**************************** ECO SUMMARY INFORMATION **************************** Release Date: 30-AUG-2006 Kit Name: AXP_DNVOSIECO03-V732.PCSI-DCX_AXPEXE Kit Applies To: OpenVMS ALPHA V7.3-2 Approximate Kit Size: 24169 blocks Installation Rating: INSTALL_2 Reboot Required: Yes Superseded Kits: AXP_DNVOSIECO02-V732 Mandatory Kit Dependencies: DEC-AXPVMS-DECNET_OSI-V0703-2-1.PCSI Optional Kit Dependencies: None Checksums: AXP_DNVOSIECO03-V732 Checksum: 3047684984 AXP_DNVOSIECO03-V732 MD5 Checksum: E9E246FDAD46E0E6D51FC942F8A79860 HP DECnet-Plus for_OpenVMS_Alpha___________________________________ Release Notes for Version 7.3-2 ECO03 August 2006 This ECO applies to DECnet-Plus V7.3-2 Alpha systems only. To verify that you are running the correct version of DECnet-Plus, enter the following NCL command: NCL SHOW IMPLEMENTATION The version should read: DECnet-Plus for OpenVMS Version V7.3-2. Hewlett-Packard Company Palo Alto, California © Copyright 2006 Hewlett-Packard Development Company, L.P. Confidential computer software. Valid license from HP required for possession, use, or copying. Consistent with FAR 12.211 and 12.212, Commercial Computer Software, Computer Software Documentation, and Technical Data for Commercial Items are licensed to the U.S. Government under vendor's standard commercial license. The information contained herein is subject to change without notice. The only warranties for HP products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. HP shall not be liable for technical or editorial errors or omissions contained herein. Intel and Itanium are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. Printed in the US 1 Installation Install this kit with the POLYCENTER Software Installation utility (PCSI) by logging into the SYSTEM account and entering the following command at the DCL prompt: $ PRODUCT INSTALL DNVOSIECO03/SOURCE=[location of the .PCSI file] ________________________ Note ________________________ For the corrections to take effect, you must reboot the system. If you are a member of a VMScluster, you must reboot the entire cluster. You can postpone rebooting if it is not convenient for you to do so now; however, the new drivers will not take effect until after a reboot. This kit contains a correction to the CDI$SHR executable (see CDI$SHR.EXE in the Corrections section). To ensure that this correction takes effect, use the following command to flush the CDI cache after you install this ECO kit: $ MCR NCL FLUSH SESSION CONTROL NAMING CACHE ENTRY "*" ______________________________________________________ 2 New PCSI INSTALL Recovery Options and UNDO Command Not Supported This ECO kit does not support the /RECOVERY_MODE and /SAVE_ RECOVERY_DATA qualifiers to the PRODUCT INSTALL command. As a result, the PRODUCT UNDO PATCH command is also not supported. Using either the UNDO command, or the /RECOVERY_MODE qualifier and pressing Ctrl-Y, causes unpredictable results. 3 DECnet-Plus Support for Mixed Case Password The following improvements have been made to the NET$OSVCM.EXE and NET$SESSION_CONTROL.EXE images: o OpenVMS version 7.3-2 and above allows you to use mixed- case and extended characters in passwords when PWDMIX is specified as login flags. 1 DECnet has two different methods for user authentication. This is controlled by the sysgen parameter, NET_CALLOUT. If this parameter is set to 255, then the verification and proxy lookups are performed in LOGINOUT rather than DECnet. In earlier versions of DECnet, when DECnet performed the user verification there were restrictions on using mixed-case and extended characters in passwords. This version of DECnet supports mixed-case password for task to task communication and remote file access. 