5    Base Operating System Software Notes

This chapter contains notes about issues and known problems with the base operating system and, whenever possible, provides solutions or workarounds to those problems.

The following topics are discussed:

5.1    Commands and Utilities

The following notes apply to commands and utilities.

5.1.1    The diskconfig Restriction

You cannot select disk partitions with an offset of zero and a size of zero using the Disk Configuration Manager utility (diskconfig). You must use the disklabel command to assign an offset and size to these partitions.

5.1.2    The vdump Restriction with Clone Filesets and NFS

If you mount a clone AdvFS fileset on a Version 4.x system across NFS to a Version 5.x system (the NFS server is Version 4.x and the NFS client is Version 5.x), you cannot use the 5.x version of vdump to back up the clone fileset. The resulting saveset from the vdump utility may not contain all of the data in the clone fileset. If both the NFS client and the NFS server are running 4.x or are running 5.x, a backup works correctly. If the NFS server is running 5.x and the NFS client is running 4.x, this also works correctly.

5.1.3    APIs and Commands Affected by Mixed-Speed CPUs

This release of the operating system supports CPUs of different speeds on some hardware platforms. Several existing APIs and commands may return incorrect information in a mixed-speed CPU environment. See getsysinfo(2) and psrinfo(1) for more information.

5.1.4    Replacing a Failed HSZ40, HSZ50, or HSZ70 Controller

This section describes how you replace a failed dual redundant controller for the following dual unit models:

If a controller fails, use the following procedure to replace the failed component and restart your configuration.

Caution

Several cautions are specified in the documentation for the controller's HSOF commands. To avoid the risk of data loss, familiarize yourself with the following CLI commands before proceeding:

The Tru64 UNIX operating system must be running when the following procedure is applied. It must not be halted and at the console prompt (>>>).

  1. Using the HSOF CLI on the good (functioning) controller, enter the following command to stop the failed controller: 

    
HSZ21> set nofailover

  2. Verify that the failed controller has stopped by ensuring that the green LED status light is no longer blinking.

  3. Physically remove the failed controller as described in the hardware documentation.

  4. At the UNIX command prompt, enter the following command: 

    # hwmgr -scan scsi -bus N

    The -scan option is asynchronous. When you issue this command, the prompt returns immediately, although the scan can still be working in the kernel, particularly if many devices are connected to the system. To test for completion of a scan, use the Event Manager (EVM) evmwatch command to monitor for an event with the following event name: sys.unix.hw.scan_completed. For example: 

    # evmwatch -A -f '[name sys.unix.hw.scan_completed]'  
A hardware scan has just completed

    The presence of tape devices increases the delay to complete the scan. Use the -bus qualifier to specify the bus you want to scan. Determine the correct bus number by examining the location field of the output from the following command: 

    # hwmgr -view devices

    Messages printed at the console will show the HSZ units going from dual to single configuration.

  5. Install a replacement controller as described in the hardware documentation.

  6. Start the replacement controller as described in the hardware documentation.

  7. Repeat the hwmgr -scan scsi command described in Step 4.

  8. At the good controller, issue either of the following HSOF CLI commands: 

    > set failover copy=this 
> set multibus_failover copy=this

    This will return the controllers to dual redundant mode in the appropriate failover mode.

Messages printed at the console will verify that the operation was successful and will show the HSZ units going from single to dual configuration.

5.1.5    Root Permissions Are Required to Use the Bootable Tape (BT) Application

If you attempt to run either the btcreate or btextract command from an account that does not have root privileges, it might produce a TCL dump script.

To avoid this problem, ensure that you always run the bootable tape commands from an account with root privileges.

5.1.6    Escaped Comment Symbols in a Makefile

The make command does not recognize escaped comment symbols as literal characters in a Makefile. Comment lines that begin with a number sign (#) and all text following this symbol up to the end of the line are considered part of a comment. This is true even if the symbol is preceded with a backslash (\).

5.1.7    Editing an HTML File with XEmacs

If you use XEmacs to edit an HTML file, the editor looks for an entry corresponding to the e-mail ID in an .emacs file. If this file does not exist or if the entry is not found, XEmacs prompts the user for the e-mail ID, and this information is updated in the .emacs file.

5.1.8    Change in the Default root crontab File

In previous releases, the cron daemon periodically cleaned files such as /var/adm/cron/log and /var/adm/messages by default.

The programs that performed these tasks have been removed from the /var/spool/cron/crontabs/root file. Therefore, the cron daemon does not clean up these files by default.

If you want to clean up these files manually, you must first stop the cron daemon and truncate the files. You can stop and restart the cron daemon by using the /sbin/init.d/cron script. If you stop the cron daemon and you have cron jobs scheduled to run during that time, they will not be executed.

5.1.9    Change to the sh, ksh, and csh Shells

The Bourne, Korn, and C shells generate random file names when opening files created using the << construct. In this release, the following changes have been implemented:

If a shell-generated file name conflicts with an existing file, the shell will retry with a different random name. After 100 failed attempts, the shell returns the error Unable to create temporary file.

5.1.10    Netscape Communicator

The following notes apply to the Netscape Communicator product.

5.1.10.1    Netscape Communicator Dumps Core Running in CDE

Netscape Communicator dumps core when the application posts a file selection dialog (XmFileSelectionBox). Typically, this occurs when you run the application in the Common Desktop Environment (CDE) and select the Save As option in the File pulldown menu of the Navigator browser. It can also occur when you select a link to download a file or save an attachment to a mail message in the Messenger Mailbox component.

To avoid this problem, invoke Netscape using the following script: 

/usr/bin/X11/netscape

If you use this script to start Netscape Communicator, the application displays the file selection dialog within CDE without dumping core. Use the -xrm '*nsMotifFSBCdeMode: True' command line option if you start Netscape Communicator using some other means.

