computing paradigm to the area of software maintenance you envision an
environment where most, if not all, tasks related to fix and maintenance level
management are automated and no longer require human intervention.
In a first approach, AIX 5L Version 5.2 provided proactive capabilities through the
compare_report command. This command and its SMIT interface allow the
comparison of installed software or fix repositories to a list of available fixes from
the IBM support Web site, enabling system administrators to develop a proactive
fix strategy.
AIX 5L Version 5.3 advances one step further and introduces automatic
download, scheduling, and notification capabilities through the new Service
Update Management Assistant (SUMA) tool. SUMA is fully integrated into the
AIX Base Operating System and supports scheduled and unattended task-based
download of Authorized Program Analysis Reports (APARs), Program
Temporary Fixes (PTFs), and recommended maintenance levels (MLs). SUMA
can also be configured to periodically check the availability of specific new fixes
and entire maintenance levels, so that the time spent on such system
administration tasks is reduced. The SUMA implementation allows for multiple
concurrent downloads to optimize performance and has no dependency on any
Web browser.

5.4.1 Packaging and installation
The Service Update Management Assistant is available by default after any AIX
5L Version 5.3 operating system installation. All SUMA modules and the suma
command are contained in the bos.suma fileset. SUMA is implemented in the


165

Perl programming language, so the Perl library extensions fileset perl.libext and
the Perl runtime environment fileset perl.rte are prerequisites to bos.suma.
The perl.libext fileset holds additional Perl modules which allow access to
AIX-specific function from native Perl code. The provided extensions yield a
significant performance improvement over system calls.
SUMA directly benefits from enhancements to the Perl runtime environment in
AIX 5L Version 5.3. The new base AIX Perl distribution (perl.rte) adds nine new
Perl modules which enable programmatic Internet access and provide the ability
to parse and generate XML code.
The lslpp -p bos.suma command can be used to verify the requisites for the
bos.suma fileset:
# lslpp -p bos.suma
Fileset
Requisites
bos.suma 5.3.0.0
*prereq bos.rte 5.1.0.0
*prereq perl.rte 5.8.2.0
*prereq perl.libext 2.1.0.0
Path: /etc/objrepos
bos.suma 5.3.0.0

*prereq bos.rte 5.1.0.0
*prereq perl.rte 5.8.2.0
*prereq perl.libext 2.1.0.0

Use the lslpp -f bos.suma command to examine the list of directories and files
which are required to install the SUMA feature.
# lslpp -f bos.suma
Fileset
File
bos.suma 5.3.0.0
/usr/suma/lib/SUMA/FixInventory.pm
/usr/suma/bin/suma_fixinv
/usr/suma/lib/SUMA/Policy.pm
/usr/suma/lib/SUMA/Download.pm
/usr/suma/lib/msg.map
/usr/suma/bin/sm_suma
/usr/suma/lib
/usr/suma/lib/SUMA/Scheduler.pm
/usr/suma/lib/SUMA/DBMStanzaDir.pm
/usr/suma/lib/SUMA
/usr/suma/lib/SUMA/SoftwareInventory.pm
/usr/suma/bin/suma_mgdb

166


/usr/suma/bin
/usr/suma/lib/SUMA/NotifyCache.pm
/usr/suma/bin/suma_swinv
/usr/suma/bin/suma
/usr/suma
/usr/sbin/suma -> /usr/suma/bin/suma
/usr/suma/lib/SUMA/PolicyFactory.pm
/usr/suma/lib/SUMA/StanzaDB.pm
/usr/suma/lib/SUMA/GConfig.pm
/usr/suma/lib/SUMA/Util.pm
/usr/suma/lib/SUMA/Messenger.pm
Path: /etc/objrepos
bos.suma 5.3.0.0

/var/suma/tmp
/var/suma
/var/suma/data

5.4.2 Functional description
The Service Update Management Assistant is implemented as a task-oriented
utility which supports a comprehensive set of features:
Automated task-based retrieval of the following fix types and categories:
– Specific APAR
– Specific PTF
– Latest critical PTFs
– Latest security PTFs
– All latest PTFs
– Specific fileset
– Specific maintenance level
Task setup supported by SMIT (fast path suma) or command line interface
(/usr/sbin/suma).
Customizable task defaults to expedite setup of SUMA tasks.
Tunable number of concurrent downloads.
Unattended task activation at flexible intervals through scheduling module
(utilizing the cron command).
Three different task actions to initiated download preview, actual code
download, or combined download and fix repository cleanup (utilizing the
lppmgr command).


167

Download filtering against local software repository, maintenance level, lslpp
command output file and installed software on the local host, or a NIM client
Summary statistics show number of downloaded, failed, and skipped update
images
Support for FTP, HTTP, or HTTPS transfer protocols and proxy servers.
(HTTPS requires OpenSSL to be installed from AIX Toolbox for Linux
Applications.)
E-mail notification of update availability and task completion.
Messaging function providing six verbosity levels (Off, Error, Warning,
Information, Verbose, and Debug) that may be uniquely set for sending
information to the screen, log file, or e-mail address.

5.4.3 Concepts and implementation specifics
The suma command can be considered as the central component of the SUMA
implementation; it provides control over all SUMA-related tasks and operations.
Therefore we also refer to the suma command as the SUMA Controller. The suma
command Perl script is the main interface among the various SUMA modules
and allows for initiation of all SUMA functions. The SUMA Controller does the
following:
Receives task information either from SMIT or directly from the command line
and distributes this information to the task, notification, and scheduler
modules.
Handles calls from SMIT or the command line to manage tasks or
notifications.
Handles all cron daemon events when called by cron and initiates the
established task.
Reports error conditions and logs operations performed.
The SUMA controller utilizes certain SUMA modules to execute SUMA
operations and functions. Refer to Figure 5-2 for a graphical illustration of the
SUMA modules and their interaction and dependencies.

168


Figure 5-2 Service Update Management Assistant: Control flow diagram

The SUMA modules supply the following services and functions:

Download module
The download module provides functions related to network activities and is
solely responsible for communicating with the IBM Sserver pSeries support
server. This communication manifests itself in two different transaction types. In
the first, a list of filesets is requested from the fix server based on the SUMA task
data passed to the download module. The second consists solely of downloading
the requested files from the IBM Sserver support server.

Manage configuration module
The manage configuration module represents a utility class containing global
configuration data and general-purpose methods. These methods allow for the
validation of field names and field values since this information is predefined,
meaning that there is a known set of supported global configuration fields and
their corresponding supported values. This module provides the interface to the
global configuration database file.


169

Messenger module
The Messenger module provides messaging, logging, and notification capability.
Messages will be logged (or displayed) when their specified verbosity level is not
greater than the threshold defined by the SUMA global configuration.
The log files themselves will be no larger than a known size (by default, 1 MB),
as defined by the SUMA global configuration facility. When the maximum size is
reached, a backup of the file will be created, and a new log file started, initially
containing the last few lines of the previous file. Backup files are always created
in the same directory as the current log file. Therefore, minimum free space for
log files should be kept in mind.
There are two log files which are located in the /var/adm/ras/ directory. The log
file /var/adm/ras/suma.log contains any messages that pertain to SUMA
Controller operations. The other log file, /var/adm/ras/suma_dl.log tracks the
download history of SUMA download operations and contains only entries of the
form DateStamp:FileName. The download history file is appended when a new
file is downloaded. The two logs are treated the same with respect to maximum
size and creation/definition.
The messenger module relies on contact information (e-mail addresses) from the
notification database file which is managed by the notify module.

Notify module
The notify module manages the file which holds the contact information for
SUMA event notifications. This database stores a list of e-mail addresses for use
by SMIT when populating the list of notification addresses as part of SUMA task
configuration.

Task module
SUMA makes use of the task module to create, retrieve, view, modify, and delete
SUMA tasks. All SUMA task-related information is stored in a dedicated and
private task database file.

Scheduler module
The scheduler module is responsible for handling scheduling of SUMA task
execution and interacts with the AIX cron daemon and the files in the
/var/spool/cron/crontabs directory.

Inventory module
The inventory module returns the software inventory (installed or in a repository)
of the local system (localhost) or a NIM client. It covers all software that is in the
installp, RPM, or ISMP packaging format.

170


If the system specified to the module is not local then the system must be a NIM
client of the local system.

Utility and database modules
Other modules supply private utilities for SUMA code and utilities for handling the
stanza-style SUMA databases.

5.4.4 Command line interface
The suma command provides task-related SUMA control functions. The
command can be used to perform the following operations on a SUMA task:
Create
Edit
List
Schedule
Unschedule
Delete
The specified operation will be performed on the task represented by a unique
task identifier (TaskID). For the create or edit cases on a SUMA task, if the
TaskID is not specified, the create operation will be assumed, and a unique
TaskID will be generated. The suma -l command displays all SUMA tasks if the
TaskID is not specified. The suma -c command entered without any additional
flag will list the SUMA global configuration settings. The usage information of the
suma command is shown in the following:
# suma -?
Usage:
Create, Edit, or Schedule a SUMA task.
suma { { [-x][-w] } | -s CronSched } [ -a Field=Value ]... [ TaskID ]
List SUMA tasks.
suma -l [ TaskID ]...
List or Edit the default SUMA task.
suma -D [ -a Field=Value ]...
List or Edit the SUMA global configuration settings.
suma -c [ -a Field=Value ]...
Unschedule a SUMA task.
suma -u TaskID
Delete a SUMA task.
suma -d TaskID


171

5.4.5 SMIT user interface
All Service Update Management Assistant related tasks and functions are
supported by SMIT menus and panels. The new main SUMA menu shown in
Figure 5-3 can be directly accessed through the SMIT fast path suma.
Alternatively, you can select the new Service Update Management Assistant
(SUMA) option in either the Software Maintenance and Utilities or the Software
Service Management menu. Both menus are listed under the Software
Installation and Maintenance option of the SMIT top-level menu.

Figure 5-3 New SMIT menu: Service Update Management Assistant

A good starting point for a SUMA novice is potentially the Download Updates
Now (Easy) SMIT menu. Some common tasks are readily accessible through
this menu of quick-and-easy-options. The options are streamlined for usability,
but consequently provide limited access to the full set of SUMA's capabilities.
The following selections are offered:
Download by APAR Number
Download by Fix Type
Download Maintenance Level
Download All Latest Fixes
Download by Fileset Name
Using the download options of this menu and closely examining the SMIT
command output will give some insight into the way SUMA retrieves updates

172


from the IBM Sserver support Web site. Use the logging facility of SMIT to
study the output in detail.
An experienced SUMA user can directly proceed to use the advanced options
under the Custom/Automated Downloads (Advanced) menu. These advanced
options allow the system administrator to exert greater control over SUMA's
function, and expose many more capabilities not available through the Download
Updates Now menu. The user will be required to possess a deeper
understanding of SUMA and the options available. The options under that menu
allow management of highly-customizable tasks, which can be scheduled for
later or repeating execution, and tailored to fit the exact needs of the system
environment. The following options are accessible through the advanced menu:
Create a New Task
View/Change an Existing Task
Remove a Task
View All Tasks
Figure 5-4 on page 174 shows the new SMIT panel to create a new SUMA task.


173

Figure 5-4 SMIT panel: Create a New SUMA Task

The third option under the main SUMA SMIT menu allows the system
administrator to customize the SUMA configuration and task defaults. The
Configure SUMA SMIT menu provides control over configuration options whose
settings affect all of the task-oriented menus previously described. Three options
are available for selection:
Base Configuration
Settings such as verbosity level, lppmgr flags, Internet protocols, log file
maximum size, and download time out can be configured here.
Task Defaults
This panel allows the system administrator to set up a basic task profile
whose settings will affect both the easy and advanced actions in the
previously described menus. The options under the Download Update Now
(Easy) menu have these default settings fixed, under the covers, and not
configurable within the panels themselves. In the advanced task panels,
these settings appear as defaults when a new task is being created.

174


Saved E-mail Addresses
The underlying panels allow the system administrator to view and delete
entries of notification e-mail addresses.
Table 5-2 provides a complete overview of all available SUMA-related SMIT
menus and panels.
Table 5-2 SUMA SMIT menus and panels
SMIT menu or panel title

Fast path

Type

Service Update Management Assistant (SUMA)

suma

Menu

Download Updates Now (Easy)

suma_easy

Menu

Download by APAR Number

suma_easy_apar

Panel

Download by Fix Type

suma_easy_fixtype

Panel

Download Maintenance Level

suma_easy_ml

Panel

Download by Fileset Name

suma_easy_fileset

Panel

Custom/Automated Downloads (Advanced)

suma_task

Menu

Select an Action to Perform

suma_task_new

Panel

View/Change an Existing SUMA Task

suma_task_edit

Panel

Remove a SUMA Task

suma_task_delete

Panel

View/Change SUMA Task Defaults

suma_task_defaults

Panel

Configure SUMA

suma_config

Menu

Base Configuration

suma_config_base

Panel

View/Change SUMA Task Defaults

suma_task_defaults

Panel

Saved Email Addresses

suma_notify

Menu

Remove a Saved Email Address

suma_notify_remove

Panel

5.5 Long user and group name support
AIX 5L Version 5.3 provides support for long user and group names. Prior to
Version 5.3, there is a limit of eight characters for user and group names. Now
the user and group name length is increased from 8 to 255. Longer than eight
character user or group name capability can be enabled on the system by
changing the system-wide configuration parameter using the SMIT panels. Note
that the change needs to be done after careful analysis and consideration. Be


175

sure to determine the right size before setting the size. It is extremely difficult to
reduce the size of the user or group names once increased because there will be
existing users with longer names who might not be able to log in if the limits are
reduced. A size should be chosen that it meets the requirements of your
particular site (across multiple systems and platforms).

Verify and change user name limit
Before changing the user name limit, you may want to verify your user name
limit. Enter smit chgsys to view the new attribute related to user name limit as
highlighted in Figure 5-5.

Figure 5-5 Change/Show Characteristics of Operating System

Notice that the Maximum login name length at boot time attribute is set to 9, but
actually it is an 8-character user name limit since it has a terminating NULL
character. You will see an error message if you try to create a user with a name
longer than 8 characters in length before you change the default. Modify the
value in this field to your desired number. You must reboot the system for
changes to take effect.
Example 5-3 on page 177 shows the steps used to verify and change the user
name limit through the command line if you wish to change the user name limit to
20 characters. It should be noted that the current user name limit (output of the
getconf command) and the limit defined in ODM (output of the lsattr command)
could be different before you reboot the system to make the changes effective.

176


Example 5-3 Verify and change user name limit through the command line
# mkuser slson1234567890
3004-694 Error adding "slson1234567890" : Name is too long.
# lsattr -El sys0 -a max_logname
max_logname 9 Maximum login name length at boot time True
# chdev -l sys0 -a max_logname=21
sys0 changed
# getconf LOGIN_NAME_MAX
9
# shutdown -Fr

After you reboot the machine, you are ready to make a user name with more
than 8 characters. Create a user named slson1234567890 and set a password for
the user.
Example 5-4 Create a long user name
21
# mkuser slson1234567890
# passwd slson1234567890
Changing password for "slson1234567890"
slson1234567890's New password:
Enter the new password again:
#

Considerations
If you decrease the limit to the original value while you have long user names
defined, the command will not tell you which user names are too long. After you
reboot the system, if there are longer names used than the new lower limit
allows, those user names cannot be used for login. This is illustrated in
Example 5-5.
Example 5-5 Cannot log in with longer user name because of the new lower limit
# chdev -l sys0 -a max_logname=9
sys0 changed


177

21
#shutdown -Fr
# login slson1234567890
slson1234567890's Password:
[compat]: 3004-616 User "slson123" does not exist.
*******************************************************************************
*
*
* 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.
*
*
*
*******************************************************************************
3004-614 Unable to change directory to "".
You are in "/home/guest" instead.
$

5.6 Dynamic reconfiguration usability
Dynamic reconfiguration (DR) implemented with AIX 5L Version 5.2 allows
physical resources such as memory and processors to be added to or deleted
from a logical partition (LPAR) without rebooting the partition. To enhance the
DR usability the new cpupstat command was added in AIX 5L Version 5.3 to
provide information to the user on configurations preventing DR operations from
being successful.
A processor dynamic reconfiguration remove request can fail for a variety of
reasons. The most common of these is that the resource is busy due to an active
processor binding. The operating system cannot ignore processor bindings and
carry on DR operations or applications might not continue to operate properly. To
ensure that this does not occur, the system administrator needs to release the
binding, establish a new one, or terminate the application. The specific process
or threads that are impacted is a function of the type of binding that is used.
There are several different commands, subroutines, and methods which can be
used to establish a processor binding:
attachrset command
binprocessor command

178

Attaches a resource set (rset) to a process
Binds or unbinds the kernel threads of a process to a
processor


bindprocessor() subroutine Binds kernel threads to a processor when used by
an application
WLM software partitions

Workload Manager (WLM) enforced binding to
processor resource sets based on WLM class
attributes

wlm_set() subroutine

Sets or queries the Workload Manager (WLM) state
from within an application and can be used to
request that WLM takes the resource set bindings
into account

AIX 5L Version 5.3 offers the new cpupstat command to analyze the system for
processor bindings and detect configurations that could cause a CPU dynamic
reconfiguration operation to fail. The cpupstat command is part of the
bos.rte.commands fileset which is always installed on an AIX system. The
command can be executed by any user.
If you enter the command without any additional parameters the following usage
message is displayed:
# cpupstat
0560-001 Usage : cpupstat [-v] -i LogicalCPU.

The command requires the index of a logical CPU ID as input; the optional -v flag
provides verbose output when specified.
The command will run through three consecutive stages:
1. Check all the WLM class rsets for rsets with only one single active CPU
matching the passed in logical CPU ID. A count of the number of classes with
such an rset is printed. If the verbose option is given, the names of the
classes will be printed as well.
2. Check all the kernel registry rsets for rsets with only one single active CPU
matching the passed in logical CPU ID. A count of the number of processes
with attachments to such rsets will be printed to the user. If the verbose option
is given, the process IDs will be printed as well.
3. A count of the bindprocessor command attachments for the highest
numbered bind ID will be printed for the user. If the verbose option is given,
the process IDs will be printed as well.
The highest bind ID will always be removed if a processor DR removal operation
succeeds. Consequently the count of bindprocessor attachments for the highest
numbered bind ID will be determined regardless of the specified index of a
logical CPU.


179

The cpupstat command can be invoked directly from the command line but it will
also be executed after a failed CPU DR removal, with the argument being the
logical CPU that failed to be removed. Thus all the necessary information is
provided to permit removal of the processor bind related inhibitors to the DR
operation.

5.7 Paging space garbage collection
Ever increasing demands in resource capacity, scalability, and flexibility require
modifications and enhancements to core AIX functions, algorithms, and services.
This general observation also applies for AIX paging space allocation policy and
management.
AIX Version 4.3.2 introduced the deferred paging space allocation policy, which
ensures that no disk block is allocated for paging until it becomes necessary to
pageout a given memory page. This allocation policy was a modification and
enhancement to the existing late allocation policy which initiated a disk block
allocation once a memory page was requested and accessed, regardless of the
need to pageout that frame. The early paging space allocation policy offers an
alternative to the deferred paging space allocation policy. If the former policy is
chosen the paging space is allocated when memory is requested – even before
the frame is accessed or scheduled for pageout. The deferred allocation policy
reduces the amount of required paging space substantially.
Independent of the allocation policy in use, the paging space slots were only
released by process termination or by the disclaim() system call. The free()
system call only releases allocated paging space if the MALLOCDISCLAIM
environment variable is set to true.
The later behavior may no longer be desirable for modern applications and
systems capable of dynamic resource optimization. The following two scenarios
illustrate this statement:
Consider the class of applications used in a large real-memory environment
and which are long running in nature. An application of that class potentially
runs in memory; however, burst activity occasionally will generate some
paging allocation. These pages eventually get re-paged in, and with the
Virtual Memory Manager’s (VMM) allocation policy of the previous AIX
releases, retain their disk blocks. Thus they may stay in memory for the life
span of the related processes, but use up paging space needlessly. This
class of applications may literally be running for months, and cannot tolerate
down time. But if, over time, an application runs out of paging space, the
operating system may be forced to kill the application to maintain system
integrity.

180


Consider a similar environment to the one just described. A long-lived
application runs in a logical partition; however, periodic dynamic
reconfiguration memory removes perform load balancing and send memory
to another partition for a relatively short amount of time. That memory may be
returned when the partition that was the target of the DR memory operation is
done with its work. Frequent small DR operations may take place as partitions
are spawned to perform one single task, then exit. The result of this memory
remove can and frequently will be paging space allocation. If memory is
subsequently re-added, these disk blocks are not freed, and this can cause
pressure on paging space, in spite of having a reasonable amount of main
memory.
AIX 5L Version 5.3 addresses these issues and introduces enhanced paging
space management algorithms to reclaim, or garbage collect (GC) paging space
as needed.
Version 5.3 implements two paging space garbage collection (PSGC) methods
for deferred allocation working segments:
Garbage collect paging space on re-pagein
Garbage collect paging space scrubbing for in-memory frames
The following modes and environments are supported:
LPAR and SMP mode
32-bit and 64-bit multi-processor kernel
All currently supported hardware platforms
Deferred allocation working storage segments only

5.7.1 Garbage collection on re-pagein
AIX 5L Version 5.3 introduces a dynamic decision-making mechanism, whereby
pagein will make a determination automatically, whether or not to free the disk
block (at the end of pagein) based upon whether paging space is low or not.
This garbage collection method is also referred to as the re-pagein disk block
free mechanism. This mechanism only applies to the deferred page space
allocation policy. Normally, the deferred page space allocation policy will, upon
paging a page back in from disk, retain its disk block, and leave the page
unmodified in memory. If the page does not get modified then, when it is selected
for replacement by the VMM it does not have to be written to disk again. This is a
performance advantage for pages that are not modified after being paged back in
from disk. The re-pagein disk block free mechanism allows the system
administrator to specify that the operating system should free a page's disk block
after pagein. The page will be left modified in memory. If the page gets selected


181

for replacement by the VMM, it will have a new disk block allocated for it, and it
will be written to disk. There are two advantages to this:
Paging space disk blocks are conserved. Pages in memory do not have
unnecessary disk space committed to them.
If such pages are selected for replacement by the VMM, new disk blocks will
be allocated for them. In the event that the Virtual Memory Manager is paging
out numerous pages at once, it is quite possible that these disk blocks may be
allocated adjacent or nearby on the paging device. This can improve the
performance of writing out the pages. If they had retained their original disk
locations, and had been modified after pagein, then if they get paged out
along with other pages, the disk blocks are likely to be scattered across the
paging devices.
This re-pagein mechanism is enabled by default on the operating system when
the system becomes low on free paging space disk blocks. At this point, by
default, pages that are paged in due to a modification will have their disk blocks
freed. A modification could be in the form of a bzero(), for example. Since the
page is being modified, there is no significant performance cost to freeing the
disk block because, if this page is selected for replacement by the Virtual
Memory Manager later on, it would have to be written to disk anyway.
The enhanced vmo command provides all tuning and control functions for the
re-pagein garbage collection. It gives the system administrator the ability to alter
the thresholds of free disk blocks at which the re-pagein disk block free
mechanism starts and stops. The mechanism may also be disabled or always
enabled. Additionally, the ability to free disk blocks on pageins due to read
accesses to a page may be specified.
The vmo command allows configuration of the following re-pagein GC parameters:
npsrpgmin

Low paging space threshold when re-pagein GC starts.

npsrpgmax

High paging space threshold when re-pagein GC stops.

rpgclean

Controls if re-pagein GC will be active only for store operations
(page modify) or for store as well as load (page read/access)
operations.

rpgcontrol

Controls if re-pagein GC will be disabled, always enabled, or only
enabled if paging space utilization is within the limits of
npsrpgmin and npsrpgmax. The scope and behavior of
rpgcontrol is influenced by rpgclean.

By default, AIX 5L Version 5.3 will activate the re-pagein garbage collection only
if paging space utilization is within the limits of npsrpgmin and npsrpgmax and it
will only free paging space disk blocks on pagein of pages that are being
modified (load operation).

182


The thresholds for GC activation are just above the traditional paging space
warning thresholds. Note that there is a static initial default value (Init) for the
thresholds, but, they are adjusted as paging spaces are added, or grown (Grow),
and shrunk (Shrink) as they are removed.
The following list provides an overview of how the new paging space tuning
parameters are determined and how they depend on the traditional paging space
thresholds for npskill and npswarn.
npskill

Init(128 psdb)
Grow(MAX(npskill, numpsblks/128))
Shrink(MAX(128 psdb, numpsblks/128))

npswarn

Init(512 psdb)
Grow(MAX(npswarn, npskill*4))
Shrink(MAX(512 psdb ,npskill*4))

npsrpgmin

Init(768 psdb)
Grow(MAX(npsrpgmin, npswarn+(npswarn/2)))
Shrink(MAX(768 psdb, npswarn+(npswarn/2)))

npsrpgmax

Init(1024 psdb)
Grow(MAX(npsrpgmax, npswarn*2))
Shrink(MAX(1024 psdb, npswarn*2))

The preceding examples used the following definitions:
psdb

Paging space disk blocks (4096 bytes)

numpsblks

Number of paging space blocks

In the example, 768 blocks are 3 MB, while 1024 are 4 MB. The default npswarn
value is 2 MB (512 blocks), but it is adjusted when a paging space grows or is
added, to where it attempts to be (npskill*4). Note, as shown, that npsrpgmin and
npsrpgmax will be adjusted in lockstep automatically with npswarn and npskill
upon adding and removing paging space.
Refer to “VMM tuning parameters for PSGC” on page 185 for additional details
about the vmo command GC controls and thresholds.
Early paging space allocation segments (PSALLOC=early) reserve all of their
paging space disk blocks when they are created, or marked for early allocation. If
the system frees an allocated disk block in an early allocation segment, the
reserved amount of paging space consumption will not be reduced and the
global system paging space count does not get updated. Because of this
behavior, early allocation segments are ignored by the re-pagein GC algorithm.


183

5.7.2 GC paging space scrubbing for in-memory frames
The re-paging disk block free mechanism is supplemented by a more powerful
but less granular method. Instead of, or in addition to the re-pagein disk block
free mechanism, already discussed in “Garbage collection on re-pagein” on
page 181, a system administrator might want to try and reclaim paging space
disk blocks for pages that are already in memory.
When using the deferred page space allocation policy, pages that are paged in
from paging space retain their paging space disk block while they are in memory.
If the page does not get modified, then when it is selected for replacement by the
Virtual Memory Manager, it does not have to be written to disk again. This is a
performance advantage for pages that are not modified after being paged back in
from disk. Over time, however, after many pageins, it is possible that there could
be many pages in memory that are consuming paging space disk blocks. This
could cause paging space to become low. To address this problem, the system
administrator, through the vmo command, can enable a mechanism to free paging
space disk blocks from pages in memory that have them. This is referred to as
garbage collection, or scrubbing of these disk blocks.
The vmo parameters scrub, scrubclean, npsscrubmin and npsscrubmax control
the garbage collection paging space scrubbing activity as follows:
npsscrubmin

Low paging space threshold for GC paging scrubbing to start.
The npsscrubmin will default to npsrpgmin.

npsscrubmax

High paging space threshold for GC paging space scrubbing to
stop. The npsscrubmax threshold will default to npsrpgmax

scrubclean

Controls if GC paging space scrubbing will free disk blocks for
unmodified pages as well as modified. By default, only paging
space disk blocks for modified pages in memory are freed.

scrub

Controls if GC paging space scrubbing will be enabled or
disabled. By default paging space scrubbing is disabled. If it is
enabled, scrubbing of in-memory paging space disk blocks will
be activated when the number of system free paging space
blocks is below npsscrubmin, and continues until above
npsscrubmax.

Note that there is a static initial default value (Init) for the thresholds, but, they are
adjusted as paging spaces are added, or grown (Grow), and shrunk (Shrink) as
they are removed.

184


The following list provides an overview of how the new paging space scrubbing
tuning parameter defaults are computed:
npsscrubmin

Init(768 psdb)
Grow(MAX(npsscrubmin, npsrpgmin))
Shrink(MAX(1024 psdb, npsrpgmin))

npsscrubmax

Init(1024 psdb)
Grow(MAX(npsscrubmax,npsrpgmax))
Shrink(MAX(1024 psdb, npsrpgmax))

The preceding examples used the following definition:
db

Paging space disk blocks (4096 bytes))

If the GC paging space scrubbing is enabled, an internal timer will trigger a
kernel service to start the garbage collection process every 60 seconds when the
number of system free paging space blocks are within the limits of the lower and
upper scrubbing thresholds.

5.7.3 VMM tuning parameters for PSGC
In support for the new re-pagein disk free mechanism and the garbage collection
paging space scrubbing, eight new controls and tuning parameters can be
configured by the vmo command. Descriptions, the default values, and the value
ranges for the new parameters are listed in this section. The traditional npskill
and the npswarn parameters are also listed for completeness and for your
reference. Consult the AIX product documentation for a detailed and full
description of the vmo command.
npskill

Specifies the number of free paging-space pages at which the
operating system begins killing processes.
Default: MAX(64, number of paging space pages/128).
Range: 0 to total number of paging space pages on the system.

npswarn

Specifies the number of free paging-space pages at which the
operating system begins sending the SIGDANGER signal to
processes.
Default: MAX(512, 4*npskill)
Range: 0 to total number of paging space pages on the system.

npsrpgmax

Specifies the number of free paging space blocks at which the
operating system stops freeing disk blocks on pagein of deferred
page space allocation policy pages.
Default: MAX(1024, npswarn*2).
Range: 0 to total number of paging space blocks in the system.

npsrpgmin

operating system starts freeing disk blocks on pagein of deferred


185

page space allocation policy pages.
Default: MAX(768, npswarn+(npswarn/2)).
npsscrubmax

operating system stops scrubbing in-memory pages to free disk
blocks from deferred page space allocation policy pages.
Default: MAX(1024, npsrpgmax).

npsscrubmin

operating system starts scrubbing in-memory pages to free disk
blocks from deferred page space allocation policy pages.
Default: MAX(768, npsrpgmin).

rpgclean

Enables or Disables freeing paging space disk blocks of deferred
page space allocation policy pages on read accesses to them.
Default: 0, free paging space disk blocks only on pagein of pages
that are being modified (write).
Range: 0-1
1, Free paging space disk blocks on pagein of a page being
modified (write) or accessed (read).

rpgcontrol

Enables or disables freeing of paging space disk blocks at
pagein of deferred page space allocation policy pages.
Default: 2, always enables freeing of paging space disk blocks
on pagein, regardless of thresholds. The value of rpgclean will
determine the scope of this setting. That means, by default, just
write accesses are always processed and read accesses will
only be additionally processed if rpgclean is 1.
Range: 0-3
0 - Disables freeing of paging space disk blocks on pagein.
1 - Enables freeing of paging space disk blocks when the number
of system free paging space blocks is below npsrpgmin, and
continues until above npsrpgmax. The value of rpgclean will
determine the scope of this setting. That means, by default,
just write accesses are always processed and read accesses
will only be additionally processed if rpgclean is 1.
3 - Always enables freeing of paging space disk blocks on pagein
of pages that are being modified (write). If rpgclean is set to 1,
re-pagein GC will additionally free paging space disk blocks of
pages that were only accessed (read), but the GC activity will
only take place when the number of system free paging space
blocks is within the threshold limits. Therefore option 3 can be
considered to be a combination of options 1 and 2.

186


scrub

Enables or disables freeing of paging space disk blocks from
pages in memory for deferred page space allocation policy
pages.
Default: 0, Disables scrubbing completely.
Range: 0-1, 1 enables scrubbing of in memory paging space disk
blocks when the number of system free paging space blocks is
below npsscrubmin, and continues until above npsscrubmax.

scrubclean

Enables or Disables freeing paging space disk blocks of deferred
page space allocation policy pages in memory that are not
modified.
Default: 0, Free paging space disk blocks only for modified
pages in memory.
Range: 0-1, 1, Free paging space disk blocks for modified or
unmodified pages.

5.8 Dynamic support for large page pools
The AIX support for large page architecture has been available since AIX 5L
Version 5.1. A white paper about this topic is on the Web at:
http://www-1.ibm.com/servers/aix/whitepapers/large_page.html

AIX 5L Version 5.3 allows dynamic, runtime resizing of the large pools without
the need for a system reboot. In previous versions large pools could be changed
only through system reboot.
The large pages are changed by the vmo command changing the lgpg_size and
the lgpg_regions attribute. Example 5-6 on page 188 shows the procedure to
change the large page pools dynamically. In the example we first check the
lgpg_regions and lgpg_size attributes of the VMM and then set them to two
regions of 16 MB size, which totals to 32 MB large pool size. You can check the
allocated large pages (alp) and the free large pages (flp) with the vmstat -l
command.


187

Example 5-6 lgpg command example

# vmo -o lgpg_regions -o lgpg_size
lgpg_size = 0
lgpg_regions = 0
# vmo -o lgpg_regions=2 -o lgpg_size=16777216
Setting lgpg_size to 16777216
Setting lgpg_regions to 2
# vmstat -l
System configuration: lcpu=4 mem=512MB ent=0
kthr
memory
page
faults
cpu
large-page
----- ----------- ------------------------ ------------ ----------------------- ----------r b
avm
fre re pi po fr
sr cy in
sy cs us sy id wa
pc
ec
alp
flp
1 1 65587 45851
0
0
0
0
0
0 14 273 75 0 0 99 0 0.01
1.1
0
2

Setting the lgpg_regions and lgpg_size back to zero disables use of the large
page pools.
There are two different conversion operations when changing the large page
segment size:
The first conversion operation creates large pages from 4 KB pages when
increasing the large page pool. If there are no free pages, the VMM migrates
the 4 KB memory area to the 16 MB large page area.
The second conversion operation creates 4 KB pages from large pages when
decreasing the large page pool. When converting from large pages to 4 KB
pages, only large pages that are not in-use (free large pages) can be
converted into 4 KB pages.
In case of an error, the conversion stops. The already converted pages will
remain converted.
The dynamic large page pool function is only available on machines that support
DLPAR. There are some environments where these machines still cannot use
DLPAR because of a DLPAR unsafe kernel extension.

5.9 Interim Fix Management
Establishing a unified framework for systems management with a common
terminology across various IBM Sserver products is one of the long-term
strategies which drives IBM’s ongoing convergence effort. As a consequence of
this strategy, AIX Emergency Fix Management has been renamed to Interim Fix
Management in AIX 5L Version 5.3.

188


Interim fixes serve only as a temporary solution to an identified software problem
to bridge the time until the program temporary fix (PTF) will have been developed
and successfully passed regression test. Thus, the term interim fix and the term
emergency fix are equivalent and can be used interchangeably.
To accommodate the name change, AIX 5L Version 5.3 introduces two new hard
links to the emgr and the epkg command which will be established during the
installation of the bos.rte.install fileset. The links introduce the fixmgr command
as a synonym for emgr and the fixpkg command as a synonym for epkg. You can
use the lslpp command to identify the new members of the command family:
# lslpp -f bos.rte.install | grep /usr/sbin/fix
/usr/sbin/fixmgr -> /usr/sbin/emgr
/usr/sbin/fixpkg -> /usr/sbin/epkg

No functional changes were implemented in conjunction with the fix management
renaming.

5.10 List installed filesets by bundle
AIX supports the concept of software bundles. A software bundle comprises a
group of software packages that support a particular environment or major
functional complex and therefore are supposed to be installed together as one
unit. The software packages are listed in the associated bundle definition file
whose file name has to end with the .bnd extension. AIX provides a set of
predefined bundles in the /usr/sys/inst.data/sys_bundles directory. User defined
bundles should be placed in the /usr/sys/inst.data/user_bundles directory
because SMIT and the Web-based System Manager will look by default in this
directory to populate relevant selection lists.
In AIX releases prior to AIX 5L Version 5.3 no one-step facility existed to enable
a system administrator to list installable software packages based on the
contents of a bundle file. Version 5.3 adds a new feature to the lslpp command
that allows the administrator to view the state of the installable software
packages listed in a bundle file.

lslpp command line interface changes
The lslpp command displays information about installed filesets or fileset
updates. The enhanced command implementation provides the new -b flag
which has to be immediately followed by the value of the File parameter. The
following paragraphs show the syntax statement of the lslpp command manual
page and describe the new flag and parameter in more detail.


189

...
lslpp Command
Purpose
Lists installed software products.
Syntax
lslpp { -d | -E | -f | -h | -i | -l | -L | -p } ] [ -a] [ -c] [ -J ] [-q ]
[ -I ] [ -O { [ r ] [ s ] [ u ] } ] [ [ FilesetName ... | FixID ... | -b
File | all ]
lslpp -w [ -c ] [ -q ] [ -O { [ r ] [ s ] [ u ] } ] [ FileName ... | all ]
lslpp -L -c [ -v]
lslpp -S [A|O]
lslpp -e
...

The FilesetName and the FixID parameter have been present in AIX releases
prior to Version 5.3. The FilesetName parameter is the name of a software
product or package. The FixID (also known as PTF or program temporary fix ID)
parameter specifies the identifier of an update to a formatted fileset. The File
parameter is new to AIX and specifies a bundle file to use as a fileset list for
input.
-b File

Specifies a bundle file to search for fileset names. The filesets listed
in the bundle are then listed as if they had been specified explicitly as
FilesetName parameters. To mimic installp behavior, the installp
image names are automatically wildcarded. For example, a bundle
file entry of I:bos.abc will behave as if bos.abc* was specified as a
FilesetName parameter.
If the file does not reside in one of the known bundle locations, the
full path and file name, including extension, must be specified.
/usr/sys/inst.data/sys_bundles/ and /usr/sys/inst.data/user_bundles/
are the known locations for bundle files.

The lslpp -l -b App-Dev command provided in the following example shows
the state of all filesets which belong to the predefined application development
system bundle /usr/sys/inst.data/sys_bundles/App-Dev.bnd. Three filesets are
listed as not installed at the end of the output. This fact may indicate that the
App-Dev bundle was never used in an install operation.

190


# lslpp -l -b App-Dev
Fileset
Level State
Description
bos.adt.base
5.3.0.0 COMMITTED Base Application Development
Toolkit
bos.adt.include
Include Files
bos.adt.lib
Libraries
bos.net.tcp.adt
5.3.0.0 COMMITTED TCP/IP Application Toolkit
bos.perf.diag_tool
5.3.0.0 COMMITTED Performance Diagnostic Tool
bos.perf.libperfstat
5.3.0.0 COMMITTED Performance Statistics Library
Interface
bos.perf.perfstat
5.3.0.0 COMMITTED Performance Statistics
Interface
bos.perf.proctools
5.3.0.0 COMMITTED Proc Filesystem Tools
bos.perf.tools
5.3.0.0 COMMITTED Base Performance Tools
bos.perf.tune
5.3.0.0 COMMITTED Performance Tuning Support
xlC.cpp
6.0.0.0 COMMITTED C for AIX Preprocessor
Path: /etc/objrepos
bos.perf.diag_tool
bos.perf.libperfstat

5.3.0.0
5.3.0.0

COMMITTED
COMMITTED

Performance Diagnostic Tool
Performance Statistics Library
Interface
bos.perf.perfstat
5.3.0.0 COMMITTED Performance Statistics
Interface
bos.perf.tools
5.3.0.0 COMMITTED Base Performance Tools
bos.perf.tune
5.3.0.0 COMMITTED Performance Tuning Support
lslpp: 0504-132 Fileset bos.loc.adt* not installed.
lslpp: 0504-132 Fileset bos.net.nfs.adt* not installed.
lslpp: 0504-132 Fileset Java14.debug* not installed.

SMIT interface support
AIX 5L Version 5.3 provides the SMIT dialog List Installed Software by Bundle in
support of the new lslpp feature. You can access this menu directly by the use of
the SMIT fast path list_installed_sw_bnd or along the following path through the
SMIT menu hierarchy: smit → Software Installation and Maintenance → List
Software and Related Information → List Installed Software and Related
Information → List Installed Software by Bundle.


191

5.11 Configuration file modification surveillance
The AIX 5L Version 5.3 enhanced geninstall command provides an easy way
for system administrators to see what modifications have been made to any of
the configuration files listed in the /etc/check_config.files file. When an
installation or update operation changes these files, the differences between the
old and new files are recorded in the /var/adm/ras/config.diff output file. If the
/etc/check_config.files file requests that the old file be saved, the old file can be
found in the /var/adm/config directory.
Version 5.3 ships a default /etc/check_config.files file which can be modified by
the system administrator at any time. The cat /etc/check_config.files
command output that follows shows an example of its contents:
#
#
#
#
#
d
d
d
d
d
d
d
s
d
d
d
d
d
d
d
d

cat /etc/check_config.files
Watched File List
d = delete old file; s = save old file
Old files are saved to /var/adm/config
This file may be edited
/etc/check_config.files
/etc/csh.cshrc
/etc/csh.login
/etc/diskpartitions
/etc/environment
/etc/filesystems
/etc/group
/etc/inittab
/etc/motd
/etc/passwd
/etc/profile
/etc/rc
/etc/services
/etc/shells
/etc/swapspaces
/etc/vfs

The first few lines of this example check_config.files file are informational
comments that are followed by a two column list. All comment lines have to begin
with the # character and blank lines are ignored. The first column of any other
line has to start with a letter d or s in either upper or lower case. The second
column holds the full path to the configuration file being monitored during the
install or update process.
If an s (either case) is specified, the old (pre-installation) version of the file will be
saved at /var/adm/config, retaining its prior path information. (For example:
/etc/inittab would be saved to /var/adm/config/etc/inittab). If a saved version of a

192


configuration file already exists in the /var/adm/config directory, the existing
version will be renamed by appending a process ID which is related to the
current geninstall command process.
If a d (either case) is specified, the pre-installation version of the file will not be
saved in /var/adm/config.
If neither d nor s is specified, that is, only the configuration file name is given, the
file will be treated as if d was specified in the first column.
Upon completion of the geninstall command execution, the new configuration
files will be compared against those saved in /var/adm/config/. If any of the files
have changed during the course of installation, a message will be printed in the
installation summary section stating which file (or files) changed, along with a
concluding message, stating that one or more files have changed. The message
also gives the location of the difference log /var/adm/ras/config.diff. The
messages are either displayed at the system console or are captured by the
smit.log if the installation process was started by SMIT.
An excerpt of a sample smit.log file that was written during the installation of
LDAP server support on an AIX 5L Version 5.3 system follows.
...
Finished processing all filesets. (Total time:

5 mins 23 secs).

