DFS(TM) for AIX(R) and Solaris
PTF 9 README
Version 3.1
Document Number 00000-00
00000
Third Edition (February 2005)
This edition applies to:
and to all subsequent releases and modifications until otherwise indicated in new editions.
This softcopy version is based on the printed edition of this book. Some formatting amendments have been made to make this information more suitable for softcopy.
Order publications through your IBM representative or through the IBM branch office serving your locality.
© Copyright International Business Machines Corporation 2001-2005. All rights reserved.
Note to U.S. Government Users -- Documentation related to restricted rights -- Use,
duplication or disclosure is subject to restrictions set forth in GSA ADP Schedule
contract with IBM Corp.
Obtaining and
Installing/Uninstalling the PTF
This README file describes the contents of the IBM DFS for AIX and Solaris Program Temporary Fix (PTF) 9 for version 3.1. It also provides instructions for obtaining and installing the PTF on each platform.
This PTF must be applied to DFS Version 3.1.
The current software changes to DFS 3.1 are cumulative. That is, this PTF includes the changes that were issued in all previous PTFs. It is not necessary to install previous updates before installing this PTF. Some exceptions to this statement are:
PTF 9 is the minimum level providing support on AIX5.3.
PTF 3 must be applied before any later updates are installed on AIX 5.1 and AIX 5.2 systems, since PTF 3 is the minimum level supporting the 64-bit kernel. As AIX 5.1 support was added with the DFS 3.1 PTF 3 release, PTF 3 is the minimum level providing DFS 3.1 support on the AIX 5.1 kernel. PTF 7 provides support for AIX 5.2 kernel. Since support for AIX 5.2 kernel is added from DFS 3.1 PTF 7 release, PTF 7 or any later PTF's provide DFS 3.1 support on AIX 5.2 kernel.
PTF 6 must be applied before any later updates are installed on Solaris 9 systems. Since Solaris 9 support was added with DFS 3.1 PTF 6 release, PTF 6 is minimum level providing DFS 3.1 support on the Solaris 9 kernel.
You can download the PTF from the IBM Web site http://www6.software.ibm.com/dl/dfs31/dfs31-p. Follow the instructions on the Web site to download the PTF, and then follow these installation instructions to install the PTF on your system.
This section describes how to install the IBM DFS Version 3.1 PTF on AIX and Solaris platforms.
This section describes the procedure for installing the DFS PTF on an AIX machine. The PTF for AIX has the following '.tar' files and free disk space requirement for different Operating Systems:
Operating System Installable '.tar' File
Free Disk Space Required
AIX 5.3 dfs-3.1-aix53-P9.tar 57 MB AIX 5.1 and AIX 5.2 dfs-3.1-aix51-P9.tar 22 MB AIX 4.3 dfs-3.1-aix43-P9.tar 13 MB
Before installing the PTF, ensure that your system meets all of the installation prerequisites described below:
The DFS 3.1 PTF 9 software includes new support for the AIX 5.3 operating system. The following upgrade paths were tested:
Note: While migrating from AIX 4.3 to AIX 5.3, you must check if APAR IY17981 is applicable for your system. Refer to AIX 5.3 Release Notes for more details.
There are three ways to upgrade machines from AIX 5.1 or AIX 5.2 to AIX 5.3:
We conducted the upgrade tests using the migration option.
The following tips are important for customers who install DFS software on AIX 5.3 systems using the DFS 3.1 GA release and PTF 9.
config.dfs -autostart no
After installing DFS 3.1 PTF 9, enable autostart again by issuing the same command with the -autostart value set to yes.
tar xvf [file_path]dfs-3.1-aix53-P9.tar
Use the file path only if the PTF tar file is not in the current directory.
On AIX 4.3.3
tar xvf [file_path]dfs-3.1-aix43-P9.tar
On AIX 5.1 or 5.2
tar xvf [file_path]dfs-3.1-aix51-P9.tar
Use of file_path is required if the PTF tar file is not in the current directory.
This section describes the procedure for installing the DFS 3.1 PTF on a Solaris machine. The PTF for Solaris has the file name dfs-3.1-solaris-P9.tar.Z.
Before installing the PTF, ensure that you have met all of the installation prerequisites described below:
uncompress [file_path]dfs-3.1-solaris-P9.tar.Z tar xvf [file_path]dfs-3.1-solaris-P9.tar
Use of file_path is required if the PTF tar file is not in the current directory.
The script automatically installs new software packages corresponding to the packages installed on the machine. If conflicts are discovered, the user is prompted to resolve conflicts. By default, the installptf script creates a back-out PTF, which is used to reverse the update PTF's installation. One exception to this statement is when PTF 6 is applied on Solaris 9 system; please refer Support for Solaris 8/Solaris 9. For information about various features of the installation script, see Installation script options.
To verify which packages were installed, issue the command pkginfo | grep IDFS
You can modify the behavior of the PTF installation script, installptf, by using command-line options or by setting up a customized environment file.
The complete format of the installptf script follows.
installptf [-e env_file] [-nolock] [-noback] [-force] [-help]
Options specified on the command line always override default options and definitions found in an environment file.
The PTF installation script reads several parameter values from its environment file, installptf.env. The environment file is located with the installation script in the IDFS-P9 directory. A customized environment file allows the administrator to change complex settings, such as file paths, that cannot be changed by using command-line options. For example, customizing the environment file to cause back-out PTFs to be uniquely named allows an administrator to store multiple uninstallation options; the installptf script's default behavior is to overwrite any existing back-out PTF with a new one.
To use a customized environment file with the PTF installation script, follow these steps.
By default, the installptf script creates a log file and a back-out PTF. Both of these artifacts are overwritten by subsequent PTF installations. The log file is stored in /var/sadm/patch/IDFS/. The back-out PTF contains archived versions of overwritten packages, a README file, and a script named backoutptf that reverses the actions taken by the installptf script for that machine. The backoutptf script is stored in /var/sadm/patch/IDFS/backout/ and must be executed from that directory.
The PTF software can be uninstalled by using the following method only if the field labeled COMMIT software updates? is set to no and the field labeled SAVE replaced files? is set to yes, as described above under Step 9. If the COMMIT option was set to yes during installation, the software cannot easily be uninstalled.
To uninstall the PTF on an AIX system, do the following:
During installation on a Solaris machine, a back-out PTF and log file are created by default. One exception to this statement is when PTF 6 is applied on Solaris 9 system. The back-out PTF includes a tar file containing replaced software packages and a script named backoutptf. To uninstall the update PTF using the back-out PTF, follow the steps below.