For more information, see the Communicator on UNIX release notes at the Netscape Web site:

http://home.netscape.com/eng/mozilla/4.0/relnotes/unix-4.0.html

5.1.10.2    Deleting Multiple Mail Messages Causes Netscape Communicator to Dump Core

Deleting multiple mail messages in Netscape Communicator's Messenger Mailbox component sometimes causes Communicator to dump core. Usually, it requires several multiple deletions of mail to make Communicator dump core. If Communicator does not dump core immediately, deleted messages might reappear in the mail folder from which they were deleted.

5.1.10.3    Netscape Communicator Dumps Core Intermittently

Netscape Communicator intermittently dumps core and returns the following error in the terminal window from which it is started: 

Memory Fault - (core dumped)

This core dump occurs with different hardware and software configurations and under different circumstances. Sometimes it hangs for a time, taking most of the CPU time, then it crashes. At other times, you must kill the process and restart the application. Numerous problems of this nature have been reported. None are resolved at this time and no workaround is available. In all cases, the behavior cannot be reproduced consistently.

5.1.10.4    Cannot Delete Mail Messages from Inbox to Trash When Using IMAP Server

After upgrading from a previous version of Communicator, an IMAP mail user cannot move messages to the Trash folder in the Messenger component. All Delete options in the user interface are insensitive (greyed-out). Setting the Move it to trash folder option in the IMAP mail server preferences window does not work. This behavior is the result of a new feature in Netscape Communicator that might require user customization after upgrading to the latest version.

Starting with Netscape Communicator Version 4.5, the Namespace extension to the standard IMAP protocol is used to locate the users' folders on the IMAP mail server. This feature does not work if you are using an older IMAP server that does not support the Namespace extension to the protocol. Use the following procedure to customize Netscape Communicator to be able to locate a user's Trash folder on an old IMAP server:

  1. Choose the Preferences option in the Edit pulldown menu and choose the Mail and News Servers option in the Preferences window.

  2. Choose the Mail Servers option from the list of Mail and News Servers options.

  3. Choose the IMAP server from the list of servers and click on the Edit button to edit the server configuration.

  4. Choose the Advanced tab in the pop-up dialog box.

  5. Ensure that the Namespace field in the tab reads as follows (quotes and period included): 

    Namespace: "INBOX."

  6. Click on the OK button in the pop-up window and again in the Preferences window to save the settings.

  7. Exit and restart Communicator.

You can now move messages to the Trash folder and use all the Delete options. Because IMAP mail server configurations differ (including the location of the user's folders on the server), check with your IMAP mail server administrator if the preceding procedure fails to resolve the problem.

5.1.10.5    Netscape Communicator Returns sh: /usr/bin/X11/showps: not found

When you select a link to a PostScript file in the Navigator component of Communicator, the following error message might be displayed: 

sh: /usr/bin/X11/showps: not found

The showps helper application has been retired from the operating system as a result of licensing changes to Adobe Display PostScript. The user might have customized the PostScript Document MIME type to use the showps helper application in $HOME/.mailcap and $HOME/.mime.types files.

To resolve this problem, you must obtain a new PostScript viewer and reconfigure the helper application for the PostScript Document MIME type in Communicator. Use the Edit option in the Edit->Preferences->Navigator->Applications pulldown menu of Communicator to edit your PostScript Document helper application and replace /usr/bin/X11/showps with the path to your new PostScript viewer.

5.1.10.6    Netscape Communicator Returns X Authorization Error

When logged in as root, if you start Netscape Communicator on display ":0.0", and your current working directory is not "/" (root), the following X authorization error is returned: 

lib: connection to ":0.0" refused by server
Xlib: Client is not authorized to connect to Server
Error: Can't open display: :0.0

To work around this problem, set the XAUTHORITY environment variable to point to the root user's .Xauthority file (Bourne shell syntax): 

# XAUTHORITY=/.Xauthority
# export XAUTHORITY
# /usr/bin/X11/netscape

5.2    SysMan System Management Applications

The following notes apply to restrictions on using the SysMan system management applications.

5.2.1    Possible Loss of Data When the root File System Is Full

When the root file system is full, Sysman applications might not be able to update the system files. This might lead to loss of data or truncated files. We recommend that you ensure that the root file system is not full or nearly full, prior to running Sysman applications.

Also, the /tmp directory usually resides on the root file system. Therefore, do not run applications that create large temporary files while you are running Sysman applications.

5.2.2    SysMan Account Manager

You cannot delete a user account by using the SysMan Account Manager application while the dxaccounts application is running. If dxaccounts is running and you try to delete a user account by using the SysMan Account Manager application, the following warning message is displayed: 

/etc/.AM_is_running existing

If you continue the delete operation, the Account Manager displays the following error message and the application hangs: 

Error: key userName UID not found in /account_management/local _passwd_table

If this occurs, kill the following process: 

sysmansh /usr/share/sysman/menu/tasks/account_management

5.2.3    Tcl Error Can Occur During DNS (BIND) Configuration

During the DNS client configuration, the following steps might result in an Out Of Order Hide Tcl error:

  1. Enter a domain name in Local Domain.

  2. Add DNS servers.

  3. Click OK in the main window.

  4. Choose Yes to update the system host name to reflect the host name with new domain name.

  5. Choose Yes to add "localhost" to access control list? option.

At this point a Tcl Stack Error can occur. However, the data is not lost.

Because all the data entered by the user is committed by the DNS client application, kill the DNS client application and restart the DNS configuration using the SysMan DNS to avoid this problem.