+-----------------------------------------------------------------------------+
Summaries:
+-----------------------------------------------------------------------------+
Installation Summary
-------------------Name
Level
Part
Event
Result
------------------------------------------------------------------------------db2_08_01.pext
8.1.1.16
USR
APPLY
SUCCESS
db2_08_01.msg.en_US.iso8859 8.1.1.16
USR
APPLY
SUCCESS
db2_08_01.jhlp.en_US.iso885 8.1.1.16
USR
APPLY
SUCCESS
db2_08_01.db2.engn
8.1.1.16
USR
APPLY
SUCCESS
db2_08_01.das
8.1.1.16
USR
APPLY
SUCCESS
db2_08_01.cs.rte
8.1.1.16
USR
APPLY
SUCCESS
db2_08_01.conv
8.1.1.16
USR
APPLY
SUCCESS
db2_08_01.conn
8.1.1.16
USR
APPLY
SUCCESS
db2_08_01.cnvucs
8.1.1.16
USR
APPLY
SUCCESS
ldap.msg.en_US
5.2.0.0
USR
APPLY
SUCCESS
ldap.html.en_US.man
5.2.0.0
USR
APPLY
SUCCESS
ldap.html.en_US.config
5.2.0.0
USR
APPLY
SUCCESS
ldap.server.java
5.2.0.0
USR
APPLY
SUCCESS
ldap.server.rte
5.2.0.0
USR
APPLY
SUCCESS


193

ldap.server.com
ldap.server.cfg
ldap.server.com
ldap.server.cfg
db2_08_01.essg

5.2.0.0
5.2.0.0
5.2.0.0
5.2.0.0
8.1.1.16

USR
USR
ROOT
ROOT
USR

APPLY
APPLY
APPLY
APPLY
APPLY

SUCCESS
SUCCESS
SUCCESS
SUCCESS
SUCCESS

File /etc/group has been modified.
File /etc/inittab has been modified.
File /etc/passwd has been modified.
One or more of the files listed in /etc/check_config.files have changed.
See /var/adm/ras/config.diff for details.
...

The content of the related /var/adm/ras/config.diff log file is shown in the
following example. Note that the /etc/inittab file was only saved to
/var/adm/config because the related entry in the /etc/check_config.files begins
with the letter s in the first column. The modifications to the configuration files are
analyzed and reported by the diff command.
# cat /var/adm/ras/config.diff
=======================================================================
/etc/group
The original file was not saved.
2c2
< staff:!:1:ipsec,aroell
--> staff:!:1:ipsec,aroell,ldap
20a21
> ldap:!:201:ldap
=======================================================================
/etc/inittab
The original file was saved to /var/adm/config/etc/inittab.
70a71,72
> fmc:2:respawn:/usr/opt/db2_08_01/bin/db2fmcd #DB2 Fault Monitor Coordinator
> ibmdir:2:once:/usr/ldap/sbin/rc.ibmdir > /dev/null 2>&1
=======================================================================
/etc/passwd
The original file was not saved.
15a16
> ldap:*:202:1::/home/ldap:/usr/bin/ksh

Should a /var/adm/ras/config.diff file already exist when beginning an install
operation, the old config.diff file will be renamed, with the current geninstall
command process ID suffixed to the name. This will keep the most current
version with a known file name that never changes. Any old files are named for
the process that moved them. This is particularly important to longer installs,

194


such as during migration, when the installp command might be called multiple
times.

5.12 DVD backup using the mkdvd command
System administrators can use the mkcd command of previous AIX releases to
create a multi-volume CD or DVD from a mksysb or savevg backup image. In
order to create a DVD ISO9660 backup it is required to specify the -L flag of the
mkcd command to ensure that the image will be in the correct format. To prevent
the command from failing because a single flag was omitted on the command
line, AIX 5L Version 5.3 provides the new mkdvd command. The mkdvd command
is actually a hard link to the /usr/sbin/mkcd script, but the script has been
enhanced to verify the command name which was used to call it. If the command
name was mkdvd then the internal DVD_FLAG will be set and the image format is
automatically adjusted to be suitable for DVDs.

5.13 NIM security
In previous versions of AIX, NIM used rsh and rcmd commands to perform
remote execution of commands on clients. These r-commands were a potential
security exposure.
AIX 5L Version 5.3 is enhanced by the nimsh environment that is part of the
bos.sysmgt.nim.client fileset. It allows the following two remote execution
environments:
NIM service handler for client communication - basic nimsh
NIM cryptographic authentication - OpenSSL
While the basic nimsh is an easy to use solution with sufficient security, the
OpenSSL provides additional up-to-date cryptographic security. In the following
sections we explain both approaches to NIM security.
The original rsh or rcmd command environments are still supported in Version
5.3 because of compatibility and ease of use reasons.
The nimsh daemon is controlled by the AIX system resource controller. You can
check the subsystem with the lssrc -s nimsh command. The recommended
way to start the nimsh subsystem is by the nimclient -C command.


195

5.13.1 NIM service handler for client communication - NIMSH
The nimsh subsystem is a restricted shell environment that allows only NIM
method execution. This increases the security of remote access compared to the
rsh or rcmd command environments.
When the NIM server intends to run a method on the NIM client, it connects to
the nimsh subsystem and sends authentication information and the requested
method in a similar way as the rcmd() service. When the nimsh authenticates the
incoming request, it checks that the request is valid and performs the requested
method. If the authentication fails or the method is not registered, the nimsh
denies the execution of the request. The use of the service ports and the
authentication is explained in the following sections.

Service ports
The client daemon has two ports registered with the Internet Assigned Numbers
Authority (IANA). Part of the /etc/services shows the port numbers here:
nimsh
nimsh
nimaux
nimaux

3901/tcp
3901/udp
3902/tcp
3902/udp

#
#
#
#

NIM Service Handler
NIM Service Handler
NIMsh Auxiliary Port
NIMsh Auxiliary Port

The primary nimsh port is used to listen for the incoming service requests. The
NIM master selects the port 1023 or the next available lower port to connect to
the nimsh (3901) port in the NIM client. Once the service request is accepted, the
primary port is used for the stdin and stdout (file descriptor 0 and 1). During the
authentication process the NIM master selects another available port that is 1022
and delivers the port number information to the NIM client. The client opens an
auxiliary connection from port nimaux (3902) to the port of the NIM master as
received from the master. The secondary nimaux port is opened for the stderr
(file descriptor 2).

Enabling secondary port
The following procedure describes how to configure existing standalone clients
to use the nimsh communication protocol with a secondary port option enabled.
By default, nimsh uses a reserved port for returning stderr output during
command execution. The default setting allows administrators to specify a
specific port for opening behind a firewall, but it can cause performance issues
when several connections are attempted in a short amount of time.
When TCP connections are closed, the closing sockets enter TIME_WAIT state.
The length of time for this state may last up to 240 seconds depending on system
settings. The secondary port option allows you to specify any specific range of
ports to cycle through during NIMSH operation.

196


For firewalls, administrators might want to open a specific range on the firewall,
and then for each machine on the internal network, ensure that the port range on
the machine coincides with the open range on the firewall. When changing the
NIMSH secondary port, you should choose a range of ports outside of the range
used for system services. Try using ports 49152 through 65535.
To configure existing standalone clients to use the NIMSH communication
protocol with a secondary port range, complete the following steps:
1. Use the smitty nim_config_services fast path on the NIM client.
2. Select nimsh as the Communication Protocol used by client.
3. Specify a start value for the secondary port number.
4. Specify an increment value for the secondary port range.
To configure existing standalone clients to use the NIMSH communication
protocol with a secondary port range from the command line, complete the
following steps:
1. Edit the /etc/environment file. Add the variable NIM_SECONDARY_PORT=60000:5
to use ports 60000 - 60005 within NIMSH.
2. Use the desired nimclient command option to restart the nimsh daemon.

Authentication process
Service requests of the nimsh are handled in a similar fashion to rcmd(). The
communicating host (NIM server) builds packets with the following data for
authentication:
Host name of NIM client
CPUID of NIM client
CPUID of NIM master
Return port for secondary (stderr) connection
Query flag
If the Query flag is set to 1 the nimshd daemon treats the incoming request as a
client discovery for information. The following data is returned:
Default host name obtained from inet0
Default route obtained from inet0
Network address obtained from host name
Network interface obtained from host name


197

If the query flag is not set, then a request for service (NIM operation) is pushed
by the NIM master. The nimshd daemon validates the method request as follows:
1. Verify the host name of the NIM master is the recognized master host name
to the client.
2. Check the client CPUID passed in the authentication data. It should match the
client’s machine ID.
3. Check the master CPUID passed in the authentication data. It should match
the master’s machine ID stored in the memory. It is read into the memory
from the /etc/niminfo file and the mktcpip -S primary_nim_interface
command outputs.
4. Verify that the operation passed in the method is a method in the path
/usr/lpp/bos.sysmgt/nim/methods.
5. Check for cryptographic authentication settings.
For additional security, nimsh supports push disablement. When push
disablement is set, nimsh does not process any NIM operation controlled by the
NIM master. Use the nimclient command to enable or disable the push from
master.
You can check the authentication procedure using the nimquery command on
the master, while at the same time watching the nimsh.log on the client. First
start the nimsh subsystem on the client and check that the /etc/niminfo and
the mktcpip command values are read and logged to the log file:
# nimclient -C
0513-059 The nimsh Subsystem has been started. Subsystem PID is 422002.
# cat /var/adm/ras/nimsh.log
Tue Jul 13 17:23:26 2004
/usr/sbin/nimsh: NIM Service Handler
started from SRC
Tue Jul 13 17:23:26 2004
no environment value for NIM_SECONDARY_PORT
Tue Jul 13 17:23:26 2004
value for route is net,-hopcount,0,,0 and
gateway is 9.3.5.41
Tue Jul 13 17:23:26 2004
value for hostname is server3
Tue Jul 13 17:23:26 2004
value for netaddr is 9.3.5.196
Tue Jul 13 17:23:26 2004
value for netif is en0
Tue Jul 13 17:23:26 2004
value for netmask is 255.255.255.0
Tue Jul 13 17:23:26 2004
obtained master's hostname:
NIM_MASTER_HOSTNAME=server4
Tue Jul 13 17:23:26 2004
NIM_MASTERID=000C91AD4C00

obtained master's id:

Tue Jul 13 17:23:26 2004
obtained master's hostname:
NIM_MASTER_HOSTNAME_LIST="server4 server2"

198


Tue Jul 13 17:23:26 2004

value for machine id is 0000316A4C00

# mktcpip -S en0
#host:addr:mask:_rawname:nameserv:domain:gateway:cost:activedgd:type:start
server3:9.3.5.196:255.255.255.0:en0:::9.3.5.41:0:no:N/A:no

Next, start a query from the server to the client:
s4 # nimquery -ahost=server3
host:server3:addr:9.3.5.196:mask:255.255.255.0:gtwy:9.3.5.41:_pif:en0:_ssl:
no:_psh:no:_res:no:asyn:no:
s4 # nimquery -ahost=server3 -p
===================
NIMSH Client Values
===================
hostname
= server3
IP address = 9.3.5.196
subnetmask = 255.255.255.0
gateway
= 9.3.5.41
interface
= en0
====================
NIMSH Client Options
====================
SSL Enabled = no
Push Denied = no
Restricted Shell = no

Finally, check the nimsh.log on the client:
# cat /var/adm/ras/nimsh.log
...
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
000000000000
Tue Jul 13 17:27:56 2004
000C91AD4C00
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004
Tue Jul 13 17:27:56 2004

file descriptor is 13
file descriptor is : 13
family is : 2
source port is : 1023
source addr is : 9.3.5.197
source hostname is : server4
getting 2nd port
count equals 0
got stderr port 0
success: we got 1st write query is 1
success: we got 2nd write local id is
success: we got 3rd write remote id is
success: we got 4th write command is
sending ack to client
buffer value is host:server3:


199

Tue Jul 13 17:27:56
Tue Jul 13 17:27:56
Tue Jul 13 17:27:56
Tue Jul 13 17:27:56
Tue Jul 13 17:27:56
Tue Jul 13 17:27:56
Tue Jul 13 17:27:56
Tue Jul 13 17:27:56
Tue Jul 13 17:27:56
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
000000000000
Tue Jul 13 17:28:18
000C91AD4C00
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18
Tue Jul 13 17:28:18

2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004

buffer value is addr:9.3.5.196:
buffer value is mask:255.255.255.0:
buffer value is gtwy:9.3.5.41:
buffer value is _pif:en0:
buffer value is _ssl:no:
buffer value is _psh:no:
buffer value is _res:no:
buffer value is asyn:no:
file descriptor is 13
file descriptor is : 13
family is : 2
source port is : 1023
source addr is : 9.3.5.197
source hostname is : server4
getting 2nd port
count equals 0
got stderr port 0
success: we got 1st write query is 1
success: we got 2nd write local id is

2004

success: we got 3rd write remote id is

2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004
2004

success: we got 4th write command is
buffer value is host:server3:
buffer value is addr:9.3.5.196:
buffer value is mask:255.255.255.0:
buffer value is gtwy:9.3.5.41:
buffer value is _pif:en0:
buffer value is _ssl:no:
buffer value is _psh:no:
buffer value is _res:no:
buffer value is asyn:no:

When comparing the nimquery and the nimsh.log files you can follow the
authentication procedure.
Note: The /etc/niminfo is read only at the nimsh startup. If you make changes
to this file, the nimsh needs to be restarted.

Configure nimsh
In order to get the nimsh running, the NIM server must be configured, for
example through smit nim_config_env, and the NIM client should be defined to
the NIM server, for example through smit nim_mkmac on the NIM master. The

200


NIM client must be configured into the NIM environment using the smit niminit
or the niminit command on the NIM client. We recommend using SMIT instead
of the regular NIM commands because the NIM performs additional checks of
the NIM environment. The default settings will create the client definitions to use
the rsh or rcmd commands.
If you want to use the nimsh environment, you have to log in to both the client
and the server and define the nimsh for each client.
On the NIM client run the smit nim_config_services command and define the
nimsh for the communication protocol used by the client. See the highlighted line
in Figure 5-6 for an example of smit nim_config_services SMIT screen.

Figure 5-6 smit nim_config_services

This SMIT task will update the NIM configuration including the NIM master and it
will start the nimshd subsystem. The changes are written to the /etc/niminfo and
the /etc/environment files.
To enable the nimsh communications, you can alternatively run the niminit
command on every NIM client as follows:
# . /etc/niminfo
# mv /etc/niminfo /etc/niminfo.bak
# niminit -aname=server2 -amaster=server4 -aconnect=nimsh


201

nimsh:2:wait:/usr/bin/startsrc -g nimclient >/dev/console 2>&1
0513-044 The nimsh Subsystem was requested to stop.
# nimclient -C

You can start the already configured nimsh by running the nimclient -C
command (note the uppercase -C) on every client. The nimclient -C command
will not update the /etc/niminfo and the /etc/environment files.
On the NIM server run the smit nim_chmac command and check that the
Communication protocol used by client has changed to nimsh. See the
highlighted line in Figure 5-7 for an example of the smit nim_chmac SMIT screen.

Figure 5-7 smit nim_chmac

The SMIT menu will update the NIM configuration on the NIM master. Now, you
can check the nimsh environment, for example with the nim -o lslpp server2
command.

202


To change the client communications for the server2 client on the NIM server
from the command line, you can alternatively run the nim command as follows:
# nim -o change -aconnect=nimsh server2

This command produces the same result as the SMIT task shown in Figure 5-7.
Once the nimshd subsystem is started, it produces log entries in the
/var/adm/ras/nimsh.log file. The log is used only for debug purposes.

BOS install and nimsh
You can select the nimsh also for the BOS install operation. At the time of writing
there was no SMIT menu to do this. Use the nim command as follows:
nim -o bos_inst -a connect=nimsh -a source=rte -a spot=spot530
-a lpp_source=aix530 server3

5.13.2 NIM cryptographic authentication - OpenSSL
AIX 5L Version 5.3 introduces a cryptographic extension for the NIM. This
cryptographic extension is based on the OpenSSL environment as it is delivered
on the Linux Toolbox for AIX media. Although you can use other sources for the
OpenSSL software, we do not recommend them because the automatic NIM
tools are prepared for the appropriate version of OpenSSL as installed from the
Linux Toolbox for AIX media.
Note: You will need to install the OpenSSL package from the Linux Toolbox for
AIX media directly from the media, or copy it to the lpp_source NIM resource
for later installation by NIM tools.

Install OpenSSL
First you need to install the OpenSSL from the Linux Toolbox for AIX package, or
copy it to the lpp_source NIM resource. We recommend copying the OpenSSL
RPM package from the CD media to the lpp_source NIM resource. Once the
OpenSSL RPM package is in the lpp_source, you can easily install the OpenSSL
in the NIM network directly from the SMIT screens for NIM.
The minimal tested version of the OpenSSL is 9.6g; version 9.7 should also work
fine. In our examples we are using OpenSSL version 9.6m. It is important to have
the same or at least compatible versions installed on the server and the clients.

Prepare a file system for /ssl_nimsh
The SSL configuration requires an exchange directory for encryption keys. By
default, NIM uses the /ssl_nimsh directory that resides in the / (root) file system.


203

We recommend creating a file system named /ssl_nimsh, of minimal size (for
example 16 MB), and mounting the file system.

Enable SSL for NIM server
You have the choice to create the SSL keys and configuration manually or run
the NIM configuration utilities from the smit nim_ssl SMIT screen. The utility
calls programs from the /usr/samples/nim/ssl directory along with the basic NIM
commands. After the utilities finish, the necessary keys are created in the
/ssl_nimsh directory. See Figure 5-8 for the SMIT screen that configures the NIM
to use the SSL environment.

Figure 5-8 smit nim_ssl

As shown in Figure 5-8, setting the Enable Cryptographic Authentication to
enable the utility will set up the SSL environment, including creating keys, for the
NIM master. If you did not install the SSL directly from the CD before, but you
copied the RPM packages to the lpp_source NIM resource, you can set the
Install Secure Socket Layer Software (SSLv3) to yes and the lpp_source which
contains RPM packages to the lpp_source you use. If your installation of the
OpenSSL has failed or is not done you will receive error messages from the NIM
configuration utility.

Enable SSL for NIM client
After the NIM server is configured for SSL you have to enable the clients to use
SSL for NIM. This is done using the smit nim_config_services SMIT screen.
See Figure 5-9 on page 205 for the SMIT screen.

204


Figure 5-9 smit nim_config_services for SSL

Set the highlighted item Communication Protocol used by client to nimsh as the
nimsh environment is responsible for the NIM communications. Set Enable
Cryptographic Authentication to enable. By selecting this item you will switch the
client to SSL communications for NIM.
The clients receive the encryption keys through the TFTP protocol. The keys
from the /ssl_nimsh directory are copied to the /tftpboot directory during the key
exchange protocol. The /tftpboot directory is used for the TFTP protocol as called
by the NIM.

Check the function
Check that the client is switched to the nimsh (secure) communication protocol
using the lsnim -l server2 command on the NIM master:
# lsnim -l server2
server2:
class
=
type
=
connect
=
platform
=
netboot_kernel =
if1
=
cable_type1
=
Cstate
=

machines
standalone
nimsh (secure)
chrp
mp
network1 server2 0
bnc
ready for a NIM operation


205

prev_state
Mstate
cpuid
Cstate_result

=
=
=
=

customization is being performed
currently running
0001810F4C00
success

You can test the communication of the secure nimsh, for example with the
nim -o lslpp server2 command.

Certificate viewing file
The following are examples from a certview certificate viewing script for
OpenSSL certificates. The script is located in the /usr/samples/nim/ssl directory.
The script is provided to help users view hash, issuer, subject, and other
certificate information available using the openssl command. The script can be
modified based on user need or preference.
To print out all readable values for certificates:
# certview certificate_names

To print out the hash value for certificates:
# certview -h certificate_names

To print out the issuer value for certificates:
# certview -i certificate_name

To print out the subject value for certificates:
# certview -s certificate_name

To print out the subject, issuer, and enddate values for certificates:
# certview -I certificate_name

Certificate password loading file
The following are examples from a certpasswd certificate password loading file
for NIM OpenSSL certificates. The file is located in the /usr/samples/nim/ssl
directory. The file is provided to help users store a desired password for
decrypting the NIM master's client key. The password provided must match the
password used to encrypt the NIM master's client key during NIM SSL
configuration.
To load the encrypted key's password in the NIM environment:
#

206

certpasswd


To unload the encrypted key's password from the NIM environment:
#

certpasswd -u

Note: Only the NIM master's client key may be password encrypted.
To password encrypt the NIM master's client key, complete the following steps:
1. On the NIM master, edit the /ssl_nimsh/configs/client.cnf config file.
2. Locate the encrypt_key variable and change the value to yes.
3. Add the output_password variable underneath encrypt_key and specify the
password. To provide the password within the config file, if output_password
is not provided, you will be prompted for the password during key generation.
4. Enter the following command:
# make -f /usr/samples/nim/ssl/SSL_Makefile.mk client

5. On each SSL client, copy the new server.pem file using the nimclient -c
command.
6. Load the password into the NIM environment using the certpasswd
command. When using password encrypted keys, NIM commands may fail
with the following error if the correct password is not loaded:
0042-157 nconn: unable to access the "clientkey.pem" file

Once loaded, the password specified will be used for client key decrypting until
the user unloads the password.

5.14 High Available NIM (HA NIM)
The most significant single point of failure in a NIM environment is the NIM
master. AIX 5L Version 5.3 introduces a way to define a backup NIM master,
takeover to the backup master, and then failback to the primary master. This
helps to create more reliable NIM environments.
Figure 5-10 on page 208 shows a simple functional diagram of the high available
NIM architecture.


207

NIM client
server3

backup
server4
primary
lpp_source
SPOT
mksysb

server2
secondary
lpp_source
SPOT
mksysb

Figure 5-10 High availability NIM

The primary NIM master, server4 is installed in the regular way. It has the regular
installation resources created. The secondary or backup server is server2, which
needs to have the NIM master filesets installed. The NIM master has to run a
backup operation regularly to synchronize the configurations from the primary to
the backup server. At the time of writing only rsh communications are supported.
The takeover is done by an administrator command from the backup server. The
backup server updates the configuration where possible and is enabled to run
operations on the objects. When a fallback is done, the administrator has to
re-synchronize the configurations since it may have changed on the backup
server.
Figure 5-11 on page 209 shows a new menu added in AIX 5L Version 5.3 to
manage the HA NIM environment. You can get this menu by running the smit
nim_altmstr command. The menu is available on the NIM primary master and
also on the NIM backup server.

208


Figure 5-11 smit nim_altmstr

The following sections describe how to configure the HA NIM environment.

Install NIM master filesets
On the backup server install the NIM master and SPOT filesets using regular AIX
procedures. Check them using the lslpp -L command as shown in the following
example.
s2 # lslpp -L "bos.sysmgt.nim*"
Fileset
Level State Type Description (Uninstaller)
---------------------------------------------------------------------------bos.sysmgt.nim.client
5.3.0.0
C
F
Network Install Manager Client Tools
bos.sysmgt.nim.master
5.3.0.0
C
F
Network Install Manager Master Tools
bos.sysmgt.nim.spot
5.3.0.0
C
F
Network Install Manager SPOT

Define the backup master to the primary master
Figure 5-12 on page 210 shows the SMIT screen that defines the backup NIM
server to the primary NIM server. Run the smit nim_mkaltmstr command on the
primary NIM server to receive this screen. The highlighted item shows that the
definition is related to alternate_master, that is, the backup NIM server.


209

Figure 5-12 smit nim_mkaltmstr

You can get the same result by running the nim command as follows:
nim -o define -t alternate_master -aplatform=chrp -aif1=’network1 server2 0’
-anetboot_kernel=mp server2

The SMIT menu will create the necessary definitions into the NIM database of
the backup NIM server.

Initialize the backup master
Figure 5-13 on page 211 shows the SMIT screen that allows you to initialize the
secondary NIM server. The core command that is run from this menu is the
niminit -a is_alternate=yes ... command with additional attributes.

210


Figure 5-13 smit niminit_altmstr

You can get the same results as displayed in the SMIT screen in Figure 5-13 by
running the niminit command as follows:
# niminit -ais_alternate=yes -aname=server2 -amaster=server4 -apif_name=en0
-aplatform=chrp

At this time, the relation between the backup NIM server and the primary NIM
master is established. Note that you will need to synchronize the resources and
the configuration from the primary NIM master to the secondary NIM server.
Tip: Initialize also the reverse relation, when the primary master is also a
backup server. Do it in the same way as you initialized the backup server here.

Perform the synchronization
Before performing the synchronization from the primary master to the backup
server, you should create the basic NIM objects, like lpp_source and SPOT,
manually on the backup server in the same fashion as on the primary server.
The synchronization procedure makes a backup of the NIM database on the
primary master and copies it to the backup server. The backup is then restored
on the backup server and recovers the NIM database using the
nim_master_recover utility.


211

Figure 5-14 shows the SMIT interface for the NIM synchronization. Start this
screen by running the smit nim_mac_op_sync_hdr command.

Figure 5-14 smit nim_mac_op_sync_hdr

Alternatively, you can use the nim command that is more simple in this case:
nim -o sync -aforce=yes server2

When the NIM objects are updated, NIM stores the last modification time in the
NIM database. This attribute is checked when the backups are performed.
Tip: The synchronization is initiated from the machine where you are logged in
to the target machines. If you log in to the backup server and start the
synchronization from there you will overwrite your primary server.

Note: You should consider scheduling regular NIM synchronizations.

Takeover
When there is a need to activate the secondary NIM server to function as the
primary NIM master, you need to use the smit nim_mac_op_takeover_hdr SMIT
screen to do it. See Figure 5-15 on page 213 for the SMIT screen.

212


Figure 5-15 smit nim_mac_op_takeover_hdr

Alternatively, you can use the nim command that is more simple in this case:
# nim -o takeover -aforce=yes server4

This will attempt to contact the primary NIM master and update it. If the contact is
not successful the backup NIM server updates its database with the
sync_required information for the primary master. The backup server attempts to
contact the clients and switch the controlling master. If the client is unreachable,
the backup NIM server updates its database with the sync_required information
for each unreachable client.
The backup server then updates its NIM database and is ready to run operations
as the NIM master.

Fallback
When the original primary NIM master is prepared to reverse takeover the
operations from the backup can be redirected back to the primary in the same
case as in the takeover from the primary to the secondary server.
Now, log in to the primary server and use the smit nim_mac_op_takeover_hdr
SMIT menu or run the nim -o takeover -aforce=yes server2 command.


213

5.15 General NIM enhancements
AIX 5L Version 5.3 provides the following enhancements to the Network Install
Manager (NIM):
Detailed output when creating a NIM lpp_source resource
Creating SPOT resource from a mksysb
Restore SPOT copy function
Adjustments in NIM to process multiple CD media
NIM interface to change network attributes
We describe these enhancements in the following sections.

5.15.1 Detailed output when creating a NIM lpp_source resource
Before Version 5.3, the NIM lpp_source creation did not give any information
about the progress of the procedure. AIX 5L Version 5.3 extends the NIM
lpp_source creation through verbose output. If the verbose output is set the
procedure that creates the lpp_source calls the bffcreate -v or gencopy -bv
commands with the verbose flag turned on.
There is a new show_progress attribute introduced to the nim command that
administrators can modify to turn the verbose attribute on or off. The default
value is to have the show_progress attribute turned on. An example of creating
the lpp_source with showing the progress output follows:
#nim -o define -t lpp_source ... -a show_progress=yes
The SMIT menus for creating the lpp_source are updated to enable the user to
set the show_progress option.

5.15.2 Creating a SPOT resource from a mksysb
Before Version 5.3 the NIM SPOT resource could be created from an existing
lpp_source resource or from the installation media. The disadvantage of such
generic SPOT is the amount of disk space it consumes and the long creation
time.
In AIX 5L Version 5.3 the SPOT can be created either from an installation source
(lpp_source or install media) or from a mksysb resource. The nim command
allows a mksysb resource to be entered as the source attribute when creating a
SPOT resource.
#nim -o define -t spot ... -a source=mksysb_resource spot_name

214


Create a mksysb NIM resource before you create the SPOT. You will need to
substitute the reference to a NIM mksysb resource if you want to create the SPOT
from a mksysb. An example that creates the SPOT from a mksysb follows:
# mksysb -i /export/images/aix530v1.mksysb
...
# nim -o define -t mksysb -aserver=master
-alocation=/export/images/aix530v1.mksysb mksysb530v1
...
# nim -o define -t spot -aserver=master -alocation=/export/spot530
-asource=mksysb530v1 spot530
...
The SMIT menus for creating the lpp_source are updated to enable the user to
set the show_progress option.
The SPOT is created from the mksysb using the /usr/lib/bootpkg/bootpkg_list file
that lists only the necessary files for the SPOT. This results in creating a quite
small SPOT, that is approximately 50 MB in size. When the SPOT is created
from the lpp_source, the only way to create a universal SPOT is to run the
installp command for all devices, which produces a large SPOT whch is about
400 MB.

5.15.3 Restore SPOT copy function
In AIX 5L Version 5.3, the restore SPOT copy function was reintroduced. The
user can select the SPOT residing on the NIM server to install BOS on the
clients. The nim and the nimclient commands are modified to accept the SPOT
as source. An example for the nim command follows:
# nim -o bos_inst -a source=spot -a spot=spot530 -a lpp_source=aix530
-a no_client_boot=yes server3
The command will allocate the SPOT resource named spot530, along with other
resources, and the client can be rebooted to perform the BOS installation from
the SPOT.
Several SMIT menus related to BOS installation are updated and can accept the
SPOT as the source for the BOS installation.

5.15.4 Adjustments in NIM to process multiple CD media
The BOS installation of previous versions of AIX was limited to one CD media.
Future releases of AIX BOS installation may require more that a single CD media


215

for the BOS installation. An enhancement to the NIM procedures to handle the
installation CD media was done so that NIM handles multiple installation media.
When NIM copies the installation files from the CD media to disk, it uses the
gencopy command. The gencopy command uses bffcreate to package the files
from the media. The gencopy command accepts a subset of flags used for
bffcreate specified after the -b flag, and then forwards these options to the
bffcreate command.
The default behavior of the bffcreate command is to handle multiple media. The
-S flag used in bffcreate suppresses multi-volume handling of the input source.
When creating the lpp_source with the nim command or through SMIT, you can
set the creation to process or not to process multiple volumes by using the
-a multiple_volumes=yes attribute to the nim command. An example of the nim
command follows:
nim -o define -t lpp_source -aserver=master -alocation=/export/aix530
-asource=/dev/cd0 -amulti_volume=yes

5.15.5 NIM interface to change network attributes
In AIX 5L Version 5.2 there was already a nim_master_recover tool that could
recover the NIM master configuration. AIX 5L Version 5.3 enhances the
nim_master_recover tool so that it can recover the NIM master after the network
attributes change. This helps the administrators to more easily change the IP
address of the NIM master. In addition there is a new SMIT menu that addresses
the NIM network changes. See Figure 5-16 for the smit nim_ch_if1 screen.

216


Figure 5-16 smit nim_ch_if1

The correct procedure to change the network attributes in a NIM environment is
as follows:
1. Add the new hostname and IP address to the /etc/hosts, DNS, or other name
resolving service.
2. Update the /.rhosts on the clients if using it.
3. Run the smit nim_ch_if1 SMIT menu that changes the NIM configuration.
4. Change the TCP/IP configuration in AIX.
5. Check the environment.
6. Delete old NIM definitions.
Changing the NIM master network configuration through the SMIT menu in
Figure 5-16 on page 217 will do the following:
The current NIM database is first saved by the m_backup_db method.
The NIM network resources are updated in the NIM database.
The NIM master’s primary interface (if1) is updated to include the new
network, hostname, and adapter information.
The NIM client /etc/niminfo files are updated on the clients.


217

If the procedure fails on a client for any unexpected reason, such as incorrectly
updated DNS or /.rhosts, you will need to re-initialize the client configuration
using the niminit command.

5.15.6 VIPA and EtherChannel support in NIM adapters
In AIX 5L Version 5.3 the nimadapter command is extended to accept definitions
of VIPA and EtherChannel secondary adapters.

5.15.7 EZ NIM enhancements
EZ NIM is an easy to understand user interface to the NIM environment. EZ NIM
does the following:
Creates an lpp_source resource when setting up the master environment.
Adds nimsh and cryptographic authentication when setting up the master and
reinstalling the clients.
Adds an rte option to be selected for client reinstall.
Adds install options of overwrite, preservation, and migration for client
reinstall.
Adds new options for viewing the NIM environment.
We do not show examples of SMIT screens here because after you run the
smit eznim fast path, you get into the easy to understand, self-explanatory
menus of the EZ NIM.

5.16 Alternate Disk Installation migration
Alternate Disk Installation Migration (ADM) is an Alternate Disk Installation option
using NIM available whenever the version or release values for an AIX release
change. These are the V and R values in the VRMF
(VersionReleaseMaintenanceFix), which is 5.3.0.0 for this release of AIX 5L.
ADM allows the users to migrate their systems on the side without requiring
downtime for installation. ADM is supported at 4.3 and higher. ADM is currently
only available for migration to 5.1 or 5.2. This function is now added to AIX 5L
Version 5.3.
Alternate disk migration installation has the following requirements:
Configured NIM master running AIX 5.1 or later with AIX recommended
maintenance level 5100-03 or later.

218


The NIM master must have bos.alt_disk_install.rte installed in its rootvg and
the SPOT which will be used.
The level of the NIM master rootvg, lpp_source, and SPOT must be at the
same level.
The client (the system to be migrated) must be at AIX 4.3.3 or later.
The client must have a disk (or disks) large enough to clone the rootvg and an
additional 500 MB (approximately) of free space for the migration. The total
amount of required space will depend on original system configuration and
migration customization.
The client must be a registered NIM client to the master.
The nim master must be able to execute remote commands on the client
using the rshd protocol.
The client must have a minimum of 128 MBs of memory.
A reliable network, which can facilitate large amounts of NFS traffic, must
exists between the NIM master and the client.
The client's hardware should support the level it is migrating to and meet all
other conventional migration requirements.
The following limitations apply to alternate disk migration installation:
If the client's rootvg has the Trusted Computing Base option enabled, either
disable it (permanently) or perform a conventional migration. TCB must
access file metadata that is not visible over NFS.
All NIM resources used must be local to the NIM master.
During the migration, the client's active rootvg may experience a small
performance decrease due to increased disk I/O, nfsd activity, and some CPU
usage associated with alt_disk_install cloning.
NFS tuning may be required to optimize performance.

5.17 Enhancements to Alternate Disk Installation
AIX 5L Version 5.3 has implemented a number of changes to make the
alt_disk_install operations easier to use, document, and maintain.
The following functional changes have been implemented:
alt_disk_install has been partitioned into separate modules with separate
syntax based on operation and function.
A library of common functions that can be accessed by the modules has been
implemented.


219

Error checking and robustness of existing alt_disk_install operations has been
improved.
Documentation has been improved by creating a separate man page for each
module (currently there is one extremely large man page).
The following three new commands have been added:
alt_disk_copy creates copies of rootvg on an alternate set of disks.
alt_disk_mksysb installs an existing mksysb on an alternate set of disks.
alt_rootvg_op performs Wake, Sleep, and Customize operations.
Also, a new library, alt_lib, has been added that serves as a common library
shared by all alt_disk_install commands.
The alt_disk_install module will continue to ship as a wrapper to the new
modules. However, it will not support any new functions, flags, or features.
Table 5-3 shows how the existing operation flags for alt_disk_install will map to
the new modules. The alt_disk_install command will now call the new modules
after printing an attention notice that it is obsolete. All other flags will apply as
currently defined.
Table 5-3 alt_disk_install command arguments
alt_disk_install command arguments

New commands

-C <args> <disks>

alt_disk_copy <args> -d <disks>

-d <mksysb> <args> <disks>

alt_disk_mksysb -m <mksysb> <args> -d
<disks>

-W <args> <disk>

alt_rootvg_op -W <args> -d <disk>

-S <args>

alt_rootvg_op -S <args>

-P2 <args> <disks>

alt_rootvg_op -C <args> -d <disks>

-X <args>

alt_rootvg_op -X <args>

-v <args> <disk>

alt_rootvg_op -v <args> -d <disk>

-q <args> <disk>

alt_rootvg_op -q <args> -d <disk>

5.18 Advanced Accounting
The Advanced Accounting subsystem is based on mainframe technology and
features interval accounting, data aggregation, and dynamic classification of
accounting data. You can customize Advanced Accounting for different

220


computing environments. You can configure Advanced Accounting to produce
the specific types of records needed for billing applications. By default, all
accounting records are produced.
Advanced Accounting provides usage-based information for a wide variety of
system resources so that you can develop comprehensive charge-back
strategies. You can collect accounting data on resources such as disks, network
interfaces, virtual devices, file systems, processors, and memory. Interval
accounting gives you the ability to view this data over system
administrator-defined time intervals in order to develop chronological views. This
has several potential applications, including capacity planning.
Interval accounting can be used to periodically record accounting data, enabling
the production of more accurate bills. You can configure Advanced Accounting to
produce intermediate process records for active processes. These records can
be added to the completed process records to produce a bill that reflects the total
use of system resources.
Data aggregation is a way to control the amount of data written to a data file. This
helps system performance by reducing the overall resource load needed to run
Advanced Accounting. Aggregation minimizes a system’s I/O requirements,
adding accounting records so that fewer records are written to the accounting
file. It is transparent to applications and middleware.
Policies are rules that provide for the automatic classification of processes.
Classification is done according to users, groups, and applications, categorizing
the use of system resources by billable entities. These categories are called
projects.
APIs are provided so that applications and middleware can describe the
transactional nature of their workloads, enabling charge-backs for server
processes. These APIs are intended to define and delineate transactions and to
identify the end user, if possible. The Advanced Accounting subsystem measures
the resource use of transactions, if possible, and records all of this information in
the accounting file.
Advanced Accounting also provides new statistics from previous accounting
tools. For example, the process record provides microsecond-level CPU times, a
memory integral based on elapsed time (standard UNIX accounting bases it on
CPU times), local and distributed logical file I/O, and local and remote socket I/O.


221

5.18.1 Data files
Every statistic recorded by Advanced Accounting is written to a data file. When
the data file reaches its capacity, the billing application can process the file. After
it is processed, that file can be reused, and the cycle repeats.
You must create your accounting data files in order to begin collecting accounting
statistics. This involves assessing your company’s needs and determining how
much data you will be collecting. You can start Advanced Accounting and let it
run for a period of time to see how much data is produced, giving you an idea of
how much disk space to reserve for accounting, as well as how many files you
will need to manage your data.
Although not required, it is a good idea to specify at least two data files so that
Advanced Accounting can remain active at all times. Advanced Accounting
writes to only one file at a time, but as it does, it requires exclusive access to this
file. With two files, Advanced Accounting can write to one file while the other file
is processed by the billing application.
The following explains the life-cycle of a data file as it moves through the
accounting process:
1. A pool of pre-allocated data files is created and registered with the Advanced
Accounting subsystem.
2. As accounting statistics are collected from the system, data is written to the
active data file.
3. The size of the data file (predetermined when the file is created) limits the
amount of data that can be written to the file. Once the data file is full, no
more data may be written to it. The Advanced Accounting subsystem will then
switch to the next empty data file and begin writing data to it.
4. Once the data file is full, you can use post-processing tools, such as a billing
application, to process the accounting data into statistics for billing purposes.
5. After the data is extracted from the data file and processed, the file is reset so
that it can be reused.
The Advanced Accounting subsystem can be configured to send the system
administrator notification when the data file is 90 percent full and when the data
files get switched. These notifications can be delivered either through the syslog
daemon or e-mail. These two notifications tell the system administrator that the
data needs to be extracted from the data file and processed.
When a data file is created, it is registered with the kernel. This is how the kernel
knows the identity of the file and can automatically switch. Table 5-4 provides a
list of the tasks and commands used with advanced accounting.

222


Table 5-4 Data file management commands
Task

Command

SMIT fast path

Allocate and define an
accounting file with specified file
name and size. The default size
is in megabytes.

acctctl fadd file size

smit create_aacct_file

Remove the specified
accounting file from the
accounting subsystem. This will
not remove the file from the file
system.

acctctl frm file

smit release_aacct_file

Indicate that the specified file
can be reused by the accounting
subsystem.

acctctl freset [file]

smit reuse_aacct_file

Query the state and current use
of the specified file. If no file is
specified, all files are queried.

acctctl fquery [file]

smit list_aacct_file

Force Advanced Accounting to
switch to a new accounting data
file. The new file may be
optionally specified.

acctctl fswitch [file]

smit switch_acct_file

5.18.2 Projects and policies
Projects represent billable entities such as users, departments, divisions,
companies, or tasks. Each project is composed of a project number, project
attribute, and project name, which collectively represent a project definition.
Project definitions are entered into the project definition database.
Projects are written to accounting records. Report and analysis commands
convert project numbers into project names by looking up entries in the system
project definition database. Logically, projects are indices into critical business
data (for example, customer name, billing address, account number, service level
agreement) that are maintained by the billing application.
Project numbers are assigned through project assignment policies, which are
composed of project assignment rules. Each rule contains classification criteria
that when fully satisfied, yield a classification result. The classification result is a
project list that is logically assigned to the object, usually a process, being
classified. The classification criteria depends on the type of policy.
Project lists enable manual project assignment. If a list of projects is specified,
users can change their current project assignment to another project in the list.


223

This provides the capability to launch jobs under different projects, which is
useful when you are performing work on behalf of multiple clients. The system
administrator can assign multiple projects to any user, then manually change the
project assignment.

