Jump to content United States-English
HP.com Home Products and Services Support and Drivers Solutions How to Buy
 Contact HP
HP.com home
HP Tru64 UNIX: Technical Updates for the Version 5.1B and Higher Operating System and Patches > Technical Updates for the Version 5.1B and Higher Operating System and Patches

Documentation Corrections

 

Content starts here

Reference Page Issues

When new and changed features are included in ongoing releases of the Version 5.1B operating system and HP TruCluster Software Products, the reference pages for the affected components are sometimes updated and then installed during the dupatch installation procedure. Some of the problems described in this section may also have been corrected in those releases.

These revised reference pages are not, however, updated in HTML format and so are not included on the Tru64 UNIX documentation Web site:

http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51B_HTML/REF_LIB.HTM

If you have installed a newer version, such as Version 5.1B-4, the new reference pages are installed on your system and you can view them using the man command. The Patch Summary and Release Notes document included with the newer Tru64 UNIX version lists the revised reference pages. You can find that document for the the following site:

http://h30097.www3.hp.com/docs/patch/

dsfmgr(8) Reference Page Contains Syntax Error (May 2007)

Example 5 in the EXAMPLES section of the dsfmgr(8) reference page contains a syntax error in step b) and several typographical errors. The example should read as follows:

  5. The following steps show how to restore previous device special 
     file names after a configuration change. This example assumes 
     that you know the previous device names and hardware IDs (HWID).
   
     a. Assuming that the former device name was dsk0, and the new 
        device name is dsk5, use the /sbin/hwmgr command to delete 
        the old database entries for each device. Specify the former 
        HWID for a device as shown in the following example:
  
        # /sbin/hwmgr delete component -id 25 
  
      b. After the component is removed, you can delete the kernel's 
         record of its device special files as follows: 
  
         # /sbin/dsfmgr -R hwid 25 
         -dsk0a -dsk0b -dsk0c -dsk0d ... dsk0h 
  
      c. You can now move the existing device special files to their 
         new locations as follows: 
  
         # /sbin/dsfmgr -m dsk5 dsk0 

RLIMIT_DATA Description in getrlimit(2) Reference Page Inaccurate (May 2007)

In the getrlimit(2) reference page, the description of the RLIMIT_DATA parameter incorrectly references the mmap function. The correct text should read as follows:

RLIMIT_DATA
The maximum size, in bytes, of a process's data segment. Exceeding 
this limit causes the brk(), malloc(), and sbrk() functions to fail 
with errno set to [ENOMEM].

gethostbyaddr(3) and gethostbyname(3) Reference Pages Outdated (May 2007)

The gethostbyaddr(3) and gethostbyname(3) reference pages displayed on systems running Version 5.1B-1 (Patch Kit 3) or higher contain outdated information. The description for these functions should read, in part, as follows:

To determine which file or files to search, and in which order, the system uses the switches in the /etc/nsswitch.conf file.

For systems not running Version 5.1B-1 or higher, the switches in the /etc/svc.conf are used.

fread(3) Reference Page Indicates Incorrect Type Pointers (May 2007)

The synopsis of the fread(3) reference page incorrectly indicates that the size and num_items parameters are pointers. The following synopsis shows the correct syntax:

size_t fread(
           void *pointer,
           size_t size,
           size_t num_items,
           FILE *stream );
 
   size_t fread_unlocked(
           void *pointer,
           size_t size,
           size_t num_items,
           FILE *stream );
 
   size_t fwrite(
           const void *pointer,
           size_t size,
           size_t num_items,
           FILE *stream );
 
   size_t fwrite_unlocked(
           const void *pointer,
           size_t size,
           size_t num_items,
           FILE *stream );

Do Not Edit .sme Files (Jan. 2007)

The sms.sme(4) reference page description says that the base.sme and clu.sme files should be edited. In fact, you should not edit any .sme file.