This section highlights general information and requirements that apply to this PTF release for IBM DFS 3.1 for AIX and Solaris.
To find the latest information about this product, please refer to the following IBM Web site:
http://www-306.ibm.com/software/stormgmt/dfs
DFS 3.1 for AIX and Solaris now supports DCE 3.2. A supported DCE product must be installed on the machine prior to installation of DFS 3.1.
This PTF supports both 32- and 64-bit DFS 3.1 kernels on AIX 5.3. Refer to Using the PTF to enable AIX 5.3 support for information about installing DFS on AIX 5.3 systems using the PTF release.
AIX 5.3 supports two types of local filesystems, JFS and JFS2. DFS currently supports only the JFS filesystem. This means that:
Note: Currently DFS 3.1 does not support NFS Version 4 on AIX.
DFS 3.1 has been recertified on AIX 5.2 kernel. DFS 3.1 PTF 7 is the minimum level providing DFS 3.1 support on the AIX 5.2 kernel. PTF 3 must exist on an A IX 5.2 system before PTF 7 or any later PTF is applied. PTF 3 is installed on AIX 5.2 using the same steps mentioned in PTF 3 Readme for AIX 5.1.
IP fragmentation problem on AIX 5.2B:
AIX 5.2B (ML-01) introduced an enhancement feature in IY39633 to improve IP fragmentation performance by allowing a chain of fragments to be sent down to the adapter device driver by the IP layer. However, this change is causing random "Connection timed out" problems on a DFS 3.1 client under heavy load. As an immediate workaround, disabling packet chaining in the driver (using ifconfig interface -pktchain) resolves the problem. Problem exists at AIX 5.2B (ML-01) and is fixed in AIX 5.2B (ML-02) onwards. Refer to IY45619 for more details about the fix.
DFS 3.1 PTF 3 is the minimum level providing DFS 3.1 support on the AIX 5.1 kernel. Both 32- and 64-bit DFS 3.1 kernels are supported on AIX 5.1. Refer to the PTF 3 Readme for information about installing DFS on AIX 5.1 systems using the PTF 3 release. PTF 3 must exist on an AIX 5.1 system before later PTFs can be applied.
AIX 5.1 supports two types of local filesystems, JFS and JFS2. DFS currently supports only the JFS filesystem. This means that:
This service update for DFS Version 3.1 includes support on the Solaris 8/Solaris 9 operating system. Refer to Using a PTF to enable Solaris 8/Solaris 9 support for information about installing DFS on Solaris 8/Solaris 9 systems using a PTF release.
Install the DFS software for Solaris on a machine with only one operating system. Systems that are configured so that the user can choose between two different versions of the Solaris operating system at startup time (dual-boot systems) are not supported.
On AIX, DFS supports TSM 5.1 and TSM 5.2. Note that there is only one butc binary which can be used with both TSM 5.1 and TSM 5.2. To run butc on AIX, it is essential to have package
'tivoli.tsm.client.api.aix43.32bit' for TSM 5.1 and
'tivoli.tsm.client.api.32bit' for TSM 5.2
installed in the system.
The following tips apply when upgrading DFS software to version 3.1 or installing a DFS 3.1 PTF.
For DFS 3.1 support on AIX 5.3, PTF 9 must be installed. Before installing any later PTF's on AIX 5.3, ensure that the General Availability (GA) version of DFS 3.1 was installed and that PTF 9 was installed with the overwrite option. Refer to section 'Using the PTF to enable AIX 5.3 support' in this document for details.
For DFS 3.1 support on AIX 5.1 and AIX 5.2, PTF 3 must be installed. Before installing any later PTF's on AIX 5.1 and AIX 5.2, ensure that the General Availability (GA) version of DFS 3.1 was installed and that PTF 3 was installed with the overwrite option.
Before upgrading to the Version 3.1 software or applying a DFS 3.1 PTF, all DCE applications must be shut down. To correctly upgrade an AIX-based machine, follow these steps:
Additional information about upgrading DCE and DFS software to Version 3.1 is available in the IBM Distributed Computing Environment for AIX Version 3.1 Quick Beginnings. (This Quick Beginnings document is available on the IBM Web site at http://www.ibm.com/software/network/dce/library/publications/dce31aix.html.) Information about installing a DFS 3.1 PTF is included earlier in this document, in Installing the PTF.
To ensure proper function, restart AIX machines immediately after installing upgraded DFS software.
Support for the Solaris 8 operating system is provided in DFS 3.1 PTF 1 and later, and support for Solaris 9 is provided in DFS 3.1 PTF 6.
Support for the Solaris 8 operating system was introduced in the first DFS 3.1 PTF; it is not available in the general availability (GA) release of DFS 3.1. The following tips are important for customers who install DFS software on Solaris 8 systems using the DFS 3.1 GA release and a PTF. Because PTF changes are cumulative, any DFS 3.1 PTF can be used to enable Solaris 8 support.
After upgrading a DFS machine's operating system from Solaris 7 to Solaris 8, you must install a DFS 3.1 PTF on the Solaris 8 system regardless of whether the PTF was previously installed on the Solaris 7 system. The PTF installation script detects the Solaris operating system version and installs appropriate software packages for that system. For this reason, the PTF installation must be done on the operating system to be used.
To do a new DFS installation on a Solaris 8 system using the GA software and the PTF, you must first install the unsupported DFS 3.1 software on the Solaris 8 system, then install the PTF software. The PTF supplies the necessary requirements for Solaris 8 support. Do not attempt to configure the DFS software until after the PTF software has been installed.
Support for the Solaris 9 operating system has been introduced in DFS 3.1 PTF 6; it is not available in the gold master (GM) release of DFS 3.1.0.1. Please refer DCE release/announcement for supported DCE version on Solaris 9. The following tips are important for customers who install DFS software on Solaris 9 systems using the DFS 3.1 GA release and PTF 6.
After upgrading a DFS machine's operating system from Solaris 7 to Solaris 9 or Solaris 8 to Solaris 9, you must install DFS 3.1 PTF 6 on the Solaris 9 system regardless of whether the PTF was previously installed on the Solaris 7/Solaris 8 system. The PTF 6 installation script detects the Solaris 9 operating system version and installs appropriate software packages for that system.