Policies
Policies automate project assignment. A policy is comprised of classification
criteria and classification results. The project assignments take place during
subroutines and kernel services, such as exec(), initp(), setuid(), and setgid().
Data can be classified by user, group, application, or a combination of these
attributes. Depending on the type of policy that is created, administrators can
specify a user name, group name, application name, and project list in the policy
file, although all four components do not have to be present for a policy to
function.
Processes can be assigned in the following two ways:
Using assignment rules when process classification attributes change. This is
the most common way that processes are classified.
Manual assignment to a class by a user with the required authority.
Advanced Accounting supports the following types of assignment policies for
processes, each with different classification criteria:
Admin policy
User and group policies
The Admin policy supports most filters and the use of Korn shell syntax wild
cards. Configuring and maintaining Admin policies requires a standalone
administration tool, such as Web-based System Manager, SMIT, or an editor.
The User and Group policies provide project assignment based exclusively on
user names or group names, depending on the policy. They are integrated with
user administration. Project lists can be specified for users and groups when
these users and groups are created.

Admin policy
The Admin policy uses the user name, group name, and application name
process attributes to classify processes. The Admin policy is application-based,
and provides the ability to gather accounting statistics at the application level.
By default, the Admin policy is located in the /etc/project directory. You may
create alternate Admin policies to use at various times. For example, you may
want to have an Admin policy that runs Monday through Friday, and another that

224


runs on Saturday and Sunday. The alternate Admin policies are stored in
subdirectories of the/etc/project/alter root directory. You must specify a name for
the subdirectory when you are creating a new Admin policy. The Admin policy
consists of one or more assignment rules that are placed in the Admin policy file,
and conform to the following syntax:
user name:group name:application name:Project List::Optional comments go here

For details on the values that are allowed in each field of the assignment rule,
see Table 5-5.
Table 5-5 User, group, and application rules
Type of rule

Description

user

May contain either a hyphen (-) or at least one valid user name as
defined in the /etc/passwd file. An exclamation point (!) can be used
before a name to exclude a given user from the class. The list of user
names is composed of one or more names, separated by a comma (,).
Patterns can be specified to match a set of user names, using full Korn
shell pattern-matching syntax. If an invalid user name is used,
Advanced Accounting will ignore the entire rule. If a hyphen (-) is used,
Advanced Accounting will skip to the next field in the rule.

group

May contain either a hyphen (-) or at least one valid group name as
defined in the /etc/passwd file. An exclamation point (!) can be used
before a name to exclude a given user from the class. The list of group
names is composed of one or more names, separated by a comma (,).
Patterns can be specified to match a set of user names, using full Korn
shell pattern-matching syntax. If an invalid user name is used,
Advanced Accounting will ignore the entire rule. If a hyphen (-) is used,
Advanced Accounting will skip to the next field in the rule.

application

May contain either a hyphen (-), a list of application path names, or the
command name of a kernel process. This is the path name of the
applications (programs) run by processes included in the class. The
application names will be either full path names or Korn shell patterns
that match path names. The list of application names is composed of
one or more path names, separated by a comma (,). An exclamation
mark (!) can be used before a name to exclude a given application. At
least one application in the list must be found at load time or the rule is
ignored. Rules that are initially ignored for this reason might become
effective later on if a file system is mounted that contains one or more
applications in the list.

For process assignment policies, the classification criteria is the user name as
listed in the /etc/passwd file, the group name as listed in the /etc/groups file, and
the fully qualified application name. The classification result is a project list. By


225

default, the first project in the list is used, unless the user changes his current
project assignment to a secondary project in the list.
Classification is done whenever an attribute value changes by comparing the
value of the process attribute against a list of possible values given in the class
assignment rules file, called rules. The comparison determines which rule is a
match for the current value of the process attributes.
To classify the process, Advanced Accounting examines the top-level admin
policy for the active configuration to determine in which class the process
belongs. For each rule in the file, Advanced Accounting checks the current
values of the process attributes against the values and lists of values specified in
the rule. Advanced Accounting goes through the rules in the order in which they
appear in the admin file, and classifies the process in the project corresponding
to the first rule for which the process is a match. Therefore, the order of the rules
in the rules file is significant.
The following is a list of the criteria used to determine whether the value of a
process attribute matches the values of the same attribute field in the admin
policy file:
If the field in the rules file has a value of hyphen (-), then any value of the
corresponding process attribute is a match.
If the value in one of the fields is preceded by an exclamation point (!), that
value is always excluded.
If the value in one of the fields is followed by an asterisk (*), then every match
with that value will be recognized.

Examples of admin policy rules
The Admin policy allows you to specify more than one user, group, or application
in their respective fields. For instance, if you wanted user1, user2, and user3 to
be given the same attributes, you would specify the following syntax:
user1,user2,user3:-:-:Project List::Comments
This syntax shows that user1, user2, and user3 will be treated the same way for
this rule. The dashes in the group name and application fields are wildcards. You
can also use an asterisk (*). As previously mentioned, you can also use
wildcards to include all the values of a certain attribute. For example, if you
wanted to include every user name beginning with B, you would type B* in the
User Name field. Full Korn shell pattern matching syntax is allowed in all of the
fields of the rule.
You can also set up your Admin policy to include certain users, but exclude
others:

226


user1,!user2,user3:-:-:Project List::Comments
This syntax shows that user1 and user3 will have the same attributes, but the
policy will exclude user2.
Note:
1. The kernel only reads numeric values. In these example, the user names
user1, user2, and user3 are converted to numeric values after they are
loaded into the kernel.
2. If any changes are made to the policy, it must be reloaded into the kernel
using the projctl command.

Aliases
You can define an alias name for a list of users or groups so that you do not have
to enter the full list of users, groups, or application names in a rule. Aliases
provides a shorthand for referring to lists of users and groups, which simplifies
admin policies, and makes them more readable and easier to maintain. A
separate alias file is used with each Admin policy and is placed in the same
directory as its associated Admin policy.
To create an alias where user1, user2, and user 3 are grouped into an alias
named dev1, define it as follows:
dev1:user1,user2,user3::Development team 1
To use the alias name in the Admin policy, the alias name must be preceded by
a dollar sign ($). For example, using the dev1 alias instead of user1, user2, and
user3 in the Admin policy looks like similar to the following:
$dev1:-:-:Project List::This is a rule within a user alias
Multiple aliases, separated by a comma, may also be used. Aliases are excluded
by placing an exclamation point (!) before the alias name in the Admin policy rule.

Alternate admin policies
It is possible to create multiple Admin policies to reflect different billing strategies
based on the time of day or the day of the week. The projctl command is used
to load alternate Admin policies. Previously loaded Admin policies do not have to
be unloaded to load a new policy. The cron facility is used to load a policy at the
given time.
Alternate Admin policies are placed in the /etc/project/alter directory. For
example, an alternate Admin policy with the name “weekend” is placed in the
Admin file /etc/project/alter/weekend/admin. If this policy used aliases, they
would be placed in the /etc/project/alter/weekend/alias file. SMIT and Web-based


227

System Manager are used to create, change, show, load, unload, and remove
Admin policies. Alternate policies are addressed in SMIT by changing the current
focus to the name of the policy.

Relative project classification
With Advanced Accounting, you can enable project assignments to be made
relative to User IDs and Group IDs. To accomplish this, use the $UID and $GID
keywords in a project list to tell the system that the project code must be
computed. A simple expression in the form “keyword” or “keyword + constant”
may be used, where “constant” is either a decimal or hexadecimal number
(0xffff).
The following rule (Table 5-6) illustrates the use of relative project classification.
Table 5-6 Relative project classification
User

Group

Application

Project

*

-

-

$UID+1000000

Disable accounting for selected processes
System administrators can disable accounting for selected processes through
the classification process. This can be accomplished by specifying a
NoAccounting project list. This attribute is inherited from the parent process to
the child process. Table 5-7 illustrates how this may be accomplished.
Table 5-7 Example of the No Accounting specification in a project list
User

Group

Application

Project

db2inst1

db2inst1

~db2inst1/sqllib/adm/db2start

NoAccounting

root

root

kbiod

NoAccounting

DB2® is classified in the first rule by the subroutine exec(). The NFS kproc in the
second rule is classified by the subroutine initp().

User and group policies
The user and group policies use the process attributes for users and groups. A
user or group name and a project list constitute a project assignment rule within
the user or group policies. There is no file associated with a user or group policy
because user policy data is stored in a security database.
Table 5-8 shows the structure of a user policy.

228


Table 5-8 Structure of a user policy
User name

Project list

user1

project1, project2

user2

biology, chemistry

To create an example group policy, complete the following:
1. Create an example group dev by typing the following:
# mkgroup dev
2. Create an example project biology_dev with the project ID 1000 by entering
the following:
# projctl add biology_dev 1000
3. Associate the project biology_dev with the newly created group dev:
# chgroup projects=biology_dev dev
To create a user policy, complete steps one through three, substituting mkuser
and chuser for mkgroup and chgroup. Once a user or group has been created, a
project is associated with it. The users or groups can have several projects
associated with them, but only one can be active at a time.

Classification semantics
For every exec(), initp(), setuid(), and setgid() subroutine, the process will be
reclassified using the project assignment rules to determine if the current project
assignment should be changed. If a project assignment rule is not loaded, or if a
rule cannot be successfully applied, then the current project identifier is used.
The default project system identifier is zero (0). It is applied to the base system
processes before accounting is enabled and it may be used to signify general
system overhead. After it is assigned, a project is inherited from the parent
process to the child process using the fork() kernel service and creatp() kernel
service.
The use of the application filter varies between the initp() kernel service and the
exec() subroutine. In the former, the command name of the kernel process that is
being started is used to perform the classification. The command name is visible
through the ps command. In the latter, classification is performed using the file
identifier (FID) of the executable, with the FID of the fully qualified path name that
is specified by the policy. Only FIDs that can be computed at policy load time are
accommodated.
If a process is classified through a rule that explicitly names an application, then
that project identifier should be applied to the process and its children, because


229

the intent is to label a block of work. This implies that subsequent exec(),
setgid(), and setuid() subroutines in the process as well as its children do not
perform project reclassification. Internally, these processes are identified with a
sticky bit that indicates they were classified through a rule that explicitly named
an application. The sticky bit is visible through the use of the new -P flag of the ps
command. The ps -P command displays the project name and subproject
identifier for the project. If the sticky bit is set for the process, the project name
will be preceded by an asterisk (*) character.
When you load a new policy file, all of the processes in the system are
reclassified, except those with a sticky bit. Processes with a sticky bit cannot be
changed. When a different project identifier is assigned to a process, a process
accounting record is written to the accounting data file, so that the use of
resources by a project may be accurately reported. Whenever this occurs, the
accounting statistics for the process are reset to zero.

Manual project classification
Users can manually change their current project assignment. For non-privileged
users, authorization is provided through the policy file. The first project listed in
the rule is considered to be the default project and is automatically chosen,
unless the user changes his current project assignment to another project in the
list. Changing the current project assignment applies only to the current session
and it is not silently applied to other jobs with the possible exception of the parent
process (for example, the shell).
Project lists may also be specified in rules that name applications, allowing
administrators to associate projects with specific instances of applications. This
is important for server processes, which may be externally configured to manage
specific resources that have different cost structures.
This mechanism enables system administrators to charge indirectly for those
resources that may not directly support usage-based accounting. For example, a
system administrator that wants to charge different rates to access specific
databases could start different instances of Oracle to manage those databases.

Commands
Table 5-9 shows the commands used to administer projects and policies.
Table 5-9 Projects and policies commands
Task

Command

SMIT fast path

Add a project definition

projctl add projname

smit add_proj

Load or reload a project definition

projctl ldprojs [-r -a]

smit load_proj

230


Task

Command

SMIT fast path

Remove the specified project
definition

projctl rm projname [-d
admnpath]

smit rename_proj

Show or change active project
definition

projctl chg projname [-p pid]

smit show_chg_proj

Merge two project definitions

projctl merge projpath

None

Start a program with a project
assignment

projctl exec projname cmd line

smit start_proj_prg

Enable or disable aggregation for
a project

projctl chattr agg projname

None

Show the policies that are loaded
on the system

projctl qpolicy

smit query_policy

List all active project definitions

projctl qprojs [-n]

None

List the specified project definition

projctl qproj projname

smit show_chg_proj

Change the project assignment
for a process

projctl chg

smit chg_proj_proc

Show project assignment for a
program

projctl qapp appname

smit show_proj_pgm

Load policies

projctl {ld |{{{{ldall | ldadm}[-d
admnpath]} | ldgrp | ldprojs |
ldusr}[-a]}

smit load_admin
smit load_users
smit load_groups

Unload policies

projctl {unld |{{{{unldall |
unldadm}[-d admnpath]} | unldgrp |
unldprojs | unldusr}[-a]}

smit unload_admin
smit unload_users
smit unload_groups

Create an admin policy

None

smit create_admin

Show or change current focus
policy

None

smit change_show_focus

Remove an admin policy

None

smit remove_admin

Add a rule

None

smit add_admin_rule

Remove a rule

None

smit remove_admin_rule

Add a user alias

None

smit add_usr_alias

Add a group alias

None

smit add_grp_alias


231

Task

Command

SMIT fast path

Show or change the specified
alias

None

smit chg_alias

Remove the specified alias

None

smit remove_alias

Manage projects

None

smit work_project

Remove project definition

None

smit remove_admin_proj

Create a project list for a user

chuser user

smit create_user

Create a project list for a group

chgroup group

smit create_group

Show or change a specified
project list for a specified user

chuser projects=projectlist
user

smit change_show_user_list

Show or change a project list for a
group

chgroup group

smit change_show_group_list

Remove a project list for a user

chuser projects=user

smit remove_user

Remove a project list for a
specified group

chgroup projects=group

smit remove_group

Show project lists for all users

lsuser -a ALL

None

Show project lists for all groups

lsgroup -a projects ALL

None

5.18.3 Transactional accounting
Applications use the Application Resource Management (ARM) interfaces to
describe the transactional nature of their workloads. Advanced Accounting
supports ARM interfaces by recording information that is presented through
these interfaces in the accounting data file. To use this information for charge
back purposes, it is necessary to understand the programming model associated
with the ARM interfaces and the mechanism for passing critical business
information to the Advanced Accounting subsystem, so that this information can
be preserved for your billing application.
The following provides a list of the known ARM interfaces:
arm_register_application
arm_start_application
arm_register_transaction
arm_start_transaction
arm_block_transaction

232


arm_unblock_transaction
arm_bind_transaction
arm_unbind_transaction
arm_stop_transaction
arm_stop_application
arm_destroy_application
The ARM APIs provide a way to delineate application transactions. This enables
the operating system to identify them and Advanced Accounting to measure and
record them in the accounting data file. The ARM APIs also enable applications
to describe their transactions, so that site-specific information can be provided.
This is accomplished largely by convention through the use of a named set of
parameters that are recognized by the ARM implementation and Advanced
Accounting.
Advanced Accounting also captures information that is generated internally. This
information cannot be changed by the user, but is needed by accounting.
The ARM programming model has a fundamentally hierarchical structure.
Transaction instances are derived from transaction definitions that are defined
when registering transactions. Application instances are derived from application
definitions that are defined when registering applications. When starting a
transaction, a transaction definition that should be used within an application
instance is specified, enabling the full set of information to be defined for each
transaction instance. For example, all transactions have application and group
names.
Advanced Accounting supports the ARM interfaces through a hierarchy of
records:
Application environment record
Transaction environment record
Transaction instance record
The application environment record describes a unique combination of
application information including the application name, application group name,
and properties. Each application environment record is assigned a unique
identifier so that it can be referred to symbolically. This identifier is called the
application environment identifier, and it is included within the transaction
instance record.
The transaction environment record describes a unique combination of
transaction information, including the transaction name and properties. Each


233

transaction environment record is assigned a unique identifier so it can be
referred to symbolically. This identifier is called the transaction environment
identifier, and it is included with the transaction instance record.
Application environment identifiers and transaction environment identifiers are
long lived, but they are not persistent in nature. Each time the system boots, the
identifiers are regenerated. Advanced Accounting circumvents this problem by
recording the application and transaction environment in each file so report and
analysis commands should use the entry in the current file to determine the
unique combination of values that should be applied to a transaction.
The transaction instance record is used to describe each individual transaction. It
includes the application environment identifier and the transaction environment
identifier defined previously. These identifiers should be used to look up the
corresponding application environment record and the transaction environment
records so that the application names, application group names, and transaction
names can be associated with the transaction instance. The intent is to minimize
the amount of data that needs to be recorded with each transaction instance
since a lot of it is static in nature.
Advanced Accounting also supports the aggregation of ARM accounting data.

5.18.4 Interval accounting
Interval accounting provides a way to collect accounting data at specified
intervals. You can configure it to produce intermediate process records for active
processes. These records can be added to the completed process records to
produce a more accurate bill reflecting the total use of the system by a user in a
given time period. Interval accounting can also be configured to periodically
capture accounting data for system resources like processors, memory, disks,
network interfaces, and file systems. This information can be used to generate a
bill that reflects the total use of a partition.
Usage information about system resources has other applications beyond
charge back. It can be used to perform capacity planning because it shows the
use of physical resources like processors, memory, disks, and network
interfaces. Interval accounting provides a time-based view of these resources, so
that load can be determined, enabling capacity-based decisions to be made
based on the data. For example, it is possible to determine the idle processor
capacity in a given interval, which can be used to determine whether more
processors are needed.
Interval accounting provides a historical view of resource use, which can be used
for performance analysis. For example, the file system records can be examined
to determine which file systems were busy when access to the system was slow.

234


The data identifies multiple busy file systems served by the same disk adapter.
This information can be used to balance file systems over disk adapters. You do
not always know which file sets are being accessed at a given time, so having the
ability to replay a log with this information is an asset.

System interval
System interval accounting collects resource use information about system
resources such as processors, disks, network adapters, file systems, and
memory usage. This information is used to profile the use of your system. For
most purposes, an interval of once per hour should be sufficient. The system
interval does not collect process accounting data.

Process interval
The process interval is used to capture information about long-running jobs such
as data modeling applications that run for months. The standard process
completion record is sufficient for most processes. Therefore, the process
interval should be relatively large, so that it does not produce unnecessary data.
The process interval writes a record for every active process, including newly
started jobs. In some situations this can amount to a large number of records, so
the process interval should be set to capture information once per day (1,440
minutes). However, if process records are being aggregated automatically by the
system, then it is recommended that the process interval be set for once per
hour.

Enabling interval accounting
By default, interval accounting is turned off. Table 5-10 shows the commands you
can use to turn on and turn off system-level and process-level interval
accounting.
Table 5-10 Interval accounting commands
Task

Command

SMIT fast path

Enable system interval accounting, specified
in number of minutes by the time parameter,
or turn off system interval accounting.

acctctl isystem {time | off}

smit system_interval

Enable process interval accounting,
specified in number of minutes by the time
parameter, or turn off system interval
accounting.

acctctl iprocess {time | off}

smit process_interval

Query the state of Advanced Accounting.

acctctl

None


235

5.18.5 Data aggregation
Data aggregation is a method of accumulating data into an accounting record
that is otherwise presented through multiple records. Aggregated data is written
periodically according to the intervals specified.
Note:
1. Interval accounting must be enabled in order to use data aggregation.
2. It is recommended that you set the process interval and system interval to
60 minutes.
Accounting data is totaled inside the kernel with no impact to applications or
middleware. The data is made available by interval accounting, which is a kernel
function that periodically writes these records to the accounting data file. When
the kernel aggregates records, it maps them to a set of aggregated records that
are managed internally. These records are pending in the sense that they have
been input to the system, but have not been committed to the recording
mechanism inside the Advanced Accounting subsystem. Interval accounting is
used to commit the aggregated records so that they are written to the accounting
data file.
Because aggregated data is recorded using different data structures, you must
verify that the billing application recognizes these structures. Consult the
documentation provided with your billing application to determine if your billing
application supports data aggregation. Data aggregation can be enabled or
disabled at the system level or project level.

Enabling system-level data aggregation
Table 5-11 lists the commands you can use to turn on and off process data
aggregation, kernel extension data aggregation, and ARM transaction data
aggregation.
Table 5-11 System-level data aggregation commands
Task

Command

SMIT fast path

Enable or disable system-wide process
aggregation

acctctl agproc {on | off}

smit system_paggr

Enable or disable system-wide aggregation for
third-party kernel extensions

acctctl agke {on | off}

smit system_kaggr

Enable or disable system-wide aggregation for
ARM transactions

acctctl agarm {on | off}

smit system_aaggr

Query the overall accounting state

acctctl

None

236


Enabling project-level data aggregation
Table 5-12 shows the commands to turn on and off aggregation for individual
projects.
Table 5-12 Project-level data aggregation commands
Task

Command

SMIT fast path

Enable or disable aggregation for a project,
specified by the projname parameter

projctl chattr agg projname
{-s|-u} [-d projpath]

None

Query the aggregation state for all projects

projctl qprojs [-n]

None

Query the aggregation state for a project, specified
by the projname parameter

projctl qproj [projname]

None

5.18.6 Report and analysis
The Advanced Accounting subsystem provides accounting data for a wide
variety of resources that are typically included within chargeback mechanisms,
but does not provide report and analysis scripts to produce customer bills. These
tools are left to you to implement.
AIX documents the format of the accounting data file as well as the format of the
individual accounting records in the /usr/include/sys/aacct.h header file. In
addition, AIX provides the sample program readaacct with source code that
parses the accounting data file. This command may be used to examine
accounting data and import accounting data into spread sheets. The following
shows the usage message for the readaacct command:
# /usr/samples/aacct/readaacct -?
/usr/samples/aacct/readaacct: illegal option -- ?
Usage : /usr/sample/bin/readaacct [-F file ] [ -t trid ] [-b begin_time ] [-e
end_time ] [-c] [-h]
-F file
: Specify Accounting File
-t trid
: Prints the details of the specific transaction ID
-b begin_time : Prints the transaction records collected on and after the
time specified
The date format should be MMddhhmmss
-e end_time
: Prints the transaction records collected before the end time
The date format should be MMddhhmmss
-c
: Prints the records in colon separated format
-h
: Prints the Accounting File Header

The source code for this command is shipped in the same directory as the
command so that you can modify the command to produce the relevant


237

information in the required format for your specific billing application. By default,
the command displays text strings.
The following is an example of the output displayed if you are running the sample
command just after the /var/aacct/acctdata accounting data file was created:
# /usr/samples/aacct/readaacct -F /var/aacct/acctdata -h
File Name=/var/aacct/acctdata
Version=1
Flags=0
Offset=4096
File Size=104857600
State=0
ID=0
First Time=0
Last Time=0
System ID=IBM,01100316A
System Model=IBM,7028-6E1
Host Name=server3
Partition Name=NULL
Partition Number=-1

5.18.7 Data file management
The Advanced Accounting subsystem continuously writes accounting data to
registered accounting data files. It is necessary to monitor the state of these files
to ensure that Advanced Accounting always has a place to record data.
Advanced Accounting produces messages to monitor the state of data files and
the subsystem status. Use messages related to the data files to trigger actions
based on events as they occur. A message is produced when a file is 90 percent
full and when it is 100 percent full, indicating that administrative action is needed.
The acctctl command is used to manage accounting data files. You can use the
cron command to periodically execute a shell script or command that examines
and manages accounting data files.
Note: By default, AIX records information about the Advanced Accounting
subsystem using the syslog daemon.
E-mail notification must be manually configured to function. Table 5-13 lists
messages that are sent through e-mail notification.

238


Table 5-13 E-mail notification messages
Subject line

Body

AACCT: File nearly full

1400-330: The accounting file is 90% full.

AACCT: File ready

1400-331: The accounting file is ready for
processing.

AACCT: Subsystem out of file

1400-332: The Advance Accounting
subsystem has run out of files for use.

AACCT: Subsystem out of kernel buffers

1400-333: The Advance Accounting
subsystem has run out of kernel buffers.

AACCT: File I/O error

1400-334: The accounting file
encountered an I/O error while writing.

Most e-mail programs provide filtering capability so that shell scripts and
commands can trigger specific actions when messages are received. Use the
string AACCT to identify incoming messages related to Advanced Accounting. To
identify the significance of the message, use the message number in the body of
the message. The readaacct command may be used to extract accounting
records from accounting data files. It is a sample command, so you should be
aware that it is not formally supported by IBM and is used here for example only.

5.18.8 Accounting records
Advanced Accounting produces several accounting records, which are defined in
the /usr/include/sys/aacct.h file. The following paragraphs describe these
records.

Pad record (type 0)
This record does not provide any meaningful accounting data. Report and
analysis tools should skip this record. It is generated for alignment purposes only.

Process record (type 1)
This record is written when a process exits, when a process is reclassified
(setuid(), chproj(), exec()), and when the system is reclassified. This record is
written by the process interval.
This record contains the following information:
User ID
Group ID
Process ID
Process flags (exited, core, killed by signal, killed by checkpoint)


239

Base command name
WLM class
Controlling terminal
Process start time (in seconds from the EPOCH)
Process elapsed time in micro seconds
Combined thread elapsed time in micro seconds
Process (combined threads) processor time in microseconds
Elapsed page seconds of disk pages
Elapsed page seconds of real pages
Elapsed page seconds of virtual memory
Local logical file I/O (JFS, J2) in bytes
Other logical file I/O (NFS, DFS™) in bytes
Local socket I/O (UNIX domain and loopback) in bytes
Remote socket I/O in bytes
The process start time and Process ID can be used to correlate interval records
for a particular process. The exit flag can be used to distinguish between interval
and exit records.

Aggregation process record (type 2)
This record is derived from the process record. A different record is produced for
each user by project. This record is produced by the process interval and
contains the following information:
User ID
Time of first record aggregated (in seconds from the EPOCH)
Number of processes aggregated
Aggregate process elapsed time in micro seconds
Aggregate thread elapsed time in micro seconds
Aggregate process (combined threads) processor time in micro seconds
Aggregate elapsed page seconds of disk pages
Aggregate elapsed page seconds of real pages
Aggregate elapsed page seconds of virtual memory
Aggregate local logical file I/O (JFS, J2) in bytes
Aggregate other logical file I/O (NFS, DFS) in bytes

240


Aggregate local socket I/O (UNIX domain and loopback) in bytes
Aggregate remote socket I/O in bytes

Aggregation application record (type 3)
This record is derived from the process record. Records are produced at the
user, project, and application level. This record is similar to the aggregated
process record, except that the application is named. This record is produced
when the process is classified with an application-specific rule, which is
supported only through the Admin policy. This record is produced by the process
interval and contains the following information:
User ID
Inode of the command that generated the project classification
Device number of the command that generated the project classification
Number of applications aggregated
Aggregate process elapsed time in micro seconds
Aggregate thread elapsed time in micro seconds
Aggregate process (combined threads) processor time in micro seconds
Aggregate elapsed page seconds of disk pages
Aggregate elapsed page seconds of real pages
Aggregate elapsed page seconds of virtual memory
Aggregate local logical file I/O (JFS, J2) in bytes
Aggregate other logical file I/O (NFS, DFS) in bytes
Aggregate local socket I/O (UNIX-domain and loopback) in bytes

Processor and memory use record (type 4)
This record provides information about the use of processors and memory at the
system level. It is generated before and after Dynamic Logical Partitioning
operations and when the size of the large page pool changes. This record is also
generated by the system interval and contains the following information:
Reason the record was generated
Number of online logical processors
Entitled physical processor capacity of the partition
Total idle time, in milliseconds
Total I/O wait time, in milliseconds
Total kernel process time, in milliseconds


241

Total user process time, in milliseconds
Total interrupt time, in milliseconds
Size of physical memory, in megabytes
Size of the large page pool, in megabytes
Large pages in use, in megabytes
Number of page ins from paging space
Number of page outs to paging space
Number of start I/Os v Number of page steals

Policy record (type 5)
This record is written when a policy file is loaded or unloaded. It is provided for
informational purposes only. This record contains the following information:
Type of policy: Admin, User, or Group
Load or unload

File system activity record (type 6)
This record describes the use of file systems at the system level. A separate
record is generated for each mounted file system. This record is produced by the
system interval and has the following information:
Total bytes transferred through read and write
Total number of read and write requests
Total number opens
Total number of creates
Total number of locks
File system type
Device name
Mount point

Network interface I/O record (type 7)
This record provides information about the use of network interfaces at the
system level. This record is produced by the system interval and contains the
following information:
Logical name of the network interface
Number of I/Os
Total bytes transferred

242


Disk I/O record (type 8)
This record provides information about the use of disks at the system level. A
separate record is written for each logical disk device. This record is produced by
the system interval and contains the following information:
Logical name of the disk
Total disk transfers
Total reads
Total writes
Block size of the disk transfer

Lost data record (type 9)
This record provides information about accounting records that were deleted
because Advanced Accounting did not have the ability to record them. This
occurs when all of the accounting data files are full. When the ability to write new
accounting records is restored, Advanced Accounting produces the lost data
record describing the outage. This record contains the following information:
Number of lost records
Number of microseconds of lost processor time associated with process
records
The time that data loss began (in microseconds from EPOCH)
Number of microseconds of lost processor time associated with third party
kernel extension records

Server I/O record (type 10)
This record is produced in hosting partitions. A separate record is produced for
each logical device that is shared with a client partition. The system interval may
be used to periodically produce this record. This record contains the following
information:
Client partition number
Server unit ID
Device logical unit ID (LUN)
Number of bytes in
Number of bytes out

Client VIO record (type 11)
This record is produced in client partitions. It describes the use of virtual devices
in client partitions. A separate record is recorded for each instance of a virtual


243

device. The system interval may be used to periodically produce this record. This
record contains the following information:
Server partition number
Server unit ID
Device logical unit ID
Number of bytes in
Number of bytes out

Third-party kernel extension common aggregation record (type 12)
This record provides accounting information for the named accounting record. It
is derived from aggregated accounting records that are produced by third-party
kernel extensions. This record is written to the Advanced Accounting subsystem
by the system interval. This record contains the following information:
Command name of the kernel extension (from u-block)
Third-party kernel extension transaction ID, in the range of 129 to 256
Number of accounting records that have been aggregated
Resource use, or accumulated processor time, for this class of transactions

ARM application environment record (type 13)
This record describes an application environment instance. It is created from
data that is passed to the operating system through the
arm_register_application() system call and the arm_start_application() system
call. This record is variable in length. All offsets are calculated relative to the start
of the record. This record contains the following information:
Character set in which the data in this record is recorded
Application environment identifier
Offset to application name
Offset to application group
Offset to application identity properties
Offset to application context properties
The operating system attempts to record the content of the application
environment in each accounting data file, so that each accounting data file can
be post-processed as a stand-alone item. This is designed to eliminate the
dependency between accounting data files.

244


ARM transaction environment record (type 14)
This record describes a transaction environment instance. It is created from data
that is passed to the operating system through the arm_register_transaction()
system call. This record is variable in length. All offsets are calculated relative to
the start of the record. This record contains the following information:
Character set in which the data in this record in recorded
Transaction environment identifier
Offset to transaction name
Offset to application identity properties
Offset to application context properties (names only)
The operating system attempts to record the content of the transaction
environment in each accounting data file (not guaranteed), so that each
accounting data file can be post-processed as a stand-alone item. This is
designed to eliminate the dependency between accounting data files.

ARM transaction instance record (type 15)
This record describes an ARM transaction instance. It is created from data that is
passed to the operating system through the arm_start_transaction() and the
arm_stop_transaction() system calls. It is variable in length. All offsets are
calculated relative to the start of the record. This record contains the following
information:
Completion status of the transaction
Offset to user identifier (not User ID)
Offset to user name (not uname)
Offset to accounting code
Response time, in milliseconds
Queued time, in milliseconds
Resource use
The application and transaction environment identifiers are defined respectively
in the application and transaction environment records. These records must be
used to associate application names, application groups, transaction names, and
properties with the transaction instance.


245

ARM aggregated transaction instance record (type 16)
This record is produced instead of the ARM transaction instance record (type
15), when aggregation is enabled for ARM transactions. This record contains the
following information:
Completion status of the transaction
Time of first record aggregated (in seconds from EPOCH)
Offset to user identifier (not User ID)
Offset to user name (not uname)
Offset to accounting code
Aggregate response time, in milliseconds
Aggregate queued time, in milliseconds
Aggregate resource use

Project definition record (type 17)
This record provides a list of project definitions. It is written when the project
definition file is loaded. Multiple records may be needed to record all project
definitions. This record is used to provide the full set of project information in
each data file, so that data files may be treated as stand-alone entities. This may
not be required by the billing application, depending on the nature of the billing
application. This feature may be disabled by disabling the project definition
accounting record. This record is variable in length and contains the following
information:
Number of projects
Number of bytes in the project definition area
Project definition area

5.19 Date APIs past 2038
The current time API does not allow the representation of times past Tuesday,
January 19, 2038 at 03:14:07 UTC (Coordinated Universal Time). UNIX systems
represent times as the number of seconds past January 1, 1970 at 00:00:00 UTC
(the epoch). The time_t type is a 32-bit signed integer on both 32- and 64-bit
machines. On January 19, 2038 03:14:08 UTC it will be more than 2^31 seconds
making it impossible for the date to be represented by a time_t type. Applications
wishing to represent dates past 2038 cannot rely on the time API because of this
limitation.

246


AIX 5L Version 5.3 has introduced a new 64-bit time API that will allow user
programs to use dates past 2038. It will use the time64_t type instead of the
time_t type. With the new API the user will be able to call the new time functions
to manipulate times up till December 31, 9999.
The following functions have been added to Version 5.3:
ctime64()
ctime64_r()
localtime64()
localtime64_r()
gmtime64()
gmtime64_r()
asctime64()
asctime64_r()
difftime64()
mktime64()
Each function mirrors its respective time32 counterpart. The only difference is
time_t has been replaced by a new datatype time64_t which is a signed 64-bit
integer.
Note: The new interface will not provide function to set the date past 2038.
AIX 5L Version 5.3 will, most likely, not be supported past 2038.
# date 070812562050
date: 0551-403 Format of date/time specification is not valid.

5.20 System V printer enhancements
AIX 5L Version 5.3 includes the following new or enhanceed functions for System
V printing:
The remote lpd printing daemon lpNet now provides improved speed and
performance.
The same JetDirect software now services both AIX printing and System V
printing. This leads to a more robust connection to the printing devices.
The lpsched and lpNet programs have been made more secure.
Performance of lpsched has been enhanced by improving the filtering
process.
Error reporting from lpsched and lpNet has been improved.


247

5.21 Mozilla browser for AIX
AIX 5L Version 5.3 introduces support for the Mozilla 1.4.2 Web browser as the
default browser for AIX. Netscape Communicator Version 4 is not supported on
AIX 5L Version 5.3.
The Mozilla Web browser for AIX is available on a CD that can be ordered with
AIX, or it can be downloaded from the following Web site:
http://www.ibm.com/servers/aix/browsers

Mozilla for AIX requires GNOME libraries, which are available on the AIX Toolbox
for Linux Applications CD or from the following Web site:
http://www.ibm.com/servers/aix/products/aixos/linux

5.21.1 Installing Mozilla for AIX
Mozilla for AIX can be installed as an option during the AIX Base Operating
System installation process, or it can be installed later. All listed installation
methods use the Mozilla installation bundle, which includes Mozilla and the
required GNOME libraries.
The Mozilla installation process fails if the required GNOME libraries are not
found. The required rpm filesets are listed in the Mozilla install log.
Use one of the following installation methods:
Install Mozilla using the following AIX BOS installation process:
– You can select Mozilla for installation during the AIX Base Operating
System installation process by selecting these options in the following
order:
•

2 = Change/Show Installation Settings and Install

•

3 = More Options

•

6 = Install More Software

•

1 = Mozilla (Mozilla CD)

– The default setting is to not install Mozilla.
– When prompted to do so, insert the Mozilla CD and the AIX Toolbox for
Linux Applications CD.
Install Mozilla as a bundle using the following Configuration Assistant
process:
– Start configassist.
– Select Manage software, and click Next.

248


– Select Install additional software, and click Next.
– Select Install by bundle, and click Next.
– Specify the device or directory that contains the installation images, and
click Next. If the location is a directory, such as /usr/sys/inst.images, verify
the following:
•

The Mozilla.base installp package is in the
/usr/sys/inst.images/installp/ppc directory

•

The toolbox rpm filesets are in the /usr/sys/inst.images/RPMS/ppc
directory

– Select the Mozilla bundle, and click Next.
– Accept the license agreement, and click Next to start the installation
process.
Install Mozilla as a bundle using the following smit or smitty process:
– Run the smitty install_bundle or smit install_bundle command.
– Specify the INPUT device and directory for software. If the location is a
directory, such as /usr/sys/inst.images, verify the following:
•

The Mozilla.base installp package is located in the
/usr/sys/inst.images/installp/ppc directory

•

The toolbox rpm filesets are located in the
/usr/sys/inst.images/RPMS/ppc directory

– Select the Fileset Bundle = Mozilla.
– In the Install Software Bundle screen, accept the license agreement, and
press Enter to start the installation process.

5.21.2 Mozilla as the browser for AIX Documentation Services
Mozilla can be configured as the default browser that is used to view the AIX
Documentation using Configuration Assistant or SMIT.
Configure Mozilla using the following Configuration Assistant process:
– Start config_assist.
– Select the Configure documentation server task.
– If Mozilla is detected as already installed, select Yes, use Mozilla as the
default browser, and click Next.
Configure Mozilla using the following smit or smitty process:
– Run the smit change_documentation_services command or the smitty
change_documentation_services command.


249

– Verify that the /usr/bin/mozilla directory is the DEFAULT_BROWSER.

5.21.3 Migrating Netscape Communicator Version 4 profile
If a Netscape Communicator Version 4 profile exists in your home directory and
Mozilla is run for the first time, Mozilla prompts whether or not it should convert
the Communicator profile, including the bookmarks to be used within Mozilla.
For more information about Mozilla for AIX, see the
file:///usr/mozilla/base/README.HTML file after Mozilla is installed.

5.22 Uniprocessor kernel support
The AIX 5L operating system previously contained both a uniprocessor 32-bit
kernel and a multiprocessor 32-bit kernel. Effective with AIX 5L Version 5.3, the
operating system supports only the multiprocessor kernel.
The AIX 5L Version 5.3 32-bit multiprocessor kernel supports the following
systems: RS/6000, Sserver pSeries, p5, or OEM hardware based on the
Common Hardware Reference Platform (CHRP) architecture, regardless of the
number of processors.
AIX 5L Version 5.2 is the last release of AIX that supports the uniprocessor 32-bit
kernel.

5.23 Enhancement of base commands and libraries
AIX 5L Version 5.3 includes improvements to the following base commands and
libraries:
The find command has been enhanced to provide information on file access
and changes in the last N minutes (instead of days).
The SMIT panels have been updated to include the latest set of flags for the
iostat command.
The following two new operators ("+=" and "%") and a special target
.SCCS_GET have been added to the make command:
+=

Appends a value to an existing variable's value

%

Implements pattern matching for substitution

.SCCS_GET target source

250


Specifies the commands for retrieving source
files from SCCS

The ps command has been enhanced to provide process hierarchy
information and a listing of descendant processes for given pid(s).
A new flag has been added to the tar command which will specify the list of
files and/or directories to be excluded from the tar file being created,
extracted, or listed.
Flags have been added to the tar command to process directory of files
recursively. An option has also been added to specify an input file for tar
extraction much like that which can be used for tar command file creation.
The fuser command has been enhanced to accept any of the signals
displayed by the kill -l command. Only the root user can kill a process of
another user.
AIX 5L text processing commands now support unlimited line lengths except
for ed/ex/vi/awk. For ed/ex/vi the line length limit is 8192 characters and for
awk it is 100 K characters.
Command buffer sizes are unlimited now. This allows an increased number of
arguments on the command line and reduces the number of failures of the
following type:
grep foo /usr/include/*/*
ksh: /usr/bin/grep: arg list too long

An option has been added to the restore command to exit on error rather
than recover and continue. An option has also been added to provide a long
style listing of backup files as in ls -l.
Options have been added to the grep command to do recursive search.
The following two formats have been added to the date command:
– %k representing the 24-hour-clock hour clock as a right-justified
space-filled number (0 to 23).
– %s representing the number of seconds since January 1, 1970,
Coordinated Universal Time (CUT).
The snap command now includes installed package information from
System V and rpm.
If Manual Pages with the same title are resident under directories in
MANPATH and /usr/share/man/info, previously it displayed all matching
entries. Now a flag has been added to display only the first matching entry.
When a new environment variable EXTENDED_HISTORY is set the shell
histories in ksh and ksh93 will have time stamp.
AIX 5L Version 5.3 now includes a restricted version of ksh and ksh93.
A backtag function has been added to the vi command. What this means is
that if you use ctl-] to tag forward to a specific line, then using ctl-T returns you


251

to the previous position from where you tagged forward. It maintains a stack
of tags. Anytime that you do ctl-] it pushes the previous position onto the
stack, and anytime you do ctl-T, it pops the previous position off of the stack
and changes the current editing position to be that position.
The cron function now accepts user-specified location, type, and size for the
cron log file when specified in the /etc/cronlog.conf configuration file.
Queues g-z are now available for user-defined queues for at jobs.
A flag has been added to the at command so that you can see both the time
of the job and the actual command that is scheduled to run.
An option has been added to nohup command to nohup an existing process.

252


6

Chapter 6.

Performance monitoring
This chapter describes the updated and new performance monitoring tools
included in AIX 5L version 5.3. The following enhancements are included:
The procmon command - process monitoring tool
Asynchronous I/O statistics
Disk service and wait time monitoring
PMAPI M:N pthreads support
Support for Micro-Partitioning and SMT
– perfstat library
– vmstat command enhancements
– iostat command enhancements
– mpstat command - CPU performance monitoring
– lparstat command
– sar command enhancements
– topas command enhancements
Per file system I/O pacing


253

6.1 General information
The stat commands (iostat, vmstat, sar, and the new mpstat and lparstat
commands) now can detect and tolerate system configuration changes. This new
function requires output changes. All five commands now display a system
configuration line, which appears as the first line displayed after the command is
invoked. Scripts that parse command output may need to be modified to detect
and handle this new output. If a configuration change is detected during a
command execution iteration, a warning line will be displayed before the data
which is then followed by a new configuration line and the header.
There are many new tunable parameters added to AIX in AIX 5L Version 5.3.
Some have also appeared in AIX 5L Version 5.2 ML5 which was made available
at the same time. It is always a good policy to check for new options when major
system service is applied as these enhancements do not always appear in
product documentation. Use the -L option with the ioo, vmo, schedo, no, and nfso
commands to get the current list of tunable parameters. Use the -h option to
display their descriptions and usage.