Maximum Block Size in vdump Is 64 KB (Jan. 2007)

The vdump and vrestore commands were tuned to disallow block sizes greater than 64 KB blocks. Therefore the description of the -b option in the vdump(8) reference page should say “Specifies the number of 1024-byte blocks per record in the saveset. The valid range is 2 through 64 blocks; the default is 60 blocks per record.”

Quotes Missing from logger(1) Example (Nov. 2006)

The first example in the examples section of the logger(1) reference page should included quotation marks around the argument as follows:

% logger -i "System rebooted"

class_disable(3) Synopsis Contains Error (Oct. 2006)

The synopsis section of the class_disable(3) reference page contains an error. The section should read as follows:

SYNOPSIS  

     #include <apar_types.h>
     #include <sys/class.h>

     class_disable(    
             class_apar_handle_t handle );

clu_get_info(3) Contains Typographical Errors (Oct. 2006)

The clu_get_info(3) reference page contains two errors in the Examples section under CLU_INF_MY_ID, where a zero (0) appears instead of a closing quotation mark. The correct text should read as follows:

switch (retval) {
	   case 0: break;
	   case CLU_NOT_MEMBER:
	   case CLU_NO_CLUSTER_NAME:
	   case CLU_NO_MEMBERID:
	   case CLU_CNX_ERROR:
	     fprintf(stderr, "clu_get_info error");
     exit(1);
	   default:
	     exit(1);
	   }

and

for (i = 0; i <= clugenptr->clu_num_of_members -1; i++) {
	   /* print or use the returned info; for example: */
	   printf("member hostname is %s", clugenptr->memblist[i].hostname);
	   }

Data Corruption Possible When Multiple Processes Use fopen( ) on Remote File Systems (Oct. 2006)

The fopen(3) reference page states the following about the fopen function:

[Tru64 UNIX] If two separate processes open the same file for append, each process can write freely to the file without destroying the output being written by the other. The output from the two processes is intermixed in the order in which it is written to the file. Note that if the data is buffered, it is not actually written until it is flushed.

This statement is true only for local file systems such as UFS or AdvFS. However, data corruption can occur when separate processes use fopen to access file systems that are NFS-mounted (or through any other form of remote mount).

Stype Parameter Not Documented in ddr_config(8) (Oct. 2006)

The ddr_config Stype parameter is not documented in the help text (using ddr_config h) or in the ddr_config(8) reference page.

When using the ddr_config s command to get information from the ddr database for Fibre Channel devices, you can optionally specify a value identifying the Stype in one of the following ways:

# ddr_config -s disk EMC SYMMETRIX "" 2

or

# ddr_config -s disk DEC HSG80 "" 2

In these examples, the "" parameter is the revision and the 2 parameter is the Stype.

There are two Stype values in use in ddr.dbase:

  • An Stype value of 1 identifies IDE/ATAPI devices

  • An Stype value of 2 identifies Fiber Channel devices

If you do not specify an Stype value, the entry will be used for all sim types.

The Stype value is also relevant when using the S option to display the DDR entry that will be used for a particular device. Without this value, the S option will show just the default entry, and not the sim-specific entry.

Correction to cfg_psm_memops(3) and cfg_psm_catops(3) (Oct. 2006)

The cfg_psm_memops(3) and cfg_psm_catops(3)(3) reference pages contain a pointer to PSM(4). The pointer should be to psm.h(4)

ksh(1) Correction (June 2003)

In the ksh(1) reference page, the following statements are incorrect:

&& Causes the list following it to be executed only if the preceding pipeline returns a 0 (zero) exit value.
|| Causes the list following it to be executed only if the preceding pipeline returns a nonzero exit value.