To do a new DFS installation on a Solaris 9 system using the GA software and the PTF 6, you must first install the unsupported DFS 3.1 software on the Solaris 9 system, then install the PTF 6 software. The PTF 6 supplies the necessary requirements for Solaris 9 support. Do not attempt to configure the DFS software until after the PTF software has been installed.
The following tips are important for customers who install DFS software on Solaris 8 systems using the DFS 3.1 GA release and a PTF, and who install DFS software on Solaris 9 systems using the DFS 3.1 GA release and PTF 6.
If the DFS autostart feature is enabled, you might see error messages during the upgrade process because the DFS 3.1 GA software attempts to start on the Solaris 8/Solaris 9 system before the PTF is installed. To avoid seeing these messages, disable the DFS autostart feature before upgrading the Solaris operating system. To disable autostart, issue the following command from a root prompt.
config.dfs -autostart no
After installing the DFS 3.1 PTF, enable autostart again by issuing the same command with the -autostart value set to yes
The Solaris operating system upgrade replaces a system file, /etc/name_to_sysnum, that contains a DFS entry. Installing a DFS 3.1 PTF on Solaris 8/Solaris 9 (as recommended) automatically places the necessary entry for Solaris 8/Solaris 9 in the new file.
The backup file created during the operating system upgrade does not contain the proper entry for Solaris 8/Solaris 9. (The backup file is stored as /etc/name_to_sysnum:2.7.) If you manually replace the name_to_sysnum file after installing the PTF, you must add or modify the dfscore statement in the file to have the following value for Solaris 8/Solaris 9 systems:
dfscore 64
cd /usr/bin rm kinit klist kdestroy ln -s /opt/dcelocal/bin/kinit kinit ln -s /opt/dcelocal/bin/klist klist ln -s /opt/dcelocal/bin/kdestroy kdestroy
If DCE is installed after the Solaris 8/Solaris 9 operating system installation, these steps are unnecessary.
In some cases, after upgrading DCE and DFS software on Solaris machines, the DCE software does not start automatically when the machine is started. This is because a link to a DCE initialization file is missing after installation.
To solve this problem, create a link named S15-20dce in the path /etc/rc3.d/. The link should reference the file /etc/init.d/dce.
Before installing DFS Version 3.1 on an AIX or Solaris machine, ensure that a supported DCE is installed on the system. If you are using DCE 3.1, then DCE 3.1 PTF 1 is required.
Installation procedures on Solaris platforms do not automatically enforce this requirement. The showrev -p command can be used to show patches that have been applied to your DCE software. If you received a CD containing DCE 3.1 and DCE 3.1 PTF 1, note that the PTF is supplied in the patches directory of the CD, and you must install the patch separately after installing the DCE software.
A command line interface similar to the UNIX man command is now available with DFS 3.1 on Solaris and AIX. The new dfsman commands enable users to read preformatted command reference files in the list3270 format. For example, issue the dfsman cm command to access the manual page on the DFS Cache Manager. For a list of all available dfsman commands, issue the dfsman help command.
Additionally, a new tool, asciiview.dfs, is now available to enable users to view the following documents:
If bak dump commands or bak dumpinfo commands are executed too closely in succession, errors may result.
These bak commands access the Backup Database for information on the dump schedule, the tape drive, and the DFS servers. To access the Backup Database, the bak process must obtain a lock. The lock is exclusive to the process and is not recursive. In other words, only one process can access the Backup Database at any given time. When a bak dump or bak dumpinfo command is issued, an instance ID is created. Simultaneous access to the Backup Database is prohibited, and the instance ID enforces this.
Each distinct dump command requests different information and uses different functions; therefore, the period of time for which the lock exists differs for each dump request. And, the period of time for which other processes are prohibited from accessing the database differs for each dump request. For this reason, the time interval between successful bak dump or bak dumpinfo commands varies.
DFS does not return an error the first time a locking conflict is encountered; instead, there is a sizable amount of time during which a locked out process retries to access the database. Therefore, if a lock is being held by another process when a bak dump or bak dumpinfo command is issued, DFS generally resolves the locking conflict. In the case that locking conflicts exist and cannot be resolved, DFS users receive an error specifying that a lock is held by another user. To resolve the problem when this error is received, stagger the execution times of the bak commands.
IBM DFS for AIX Quick Beginnings and IBM DFS for Solaris Quick Beginnings, included with DFS Version 3.1, give an incorrect file name in the file path for installed HTML documentation. The installed file has the extension .htm, but the file is incorrectly listed with the .html extension in the section titled "Viewing the HTML documentation."
The correct paths follow.
For AIX:
/usr/lpp/dfsdoc/html/en_US/index.htm
For Solaris:
/opt/dce/dfsdoc/html/en_US/index.htm
Instructions for creating the root.dfs fileset, which is the root of a DFS cell's filespace, were omitted from the DFS Version 3.1 Quick Beginnings guides. The instructions can be found in the DFS Administration Guide, in the "DFS configuration issues" section, under the heading "Setting up filesets."
The Quick Beginnings guides included with DFS Version 3.1 inappropriately recommend restarting the system as a way of shutting down DFS processes before a software upgrade. This method is not effective in most cases because DFS automatically restarts when the system restarts.
This error appears with the heading "Stopping Software Processes" in the "Installing DFS 3.1" section of the AIX and Solaris Quick Beginnings guides.
Before upgrading DFS software, stop all DCE applications, then stop any remaining DCE and DFS processes. After upgrading DFS software, restart the system before using DFS.
When migrating AIX machines from IBM DCE DFS 2.2 to DFS 3.1 by using the SMIT or smitty utility, choose the Install Latest option to upgrade files. Choosing the Update All option does not properly update all the necessary files.
This note does not apply to the DFS 3.1 PTF installation procedures. Follow the instructions in Installing the PTF on an AIX machine to install a DFS 3.1 PTF.
A code error in DFS 3.1 for AIX sometimes causes the installation program (SMIT or installp) to return an error message similar to the following:
missing prerequisite dce.client.runtime 3.1.0.125
This error has two possible causes. To determine which cause applies to your case, mount the CD-ROM and issue the following command:
grep 125 /cd mount point/usr/sys/inst.images/.toc
If no value is returned, you might be trying to install from a copy of the installation images that does not have a valid table of contents (.toc) file.
If the line *prereq dce.client.rte 03.01.0000.0125 is returned when you issue the grep command, you might have a defective CD-ROM that has an invalid table of contents (.toc) file.
To install DFS 3.1 from a copy of the installation images, copy the .toc file from the CD-ROM to the path where you copied the installation images. Using the inutoc command to create a .toc file after copying the installation images creates an invalid .toc file.