6.2 The procmon command - process monitoring tool
procmon is an Eclipse-based tool that graphically displays performance
characteristics of processes in the system. There are two different uses of this
tool:
It is a monitoring tool that displays a dynamic, sorted list of processes, and
information about them.
It provides the ability to carry out actions on the processes listed. These
actions are basic administration commands such as kill, renice for
example, as well as other advanced tools such as svmon and proctools.
The output displayed (Figure 6-1 on page 255) has the following flexibility:
Columns can be configured
– Select sorting column
– Select ascending/descending order
– Change order of columns
– Add or delete columns
Table can be searched, for example to highlight:
– All processes with a common ppid
– All processes belonging to a given user

254


Highlights are maintained after refresh
Actions available on processes include:
– kill
– renice
– Show list of threads (also in a sorted table)
– Run performance tools (svmon and proctools)
Shows text results in a new window.

Figure 6-1 procmon - new process monitoring tool

Figure 6-2 on page 256 displays a sample of global metrics for partition
performance.

Chapter 6. Performance monitoring

255

Figure 6-2 Global metrics for partition performance

6.3 Asynchronous I/O statistics
AIX 5L Version 5.3 introduces enhancements to the iostat command that allow
the user to obtain AIO statistics. To give you a better understanding of the new
features, we begin this discussion with an overview of the already existing AIO
device driver logic, followed by an explanation of the new enhancements.

6.3.1 Asynchronous I/O basics
The Asynchronous I/O (AIO) as well as its POSIX version are already available
in the previous versions of AIX. AIO is an environment that speeds up the
application access to the files by using separate kernel processes (aioserver) to
do the real I/O transfer. There are two AIO drivers available in the system:
Legacy AIO
POSIX AIO

256


Both of the drivers are similar in function. The implementation of the POSIX AIO
complies with the POSIX standards and also implements its extensions. Since
the internal logic of both drivers is the same, we are describing the AIO logic
using the legacy AIO driver and point out the differences for the POSIX AIO
driver.
Figure 6-3 shows how the AIO flow goes from the application to the logical
volume.

aio_write(fs) aio_write(fs) aio_write(fs) aio_write(raw_lv)
Q[0]

Q[1]

Q[2]

Q[3]

Q[255]

...

aioserver[0]

FS1

aioserver[0]

FS2

aioserver[0]

fastpath

LV3

Figure 6-3 AIO flow

The user program calls the aio_write(), aio_read(), or the lio_listio() system call to
write or read a data buffer to or from a file. The file can be a regular file or also a
raw device.
Today, there is a fixed number of preallocated AIO request queues in the
system, but this number may change in the future. The AIO device driver
allocates one of those queues and puts the data from the buffer into the allocated
queue. There is an internal logic of assigning the queues to specific AIO
requests. Usually writes to the same file system go into the same queue, while
writes to different file systems are preferred to go to different queues. However,
there may be situations when two or more different file system requests go into
the same queue or requests to the same file system are spread to several
queues.


257

The AIO requests are then picked up by aioserver kernel processes, which
transfer each AIO request to the output file. The POSIX AIO kernel processes
are named posix_aioserver. You can check them by using, for example, the
ps aux command.
If the destination file is a raw device (raw logical volume), the AIO device driver
will use the fastpath routine instead of the AIO queues. The fastpath routine
moves the input buffer of the AIO system call directly to the raw device, thus
bypassing the AIO and its queues.
In order to use the AIO drivers we have to enable the AIO device driver using the
mkdev -l aio0 command or through smit aio. In order to enable the POSIX AIO
device drivers, you have to use the mkdev -l posix_aio0 or smit posixaio
commands.
Using SMIT and going through menus Devices → Asynchronous I/O →
Asynchronous I/O (Legacy) → Change/Show Characteristics of
Asynchronous I/O you can set the characteristics of the AIO, like minservers,
maxservers, maxreqs, kprocprio, autoconfig, or fastpath.

6.3.2 AIO enhancement of the iostat command in Version 5.3
In the previous versions of AIX there were no tools available to monitor the AIO;
from Version 5.3 the iostat command is enhanced to monitor the AIO.
The iostat command reports CPU and I/O statistics for the system, adapters,
TTY devices, disks, and CD-ROMs. This command is enhanced by new
monitoring features and flags for getting the AIO and the POSIX AIO statistics.
The following new flags are added to the iostat command:
-A

Reports legacy AIO statistics along with utilization metrics.

-P

Reports POSIX AIO statistics along with utilization metrics.

-q

Reports each AIO queue’s request count.

-Q

Reports AIO queues associated with each mounted file system and the
queue request count.

-l

Displays the data in a 132 column width. This flag is simply a formatting
flag.

The iostat -A option
When using the -A option, the output of the iostat command gives additional
statistics of the AIO. The following information is added:
avgc

258

Average global non-fastpath AIO request count per second for the
specified interval.


avfc

Average AIO fastpath request count per second for the specified interval.

maxg Maximum non-fastpath AIO request count since the last time this value
was fetched.
maxf

Maximum fastpath request count since the last time this value was
fetched.

maxr

Maximum AIO requests allowed. This is the AIO device maxreqs
attribute.

When the AIO device driver is not configured in the kernel, the iostat -A
command gives an error message that the AIO is not loaded, such as in the
following example:
# iostat -Aq
aio: avgc avfc maxg maxf maxr avg-cpu: %user %sys %idle %iow physc %entc
iostat: 0551-157 Asynchronous I/O not configured on the system.

Example 6-1 demonstrates the use of the iostat -A command.
Example 6-1 iostat command AIO statistics in a Micro-Partitioning environment
# iostat -A 1 1
System configuration: lcpu=4 drives=1 ent=0.50
25
6 29
10 4096
30.7 36.3 15.1 17.9 0.0 81.9
Disks:
hdisk0

% tm_act
100.0

Kbps
61572.0

tps
Kb_read Kb_wrtn
484.0
8192
53380

Notice the avgc column, which it shows the average number of AIO requests in
the queues; and the maxg column, which shows the maximum number of AIO
requests in the queues for the last measuring period. If the avgc or maxg is
getting close to the maxr then tuning of the maxreqs and maxservers attributes is
required.
If you use raw devices, the avfc is interesting because it reflects the use of the
average AIO fastpath calls, as is the maxf, which shows the maximum value of
fastpath count value. Since the fastpath calls bypass the AIO queues, these
statistics give only information on how fastpath AIO is used.

The iostat -Aq option
If you are interested in the allocation of the AIO queues and their use, then the
iostat -Aq command is useful. Example 6-2 on page 260 shows output from this
command.


259

Example 6-2 Output of the iostat -Aq command with AIO queue statistics
# iostat -Aq 1 1
25
10 29
10 4096
29.5 37.9 23.0 9.5 0.0 83.6
q[ 0]=0
q[ 5]=0
...
q[120]=0
q[125]=0
q[130]=0
q[135]=0
q[140]=0
...
q[250]=0
q[255]=0

q[
q[

1]=0
6]=0

q[ 2]=0
q[ 7]=0

q[
q[

3]=0
8]=0

q[ 4]=0
q[ 9]=0

q[121]=0
q[126]=0
q[131]=0
q[136]=0
q[141]=0

q[122]=0
q[127]=0
q[132]=0
q[137]=0
q[142]=0

q[123]=0
q[128]=0
q[133]=0
q[138]=0
q[143]=0

q[124]=0
q[129]=0
q[134]=0
q[139]=25
q[144]=0

q[251]=0
q[256]=0

q[252]=0
q[257]=0

q[253]=0
q[258]=0

q[254]=0

If statistics for any single queue are significantly higher than the others, this
indicates that applications are using one file system significantly more than other
file systems. The queues are usually allocated one per file system.

The iostat -AQ option
If a specific AIO queue is filling up unusually and you want to know to which file
system the queue is related, then the iostat -AQ command is useful. Figure 6-3
shows the distribution of the AIO queues to specific file systems, as shown in
Example 6-3.
Example 6-3 The iostat -AQ command output
sq 156 # /ad/iostat.fix -AQ 1 1
26
7 29
10 4096
27.1 36.4 19.6 16.8 0.0 78.1
Queue#
129
130
132
133
136
137
138
139

260

Count
0
0
0
0
0
0
0
26

Filesystems
/
/usr
/var
/tmp
/home
/proc
/opt
/ad


The iostat -P option and related options for POSIX AIO
In order to get the basic POSIX AIO statistics we have to use the iostat -P
command. The output is in the same format as that for the iostat -A legacy AIO
command; the meaning of the metrics is the same as well, but related to the
POSIX AIO calls.
The output format of the iostat -Pq command corresponds with the iostat -Aq
and the iostat -PQ corresponds with the iostat -AQ command.

Other iostat -A usage
There may be several possible combination of the iostat flags, but it is better to
run the iostat command twice in two different windows then use too many flags
and get too much information. There are also several unsupported combinations
of flags. One example of a reasonable use would be the iostat -A -l hdisk3
hdisk4 1 10 command that has combined flags.

6.4 Disk service and wait time monitoring
The following sections discuss the enhancements made in AIX 5L Version 5.3 in
the area of disk service and wait time monitoring.

6.4.1 dkstat and device support
The dkstat structure is the base statistical interface defined in the <sys/iostat.h>
header file that is used by the device drivers to provide the statistical data and
the performance monitoring tools, such as iostat -D or sar -d, to read the data.
The dkstat structure is updated in AIX 5L Version 5.3 to support additional
statistics relative to disk service times. The dkstat structure is referenced in the
/usr/include/sys/iostat.h header file.

6.4.2 avwait and avserv enhancements for the sar -d
Starting with Version 5.3, the -d flag is modified in the sar command. The -d flag
of the sar command is intended to give statistics about the I/O activity on the
disks. The sar -d command reads the dkstat structures of the devices that
supply the necessary information.
In older versions of AIX, avwait and avserv are always reported as zero. From
Version 5.3 the avwait and avserv are implemented. The meaning of the avque
statistics has changed from Version 5.3. See Example 6-4 on page 262 for
sample output of the sar -d command.


261

Example 6-4 sar -d command output
# sar -d 1 2
AIX sq1test1 3 5 00CDDEDC4C00
06/22/04
10:01:37

device

%busy

avque

r+w/s

Kbs/s

avwait

avserv

10:01:38

hdisk0

100

36.1

363

46153

51.1

8.3

10:01:39

hdisk0

99

38.1

350

44105

58.0

8.5

Average

hdisk0

99

37.1

356

45129

54.6

8.4

There are two new measured statistics and one that has been modified:
avwait

Average wait time per request in milliseconds. This is the time that the
request is waiting in the device driver I/O queue to be passed to the I/O
device.

avserv

Average service time per request in milliseconds. This is the length of
time between when the I/O is passed to the device to when the result is
returned back from the device.

avque

Average number of requests waiting to be sent to disk. Imagine this as
the average length of the occupied I/O queue.

There is a change in the avque interpretation introduced in Version 5.3. The
avque statistics used to represent the instantaneous number of requests sent to
disk but not completed yet in the previous versions of AIX.
The other columns, like Kbs/s, r+w/s, %busy, device and so on have the same
meaning as before.

6.5 PMAPI M:N pthreads support
Under the M:N threading model, M user threads are mapped to N kernel threads,
with M being typically considerably bigger than N to allow large numbers of
pthreads to run. Making PMAPI calls from a program running in this mode was
previously not supported.
The PMAPI library has been updated to handle the M:N thread model; the
current unchanged interfaces simply work in M:N mode. The only significant
change is for third party API callers, for example debuggers, where new
interfaces with pid, tid, and ptid must be used.

262


6.6 Support for Micro-Partitioning and SMT
The introduction of Micro-Partitioning and Simultaneous Multi Threading (SMT)
required additional methods for recording utilization to be added to hardware and
used by AIX.
The Process Utilization Resource Register (PURR) is a new register, provided by
the POWER5 processor, which is used to provide an actual count of physical
processing time units that a logical processor has used. All performance tools
and APIs utilize this PURR value to report CPU utilization metrics for
Micro-Partitioning and SMT systems.

6.6.1 perfstat library
The performance library API libperfstat is enhanced to capture the new
PURR-based utilization statistics.
The following enhancements are introduced by AIX 5L Version 5.3 to the
libperfstat library:
perfstat_cpu_t

The structure defines the basic processor statistics on a
per processor basis. The structure is extended with
PURR-based utilization and Micro-Partitioning statistics. It
has also been extended with statistics displayed by the
mpstat command, like affinity and interrupt statistics.

perfstat_cpu_total_t The structure defines the basic processor statistics
system-wide. The structure is extended with PURR-based
utilization and Micro-Partitioning statistics. It has also
been extended with more detailed interrupt statistics.
perfstat_partition_total_t
This is a new structure that contains partition-specific
information and statistics similar to what the lparstat
command can display.
perfstat_partition_total()
This is a new function that returns information in a
perfstat_partition_total_t structure.

6.6.2 vmstat command enhancements
The vmstat command has been enhanced to support Micro-Partitioning and can
now detect and tolerate dynamic configuration changes.
The vmstat command has two new metrics that are displayed: physical
processor consumed and percentage of entitlement consumed. They are


263

represented as pc and ec in the output format. The physical processor consumed
represents the number of physical processors consumed by the partition during
an interval. The percentage of entitlement consumed is the percentage of
entitled capacity consumed by a partition during an interval (pc/ent)*100. These
new metrics will be displayed only when the partition is running as a shared
processor partition. If the partition is running as a dedicated processor partition
the new metrics will not be displayed.
Changed from previous releases, the first interval of the command output is now
meaningful and does not represent statistics collected from system boot. Internal
to the command the first interval is never displayed, and therefore there may be a
slightly longer wait for the first displayed interval to appear. Scripts that discard
the first interval should function as before.
With these new changes, the output of the vmstat command will be as shown in
the following examples.
On a dedicated partition:
# vmstat
kthr
memory
page
faults
cpu
---- ------------- ----------------------- ------------ -----------------------r b avm
fre
re pi po fr sr cy in
sy
cs us sy id wa
1 1 19782
32762
0
0
0
0
0
0 125
10
47 50 50 0
0

On a Micro-Partitioning machine:
# vmstat
kthr
memory
---- ------------r b avm
fre
1 1 19782
32762

page
----------------------re pi po fr sr cy in
0
0
0
0
0
0 125

faults
cpu
------------ ---------------------------------sy
cs
us sy id wa
pc ec
10
47
50 50
0
0
1.5 85

6.6.3 iostat command enhancements
Beginning with AIX 5L Version 5.3, the iostat command reports the number of
physical processors consumed (physc), the percentage of entitled capacity
consumed (% entc), and the processing capacity entitlement when running in a
shared processor partition. These metrics will only be displayed on shared
processor partitions.
Changed from previous releases, the first interval of the command output is now
meaningful and does not represent statistics collected from system boot. Internal
to the command, the first interval is never displayed, and therefore there may be
a slightly longer wait for the first displayed interval to appear. Scripts that discard
the first interval should function as before.

264


The following iostat command output shows an example. The newly introduced
information is in bold type.
# iostat
tty:

Disks:
hdisk1
hdisk0
#

tin
0.3

tout
avg-cpu:
98.2
% tm_act
0.2
0.0

Kbps
5.9
0.0

% user
8.9

tps
1.5
0.0

% sys
0.2

Kb_read
128
0

% idle
90.8

% iowait physc
0.2 0.03

% entc
10.2

Kb_wrtn
14360
0

Percentage of entitled capacity consumed is display in the %entc column. The
number of physical processors consumed is shown in the physc column.
In the system configuration information you can see the currently assigned
processing capacity specified as ent.

6.6.4 mpstat command - CPU performance monitoring
The mpstat command is the basic monitoring tool for logical CPU usage. It
accepts the following flags:
no flag

Basic statistics

-d

Dispatcher statistics

-i

Interrupt statistics

-s

SMT statistics

-a

Display all statistics (wide output)

-w

Format output to wide columns. The -a flag implicitly turns on this
flag.

In the following sections we describe the use of the main mpstat command flags
to obtain basic (no flag), dispatcher (-d), interrupt (-i) and SMT (-s) statistics. The
-a and -w flags are focussed on the formatting of the output; they combine the
statistics received using the described flags.

Enhanced basic statistics
The calculation of the performance statistics has gone through significant
changes because of the Micro-Partitioning and SMT technologies. In previous
versions of AIX the percentage of the processor usage calculation was based on
calculating the average usage. The calculation was simple:
avg_usage = (cpu1_usage + cpu2_usage + cpun_usage) / n


265

In the case of Micro-Partitioning architecture we have to consider that the
system-wide utilization is reported as a percentage of the allocated entitled
capacity and that there could be an unused slice of each entitled processor
capacity. This unused part needs to be integrated into the calculations.
Figure 6-4 shows the basics for statistics calculations.
Note: The enhanced statistics also apply to the sar -P command.

lcpu=2, ent = 1.50
0.10

0.50

1.00

cpu 0
pc = 0.20

cpu 1
pc = 0.40

CPU busy
Figure 6-4 CPU statistics for POWER5 systems

lcpu, ent, pc, %ec
In Figure 6-4 we have one physical processor that maps to two logical
processors (cpu0, cpu1), using SMT. The number of logical processors is marked
by lcpu.
The processor entitlement is the CPU capacity we are entitled to use. We are
using the ent metric for the total entitlement. In Figure 6-4 we have a 1.50
processor entitlement.
The entitlement does not need to be consumed entirely. If there are idle
processor cycles, they can be used by other partitions. The processors can use
different fractions of the entitlement. The fraction of the processor that is
consumed by the partition using Micro-Partitioning technology is marked by pc.
In Figure 6-4 the fraction of processor consumed is pc(0)=0.20 for processor 0

266


and pc(1)=0.40 for processor 1. The total fraction of processor consumed is
pc(all)=0.60. There is also an unused fraction of the processors
pc(u)=ent - pc(all)=0.90. The processing capacity consumed is the sum of the pc:
pc(0)+pc(1)=pc(all), and pc(all)+pc(u)=ent.
The %ec is the percentage form of the usage of the entitlement. The cpu0 is
using %ec(0)=13.3%=100*pc(0)/ent and cpu1 is using
%ec(1)=26.7%=100*pc(1)/ent. The unused part of the entitlement is
%ec(u)=60%=100*pc(u)/ent. The total percentage of entitlement consumed, or
%ec is the sum %ec(0)+%ec(1)=ec(all), and ec(all)+%ec(u)=100%

us, sy, wa, id
The calculation of the percentage of the CPU time consumed by user (us), kernel
(sy), waiting for I/O operation to complete (wa) or idle (id) is calculated relative to
the consumed entitlement.
In Figure 6-4 on page 266 we assume that the bar shows the CPU use by the
user space (us) and the rest is idle (id). In order to simplify the example, we
assume that the CPU use by kernel (sy) or waiting for I/O (wa) is zero. The cpu0
is running 0.10 processor fraction in user space that is 50% of the used
processor fraction by this processor. The remaining part is 0.10 processor
fraction, that is 50% of the processor fraction used by this processor. The cpu1 is
using 0.30 of 0.40 processor fraction that is 75%.
The per CPU sy, wa, and id statistics are calculated in the same way as the us
statistics, just using the kernel space, wait I/O, or idle values.
Note: While the per CPU us, sy, wa, and id statistics are calculated as
percentages of the fraction of the physical CPU consumed by a particular
logical CPU, the all CPU us, sy, wa, and id statistics are calculated as
percentages of the allocated entitlement.
The wa and id statistics for the U line are related to the allocated entitlement.
The us and sy statistics are not applicable for the line representing unused
entitlement.
The statistics for all CPUs used by the user space is the processor fraction used
for user space of all processors, that is 0.40. This fraction is than related to the
total entitlement in percents that is the value of the us: us=26.7%=100*0.40/1.50.
The statistics for all CPUs are calculated as the sum of the idle processor
fractions plus the unused entitlement related to the total entitlement. The cpu0
spent 0.1 processor fraction and cpu1 spent also 0.1 processor fraction within
the used entitlement, but running the wait kernel process. Both processors were
using 0.2 processor fractions in idle state from the kernel point of view. But the


267

total time the processors were idle is more when you add in the unused
entitlement of 0.90. So the total idle time was 0.10+0.10+0.90=1.10. Calculating
a percentage related to the total entitlement of 1.50 results in
id(all)=73.3%=100*1.10/1.50.
The idle statistics for the U line (representing unused entitlement) is calculated
as the percentage of the unused entitlement which was unused because the
processor was idle. In our case all the unused processor entitlement was
because the processor was idle and so id(u)=100.0%.
If there were some unused entitlement because of I/O wait, then the wa(u) would
be non zero.
Table 6-1 shows a sample calculation of the statistics with easy to calculate
numbers. In the pc and %ec columns, the sum of the CPUs is reflected in the
ALL row.
Table 6-1 Sample CPU usage calculation
CPU

us

sy

wa

id

pc

%ec

0

50.0

0.0

0.0

50.0

0.20

13.3

1

75.0

0.0

0.0

25.0

0.40

26.7

U

n/a

n/a

0.0

100.0

0.90

60.0

ALL

26,7

0.0

0.0

73.3

0.60

40.0

Example 6-5 shows how the processor usage and the entitlement goes together
on a real system. The example was created by running the mpstat command on
a slightly loaded system.
Example 6-5 mpstat basic statistics
# mpstat 1 2
cpu min maj mpc int cs ics rq mig lpa sysc us sy wa id
pc %ec lcs
0
0
0
0
31 148
74
1
1 100 204 68 22 0 10 0.01 3.5 197
1
0
0
0 177
0
0
0
0 0 0 11 0 89 0.00 1.2 150
U
- - - - 0 95 0.29 95.3
ALL
0
0
0 208 148
74
1
1 100 204 2 1 0 97 0.01 4.7 173
------------------------------------------------------------------------------0
0
0
0
32 151
75
1
0 100 173 66 24 0 10 0.01 3.2 194
1
0
0
0 172
0
0
0
0 0 0 12 0 88 0.00 1.1 144
U
- - - - 0 96 0.29 95.7
ALL
0
0
0 204 151
75
1
0 100 173 2 1 0 97 0.01 4.3 169

268


This example shows the basic statistics obtained from the mpstat command. The
first line tells us the number of logical processors available to the operating
system and the total processor entitlement assigned to this partition using
Micro-Partitioning technology.
The cpu column represent the ID of the logical processor. The U line stands for
the unused entitlement. The line U shows up only for POWER5 systems and only
if part of the entitled capacity is not used by this partition. The last line “ALL”
stands for all processors’ statistics.
Note: Many statistics for the U line statistics are not applicable. For example,
an unused time slice cannot generate a context switch.
The us, sy, wa, id columns stand for the processor usage statistics in user mode,
kernel mode, waiting for I/O completion or idle. Their meaning and calculation is
explained earlier in this section.
The pc stands for the fraction of processor consumed. Its meaning and
calculation is explained earlier in this section.
The %ec stands for the percentage of the consumed entitled capacity. Its
meaning and calculation is explained earlier in this section.
The min stands for the minor page faults. Minor page faults do not involve disk
I/O activity.
The maj stands for the major page faults. Major page faults involve disk I/O
activity.
The mpc stands for Multi Processor Communication interrupts and is the sum of
the mpcs and mpcr as described in “Interrupt statistics” on page 270.
The int stands for the sum of all interrupts as described in “Interrupt statistics” on
page 270.
The cs stands for the context switches and the ics stands for involuntary context
switches. The cs is the sum of voluntary context switches and involuntary context
switches. The context switches are described in “Affinity and migration statistics”
on page 271.
The rq stands for the run queue that is maintained per processor. The value for
all CPUs is the sum of each processor’s run queue. This is also described in
“Affinity and migration statistics” on page 271.
The mig stands for the total number of thread migrations to another processor.
This is the sum of the migrations in affinity domains S1 to S5, that is,


269

mig=S1rd+S2rd+S3rd+S4rd+S5rd. Affinity domains are described in “Affinity and
migration statistics” on page 271.
The lpa stands for the logical processor affinity. It is the percentage of logical
processor redispatches within the scheduling affinity domain 3. Affinity domains
are described in section “Affinity and migration statistics” on page 271.
The sysc stands for the number of system calls per each processor and the sum
for all processors.
The lcs stands for the number of logical processor (hardware) context switches.
This is the sum of involuntary and voluntary context switches as described in
“Affinity and migration statistics” on page 271.
The min, maj, mpc, int, cs, rq, mig, lpa, sysc, and lcs statistics are per second
within the interval.

Interrupt statistics
The interrupt-oriented statistics are obtained by the mpstat -i command.
Example 6-6 mpstat -i output
# mpstat -i 1 1
cpu
0
1
2
3
ALL

mpcs
0
0
0
0
0

mpcr
0
0
0
0
0

dev
2
1
2
3
8

soft
2160
1
2342
74
4577

dec
25
100
114
10
249

ph
14
126
8
6
154

The lcpu, ent, and cpu columns are the same as in all mpstat command outputs.
The mpcs, mpcr, dev, soft, dec, and ph columns are the interrupt statistics per
second during the sample interval if specified, or the total since boot if the
interval is not specified.
The mpcs and mpcr columns show the multiprocessor communication send and
receive interrupts. The MPC interrupts are forwarded among processors in
similar manner as for the interprocess communications (IPCS), but on the
processor level. For example, whenever a TLB line is invalidated, the processor
that invalidates the line sends MPC interrupts to other processors using that line.
Their sum is displayed in the mpc column of the basic statistics.

270


The dev column stands for the interrupts initiated by a device. Devices generate
these interrupts, for example, when the buffer is ready for the kernel.
The soft column stands for the software interrupts. These interrupts are
generated whenever a device driver needs more time to finish processing a
device interrupt.
The dec column stands for the number of decremeter interrupts. These interrupts
are generated from timer counters.
The ph column stands for the number of phantom interrupts. The best way to
describe these interrupts is with an example. On a system running two partitions,
SPLPAR1 and SPLPAR2, a device originates an interrupt to processor belonging
to SPLPAR1. In the time when the interrupt is initiated, the SPLPAR1 is outside
its entitled capacity and SPLPAR2 is running on the processor. SPLPAR2 gets
the interrupt, detects that it is not for it, counts it as a phantom interrupt, and calls
the hypervisor to forward it to SPLPAR1. When the SPLPAR1 gets the processor
again, the interrupt is forwarded to the SPLPAR1. Phantom interrupts show that
other partitions are doing I/O operations.

Affinity and migration statistics
The operating system kernel dispatches the processes with an attempt to
optimize performance. The process performance is dependent on which
processor is dispatched to. If the process is dispatched to the same processor as
it was dispatched to in the previous time slice, the performance is much better
than if it were dispatched to a different MCM module due to cache and
addressing requirements. There is a need to measure the dispatching of the
processes and evaluate the efficiency of the dispatches.


271

S0
v-cpu0

v-cpu2

v-cpu1

v-cpu3

proc0

S1

proc2

chip3 chip2
chip0 chip1

L2 cache

chip0

MCM0

S2

S3

S5

IB M
H C R
U
s e r v e r6
p S e r ie s

IB M

MCM0 MCM1

HCR
U6
server
IB M
H C R
U
s e r v e r6

pSeries

p S e r ie s

S4
Figure 6-5 Affinity domains and process dispatches

Figure 6-5 shows the logic of the affinity domains. The IBM Sserver p5 systems
with SMT turned on may feature the following affinity domains:
S0
S1

The process redispatch occurs within the same physical processor,
among different logical processors. This can happen in the case of
SMT-enabled systems. This involves sharing of the L1 cache.

S2

The process redispatch occurs within the same processor chip, but
among different physical processors. This involves sharing of the L2
cache.

S3

The process redispatch occurs within the same MCM module, but among
different processor chips. This involves sharing of the L3 cache.

S4

The process redispatch occurs within the same CEC plane, but among
different MCM modules. This involves access to the main memory or
L3-to-L3 transfer.

S5

272

The process redispatch occurs within the same logical processor.

The process redispatch occurs outside of the CEC plane. This domain is
not currently used.


The mpstat -d command gives detailed dispatcher statistics. The statistics focus
on process dispatches and the affinity of them related to each logical processor
as well as averaged to global view.

Figure 6-6 mpstat -d command output

Figure 6-6 shows two samples taken by the mpstat -d command. The first
sample was taken when the system was slightly loaded with two processes
running in cycle. The second example was taken when the system was loaded
with more processes.
In our examples we have two physical processors of an p5-520 server configured
to run SMT. SMT architecture creates four logical processors above the two
physical processors. The cpu column stands for the logical CPU ID and does not
change until a DLPAR reconfiguration takes place. The ALL stands for the sum
of processor statistics.
The context switch statistics, shown in the cs column, are counted per each
processor and the ALL processor cs value is the sum of each cs line.
The ics column stands for the involuntary context switches. These context
switches are due to the exhaustion of available time slices.
The bound column stands for the total threads bound to the processor. This
happens if the user runs the bindprocessor command or when the program itself
binds threads to a processor. When there are processor bound threads, the push
column may show non zero values.
The rq column stands for the process run queue. The ALL processor rq value is
the sum of all rq lines.


273

The push column stands for the number of thread migrations due to starvation
load balancing. This happens when a thread cannot get processor time for a long
period of time. The scheduler determines if there is a better processor to run the
process, then it pushes the thread to that processor. If the thread is pushed
outside the affinity domain S3 (outside MCM) then this counter is incremented.
The S3pull column stands for the number of process migrations outside the
affinity domain S3 due to idle stealing. This happens when one particular CEC
board (S3 affinity domain) is underloaded and it is stealing threads from CPUs
from other CEC boards.
The S3grd column stands for the number of dispatches from the global run
queue, outside the affinity domain S3. This happens on multiple-CEC systems,
like the p5-570, when a process is dispatched to a different CEC board. Such
dispatches may occur when the loads to the CEC boards are significantly
unbalanced.
The S0rd, S1rd, S2rd, S3rd, S4rd and S5rd stand for the thread redispatches
within the affinity domains S0 (same logical cpu), S1 (same physical processor),
S2 (same processor chip), S3 (same MCM), S4 (same CEC plane) and S5
(outside CEC plane). The values for ALL are averaged per each column.
When the processor was slightly loaded, all the thread redispatches were
dispatched back to the same logical CPU and there were no available threads on
the run queue that could be dispatched to cpu0 or cpu2. In this case cs=0 for the
logical cpu, and the SXrd values show only a dash.
When the processor is loaded by more processes (or threads), the threads
cannot stay on the same logical CPU and sometimes they need to be
redispatched to different logical CPUs. This will cause the S0rd values to drop
from 100% to 99% or even more, and the S1rd values will start to grow as the
threads are redispatched to the closest available logical CPU. The S1rd column
is zero in our example, because we are using single chip modules of the p5-520
system and there is just one chip per processor module (SCM). When we
dispatch threads to the other physical processor, it is dispatched to the second
SCM module. These thread redispatches appear in the S3rd column.
The ilcs and vlcs columns show, respectively, involuntary and voluntary logical
context switches. These are hardware context switches due to partitioning using

SMT utilization
The addition of Simultaneous Multi Threading (SMT) requires statistics to be
reported on the use of the particular logical processors. Example 6-7 on
page 275 shows a sample of how to use the command and explanation follows.

274


Example 6-7 mpstat -s command output
# mpstat -s 1
Proc1
0.27%
cpu0
cpu2
0.17%
0.10%

Proc0
49.63%
cpu1
cpu3
3.14% 46.49%

The mpstat -s 1 2 command is run to get SMT statistics in 1 second samples.
First, the output gives the system information about the number of logical
processors and the entitlement assigned to the partition, as with any other flag.
The statistics are then organized into as many columns as there are logical
processors assigned to the partition.
The meanings of the columns are best described by an example: The total
assigned entitlement is 0.5. This entitlement is distributed across the physical
processors as 0.27% + 49.63% = 49.90%, or approximately 0.5 of the total
entitlement in the system configuration line of the output. Each physical
processor’s consumed entitlement is then redistributed again to logical
processors, for example Proc0 = cpu1 + cpu3, that is 3.14% + 46.49% = 49.63%.

6.6.5 lparstat command
The lparstat command is new in AIX 5L Version 5.3. It reports LPAR-related
information and statistics.
The optional command flags for the lparstat command are:
-i

Lists details on the LPAR configuration

-h

Adds summary hypervisor statistics to the default lparstat command
output

-H

Provides detailed hypervisor information, including statistics for each of
the hypervisor calls

The lparstat command with no options will generate a single report containing
utilization statistics related to the LPAR since boot time. The lparstat command
using the -h flag will add hypervisor summary statistics to the default lparstat
output. If an interval and count are specified, the output report will repeat every
interval seconds for count iterations.


275

The following is an example of this command, collecting statistics for one
five-second interval, with the hypervisor summary data highlighted.
# lparstat -h 5 1
%user %sys %wait %idle physc %entc lbusy
app vcsw phint %hypv hcalls
----- ---- ----- ----- ----- ----- -------- ---- ----- ----- -----0.0
0.5
0.0
99.4 0.00 1.0
0.0
- 1524
0
0.5
1542

Using the -H flag displays detailed hypervisor information, including statistics for
many hcall functions. The command shows the following for each of these hcalls:
Number of calls

Number of hypervisor calls made

Total Time Spent

Percentage of total time spent in this type of call

Hypervisor Time Spent

Percentage of hypervisor time spent in this type of call

Average Call Time

Average call time for this type of call in nanoseconds

Maximum Call Time

Maximum call time for this type of call in nanoseconds

# lparstat -H 5 1
Detailed information on Hypervisor Calls
Hypervisor
Call
remove
read
nclear_mod
page_init
clear_ref
protect
put_tce
xirr
eoi
ipi
cppr
asr
others
enter
cede
migrate_dma
put_rtce
confer
prod
get_ppp
set_ppp

276

Number of
Calls

%Total Time
Spent

%Hypervisor
Time Spent

Avg Call
Time(ns)

Max Call
Time(ns)

2
21
0
3
0
0
14
10
10
0
0
0
0
5
1595
0
0
0
27
1
0

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.5
0.0
0.0
0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.0
0.0
0.0
0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
99.6
0.0
0.0
0.0
0.1
0.0
0.0

417
148
1
774
1
1
544
536
526
1
1
1
1
397
7918
1
1
1
672
1502
1

550
294
0
2280
0
0
908
758
695
0
0
0
0
685
1343207
0
0
0
922
2579
0


purr
pic
bulk_remove
send_crq
copy_rdma
get_tce
send_logical_lan
add_logicl_lan_buf

0
1
0
0
0
0
0
0

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

1
309
1
1
1
1
1
1

0
410
0
0
0
0
0
0

To show the actual status of the partition you can use the lparstat -i command
in the AIX command line interface of the partition:
# lparstat -i
Node Name
Partition Name
Partition Number
Type
Mode
Entitled Capacity
Partition Group-ID
Shared Pool ID
Online Virtual CPUs
Online Memory
Maximum Memory
Minimum Memory
Minimum Capacity
Maximum Capacity
Capacity Increment
Active CPUs in Pool

:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:

applsrv
Apps_Server
4
Shared-SMT
Uncapped
0.30
32772
0
2
10
1
512 MB
1024 MB
128 MB
128
0.20
1.00
0.01
2
2
0.00
15.00%

To verify the dynamic action use the lparstat -i command on the selected
partition again. We changed the partition mode from uncapped to capped for
illustrative purposes.
# lparstat -i
Node Name
Partition Name
Partition Number
Type
Mode

:
:
:
:
:

applsrv
Apps_Server
4
Shared-SMT
Capped


277

Entitled Capacity
Partition Group-ID
Shared Pool ID
Online Virtual CPUs
Online Memory
Maximum Memory
Minimum Memory
Minimum Capacity
Maximum Capacity
Capacity Increment
Active CPUs in Pool
Unallocated Weight

:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:

0.30
32772
0
2
10
1
512 MB
1024 MB
128 MB
128
0.20
1.00
0.01
2
2
0.00
15.00%
0

6.6.6 sar command enhancements
Beginning with AIX 5L Version 5.3, the sar command reports utilization metrics
physc and %entc, which are related to Micro-Partitioning and simultaneous
multi-threading environments. These metrics are only be displayed on
Micro-Partitioning and simultaneous multi-threading environments. The physc
metric indicates the number of physical processors consumed by the partition (in
case of system-wide utilization) or logical CPU (if the -P flag is specified) and
%entc indicates the percentage of the allocated entitled capacity (in case of
system-wide utilization) or consumed entitled capacity (if the -P flag is specified).
When the partition runs in capped mode, the partition cannot get more capacity
than it is allocated. In uncapped mode, the partition can get more capacity than is
actually allocated to it. This is called consumed entitled capacity. If the -P flag is
specified and there is unused capacity, the sar command prints the unused
capacity as a separate CPU with cpu id U similar to what the mpstat command
does.
Highlights of command options that have changed in AIX 5L V5.3 are the
following:
-u

278

Reports per processor or system-wide statistics. When used with the
-P flag, the information is provided for each specified processor;
otherwise, it is provided only system-wide. Because the -u flag
information is expressed as percentages, the system-wide information
is simply the average of each individual processor's statistics except
when using a shared pool of processors using Micro-Partitioning


technology, where the percentage are relative to the entitled capacity.
The following values are displayed:
%idle

Percentage of time the cpu or cpus were idle with no
outstanding disk I/O requests.

%sys

Percentage of time the cpu or cpus spent in execution at the
system (or kernel) level.

%usr

Percentage of time the cpu or cpus spent in execution at the
user (or application) level.

%wio

Percentage of time the cpus were idle during which the
system had outstanding disk/NFS I/O requests.

physc

Number of physical processors consumed. This is reported
only if the partition is running with shared processors or
simultaneous multi-threading enabled (-P only).

%entc

Percentage of entitled capacity consumed (physc/ent)*100.
This is reported only if the partition is running with shared
processors or simultaneous multi-threading enabled
(-P only).

Note: The sar command reports system unit activity if no other specific
content options are requested. If the -P flag is used and the partition is running
with shared processors, and if the partition capacity usage is less than what is
allocated, then a CPU row with cpuid U will be reported to show the
system-wide unused capacity. If the partition is running with shared
processors in uncapped mode, then %entc will report the percentage of
consumed entitled capacity against each CPU row and percentage of
allocated entitled capacity in the system-wide CPU row.

6.6.7 topas command enhancements
Figure 6-7 on page 280 displays a sample default output of the topas command.
If topas runs on a partition with shared processor allocation (Micro-Partitioning),
beneath the CPU utilization there are two new values displayed:
Physc
Number of physical processors utilized by the partition (if using
Micro-Partitioning technology).
%Entc
Percentage of Entitled Capacity utilized by a partition (if using
Micro-Partitioning technology).


279

Figure 6-7 topas default sample output

Figure 6-8 on page 281 shows the output displayed when using the new -L flag of
the topas command. The -L flag switches the output to logical partition display.
You can either use -L when invoking the topas command, or as a toggle during
running of topas. In this mode topas displays data similar to the mpstat and
lparstat commands.

280


Figure 6-8 topas logical partition sample output

6.7 Per file system I/O pacing
I/O pacing is used to prevent a large number of I/O page outputs for one file from
monopolizing the I/O bus. There are two tunable parameters you can set up:
minpout

Minimum number of outstanding I/O pages.

maxpout

Maximum number of outstanding I/O pages.

When the number of outstanding output pages for a particular file reaches the
maxpout value, any other threads waiting to do additional pageouts on that
segment will sleep until the number of outstanding output pages falls to the
minpage value for that file.
In previous versions of AIX this behavior was tunable only system wide using the
smit chgsys fastpath or the chdev -l sys0 -a maxpout=24 -a minpout=32
command. There were cases when some file systems, like database file
systems, required different values then other file systems, like temporary file
systems.


281

In Version 5.3 the I/O pacing can be tuned on a per file system basis. This tuning
is done when using the mount command:
mount -o minpout=40,maxpout=60 /fs

A more user friendly way to do this is to use SMIT or to edit the /etc/filesystems.
The highlighted mount options of the smit crjfs2lvstd menu in Figure 6-9 show
how to set the minpout and maxpout parameters on a per file system basis.