The correct statements are:

&& Causes the list following it to be executed only if the preceding pipeline returns a nonzero exit value.
|| Causes the list following it to be executed only if the preceding pipeline returns a 0 (zero) exit value.

ypset(8) Correction (Feb. 2003)

The ypset(8) reference page indicates that both V1 and V2 are allowed as options. This is incorrect. V2 is not a supported option.

aio_return(3) Correction (Feb. 2003)

The following information should be appended to the RETURN VALUES section of the aio_return(3) reference page:

On an unsuccessful call, the value of -1 is returned and errno is set to indicate the error. If the operation did not complete, but it terminated normally (because, for example, the call was purposely interrupted by the aio_cancel function), errno is set to 0.

disklabel(8) Correction (Feb. 2003)

The following definition of the output from the disklabel command is missing from some versions of the disklabel(8) reference page:

An asterisk (*) is sometimes shown in the output from the disklabel command, under the column headed cylinders grouped for a partition (cpg).

This asterisk indicates that the start or the end of a cylinder does not fall exactly on a block boundary.

dxshutdown(8) Correction (Feb. 2003

Since the release of Tru64 UNIX Version 5.0, the command /usr/bin/X11/dxshutdown is a wrapper shell script that runs the SysMan shutdown program. Prior to Version 5.0, dxshutdown was an X motif application. The X motif version of dxshutdown is shipped in an obsolete subset. The new dxshutdown shell script can run the old version when it is installed as /usr/bin/X11/dxshutdown_old. Use the following command:

# /usr/bin/X11/dxshutdown -old

The current OPTIONS section of the reference page is no longer applicable because suitlets use Tk and Tk uses X, not Xt. The only useful argument is focus hostname when running on a cluster.

The SysMan application is no longer called Shutdown Manager. If invoked from the SysMan menu, the leaf is labelled "Shutdown the system" and the application is labelled "Shutdown targeted on <hostname>".

In the EXAMPLES section, the /usr/dt/appconfig/help/C/DXshutdown.sdl help file is no longer used. In the FILES section, /usr/dt/appconfig/help/C/Dxshutdown.sdl and $HOME/Dxshutdown are no longer used.

emx(7) Correction (Feb. 2003)

The emx(7) reference page provides an example of how to turn off I/O limiting by using the following run-time configuration command:

# /sbin/sysconfig -r io NPort_Max_IOs = 0xFFFFFFFF

This example command should read as follows:

# /sbin/sysconfig -r emx NPort_Max_IOs=0xFFFFFFFF

dd(1) Correction (Feb. 2003)

Note 1 in the dd(1) reference page provides an example of zeroing a disk label. The syntax of the disklabel command used in that example is incorrect and should read as follows:

# disklabel -r /dev/rdisk/dsk1a
# disklabel -z  /dev/rdisk/dsk1a
# disklabel: Disk /dev/rdisk/dsk1a is unlabeled

EvmEvent(5) Corrections (Nov. 2002)

The following information was omitted from the EvmEvent(5) reference page.

Reserved Component Names

The convention of reserved component names makes it possible to select related events, regardless of which subsystem or application posts them. See the Programmer's Guide for a description of reserved component names. The following table lists the reserved component names used by EVM:

Component NameFollowing Value Comments
_catname Processor set categoryThe name of the processor set managed category that was created or removed.
_hostnameSystem host nameThe system host name, referred to by the event, which may not be the posting host.
_hwcomponentHardware componentThe name of the hardware component that has undergone a state change.
_hwidHardware IDIdentifies events referring to hardware device IDs.
_physical_addressMemory addressThe address of a physical memory location experiencing a problem.
_pidProcess ID The proccess ID involved in a process set manager operation.
_process Process nameThe name of a process that was started or terminated.

sysman_menu(8) Correction (Nov. 2002)

The following notes apply to the sysman_menu(8) reference page:

  • The host option is not supported.

  • The menu flag cannot be used with the list flag (/usr/sbin/sysman [menu] list).

  • Examples 5 and 6 are not valid.

  • Example 4 does not specify a cluster member for the focus option.