If the results of the grep command indicate that you might have a defective CD-ROM, check the part number. Proper disks have the part number CV6Z7ML printed on the front of the disk below the version number. Some preproduction disks with defective .toc files were accidentally released; these have the part number CV6F5ML. If you have a defective CD-ROM, contact your IBM sales representative for a replacement.
The need for the en_US locale on Solaris was not documented. Users should be aware that DFS 3.1 on Solaris 7, 8 and 9 requires that the North American English locale (en_US) be installed on the host. This locale can be installed by the Solaris administrator at system installation. Alternatively, en_US locale support can be added to a pre-existing Solaris system. To find the Sun packages needed, go to the docs.sun.com Web site and search for "North America OS Support." Depending on the hardware architecture, the host might also need the package referred to as "North America 64 bit OS Support."
This section describes the updates to IBM DFS 3.1 for AIX and Solaris that are included in the service updates.
PTF 9 includes all updates that were in previous PTFs, plus the following items.
A Reference APAR exists for each platform that supports PTF 9 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
APAR IY60271 (Solaris)
The DFS server machine panics as the default threadstack size on solaris 9
machines is falling short. The problem has been fixed.
APAR IY60523 (AIX 5.x)
On Aix 5.x DFS was causing machine crash in dmlfs. This change is relevant only
to those customers, using the DFS Storage Management Toolkit. The problem has been fixed.
APAR IY60864 (AIX)
The current value of buffer cache is falling short for larger aggregates on
heavily used file servers. The problem has been fixed.
APAR IY61062 (AIX 5.x)
On Aix 5.x, 64 bit mode, 'fts move' causes system crash for filesets which
contains special device files. The problem has been fixed.
APAR IY63450
The fxd daemon is picking up out of range ports even after setting
RPC_RESTRICTED_PORTS variable. The problem has been fixed.
APAR IY66881 (AIX)
The DFS memcache limit is increased to 128 MB from the existed 16 MB limit.
APAR IY68104 (AIX 5.x)
Access to DFS exported JFS file system fails with EACCES error. The problem has been fixed.
PTF 8 includes all updates that were in previous PTFs, plus the following items.
A Reference APAR exists for each platform that supports PTF 8 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
DFS client configuration fails when RPC_RESTRICTED_PORTS is set. This is when config scripts trying to change mode on links /:/ and /.:/ before their creation. Problem is fixed.
DFS client hangs in 'ls' command, due to a leak in dcache refcount. Problem is fixed.
bosserver coredumps in 'bos lsadmin' command when used with an invalid adminlist file. Problem is fixed.
Due to un-initialized memory, machine panics while exporting an aggregate into DFS space. Problem is fixed.
DFS Client hangs when a thread tries to reacquire an scachelock. The deadlock has been fixed.
'bak restoreft' command fails with the error "illegal host id". Problem is fixed.
When a user belonging to more than 30 groups accessed DFS space, machine crashed. This problem is fixed.
Machine panics in rpc call due to improper VM handle on AIX POWER4 Processor. Problem is fixed.
'fts syncserv' output is not formatted properly when fldb entries are greater than or equal to 25. Problem is fixed.
EFBIG error when uncompressing a file, resulting in the creation of a file larger than 4GB. Problem is fixed.
DFS 3.1 is now compatible with Tivoli Storage Manager (TSM) server release 5.2. The butc code in DFS 3.1 is recertified to support this new release.
PTF 7 includes all updates that were in previous PTFs, plus the following items.
A Reference APAR exists for each platform that supports PTF 7 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
Due to the file attribute synchronization problem in DM managed episode filesystem, Unix sum command sometimes used to show wrong output. This problem has been fixed.
Repserver coredumps due to invalid connection pointer caused by wrong IP address. This problem has been fixed.
Repserver coredumps while trying to mark a fileset for deletion. This problem has been fixed.
The information about the locked filesets as shown by the fts fldb command is now being directed to stdout instead of stderr.
DFS Client machine crashed while trying to decommission FLDB machine from cell. This problem has been corrected.
In DFS Storage Management environment, changes made to files by DM applications were not visible to the DFS Clients. This problem has been fixed.
The DFS Fileserver machine panics when client sends wrong packet. This problem has been fixed.
DFS 3.1 is now compatible with Tivoli Storage Manager (TSM) server release 5.1. The butc code in DFS 3.1 was recertified to support this new release. A single butc binary will now work with all the TSM versions.
DFS Fileserver used to panic while writing to a fileset having an exceeded disk space or fileset quota. This problem has been resolved to return appropriate error codes.
The time field for the read-only copy of a fileset was not initialized by the fileserver. This stopped DFS clients from accessing the fileset. This problem has been fixed.
PTF 6 includes all updates that were in previous PTFs, plus the following items.
A Reference APAR exists for each platform that supports PTF 6 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
dfsiauth command now show authentication mappings on AIX 51 machine in 64 bit mode.
Credential files generated by "list" and "delete" option of dfsiauth command were not purged after these commands completed execution. This problem has been fixed.
dfsiauth command will now show complete hostname.
System performance decreases while copying files in DFS space after increasing the cache size beyond 2.6 GB. This problem is now corrected.
On a mutlitple CPU sparc machine running Solaris 8, "cd" to dfs space hangs while accessing the rootvnode of the dfs filesystem. This problem has now been fixed.
The DFS cache manager was returning the error code of the parent directory, instead the file being looked upon. The code has been modified so that the DFS cache manager will return the correct error code to the user.
On an AIX 51, 64 bit kernel, the "df/mount" command stops working if any of the episode fileset is mounted locally. This problem is now corrected.
DFS server crashes on AIX5.1 machine, while doing any operation in a local file system which is exported through DFS. This problem has been fixed.
When new server entry name is part of already entered server name , then 'bak addftentry' command erroneously returns with duplicate server entry. The problem has been fixed.
Configuring dce with "autostart" and "clean_autostart" option and then configuring dfs with "autostart" option used to cause dce services to fail at startup. This problem has been corrected.
Stop.dfs sometimes fails to stop running upclient process. This problem has been corrected.
On a machine with dce hostname different from the machine local host name, dfsexport used to fail while detaching aggregate. This problem has been fixed.
If we run dfsstat on a Solaris 7/8 DFS 3.1 PTF5 client machine, it tries to retrieve File Server thread pool values and fails. This problem has been fixed.