Figure 6-9 smit crjfs2lvstd

You can check that the file system is mounted correctly by using the mount
command as follows:
s4 # mount
node
mounted
mounted over
-------- --------------- --------------/dev/hd4
/
/dev/hd2
/usr
/dev/hd9var
/var
/dev/hd3
/tmp
/dev/hd1
/home
/proc
/proc
/dev/hd10opt
/opt
/dev/fslv00
/fs
rw,maxpout=60,minpout=40,log=/dev/loglv01

282


vfs
-----jfs2
jfs2
jfs2
jfs2
jfs2
procfs
jfs2
jfs2

date
-----------Jul 09 18:19
Jul 09 18:19
Jul 09 18:19
Jul 09 18:19
Jul 09 18:20
Jul 09 18:20
Jul 09 18:20
Jul 15 11:06

options
--------------rw,log=/dev/hd8
rw,log=/dev/hd8
rw,log=/dev/hd8
rw,log=/dev/hd8
rw,log=/dev/hd8
rw
rw,log=/dev/hd8

Advanced support can check the file system in the kernel using the kdb kernel
debugger. First, with the pdt * sub-command get the file system slot ID from the
list of mounted file systems, then get the entries related to this slot ID. The
following is an example of this:
# kdb
...
(0)> pdt *
SLOT
0000
0080
0081
0082
0083
0084
0085
0086
0087
0088
0089
008C
008F

vmmdseg+90A0000
vmmdseg+90A5400
vmmdseg+90A54A8
vmmdseg+90A5550
vmmdseg+90A55F8
vmmdseg+90A56A0
vmmdseg+90A5748
vmmdseg+90A57F0
vmmdseg+90A5898
vmmdseg+90A5940
vmmdseg+90A59E8
vmmdseg+90A5BE0
vmmdseg+90A5DD8
(0)> pdt 8f

NEXTIO
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF
FFFFFFFF

DEVICE DMSRVAL
IOCNT <name>
8000000A00000002 00000000 00000000 paging
01460CA8 00000000 00000000 remote
0408D100 00000000 00000000 remote
8000000A00000007 00003022 00000000 local client
8000000A00000004 00000000 00000000 local client
8000000A00000005 00000000 00000000 local client
8000000A00000006 00000000 00000000 local client
0408D040 00000000 00000000 remote
0408B708 00000000 00000000 remote
8000000A00000008 00000000 00000000 local client
8000000A00000009 00000000 00000000 local client
040ECC18 00000000 00000000 remote
8000002D00000008 00000000 00000000 local client

PDT address F1000100090A5DD8 entry 008F of 1FFF, type: LOCAL CLIENT -- XPAGER
next pdt on i/o list (nextio) : FFFFFFFF
dev_t or strategy ptr (device) : 8000002D00000008
last frame w/pend I/O (iotail) : FFFFFFFFFFFFFFFF
free buf_struct list (bufstr) : F10006000D862E90
total buf structs
(nbufs)
: 0000
available (PAGING)
(avail)
: 0000
disk map srval
(dmsrval) : 00000000
i/o's not finished
(iocnt)
: 00000000
device wait list (devwait)
: 0000000000000000
buffer wait list (bufwait)
: 0000000000000000
logical volume lock (lock)
:@F1000100090A5E30 00000000
buffer list lock
(buf_lock) :@F1000100090A5E38 00000000
flag bits
(devflags) : 84000000
max phys Xlation ent (maxphys) : 00000020
pageout down limit
(minpout) : 00000028
pageout up limit
(maxpout) : 0000003C
external pager (pager) : 0000000001458158
(0)>


283

284


7

Chapter 7.

Networking
AIX 5L provides many enhancements in the networking area. Described in this
chapter, they include:
NFS Version 4
Multipath routing enhancements
PMTU enhancements
DHCPv6 support
IPv6 functional updates
Interface layer support for hotplug network adapters
Support for SLP client service
Stream Control Transmission Protocol (SCTP)
Network RAS enhancements
Enable IP Security with intervening NAT devices
AIX SNMP subagent enhancements
SMBFS enhancements
Ported 64-bit DES NFS kernel extension
BIND domain search improvement
Network API support for ARP


285

7.1 NFS Version 4
The Network File System (NFS) is a distributed file system that allows users to
access files and directories located on remote computers and treat those files
and directories as if they were local. Prior to Version 5.3, AIX supports the NFS
Version 2 and 3 protocols. Starting with Version 5.3, AIX supports NFS Version 4
as well as NFS Version 2 and 3. NFS Version 4 includes the following features:
More focus on security
– RPCSEC-GSS RPC
– Support for foreign domain access using identity mapping (requires
deployment of LDAP and Enterprise Identity Mapper)
RPC operations for file locking move into the main NFS protocol
– No rpc.lockd, rpc.statd and rpc.mountd in NFS V4
– Friendlier to firewalls
NFS V4 only supports TCP
– No support for UDP
NFS V4 ACL support in JFS2
– Fine-grained access control for file system objects
– Support inheritance features
You can configure NFS V4 through SMIT, command line, and Web-based System
Manager.

RPCSEC-GSS RPC authentication
NFS V4 can be configured with RPCSEC_GSS to provide stronger security for
the protocol. While not mandated, other more traditional methods such as
AUTH_SYS are also be provided. RPCSEC_GSS is based on the function of
GSS-API. This allows for support of multiple security mechanisms without the
requirement of adding new RPC authentication methods.
In addition to the authentication aspects of these mechanisms, each offers
selectable levels of protection such as integrity or privacy on the message data.
The RPCSEC_GSS service attribute specifies the level of message protection
and is mandatory for RFC compliance. The Version 4 protocol provides support
for the NFS client and server to negotiate both the mechanisms and the service
levels.

286


File locking and open share reservations
NFS V4 brings significant changes for file locking when compared to earlier NFS
versions. The RPC operations for file locking move into the main NFS protocol.
The separate Network Lock Manager (NLM) and status monitor protocols in
earlier NFS versions are eliminated, along with the corresponding rpc.lockd and
rpc.statd daemons in NFS V4. Note that the rpc.statd and rpc.lockd are still used
for NFS V2 and V3 for file locking. In addition, the connection between client and
server is now stateful in NFS V4, while the previous versions of NFS were
stateless.

NFS V4 ACL support
The NFS V4 protocol defines an ACL model and data types, and provides RPC
operations for passing Access Control List (ACL) specifications between the
server and client. The defined ACL format closely resembles the ACLs of the PC
environment. The ability to provide for ACL query and specification as part of the
NFS V4 protocol is a valuable feature. ACLs are an optional feature of the
protocol. They are communicated in the recommended portion of the NFS V4
file.
Server support for NFS V4 ACL
When an AIX is an NFS V4 server, it can service a variety of clients. A client
can be an AIX or third party NFS. One of the NFS V4 server’s responsibilities
is to service NFS V4 ACL requests. These requests are setting and getting
ACLs of the underlying file system. These requests can only be serviced
when the underlying file system supports NFS V4 ACL. JFS2 is enhanced to
support NFS V4 ACL in Version 5.3.
Note: Refer to 3.2.6, “JFS2 ACL support for NFS V4” on page 131 for further
details.
Client support for NFS V4 ACL
When an AIX is an NFS V4 client, it is able to set and retrieve NFS V4 ACL
from its NFS V4 server if the server supports the NFS V4 ACL.
AIX ACL support outside of NFS V4 protocol
AIX ACLs are supported in NFS V2 and NFS V3 when both the client and
server are AIX. This service is not defined in the NFS protocol. When an NFS
client file system is mounted, a handshake is attempted to establish the AIX
ACL support communication channel. When both the client and server are
AIX, the handshake is successful and a TCP port is set up to support the
off-protocol AIX ACL communication. NFS V4 will continue this off-band
support for AIX ACL between client and server.

Chapter 7. Networking

287

NFS V2 and NFS V3 ACL support
The NFS V2 and NFS V3 code will not be changed to support any new ACL
function.
Note: NFS clients use the NFS version 3 protocol by default.

The new daemon processes
The following new daemon processes, highlighted in Example 7-1, are
introduced:
nfsrgyd

The nfsrgyd daemon provides a name translation service for NFS
servers and clients. This daemon must be running in order to
perform translations between NFS string attributes and UNIX
numeric identities. The nfsrgyd daemon should be up and running
from servers and clients using NFS V4 or RPCSEC-GSS.

gssd

Some NFS security methods, such as Kerberos 5, are provided
under a more general mechanism called General Security
Services, or GSS. In AIX, GSS services are provided by a library in
the IBM Network Authentication Service (NAS) fileset. NAS is
shipped on the expansion pack. If the gssd daemon is not running,
then efforts to access files using NFS using GSS security methods
such as Kerberos 5 will fail.

Example 7-1 The nfsrgyd and gssd daemons
# lssrc -g nfs
Subsystem
biod
nfsd
nfsrgyd
gssd
rpc.mountd
rpc.lockd
rpc.statd

Group
nfs
nfs
nfs
nfs
nfs
nfs
nfs

PID
258180
413858
266276
446648

Status
active
active
active
active
inoperative
inoperative
inoperative

New command utilities
The following are new command utilities:
chnfsdom

The chnfsdom command changes the local NFS domain of the
system. The local NFS domain is stored in the
/etc/nfs/local_domain file.

nfs4cl

Displays or modifies current NFSv4 statistics and properties.

nfshostkey An NFS server (or full client) using RPCSEC_GSS RPC security
must be able to acquire credentials for its host principal to accept

288


requests. The nfshostkey command is used to configure this
information; it takes the principal and the path to a keytab file for
that principal. This information goes into /etc/nfs/hostkey.
chnfsim

Changes NFS foreign identity mappings.

Setting up a network for RPCSEC-GSS
Figure 7-1 shows the setup of an NFS V4 server and client in which a network is
configured for RPCSEC-GSS.
Note:
1. RPCSEC_GSS also can be configured with NFS V3 in AIX 5L Version 5.3.
2. Although NFS V4 can be configured without RPCSEC-GSS in Version 5.3,
the following example uses RPCSEC-GSS for better security.

Server3

Authentication

KDC (Kerberos Server)
Kerberos domain:kr.ibm.com
Authentication
IP: 9.3.5.196

Server1

NFS data

NFS Client
Kerberos client
Kerberos domain: kr.ibm.com
IP: 9.3.5.194

Server2

NFS Server
Kerberos client
Kerberos domain:kr.ibm.com
IP: 9.3.5.195

Figure 7-1 An example configuration

The system server3.kr.ibm.com® will be configured as the Key Distribution
Center (KDC) server, the Kerberos realm KR.IBM.COM will be created, and
server2.kr.ibm.com will be NFS server offering file systems exported with
RPCSEC-GSS.
In addition, this network has the following users: slson and myahn.
Create the /etc/exports file through SMIT or using the vi command.
# vi /etc/exports


289

/slson -vers=4,sec=krb5:krb5i:krb5p,rw

Step1: Set up the KDC server
Note: Kerberos requires that the system time be reasonably close throughout
the network. Before beginning this procedure, you should set up a mechanism
to automatically synchronize time throughout the network, such as the AIX
timed daemon or an NTP setup.
In this scenario, server3.kr.ibm.com will be configured as the KDC server
(Example 7-2).
1. Install the krb5.server and modcrypt.base file set on server3. Install the
krb5.client and modcrypt.base file set on server2 and server1. These file sets
are available from AIX 5L Version 5.3 expansion pack.
Run the following command:
config.krb5 -S -d kr.ibm.com -r KR.IBM.COM

After running this command, the system will ask for a Master Database
password and a password for the administrative principal.
Example 7-2 Setting up KDC server
# lslpp -L |grep krb5
krb5.client.rte
krb5.client.samples
krb5.server.rte
# lslpp -L |grep modcrypt
modcrypt.base.includes
modcrypt.base.lib

1.4.0.0
1.4.0.0
1.4.0.0

5.3.0.0
5.3.0.0

C
C
C

C
C

F
F
F

F
F

Network Authentication Service

Cryptographic Library Include
Cryptographic Library
(libmodcrypt.a)

# config.krb5 -S -d kr.ibm.com -r KR.IBM.COM
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 'KR.IBM.COM'
master key name 'K/M@KR.IBM.COM'
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/admin@KR.IBM.COM;
defaulting to no policy. Note that policy may be overridden by
ACL restrictions.

290


Enter password for principal "admin/admin@KR.IBM.COM":
Re-enter password for principal "admin/admin@KR.IBM.COM":
Principal "admin/admin@KR.IBM.COM" 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.

2. Create principals for each user and host by running the
/usr/krb5/sbin/kadmin.local command on the KDC server. This example
creates Kerberos principals (Example 7-3) that match the UNIX user name of
the associated user. The principal name will be mapped to the user name by
NFS to determine the UNIX credential associated with the principal. For this
network, we created the following principals:
– slson
– myahn
– nfs/FQDN (principals for hosts)
The chosen user principal names must match the corresponding user names
in the local registry (/etc/passwd, for example). NFS uses the principal name
as a user name to obtain user and group IDs on the local system. If the
names do not match, the access will be treated as an anonymous access.
Example 7-3 Add principal on KDC server
# /usr/krb5/sbin/kadmin.local
kadmin.local: addprinc slson
WARNING: no policy specified for slson@KR.IBM.COM;
ACL restrictions.
Enter password for principal "slson@KR.IBM.COM":
Re-enter password for principal "slson@KR.IBM.COM":
Principal "slson@KR.IBM.COM" created.
kadmin.local: addprinc myahn
WARNING: no policy specified for myahn@KR.IBM.COM;
ACL restrictions.
Enter password for principal "myahn@KR.IBM.COM":
Re-enter password for principal "myahn@KR.IBM.COM":
Principal "myahn@KR.IBM.COM" created.
kadmin.local: addprinc nfs/server2.kr.ibm.com
WARNING: no policy specified for nfs/server2.kr.ibm.com@KR.IBM.COM;
ACL restrictions.


291

Enter password for principal "nfs/server2.kr.ibm.com@KR.IBM.COM":
Re-enter password for principal "nfs/server2.kr.ibm.com@KR.IBM.COM":
Principal "nfs/server2.kr.ibm.com@KR.IBM.COM" created.

Step 2: Configure the NFS client and server
The NFS client and server will now be configured as Kerberos clients
(Example 7-4) by using the config.krb5 command. How this is done will depend
on how the KDC was configured. In this scenario we ran the following command
on each NFS system:
#config.krb5 -C -d kr.ibm.com -r KR.IBM.COM -c server3.kr.ibm.com -s
server3.kr.ibm.com

It is now possible to run kinit as any of the user principals on any of the
configured systems. For example, to kinit as user myahn, run the following
command:
#/usr/krb5/bin/kinit myahn

You will need to specify myahn’s Kerberos, not AIX, password. This example
uses kinit to authenticate the user. It is possible to configure AIX to use Kerberos
authentication during system login. For more information, see Authenticating to
AIX Using Kerberos in AIX 5L Version 5.3 Security Guide.
Example 7-4 Setting up the Kerberos client
#Run the following commands from NFS server(server2) and NFS client(server1).
# config.krb5 -C -d kr.ibm.com -r KR.IBM.COM -c server3.kr.ibm.com -s server3
server3.kr.ibm.com
Initializing configuration...
Creating /etc/krb5/krb5.conf...
The command completed successfully.
# /usr/krb5/bin/kinit myahn
Password for myahn@KR.IBM.COM:

Step 3: Configure the NFS server keytab entry
The NFS server will now be configured with the appropriate keytab entry. In this
scenario (Example 7-5), we configured the keytab entry for server2.kr.ibm.com.
1. From server2.kr.ibm.com, run the kadmin command. Then, run the following
command:
kadmin:ktadd nfs/server2.kr.ibm.com

This creates a keytab file.

292


2. Set up the gssd daemon to use the keytab file you just created with the
nfshostkey command. In this scenario, we ran the following:
#nfshostkey -p nfs/server2.kr.ibm.com -f /etc/krb5/krb5.keytab

3. Set up the gssd daemon to start up automatically by running the following
command:
#chnfs -S -B
Example 7-5 Creating keytab file and setting up the gssd daemon for NFS Server
# /usr/krb5/sbin/kadmin -p admin/admin
Authenticating as principal admin/admin with password.
Password for admin/admin@KR.IBM.COM:
kadmin: ktadd nfs/server2.kr.ibm.com
Entry for principal nfs/server2.kr.ibm.com with kvno 2, encryption type Triple
DES cbc mode with HMAC/sha1 added to keytab WRFILE:/etc/krb5/krb5.keytab.
Entry for principal nfs/server2.kr.ibm.com with kvno 2, encryption type ArcFour
with HMAC/md5 added to keytab WRFILE:/etc/krb5/krb5.keytab.
Entry for principal nfs/server2.kr.ibm.com with kvno 2, encryption type AES-256
CTS mode with 96-bit SHA-1 HMAC added to keytab WRFILE:/etc/krb5/krb5.keytab.
Entry for principal nfs/server2.kr.ibm.com with kvno 2, encryption type DES cbc
mode with RSA-MD5 added to keytab WRFILE:/etc/krb5/krb5.keytab.
# nfshostkey -p nfs/server2.kr.ibm.com -f /etc/krb5/krb5.keytab
# chnfs -S -B
0513-059 The gssd Subsystem has been started. Subsystem PID is 434406.

Step 4: Configure the NFS registry daemon
At this point, the NFS server will work, although all users will appear as nobody.
Any users that do not exist will have access to the exported directory only as
nobody. To get user names to map properly, you must configure the NFS registry
daemon, as follows:
1. Set up the domain using the chnfsdom command.
2. Set up the /etc/nfs/realm.map file (Example 7-6); this file should contain one
line, with the realm name followed by the local domain. The realm entry in this
file is not case-sensitive, so technically, this entry is not required.
Example 7-6 Change nfs domain and set up map file
# chnfsdom kr.ibm.com
# vi /etc/nfs/realm.map
realm.map KR.IBM.COM kr.ibm.com
#refresh -s nfsrgyd


293

3. For server1.kr.ibm.com, which will be the NFS client, start up the gssd
daemon using the chnfs -S -B command (Example 7-7). Before trying any
Kerberos client operations, the user must use kinit to obtain valid
credentials. You are now able to mount the NFS file system with the nfs4 and
Kerberos option.
Example 7-7 Mount NFS4 file system with Kerberos on the NFS client
# chnfs -S -B
# mount -o sec=krb5,vers=4 server2.kr.ibm.com:/slson /mnt
# mount
node
mounted
mounted over
vfs
date
options
-------- --------------- --------------- ------ ------------ --------------/dev/hd4
/
jfs
Jul 09 11:03 rw,log=/dev/hd8
/dev/hd2
/usr
jfs
/dev/hd9var
/var
jfs
/dev/hd3
/tmp
jfs
/dev/hd1
/home
jfs
/proc
/proc
procfs Jul 09 11:05 rw
/dev/hd10opt
/opt
jfs
/dev/nimlv
/export
jfs
Jul 09 10:14
rw,log=/dev/loglv00
server2.kr.ibm.com /slson
/mnt
nfs4
Jul 13 10:41
sec=krb5,vers=4

If you are trying to mount the NFS file system from a machine in which Kerberos
has not been set up properly, the attempt will fail with the following error
messages (Example 7-8).
Example 7-8 Mount failure without Kerberos configuration
# mount server2.kr.ibm.com:/slson /mnt
mount: 1831-011 access denied for server2.kr.ibm.com:/slson
mount: 1831-008 giving up on:
server2.kr.ibm.com:/slson
The file access permissions do not allow the specified action.

7.2 Multipath routing enhancements
Starting with AIX 5L Version 5.1, you are allowed to use multiple routes to the
same destination (including multiple default routes) through the Multipath
Routing feature. The cost (hopcount) is used to determine the route to use when
there are multiple routes to the same destination. A round-robin scheme is used
to select a route when there are multiple routes with the same destination and

294


cost. The following feature is added to provide you with more flexibility for
multipath routing in Version 5.3.

Configure multipath routing policy with SMIT
Configurable multipath routing methods are now included in the SMIT mkroute
fastpath. Notice that there are six items you can configure for multipath routing
as shown in Figure 7-2.

Figure 7-2 Add Static Route SMIT panel

If the policy is not explicitly set during the creation of a static route and multipath
routing is used, then the global no command option mpr_policy determines the
policy that will be used.
The default policy is Weighted Round-Robin, which behaves just like
Round-Robin when the weights are all 1. If the default policy is chosen when the
route is created, then the no option “ mpr_policy” takes precedence and its value
is used to determine the policy that will be used. On the other hand, if the policy
is explicitly set to Weighted Round-Robin then this setting overrides the
mpr_policy setting. Table 7-1 explains each routing method for multipath routing.
Table 7-1 Routing method for multipath routing
Policy for
multipath routing
Default(Global)

Policy
number
N/A

Descriptions
no command option named mpr_policy determines
the policy that will be used.


295

Policy for
multipath routing

Policy
number

Descriptions

Weighted RR

1

User-configured weights (optional when routes are
added) will be combined with round-robin.

Random

2

A route will be randomly picked among the multiple
routes.

Weighted Random

3

A route will be randomly picked among the multiple
routes but with weights.

Lowest Utilization

4

The route with the lowest utilization (number of
active connections) will be picked.

Hash-Based

5

Hash based on Source/Dest IP address and
TCP/UDP port number.

Weights
In the previous SMIT menu, shown in Figure 7-2 on page 295, you can find a
new item of Weight (for Weighted Multipath Routing Policy). If there are weights
assigned to multiple duplicate routes (same netmask and cost) and weighted
round-robin (5) is chosen as the routing policy, then the system round-robins
between these routes based on the weights. For example, consider the case
where there are two routes to a specific destination and there are weights of 3
and 1 assigned to the first and the second route respectively. Now when
weighted round-robin is chosen, then the first route is used three times in
succession before the packets are routed using the second route.

Configure multipath routing policy through command line
In order to add routing policy and/or weights use the route add command:
route add .. -policy <routing policy> -weight <weight of the route>

Example 7-9 shows the command for adding a routing table to the destination of
9.3.5.194 and gateway 192.168.100.1 with a Weighted Round-Robin for policy
and 3 for weight.
Example 7-9 Setting policy and weight for multipath routing
# route add 9.3.5.194 192.168.100.1 -policy 1 -weight 3
# netstat -rn
Routing tables
Destination
Gateway
Flags Refs
Use
Route Tree for Protocol Family 2 (Internet):
9.3.5.194
192.168.100.1
UGH
0
127/8
127.0.0.1
U
12

296

If

en2
lo0


0
69940

PMTU Exp Groups

-

-

192.168.100.0
192.168.100/24
192.168.100.11
192.168.100.255

192.168.100.11
192.168.100.11
127.0.0.1
192.168.100.11

UHSb
U
UGHS
UHSb

0
2
0
0

Route Tree for Protocol Family 24 (Internet v6):
::1
::1
UH
0

0
97
0
0

en2
en2
lo0
en2

-

-

28

lo0

-

=>

-

To look at the routing tables including the policy and weights use the netstat -C
command.
Note: If the policy is not explicitly set and multipath routing is used, then the
global no command option called mpr_policy determines the policy that will be
used.

7.3 PMTU enhancements
The following sections discuss the Path Maximum Transmission Unit
enhancements found in AIX 5L Version 5.3.

PMTU discovery in AIX 5L Version 5.3
The current Path Maximum Transmission Unit (PMTU) discovery implementation
uses ICMP Echo Request and ICMP Echo Reply packets to discover PMTU
using IPv4. This involves sending extra packets through the network. Some
system administrators set up their firewall to drop ICMP Echo packets, resulting
in the failure of this method of PMTU discovery. Current implementation clones a
route to store the PMTU value in the routing table. This cloning mechanism does
not work for multipath routing because the cloned route is always used instead of
alternating between the two multipath network routes.
The new feature is implemented in Version 5.3 to have the Path MTU discovery
mechanism use TCP packets and UDP datagrams rather than ICMP Echo
packets. Also, the discovered MTU will not be stored in the routing table. This will
enable multipath routing to work with PMTU discovery since no cloning will be
necessary.
IPv6 never sends an ICMPv6 packet to detect the PMTU. The first packet of a
connection always starts the process. Also, IPv6 routers must never fragment
packets and must always return an ICMPv6 Packet too big message if they are
unable to forward a packet because of a smaller outgoing MTU. Thus, for IPv6
the only change necessary is to make PMTU discovery work with Multipath
routing.


297

Table 7-2 on page 298 lists the PMTU enhancements in Version 5.3.
Table 7-2 PMTU enhancements
Version 5.2 or
earlier

Version 5.3

Improvements

Method of PMTU
discovery (IP V4)

Extra ICMP echo
request and ICMP
echo reply

No additional
packets (Just
uses TCP/UDP
itself)

Network
performance

PMTU data
(IP V4 and IP V6)

Part of routing table

Separate PMTU
table

No more cloning
data in routing table
so that we can use
multipath routing
with PMTU

Command to find
PMTU data

netstat -rn

pmtu display

Take note!

Command to display PMTU table
Beginning with Version 5.3, when the Path MTU discovery is attempted for a
destination, a pmtu entry gets created in a Path MTU (PMTU) table. This table
can be displayed using the pmtu display command. Accumulation of pmtu
entries can be avoided by allowing unused pmtu entries to expire and be deleted.
PMTU entry expiration is controlled by the pmtu_expire option of the no
command. pmtu_expire is set to 10 minutes by default.
The syntax of the pmtu command is as follows:
pmtu [-inet6] display/[delete [-dst destination] [-gw gateway] ]

Flags used with the pmtu delete command are the following:
-dst

Specifies the destination of the pmtu entry to be deleted.

-gw

Specifies the gateway of the pmtu entry to be deleted.

-inet6

Specifies to display or delete Ipv6 pmtu entry.

If you want to see the pmtu table, simply type a pmtu display command and the
output will be shown as in Figure 7-3.

298


Figure 7-3 Output of the pmtu display command

7.4 DHCPv6 support
DHCP is an application-layer protocol that allows a client machine on the network
to get IP addresses and other configuration parameters from the server. These
parameters are defined in options. Options are obtained by exchanging packets
between a daemon on the client and another on the server. DHCP as
implemented in the previous AIX versions only supports IPv4. Starting with
Version 5.3, support for DHCPv6 that uses IPv6 is now available. You need to
configure /etc/dhcpv6/dhcpsdv6.cnf for the dhcpv6 server and
/etc/dhcpv6/dhcpc6.cnf for the dhcpv6 client. The name of the /etc/dhcprd.cnf file
remains unchanged, with newly added key words for DHCPv6 support.

Configuration for DHCPv6 server
There is no dhcpsdv6 file by default and you need to create one for your
configuration. If you go to the /usr/samples/tcpip/dhcpv6 directory, you can see
some sample files that might help you to understand dhcpv6 configurations.
Example 7-10 shows a basic dhcpv6 server configuration.
Example 7-10 dhcpv6 server configuration file /etc/dhcpv6/dhcpsdv6.cnf
logging_info{
logFileSize
logItem
logItem
logItem

4000
SYSERR
PROTERR
WARNING


299

logItem
logItem
logItem
logItem
logItem
numLogFiles
logFileName

EVENT
ACTION
INFO
ACNTING
TRACE
3
"/var/tmp/dhcpsdv6.log"

}
duid 1 en0
numprocessthreads 10
preference-number 255
reconfig-policy no
unicast-option
yes
leaseExpiredInterval 3000 seconds
unicast-enable
yes
saveInterval
60 seconds
clientpruneintv
20 seconds
subnet dead:dead:aaaa:: 48 dead:dead:aaaa:aaaa::0006-dead:dead:aaaa:aaaa::000a
{
interface-id "en1"
preferred-lifetime
100 seconds
valid-lifetime
200 seconds
rapid-commit yes
option 23 dead::beef beef:aaaa::bbbb:c aaaa:bbbb::cccc
option 24 ibm.com austin.ibm.com
}

Table 7-3 defines the keywords used in the /etc/dhcpv6/dhcpsdv6.cnf file.
Table 7-3 Keywords in /etc/dhcpv6/dhcpsdv6.cnf
Keyword
duid

Used to identify the server.
The following values are allowed:
duid 1 interface
duid 2 interface
duid 3 enterprise number identifier
duid number 0xhexdigit

numprocessthreads

Specifies the number of packet processors threads to create.
Minimum of one. Each process thread handles one client. The
default is 30.

preference-number

300

Description

Allows the client to identify the server it prefers to obtain
information from. The higher the value, the greater the chance
the client will use this server for its services. The default and
maximum value is 255.


Keyword

Description

reconfig-policy

Allows the server to send reconfiguration message to the client.
By default this is not set and is considered off.

unicast-option

Allows containers to offer message exchange by unicasting.
This can use used to turn on and off individual containers and
subnets even if the server policy differs. By default this is not set
and is considered off.

leaseExpiredInterval

Specifies how often addresses in the BOUND state are
checked to see if they have expired. If the address has expired,
the state is moved to EXPIRED. If a unit is not set, the system
default is set to seconds. The default value is 900 seconds.

unicast-enable

Unicast policy for the server, this allows the server to
communicate using unicast. By default this is turned on.

saveInterval

Specifies how often the DHCP server should force a save of the
open databases. For heavily loaded servers, this should be 60
or 120 seconds. If unit is not set, the system default is set to
seconds. The default value is 3600 seconds.

clientpruneintv

Specifies how often the DHCP server removes clients that are
not associated with any address (in the UNKNOWN state) from
the database. This reduces the memory use of the DHCP
server. If units is not set, the system default is set to seconds.
The default value is 3600 seconds.

subnet

Specifies subnet to be used. The subnetid must be an IPv6
address. The prefix-length must be a positive integer less than
128.

interface-id

This can only be listed under subnet. Client requests received
on this interface will be allowed to obtain addresses.

preferred-lifetime

The preferred lifetime of the IANA or IATA. The default is 43200
seconds.

valid-lifetime

The valid lifetime of the IANA or IATA. The default is 86400
seconds.

rapid-commit

Allows for the server to do rapid commit for the container or
globally set. By default this is not set and is considered off.

option 23

DNS servers (IPv6 address).

option 24

Domain list (Domain name).


301

Configuration for DHCPv6 client
The /etc/dhcpv6/dhcpc6.cnf file is used to configure the DHCPv6 client. There
also is a sample file under the /usr/samples/tcpip/dhcpv6 directory. Example 7-11
shows a simple dhcpv6 client configuration.
Example 7-11 dhcpv6 client configuration file: /etc/dhcpv6/dhcpc6.cnf
interface

}
option

en1 {
option ia-na {
ia-id01
renew-time0x40
rebind-time0x60
request-option { 3 23 24 }

Table 7-4 defines the keywords used in the /etc/dhcpv6/dhcpc6.cnf file.
Table 7-4 Keywords in /etc/dhcpv6/dhcpc6.cnf
Keyword

Description

ia-na

Specifies option 3. If specified, the client requests
non-temporary addresses from the server.
Option ia-na takes the following parameters:
ia-id
<value>
renew-time <value>
rebind-time <value>
These parameters specify the user-preferred values and are
optional. The value specified can be a decimal number or a hex
number prefixed with '0x'.

request-option

If specified, the client requests a list of options from the server.

Configuration for DHCPv6 relay agent
The /etc/dhcprd.cnf file is the configuration file for the DHCP and BOOTP relay
agent. Even though the file name is the same as the previous version, some key
words are added to support IPv6. Example 7-12 shows a simple dhcpv6 relay
agent configuration.
Example 7-12 dhcpv6 relay agent configuration file:/etc/dhcprd.cnf
relay IPv6
server6 9.3.5.196

Table 7-5 provides the keywords newly added in the /etc/dhcprd.cnf file.

302


Table 7-5 Key words in /etc/dhcprd.cnfi
Keyword

Description

relay

IPv4, IPv6, or ALL

server6

Specifies the IPv6 address of the dhcpv6 server.

option6

Specifies the dhcpv6 relay agent options. The keyword is valid
only if the relay mode is set to IPv6 or ALL.

7.5 IPv6 functional updates
AIX 5L Version 5.3 is fully compliant with RFC 3542: Advanced Sockets
Application Program Interface (API) for IPv6. This API supports advanced
applications which typically use raw sockets to access IPv6 or ICMPv6 header
fields. The main changes implemented in Version 5.3 for IPv6 are:
Structure and constant definitions required for applications to use IPv6 raw
sockets have been updated in the AIX header files to conform to RFC 3542.
A number of socket options and accompanying ancillary data types have
been added to allow UDP/raw socket based applications to obtain information
about incoming IPv6 datagrams. For example, applications can choose to
receive data about the received interface, destination address, and received
hop limit of an incoming IPv6 datagram. Applications can also receive
extension headers from incoming IPv6 datagrams. The following are socket
options: IPV6_RECVPKTINFO, IPV6_RECVHOPLIMIT,
IPV6_RECVTCLASS, IPV6_RECVRTHDR, IPV6_RECVHOPOPTS,
IPV6_RECVDSTOPTS.
Applications can also set socket options to control fields in outgoing IPv6
datagrams. They may also specify extension headers to be included in
outgoing datagrams. The following are socket options: IPV6_PKTINFO,
IPV6_NEXTHOP, IPV6_TCLASS, IPV6_RTHDR, IPV6_HOPOPTS,
IPV6_DSTOPTS, IPV6_RTHDRDSTOPTS.
A number of additional features can be enabled using IPv6 socket options:
IPV6_USE_MIN_MTU

Controls Path MTU discovery.

IPV6_DONTFRAG

Prevents fragmentation of outgoing packets on this
socket.

IPV6_RECVPATHMTU

When an application sends packets too big for the
outgoing path MTU, this options allows the
application to receive the actual path MTU as
ancillary data.


303

IPV6_PATHMTU

Allows application to call this socket option to
obtain the path MTU at any time.

A number of new library functions have been implemented that allow
applications to easily construct IPv6 extension headers.
Functions to build/read a routing header:
– socklen_t inet6_rth_space(int, int);
– void *inet6_rth_init(void *, socklen_t, int, int);
– int inet6_rth_add(void *, const struct in6_addr *);
– int inet6_rth_reverse(const void *, void *);
– int inet6_rth_segments(const void *);
– struct in6_addr *inet6_rth_getaddr(const void *, int);
Functions to build/read hop-by-hop/destination options header:
– int inet6_opt_init(void *, socklen_t);
– int inet6_opt_append(void *, socklen_t, int, uint8_t, socklen_t, uint_t, void **);
– int inet6_opt_finish(void *,socklen_t, int);
– int inet6_opt_set_val(void *, int, void *, socklen_t);
– int inet6_opt_next(void *, socklen_t, int, uint8_t*, socklen_t *, void **);
– int inet6_opt_find(void *, socklen_t, int, uint8_t, socklen_t *, void **);
– int inet6_opt_get_val(void *, int, void *, socklen_t);
Version 5.3 modifies the behavior of AF_INET6 IPPROTO_RAW sockets to
conform to RFC 3542. Only packet data will be delivered on these sockets
when an application performs a recv call. The IPv6 header will no longer be
delivered along with incoming data.
Previous versions of AIX delivered IPv6 headers along with packet data on
AF_INET6 IPPROTO_RAW sockets. If an application wants to continue to
receive the IPv6 header along with the data, it must set socket option
IPV6_AIXRAWSOCKET and recompile.

Support for IPv6 over PPP
Version 5.3 now provides IPv6 functional support over PPP. Refer to AIX 5L
Version 5.3 documentation for further details.

7.6 Interface layer support for hotplug network adapters
Prior to AIX 5L Version 5.3, the number of network interfaces defaults to 256. It is
a configurable network option (ifsize). This determines the maximum number

304


of interface layer structures supportable for that interface type. The maximum
acceptable value for ifsize is 1024. With the introduction of virtualization
technologies, where there can be thousands of interfaces on a machine, each
requiring interface layer structure support, the limitation needs to be increased.
Starting with Version 5.3, any number of interfaces (real or virtual) can be
supported by implementing if_layer hotplug support. Since most of the current
focus of interface virtualization is on Ethernet, it introduces the hotplug features
for Ethernet only. The other interface types such as token-ring will still continue
to use the current (Version 5.2 and earlier) allocation strategies where the
network option ifsize (default 256, maximum size 1024) determines the maximum
number of supportable interfaces of that type at the interface layer. Table 7-6
contrasts the network option ifsize differences between releases.
Table 7-6 Differences of network option ifsize
AIX 5L Version 5.2 or earlier

AIX 5L Version 5.3

Network option ifsize defaults to 256 and
maximum of 1024.

No ifsize limitation for Ethernet.

You can still see network option ifsize with the no command. However, it does not
mean anything for Ethernet. The help text for no -h ifsize is shown in
Example 7-13; most of the explanation applies for token-ring as well.
Example 7-13 Network option ifsize
# no -o ifsize
ifsize = 256
#
# no -h ifsize
Help for tunable ifsize:
Specifies the maximum number of network interface structures per interface.
This limit does not apply for ethernet interface structures for which the
infrastructure expands dynamically to handle any number of ethernet interface
structures. So the rest of the message applies to other interface types. The
default value is 256. If the system detects at boot time that more adapters of
a type are present than would be allowed by the current value of ifsize, it
will automatically increase the value to support the number of adapters
present, so there is no reason to change it in runtime. ifsize is a Reboot
attribute. ifsize ranges from 8 to 1024.

It needs to be noted that the interface layer for hotplug network adapters grows
dynamically on an as needed basis for efficient resource allocation.


305

7.7 Support for SLP client service
Service Location Protocol Version 2 (SLPv2) provides a flexible and scalable
framework for providing hosts with access to information about the existence,
location and configuration of networked services. Any TCP/IP application can
take advantage of this protocol, which requires little or no static configuration to
discover the network services. SLP eliminates the need for a user to know the
name of a network host supporting a service.
For example, if a user wants to resolve a hostname-to-ipaddress mapping or
ipaddress-to-hostname mapping, the user has to know the DNS server IP
address first so that the resolver can be configured in /etc/resolv.conf file before
any name resolution can be performed properly. However, if the application is
using SLP protocol, the user only needs to supply the desired type of service and
a set of attributes which describe the service. Based on the description, the SLP
resolves the network address of the host which provides the service that the user
wants. SLP provides a dynamic configuration mechanism for applications in local
area networks. Applications are modeled as clients that need to find servers
attached to any of the available networks within an enterprise.
Version 5.3 supports the SLP client part of the APIs and configuration file. The
SLP client APIs are the following:
SLPOpen
SLPClose
SLPFree
SLPFindSrvs
SLPFindSrvTypes
SLPFindAttrs
SLPEscape
SLPUnescape
SLPParseSrvURL
SLPFindScopes
SLPGetProperty
SLPSrvTypeCallback
SLPSrvURLCallback
SLPAttrCallback
The /etc/slp.conf file is the configuration file for SLP; Table 7-7 on page 307
explains the parameters used in this file.

306


Table 7-7 Parameters for /etc/slp.conf
Parameters

Descriptions

net.slp.maxResults

A 32-bit integer that gives the maximum number
of results to accumulate and return for a
synchronous request before the time out. Positive
integers and -1 are legal values. A value of -1
indicates that all results should be returned. The
default value is -1.

net.slp.useScopes

A value-list of strings that indicate the only scopes
a UA or SA is allowed to use when making
requests or registering. Otherwise, indicates the
scopes a DA must support. The default scope
DEFAULT is used if no other information is
available.

net.slp.DAAddress

A value-list of IP addresses or DNS resolvable
host names giving the SLPv2 DAs to use for
statically configured UAs and SAs. The default is
none.

net.slp.isBroadcastOnly

A boolean indicating whether broadcast should be
used instead of multicast. The default is false (that
is, multicast is used).

net.slp.multicastTTL

A positive integer less than or equal to 255, giving
the multicast TTL. The default is 255 (in seconds).

net.slp.DAActiveDiscoveryInterval

A 16-bit positive integer giving the number of
seconds between DA active discovery queries.

net.slp.multicastMaximumWait

A 32-bit integer giving the maximum amount of
time to perform multicast, in milliseconds. This
property corresponds to the CONFIG_MC_MAX
parameter in the protocol specification. The
default is 15000 (in ms).

net.slp.multicastTimeouts

A value-list of 32-bit integers used as time outs, in
milliseconds, to implement the multicast
convergence algorithm.

net.slp.DADiscoveryTimeouts

milliseconds, to implement the multicast
convergence algorithm during active DA
discovery.

net.slp.datagramTimeouts

milliseconds, to implement unicast datagram
transmission to DAs.


307

7.8 Stream Control Transmission Protocol (SCTP)
The Stream Control Transmission Protocol (SCTP) is a new reliable transport
protocol. It was proposed to the Internet Engineering Task Force (IETF) as a
telephony signaling transport above the IP protocol layer.

7.8.1 SCTP basics
SCTP is a message-oriented transport layer protocol that can be compared to
the UDP or TCP protocols. In general, SCTP provides more flexibility for certain
applications than TCP or UDP. Figure 7-4 shows the network layers and where
the SCTP fits.

App3

App2

App1

Application

Sockets API
UDP

TCP
IP
MAC

SCTP

Transport
Internet
Link

Figure 7-4 The IP network protocol layers

The SCTP architecture is based on IETF RFC 2960 available at:
http://www.ietf.org/rfc/rfc2960.txt

The implementation of the sockets API extensions for SCTP are based on a
subset of the IETF draft draft-ietf-tsvwg-sctpsocket-04. At the time of writing, the
latest version of the draft is available at:
http://www.ietf.org/

There are other related RFCs that describe the User Adaptation Layers and
enhancements, such as RFC 3057, RFC 3331, RFC 3332, RFC 3309 or
RFC 3286.
SCTP is a connection-oriented protocol, similar to TCP, but provides
message-oriented data transfer, similar to UDP. For applications like Voice over
IP (VoIP) the SCTP protocol works out to be ideal because TCP with its inherent

308


properties like byte-oriented transfer and strict in-order delivery can prove to be
very slow, and UDP is too unreliable to be used directly without application
overhead to provide reliability. Table 7-8 highlights the major differences between
the three transport protocols.
Table 7-8 TCP, UDP, and SCTP comparison
TCP

UDP

SCTP

Reliability

Reliable

Unreliable

Reliable

Connection
management

Connection
oriented

Connectionless

Connection
oriented

Transmission

Byte oriented

Message oriented

Message oriented

Flow control

Yes

No

Yes

Congestion control

Yes

No

Yes

Fault tolerance

No

No

Yes

Data delivery

Strictly ordered

Unordered

Partially ordered

Security

Yes

Yes

Improved

7.8.2 SCTP API in AIX
AIX implements the SCTP at kernel level as a loadable kernel extension
accessible through SCTP socket APIs. The SCTP socket APIs were designed to
maintain consistency with existing socket APIs, like TCP or UDP, but at the same
time, provide a base for access to new SCTP features. The SCTP APIs provide
compatibility so that most existing TCP and UDP applications can be migrated to
SCTP with few changes.
At the time of writing, only UDP-style SCTP sockets for IPv4 are implemented in
Version 5.3.
The programming of an SCTP communication is very similar to classic UDP.
Before you start to implement SCTP sockets, you need to include the necessary
header files:
#include <sys/socket.h>
#include <netinet/sctp.h>

First you have to create the socket with type of SOCK_SEQPACKET and with
protocol of IPPROTO_SCTP. The socket() function creates an SCTP-type
socket and initializes the SCTP structures with default values. The socket has to
be created on the client and server sides, as in the following example:
so = socket(AF_INET, SOCK_SEQPACKET, IPPROTO_SCTP);


309

You can then use the bind(), listen(), connect(), sendmsg(), recvmsg(), and
close() system calls in the same way as for UDP protocol.
Since SCTP is a connection-oriented protocol, unlike UDP, multiple incoming
connections from clients to a single server are handled on a single server socket.
The socket buffer is, therefore, shared among all these connections. Since this
can be limiting for some applications, SCTP provides new APIs in addition to the
standard socket APIs to manipulate this. The two new APIs supported in this
release of AIX 5L Version 5.3 are:
sctp_peeloff()
sctp_opt_info()
The sctp_peeloff() call is used by applications to branch off an existing
association into a separate socket.
The sctp_opt_info() call is used by applications to pass information in or out of
the SCTP stack.
More information about the socket API for SCTP can be found in the SCTP
Socket API draft at the following URL:
http://www.ietf.org

7.8.3 SCTP management in AIX
Starting with AIX 5L Version 5.3, SCTP is delivered as the bos.net.sctp LPP
fileset. The fileset is not installed at initial system installation and you will need to
install it in order to use SCTP.
Once the SCTP is installed, you get the sctpctrl command that handles the
SCTP device driver and its environment. This command accepts the following
flags:
load

Loads and initializes the SCTP kernel extension.

unload

Unloads the SCTP kernel extension.

dump

Dumps various data structures used by the SCTP kernel extensions.
This flag is intended to be used by IBM service.

set

Sets attributes of the SCTP driver.

get

Displays specified or all SCTP attributes.

stats

Displays SCTP statistics.