Corrections to Manuals

Tuning Guide Table Lists Tools in Error (Sept. 2005)

Table 5.1 of the System Configuration and Tuning guide lists two tools to aid in detecting poor NFS performance that should have been removed from the table. Attempting to use these tools causes no harm, but does result in error messages being generated. Those tools and the error messages they generate are as follows:

(dbx) p nfs_sv_active_hist
"nfs_sv_active_hist" is not defined or not active

(dbx) p nchstats
can't evaluate a Block

Information on Default Private Region Size Is Incorrect in LSM Manual (March 2005)

The second paragraph of Section 1.1.2 of the Logical Storage Manager manual (September 2002 version), incorrectly states that the default private region size guarantees space for a configuration database that tracks 8192 objects (LSM disks, subdisks, plexes, and volumes). The statement assumes that the private region contains only the configuration database and does not account for the kernel change log and other LSM metadata.

To determine the size of the configuration database it is necessary to issue the voldg list diskgroupname command. For example:

# voldg list rootdg
Group:     rootdg
dgid:      1105649937.1026.hero
import-id: 0.1
flags:
copies:    nconfig=default nlog=default
config:    seqno=0.1171 permlen=2993 free=2987 templen=4 loglen=453
config disk dsk2 copy 1 len=2993 state=clean online
config disk dsk3 copy 1 len=2993 state=clean online
log disk dsk2 copy 1 len=453
log disk dsk3 copy 1 len=453

The output from this example shows that the configuration database length is 2993. The private region size for the disks in this example is the default size of 4096.

Additional Step Required to Recover from a Temporary Disk Failure Under LSM Control (Nov. 2003)

In Section 6.4.3 of the Logical Storage Manager manual — Recovering From Temporary Disk Failures — there is an extra step required before recovering the volumes. You must readd the failed disks to the affected disk group(s), using the option to restore the last-known disk media names.

The complete process to recover from temporary disk failures includes the following steps:

  1. Make sure the disk is back on line and accessible; for example:

    • Confirm that the disk is firmly snapped into the bay.

    • Reconnect any loose cables.

    • Perform any other checks appropriate to your system.

  2. Scan for all known disks to ensure the disk is available:

    # voldctl enable
  3. [NEW] Readd the failed disk (or disks) to the appropriate disk group(s):

    # voldg -[-g diskgroup] -k adddisk \
    disk_media_name=disk_access_name

    For example, to readd disk dsk14 to the dg1 disk group, where the disk media name is datadisk:

    # voldg -g dg1 -k adddisk datadisk=dsk14

    Or, for example, to readd disk dsk22 to the rootdg disk group, where the disk media name and disk access name are the same:

    # voldg -k adddisk dsk22=dsk22
  4. Recover the volumes on the disk:

    # volrecover -sb disk

AdvFS Administration Manual Correction — Extend an AdvFS File System When Increasing the Size of the Underlying Volume (April 2003)

In Section 2.3.4.3 of the AdvFS Administration manual — Increasing Storage in Domains by Extending an Existing Volume — there is an extra step required when the underlying storage volume is a hardware RAID device. You must modify the volume's disk label information to reflect the new, increased size of the partition supporting the domain, and then apply the updated disk label to the volume before extending the file system.

The complete process to extend a domain by increasing the size of an underlying hardware RAID volume includes the following steps:

  1. Using HSG80 commands, extend the hardware RAID volume.

    This might involve adding another stripeset to an existing stripeset, or creating a concatset from the original hardware RAID volume and adding another volume to it.

    For example, assume the AdvFS domain uses disk dsk25c, which is a single hardware RAID volume. To extend the capacity of the disk, create a concatset from it and another single hardware RAID volume, as shown in the following example:

    HSG80> show disks  1
    Name          Type                      Port Targ  Lun        Used by
    ------------------------------------------------------------------------------
    