5.2.4    Large Integer Values in Configuration Applications

Entering a very large integer value (on the order of 1019) in numeric fields in some system configuration applications can cause a stack trace. Such large integers are not appropriate values for these applications. Therefore, this problem is not expected to impede you from configuring your system.

5.2.5    Error Message When Using sysman -cli -set values Command

If you use the sysman -cli -set values command to change specific values for an existing row in the table defined by the staticRoutes group, you might receive an error message. For example: 

# sysman -cli -set values -comp routing -group staticRoutes\"
 
 -attr gateway=1.2.3.4 -key1 "dummy system 1.1.1.1"
Error: "SYSMAN_NO_DATA"
No row exists with the specified key: 'dummy system 1.1.1.1'

If a row with the defined key is present in the staticRoutes group, you can ignore this message. In any case, to verify that the row was modified properly, issue the following command: 

# sysman -cli -list values -comp routing -group staticRoutes

5.2.6    Problem When Reconfiguring Network Interface Cards

If you use a SysMan application to reconfigure a network interface card (NIC) and you change the host name, the HOSTNAME variable in the /etc/rc.config file is not updated.

If the system has a single network interface card, you can correct this problem by performing the steps in the following procedure. If the system has more than one network interface card and you changed the host name of the primary card (that is, the card with the same host name as the system's host name), do the following to correct the problem:

  1. Use the rcmgr set HOSTNAME command to set the HOSTNAME to the correct name. For example: 

    #  rcmgr set HOSTNAME abcxyz.com

  2. Use the hostname command to change the host name to the correct value in the kernel. For example: 

    #  hostname abcxyz.com

  3. Use the xhost command to add localhost to the access control list of the local X server, as follows: 

    #  xhost + localhost

5.2.7    SysMan Command Line Interface

The -mib option of the SysMan command line interface does not work properly. The /usr/sbin/sysman -cli -mib command generates a Management Information Base file for use by SNMP applications. If you use this command, the following error message is displayed: 

ERROR: inrange longer than outrange

5.2.8    SysMan Web Page Does Not Display or Work Properly

If you navigate to the SysMan web page from the default web page /usr/doc/netscape/Welcome.html, the SysMan page may not display or function correctly. (For example, downloads may not work.)

A workaround for this problem is to navigate to this page by opening http://localhost:2301 and selecting the Tru64 UNIX SysMan icon.

5.2.9    SysMan Configuration and Administration Restriction

The sysman configuration and administration utility does not work on hardware configurations within the following locales:

To avoid this problem, set the LC_ALL and LANG environment variables to C when you run the sysman utility.

5.2.10    NTP Configuration Restriction

If you edit the /etc/ntp.conf file manually and subsequently run the SysMan NTP client configuration utility, your changes might be lost.

The SysMan NTP client configuration utility understands only a small subset of the commands that can be used in the ntp.conf file. When the NTP client configuration utility reads the /etc/ntp.conf file, it ignores commands it does not understand and it does not output those commands when rewriting the file. It also does not allow you to enter commands it does not understand. For example, it does not allow you to enter commands using the sysman -cli command.

If you want to configure your system as an NTP sever or your configuration requires a more complex ntp.conf file than SysMan can produce, edit the ntp.conf file manually and do not use the SysMan utility to modify it. For more information, see the Network Administration: Services manual, ntp.conf(4), and ntp.conf(4).

5.2.11    Configuring an NFS Server

When configuring an NFS server, you cannot leave the Number of TCP server Threads field and the Number of UDP Server Threads field blank. If you leave either field blank, the configuration application will stack trace. You must specify a number for each field so that the combined sum is greater than 0 and less than or equal to 128.

5.2.12    Display Problems on Systems with Low-Resolution Graphics Cards

If your system has a low-resolution graphics card and you use large fonts, the images of some graphical applications might be displayed longer than the display area on the monitor. Images that are larger than the display area are truncated at the bottom, often resulting in the buttons being cut off. Some windows in Quick Setup have exhibited this behavior.

To avoid this problem, reduce the size of the font. See the documentation for the window manager you are using. Also, applications with a curses (character) mode fit the screen better than graphical applications using large fonts.

For more information, see X(1X), dtstyle(1), curses(3), and sysman_intro(8).

If you are running these applications in Asian locales, see Section 5.2.13.

5.2.13    SysMan Applications Do Not Work Properly at VGA Resolution

If your monitor is set to VGA resolution (640x480), Quick Setup and other SysMan applications might not work correctly. The problem is that the large fonts used by some locales (for example, Asian locales such as Chinese and Japanese) cause dialog boxes to be larger than 640x480, and control buttons such as OK and Cancel are no longer accessible. This is a more specific problem than that described in Section 5.2.12, and the workarounds in that note do not work for this problem.

To avoid this problem, you have the following options:

5.2.14    SysMan Menu

The following notes apply to the SysMan Menu application. Also see Section 8.8.1 for information related to online help.

5.2.14.1    Installation Branch Hangs When Run in Background

The Install software, List installed software, and Remove installed software tasks in the Installation branch of the SysMan Menu hang if you run the SysMan Menu in the background. Do not run the SysMan Menu in the background if you plan to use these tasks in Installation branch.

5.2.14.2    Installation Branch Is Not Supported for Clusters

Do not run the SysMan Menu Installation branch in a cluster environment. The Install software, List installed software, and Remove installed software tasks in the Installation branch of the SysMan Menu do not work on a cluster system.