Example 7-14 shows the procedure of loading the kernel SCTP extension using
the sctpctrl load command, listing the SCTP settings using the sctpctrl get
command, and unloading the SCTP kernel extension using the sctpctrl unload

310


command. The output of the stpctrl stats command gives detailed device
driver statistics and it is not displayed in the example.
Example 7-14 sctpctrl command
# sctpctrl load
# sctpctrl get
sctp_assoc_maxerr
sctp_cookie_life
sctp_delack_timer
sctp_dontdelayack
sctp_ecn
sctp_ephemeral_high
sctp_ephemeral_low
sctp_instreams
sctp_maxburst
sctp_outstreams
sctp_path_maxerr
sctp_pmtu_discover
sctp_rttmax
sctp_rttmin
sctp_recvspace
sctp_sendspace
sctp_send_fewsacks
# sctpctrl unload

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

10
60
4
1
1
65535
32768
2048
8
10
5
1
60
1
65536
65536
0

The sctpctrl command is not persistent across reboots and you may need to
call it at the system startup, for example from the /etc/inittab or the /etc/rc.tcpip
script.

7.9 Network RAS enhancements
In order to improve network RAS, the following features are included in AIX 5L
Version 5.3:
For better display of the internal structures: KDB subcommands tcb, udb,
tcpcb, sock, sockinfo, netm, ndd and ifnet are modified.
For improving network memory debug: KDB subcommands netm and
kmbucket have new options.
For ns_alloc/free event tracing:
New options are added in the no command to turn on/off and configure filters.
(type of buffer, size of buffer).
New kdb subcommand nsdbg is introduced to display events.


311

For checking network options based on their dependence: new options are
added in the no command.

KDB subcommands enhancements
The following KDB subcommands are modified to produce better display or
improve network memory debug. The options in bold characters are new or
modified in Version 5.3.
tcb [-s ]
udb

[-s ]

|

[ -b <bucket_index> ] [ symbol | address ]

|

[ -b <bucket_index> ] [ symbol | address ]

tcpcb

[ tcp | udp ] [ symbol | address ]

sock

[ -d ] [ tcp | udp ] [ symbol | address ]

sock

[ -s ]

sock

[ -f ]

[ -s ]

[ -s [ udp | tcp ] ]

[ tcp | udp ]

sockinfo [ -d ] [ address ]
ndd

|

|

[ <type_of_address> ]

[ -n nddname]

netm [ [-c display_count] [-t type[,type]*] [-s size[,size]*] ]
kmbucket -k <bucket_address>

New kdb nsdbg subcommand
The nsdbg subcommand displays the ns_alloc and free event records stored in
the kernel. This function is only available if the ndd_event_tracing parameter is
turned on using the no command.
#nsdbg [-i starting_index] [-c display_count] [-n nddname[,nddname[,...]] ]

Example 7-15 shows the output of the nsdbg subcommand.
Example 7-15 The output of nsdbg subcommand
# no -o ndd_event_tracing=1
Setting ndd_event_tracing to 1
#In order to create some events for nsdbg, do the following
#ifconfig en0 down
#ifconfig en0 up
#kdb
(0)> nsdbg
ndd_name=ent0, ndd_addr=F10006000C662030
index =000503, type =Free , cpu=0000
refcnt=000000, flags=00000000
Stack Trace back
.soclose2+000318

312


.soclose+000014
.soo_close+0000C8
.closef+00006C
.closefd+0000F8
.close+000284
.sys_call_ret+000000
ndd_name=ent0, ndd_addr=F10006000C662030
index =000502, type =Alloc, cpu=0000
refcnt=000000, flags=00000000
Stack Trace back
.soconnect+000358
._connect+000260
.connext+000018
.sys_call_ret+000000
(0)>

New options for the no command
For network RAS support, the network options described in Table 7-9 are added
to Version 5.3.
Table 7-9 New network options added in Version 5.3
no option

Description

ndd_event_tracing

Turn on/off ns_alloc/free event tracing and define event buffer
size for ns_alloc/free event. (similar to net_malloc_police)

ndd_event_name

Limit ns_alloc/free event tracing to interface name/alias
containing one element of the string list. If the string is all or
empty, events for all interfaces will be captured. (Up to 8 different
names in a list.)

net_buf_size

Limit net_malloc/free event tracing for the size list. A size of 0
means all.

net_buf_type

Limit net_malloc/free event tracing for type contained in this
string list. If the string is all or empty, events for all type will be
captured.

7.10 Enable IP Security with intervening NAT devices
In order to establish a secure communication network, IP Security is widely
chosen in TCP/IP network. In previous versions of AIX, IP Security has been
supported without support for NAT. However, Network Address Translation (NAT)
protocol is also widely used in Internet Connection sharing and ships as a
standard feature in every router and edge device. The IP Security protocol


313

depends on identifying remote endpoints and their policy based on the remote IP
address. When intermediate devices such as routers and firewalls translate a
private address to a public address, the required authentication processing in IP
Security might fail because the address in the IP packet has been modified after
the authentication digest was calculated. With the new IP Security NAT support,
devices that are configured behind a node that performs network address
translation are able to establish an IP Security Tunnel. The IP Security code is
able to detect when a remote address has been translated. Using the new IP
Security implementation with support for NAT allows a VPN client to connect
from home or on the road to the office through an Internet connection with NAT
enabled. Figure 7-5 illustrates the difference between a NAT-enabled IP Security
implementation, with UDP encapsulated traffic, and an implementation that is not
NAT-enabled.

Private
Network

IPsec traffic

Fail

VPN Security
Gateway

NAT Router

Private
Network

UDP Encapsulated
IPsec traffic

Pass

VPN Security
Gateway

NAT Router

Figure 7-5 NAT-enabled IP security

NAT devices support for IP Security network
Starting with AIX 5L Version 5.3, NAT devices support for IP Security is now
available. In order to use the NAT support in IP Security, you need to set the
ENABLE_IPSEC_NAT_TRAVERSAL variable in the /etc/isakmpd.conf file as
shown in Example 7-16.
Example 7-16 /etc/isakmpd.conf used or NAT support
~
#
#
#

314

MULTI_LOGIN_OK


# 7) Enable IPsec NAT Traversal
#
You can enable IPsec NAT Traveral and UDP encapsulation functionality
#
by specifying following text:
ENABLE_IPSEC_NAT_TRAVERSAL
#
# 8) NAT Keep Alive Interval
#
#
The syntax for this option is:
#
#
NAT_KEEPALIVE_INTERVAL = keepalive interval value (seconds)
#
#
The NAT_KEEPALIVE_INTERVAL is optional. The default value is 20 seconds.
#
NAT_KEEPALIVE_INTERVAL=30
#

When this variable is set, filter rules are added to send and receive traffic on port
4500. Example 7-17 shows the filter rules when the
ENABLE_IPSEC_NAT_TRAVERSAL variable is set.
Example 7-17 Filter rules added with ENABLE_IPSEC_NAT_TRAVERSAL
Dynamic rule 2:
Rule action
Source Address
Source Mask
Destination Address
Destination Mask
Source Routing
Protocol
Source Port
Destination Port
Scope
Direction
Fragment control
Tunnel ID number

:
:
:
:
:
:
:
:
:
:
:
:
:

permit
0.0.0.0 (any)
0.0.0.0 (any)
0.0.0.0 (any)
0.0.0.0 (any)
no
udp
0 (any)
4500
local
inbound
all packets
0

Dynamic rule 3:
Rule action
Source Address
Source Mask
Destination Address
Destination Mask
Source Routing
Protocol
Source Port
Destination Port
Scope
Direction

:
:
:
:
:
:
:
:
:
:
:

permit
0.0.0.0 (any)
0.0.0.0 (any)
0.0.0.0 (any)
0.0.0.0 (any)
no
udp
4500
0 (any)
local
outbound


315

Fragment control
Tunnel ID number

: all packets
: 0

Setting the ENABLE_IPSEC_NAT_TRAVERSAL variable also adds some
additional filter rules in the filter table. Special IPSEC NAT messages use UDP
encapsulation and filter rules must be added to allow this traffic to flow. In
addition, in phase 1 signature mode is required. If IP Address is used as the
identifier in the certificate, it should contain the private IP address. IP Security
also needs to send NAT keep alive messages to maintain the mapping of the
original IP Address and the NAT address. The interval is specified by the
NAT_KEEPALIVE_INTERVAL variable in /etc/isakmpd.conf file. This variable
specifies how frequently, in seconds, NAT keepalive packets are sent. If you do
not specify a value for NAT_KEEPALIVE_INTERVAL, a default value of 20
seconds is used.

Limitations when using NAT exchanges
NAT exchanges supports ESP only. AH (Authentication Header) and ESP
(Encapsulating Security Payload) can be used to implement IP security.
However, AH does not a change in IP address in IP packet which is necessary
for NAT. Additionally, a connection using NAT must select tunnel mode so that
the original IP address is encapsulated in the packet. Transport mode and
addresses with NAT are not compatible.
Refer to AIX 5L Version 5.3 Documentation Security Guide for further detail.

7.11 AIX SNMP subagent enhancements
SNMP manager, agent, and subagent are the primary functional entities. SNMP
manager is the system management application, such as IBM Director, IBM
Netview and other network management systems (NMS). SNMP agent accepts
the requests from the manager, verifies the authentication, passes the requests
to the appropriate SNMP subagent, and responds to the manager with the data
or execution result from the subagent. SNMP subagent executes the specific
system management tasks. AIX Enterprise SNMP MIB Support, SNMP subagent
aixmibd is first shipped with AIX 5L Version 5.2. In Version 5.2, about half of the
MIB objects defined in the MIB were implemented, while all are now
implemented in Version 5.3.

7.11.1 MIB objects added in Version 5.3
SNMP subagent aixmibd provides the system management data, and manages
the system resources through the standard SNMP. It also generates traps for

316


some critical system resources. It manages and provides data for the system
resources shown in Table 7-10.
Table 7-10 New MIB resources implemented in Version 5.3
Supported resources

Detail

System environment
information

Number of CPUs, machine type, machine serial number,
system date and time, maximum number of processes
allowed per user, maximum number of fixed licenses,
number of available fixed licenses

Substystem

Name, PID, status

Subserver

Name, description, status, the subsystem name which the
subserver belongs to

print queue

Name, device, status, start/stop queue, discipline,
accounting data file, host name where the queue is
running, remote queue name

system user

Name, UID, home directory, shell, primary group, groups
which the user belongs to, enable or disable local login,
enable/disable remote login, allowable attempts,
password maximum age, status, reset login count

System group

Name, GID, admin group or not, admin users for this
group, the users who are in this group

Printer device

Name, type, interface, status, description, location, port
number

Tape drive

Name, type, interface, status, description, location

Hard disk

Name, type, interface, status, description, location, size

Memory

Name, type, description, location, size

CDROM

Name, type, description, location, interface

SCSI

Name, description, location, status, adapter ID

Processor

Name, type, description, speed

Network interface

Name, type, interface, status, location, description

Adapter

Name, description, type, interface, status

You might want to take a look at the MIB resources already supported with
Version 5.2. Version 5.3 continues to support these resources, which are
identified in Table 7-11 on page 318.


317

Table 7-11 MIB resources implemented in version 5.2
Supported resources

Detail

System environment
information

CPU utilization, CPU utilization threshold, system run
level, system state, latest trap message

Agent control variables

Polling intervals, trap thresholds, polling
enable/disable, agent access

File systems

Name, mount point, total size, free size, total inodes,
free inodes, type (NFS, jfs, jfs2)

Volume group

Name, identifier, state, total size, free size, number of
total logical volumes, number of open logical
volumes, number of active physical volumes

Logical volume

Name, the volume group name which the logical
volume resides in, type, label, mount point, size, state

Paging space

Name, volume group name, physical volume name,
size, percent used, status, type

Process

Process PID, UID, PPID, group name, priority,
command, status, tty

Current logged in user

Name, tty, the host name from which the user got
logged in, date and time

Important resource monitoring
traps

File system full trap, CPU utilization high trap, volume
group full trap, paging space full trap, too many
attempts to get logged in trap

7.11.2 The snmptrap command enhancement
Starting with Version 5.3, you have a new option with snmptrap command:
snmptrap -h. The -a flag specifies a host where the AIX SNMP agent (snmp)
must be running and the SNMP agent forwards this trap to network mangers.
However, the -h flag does not require the AIX SNMP agent to forward the trap
message to network managers, and it sends the trap directly to the network
manager. If there are no -h and -a flags, the trap will be sent to the AIX SNMP
agent on the local host.
snmptrap

[-d] [-a hostname|-h dest] [-c community] -m message

-d
-a hostname

Connect to the snmp agent running on the specified AIX
host.

-h dest

318

Enable the debug facility.

Any destination hostname on which the trap will be
received. This option makes it possible to send the


message to any platform. Either -a hostname or -h dest
can appear in the command line, but not both.
-c community

Use specified community name.

-m message

Define the message sent by the snmptrap command. the
-m option must be the last option.

If you want to send a trap to the network manager running on a Linux platform
and where the host name is myahn, use the following command:
# snmptrap -h myahn -m hello world

7.12 SMBFS enhancements
Server Message Block File system (SMBFS) allows access to shares on SMB
servers as local file systems on AIX. In this file system, the user can create,
delete, read, write, and modify the access times of files and directories. The
owner or access mode of files and directories cannot be changed. SMBFS can
be used to access files on an SMB server. The SMB server is a server running
Samba; an AIX server running AIX Fast Connect; or a Windows® XP, Windows
NT®, or Windows 2000 server or workstation. Each of these server types allows
a directory to be exported as a share. This share can then be mounted on an AIX
system using SMBFS. In AIX 5L Version 5.3, SMBs following newer protocol are
used for file open, allowing the AIX file attributes to be mapped to a larger set of
Windows file attributes. To install SMBFS on an AIX system, install the
bos.cifs_fs package from AIX install CD-ROMs (Example 7-18). When the
bos.cifs_fs package in installed, the device nsmb0 is created. This device allows
the mount command to establish a connection between the SMB server and the
client.
Example 7-18 Configuring SMB clients
# lslpp -L |grep cifs
bos.cifs_fs.rte
bos.cifs_fs.smit
# lsdev -C|grep nsmb
nsmb0
Available

5.3.0.0
5.3.0.0

C
C

F
F

Runtime for SMBFS
SMIT Interface for SMBFS

N/A

7.13 Ported 64-bit DES NFS kernel extension
The DES encryption kernel library used for the secure NFS is enabled for the
64-bit environment. This library is part of the DES LPP that is part of the AIX 5L
Version 5.3 Expansion pack.


319

7.14 BIND domain search improvement
Before a user can use a nameserver to map hostname to IP address, and vice
versa, on the client side the user must configure the /etc/resolv.conf file. In this
file, the user specifies the nameserver by configuring sets of nameserver
IP_address values for each nameserver he wants to use. The user also specifies
how many domain names to use in the DNS queries by configuring the search
DomainName ... statement. A typical file looks like the following:
nameserver 9.53.159.2
nameserver 9.3.149.2
search
ibm.com
austin.ibm.com

aoot.austin.ibm.com

A user can specify up to three nameserver lines in the /etc/resolv.conf file. In
cases where there is only one nameserver, it opens a UDP socket, sends out
query, receives the response, and then closes the socket. This is the sequence
that takes place for every DNS query that the client makes. This can be very
inefficient in cases where there are multiple queries.
Moreover, a user can specify a maximum of six domain names on the search
line. This does not always meet user needs.
AIX 5L Version 5.3 has made enhancements in both these areas.
The DNS resolver uses either TCP socket or UDP socket for the hostname and
IP address resolution. In any user application, if there is no sethostent(1) call
which explicitly asks the resolver to use TCP socket, by default, the resolver will
use UDP connection. There are cases when the resolver may send out multiple
queries using UDP connection before it can resolve the name. For each query, it
will do socket(), sendto(), poll(), recvfrom(), and close() in sequence. For multiple
queries, there will be multiple socket() and close() calls. This will cause
performance degradation.
In Version 5.3 the UDP socket is kept open for multiple queries against the same
nameserver until it finishes the name resolution, then calls close() once to close
the UDP socket. In general this results in performance improvement.
In cases where the answer returned from the nameserver is truncated, the UDP
socket is closed and a TCP socket is opened to send out the query to the same
nameserver.
As for the domain names, the limit of six on the search keyword has been
replaced by the maximum number of characters in the search string to be 1024.
The search keyword now can support any number of domain names as long as
the number of characters do not exceed 1024.

320


7.15 Network API support for ARP
AIX 5L Version 5.3 provides new APIs for ARP: arpresolve_common and
arpupdate. Prior to Version 5.3, whenever you change the arptab structure, you
will break the interface (if) layer code for all the devices using ARP. Version 5.3
now provides a set of APIs to access the ARP table to avoid breaking the if code.

The arpresolve_common function
The arpresolve_common function is called by arpresolve from the if layer of the
interface. It does the following, with the parameters provided in Table 7-12:
If the arp entry is complete, then it will return the hwaddr and if_dependent
data if the if_dependent is true.
If entry is not complete, then add the mbuf to at->at_hold.
If entry does not exist, then create a new entry by calling arptnew and adding
the mbuf to at->at_hold.
The arpresolve_common calls arpwhohas when it creates a new arp entry or
when the timer for the incomplete arp entry (whose IP address is as per
pointer dst) has expired.
int arpresolve_common ( register struct arpcom *ac, struct mbuf *m, int
(*arpwhohas)(register struct arpcom *ac, struct in_addr *addr,
int skipbestif, void *extra), struct sockaddr_in *dst, u_char * hwaddr,
int szhwaddr, void *extra, union if_dependent *if_dependent)
Table 7-12 Parameters for arpresolve_common
Parameter

Description

ac

Pointer to the arpcom structure.

m

Pointer to the mbuf which will be added to the list awaiting the arp
table entry completion.

dst

Pointer to the sockaddr_in structure this structure has the destination
IP address.

int
(*arpwhohas)
(....)

Function pointer to arpwhohas.

hwaddr

Pointer to the buffer. This buffer is filled with the hardware address if
it finds a completed entry in the arp table for the destination IP
address.

szhwaddr

Size of buffer hwaddr.


321

extra

A void pointer which can be used in future for IF layer to pass extra
structures to arpwhohas.

if_dependent

Pointer to the if_dependent structure. The function
arpresolve_common uses this to pass the if_dependent data which
is part of arptab entry to the calling function.

The arpupdate function
The arpupdate function is called by arpinput functions from the if layer. The
function does the following, with the parameters provided in Table 7-13:
Looks at the arp table for h/w address for the given IP address.
Updates the arp entry for a given IP address.
Does reverse arp lookup.
int arpupdate (ac, m, hp, action, prm)
struct arpcom *ac;
struct mbuf *m;
int action;
struct arpupdate_parm
*prm;
Table 7-13 Parameters for arpupdate
Parameter
ac

Pointer to the arpcom structure.

m

Pointer to the mbuf which contains the arp response packet which
was received by the interface.

hp

Pointer to buffer.

action -

LOOK (1)Look for the IP address [isaddr] in the arp table and return
h/w address & if_dependent structure.
LKPUB (2) - Look for the IP address [isaddr] in the arp table and
return h/w address & if_dependent structure only if flag set to
ATF_PUBL.
UPDT (3) - update the arp entry for an IP address [isaddr] or if no arp
entry is there then create a new one and also update the
if_dependent structure using function ptr passed in prm structure.
REVARP (4) - Reverse arp request. hwaddr contains the h/w
address, szhwaddr is size and the IP address will returned in
saddr if an entry is found.

prm

322

Description

Pointer to structure arpudpate_parm


7.16 Withdraw older versions of TCP applications
Starting with Version 5.3, the following services will not be supported:
The previous versions of AIX support BIND Versions 4, 8 and 9. With Version
5.3, AIX does not support BIND Version 4 and only supports BIND Versions 8
and 9.
TCPIP LDAP name resolution that used the.IBM SecureWay® Directory
schema will no longer be supported in 5.3. In 5.3, TCPIP LDAP name
resolution uses only RFC 2307 schema. See "TCP/IP Name Resolution" in
the System Management Guide: Communications and Networks for detailed
information.


323

324


8

Chapter 8.

Security
This chapter is dedicated to the latest security topics as they apply to AIX 5L.
Topics include:
LDAP user authentication enhancements
Full PAM exploitation
Prioritizing LDAP servers
Rsh login control
chpasswd command
Initial login license limit increased
NIS/NIS+ enhancements
Support for passwd.adjunct NIS map


325

8.1 LDAP user authentication enhancements
The LDAP authentication package in AIX 5L Version 5.2 authenticates users by
retrieving the user password from the LDAP server database and doing
authentication locally. In order to retrieve a user password from the LDAP server,
each client needs to have privileged access to the server. This privilege is also
used by client systems for password change or user login activity logging. The
privileged access is made available by storing the LDAP server privileged DN
and password on each client system. In some environments, it may be a concern
to allow any local root user of a configured client system to have privileged
access to the LDAP database. AIX LDAP package in Version 5.3 now provides a
mechanism to eliminate the need for privileged access rights from client systems
for better security.

UNIX-type and LDAP-type authentication
The AIX LDAP authentication package was first implemented in AIX 4.3. In order
to keep the same AIX user authentication and management behavior in the
LDAP environment as using flat ASCII files (/etc/passwd, for example), client
systems connect to the server using a privileged account (LDAP server
administrator account). Such privileged connection serves the purpose of
maintaining the AIX user management behavior, such as user creation and
removal, password change, user login log, and so on. The only trade off is
security: Each local root user of a client system has full control of the LDAP
server database. This could be a major concern for LDAP deployment, especially
if multiple clients are involved. An alternative user authentication mechanism is
introduced in Version 5.3: LDAP server authentication, or simply LDAP-type
authentication. LDAP-type authentication does not require privileged
connections between client systems and the LDAP server. LDAP-type
authentication involves constructing a distinguished name (DN) from the user
account name, sending the constructed DN and the login password to the LDAP
server, and requesting that the server determine if the user password matches
the one stored in the server database.
Table 8-1 Difference between UNIX-type and LDAP-type authentication
UNIX-type authentication
Authentication mechanism of fetching
user's encrypted password and
authenticating user locally.

326

LDAP-type authentication
Authentication mechanism of sending
user-entered password and constructed
DN (from user name) to the LDAP server
and having the server do the
authentication.


Advantages of LDAP-type authentication
There are two advantages if you choose LDAP-type authentication:
A more secure LDAP database. No privileged connection between server and
client is required. It can be configured such that the local root cannot change
password, create, delete, or modify user accounts. The current attribute-level
access control feature of IBM Directory server (Version 4.1 and later)
supports fine grained configuration with respect to what can be done by a
local root user to the LDAP database.
Multiple encryption mechanism support by the LDAP server. Prior to Version
5.3, only support for the crypt mechanism for user password encryption
existed. With LDAP-type authentication, password encryption is done by the
LDAP server, and the password encryption is hidden from users. The IBM
Directory currently supports imask, crypt, sha, as well as clear text password.
Note: The password is sent in clear text over the wire to the LDAP server for
LDAP-type authentication. You may need to configure SSL when using
LDAP-type authentication, if you are concerned about passwords being sent
in clear text form.

Configuration file
Example 8-1 shows new controls added to the /etc/security/ldap/ldap.cfg file. The
explanation for each of the controls is also added to this file. The old credential
definition of ldapadmindn and ldapadmpwd will be removed from the file.
However, the secldapclntd will keep its ability to read ldapadmindn and
ldapadmpwd for backward compatibility. In case both binddn and ldapadmindn
exist in the ldap.cfg file, secldapclntd will use ldapadmindn first.
Example 8-1 New controls added in /etc/security/ldap/ldap.cfg
# LDAP server bind DN.
#binddn: cn=admin
# LDAP server bind DN password
#bindpwd:
# Authentication type. Valid values are unix_auth and ldap_auth. Default is
unix_auth.
# unix_auth - retrieve user password and authenticate user locally.
# ldap_auth - bind to ldap server to authenticate user remotely by the LDAP
server.
#authtype: unix_auth
# Search mode. Valid values are ALL and OS. Default is ALL.
# ALL - returns all attributes of an entry.
# OS - returns only the OS required attributes of an entry.
#
Non-OS attributes like telephone number, binary images,
#
will not be returned.

Chapter 8. Security

327

# Note: use OS only when user entry has many non-OS required
# attributes or attributes with large values, e.g., binary data,
# to reduce sorting effort by the LDAP server.
#
#searchmode: ALL
# Whether to use Kerberos for initial bind to the server.
# useKRB5
- valid value is either "yes" or "no". Default is "no".
# krbprincipal - Kerberos principal used to bind to the server.
# krbkeypath
- Path to the kerberos keytab, default to
#
/etc/security/ldap/krb5.keytab
# krbcmddir
- Directory that contains the Kerberos kinit command.
#
Default to /usr/krb5/bin/.
#useKRB5:no
#krbprincipal:principal
#krbkeypath:/etc/security/ldap/krb5.keytab
#krbcmddir:/usr/krb5/bin/
#
# Default-entry location. Valid values are "ldap" and "local".
# The default is "ldap".
# ldap - use the default entry in LDAP.
# local - use the default entry from /etc/security/user.
#defaultentrylocation: ldap
# Timeout period (seconds) for LDAP client requests to the server. This value
# determines how long the client will wait for a response from the LDAP server.
# Valid range is 0 - 3600 (1 hour). Default is 60 seconds. Set this
# value to 0 to disable the timeout.
#ldaptimeout:60

Proxyuser
This is the LDAP user account, the alternative of the LDAP administrator
account, that a client system uses to bind to the LDAP server. The access right
of this account to the LDAP server determines the behavior of the client systems.
The mksecldap command will support the creation of this account at directory
server setup. By default, this proxy user just has limited permissions and needs
to be modified if necessary. The default ACL that is set for the proxy user during
server setup can be configured by modifying the entries in
/etc/security/ldap/proxy.ldif.template. The following shows new options to
mksecldap command added in Version 5.3.
For server setup (New options: -x, -X):
#mksecldap -s -a adminDN -p adminpasswd -S schematype [ -d baseDN ] [ -n
port ] [ -k SSLkeypath] [ -w SSLkeypasswd ] [ -u NONE ] [ -U ] [ -x
proxyuserDN ] [ -X proxyuserPasswd ]

328


For client setup (New options : -A, -D, -m):
#mksecldap -c -h serverlist -a bindDN -p bindPasswd [ -A AUTHTYPE ] [ -d
baseDN ] [ -n serverPort ] [ -k SSLKeyPath ] [ -w SSLKeyPasswd ] [ -t
cacheTimeout ] [ -C cachesize ] [ -P NumberofThreads ] [ -T heartBeatInt ]
[ -u userlist ] [ -U ] [ -m searchMode ] [ -D defaultEntryLocation ]

The secldifconv command was added in Version 5.3 to provide a utility to
convert user and group entries between the schema types (AIX, RFC2307,
RFC2307AIX) supported by AIX. Example 8-2 shows how to generate output
from the LDAP database and convert the data to RFC2307 schema.
#secldifconv -S schematype -i infile [ -r ]
Example 8-2 Converting AIX schema to RFC2307
#To generate the input file for secldifconv
#db2ldif -o ldif.out
#To convert to RFC2307 schema type
#secldifconv -S RFC2307 -i ldif.out > newldif.out
#The converted data could then be added to an LDAP server through the following
command.
#ldif2db -i newldif.out

8.2 Full PAM exploitation
Over the recent past the Pluggable Authentication Module (PAM) framework
became a standard method of authentication in the industry. PAM provides a
means of separating authentication technologies from authentication services,
allowing different services to potentially follow independent authentication paths.
The library /usr/lib/libpam.a was provided in AIX 5L Version 5.1 to allow PAM
applications the ability to make use of the PAM framework. This solution did not
allow existing AIX security services to utilize the PAM framework. Further PAM
function was integrated into AIX 5L Version 5.2 through the use of the existing
AIX loadable authentication module (LAM) scheme and creation of a PAM AIX
LAM module /usr/lib/security/PAM. Users defined to use the PAM AIX LAM
module for their registry could then be routed to PAM for authentication,
effectively PAM-enabling all existing AIX security services. The module
/usr/lib/security/pam_aix was also provided to allow existing PAM applications
access to AIX security services through the PAM interface.

Chapter 8. Security

329

The enhancements to the AIX 5L Version 5.3 PAM implementation are focused
on PAM application compatibility, better AIX base operating system integration,
and improved configuration support.
The PAM implementation in AIX 5L Version 5.3 features the following:
System-wide specification to use the PAM library for authentication.
Default configuration file /etc/pam.conf shipped with AIX.
PAM enablement of native AIX commands by direct calls to the PAM library.
Additional PAM modules provided in /usr/lib/security.
Additional control option and new APIs for the PAM library.
Comprehensive documentation in the AIX standard publications covering the
PAM application programming interface (PAM API), the PAM service
programming interface (PAM SPI), all PAM modules, and PAM-enabled
applications.
The AIX 5L Version 5.3 PAM implementation adheres to the PAM specification
as defined in the Open Group standard X/Open Single Sign-on Service (XSSO)
and in RFC 86.0.

8.2.1 AIX authentication through PAM library
In AIX 5L prior to Version 5.3 there are two separate paths of PAM integration.
One path uses the typical PAM behavior of an application invoking a PAM library
call and then correspondingly calling configured modules in /etc/pam.conf. The
related process flow is similar to the following:
Application → PAM library → PAM module
The second path is unique to AIX and provides existing applications the ability to
make use of PAM with minimal modification through use of the PAM AIX LAM
module /usr/lib/security/PAM. This path requires users to be configured to utilize
the PAM compound module database (PAMFiles) for their registry and SYSTEM
attributes, with the additional restriction that the module used by the service
could not be the pam_aix module to avoid loop conditions. The related process
flow is similar to the following:
Application → AIX Security Library →
→ PAM AIX loadable authentication module → PAM library → PAM module
AIX 5L Version 5.3 obtains greater PAM compatibility and eliminates the need for
the second path of integration by calling the PAM library directly from AIX

330


authentication applications rather than going through the AIX loadable
authentication module framework.
This function requires the configuration of the new system-wide attribute named
auth_type which determines whether to call the PAM or AIX security library for
user authentication tasks. This attribute is configured in the usw stanza of the
/etc/security/login.cfg file and has the following characteristics:
auth_type: Determines whether PAM or the standard AIX authentication
mechanism will be used. The two values to which auth_type can be set are:
– PAM_AUTH - Use the PAM library for user authenticate tasks as
configured by the /etc/pam.conf file.
– STD_AUTH - Use the standard AIX security library for user authentication
tasks. This is the default value.
System administrator can use the chsec command to enable PAM authentication
for all users system wide:
# chsec -f /etc/security/login.cfg -s usw -a auth_type=PAM_AUTH
Setting auth_type = PAM_AUTH will configure PAM-enabled commands to
invoke the PAM library functions directly for authentication rather than use the
standard AIX authentication mechanism. This configuration is a run-time
decision and does not require a reboot of the system to take affect. Note, users
will no longer need their SYSTEM and registry attributes set to use PAM
because the configuration is now a system setting.
The lssec command will show the current setting for the auth_type parameter:
# lssec -f /etc/security/login.cfg -s usw -a auth_type
usw auth_type=PAM_AUTH

The following native AIX commands and applications have been changed to
recognize the auth_type attribute and to directly call the PAM library if the system
is configured to use PAM for authentication:
login
rlogind
telnetd
su
passwd
ftpd
rexecd
rshd

Chapter 8. Security

331

snappd
imapd
Conversation functions were also added to each modified command and
application in order to handle prompting and messaging.
In order to enable the PAM exploitation, the PAM stacks for all PAM-enabled
services have to be configured in the /etc/pam.conf file. AIX 5L Version 5.3
delivers a pre-configured default PAM configuration file which defines the PAM
stacks for all previously listed PAM-enabled applications and commands to
closely resemble the behavior of the standard AIX authentication mechanism.
The default /etc/pam.conf file on an AIX 5L Version 5.3 system lists the PAM
stack definitions by module type. By default, PAM-enabled services not explicitly
defined will fail authentication because the OTHER service name entry at the
end of each module type stack is configured to use the
/usr/lib/security/pam_prohibit module.
The following sections show the PAM stack configurations consolidated by the
service name:

login service
login
login
login
login

auth
account
password
session

required
required
required
required

/usr/lib/security/pam_aix

rlogin service
rlogin
rlogin
rlogin
rlogin
rlogin

auth
auth
account
password
session

sufficient
required
required
required
required

/usr/lib/security/pam_rhosts_auth

required
required
required
required


telnet service
telnet
telnet
telnet
telnet

auth
account
password
session

su service
su
su
su
su
su
su

332

auth
auth
account
account
password
session

sufficient
required
sufficient
required
required
required

/usr/lib/security/pam_allowroot
/usr/lib/security/pam_allowroot


passwd service
passwd password

required


rsh service
rsh auth
rsh account
rsh session

required
required
required

/usr/lib/security/pam_rhosts_auth

rexec service
rexec auth
rexec account
rexec session

required
required
required


ftp service
ftp auth
ftp account
ftp session

required
required
required


snapp service
snapp auth
required
snapp session required


imap service
imap auth
required
imap session required


Stacking is implemented in the configuration file by creating multiple entries for a
service with the same module type field. The modules are invoked in the order in
which they are listed in the file for a given service, with the final result determined
by the control flag field specified for each entry. Valid values for the control flag
field are: required, requisite, sufficient, and optional. The requisite control option
for the PAM library was introduced with AIX 5L Version 5.3 and the
corresponding behavior in the stack is as follows:
requisite

If a requisite module fails, the PAM framework immediately returns to
the application with the error returned by the requisite module and
stops processing the module stack. If all requisite and required
modules in the stack succeed, then success is returned. If none of
the service modules in the stack are designated as required or
requisite, then the PAM framework requires that at least one optional
or sufficient module succeed.

Most PAM modules accept a debug option to be specified in the module options
field of the related entry in the /etc/pam.conf file. Consult the module
specification in the AIX product documentation for valid options. Debug syslog
messages from the PAM library will be logged as soon as the system

Chapter 8. Security

333

administrator creates an empty /etc/pam_debug file (touch /etc/pam_debug
command), configured /etc/syslog.conf file to contain the following line:
auth.debug

/tmp/syslog_auth.log

and refreshes the syslogd daemon configuration (refresh -s syslogd
command). You can specify any other syslog output file name in /etc/syslog.conf
as desired.
Note: The AIX loadable authentication module for PAM
/etc/security/security/PAM is no longer the recommended means to provide
PAM authentication to native AIX applications. System administrators should
use the system-wide configuration parameter auth_type=PAM_AUTH in
/etc/security/login.cfg to configure authentication applications to directly
access the PAM library for user authentication tasks. This configuration
provides full PAM compliancy, enhanced module compatibility, and improved
configuration support.

8.2.2 Additional PAM modules
In previous AIX releases, only the pam_aix PAM module was shipped to support
the configuration of PAM services. The pam_aix module provides PAM-enabled
applications access to AIX security services by providing interfaces that call the
equivalent AIX services where they exist.
AIX 5L Version 5.3 adds six new PAM modules in the /usr/lib/security/ directory
to simplify PAM configuration and setup:
pam_allow

Always returns PAM_SUCCESS

pam_allowroot

Returns success if the user ID (UID) is 0 (root user)

pam_ckfile

Performs a file check, similar to that done for /etc/nologin

pam_permission

Checks the user name or group name against a control
list in a file, similar to the /etc/ftpusers check

pam_prohibit

Always returns a PAM failure code

pam_rhost_auth

Performs rhosts authentication

The modifications and additions made in AIX 5L Version 5.3 allow the pam_aix
module to be used by default on the system, thereby providing a fully integrated
and pre-configured PAM solution.
The following sections give a brief description of each of the new PAM modules
and also provide some information about the pam_aix module for completeness.

334


pam_aix module
The pam_aix module allows the authentication path to be routed back to the AIX
security services from PAM through use of the AIX security library routines. This
allows PAM to closely mimic the behavior of many existing applications as well
as allowing new applications to use AIX authentication. This module supports all
user types (except for PAMfiles) including: local, NIS, NIS+, DCE, and LDAP
users. Users configured to use the PAM AIX LAM module (registry=PAMfiles)
supplied in Version 5.2 will fail authentication when using pam_aix due to loop
detection performed by the /usr/lib/security/PAM module. pam_aix accepts the
debug and nowarn options for all SPIs, and additionally accepts try_first_pass
and use_first_pass for the pam_sm_authenticate SPI.
AIX 5L Version 5.3 adds one new option use_new_state to the pam_aix module:
use_new_state

AIX builds and maintains state information when
authenticating a user. By default, the pam_aix module will
use the same state information throughout a PAM session.
This can produce results that are correct in terms of AIX
authentication but are unexpected within the PAM
framework. For example, pam_authenticate requests may
fail due to access restrictions. If this behavior is not desired
for a given module type, specify the use_new_state option to
use new state information for each invocation.

Typical usage for the pam_aix module is as a backup, or OTHER service. This
way, if a specific authentication stack is not defined for a service, local AIX
authentication is used.

pam_allow module
This module returns PAM_SUCCESS for all invocations. Support for
authentication, account management, password, and session module types is
provided. Use of this module is intended for administrator testing and for
disabling the checks performed by a given PAM module type. This module only
accepts debug and nowarn as specified options in the configuration file.
This module should be used with caution and should often only be used for PAM
debugging purposes. Placing this module in the PAM stack for a service could
potentially grant access to all users.

pam_allowroot
The pam_allowroot module checks the real user ID (UID) under which the PAM
application was run. If the UID of the authenticating user is 0 (zero), then it is the
root user and PAM_SUCCESS is returned. Otherwise the module returns a
failure code. Support for authentication and account management module types

Chapter 8. Security

335

is provided. This module only accepts debug and nowarn as specified options in
the configuration file.
The pam_allowroot module only checks the real user ID. Many applications that
require root access will set the effective user ID to 0. For this reason, the
effective ID is not used in determining whether or not the user executing the
authenticating application is a root user.
It is recommended that pam_allowroot be used as sufficient in conjunction with
other modules. This allows the root user to bypass the rest of the modules in the
stack and for a failure not to impact the result of other authenticating users. An
example authentication stack configuration which mimics the historic behavior of
the su command follows:
#
# The PAM configuration for standard su behavior.
#
su auth sufficient /usr/lib/security/pam_allowroot
su auth required

pam_ckfile
The pam_ckfile module allows or denies authentication, based on the existence
of a file. The file checked for existence can be set with the file module option. If
not specified the file defaults to /etc/nologin.
If the specified file exists, only root users (those with a user ID of 0) may
authenticate. All other users are denied access for the service, and pam_ckfile
will echo the contents (if any) of that file. If the specified file does not exist, the
module returns PAM_IGNORE. System administrators should ensure that
success or failure of the module stack for a service does not depend solely on
the result of this module.
It is recommended that pam_ckfile is used with the required or requisite control
flag in conjunction with other modules.

pam_permission
The pam_permission module is an authentication and account-service PAM
module that uses an access-control list to determine whether or not to permit or
deny authentication requests. The file to use for the control list is configured
using a module option and defaults to /etc/ftpusers if not specified.
If the access-control file exists, the pam_permission module will scan the file
using the authenticating user name and group(s). The first match will then be

336


used to determine the result. The general syntax for an entry in the
access-control file is as follows:
[+|-][@]<name>
The optional first character controls whether to allow(+) or deny(-) the request for
the user or group specified by <name>. If a + or - is not the first character in an
entry, then the value of the found module option determines the behavior.
Preceding a name with an @ symbol designates the entry as a group. Otherwise
the entry is used as a user name. The first match found to a user name or group
entry is used to determine access.
All spaces in an entry are ignored. Comments may be added to the file using the
# character as the first character in the line. Only one entry or comment is
allowed per line and the entries are processed one at a time, sequentially,
starting at the top of the file.
Using the keyword ALL will match all users. Since the file is parsed sequentially,
use of the ALL keyword should be reserved for the end of the file since any
entries after it are ignored.
Upon reaching the end of the access-control file, if a match to a user name or
group has not been made, the result will be the opposite value of the found
module option. For example, if found=prohibit is set and the user is not found
within the file, then the result for that user would be allow.
If the specified access control file does not exist, the module will return
PAM_IGNORE and have no affect on the module stack. It is not recommended
that the overall success or failure of the module stack depend solely on
pam_permission.
It is recommended that pam_permission be used with the control flag required or
requisite in conjunction with other modules.

pam_prohibit
The pam_prohibit module returns a failure for all PAM module types. If used as a
required or requisite module for a service, the stack that this module is
incorporated into will always fail. It is recommended that individual services be
explicitly configured in /etc/pam.conf and then the pam_prohibit module used for
the OTHER service entries. Configuring the system in this way ensures that only
known PAM-enabled applications are capable of successfully authenticating
users. Following is an example of how to configure the OTHER service keyword
in /etc/pam.conf to use the pam_prohibit module:
#
# Fail for all PAM services not explicitly configured

Chapter 8. Security

337

#
OTHER
OTHER
OTHER
OTHER

auth
account
password
session

required
required
required
required

/usr/lib/security/pam_prohibit

pam_rhost_auth
The pam_rhosts_auth module provides rhost authentication services similar to
the rlogin, rsh, and rcp commands. The module queries the PAM handle for the
remote user name, remote host, and the local user name. This information is
then compared to the rules in /etc/hosts.equiv and $HOME/.rhosts.
For a typical user, the module first checks /etc/hosts.equiv. If a match is not
found for the user name and host name, the module will continue on to check the
$HOME/.rhosts file. If a user name and host name match is still not found, the
module returns the PAM_AUTH_ERR failure code. Otherwise, the result
depends on the first rule found matching the specified user name and host name.
When authenticating to the root user (user with the UID of 0), the first check of
the /etc/hosts.equiv file is skipped. Success of the rhosts authentication is based
solely on the contents of the root user's $HOME/.rhosts file.
This module requires that a PAM application, before making the call to
pam_authenticate, call pam_set_item and at least set the values of
PAM_RHOST and PAM_RUSER. If the PAM_USER item is not set, the module
will prompt for the user name through the conversation function provided in the
PAM handle.
Further description on how rhosts authentication works can be found in the
documentation for the ruserok() subroutine. Information regarding the syntax of
rhost configuration files can be found in the $HOME/.rhosts or /etc/hosts.equiv
files description.
For expected behavior, pam_rhosts_auth should be used as one of the first
authentication modules in the stack and designated as sufficient.
#
# PAM authentication stack for typical rlogin behavior.
#
rlogin auth sufficient /usr/lib/security/pam_rhosts_auth
rlogin auth required
/usr/lib/security/pam_unix

338


8.2.3 PAM application programming interface changes
Support for the requisite control option in a PAM configuration file was added to
the PAM library in AIX 5L Version 5.3.
In addition to the new control option, the PAM library was further enhanced by
three new PAM environment APIs in accordance with the PAM specification as
defined in the Open Group standard X/Open Single Sign-on Service (XSSO) and
in RFC 86.0:
pam_getenv

Retrieve the value of a defined PAM environment
variable.

pam_getenvlist

Retrieve a list of all of the defined PAM environment
variables and their values.

pam_putenv

Set a PAM environment variable.