    DISK10000     disk                         1    0    0        DELI-5.1A
    DISK10100     disk                         1    1    0        SPARESET
    DISK10200     disk                         1    2    0        D8          2
    DISK20000     disk                         2    0    0        DELI-5.1A
    DISK20100     disk                         2    1    0        SPARESET
    DISK20200     disk                         2    2    0                    3
    DISK30000     disk                         3    0    0        GALLO-M1
    DISK30200     disk                         3    2    0
    DISK40000     disk                         4    0    0        GALLO-M1
    DISK40200     disk                         4    2    0
    DISK50000     disk                         5    0    0        GALLO-M2
    DISK50200     disk                         5    2    0
    DISK60000     disk                         6    0    0        GALLO-M2
    DISK60200     disk                         6    2    0
    HSG80> add concatsets C1 DISK10200  4
    HSG80> set C1 add=DISK20200  5
    HSG80> show C1  6
    Name          Storageset                     Uses             Used by
    ------------------------------------------------------------------------------
    
    C1            concatset                      DISK10200        D8
                                                 DISK20200
            State:
              NORMAL
              DISK10200 (member  0) is NORMAL
              DISK20200 (member  1) is NORMAL
            Size:             71112778 blocks  7

    The following list explains each step:

    1

    Find unused disks — those with an empty (blank) Used by field.

    2

    DISK10200 (also called D8) is used by the AdvFS domain that will be extended. This storage volume is recognized as dsk25 on Tru64 UNIX.

    3

    DISK20200 is unused.

    4

    Create a concatset called C1 from DISK10200.

    5

    Add DISK20200 to concatset C1 to create a larger disk.

    6

    Display the size of the concatset C1.

    7

    Note the size, as you will need to modify the disk label for dsk25 on Tru64 UNIX to match it.

  2. Return to the Tru64 UNIX prompt.

  3. Save a copy of the old disk label information for the volume:

    # disklabel -r dskN > /tmp/label

    For example:

    # disklabel -r dsk25 > /tmp/dsk25MOD
  4. Edit the saved label and increase the size of the partition used by AdvFS:

    # vi /tmp/label

    For example:

    # vi /tmp/dsk25MOD
  5. Write the edited disk label back to the hardware RAID volume:

    # disklabel -R dskN /tmp/label

    For example:

    # disklabel -R dsk25 /tmp/dsk25MOD
  6. Optionally, display the size of the domain before extending it:

    # showfdmn domain

    For example, for a domain called clinical_trials, enter:

    # showfdmn clinical_trials
                  Id              Date Created  LogPgs  Version  Domain Name
    3e8ca76d.040d38c8  Thu Apr  3 16:28:13 2003     512        4  clinical_trials
    
      Vol   512-Blks        Free  % Used  Cmode  Rblks  Wblks  Vol Name
       1L   35556384    35547280      0%     on    256    256  /dev/disk/dsk25c
  7. Extend the AdvFS file system:

    # mount -u -o extend /file_system

    For example, if the AdvFS domain that uses dsk25 is mounted on /test_data, enter:

    # mount -u -o extend /test_data
  8. Optionally, verify that the domain now shows the larger size:

    # showfdmn domain

    For example:

    # showfdmn clinical_trials
                   Id              Date Created  LogPgs  Version  Domain Name
    3e8ca76d.040d38c8  Thu Apr  3 16:28:13 2003     512        4  clinical_trials
    
      Vol   512-Blks        Free  % Used  Cmode  Rblks  Wblks  Vol Name
       1L   71112768    71103120      0%     on    256    256  /dev/disk/dsk25c

System Configuration and Tuning Correction (March 2003)

Section 4.2 provides an example demonstrating how to enable access to the system's real time clock. This example is incorrect. The correct command is:

# mknod /dev/timedev c 15 0

Section 4.4.8.4 states: The max_async_req attribute specifies the maximum number of sessions within any given RDG context table. The recommended value is at least the number of Oracle processes plus two. This is incorrect. The max_sessions specifies the maximum number of sessions within any given RDG context table.

Section 4.4.8.5 states that the max_async_req attribute specifies the maximum number of pages automatically wired in memory for message packets. This is incorrect. The rdg_max_auto_msg_wires attribute specifies the maximum number of pages automatically wired in memory for message packets. We recommend setting this attribute to 0.