The dcecp is unable to handle multi piped commands particularly on faster machines. This led to a problem while trying to configure DFS servers. The problem has been fixed.
Modification of a fragmented file in episode filesystem causes crash on AIX and solaris m/c in 64 bit mode. This problem is fixed.
After changing the IP address of any fileserver machine, DFS client is unable to access any of the fileset residing on that server for few minutes resulting in an error. This problem is now being fixed.
DFS client machine hangs due to a deadlock between threads. The problem has been fixed.
DFS client machine panics due to a deadlock in the cache manager code. The problem has been fixed.
asciiview program in IDFSenUSd package has been renamed to asciiview.dfs, due to a conflict with the same binary name in DCE docs package. Users can access the program with the renamed version.
While running many concurrent fts dump commands, some of those dump command fail with the error EBADF. This is due to the volume descriptor getting garbage collected incorrectly before completion of operation on that volume. The problem has been fixed.
DFS3.1 has now been recertified to work on Solaris 9.
Machine crashed while handling rpc exception on AIX 51 machine in 64 bit mode. This problem has been fixed.
PTF 5 includes all updates that were in previous PTFs, plus the following items.
A Reference APAR exists for each platform that supports PTF 5 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
When the RPC_RESTRICTED_PORTS environment variable was set, the message that was logged in the cfgdce.log file was incorrectly interpreted as an error message. The message is now correctly interpreted as an informative entry indicating that ports are restricted to a given range.
When a replicated fileset is renamed and mount point re-evaluated from one client box, "cm whereis ." command returns the old fileset name for the new read-only mount-point, if run from the same client. This problem has been fixed.
DFS client machine panicked when setting a pag for a user. The panic occured while attempting to release an already-released mutex. This problem has now been corrected.
Reconfiguring fileserver on a machine sometimes resulted in an orphaned UUID in the bosserver administrative list. The problem has been corrected.
The dfsstat tool is now available on both 32-bit and 64-bit environment for AIX 5.1 systems.
Repeated starting or stopping of DCE/DFS causes the dcecp program to coredump. This problem was due to memory reference being freed by both DCE and DFS code. The problem has been fixed.
In the binary distribution process from upserver to upclient, file attributes were not being properly transferred from server to client. This problem occurred because of a change in the variable name used for file size in the stat structure for AIX. The problem has been corrected.
DFS Fileserver service threads caused a system hang, while trying to process a who-are-you challenge validation. This problem was due to solaris kernel not supporting time sliced preemption of SYS class threads. If such a thread is looping around a code path, that would result in a CPU starvation for other threads waiting for a CPU cycle, resulting in a hang. This problem has been fixed.
dfsstat command has been enhanced to display the following values from the
Fileserver threadpools represented by the fxd's mainprocs and tokenprocs parameters :-
Executing dfsstat binary would now invoke the appropriate 32-bit or 64-bit binary in respective environment.
DFS access failures were encountered when client sets the machine context with the Fileserver and receives an EINTR error code. This happened because cache manager did not handle the return code properly, and erroneously used to mark the server down. This code is now handled correctly.
When configuring the DFS client on an AIX 5.1 system on which the cache directory does not reside in the JFS file system, the DFS client took more than two minutes to display a failure message. Now, in this circumstance, an error message is reported immediately.
When a fileset is created by a DCE principal of a foreign cell, the owner of the fileset was set to foreign DCE principal creating it. This has been modified to set the fileset owner as local root.
Garbage collection, a feature in NFS-DFS gateway, which periodically deletes unauthenticated and expired mappings from the gateway machine, has been enhanced to run according to user specified parameters. Note that the garbage collector runs once the total number of mappings ( mapping threshold ) exceeds a specific limit. Two new command-line options have been added to the "dfsiauth" command suite :-
On a machine running DCE 3.2 components and having DFS components installed or configured, restarting the following DCE components failed: gdad, dts client, dts global, audit, ems, password strength server. The problem has been identified and corrected.
DFS client machine panicked when trying to display an error message. The panic occured while attempting to release an already-released mutex. This problem has now been corrected.
Performing a store operation on a fileset having an exceeded disk quota sometimes caused the DFS machine to panic. This problem has been fixed to return an error message and the machine does not panic.
Following command line options have been introduced in "cm" command suite to activate or deactivate the DFS cache manager bulkstat feature from a client machine at runtime :
The bulkstat feature provided by cache manager collects bulk data to be added to the caches for the names and statuses of files in a directory on which a lookup is to be made. This happens only if the directory lookup is being made for the first time.
Note : We have added this cm bulkstat feature in the belief that it may improve performance for some customers depending on their specific data access patterns. However, please note that the messages for this new feature have not been translated to any languague other than English.
Users were experiencing DFS access failures sporadically due to the self credentials being marked as expired by DFS Cache Manager. This problem has been corrected.
If a parent directory is having more than 65533 sub-directories, then removing one of those sub-directories panics the machine. The linkcount of parent directory overflows when there are more than 65533 sub-directories. This problem has been corrected by putting a check on the linkcount of the parent directory. Now, creation of a new sub-directory in a parent directory , already having 65533 sub-directories, fail with error EMLINK.
On AIX 5.1 systems, the value of the SYS_NAME parameter has been modified. Now, rs_aix51 is returned when the cm sysname command is issued.
PTF 4 includes all updates that were in previous PTFs, plus the following items.
A Reference APAR exists for each platform that supports PTF 4 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
After restarting a DFS Client machine, the command start.dce all failed to start the DFS client and returned the following error message.
dfsd: failed to invoke the TKN server (code 22)
The problem has been identified and corrected.
When using the SMIT or smitty utility, the following DFS Fileset Replication Server operations did not function:
This occurred because the operational name of the DFS Fileset Replication Server component is dfs_repsvr; however, the name of the component being used by the smitty utility was dfs_rep_svr. This inconsistency has been fixed and the smitty utility can now be used for DFS Fileset Replication Server operations.
Numerous unnecessary messages were returned when a user attempts to start DCE on a Solaris 7 or Solaris 8 machine on which DFS 3.1 is installed but not configured. The unnecessary messages are no longer displayed.
A Solaris 8 system panicked due to a message ID mismatch. Now, the correct message IDs are generated and the problem has been corrected.
Users were sometimes unable to transfer files to DFS space using FTP after upgrading to AIX ML09. This problem has been corrected.
Users sometimes received no response when performing directory operations, such as ls, in the DFS filespace. The DFS buffer cache implementation has been modified to correct this problem.