8.3 Prioritizing LDAP servers
AIX supports multiple LDAP servers and offers a failover mechanism. When the
current server it is talking to becomes unavailable, it automatically talks to the
next available server in the list. Since Version 5.3, AIX is also capable of
detecting when a higher priority server is back on line, and reconnects to it.
The list of the servers as defined in the /etc/security/ldap/ldap.cfg configuration
file is prioritized. The first server in the list has the highest priority. The
connection is managed by the secldapclntd daemon that reads the configuration
file. An example of a section of the /etc/security/ldap/ldap.cfg file follows:
/etc/security/ldap/ldap.cfg:
# Comma separated list of ldap servers this client talks to
ldapservers:ldapsrv1.ibm.com,ldapsrv2.ibm.com

8.4 Rsh login control
The classic way of disabling execution of r-commands has been to set the ttys
user attribute for specific users:
chuser ‘ttys=ALL, !RSH’ username
In AIX 5L Version 5.3, a new user attribute rcmds is introduced. This attribute
controls the r-command execution. It accepts the following values:
allow

Allows the user to execute r-commands.

Chapter 8. Security

339

deny

Denies the user to execute r-commands. This is the same
as “tty=ALL, !RSH”.

hostlogincontrol

Specifies that the ability of remote command execution is
under the control of the hostsallowedlogin and
hostsdeniedlogin attributes. Those attributes control the
user login.

An example of use of the rcmds user attribute follows:
s4 # chuser rcmds=hostlogincontrol adrian
s4 # chuser hostsdeniedlogin=server4 adrian
s4 # su - adrian
$ rsh server4 date
3004-306 Remote logins are not allowed for this account.
$ exit
s4 # chuser hostsdeniedlogin='' adrian
s4 # chuser hostsallowedlogin=server4 adrian
s4 # su - adrian
$ rsh server4 date
Thu Jul 8 15:47:59 CDT 2004

8.5 chpasswd command
The chpasswd command is introduced that allows the user to change the
password in non-interactive form, for example from a script.
An example on how to change the password of user adrian to passwd follows:
# chpasswd -?
chpasswd: Not a recognized flag: ?
Usage:
chpasswd [-R load_module] [-e] [ -f flags | -c ]
# chpasswd
adrian:passwd
^D
#

8.6 Initial login license limit increased
The license policy at the time of writing is based on per CPU licensing. The per
user licensing is irrelevant in this case.

340


The initial login license limit has been changed to 32,767, which is the largest
signed 16-bit number. This change is also included in the later versions of AIX 5L
Version 5.2. The previous initial login license limit was set to 2. Changes to this
attribute do not affect the system performance.
The login license limit is defined in the maxlogins attribute in the
/etc/security/login.cfg file.

8.7 NIS/NIS+ enhancements
The following enhancements have been made in AIX 5L Version 5.3:
AIX security support of netgroups in LDAP
Client support for shadow passwords (encrypted passwords in the
/etc/security/passwd file)
A popular abstraction known as netgroups is used to name sets of users, hosts,
and domain names. As a concept they are very useful for accessing files that
require permissions. For example, used in /etc/exports, they can specify which
hosts may mount certain file systems. Used in /etc/hosts.equiv or in user’s
.rhosts file, they can grant or revoke login permissions for groups of users or
machines.
Prior versions of AIX supported netgroups only through NIS. AIX 5L Version 5.3
has extended this support to LDAP, and any customizable load modules.
Version 5.3 has made access control more secure by tightening the shadow
passwords and authentication maps through use of the passwd.adjunct map.

8.8 Support for passwd.adjunct NIS map
The previous versions of AIX do not support the passwd.adjunct map file in NIS.
Starting with Version 5.3, if you have a NIS server with the passwd.adjunct file,
you are now able to be authenticated through this map file. Example 8-3 shows
the steps to test the passwd.adjunct map support.
Example 8-3 Authentication through the passwd.adjunct NIS map file
# Changes the current domain name of the system to your domain. (“nis-dom”, for
example)
# chypdom -B nis-dom
# Configure a NIS client using your NIS server address (nismaster.itso.ibm.com,
in this example)

Chapter 8. Security

341

# mkclient -B -S nismaster.itso.ibm.com
# After configuring the NIS client, use ypcat to display the NIS passwd.adjunct
database:
# ypcat passwd.adjunct.byname
adjtest:AgRjp3PiYTFHc:::::
# telnet to the NIS client (your AIX box) using
login_name=adjtest, passwd=adjtest. You should get
authenticated.

342


9

Chapter 9.

Clustering technologies
This chapter focuses on the enhancements to the clustering technologies
introduced by AIX 5L Version 5.3, namely:
RSCT support for Micro-Partitioning technology
Least privilege resource manager
Cluster Systems Management Version 1.4 for AIX


343

9.1 RSCT support for Micro-Partitioning technology
The host resource manager is a component of the RSCT that provides the
capability to monitor resources that are tied to an individual machine. This is
similar to the function of the aixos resource monitor that is available in event
management. However, in this component the values to be monitored are
implemented within the new resource management model that allows us to
extend these capabilities to configure and control these resources in the future.
The types of values that will be provided include those related to the operating
system load and status (paging space, memory usage, processor load, and so
on). In addition, this resource manager includes the capability to monitor
programs for creation, death, and other characteristics.
AIX 5L Version 5.3 uses RSCT version 2.4. The following resource classes are
updated in RSCT version 2.4:
IBM.Host
IBM.Processor

9.1.1 Updated Host resource class - IBM.Host
The following three persistent resource attributes are added to the IBM.Host
class in RSCT version 2.4.0:
NumOnVProcessors

Indicates the number of virtual processors currently
online in this partition. This attribute is available only
on partitions using Micro-Partitioning technology.

EntProcCapacity

Indicates the number of processor units this LPAR is
entitled to receive. This attribute represents the
partition’s percentage of shared physical processors.
This attribute is available only on partitions using

NumActPProcessors

Indicates the current number of physical CPUs in the
system that contains this partition. This attribute is
available only on partitions using Micro-Partitioning
technology.

9.1.2 Updated Processor resource class - IBM.Processors
The following persistent resource attribute is added to the IBM.Processors class
in RSCT 2.4.0:
LogicalId

344

Identifies SMT threads ID. On SMT-enabled system, one
physical or virtual processor can support multiple thread


contexts and execute instructions for them in parallel,
thereby improving the utilization of the functional units
and otherwise idle processor cycles due to cache delays
or other wait states imposed on an individual thread of
execution. This attribute refers to this SMT thread’s ID.

9.2 Least privilege resource manager
The Least privilege resource manager (LPRM) provides a way through which
non-root users and processes can access least privilege commands and
executable scripts based on their authenticated identity and authorizations in the
access control lists. Least privilege commands/scripts are root-only executables
to which a normal user, who otherwise cannot run such commands, is given
access in the most restricted sense for certain purposes only. Such a facility
reduces the need for users and processes to run with total root authority when, in
reality, they need only execute one or two commands that require root authority.
The LPRM is modeled after PSSP's sysctl facility, which grants non-root users
and processes access to root commands by wrapping the root commands in
sysctl procedures. These procedures, in turn, become the actual commands
seen and issued by users or processes. This is accomplished in the LPRM by
modeling a resource class, the resources of which are such root only commands.
Using CT security services for authentication and RSCT RMC ACL file for
authorization users are granted access to these resources.
LPRM functions as an authenticated client server system, enabling remote and
parallel execution of root commands. A command line interface, that is the LPRM
client, is provided which can perform operations on the commands such as
creation, execution, deletion, modification, listing, history of previous commands.
The LPRM daemon maintains and manages the resource class and its
instances.
LPRM supports IW, Distributed management, and Peer modes of operation. Not
only does the LPRM give the most restricted access to root-only commands, it
also logs the detailed usage information of such commands using AuditLog.
Also, when RMC's Full ACL support is available, LPRM will be able to give least
privilege access to users at a command (resource) level.

9.2.1 LPRM commands
The LPRM is related to the rsct.core.lprm LPP file set. This section describes the
basic commands of the LPRM environment.

Chapter 9. Clustering technologies

345

mklpcmd command
The mklpcmd command defines a new least-privilege root command or script to
the resource monitoring and control (RMC) subsystem.
To define the /mycommand script as the LP1 LPRM command locally on your
node, run the mklpcmd command as follows:
# mklpcmd LP1 /mycommand

lslpcmd command
The lslpcmd command lists the least-privilege resources on one or all nodes in a
domain.
To list the LPRM commands locally on your node with all their attributes, run the
lslpcmd -A command as follows:
# lslpcmd -A
Name = LP1
ActivePeerDomain =
CheckSum = 1723032027
CommandPath = /mycommand
ControlFlags = 1
Description =
Lock = 0
NodeNameList = {server4}

runlpcmd command
The runlpcmd command runs the least-privilege resource.
The following example shows how to run the LP1 command and substitute an
argument to it:
# cat /mycommand
#!/usr/bin/ksh
echo "$0: Starting my program"
echo "$0: My program says $1"
echo "$0: My program finished"
exit 0
# runlpcmd LP1 "Hello"
Stdout::/mycommand: Starting my program
/mycommand: My program says Hello
/mycommand: My program finished

346


chlpcmd command
The chlpcmd command changes the read/write attribute values of a least
privilege root command or script.
The following example shows how to use the chlpcmd command to change the
command path attribute:
# chlpcmd LP1 CommandPath=/mycommand2
# lslpcmd -A
Name = LP1
ActivePeerDomain =
CheckSum = 1512195727
CommandPath = /mycommand2
ControlFlags = 1
Description =
Lock = 0
NodeNameList = {server4}
# runlpcmd LP1 "Hello"
Stdout::/mycommand2: Starting my program
/mycommand2: My program says Hello
/mycommand2: My program finished

lphistory command
The lphistory command lists or clears the history of a certain number of
least-privilege commands that were previously issued during the current
resource monitoring and control (RMC) session.
The following example shows how to use the lphistory command to show the
history of commands and the lphistory -c to clear the history:
# lphistory
/mycommand
/mycommand
/mycommand
/mycommand Hello
/mycommand Hello
/mycommand2 Hello
# lphistory -c

rmlpcmd command
The rmlpcmd command removes one or more least-privilege resources from the
resource monitoring and control (RMC) subsystem.


347

An example of removing the LP1 command from the LPRM environment follows:
# rmlpcmd LP1

9.2.2 LPRM security
The security to execute or modify the LPRM commands is based on the RMC
ACLs. To enable the execution or creation of the commands configured to the
LPRM you need to modify the /var/ct/cfg/ctrmc.acls file and edit the
IBM.LPCommands class. If it does not exist then create a new stanza. To enable
the user to create, modify, and execute the commands defined to the LPRM, add
the following line to the IBM.LPCommands stanza:
user@LOCALHOST

*

rwx

To enable the user only to execute the commands already defined to the LPRM,
add the following line to the IBM.LPCommands stanza:
user@LOCALHOST

R

x

An example that enables user adrian to run the LPRM command, but does not
allow this user to create new ones or modify existing ones, follows.
# more /var/ct/cfg/ctrmc.acls
...
IBM.LPCommands
root@LOCALHOST
adrian@LOCALHOST

*
R

rwx
x

# stopsrc -s ctrmc
# startsrc -s ctrmc
# su - adrian
# su - adrian
$ runlpcmd LP1 "Hello"
Stdout::/mycommand: Starting my program
/mycommand: My program says Hello
/mycommand: My program finished
$ mklpcmd LP2 /mycommand2
2610-418 Permission is denied to access the resources or resource class
specified in this command.
$ exit
# su - guest
$ runlpcmd LP1 Hello
2610-418 Permission is denied to access the resources or resource class
specified in this command.

In the example, user adrian started the /mycommand through the LPRM
environment and the command is executed under the root user.

348


9.3 Cluster Systems Management Version 1.4 for AIX
AIX 5L Version 5.3 introduces a new version of the Cluster Systems
Management (CSM) software. The version of CSM that comes with AIX 5L
Version 5.3 is CSM Version 1.4.
CSM Version 1.4 for AIX requires, and ships with, AIX 5L Version 5.3 or AIX 5L
Version 5.2 Maintenance Level 04. If the system is currently at a lower level, then
the upgrade procedure described later in this section should be followed.
Note: Prerequisites for CSM are AIX installp file sets and also RPM packages.
All prerequisite installp file sets and RPM packages are available on the basic
AIX installation CD.
The CSM that comes with the AIX 5L Version 5.3 does not contain the full license
of the product. It contains only a time-limited license. The full license key is
delivered on separate CD media.

9.3.1 CSM 1.4 new features
CSM Version 1.4 has the following new features:
Support for AIX 5L Version 5.3.
HW support of the new POWER5 systems.
Support for JS20 Blade servers running Linux.
Enhanced cluster scaling. POWER4 HMC scales to 32 physical systems and
64 operating system images.
Utilities and documentation for the transition from PSSP to CSM.
New version of OpenCIMOM is used. Version 0.8-1 is required to support
new hardware and is necessary for CSM Version 1.4.
The rpower command is updated with new function.
Java 1.4.1 is used by CSM as it is the default on AIX 5L Version 5.3.

Hardware management console power method
The rpower command is updated to have ability to wait for on, off, reboot,
cec_on, cec_off and open_firmware operations to complete before returning.
The rpower -w will wait for all these commands to complete, but only when used
with the newly introduced -w option. The command completes means that it
returns when the operation is accepted by the HMC CIMOM, but it does not wait
until the node changes state.


349

In previous versions, the command returned before the operation was accepted
by the HMC CIMOM infrastructure and it was even before the CIMOM operation
has ever begun. With the new -w option, the command returns when the
operation is accepted by the HMC CIMOM.
The rpower command executes the operations in parallel for nodes on multiple
HMCs.

9.3.2 CSM migrations to Version 1.4
Supported migration to CSM Version 1.4 is from CSM Version 1.3. Customers
with lower versions of CSM must upgrade to CSM Version 1.3 at least.
Use the following procedure to upgrade to CSM Version 1.4:
1. Verify that the environment before upgrade corresponds to the requirements.
2. Back up the CSM data. The migration preserves the CSM database, but it is
prudent to create a backup, just in case. Run the csmbackup to save the CSM
environment and archive the file outside the server that is being upgraded.
3. Perform a regular system migration of the AIX as documented in the AIX 5L
Version 5.3 Installation Guide and Reference, in the section titled “Migration
Installation.”
As part of the migration, all the CSM LPP file sets will be updated. If NIM is
configured, all the NIM data and definitions will be preserved.
4. Update the open source prerequisites. The AIX migration does not
automatically update the RPM packages. The following CSM prerequisite
RPM packages need to be upgraded:
– tk
– tcl
– expect
– conserver
The installation packages of the listed packages are available on one of the
AIX 5L Version 5.3 installation CD media. For example, you can use the AIX
geninstall command to install the RPMs as follows:
geninstall -laX -d /dev/cd0 R:expect R:tcl R:tk R:conserver
5. Accept the CSM license. The license key is shipped separately on the CSM
license CD media. If you are migrating the operating system, you will be
getting a new version of CSM and you will have to accept the new license.

350


To accept the license and activate the license key, use the mount and
csmconfig commands as follows:
mount -v cdrfs -o ro /dev/cd0 /mnt
csmconfig -L /mnt/csmlum.full
At the prompt, follow the instructions to accept the CSM license. Check the
success of the csmconfig command by running it with no flags, and then
checking the output.
6. Copy the CSM files into the /csminstall subdirectories. Use the following
command to do this:
csmconfig -c
These files are primarily used when CSM system management scripts are run
on the nodes of the cluster.
7. Verify the installation. To verify that the management server has installed
correctly and is ready for use, you can run the ibm.csm.ms probe, which is
shipped with CSM. To run the probe issue the following command:
probemgr -p ibm.csm.ms -l 0
8. Migrate the nodes of the cluster. The operating system migration of the
cluster nodes is accomplished using standard AIX support. This includes the
AIX Network Installation Manager (NIM).
Migrating the CSM management server does not necessarily mean that the
nodes must also be migrated. The management server will still be able to
manage the nodes even if they are not upgraded, if the nodes are at compatible
levels with the CSM management server.
When the nodes are migrated to a new operating system level, keep in mind that
there will also be a new level of CSM file sets installed on the nodes. You should
therefore update the InstallCSMVersion and InstallDistributionVersion node
attributes to be consistent with what is installed. This is done by using the chnode
command. For example, if you would like to update all the node definitions to
have a CSM version of 1.4.0 and an operation system version of 5.3.0 you could
issue the following command:
chnode -a InstallCSMVersion=1.4.0 InstallDistributionVersion=5.3.0


351

352


10

Chapter 10.

National language support
This chapter discusses the enhancements made in AIX 5L Version 5.3 for
national language support, including the following topics:
Gujarati NLS enablement
Tamil NLS enablement
Kazakh NLS enablement
Telegu NLS enablement
Unicode extension
Update AGFA TrueType font rasterizer
Unicode 4.0 support


353

10.1 Gujarati NLS enablement
AIX 5L Version 5.3 provides full locale support within the base operating system
for the Gujarati language as spoken in western India. Hindi is the country’s
national language spoken by nearly 500 million people. However, there are 17
other officially recognized languages, of which Gujarati is one. The Gujarati script
is written from left to right and top to bottom in the same fashion as English.
However, the language takes on a greater complexity because of the various
ways in which Gujarati characters combine to form connected ligatures and
syllables. The rules for rendering Gujarati are very similar to those for Hindi,
although some differences exist primarily due to differences in the available
glyphs for rendering the language.

10.2 Tamil NLS enablement
for the Tamil language as spoken in India. Hindi is the country’s national
language spoken by nearly 500 million people. However, there are 17 other
officially recognized languages, of which Tamil is one. Tamil is the main language
spoken in the state of Tamil Nadu and the union territory of Pondicherry. Tamil
language is a member of the Dravidian/South Indian family of languages.
Tamilians alone number about 64 million people. The Tamil script is written from
left to right and top to bottom in the same fashion as English. However, proper
rendering of the language is somewhat difficult due to the various ways in which
Tamil letters can change shape and position depending on their context.

10.3 Kazakh NLS enablement
AIX 5L Version 5.3 provides full locale support within the AIX base operating
system for the Kazakh language as spoken in Kazakhstan. Kazakh is the official
language and principle native language of the Republic of Kazakhstan. It is also
spoken in southern Siberia, northwestern China (Sinkiang-Uighur) and
northwestern Mongolia. It is one of the most widely spoken languages in central
Asia. An estimated 8 million people speak Kazakh.

10.4 Telegu NLS enablement
for the Telegu language as spoken in India. Hindi is the country’s national
language spoken by nearly 500 million people. There are 17 other officially

354


recognized languages, of which Telegu is one. It is the main language spoken in
the state of Andhra Pradesh and is the second most widely spoken language in
southern India. Telegu language is a member of the Dravidian/South Indian
family of languages. It is written using the Telegu script and is written from left to
right and top to bottom in the same fashion as English. However, proper
rendering of the language is somewhat difficult due to the various ways in which
Telegu letters can change shape and position depending on their context.

10.5 Unicode extension
for the GB18030-2000 codeset standard. GB 18030-2000 is a new Chinese
standard that specifies an extended codepage and a mapping table to Unicode.
GB 18030 was first published on March 17, 2000. After feedback from the
worldwide software industry, the codepage was changed, and a new mapping
table was released.
This codepage standard is important for the software industry because China
has mandated that any software application that is released for the Chinese
market must support GB 18030.

10.6 Update AGFA TrueType font rasterizer
The UFST (Universal Font Scaling Technology) font rasterizer from
AGFA/Monotype has been updated from version 4.4 to version 4.6 for AIX 5L
Version 5.3.
UFST version 4.6 includes support for 4-byte font indexing. Previous versions
only supported 2-byte font indexing, so were limited to fonts that contain no more
than 65536 characters. New international standards such as Chinese GB18030
require thousands of additional characters, thus the bigger font index is needed.

10.7 Unicode 4.0 support
AIX 5L Version 5.3 includes enhancements to the existing Unicode locales in
order to bring them up to compliance with the latest published version of the
standard, which is the 4.0 version. This version adds additional characters and
scripts to the standard, bringing the total number of defined Unicode characters
to 96,382. Some of these characters are required in order to achieve support for
Japanese Industrial Standard JISX0213.

Chapter 10. National language support

355

356


A

Appendix A.

Web-based System Manager
enhancements
AIX 5L Version 5.3 supports the following features in Web-based System
Manager, most of which are already available with Version 5.2 ML3:
Now using Java 1.4.2 (Version 5.3) instead of 1.4.1 (Version 5.2 ML3)
Support for Java Web Start (Version 5.2 ML3)
Voicing support on Windows (Version 5.2 ML3)
Improved keyboard navigation (Version 5.2 ML3)
Support for Windows Native theme on Windows (Version 5.2 ML3)
In addition to these topics, enhancements in the base operating system have
been reflected by modifications in Web-based System Manager in the areas of:
Advanced Accounting
NFS Version 4
SMT, job scheduling, resource sets
Performance tuning and monitoring
Configuration assistant
Remote client for firewall configuration
For a discussion of these topics, consult the product documentation.


357

A.1 Java Web Start
Users of the Linux or Windows client have the choice of using Java Web Start
instead of installing the client using InstallShield. The URL for downloading the
remote client is:
http://<hostname>/remote_client.html

The page displayed will allow the following choices:
InstallShield

This remote client is installed using an InstallShield
wizard and it must be re-installed to obtain updates. This
client is useful when running the Web-based System
Manager over a broadband connection (cable modem or
DSL), because updates to the console are not
automatically downloaded.

Java Web Start

This remote client is loaded by Java Web Start, which
must be installed on the client system prior to installing
the remote client. This version of the remote client will
check for updates on the server every time it is invoked
and download updates automatically.

A.2 Voicing support on Windows
Installation of the remote client now automatically installs the IBM Accessibility
Speech Interface V1.2 and IBM ViaVoice® TTS Runtime V6.740. This
technology preview provides for a limited voicing capability of Web-based
System Manager. In order to run with Voicing enabled, the startup file
wsmsvk.bat sets the required environment variables and launches Web-based
System Manager.
The startup file wsm.bat that starts Web-based System Manager without voicing
turned on and wsmsvk.bat are located in the bin sub-directory of the
Web-based System Manager install directory (normally C:Program
Fileswebsmbin).
The desktop shortcut that was created above launches Web-based System
Manager without voicing turned on. You can create a copy of that shortcut and
change the startup file to wsmsvk.bat if you would like to be able to launch the
voicing version of Web-based System Manager using a short-cut.

358


A.3 Keyboard accessibility
The goal of keyboard accessibility is for the user to be able to use the Web-based
System Manager without having to use a mouse (Figure A-1). The following
keyboard accessibility features are available:
Menu mnemonics

All menu choices can be selected from the keyboard by
typing the letter indicated in the menu title. To open the
menu, type the underlined letter while pressing the Alt key
on the keyboard. This is true only for opening the menu.
Once the menu is open, release the Alt key. For example,
to select the Properties option in the Selected menu, open
the menu by typing s while holding the Alt key, then
release the Alt key and type r to select the Properties
option. When using mnemonics of the Web-based
System Manager menu bar, be sure to move the mouse
cursor into the console frame.

Figure A-1 Improved keyboard navigation in Web-based System Manager

Menu accelerators or shortcut keys
Key combinations are available for common actions, for
example, Ctrl + Q to quit and F9 for Key Help.

Appendix A. Web-based System Manager enhancements

359

A.4 Inherit desktop theme on Windows remote client
Web-based System Manager supports the Windows Native theme on the
Windows client (Figure A-2).
The Native theme causes Web-based System Manager to inherit a number of
desktop theme properties. When you set your theme type to Native and you
subsequently run Web-based System Manager on a non-Windows client, your
theme will be set to Classic. As long as you do not change your theme (and save
your preferences), Web-based System Manager will remember to set the theme
back to Native when you run again from a Windows client. If you run an older
version of the Web-based System Manager client and save your preferences,
then any Native theme setting is lost and will not be remembered the next time
you launch Web-based System Manager using that preference file.

Figure A-2 Native theme support in Web-based System Manager

360


Abbreviations and acronyms
ABI

Application Binary Interface

BCT

Branch on CounT

AC

Alternating Current

BFF

Backup File Format

ACL

Access Control List

BI

Business Intelligence

ADSM

ADSTAR Distributed Storage
Manager

BIND

Berkeley Internet Name
Domain

ADSTAR

Advanced Storage and
Retrieval

BIST

Built-In Self-Test

BLAS

AFPA

Adaptive Fast Path
Architecture

Basic Linear Algebra
Subprograms

BLOB

Binary Large Object

AFS®

Andrew File System

BLV

Boot Logical Volume

AH

Authentication Header

BOOTP

Boot Protocol

AIO

Asynchronous I/O

BOS

Base Operating System

AIX

Advanced Interactive
Executive

BPF

Berkeley Packet Filter

ANSI

American National Standards
Institute

BRX

Branch Execution Unit

BSC

APAR

Authorized Program Analysis
Report

Binary Synchronous
Communications

BSD

Berkeley Software Distribution

Application Programming
Interface

CA

Certificate Authority

CAD

Computer-Aided Design

AppA

Application Audio

CAE

Computer-Aided Engineering

AppV

Application Video

CAM

ARP

Address Resolution Protocol

Computer-Aided
Manufacturing

ASCI

Accelerated Strategic
Computing Initiative

CATE

Certified Advanced Technical
Expert

ASCII

American National Standards
Code for Information
Interchange

CATIA

Computer-Graphics Aided
Three-Dimensional
Interactive Application

ASR

Address Space Register

CCM

Common Character Mode

ATA

Advanced Technology
Attachment

CD

Compact Disk

CDE

Common Desktop
Environment

API

ATM

Asynchronous Transfer Mode

AuditRM

Audit Log Resource Manager

CDLI

Common Data Link Interface

AUI

Attached Unit Interface

CD-R

CD Recordable

AWT

Abstract Window Toolkit


361

CD-ROM

Compact Disk-Read Only
Memory

CWS

Control Workstation

DAD

Duplicate Address Detection

CE

Customer Engineer

DAS

Dual Attach Station

CEC

Central Electronics Complex

DASD

Direct Access Storage Device

CFD

Computational Fluid
Dynamics

DAT

Digital Audio Tape

DBCS

Double Byte Character Set

DBE

Double Buffer Extension

DC

Direct Current

CFM

Configuration File Manager

CGE

Common Graphics
Environment

CHRP

Common Hardware
Reference Platform

DCE

Distributed Computing
Environment

CIM

Common Information Model

DCM

Dual Chip Module

CISPR

International Special
Committee on Radio
Interference

DCUoD

Dynamic Capacity Upgrade
on Demand

DDC

Display Data Channel

CIU

Core Interface Unit

DDS

Digital Data Storage

CLI

Command Line Interface

DE

Dual-Ended

CLIO/S

Client Input/Output Sockets

DES

Data Encryption Standard

CLVM

Concurrent LVM

DFL

Divide Float

CMOS

Complimentary Metal-Oxide
Semiconductor

DFP

Dynamic Feedback Protocol

DFS

Distributed File System

CMP

Certificate Management
Protocol

DGD

Dead Gateway Detection

DH

Diffie-Hellman

DHCP

Dynamic Host Configuration
Protocol

DIMM

Dual In-Line Memory Module

DIP

Direct Insertion Probe

DIT

Directory Information Tree

DIVA

Digital Inquiry Voice Answer

DLPAR

Dynamic LPAR

DLT

Digital Linear Tape

COFF

Common Object File Format

COLD

Computer Output to Laser
Disk

CPU

Central Processing Unit

CRC

Cyclic Redundancy Check

CRL

Certificate Revocation List

CSID

Character Set ID

CSM

Cluster Systems
Management

CSR

Customer Service
Representative

DMA

Direct Memory Access

DMT

Directory Management Tool

CSS

Communication Subsystems
Support

DMTF

Distributed Management Task
Force

CSU

Customer Set-Up

DN

Distinguished Name

CUoD

Capacity Upgrade on
Demand

DNLC

Dynamic Name Lookup
Cache

362


DNS

Domain Naming System

DOE

Department of Energy

DOI

ELF

Executable and Linking
Format

Domain of Interpretation

EMU

European Monetary Union

DOM

Document Object Model

EOF

End of File

DOS

Disk Operating System

EPOW

DPCL

Dynamic Probe Class Library

Environmental and Power
Warning

DRAM

Dynamic Random Access
Memory

ERRM

Event Response resource
manager

DRM

Dynamic Reconfiguration
Manager

ESID

Effective Segment ID

ESP

DS

Differentiated Service

Encapsulating Security
Payload

DSA

Dynamic Segment Allocation

ESS

Enterprise Storage Server®

DSE

Diagnostic System Exerciser

ESSL

Engineering and Scientific
Subroutine Library

DSMIT

Distributed SMIT

ETML

DSU

Data Service Unit

Extract, Transformation,
Movement, and Loading

DTD

Document Type Definition

F/C

Feature Code

DTE

Data Terminating Equipment

F/W

Fast and Wide

DVD

Digital Versatile Disk

FC

Fibre Channel

DW

Data Warehouse

FCAL

Fibre Channel Arbitrated Loop

DWA

Direct Window Access

FCC

EA

Effective Address

Federal Communication
Commission

EC

Engineering Change

FCP

Fibre Channel Protocol

ECC

Error Checking and
Correcting

FDDI

Fiber Distributed Data
Interface

ECN

Explicit Congestion
Notification

FDPR

Feedback Directed Program
Restructuring

EEH

Extended Error Handling

FDX

Full Duplex

EEPROM

Electrically Erasable
Programmable Read Only
Memory

FIFO

First In/First Out

FLASH EPROM

Flash Erasable
Programmable Read-Only
Memory

FLIH

First Level Interrupt Handler

FLOP

Floating Point Operation

FMA

Floating point Multiply Add
operation

FP

Fixed Point

FPR

Floating Point Register

FPU

Floating Point Unit

EFI

Extensible Firmware Interface

EHD

Extended Hardware Drivers

EIA

Electronic Industries
Association

EIM

Enterprise Identity Mapping

EISA

Extended Industry Standard
Architecture

ELA

Error Log Analysis


363

FRCA

Fast Response Cache
Architecture

HP-UX

Hewlett-Packard UNIX

HTML

Hyper-text Markup Language

FRU

Field Replaceable Unit

HTTP

Hypertext Transfer Protocol

FSRM

File System Resource
Manager

Hz

Hertz

FTP

File Transfer Protocol

FTP

File Transfer Protocol

GAI

I/O

Input/Output

I2 C

Inter Integrated-Circuit
Communications

Graphic Adapter Interface

IAR

Instruction Address Register

GAMESS

General Atomic and
Molecular Electronic Structure
System

IBM

International Business
Machines

GID

Group ID

ICCCM

Inter-Client Communications
Conventions Manual

GPFS

General Parallel File System

ICE

Inter-Client Exchange

GPR

General-Purpose Register

ICElib

Inter-Client Exchange library

GUI

Graphical User Interface

ICMP

GUID

Globally Unique Identifier

Internet Control Message
Protocol

HACMP

High Availability Cluster Multi
Processing

ID

Identification

IDE

Integrated Device Electronics

HACWS

High Availability Control
Workstation

IDL

Interface Definition Language

HBA

Host Bus Adapters

IDS

Intelligent Decision Server

HCON

IBM AIX Host Connection
Program/6000

IEEE

Institute of Electrical and
Electronics Engineers

HDX

Half Duplex

IETF

Internet Engineering Task
Force

HFT

High Function Terminal

IHS

IBM HTTP Server

HIPPI

High Performance Parallel
Interface

IHV

Independent Hardware
Vendor

HiPS

High Performance Switch

IIOP

Internet Inter-ORB Protocol

HiPS LC-8

Low-Cost Eight-Port High
Performance Switch

IJG

Independent JPEG Group

HMC

Hardware Management
Console

IKE

Internet Key Exchange

ILMI

HMT

Hardware Multithreading

Integrated Local Management
Interface

HostRM

Host Resource Manager

ILS

International Language
Support

HP

Hewlett-Packard

IM

Input Method

HPF

High Performance FORTRAN

INRIA

HPSSDL

High Performance
Supercomputer Systems
Development Laboratory

Institut National de Recherche
en Informatique et en
Automatique

364


IOCTL

I/O Control

L1

Level 1

IP

Internetwork Protocol (OSI)

L2

Level 2

IPL

Initial Program Load

L3

Level 3

IPSec

IP Security

LAM

IrDA

Infrared Data Association
(which sets standards for
infrared support including
protocols for data
interchange)

Loadable Authentication
Module

LAN

Local Area Network

LANE

Local Area Network
Emulation

LAPI

Low-Level Application
Programming Interface

IRQ

Interrupt Request

IS

Integrated Service

LDAP

ISA

Industry Standard
Architecture, Instruction Set
Architecture

Lightweight Directory Access
Protocol

LDIF

LDAP Directory Interchange
Format

ISAKMP

Internet Security Association
Management Protocol

LED

Light Emitting Diode

ISB

Intermediate Switch Board

LFD

Load Float Double

ISDN

Integrated-Services Digital
Network

LFT

Low Function Terminal

LID

Load ID

ISMP

InstallSheild Multi-Platform

LLNL

ISNO

Interface Specific Network
Options

Lawrence Livermore National
Laboratory

LMB

Logical Memory Block

International Organization for
Standardization

LP

Logical Partition

LPAR

Logical Partition

ISV

Independent Software Vendor

LP64

Long-Pointer 64

ITSO

International Technical
Support Organization

LPI

Lines Per Inch

LPP

Licensed Program Product

LPR/LPD

Line Printer/Line Printer
Daemon

ISO

IXFR

Incremental Zone Transfer

JBOD

Just a Bunch of Disks

JCE

Java Cryptography Extension

LRU

Least Recently Used

JDBC

Java Database Connectivity

LTG

Logical Track Group

JFC

Java Foundation Classes

LUN

Logical Unit Number

JFS

Journaled File System

LV

Logical Volume

JSSE

Java Secure Sockets
Extension

LVCB

Logical Volume Control Block

LVD

Low Voltage Differential

LVM

Logical Volume Manager

MAP

Maintenance Analysis
Procedure

JTAG

Joint Test Action Group

JVMPI

Java Machine Profiling
Interface

KDC

Key Distribution Center


365

MASS

Mathematical Acceleration
Subsystem

MX

Mezzanine Bus

NBC

Network Buffer Cache

MAU

Multiple Access Unit

NCP

Network Control Point

MBCS

Multi-Byte Character Support

ND

Neighbor Discovery

Mbps

Megabits Per Second

NDP

Neighbor Discovery Protocol

MBps

Megabytes Per Second

NDS

Novell Directory Services

MCA

Micro Channel® Architecture

NFB

No Frame Buffer

MCAD

Mechanical Computer-Aided
Design

NFS

Network File System

MCM

Multichip Module

NHRP

Next Hop Resolution Protocol

MDF

Managed Object Format

NIM

Network Installation
Management

MDI

Media Dependent Interface

NIMOL

NIM on Linux

MES

Miscellaneous Equipment
Specification

NIS

Network Information Service

MFLOPS

Million of FLoating point
Operations Per Second

NL

National Language

NLS

National Language Support

MII

Media Independent Interface

NT-1

Network Terminator-1

MIB

Management Information
Base

NTF

No Trouble Found

NTP

Network Time Protocol

MIP

Mixed-Integer Programming

NUMA

Non-Uniform Memory Access

MLR1

Multi-Channel Linear
Recording 1

NUS

Numerical Aerodynamic
Simulation

MMF

Multi-Mode Fibre

NVRAM

MODS

Memory Overlay Detection
Subsystem

Non-Volatile Random Access
Memory

NWP

Numerical Weather Prediction

MP

Multiprocessor

OACK

Option Acknowledgment

MPC-3

Multimedia PC-3

OCS

Online Customer Support

MPI

Message Passing Interface

ODBC

Open DataBase Connectivity

MPIO

Multipath I/O

ODM

Object Data Manager

MPOA

Multiprotocol over ATM

OEM

MPP

Massively Parallel Processing

Original Equipment
Manufacturer

MPS

Mathematical Programming
System

OLAP

Online Analytical Processing

OLTP

MSS

Maximum Segment Size

Online Transaction
Processing

MST

Machine State

ONC+

Open Network Computing

MTU

Maximum Transmission Unit

OOUI

MWCC

Mirror Write Consistency
Check

Object-Oriented User
Interface

366


OSF

Open Software Foundation,
Inc.