System Configuration and Tuning Guide (Feb. 2003)

In Section 6.2.2.2, it states that if you increase the value of the max_proc_per_user attribute, you increase the amount of wired memory.

This statement is false. Increasing this attribute value does not increase the amount of wired memory.

Installation Guide Contains Incorrect Java Version (Feb. 2003)

The Installation Guide states that Version 1.3.1-1 of Java™ is provided with Version 5.1B of the operating system. This is incorrect. Java Version 1.3.1-2 is provided with this release of the operating system.

Correction to Configuration Cloning Restrictions (Nov. 2002)

In Section 7.4 of the Installation Guide — Advanced Topics, the second bullet explains that software patches are not cloned. There are typographical errors in the procedure to use if your model system has already been patched; the -f option is incorrect and should be -force. The following is the corrected text:

  • Software patches are not cloned. The reason for this restriction is that installed patches change serial numbers, which will cause validation errors when the config.cdf file is applied to the target system. The config.cdf file you create must be created from a model system that has not been patched.

    If the model system has already been patched, there are two alternatives you can follow, although they are not recommended:

    • Create the config.cdf from the patched system anyway. Use the -force option when applying the config.cdf file to the target system to force the config.cdf onto the target system. For example, at the target system, enter:

      #  sysman -clone -apply -force config.cdf

      Then, at the newly cloned target system, create another config.cdf to use for cloning additional systems. The new config.cdf will contain the correct checksums, and cloning other systems from this file will be valid.

      Caution:

      The -force option to force a CDF onto a target system is typically used for internal debugging purposes. Be aware that should the config.cdf contain additional problems other than the patched model system issue, forcing the CDF could cause problems on the target system from which there may be no recovery.

    • Use the Update Installation process to update the patched model system to the next version of the operating system, if the next version has been released, thereby bypassing the patch issue.

Other Documentation Issues

HSZ70 Storage Array Controller is Supported on Tru64 UNIX Version 5.1B (May 2003)

The Tru64 UNIX Version 5.1B Release Notes incorrectly announced the retirement of the HSZ70 storage array controller in conjunction with the release of Version 5.1B of the Tru64 UNIX operating system.

The HSZ70 is supported with Version 5.1B. The controller driver software is included on the kit, and the Version 5.1B documentation includes troubleshooting information for the HSZ70.

DEC Ada Retirement (May 2003)

DEC Ada was retired in March, 2000. The Tru64 UNIX Version 5.1B Release Notes incorrectly states that DEC Ada (UPI - 0HM) and DEC Ada PDO (UPI - 0VS) will be retired in a future release of the operating system.

Motif 1.2 Input Methods Work (Sept. 2004)

The README file included with the Motif 2.1 package inaccurately states that input methods that “currently work will not work with Motif 2.1.” In fact, input methods will work in most cases. When Motif 2.1 is installed, the README file is available at /usr/opt/motif2.1/READM.

Book on Protecting Against Spoofing Attacks Available (Oct. 2002)

Protecting Your System Against File Name Spoofing Attacks is an engineering practices document that describes how to address the problem of name space collisions (spoofing, or symlink attacks) by using system administration tools and safe coding practices in local scripts and programs.

The engineering practices described in the document focus on writing safe, privileged code and scripts. In particular, the document addresses the use of temporary files, which is a commonly exploited vulnerability. This document is available on the Tru64 UNIX Documentation Web page: http://www.tru64unix.compaq.com/docs/wpapers/spoof_wp/TITLE.HTM

OpenLDAP Directory Server Book Available (Dec. 2002)

The OpenLDAP Directory Server Installation and Administration guide is now available on the Tru64 UNIX Documentation Web page: http://www.tru64unix.compaq.com/docs/pub_page/V51B_DOCS/V51B_DOCLIST.HTM

This guide describes how to install and configure the Lightweight Directory Access Protocol (LDAP) Directory Server in a Tru64 UNIX environment.

Printable version
Privacy statement Using this site means you accept its terms Feedback to webmaster
2002–2009 Hewlett-Packard Development Company, L.P.