Previously, online manual pages were not available with DFS 3.1. The dfsman utility has been created and can now be used to access DFS manual pages. See Online manual pages now available for details.
When a cloned fileset's quota limit was exceeded, delete operations were not permitted. This problem has been identified and corrected.
The DFS 3.1 client sometimes failed due to a null pointer. A conventional check for a null pointer has now been introduced in the client code and the failure no longer occurs.
DFS file server configuration failed when replica security servers were out of sync with the master server. The problem has been identified and corrected.
In a DCE cell with replicated security servers, attempting to unconfigure a DFS file server machine sometimes resulted in an incomplete unconfiguration. This problem has been identified and corrected.
Previously, when configuring the DFS client to use a memory cache with a size greater than 64 MB, the DFS client failed and returned an incorrectly formatted and misleading error message. The following changes have been made to resolve this problem.
The DFS code has been modified so that the ftserver extracts the DCE principal and group from the binding handle and passes it through the volume status structures, ensuring that the DCE principal that created the fileset is the owner of the fileset.
A Solaris 8 system panicked when the dfsgw add command was executed. This problem has been corrected.
The DFS configuration script incorrectly disallowed the creation of a DFS client cache greater than 1.5 GB. The DFS client cache limit is now correctly calculated based on the chunk size setting, specifically 32000 * chunk size.
The dfsstat tool for collecting and analyzing performance data is now available with DFS 3.1 on Solaris and AIX 4.x systems. This tool can be used to gather various file server, Cache Manager, and RPC statistics. For Solaris, there are both 32-bit and 64-bit versions of the tool available in the /opt/dcelocal/bin and /opt/dcelocal/bin/sparcv9 directories, respectively.
The environment variables RPC_UNSUPPORTED_NETIFS and RPC_UNSUPPORTED_NETADDRS were being ignored. This occurred because the DFS variables created to hold the values of these environment variables were being incorrectly set to NULL instead of the environment variables' values. The problem has been corrected and the environment variables now function correctly to block a network interface or an IP address from being picked up by the File Exporter daemon.
When executing the bos lsadmin command, the boserver failed and wrote to a core file. A memory leak in the BOSSVR_ListSUsers function was found and corrected.
An AIX 4.3.3 system running as an NFS/DFS Authenticating Gateway sometimes failed. This occurred because DFS code was using a field that the NFS code now uses. The problem has been corrected and the failure no longer occurs.
DFS 3.1 is now compatible with Tivoli Storage Manager (TSM) server release 4.2. The butc code in DFS 3.1 was recertified to support this new release. For AIX 4.x and for Solaris, versions of butc corresponding to TSM 3.7, TSM 4.1, and TSM 4.2 are being delivered. For AIX 4.x, select the appropriate butc version by changing the link from /usr/bin/butc to /usr/lpp/dce/bin/<butc37 | butc41 | butc42>. For Solaris, select the appropriate butc version by changing the link from /usr/bin/butc to /opt/dcelocal/bin/<butc37 | butc41 | butc42>. For AIX 5.1, only butc42 is being delivered because only TSM 4.2 is supported.
Documentation does not exist to assist users in determining an appropriate minimum time limit between successive bak dump or bak dumpinfo commands. If bak dump or bak dumpinfo commands are executed too closely in succession, errors may result. See Specifying a time interval between successive bak dump or bak dumpinfo commands for more information.
In a cell with a System Control server, unconfiguration of the DFS file server using the -config_type value admin left stale entries in the admin.up and admin.ft files. These stale entries corresponded to the already deleted server principal UUIDs during the unconfiguration process. This problem has been corrected.
Changes were made to the DFS 3.1 configuration scripts to support successful configuration of DFS 3.1 on both DCE 3.1 and DCE 3.2.
The DFS client had a leak in the vnode reference count, which caused several resource problems. For example, deleting a tarred file while the tar process was running did not free space on the fileset, even if the tar was killed. The leak has been repaired and the resource problems no longer occur.
The DFS flserver stopped receiving read requests and generated errors for an unapparent reason. This was due to the error handling in Ubik transaction commit processing that was not allowing the freeing of internal resources, leading to an exhausted resource pool and the blockage of normal processing. Ubik error handling has been corrected and internal resources are now properly freed.
It was possible to experience sporadic DFS Client access failures. This occurred when the client made a call to either the file server or the replication server because the self credentials were not being refreshed. Now, the self credentials are being properly refreshed and access failures are no longer experienced.
An incorrect address count was being reported in fshost trace messages because an internal count was not being updated consistently. The count maintenance has been repaired.
If RPC variables were not specified in the list of exported variables in the /etc/default/init file, then the DFS file server fxd process did not pick up the information. Now, when the RPC variables are not specified in the environment, the dfsbind process reads the /opt/dcelocal/etc/environment file and passes the RPC variable information to the fxd process.
PTF 3 includes all updates that were in PTF 1 and in PTF 2, plus the following items.
A Reference APAR exists for each platform that supports PTF 3 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
The system sometimes stopped responding while running some set attribute calls on DB2 datalinks, because the changes were not being flagged properly. Code changes were made in the DFS Cache manager functions to check for the setAttr flag when storing modified information for a file.
The fts lsfldb command did not work properly on detached aggregates, using the aggregate name or id to list filesets. This problem has been corrected so that the aggregate id may be used.
The DFS 3.1 configuration script displayed the following spurious error message:
0x113152c8: The value, admin, is not recognized for the keyword, config_type.
This problem has been corrected.
The DFS 3.1 server panicked when trying to write to a DM-managed aggregate on a 64-bit Solaris platform. This problem has been corrected.
A Solaris 64-bit platform panicked while trying to write to a DM-managed aggregate, because it attempted to release a mutex lock before it was obtained. This problem has been corrected.
This is an enhancement to the unsupported component dfsstat, which will be made available with PTF 3. The dfstat program collects various Remote Procedure Call (RPC) and DFS statistics that may be useful in performance analysis, load monitoring, or environment characterization. This program has been further enhanced to display the tokenprocs and mainprocs thread counts (total and idle) of the DFS fileserver process called fxd.
DFS 3.1 is now compatible with Tivoli Storage Manager (TSM) server releases 3.7 and 4.1, and TSM client releases 3.7.1, 3.7.2, 4.1.1, and 4.1.2. The butc code in DFS 3.1 was recertified to support these new releases.
The fts addsitefileset server_name aggregate2 command returned an incorrect error message as follows:
Cannot create another R/O replica on server name (aggregate1) There is already one on server name (aggregate1)
In this error message, the first aggregate name listed (aggregate1) should display as (aggregate2). This has been corrected.