5.2.14.3    Running the SysMan Menu Standalone on a PC Fails to Launch Tasks

When you run the SysMan Menu from a PC, you might encounter the following problems:

To avoid these problems, run the SysMan Menu from within the SysMan Station by doing the following:

  1. Start the SysMan Station either from the Start menu or from a web browser.

  2. Choose the Hardware view.

  3. Right click on a host icon and choose the SysMan Station.

5.2.14.4    Manage Local and NIS Users

The Manage local users and Manage NIS users tasks on the SysMan Menu are front ends for the useradd, usermod, and userdel commands. If you use the Manage local users and Manage NIS users tasks to change the /etc/passwd file, a warning message is displayed, even though the operation completes successfully. However, the changes do not display in the Manage local users and Manage NIS users tasks dialog boxes. This happens in the following cases:

To correct this problem, exit the task and restart it. The correct attributes for the user will be displayed.

5.2.14.5    Display Currently Mounted File Systems

The SysMan Menu Display Currently Mounted File Systems option works only in the English version of the operating system. (You can access the Display Currently Mounted File Systems option from the SysMan Menu by selecting Storage --> File Systems Management Utilities --> General File System Utilities.)

To avoid this problem, set the LC_ALL environment variable to C before running the sysman command. This forces the SysMan Menu to work in English. Restore the LC_ALL environment variable to its prior setting when you are done using the SysMan Menu.

5.2.15    SysMan Station

The following notes apply to the SysMan Station. Also see Section 8.8.2 for information related to online help.

5.2.15.1    Incorrect Launch Status

SysMan Station checks the status returned by all the applications that it launches. A few applications incorrectly exit with nonzero (failure) status returns even though the tool has launched successfully.

5.2.15.2    Objects Might Not Display Properly with Internet Explorer

Objects might not display properly in SysMan Station View windows when you run SysMan Station from a PC using the Internet Explorer web browser. Sometimes objects are overlayed in the upper left corner of the display window.

To correct this problem, choose the Show All option from the Action menu to redraw the display properly.

5.2.15.3    Cannot Restart the Client in a Web Browser

When you run the SysMan Station client from a web browser, if you exit the SysMan Station and attempt to restart it by returning to the URL (http://your_machine:2301), the client will not restart.

You can correct the problem by restarting the browser. The client will load properly from the URL.

5.2.15.4    Client 5-Minute Timeout

If you restart the SysMan Station daemon (smsd) while there are active SM Station clients, the clients will keep an active network connection that times out after 5 minutes have elapsed. Attempts to use the /sbin/init.d/smsd start command during this 5-minute interval will fail to restart the server because it cannot access the required network port. You must wait for the 5-minute timeout to elapse before you can restart the SysMan Station server.

You can use the following command to see if any clients are using the network port: 

# /usr/sbin/netstat -a | grep 596

If no matches are found, you can restart the SysMan Station daemon.

5.2.15.5    Physical_Filesystems View Displays Two Disk Objects for LSM

Two disk objects are displayed in the Physical_Filesystems view for each file system that uses LSM. One disk object represents the LSM private region, the other represents the LSM public region.

5.2.15.6    Icons Indicating Warning or Failed States

Objects in a failed or warning state are depicted in the SysMan Station's Hardware view using a red or yellow highlight for the object's icon. A small number of objects do not have warning or failed icons. In this case, the object's label correctly indicates that it is in a warning or failed state.

5.2.15.7    Group Icons Are Not Available for Some Objects

When objects are grouped together, a special group icon is used to represent the grouping. A small number of objects do not display a group icon when an object group is formed. In these instances, the group's label correctly indicates that the icon represents a group.

5.2.15.8    Multiple AdvFS Volumes Might Not Appear Properly

When multiple volumes are added to AdvFS file domains, the new AdvFS volume objects might not appear in the SysMan Station AdvFS Filesystem and Physical Filesystem view windows.

You can correct this problem by restarting the SysMan Station daemon (smsd). To restart the smsd daemon, exit all connected SysMan Station client sessions and issue the following command: 

 # /sbin/init.d/smsd restart

5.2.15.9    SysMan Station Does Not Update Changes to Device base_name Dynamically

If you use the dsfmgr utility to change the base_name of a device while the SysMan Station daemon (smsd) is running, any clients connected to the running daemon will continue to reflect the old base_name in their views.

To obtain a correct view, close all open SysMan Station sessions and restart the SysMan Station daemon with the following command: 

# /sbin/init.d/smsd restart

If the system is a member of a cluster, perform these steps on all affected cluster members.

5.2.15.10    Some Tools Might Fail When Launched From a Group Object

If you launch a tool from a SysMan Station group object, the SysMan Station attempts to invoke an instance of the tool for each object contained in the group. For certain tools, not all instances of the application will launch successfully. An application error or SysMan Station Authentication server core dump might result.

To work around this problem, do the following:

  1. Close all open SysMan Station sessions.

  2. Enter the following commands: 

    # /sbin/init.d/smauth restart
# /sbin/init.d/smsd restart

On a cluster, perform these steps on all affected cluster members.

5.2.15.11    Unable to Expand the Host Object in a Cluster

The Sysman Station client might occasionally encounter a Java class exception error when a user attempts to expand a Host object.

If you encounter this problem, restart the SysMan Station client and retry the expand operation.

5.2.15.12    Unable to Select Other Objects After a Vertical Bus Is Selected

If you select a vertical bus within the SysMan Station Hardware view, you are unable to select any other object in the view. If this problem occurs, exit the view and then return to it, or restart the client.

5.2.16    Configuring Tru64 UNIX from Linux Systems

You can now configure Tru64 UNIX from systems running Linux using Java. The following procedure describes how to install the SysMan client on these systems:

  1. Download the /usr/share/sysman/web/classLib/suit.jar file from the Tru64 UNIX system to your Linux system and add the full path of that file to your CLASSPATH environment variable.