POP

Power-On Password

POSIX

Portable Operating Interface
for Computing Environments

POST

Power-On Self-test

POWER

Performance Optimization
with Enhanced Risc
(Architecture)

OSL

Optimization Subroutine
Library

OSLp

Parallel Optimization
Subroutine Library

P2SC

POWER2™ Single/Super
Chip

PPC

PowerPC®

PAG

Process Authentication Group

PPM

Piecewise Parabolic Method

PAM

Pluggable Authentication
Mechanism

PPP

Point-to-Point Protocol

PAP

Privileged Access Password

PREP

PowerPC Reference Platform

PBLAS

Parallel Basic Linear Algebra
Subprograms

PRNG

Pseudo-Random Number
Generator

PCB

Protocol Control Block

PSE

Portable Streams
Environment

PCI

Peripheral Component
Interconnect

PSSP

Parallel System Support
Program

PDT

Paging Device Table

PTF

Program Temporary Fix

PDU

Power Distribution Unit

PTPE

PE

Parallel Environment

Performance Toolbox Parallel
Extensions

PEDB

Parallel Environment
Debugging

PTX®

Performance Toolbox

PV

Physical Volume

PEX

PHIGS Extension to X

PVC

Permanent Virtual Circuit

PFS

Perfect Forward Security

PVID

Physical Volume Identifier

PGID

Process Group ID

QMF™

Query Management Facility

PHB

Processor Host Bridges

QoS

Quality of Service

PHY

Physical Layer

QP

Quadratic Programming

PID

Process ID

RAID

PID

Process ID

Redundant Array of
Independent Disks

PIOFS

Parallel Input Output File
System

RAM

Random Access Memory

RAN

Remote Asynchronous Node

PKCS

Public-Key Cryptography
Standards

RAS

Reliability, Availability, and
Serviceability

PKI

Public Key Infrastructure

RDB

Relational DataBase

PKR

Protection Key Registers

RDBMS

PMTU

Path MTU

Relational Database
Management System

POE

Parallel Operating
Environment

RDF

Resource Description
Framework


367

RDISC

ICMP Router Discovery

SCB

Segment Control Block

RDN™

Relative Distinguished Name

SCSI

RDP

Router Discovery Protocol

Small Computer System
Interface

RFC

Request for Comments

SCSI-SE

SCSI-Single Ended

RIO

Remote I/O

SDK

Software Development Kit

RIP

Routing Information Protocol

SDLC

RIPL

Remote Initial Program Load

Synchronous Data Link
Control

RISC

Reduced Instruction-Set
Computer

SDR

System Data Repository

SDRAM

RMC

Resource Monitoring and
Control

Synchronous Dynamic
Random Access Memory

SE

Single Ended

ROLTP

Relative Online Transaction
Processing

SEPBU

Scalable Electrical Power
Base Unit

RPA

RS/6000 Platform
Architecture

SGI

Silicon Graphics Incorporated

SGID

Set Group ID

RPC

Remote Procedure Call

SHLAP

RPL

Remote Program Loader

Shared Library Assistant
Process

RPM

Redhat Package Manager

SID

Segment ID

RSC

RISC Single Chip

SIT

Simple Internet Transition

RSCT

Reliable Scalable Cluster
Technology

SKIP

Simple Key Management for
IP

RSE

Register Stack Engine

SLB

Segment Lookaside Buffer

RSVP

Resource Reservation
Protocol

SLIH

Second Level Interrupt
Handler

RTC

Real-Time Clock

SLIP

Serial Line Internet Protocol

RVSD

Recoverable Virtual Shared
Disk

SLR1

Single-Channel Linear
Recording 1

SA

Secure Association

SM

Session Management

SACK

Selective Acknowledgments

SMB

Server Message Block

SAN

Storage Area Network

SMIT

SAR

Solutions Assurance Review

System Management
Interface Tool

SAS

Single Attach Station

SMP

Symmetric Multiprocessor

SASL

Simple Authentication and
Security Layer

SMS

System Management
Services

SBCS

Single-Byte Character
Support

SNG

Secured Network Gateway

SNIA

ScaLAPACK

Scalable Linear Algebra
Package

Storage Networking Industry
Association

SNMP

Simple Network Management
Protocol

368


SOI

Silicon-on-Insulator

TOS

Type Of Service

SP

IBM RS/6000 Scalable
POWER parallel Systems

TPC

Transaction Processing
Council

SP

Service Processor

TPP

Toward Peak Performance

SPCN

System Power Control
Network

TSE

Text Search Engine

TSE

Text Search Engine

SPEC

System Performance
Evaluation Cooperative

TTL

Time To Live

SPI

Security Parameter Index

UCS

Universal Coded Character
Set

SPM

System Performance
Measurement

UDB EEE

Universal Database and
Enterprise Extended Edition

SPOT

Shared Product Object Tree

UDF

Universal Disk Format

SPS

SP Switch

UDI

Uniform Device Interface

SPS-8

Eight-Port SP Switch

UIL

User Interface Language

SRC

System Resource Controller

ULS

Universal Language Support

SRN

Service Request Number

UNI

Universal Network Interface

SSA

Serial Storage Architecture

UP

Uniprocessor

SSC

System Support Controller

USB

Universal Serial Bus

SSL

Secure Socket Layer

USLA

User-Space Loader Assistant

STFDU

Store Float Double with
Update

UTF

UCS Transformation Format

STP

Shielded Twisted Pair

UTM

Uniform Transfer Model

SUID

Set User ID

UTP

Unshielded Twisted Pair

SUP

Software Update Protocol

UUCP

UNIX-to-UNIX
Communication Protocol

SVC

Switch Virtual Circuit

VACM

SVC

Supervisor or System Call

View-based Access Control
Model

SWVPD

Software Vital Product Data

VESA

SYNC

Synchronization

Video Electronics Standards
Association

TCB

Trusted Computing Base

VFB

Virtual Frame Buffer

TCE

Translate Control Entry

VG

Volume Group

Tcl

Tool Command Language

VGDA

TCP/IP

Transmission Control
Protocol/Internet Protocol

Volume Group Descriptor
Area

VGSA

Volume Group Status Area

TCQ

Tagged Command Queuing

VHDCI

TGT

Ticket Granting Ticket

Very High Density Cable
Interconnect

TLB

Translation Lookaside Buffer

VIPA

Virtual IP Address

TLS

Transport Layer Security

VLAN

Virtual Local Area Network

VMM

Virtual Memory Manager


369

VP

Virtual Processor

VPD

Vital Product Data

VPN

Virtual Private Network

VSD

Virtual Shared Disk

VSM

Visual System Manager

VSS

Versatile Storage Server™

VT

Visualization Tool

WAN

Wide Area Network

WBEM

Web-based Enterprise
Management

WLM

Workload Manager

WTE

Web Traffic Express

XCOFF

Extended Common Object
File Format

XIE

X Image Extension

XIM

X Input Method

XKB

X Keyboard Extension

XLF

XL Fortran

XML

Extended Markup Language

XOM

X Output Method

XPM

X Pixmap

XSSO

Open Single Sign-on Service

XTF

Extended Distance Feature

XVFB

X Virtual Frame Buffer

370


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 372. Note that some of the documents referenced here may be available
in softcopy only.
Advanced POWER Virtualization on IBM eServer p5 Servers Architecture
and Performance Considerations, SG24-5768 available in December 2004
Advanced POWER Virtualization on IBM eServer p5 Servers: Introduction
and Basic Configuration, SG24-7940
A Practical Guide for Resource Monitoring and Control, SG24-6615
Linux Applications on pSeries, SG24-6033
Managing AIX Server Farms, SG24-6606
The Complete Partitioning Guide for IBM Eserver pSeries Servers,
SG24-7039
Effective System Management Using the IBM Hardware Management
Console for pSeries, SG24-7038
Problem Solving and Troubleshooting in AIX 5L, SG24-5496
Understanding IBM Sserver pSeries Performance and Sizing, SG24-4810
AIX 5L Workload Manager (WLM), SG24-5977
AIX Reference for Sun Solaris Administrators, SG24-6584
AIX Version 5.2 Differences Guide, SG24-2014
Introducing VERITAS Foundation Suite for AIX, SG24-6619

Other publications
These publications are also relevant as further information sources:
The following types of documentation are located through the Internet at the
following URL:


371

http://www.ibm.com/servers/eserver/pseries/library

– User guides
– System management guides
– Application programmer guides
– All commands reference volumes
– Files reference
– Technical reference volumes used by application programmers

Online resources
These Web sites and URLs are also relevant as further information sources:
A location to download and explore the Mozilla browser for AIX
http://www.ibm.com/servers/aix/browsers

Information on RFCs for Internet standards
http://www.ietf.org

How to find product documentation
http://publib16.boulder.ibm.com/pseries/en_US/infocenter/base/

What is new with BSD UNIX
http://www.openbsd.org

Inoformation on large page support for AIX
http://www-1.ibm.com/servers/aix/whitepapers/large_page.html

The OpenGroups home page
http://www.opengroup.org

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

372


Help from IBM
IBM Support and downloads
ibm.com/support

IBM Global Services
ibm.com/services

Related publications

373

374


Index
Symbols
$mapformat 88
'@' (at-sign) 88
./lib/5.8.2/aix-thread-multi/CORE/libperl.a 77
./lib64/5.8.2/aix-thread-multi-64all/CORE/libperl.a
77
.rhosts 341
/etc/check_config.files 192
/etc/dhcprd.cnf 302
/etc/dhcpv6/dhcpc6.cnf 302
/etc/dhcpv6/dhcpsdv6.cnf 299
/etc/exports 289, 341
/etc/hosts.equiv 341
/etc/nfs/realm.map 293
/etc/pam.conf 332
/etc/resolv.conf 320
/etc/security/ldap/ldap.cfg 327
/etc/security/login.cfg 331
/usr/bin 77
/usr/lpp/perl.rte/README.perl.aix 77
/usr/opt/perl5 77
/usr/opt/perl5/bin 77
/usr/opt/perl5/link_perl_32 78
/usr/opt/perl5/link_perl_64 78
/usr/share/man/info 251
/usr/sys/inst.images 249
/usr/sys/inst.images/installp/ppc 249
/usr/sys/inst.images/RPMS/ppc 249
__pthrdscreds structure 83

Numerics
3.1 73
9111-520 Feature 7940 3
9113-550 Feature 7941 3
9117-570 Feature 7942 3

A
AACCT 220
abort() 73
access control entry 133
accessibility 358
accounting data files 222


accounting records 239
acctctl command 223
ACL 348
acl
ace 136
aclconvert 139
AIXC ACL 140
eav2 138
inheritance 138
NFS4 ACL 139
aclconvert 139
adaptive heap cache 72
ADM 218
Advanced Accounting 220
accounting records 239
aggregation application record 241
aggregation process record 240
ARM aggregated transaction instance record
246
ARM application environment record 244
ARM transaction environment record 245
ARM transaction instance record 245
client VIO record 243
disk I/O record 243
file system activity record 242
lost data record 243
pad record 239
policy record 242
process record 239
processor and memory use record 241
project definition record 246
server I/O record 243
third-party kernel extension record 244
acctctl command 223
Application Resource Management (ARM) 232
data aggregation 236
project-level data aggregation 237
system-level data aggregation 236
data files 222
data file management 238
data file management commands 223
e-mail notification messages 239
interval accounting 234
interval accounting commands 235

375

process interval 235
system interval 235
policy 223
admin policy 224
alternate admin policy 227
group policy 228
user policy 228
project 223
classification semantics 229
manual project classification 230
projctl command 230
projects and policies commands 230
relative project classification 228
report and analysis 237
readaacct command 237
transactional accounting 232
Advanced POWER Virtualization feature 2
Advisory Info 69
affinity 11, 30
AGFA TrueType Font Rasterizer 355
AGFA/Monotype 355
AIO 256
fast path 258
AIO enhancement 258
AIO fastpath request count 259
AIO queue 259–260
AIO request count 258
AIO request queues 257
aioserver 256, 258
AIX Documentation Services 249
AIX Fast Connect 319
AIX printing 247
AIX security support 341
AIX Toolbox for Linux Applications 248
allocator options
buckets 73
disclaim 73
multiheap 73
threadcache 73
allocator type 73
3.1 73
default allocator 73
user 73
Watson 73
alt_disk_copy command 220
alt_disk_mksysb command 220
alt_lib 220
alt_rootvg_op command 220
Alternate Disk Installation enhancements 219

376


Alternate Disk Installation migration 218
Application Resource Management (ARM) 232
arc count 87
ARM 232
arp 321
arpresolve_common 321
arpupdate 322
ARP, Virtual Ethernet 41
arpresolve_common 321
arpupdate 322
asctime64() function 247
asctime64_r() function 247
Asynchronous I/O 256
Asynchronous I/O statistics 256
at command 252
auth_type attribute 331
authenticatex() 71
authentication 326
authentication maps 341
Authentication process 197
automatic logging of allocations 73
avfc statistics 259
avgc statistics 258
avque statistics 262
avserv statistics 261–262
avwait statistics 261–262

B
backtag functionality to vi 251
backup on DVD 195
Barriers 67
barriers 90
base commands and libraries enhancements 250
at command 252
cron command 252
date command 251
find command 250
fuser command 251
grep command 251
iostat command 250
make command 250
nohup command 252
ps command 251
restore command 251
snap command 251
tar command 251
Berkeley Packet Filter 90
bffcreate command 214, 216

BIND 323
BIND domain search improvement 320
Block device mapping 74
subroutines 75
disclaim() 75
mmap() 75
msync() 75
open() 75
bmap 127
BOS install 203
bos.rte.commands 76
bos.suma fileset 165
bos.sysmgt.quota fileset 115
bosinst.data 164
ERASE_ITERATIONS 165
ERASE_PATTERNS 164
INSTALL_METHOD 165
BPF 90
breakpoints 88
bucket allocator 72
buckets 73
buckets-based front-end 72
buffer threshold 154

C
cache 72
cache bounces 84
caching mechanisms 72
call graph 91
capped mode 17
catch_overflow 73–74
CDE 161
CD-ROM support, vSCSI 51
cede, hcall 7
Certificate 206
Certificate Password 206
certpasswd command 206
certview command 206
chcore command 144
chdev -l sys0 command 281
checksum 76
chfs command 117
Chinese standard 355
chlpcmd command 347
chlv command 108
chlvcopy command 96
chnfs 293
chnfsdom 288, 293

chnfsim 289
chnode command 351
chpasswd command 340
chpassx() 71
CHRP 250
chuser command 84, 339
rcmds user attribute 339
chvg command 102, 112
CIMOM 350
classes of errors 73
classical concurrent mode VG 97
Clocks 63
close() 320
Cluster Systems Management 349
codepage 355
codeset
GB 18030-2000 355
commands
acctctl 223
aclconvert 139
alt_disk_copy 220
alt_disk_mksysb 220
alt_rootvg_op 220
bffcreate command 214, 216
certpasswd command 206
certview command 206
chcore command 144
chdev -l sys0 command 281
chfs 117
chlpcmd command 347
chlv 108
chlvcopy 96
chnfsdom 288, 293
chnfsim 289
chnode command 351
chpasswd command 340
chuser command 84, 339
chvg 102, 112
config.krb5 290
configassist 248
cpupstat 178
csmbackup command 350
csmconfig command 351
csum 76
dmpfmt command 149
dump command 81
emgr 189
epkg 189
exportvg 113

Index

377

extendlv 108
extendvg 96, 112
fixmgr 189
fixpkg 189
fsck 129
gencopy command 214, 216
gencore command 146
getconf 177
getea 130
gprof 91
importvg 96–97, 113
ioo 109
iostat command 256, 258, 261
j2edlimit 118
kadmin.local 291
kdb command 283
kinit 292
ld command 79, 81
ldedit command 82
lparstat 7, 25, 27
lphistory command 347
lpNet 247
lpsched 247
lquerypv 113
lscore command 144
lslpcmd command 346
lslpp 189
lslv 96
lspv 96, 113
lsvg 104
lvmo 109
mkdvd 195
mklpcmd command 346
mklv 108
mklvcopy 96, 108
mksecldap 328
mkuser command 84
mkvg 96, 100, 112
mount command 282
mpstat 11
nfs4cl 288
nfshostkey 288, 293
nim command 203, 212, 214–216
nim_master_recover command 211, 216
nimadapter command 218
nimclient -C command 195
nimclient command 198
nimclient commands 215
niminit command 210, 218

378


nimquery command 198
no 313
pax command 150–151
perl 76
perldoc 77
pmtu 298
probemgr command 351
procmon 76
projctl 230
quota 118
quotacheck 117
quotaoff 118
quotaon 117
readaacct 237
reducevg 112
repquotas 118
rmlpcmd command 347
rpower command 349
rpower -w command 349
runlpcmd command 346
sar -d command 261
sctpctrl command 310
secldifconv 329
setea 130
snap command 149, 151
snapsplit command 153
snmptrap 318
sum 76
suma 171
sysdumpdev command 147, 149–150
sysdumpstart command 148
tcpdump 90
trace command 156
trcctl command 154
ulimit command 85
varyoffvg 113
varyonvg 96, 112
vmo 182
vmo command 86, 187
vmstat 109
vmstat -l command 187
commands fsck 129
commands genintall 192
commands man 92
Common Hardware Reference Platform 250
communication with external networks 36
compression 144
condition 90
confer, hcall 7

config.krb5 290
configassist command 248
Configuration Assistant 248
configuration file modification surveillance 192
/etc/check_config.files 192
continue debug option 73
Coordinated Universal Time 246
Core file 144
core file destination directory 144
core file naming 144
Core size 85
CORE_NAMING environment 145
corefile 89
corefile subcommand 90
CPU time 85
CPU time limits 85
cpupstat command 178
credentials 82
cron command 252
cryptographic authentication 203, 218
Cryptographic Library 76
cryptographic sum command 76
CSM 349
management server 351
csmbackup command 350
csmconfig command 351
csum command 76
CT security services 345
ctime64() function 247
ctime64_r() function 247
ctrmc.acls file 348
customizable load modules 341

D
daemon
lpd 247
Data size 85
data structures 89
data_stagger_interval option to vmo 86
Date APIs past 2038 246
date command 251
DBX 88–89
DBX functionality 87
DBX malloc command 74
DBX subcommands
corefile 90
disable 88
enable 88

fd 89
handler 89
kthread 89
list 88
map 88
onceblock 89
proc 89
resource 89
stop 88
trace 88
use 88
debug a corefile 89
debug events 88
debug session 89
debugged process 88
debuggee 90
debugging core files 89
debugging options
catch_overflow 73
Malloc Log 73
Malloc Trace 73
report_allocations 73
validate_ptrs 73
debugging pthreaded code 89
decrypting 206
dedicated memory 16
dedicated processor partition 16
dedicated processors 18
default allocator 72–73
DES NFS kernel extension 319
device driver 111
device support 261
DHCPv6
DHCPv6 client 302
DHCPv6 relay agent 302
DHCPv6 server 299
difftime64() function 247
disable subcommand 88
disclaim 73
disclaim() subroutine 75
disk I/O performance 112
disk quotas 114
bos.sysmgt.quota 115
chfs command 117
j2edlimit command 118
limits classes 114, 116
quota command 118
quotacheck command 117
quotaoff command 118

Index

379

quotaon command 117
repquotas command 118
SMIT interfaces 119
system configuration 120
distinguished name (DN) 326
dkstat structure 261
dlfcn.h file 78
dlopen() function 78
dlsym() function 78
RTLD_ENTRY handle 79
RTLD_NEXT handle 79
RTLD_SELF handle 79
dmpfmt command 149
DNS query 320
domain names 320
domain search improvement 320
donor, PLM 54
dump command 81
dump-time 90
DVD 149
Dynamic 114
dynamic large page pool 187–188
dynamic linking 79
dynamic partitiong
processing units 27
dynamic partitioning
processor 25
shared processor partition 22
virtual SCSI 51
weight 27
dynamic reconfiguration 178
cpupstat command 178

E
eav2 138
getea 130
setea 130
Eclipse 161
Eclipse runtime environment 76
Eclipse-based
Procmon 254
Eclipse-based tools 76
ed/ex/vi 251
ed/ex/vi/awk 251
edit subcommand 88
e-fix upgrade 76
ELF file format 78
emgr command 189

380


enable subcommand 88
encrypt 206
Enhanced DBX functionality 87
Enhanced Journaled File System
disk quotas 114
Enhanced libc.a 70
entitled capacity 7
EntProcCapacity, regarding IBM.Host 344
environment
CORE_NAMING environment 145
environment file 197
environment variable
GPROF 91
epkg command 189
Erasure 163
errdemon 142
errlog() function 142
errlog.save file 142
error log
CORRUPT_LOG entry 142
DUMP_STATS entry 149
error log hardening 142
LOG_COPY_FAILEDentry 142
LOG_COPY_IO entry 142
Error messages 73
Error reporting
lpNet 247
lpsched 247
EtherChannel 43
Etherchannel 218
Event Management 344
executable’s read/write sections 81
Expansion pack 319
expired, virtual processor state 7
exportvg command 113
extended attribute version 2 138
Extended Log 73
EXTENDED_HISTORY 251
extendlv command 108
extendvg command 96, 112
external networks 40, 43
routing 40
Shared Ethernet Adapter 40
EZNIM 218

F
F_NONEXEC bit in XCOFF 81–82
failed dump 147

fast path 258
fatal error 89
fatal flaws 73
fd 89
fd subcommand 89
file
///usr/mozilla/base/README.HTML 250
File size 85
file stream 73
file subcommand 88
file system check 129
file system I/O pacing 281
file system shrink 123
filesets
bos.suma 165
find command 250
firewall 197
fixmgr command 189
fixpkg command 189
fragmentation 72
free pool 22
fsck 129
fsck command 129
FULLCORE 90
functions
authenticatex() 71
chpassx() 71
ctime64() 247
ctime64_r () 247
ctime64_r() 247
difftime64() 247
dlopen() function 78
dlsym() function 78
errlog() function 142
getgroupattrs() 71
getuserattrs () 71
gmtime64() 247
gmtime64_r() 247
localtime64_r() 247
loginrestrictionsx() 71
mktime64() 247
passwdexpiredx() 71
perfstat_partition_total() function 263
pthread_create() function 83
pthread_create_withcred_np() function 82
putgroupattrs () 71
sctp_opt_info() function 310
sctp_peeloff() function 310
simple_lock() function 84

socket() function 309
vmgetinfo() function 86
fuser command 251

G
GB18030 355
GB18030-2000 codeset 355
gencopy command 214, 216
gencore command 146
geninstall command 192
getconf 177
getea 130
getgroupattrs() 71
getuserattrs () 71
global_blocked_io_count parameter 111
global_pbuf_count parameter 110
glyphs 354
gmon.out
process level interpretation 91
thread suport 87
gmtime64() function 247
gmtime64_r() function 247
GNOME 161
GNOME libraries 248
GPROF 91
gprof 87
gprof command 91
grep command 251
gssd daemon 288
guard page 74
Gujarati language 354
Gujarati NLS enablement 354

H
handler subcommand 89
HANIM 207
Fallback 213
Takeover 212
HARD limit 85
hash-based 296
hcall, hypervisor call 5
cede 7
confer 7
prod 7
HDECR, hypervisor decrementor 6
header of a corefile 90
heap 72
heap lock 73

Index

381

heap size 86
Hindi 354
histogram 91
HMC 350
Host resource class 344
Host Resource Manager 344
hosted partition 47
hosting partition 47
hypervisor decrementor, HDECR 6
hypervisor mode 6

I
I/O pacing 281
IBM Directory 327
IBM.Host resorce class 344
IBM.LPCommands class, regarding LPRM 348
IBM.Processors resource class 344
ICMPv6 297, 303
Identification and Authentication 70
IEEE 802.1Q VLAN 13, 34
IETF 308
IETF draft 308
importvg command 96–97, 113
increased number of arguments 251
increased performance 72
InfoCenter 160
inheritance 138
Initial login license limit 340
initiator, vSCSI 47
inline log 125
in-memory channel 37
Installing Mozilla for AIX 248
Interim Fix Management 188
fixmgr command 189
fixpkg command 189
Internet Engineering Task Force 308
inter-partition communication 32
interpreter threads 77
interval accounting 234
ioo command 109
pv_min_pbuf parameter 111
iostat command 250, 256, 258, 261
iostat.h file 261
IP fragmentation 42
IP protocol layer 308
IP Security 313
IPC limits 86

382


message queues 87
semaphores 87
shared memory segments 87
ipforwarding 40
IPPROTO_SCTP 309
iptrace/ipreport 90
IPv4 309
IPv6 299, 303
DHCPv6 299
ICMPv6 297, 303
ithreads 77

J
j2edlimit command 118
Japanese Industrial Standard
JISX0213 355
Java 92
Java 1.4.1 92
Java 1.4.2 92
Java 3D version 1.3.1 92
JavaTM Web Start 357
JetDirect software 247
jfs2
extended attributes version 2 129
file system shrink 123
logredo 128

K
kadmin
ktadd 292
kadmin.local 291
Kazakh language 354
Kazakh NLS enablement 354
Kazakhstan 354
kdb command 283
kdb kernel debugger 283
KDB subcommands
ifnet 311
kmbucket 311
ndd 311
netm 311
nsdbg 311
sock 311
sockinfo 311
tcb 311
tcpcb 311
udb 311
KDC 289

KDE 161
Kerberos 289–290
config.krb5 290
kadmin.local 291
KDC 289
kinit 292
krb5.client 290
krb5.server 290
modcrypt.base 290
Kernel round robin locks 84
kernel threads 262
Key Distribution Center 289
key-based searches 74
kinit 292
krb5.client 290
krb5.server 290
krlock 84
ksh 251
ksh93 251
ktadd 292
kthread subcommand 89

L
large page pool 187
late staggering 86
latency, virtual processor 30
layer 2 bridge 40, 42
LCD display 149
ld command 79, 81
LDAP 341
prioritizing LDAP servers 339
ldap 326
authentication 326
distinguished name (DN) 326
IBM Directory 327
mksecldap 328
proxyuser 328
secldifconv 329
ldap.cfg file 339
ldedit command 82
Least privilege resource manager 345
Legacy AIO 256
lgpg_regions attribute 187
lgpg_size attribute 187
libcap library 90
libpthreads.a 89
licensed software components 3
licensing 340

ligatures 354
limitations and considerations 51
Micro-Partitioning 30
Partition Load Manager, PLM 57
Virtual Ethernet 39
virtual SCSI, vSCSI 51
limits classes 114, 116
limits file 84
linker 78
list subcommand 88
loaded module information 88
loaded modules 88
locale support 354–355
localtime64() function 247
localtime64_r() function 247
logical states, virtual processor 7
logical track group 111
logical volume
striped column support 106
Logical Volume Manager 111
command performance 96
chlvcopy command 96
extendvg command 96
importvg command 96
lslv command 96
lspv command 96
mklvcopy command 96
mkvg command 96
varyonvg command 96
pbuf pools 109
LogicalId, regarding IBM.Processors 344
login permissions 341
login.cfg file 341
loginrestrictionsx() 71
logredo 128
long user and group name 175
lowest utilization 296
lparstat, command 7, 25, 27
lpd printing daemon 247
lphistory command 347
lpNet 247
lpNet command 247
lpp_source 214
LPRM 345
commands 345
security 348
lpsched command 247
lquerypv command 113

Index

383

LRDMA,Logical Remote Direct Memory Access 48
lscore command 144
lslpcmd command 346
lslpp command 189
lslv command 96
lspv command 96, 113
lsvg command 104
LTG 111
lvmo command 109
max_vg_pbuf_count parameter 110
pervg_blocked_io_count parameter 110
pv_pbuf_count parameter 110
total_vg_pbufs parameter 110

M
MAC address 37, 41
make command 250
malloc allocations 74
malloc allocator 72
Malloc Buckets 72
Malloc Log 73–74
malloc subsystem 72
Malloc Trace 73
malloc() 72
MALLOCBUCKETS 73
MALLOCDEBUG 73
=continue 73
=output
/dev/null 73
stdout 73
MALLOCDISCLAIM 73
MALLOCMULTIHEAP 73
MALLOCOPTIONS 73
MALLOCTYPE 72–73
man command 92
man command presentation enhancement 92
MANPATH 251
Manual Pages 251
map subcommand 88
mappings 88
MAX REQUEST 113
max_logname 177
maxdata 77
maxf statistics 259
maxg statistics 259

384


maximum transfer size 111
maxpout 281
maxr statistics 259
MD5 algorithm 76
Memlock 60
memory fragmentation 72
memory management, PLM 56
Memory size 85
Message Passing 68
message queue identifiers 87
message-digests 76
metadata 73
mib 316
Micro-Partitioning
dedicated memory 16
dedicated processor partition 16
dedicated processors 18
entitled capacity 7
Firmware enablement 2
introduction 14
overview 15
processing capacity 16
processing units 16
virtual processors 18
minpout 281
mismatched libraries 89
mismatched module 89
mkdvd command 195
mklpcmd command 346
mklv command 108
mklvcopy command 96, 108
mksecldap 328
mksysb 214
mktime64() function 247
mkuser command 84
mkvg command 96, 100, 112
mmap() subroutine 75
mnemonics 359
modcrypt.base 290
mount command 282
mounted 88
Mozilla 248
mpio 143
mpr_policy 295
mpstat, command 11
msgget() subroutine 87
msync() subroutine 75

multi 91
multi-column tables for man command 93
multiheap 73
Multipath routing 294
multipath routing
hash-based 296
lowest utilization 296
random 296
weight 296
weighted random 296
weighted RR 296
Multiple Desktop 161
multiprocessor kernel 250
multithread 91
multi-threaded applications 72, 91

N
NAT 313
NDP, Virtual Ethernet 41
netgroups 341
Netscape Communicator 248
Network Address Translation 313
Network Install Manager 214
network layer 308
networking enhancements 285
new flags 90
new flags for tcpdump 90
New malloc() 72
NFS kernel extension 319
NFS V4 286
/etc/exports 289
chnfsdom 288
chnfsim 289
gssd daemon 288
NFS V4 ACL 286
nfs4cl 288
nfshostkey 288
nfsrgyd daemon 288
RPCSEC-GSS 286
NFS4 ACL 139
nfs4cl 288
nfshostkey 288, 293
nfsrgyd daemon 288
NIM 206, 214, 218
backup master 209
HANIM 207
High Available NIM 207
install options 218

lpp_source 214
lpp_source resource 218
NIM Security 195
primary master 209
SPOT 214–215
SPOT copy 214–215
NIM client 196
nim command 203, 212, 214–216
show_progress attribute 214
NIM cryptographic authentication 195, 203
NIM master 196
NIM Security 195
NIM service handler 195–196
nim_master_recover command 211, 216
NIM_SECONDARY_PORT 197
nimadapter command 218
nimclient -C command 195
nimclient command 198
nimclient commands 215
niminfo file 198, 217
niminit command 210, 218
NIMOL, NIM on Linux 4
nimquery command 198
nimsh 195–196, 203, 218
nimsh.log log file 199
nimshd 197–198
NIS 341
passwd.adjunct 341
NIS/NIS+ enhancements 341
NLS enablement
Gujarati 354
Kazakh 354
Tamil 354
no
ifsize 304
mpr_policy 295
no command
ndd_event_name 313
ndd_event_tracing 313
net_malloc_size 313
net_malloc_type 313
nohup command 252
noswitchgid 84
noswitchuid 84
notification message 89
not-runnable, virtual processor state 7
npsrpgmax parameter 182
npsrpgmin parameter 182
npsscrubmax parameter 184

Index

385

npsscrubmin parameter 184
NumActPProcessors, regarding IBM.Host 344
NumOnVProcessors, regarding IBM.Host 344
NVRAM 147

O
onceblock subcommand 89
Open files 85
open() subroutine 75
OpenSSL 203, 206
optimal LTG size 112
ordering, Advanced POWER Virtualization feature
2
output 254
output debug option 73

P
packet timestamps 90
paging space garbage collection 180
paging space scrubbing 184
re-pagein garbage collection 181
tuning and control 185
rpgclean parameter 182
rpgcontrol parameter 182
scrub parameter 184
scrubclean parameter 184
vmo command 182
paging space scrubbing 184
PAM 329
PAM modules 334
pam_aix module 335
pam_allow module 335
pam_allowroot module 335
pam_ckfile module 336
pam_getenv subroutine 339
pam_getenvlist subroutine 339
pam_permission module 336
pam_prohibit module 337
pam_putenv subroutine 339
pam_rhost_auth module 338
partition isolation 12
donor, PLM 54
software license charge 3

386


Partition Load Manager, PLM 2
entitlements 54
introduction 52
memory management 56
ordering 2
partition priority 54
policy file 54
processor management 56
requester 54
thresholds 54
PASSNO system environment 153
passwd.adjunct 341
passwdexpiredx() 71
Path MTU 297
Pattern 163
pax command 150–151
pax format 150–151
pbuf pools 109
ioo command 109
lvmo command 109
vmstat command 109
performance 72
performance considerations
EtherChannel 43
Virtual Ethernet 38
virtual SCSI 51
performance degradation 320
performance improvement 320
Performance of lpsched 247
perfstat_cpu_t structure 263
perfstat_cpu_total_t structure 263
perfstat_partition_total() function 263
perfstat_partition_total_t structure 263
Perl 5.8.2 76
perl command 76
perl thread 77
perl.man.en_US fileset 77
perl.rte fileset 77
perl581delta 77
perldelta 77

perldoc command 77
PerlIO support 77
pid, regarding PMAPI 262
PLM
Partition Load Manager 2
Pluggable Authentication Module framework 329
/etc/pam.conf configuration file 332
/etc/security/login.cfg configuration file 331
auth_type attribute 331
PAM Application Programming Interface 339
pam_getenv 339
pam_getenvlist 339
pam_putenv 339
PAM modules 334
pam_aix 335
pam_allow 335
pam_allowroot 335
pam_ckfile 336
pam_permission 336
pam_prohibit 337
pam_rhost_auth 338
PAM-enabled AIX commands 331
requisite PAM control option 333
RFC 86.0 330
X/Open Single Sign-on 330
PMAPI pthreads 262
pmtu 297–298
poll() 320
Pondicherry 354
Port virtual LAN ID, PVID 34
POSIX
Realtime functions 60–69
Realtime options 60
Advisory Info 60
Barriers 60
Clocks 60
Memlock 60
Message Passing 60
Priority Scheduling 60
Semaphores 60
Shared Mamory Objects 60
Spin Locks 60
Thread Options 60
Timers 60
POSIX AIO 256
posix_aioserver 258
POWER Hypervisor
introduction 5

processor dispatch 6
tasks 5
virtual Ethernet 13
virtual I/O adapter types 12
virtual I/O implementation 12
virtual I/O operations 12
virtual SCSI 13
virtual TTY console support 14
primary dump device 150
printer enhancements 247
Priority Scheduling 64
probemgr command 351
proc subcommand 89
process level interpretation
gmon.out 91
Process Utilization Resource Register 263
processing capacity 16
processing units 16
processor dispatch concepts 9
processor dispatch mechanism 10
processor management 56
Processor resource class 344
Processor Utilization Resource Register, PURR 6
Procmon 254
actions on processes
kill, renice, svmon, proctols 254
procmon
output 254
Procmon - Process monitoring tool 254
procmon command 76
prod, hcall 7
profiling information 87, 91
projctl command 230
proxyuser 328
ps command 251
PSSP 345
pthread 82
pthread information 89
pthread_create() function 83
pthread_create_withcred_np() function 82
pthreads 262
ptid, regarding PMAPI 262
PURR 263
PURR, Processor Utilization Resource Register 6
putgroupattrs () 71
PVID, Port virtual LAN ID 34

Index

387

Q
quota command 118
quotacheck command 117
quotaoff command 118
quotaon command 117

R
RAID 106
random 296
RDMA, Remote Direct Memory Access 48
readaacct command 237
reclaimed 72
recvfrom() 320
Redbooks Web site 372
Contact us xxi
reducevg command 112
redundant array of independent disks 106
Remote Direct Memory Access, RDMA 48
re-pagein garbage collection 181
report_allocations 73
repquotas command 118
request for an I/O 111
Request for Comments
RFC 86.0 330
requisite control option 333
resource limit values 84
Resource Manager 344
resource manager 345
resource overhead 72
resource subcommand 89
restore command 251
restricted version
ksh 251
ksh93 251
RFC 2960 308
RFC 3057 308
RFC 3286 308
RFC 3309 308
RFC 3331 308
RFC 3332 308
RFC 86.0 330
RMC 345, 348
rmlpcmd command 347
RPCSEC-GSS 286
rpm filesets 248
rpower command 349

388


rpower -w command 349
RSCT 344
Rsh support 339
runlpcmd command 346
runnable, virtual processor state 7
running, virtual processor state 7
RWNONEXEC 81–82

S
sar -d command 261
scalable volume group 97
chvg command changes 102
lsvg command changes 104
meta data structure 98
mkvg command changes 100
scaling_factor 91
SCRIPTLOG system environment 153
SCRIPTSIZE system environment 153
scrub parameter 184
SCSI RDMA 48
SCTP 308
SCTP API in AIX 309
SCTP management 310
sctp.h file 309
sctp_opt_info() function 310
sctp_peeloff() function 310
sctpctrl command 310
search keyword 320
secldifconv 329
secondary dump device 150
secondary port 196
Security
NIM Security 195
segfault 74
semaphore identifiers 87
Semaphores 65
semget() subroutine 87
sendto() 320
server adapter, vSCSI 47
Service Location Protocol 306
Service ports 196
Service Update Management Assistant 165
bos.suma fileset 165
SMIT user interface 172
suma command 171
services file 196
setea 130

sethostent(1) 320
SHA-1 algorithm 76
shadow passwords 341
/etc/security/passwd 341
external networks 40
performance considerations 42
Shared Memory Objects 64
shared memory segment identifiers 87
shared processing pool 18
shell histories enhancements 251
shmget() subroutine 87
signaling transport 308
SIGXCPU signal 85
simple_lock() function 84
Single thread trace 156
SLP 306
SMB 319
SMIT
disk quotas support 119
smit change_documentation_services 249
smit install_bundle 249
smitty change_documentation_services 249
smitty install_bundle 249
snap command 149, 151, 251
external scripts 151
snap.pax.Z file 153
SNAPDIR system environment 153
snapscripts directory 151
snapshots 128
snapsplit command 153
snmp 316
IBM Netview 316
mib 316
snmptrap 318
subagent 316
SOCK_SEQPACKET 309
socket 309
socket() 320
socket() function 309
socket.h file 309
sockets 308
SOFT limit 85
software bundles 189
lslpp command 189
PLM 3

software maintenance
installed filesets by bundle 189
lslpp command 189
Service Update Management Assistant 165
source code locations 88
SP switch 90
speed performance 72
Spin Locks 62
spinlocks 90
SPOT 214–215
SPOT copy 214–215
SSA support
Stack size 85
static linking 79
statistical interface 261
statistics
Asynchronous I/O statistics 256
avfc statistics 259
avgc statistics 258
avque statistics 262
avserv statistics 261–262
avwait statistics 261–262
maxf statistics 259
maxg statistics 259
maxr statistics 259
stderr 73
stop subcommand 88
Stream Control Transmission Protocol 308
striped column 106
chlv command 108
extendlv command 108
mklv command 108
mklvcopy command 108
structures
__pthrdscreds structure 83
dkstat structure 261
perfstat_cpu_t structure 263
perfstat_cpu_total_t structure 263
perfstat_partition_total_t structure 263
subroutines
disclaim() 75
mmap() 75
msgget() 87
msync() 75
open() 75
pam_getenv 339
pam_getenvlist 339
pam_putenv 339

Index

389

semget() 87
shmget() 87
sum command 76
SUMA 165
suma command 171
SVR4 linking affinity 78
syllables 354
synchronous errors 73
Sysctl 345
sysdumpdev command 147, 149–150
verbose flag 147
sysdumpstart command 148
system dump 146, 149
dump device 150
dump information 146
failed dump 147
trace back information 148
verbose flag 147
system environment 153
PASSNO 153
SCRIPTLOG 153
SCRIPTSIZE 153
SNAPDIR 153
System V printer enhancements 247
System V printing 247

xprofiler command 91
thread trace 156
Threadcache 73
threadcache 73
threads
thread based credentials 82
thread-safe 87
tid, regarding PMAPI 262
time_t type 246
TIME_WAIT 196
time32 247
time64_t type 247
Timers 66
trace 156
trace back information 148
trace buffers 154
trace command 156
trace subcommand 88
tracepoints 88
transactional accounting 232
transfer sizes 112
trcctl command 154
trunk flag 41
TTY 146

T

U

tagged packets 34
Tamil language 354
Tamil Nadu 354
Tamil NLS enablement 354
tape support
tar command 251
target, vSCSI 47
TCP 196, 308
TCP socket 320
tcpdump upgrade 90
telephony signaling transport 308
TFTP protocol 205
thread arc execution 91
thread based credentials 82
thread cache front-end 72
thread IDs 73
thread level profiled programs 91
Thread Options 67
thread support
gprof command 91

UDP 308
UDP connection 320
UDP socket 320
UFST 355
ulimit command 85
uncapped mode 17
Unicode 4.0 support 355
Unicode extension 355
uniprocessor kernel 250
uniprocessor kernel support 250
Universal Font Scaling Technology 355
unlimited command buffer sizes 251
unlimited line lengths support 251
untagged packets 34
use subcommand 88
user 73
User Adaptation Layer 308
user customization 88
user defined queues for at jobs 252
user notification 88
user threads 262

390


UTC 246

V
validate_ptrs 73
Variable logical track group 111
varyoffvg command 113
varyonvg command 96, 112
verbosity level 88
VIPA 218
Virtual Ethernet 13
ARP 41
benefits 38
broadcast 41
communication with external networks 36
interpartition communication 35
introduction 32
multicast 41
NDP 41
Virtual Ethernet adapter 37
boot device 38
in-memory channel 37
MAC address 37
transmission speed 38
trunk flag 41
virtual host bridge 49
Virtual I/O Server
ordering 2
virtual LAN
AIX support 34
overview 32
virtual LAN, VLAN 32
Virtual Memory Manager
paging space garbage collection 180
virtual processor 7, 18
latency 30
maximum number 18
move 29
reasonable settings 31
state
expired 7
not-runnable 7
runnable 7
running 7
AIX device configuration 49

architecture 47
CD-ROM support 51
client adapter
client adapter, vSCSI 47
dynamic partitioning 51
hosting partition 47
initiator 47
introduction 46
protocols
protocols, vSCSI 47
server adapter 47
SSA support 51
tape support 51
target 47
virtual TTY console support 14
virtualization systems technologies 1
VLAN, Virtual LAN 32
VM_NEW_HEAP_SIZE 86
VM_STAGGER_DATA 86
vmgetinfo() function 86
vminfo.h file 86
VMM 187
vmo command 86, 182, 187
lgpg_regions attribute 187
lgpg_size attribute 187
scrub parameter 184
vmstat command 109
vmstat -l command 187
Voice over IP 308
VoIP 308
volume group
classical concurrent mode 97
pbuf pools 109
scalable volume group 97
VRMF 218

W
watchpoints 88

Index

391

Watson 73
Watson Malloc 72
weight 296
weight, uncapped 17
weighted random 296
weighted RR 296
wsm 357
JavaTM Web Start 357
mnemonics 359
wsmsvk.bat 358

X
X/Open Single Sign-on 330
XCOFF file format 78
XSSO 330

Y
Yorktown allocator 72

392



(0.5” spine)
0.475”<->0.875”
250 <-> 459 pages

Back cover

®

AIX 5L Differences Guide

Version 5.3 Edition
AIX - The industrial
strength UNIX
operating system
An expert’s guide to
the new release
AIX 5L Version 5.3
enhancements
explained

This IBM Redbook focuses on the differences introduced
in AIX 5L Version 5.3 when compared to AIX 5L Version
5.2. It is intended to help system administrators,
developers, and users understand these enhancements
and evaluate potential benefits in their own
environments.
AIX 5L Version 5.3 introduces many new features,
including NFS Version 4 and Advanced Accounting, and
exploits the advanced capabilities of POWER5 equipped
severs, such as Virtual SCSI, Virtual Ethernet SMT,
Micro-Partitioning, and others. There are many other
enhancements available with AIX 5L Version 5.3, and
you can explore them in this redbook.
For customers who are not familiar with the
enhancements of AIX 5L through Version 5.2, a
companion redbook SG24-5765 is available.

INTERNATIONAL
TECHNICAL
SUPPORT
ORGANIZATION

BUILDING TECHNICAL
INFORMATION BASED ON
PRACTICAL EXPERIENCE
IBM Redbooks are developed by
the IBM International Technical
Support Organization. Experts
from IBM, Customers and
Partners from around the world
create timely technical
information based on realistic
scenarios. Specific
recommendations are provided
to help you implement IT
solutions more effectively in
your environment.

For more information:
ibm.com/redbooks
SG24-7463-00

ISBN 0738491535

AIX 5L Differences Guide Version 5.3 Edition

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to AIX 5L Differences Guide Version 5.3 Edition (20)

More from IBM India Smarter Computing (20)

Recently uploaded (20)

AIX 5L Differences Guide Version 5.3 Edition