On Solaris systems with 64-bit kernel, the DFS 3.1 server sometimes panicked while trying to write to a DM-managed aggregate. The problem resulted from trying to store the 64-bit structure address in a 32-bit thread id in DM. The problem has been corrected and 64-bit values are now properly ported.
On Solaris systems with 64-bit kernel, a call to the dm_path_to_fshandle API on an exported DM-managed aggregate returned an ENODEV error (which translates to "No such device"). This problem occurred in a 64-bit environment because the device number in the aggregate was specified as a 32-bit integer. This defect has been fixed.
On Solaris, a core dump sometimes resulted when unmounting the dmlfs aggregate. The problem was resolved by adding a new field, freevfs, to the DM vfs operations, to prevent accessing the wrong pointer while doing the unmount.
The DFS client crashed when the DFS Cache Manager attempted to release a write lock before it was obtained. The code was modified so that the lock is obtained before the call to release it is made.
During automatic tape backups, the wrong date was entered in the DFS central log. The butc code was modified to fix this problem.
When using the smitty utility for unconfigure, restart, and stop operations for all components, these operations were carried out on DCE components only, and not on DFS components. This problem occurred only on systems with DFS 3.1 PTF 1 and above. These operations now function correctly.
The dfsexport function was unable to detach the dmlfs aggregate when the -force flag was specified. The solution involved code modifications to check for force flags in addition to other conditions.
The smitty utility did not display the correct id for a new aggregate. Instead, it passed the value equal to the one following the highest used aggregate id. The logic was changed to generate the correct default value as the first unused id.
If a df command was issued when DCE was already running, some of the DFS filesystem entries were displayed two or more times. This problem occurred if the number of aggregates was more than 24. To fix it, the allowable number of aggregates was increased.
The DFS backup process failed because a function call from the butc command erroneously returned loopback as the host name, and there is no matching entry in the database. The code was changed to return the correct host name.
The need for the en_US locale on Solaris was not documented. Users should be aware that DFS 3.1 on Solaris 7 and 8 requires that the North American English locale (en_US) be installed on the host. This locale can be installed by the Solaris administrator at system installation. Alternatively, en_US locale support can be added to a pre-existing Solaris system. To find the Sun packages needed, go to the docs.sun.com Web site and search for "North America OS Support." Depending on the hardware architecture, the host might also need the package referred to as "North America 64 bit OS Support."
A server panic prevented users from accessing DFS space when the authentication level was set to pkt_privacy. The code was modified to correct this problem.
The replica server core dumped when trying to set the delete flag for a volume that was already marked for deletion. This problem was caused by an inconsistency between the Fileset Location Database (FLDB) and the local replica information. The problem was corrected by providing an extra conditional check for the delete flag.
PTF 2 includes all updates that were in PTF 1, plus the following items.
A Reference APAR exists for each platform that supports PTF 2 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
When a non-local user was added to a local group, the user was not automatically assigned all the group's permissions. This problem has been corrected.
In the GA software, processes controlled by the BOS (basic overseer) server all started from the same path, /opt/dcelocal/var/dfs/adm, and all wrote core files in that path if they failed. To distinguish the core files, the BOS server renamed the files to core.process_name after they were written. It was possible for core files to be corrupted if more than one process created a core file at the same time because the BOS server process was not able to complete the renaming. In this situation, the BOS server sometimes gave an error message like the following: Core file program (ftserver) does not match current program (core ignored).
This change causes the BOS server to start each process from a different path corresponding to its process name. If a process writes a core file, that file will be named core and stored in the process-specific directory. A symbolic link will be created in the /opt/dcelocal/var/dfs/adm/ path with the old-style name, core.process_name.
For example, a BOS server-controlled process that is named ftserver will be started with the path /opt/dcelocal/var/dfs/adm/ftserver. If it fails and writes a core file, the file will be /opt/dcelocal/var/dfs/adm/ftserver/core. A symbolic link to the core file will be created as /opt/dcelocal/var/dfs/adm/core.ftserver.
It was possible for the cache manager of a DFS client machine to classify a file server as nonfunctional (down) because of user authentication failures at the file server machine. This behavior has been eliminated.
DFS servers on Solaris systems sometimes panicked during a backup operation while trying to fetch directory status as a bulk operation. While trying to fetch directory status, a function attempted to release a preemption lock that already had been released; the error message mutex_exit: not owner resulted. This problem has been corrected.
DFS failed to start on Solaris machines after the Solaris 7 operating system kernel patch 106541-14 had been applied. The Solaris patch overwrites a system file, /etc/name_to_sysnum, which contains a DFS-related entry. DFS now checks for the required entry at initialization and writes the appropriate value if it does not exist. For additional information about /etc/name_to_sysnum, refer to Using a PTF to enable Solaris 8 support.
Setting the environment variable RPC_UNSUPPORTED_NETADDRS on a DFS file server after configuring DCE and DFS sometimes caused DFS clients accessing the file server machine to crash. Modifying the environment variable changed the flserver binding information in the cell directory service, and client dfsbind queries returned an unacceptable value. This problem has been corrected.
The Solaris 7 versions of two binaries, epimount and epiunmount, were included in the Solaris 8 installation of PTF 1. The code did not function correctly in a Solaris 8 environment. The correct versions of these binaries are supplied in PTF 2.
It was possible for data on a Solaris DFS 3.1 file server to be corrupted by a write operation from an IBM DFS for Windows Version 3.0 client. The problem was caused by an unacceptable value that was passed to the Solaris machine with a buffer file. The file server now checks the value before passing it to the Solaris routine.
A runtime error sometimes occurred when an IBM DFS client attempted to store data on a DFS server. The error occurred because the server attempted to change a transmission setting that the client was unable to change. The server code no longer attempts to renegotiate that setting for the duration of communication with the client.
The following updates were included in PTF 1. They also are included in PTF 2.
A Reference APAR exists for each platform that supports PTF 1 for DFS 3.1. A Reference APAR provides a list of the APARs included in the platform's PTF.
This change eliminates a synchronization error that sometimes caused a system failure on AIX machines.
Detaching a DM (data management) aggregate caused a system failure when the aggregate had events pending. The abandon_Seq() procedure is rewritten to send a CANCEL command for pending events when the related aggregate is detached.
This change is relevant only to those customers using the DFS Storage Management Toolkit.