    If you use csh and you downloaded the suit.jar file to /usr/local/lib, use the following syntax: 

    setenv CLASSPATH $CLASSPATH:/usr/local/lib/suit.jar

    If you use ksh and you downloaded the suit.jar file to /usr/local/lib, use the following syntax: 

    CLASSPATH=$CLASSPATH:/usr/local/lib/suit.jar
export CLASSPATH

  2. Run SysMan Menu with the following command, substituting the name or IP address of your Tru64 UNIX computer for HOST: 

    # java suit HOST sysman

    Alternatively, you can run a SysMan task directly by substituting the accelerator for sysman. For example: 

    # java suit HOST ntp_config

You can redirect the standard output to /dev/null if you do not want to see the diagnostic messages that SysMan prints when run in this fashion.

SysMan has been tested on RedHat Linux Version 6.0 and SuSE Linux Version 6.0 on Intel using Version 1.1.7 of the Java Run-time Engine (JRE). Other versions of Linux and Java might also work.

5.3    System Administration

The following notes apply to system administration.

5.3.1    Binary Error Log Event May Be Incorrectly Reported As Double Error Halt

A binary error log (binlog) event with type 113 is reported as a Double Error Halt event when reported by the Event Manager (EVM), but is reported as a Console Data Log event by Compaq Analyze. EVM might report this event by mailing it to the root user and by displaying it on the system console.

The event is actually a Console Data Log event. This event type is posted when any of several different errors occur, including double error halts, uncorrectable environmental errors, and platform-specific system faults. See the the Compaq Analyze translation of the event for information about its cause.

5.3.2    Boot Sequence Stops in Single-User Mode After Core Dump

A problem exists that causes the boot sequence to stop in single-user mode and display the following message: 

/sbin/dn_setup: 1048647 Memory fault - core dumped
bcheckrc: Device Naming failed boot configure or verify.
Please correct the problem and continue or reboot
 
INIT: SINGLE-USER MODE
#

This problem is most likely to occur every time you boot systems with Fibre Channel devices or once if you have changed the hardware configuration between boots.

The problem occurs after the dsfmgr command successfully completes during the exit cleanup routines.

Data integrity is not compromised and no corrective action is required, except to remove the core file from the root directory (/core or /core.dsgmgr*).

You can continue the boot process by pressing [Ctrl/d] to exit single-user mode or by initiating multiuser mode with the init 3 command. Alternatively, you can reboot the system, provided no Fibre Channel devices are connected to the system.

5.3.3    Account Manager

The following notes apply to the Account Manager (dxaccounts) applications.

5.3.3.1    General Restrictions

The Account Manager has the following restrictions on both base security and enhanced security (C2) systems:

5.3.3.2    Account Manager and Enhanced Security

The following problems apply to the Account Manager application when running on systems with enhanced security:

5.3.4    EISA Configuration Utility Revision Requirements

For Tru64 UNIX and its software supplements, the supported version of the EISA Configuration Utility (ECU) is Version 1.10 or higher. If your system is configured with an EISA bus, update the ECU to this supported version.

5.3.5    Alternate Root Installation May Change Host File Dates

During an alternate root installation of base operating system subsets, such as is done using the dmu utility to set up a Dataless Management Services environment, the file access dates on some of the files in the host server's file system might be changed to correspond to those from the subset's file inventory. When the release installed into the alternate root is different from that installed on the host system, these changed dates appear invalid because they may be newer (or older) than the actual file dates from the host system's installation kit.

This occurs when the pax utility is invoked by the setld utility to copy symbolic links from the kit subsets, and the symbolic links target absolute paths that correspond to actual files in the host system's file system. The pax utility attempts to adjust the dates for the symbolic link, but the file system actually adjusts the dates for the target of the symbolic link.

The changed dates have no operational impact on the host system. The content of the affected files is not changed. However, because the dates have changed, the behavior of utilities that examine file dates (such as the find command or archivers) might be affected.

5.3.6    Prestoserve Restriction on Devices Larger Than 1 TB

The Prestoserve product has a restriction in its ability to cache blocks on a device that is larger than 1 TB. This applies both to physical devices (RAID) and logical devices (LSM). Any device that is larger than 1 TB can be cached via the Prestoserve product but blocks with an address that exceeds the 1 TB boundary will not be cached.

5.3.7    Possible Error Updating /etc/doprc

If you use the dop -W command and the following error occurs, unset the DISPLAY variable: 

Error occurred trying to update /etc/doprc:
        Dop_write_actions verbose ; Dop_update_binary /etc/doprc
        child process exited abnormally

5.3.8    Persistent Reservation Error

Under certain circumstances, you can encounter persistent reservation problems. If this occurs, see Appendix B.

5.3.9    Security

The following notes have to do with system management and security.

5.3.9.1    Authentication Problem with Multihreaded Applications

Third-party applications that perform user authentication or impersonation from multiple threads, such as PMDF, will correctly verify a user's group membership only from the first thread. All other threads that call the sia_get_groups routine receive a failure status. This can lead to seemingly random behavior, in which a user's membership in a group of which the user is a legitimate member is sporadically denied.

5.3.9.2    Security and Compaq Management Agents

The Compaq Management Agents for Tru64 UNIX is configured by default when you install the operating system. Anonymous login to WebAgent applications, enabled by default, allows nonprivileged users to invoke the Management Agents and view details of any connected devices in the local area network, although users cannot perform any operations unless authorized. Many site security policies strongly discourage such anonymous access. To disable these agents, use the Compaq Management Agents Configuration icon on the Monitoring/Tuning menu.

5.3.9.3    Behavior of useradd, usermod, and userdel Commands

The useradd command correctly honors the default administrative lock value found in the /.sysman/Account_defaults file. If the Account_defaults file does not exist, the internal default for the useradd command is to create locked accounts. Use the administrative_lock_applied extended command line option to override the default. In the following example, the useradd command creates a locked account for foo regardless of the default value for administrative lock: 

useradd -x administrative_lock_applied=1 foo
 

For base security, a locked account has the text Nologin in the password field in the /etc/passwd file. If an account is unlocked and has no password, that account has no value in the password field. The account is open and accessible to anyone. A warning is displayed if an unlocked account with no password is created.

For enhanced security, all accounts have an asterisk (*) in the password field in the /etc/passwd file, but the lock flag in the protected password database is correctly set to reflect the lock status. As with base security, an unlocked account with no password is accessible to anyone.

The usermod command correctly sets the lock flags for enhanced security when the administrative_lock_applied option is given on the command line. If you use the usermod command to unlock a locked account with no password, a warning is displayed.

The userdel command will either retire or remove accounts on a system running enhanced security.

5.3.9.4    Prevent IP Spoofing Attacks

To detect and prevent an IP spoofing attack that can potentially result in a denial of service, configure the ifaccess.conf file to disable localhost as a source address.

For all adapters except the local loopback adapter (lo0), disable incoming packets with a source address of localhost (127.0.0.1). For example, add the following entry to the /etc/ifaccess.conf for tu0: 

tu0     127.0.0.1   255.255.255.255  denylog
 

Then enable access filtering on tu0: 

# ifconfig tu0 filter

Note that localhost is now disabled by default for new installations.

5.3.9.5    Use db_checkpoint for Log Trimming

A customized version of the Berkeley Database (Berkeley DB) is embedded in this version of the operating system to provide high-performance database support for critical security files. The database includes full transactional support and database recovery, using write-ahead logging and checkpointing to record changes.

The secconfig utility enables you to create a cron job to perform log file trimming; that is, to delete log files no longer involved in active transactions.

The db_archive utility requires a log file checkpoint to determine when a log file is no longer in use. Under some circumstances, security activity may not generate checkpoints for long intervals. Therefore, add the following line to the /var/spool/cron/crontabs/root before the db_archive entry: 

 /usr/tcb/bin/db_checkpoint -1 -h /var/tcb/files

5.3.10    OSFJAVA Subsets Are Required for Bundled Applications

Java Version 1.1.8-7 is included with this version of the operating system. Other versions of Java are available, however, SysMan Station and other system components will fail to operate if Java Version 1.1.8-7 is removed. Therefore, do not remove Java Version 1.1.8-7 from your system.

5.3.11    Argument Size Limit for the exec System Call

The amount of memory used by the arguments to the exec system call is limited by sysconf(_SC_ARG_MAX), which is about 38 KB. You can exceed this limit systemwide by setting the exec_disable_arg_limit argument in the sysconfigtab file to 1, as follows: 

# sysconfig -r proc exec_disable_arg_limit=1

When you set this argument to 1, the limit becomes an amount that is slightly less than the maximum stack size for the process, which is typically 8 MB or more. When you set the exec_disable_arg_limit argument to 1, sysconf(_SC_ARG_MAX) incorrectly reports that the limit is 38 KB. However, programs that rely on this value will not be limited to 38 KB and will function normally.

It is unlikely that programs will require more than 38 KB of memory; however, test suites that test this limit and expect an error return when sysconf(_SC_ARG_MAX) is exceeded will not obtain their expected result. If you are running test suites that expect an error return when this limit is exceeded, leave the exec_disable_arg_limit argument set to 0. Otherwise, we recommend that you set this argument to 1.

5.3.12    Hardware Manager Incorrectly Reports the Presence of the Keyboard and Mouse

If you boot a system with no keyboard and no mouse, the system incorrectly registers default keyboard and mouse information with the hardware manager, and displays keyboard0 and mouse0 when you view the hardware hierarchy. This information is not correct and will not match the correct information reported by the sizer command.

To view the correct information, use the sizer -wk and sizer -wm commands.

5.3.13    Compaq Management Agents for Tru64 UNIX (Compaq Insight Manager)

The following notes provide information on the Compaq Management Agents for Tru64 UNIX (formerly, Insight Manager). Version 2.1B agents are provided with this release of the operating system.

5.3.13.1    Software Notes

The following notes apply to the Compaq Management Agents for Tru64 UNIX software:

5.3.13.2    Known Problems

The following problems exist in the current version of the Compaq Management Agents:

5.3.14    Event Manager

The notes in this section apply to the Event Manager.

5.3.14.1    Event Viewer and evmget Display a Message When the binlog File Is Invalid

If the binary error log file, /var/adm/binary.errlog, contains invalid log entries, an error message similar to the following is displayed when you run evmget: 

binlog2evm: Invalid event data encountered at offset 80216
binlog2evm: Error occurred while reading from
"/.local../usr/var/adm/binary.errlog"
binlog2evm: Skipped invalid data - restarted at offset 85248

If you see this message, follow your normal investigation and reporting procedures to determine the source of the corruption.

A short-term solution to prevent the message from being displayed is to redirect stderr to /dev/null.

If you are certain that the error log is properly backed up and does not contain required event information, you can permanently remove the invalid data by cleaning up the log file as described in the directions in binlogd(8). This operation removes the log file and creates a new one. Because two generations of the error log are held, the message continues to be displayed until you run the cleanup procedure twice.

5.3.14.2    Event Manager Reports Kernel Messages with Critical Priority

Event Manager reports all messages that are posted from the kernel through the syslog event facility as having critical priority. This incorrectly includes many informational messages that are posted when you start the system.