An error caused a failure in the DFS core process when an incorrect encryption key was received from a DFS client during a remote procedure call (RPC). This problem has been corrected.
The flserver command, which is used to instantiate fileset servers, failed because of a code error. This error has been eliminated.
An incorrectly set buffer pointer in an internal function caused a DFS server failure. The problem has been fixed.
In the previous code release, it was possible for the fts release command to fail to complete when the -wait option was used and a remote procedure call (RPC) failed. This problem has been remedied.
The maximum size of the volume descriptors table maintained by the DFS kernel has been increased from 50 to 250 entries.
This change allows administrators to disable Fileset Location Database (FLDB) locking for replication server updates.
Normally, a replication server locks the FLDB entry for a fileset before loading updates from that fileset, and unlocks it when finished. Locking the FLDB entry makes the fileset unavailable to other administrative operations until the replication server has finished retrieving updates. Greater speed can be achieved if the FLDB entry is not locked during replication, but the replication server can possibly interfere with other administrative actions on that fileset if the FLDB entry is not locked. Consider the possible risks before using this option.
The repserver command now recognizes the option -unlocked. If this option is used, the replication server does not lock the FLDB entry while performing updates.
The fts move command sometimes failed for large filesets (60 GB or more) because a code error allowed the ftserver garbage collection process to delete the transaction identifier before the original fileset was completely removed. This problem has been corrected.
The code that interacts with VMM (virtual memory manager) functions in AIX has been modified to provide more information for debugging purposes.
A code error sometimes caused multiprocessor DFS client machines to stop responding. The problem, which occurred only on the AIX operating system, has been corrected.
An incorrect call to the AIX system function naix_create_attr() caused a server panic. This problem has been eliminated.
Spurious error messages, indicating that the DFS file server component had not started, appeared when DFS was started and the RPC_RESTRICTED_PORTS environment variable was set. The erroneous messages follow.
Unable to start DFS file server The components on DFS host did not start correctly
This behavior has been eliminated.
An omitted header file caused migration from DFS Version 2.1 to DFS Version 3.1 to fail. This problem has been corrected.
An uninitialized variable sometimes caused a system failure when Episode vnodes were accessed. The variable is now initialized when the system starts.
SMIT panels for configuring the DFS client were missing. The DFS client configuration options now appear on the Configure, Unconfigure, Stop, and Restart panels.
Additionally, an error that prevented fileset mount points from being deleted when the SMIT panel was used for this task has been corrected.
A code error sometimes caused a server panic when DFS servers with class B or class C Internet protocol (IP) addresses attempted to access the DFS file space. This problem has been corrected.
A class B IP address is one in which the first and second fields are fixed values assigned by Internet authorities. A class C IP address has its first three fields assigned by Internet authorities. For example, a company with a class B IP address might use IP addresses of the form 130.0.*.*, where * represents any number between 0 and 255.
This problem did not affect servers with class A IP addresses, in which only the first field contains a fixed value.
A system panic resulted when a nonroot process attempted to set its DFS authentication credential to one not owned by that process. For example, this might happen if a Web server process received a user name and password and attempted to use that credential to access DFS filespace. The problem has been eliminated.
A discrepancy between the response packet size expected by user-available functions and the size expected by kernel functions sometimes caused functions in the dm_ group to fail with the error code EFBIG. This problem has been eliminated.
This change is relevant only to those customers using the DFS Storage Management Toolkit.
It was possible for a null pointer to be dereferenced during execution of the dmepi_set_acl function, causing a system panic. This problem has been eliminated.
This change is relevant only to those customers using the DFS Storage Management Toolkit.
The dmepi_set_acl function worked incorrectly when used to modify a directory's initial object access control list (ACL). The function incorrectly modified the directory's initial container ACL instead of its initial object ACL. This error has been corrected.
This change is relevant only to those customers using the DFS Storage Management Toolkit.
The cfgdmepi system function failed to load the kernel extension dmlfs.ext because of an uninitialized variable. This problem has been eliminated.
This change is relevant only to those customers using the DFS Storage Management Toolkit.
The -noautoquery option in the backup tape coordinator (butc) command did not work as specified. The tape coordinator process prompted the administrator to mount the first tape regardless of whether the -noautoquery option was used. This error has been corrected.
A code error allowed the dfsd process to return the message that dfsd had started successfully before cache initialization was complete. The process now waits for confirmation that the cache is initialized before issuing the success message.
The dfsexport -detach command failed undetectably because of a corrupted value in the agfs_fstype field of the fileset identifier (fsid). This problem has been eliminated.
Attempting a chmod() action that was forbidden by a data management (DM) application on the server caused the cache manager to emit a continuous sequence of AFS_StoreStatus remote procedure calls until the client was restarted. This problem has been eliminated.
This change is relevant only to those customers using the DFS Storage Management Toolkit.
The rmdir command failed without giving an error message when executed on a nonempty DFS directory. The command now returns the error code ENOTEMPTY in this case.
It was possible to overflow the allotted storage space for reference count values in the cm_volume.c file by setting the allowed number of status cache entries in the DFS cache manager to a value greater than 30,000. (The number of status cache entries in the DFS cache manager is configured by using the dfsd -stat command option.) A server panic, documented by a failed assertion message, resulted from the overflow.
Additional storage space has been allotted for the reference count variable to correct this problem.
An error prevented the links agfs, lfs, and dmlfs from being created in the /usr/lib/fs/ path during installation. This error has been eliminated.
A programming error allowed a data management lock structure to be referenced after its destruction if a kernel module removal function failed. A system panic resulted in Solaris systems. This error has been eliminated.
A system panic occurred in 64-bit Solaris systems during removal of a file from a type DMLFS aggregate. The problem occurred because of a recursive procedure call, which has been eliminated.
The default location for the DFS client cache on AIX systems was erroneously set as /var/adm/dfs/cache. The default location now is set to the correct value, which is /var/dce/adm/dfs/cache.
Migrating files from a DM-managed aggregate failed on 64-bit Solaris systems, with an EFAULT error code being returned from the DMAPI calls dm_read_invis() or dm_write_invis(). This problem has been eliminated.
This change is relevant only to those customers using the DFS Storage Management Toolkit.
A code error caused the dm_get_bulkattr() function to return incorrect status information on DM-managed aggregates. This problem, which occurred only on Solaris systems, has been eliminated.
This change is relevant only to those customers using the DFS Storage Management Toolkit.
(c) IBM Corporation 2001-2005. All Rights Reserved