5.3.14.3    Event Manager Fails to Detect syslogd and binlogd During Boot

Occasionally, Event Manager displays either or both of the following messages on the console: 

S97evm: Communication with syslogd is not functioning
S97evm: Communication with binlogd is not functioning

In most cases, communication with syslogd and binlogd works correctly. The S97evm script subscribes to binlog and syslog events and then posts a syslog and binlog event. The S97evm script expects to see the events it just posted, but a synchronization issue in the script might cause the posted events to be missed. As a result, the script times out and displays the error messages.

You can correct this problem by modifying the sleep time-out parameter in the /sbin/rc3.d/S97evm script by changing sleep 1 to sleep 3. Then rerun the script by executing the /sbin/rc/3.d/S97evm start command.

5.4    Network and Communications

The following notes apply to network and communications software.

5.4.1    Performance Tuning for Gigabit Ethernet Adapters

The following attributes are available for tuning the performance of driver software for Gigabit Ethernet adapters. These attributes are included in the alt kernel subsystem even though they are not visible through the sysconfig utility or the dxkerneltuner application:

minimal_isr

A value that enables (1) or disables (0) the awakening of separate threads by the driver's interrupt service routine to handle completion of transmit and receive operations. When separate worker threads are enabled, the system load of the Gigabit Ethernet driver can be spread over multiple CPUs. When separate worker threads are disabled, the master system CPU (or the master CPU in each RAD of a NUMA system) services all the transmit and receive operations for the system's Gigabit Ethernet adapters.

Default: 1 (multiple threads enabled)

A 0 setting improves the performance of the driver software itself. However, overall system performance is likely to suffer during times of high network load because kernel applications other than the driver can have difficulty accessing the master CPU. Furthermore, the likelihood of overall system performance degradation goes up with each additional Gigabit Ethernet adapter supported by the system. For this reason, the default setting (1) is recommended for most systems and applications.

recv_coal

The number of packets that are sent before an interrupt is generated (assuming that the recv_ticks timer does not expire first).

Default: 32 (packets)

Minimum: 0

Maximum: 511

This attribute can be modified at run time.

* recv_ticks

The number of microseconds (usec) that elapse after receipt of the last packet before an interrupt is generated (assuming that the recv_coal threshold is not reached first).

Default: 200 (usec)

Minimum: 0

Maximum: 888,888

This attribute can be modified at run time.

send_coal

The number of packets that are sent before an interrupt is generated (assuming that the send_ticks timer does not expire first).

Default: 32 (packets)

Minimum: 0

Maximum: 511

This attribute can be modified at run time.

send_ticks

The number of microseconds (usec) that elapse after transmission of the last packet before an interrupt is generated (assuming that the send_coal threshold is not reached first).

Default: 125,000 (usec)

Minimum: 0

Maximum: 888,888

This attribute can be modified at run time.

The default value of the send_ticks attribute might have a negative performance impact on applications using NFSv3. For these applications, it is recommended that send_ticks be lowered to 10. Depending on the results from continued testing with a wide variety of applications and systems, 10 might be the default value for send_ticks in a future release.

5.4.2    The autofsd Utility

The following problems can occur when you use the autofsd utility:

5.4.3    The rcinet stop inet Command Stops IPv6 Communications

Issuing the /usr/sbin/rcinet stop inet command marks all network interfaces as down. Therefore, applications that use IPv6 will be unable to transmit messages through the interfaces.

5.4.4    Mail

This following notes provide information on problems that can occur when configuring and running mail on your systems.

5.4.4.1    The mailcv -I -t and -M -t Commands Do Not Work As Expected

If you are converting a dtmail folder hierarchy to IMAP, or if you are converting a single folder that does not already exist in the IMAP hierarchy, you receive the following error message and the conversion of the hierarchy stops: 

Mailcv: Can't create output file {foldername}, ignoring conversion.
 
foldername is the new name of the folder.

Use Netscape to migrate your folders to IMAP, as follows:

  1. Set the Local Mail folder to point to the directory that contains the dtmail folder hierarchy.

  2. From the Preferences menu, choose the Mail & Newsgroup subtree, then choose Mail Servers.

  3. Choose the Local Mail Directory and change the directory to the UNIX folder directory you want to convert.

  4. Click OK and restart Netscape.

  5. Choose the Netscape Messenger window to display your mail folders. Drag and drop the mail folders from the local folders to the IMAP folders or select all the messages in a folder and use the move command to move all the messages to the IMAP folder.

If you are converting dxmail or MH mail to IMAP folders, you receive the following error message: 

Mailcv: Can't create output file {foldername}, ignoring conversion.
 
foldername is the new name of the folder.

To migrate folders from dxmail or MH mail folders to IMAP, do the following:

  1. Migrate the folders to UNIX style by using the mailcv command with the -A option.

  2. Use Netscape, as described in the previous procedure, to migrate the UNIX mail folders to your IMAP folders.

5.4.4.2    IMAP Server: Preserving Uppercase User Names

If your system is configured as an IMAP server and you want to preserve uppercase for user names, do the following:

  1. Add the F=u flag for IMAP mailer in your sendmail configuration file.

  2. Edit the /var/adm/sendmail/sendmail.cf.pd file and /var/adm/sendmail/sendmail.m4 file (if it exists) before you run either the mailsetup script or mailconfig application.

    Search for the line with Mimap and add the u flag to its F= option. The original line appears as follows: 

    Mimap,   P=/usr/bin/deliver, F=nsmFDM, S=10, R=20/50, A=deliver $u
 

    After you update the line, it should appear as follows: 

    Mimap,   P=/usr/bin/deliver, F=nsmFDMu, S=10, R=20/50, A=deliver $u

If you have already configured sendmail using either the mailsetup script or the mailconfig application, apply these changes to the /var/adm/sendmail/sendmail.cf file in addition to the /var/adm/sendmail/sendmail.cf.pd and /var/adm/sendmail/sendmail.m4 files.

5.4.4.3    The sendmail Warning Message

The permissions on the /var directory do not satisfy the checks by the sendmail binary. The sendmail utility expects the permissions on the /var directory to be 755. However, the permissions are 775. Therefore, sendmail logs the following warning message in the syslog file every time it checks the mode of the /var directory: 

WARNING: writable directory /var/adm/sendmail

This does not impact the functionality of sendmail, so you can ignore this warning. If you want, you can change the permissions on the directory to 755 by logging in as root and entering the following command: 

# chmod go-w /var

5.4.4.4    Problem Starting the sendmail Daemon

If you manually edit the /var/adm/sendmail/sendmail.cf file and the file contains errors, the Sendmail startup script might display a message that the daemon started when it has not.

To verify whether the sendmail daemon has actually started, issue the following command: 

# ps -aef | grep sendmail

If the sendmail process is not present, check the /var/adm/syslog.dated/current/mail.log file for any errors associated with the start of the daemon. Correct any errors recorded in this file before starting the sendmail daemon again with the following command: 

# /sbin/init.d/sendmail start

5.5    Local Area Transport

The following notes apply to Local Area Transport (LAT).

5.5.1    Duplicate Minor Numbers and latsetup

The latsetup utility sometimes creates devices with duplicate minor numbers. If you manually create LAT BSD devices that do not match the valid BSD tty name space convention, latsetup can create devices with duplicate minor numbers. For example, creating device tty0 with a minor number 2 instead of 1 can cause this problem.

5.5.2    Simultaneous llogin Connections

When doing a number of simultaneous llogin connections, use llogin with the -p option. To speed up an llogin connection, add the target host name as a reserved service.

5.6    File Systems

The following notes apply to file systems.

5.6.1    Advanced File System (AdvFS)

The following notes discuss features, problems, and restrictions of the Advanced File System (AdvFS).

5.6.1.1    New AdvFS Mount Option Activates Data Logging on All Files in a Fileset

A new AdvFS mount option, -o adl, causes all files in the mounted fileset to use atomic write data logging for the duration of the mount. See chfile(8) for an explanation of atomic write data logging. This data logging is not persistent across mounts, unlike the data logging provided by chfile(8).

Additionally, files using temporary (-o adl) data logging may be mmapped(), unlike files that have had persistent data logging activated on them. The temporary data logging is suspended until the last thread using the mmaped file unmaps it. The use of chfile on a file that is using temporary data logging causes the chfile command to override the temporary data logging provided by the new mount option, and the file's I/O mode is changed persistently according to the arguments given to the chfile command.

5.7    Logical Storage Manager

The following notes describe problems and restrictions of the Logical Storage Manager (LSM).

5.7.1    The diskconfig LSMnopriv Option Does Not Work

The LSMnopriv partition type in diskconfig does not work and will generate an error when used. To create this type of partition, invoke the disklabel command directly with an fstype of LSMnoprv.

5.7.2    Cannot Enable Logging on RAID 5 Volumes Using the LSM Bottom-Up Commands

You can create RAID 5 volumes using either the LSM top-down or bottom-up commands. However, you cannot enable logging using the bottom-up commands, such as volsd aslog, to associate a log subdisk to a plex or volplex att to attach a logging plex to a RAID 5 volume.

The volassist top-down command does work. Therefore, use the volassist addlog command to add logging to RAID 5 volumes. Note that if you create a RAID 5 volume using the volassist make command, logging is configured and enabled automatically.

5.7.3    LSM Dirty Region Logging (DRL) Cannot Be Used with rootvol

LSM Dirty Region Logging (DRL) cannot be used with a mirrored rootvol. If a system with a mirrored rootvol is not brought down cleanly, the system automatically recovers the rootvol by doing a complete resynchronization. Attaching a logging subdisk might degrade the rootvol write performance with no benefit in recovery time.

5.7.4    LSM Initialization Hangs When Accessing Failed Fibre Channel Devices

Due to a problem with error recovery on failed Fibre Channel devices, systems might block for long periods of time while booting during the LSM startup. The system might appear to be deadlocked during the boot after the starting LSM message, but the system will complete the boot process after the device driver error recovery times out. This might take some time and will take longer on clusters with more members or more failed Fibre Channel devices.

A similar delay might occur when attempting to operate on failed devices, for example, importing a diskgroup with a failed disk.

If this problem occurs, either remove (delete) or replace and recover the failed Fibre Channel devices so that all Fibre Channel units are available at boot time. There should be no unreasonable delays during normal operations or LSM I/O, only booting (LSM initialization) and other operations on failed disks are affected. If you cannot remove the failed units from the system prior to booting, the system will boot; however, it will take longer than expected.

5.8    Printing

The notes in this section apply to BSD printing and Advanced Printing Software.

5.8.1    Advanced Printing Software GUI Dialogs Appear on All Workspaces

A subtle difference in behavior in the X11R6.5 software, distributed with this version of the operating system, causes some Motif and CDE applications to display windows on all CDE workspaces. The pdprintinfo and pdprintadmin programs supplied with Advanced Printing Software exhibit this behavior.

For example, if you use the pdprintadmin program to display attributes of a logical printer in your environment, and then switch to a different CDE workspace on your CDE console, the pdprintadmin tool displays the printer property sheet on the new workspace.

5.8.2    Printing Entries in the syslog Now Log at info Priority

In this release of the operating system, the lpr facility in the /etc/syslog.conf file is set to report messages at theinfo level. In previous releases, it was set to the error level. By setting the level to info, generally informative system logging occurs for both the lpr/lpd subsystem and Advanced Printing Software. You can change the syslog setting of this and other facilities by modifying the /etc/syslog.conf file and then restarting the syslog daemon.