4 Corrections for Base Components ________________________ Note ________________________ A few of the corrections included in these Release Notes describe late corrections made for V7.3-2 that were not included in the V7.3-2 Release Notes. ______________________________________________________ 4.1 SYS$LIBRARY:CDI$SHR.EXE: Image Identification: "V7.3-2 ECO03" 25-AUG-2006 21:45:07.84 ECO03: - This release fixes the creation of multiple BG devices that are never deleted, eventually consuming the entire BYTLM quota. ECO02: - Prevented an access violation within CDI that can lead NET$ACP into a CPU hogging loop. ECO01: - Corrected a problem with Domain proxies brought about by using Session Control V3 (SC3) with RFC1006 connections. The problem was corrected by reverting IP address towers back to SC2 (see SYS$LOADABLE_ IMAGES:NET$SESSION_CONTROL.EXE corrections). CDI is now used to look up the node name for RFC1006 connections. This can cause rejected connections if the backtranslation entries in DECdns are incorrect. To ensure that this change takes effect, use the following 2 command to flush the CDI cache after you install this ECO kit: $ MCR NCL FLUSH SESSION CONTROL NAMING CACHE ENTRY "*" 4.2 SYS$SYSTEM:CTF$DCP.EXE: Image Identification: "V7.3-2 ECO03" 25-AUG-2006 21:47:21.58 ECO03: - This release fixes the system hang while executing the TRACE START/LIVE command as the CTF$DCP process completely utilizes the CPU in kernel mode. ECO01: - Prevent possible system crash by correcting deallocation logic for trace records and CTF work queue elements in CTF$DISPATCH. 4.3 SYS$LIBRARY:CTF$ROUTING_ANALYZE.EXE: Image Identification:"V1.0" 25-AUG-2006 21:48:02.31 ECO03: - This release fixes the problem of decoding LAN Link State Packet (LSP) messages. This release also supports Dynamic Hostname, Router ID and padding of messages. 4.4 SYS$SYSTEM:DECNET_REGISTER.EXE: Image Identification:"V7.3-2 ECO03" 25-AUG-2006 21:47:20.61 ECO03: - The wildcard node name lookup through DECNET_REGISTER is modified to continue the lookup even on encountering a failure in one of the node name lookups. - This release allows to modify/update/register the node information related to a host within the same subnet. 3 4.5 SYS$COMMON:DTSS$PROVIDER_NTP.EXE: Image Identification: "V1.0" 20-OCT-2005 01:57:15.18 ECO02: - Corrected a problem in DTSS time provider program which resulted in failure to take time from NTP server when more than one NTP server is specified in the list and if at least one of them does not respond. 4.6 SYS$SYSTEM:DTSS$SERVICE.EXE: Image Identification: "DECDTS V2.1-0" 25-AUG-2006 21:47:37.05 ECO03: - This release enhances DTSS to work efficiently on multiple interface systems. Due to this change; a server visible on more than one interface will be listed with duplicate entries with the following command: $MCR NCL SHOW DTSS DECNET LOCAL SERVER *. However only one instance of a particular server will be shown as used for synchronization as TRUE. The duplicate entries will be set to FALSE. The entries will be considered alternatively in each synchronization cycle. This also fixes the server running on a multihomed system being not visible on some systems (Mostly Clerks). - DTSS will request the TDF change notification as absolute timer requests thus preventing spurious notification leading to delayed daylight switches. - The values parsed for the system wide logical, DTSS$ETHERNET_DEVICE is not case sensitive. In earlier versions, the values parsed to this logical were accepted in upper case only. - This release has a fix to display the value of the "Next TDF change" field as "" when there is no Daylight Saving Time (DST) for the configured subregion and to display the time for the next TDF change when DST is present. 4 ECO02: - This release fixes memory leak in DTSS clerk during the execution of DTSS SHOW and SET commands. All of the SHOW and SET DTSS commands used to leak about 655 bytes per iteration. - Modified the command,"show dtss all status" to display some of the missing attributes. - DTSS will now read the zone information from binary timezone files. Thus the problem of changing the TDFs wrongly based on previous year's rule is rectified. If DTSS cannot find the applicable rule for the current timezone from binary timezone files, it will go ahead with the rule that the SYS$TIMEZONE_RULE says. - This release fixes the loss of Global DTSS servers loaded from config file, when these Global servers are to be reached using TCP/IP. - DTSS will request the TDF change notification as absolute timer requests thus preventing spurious notification leading to delayed daylight switches. ECO01: - Removed the static device tables formerly used determine the devices used by DTS. Now, DTS uses the $DEVSCAN and $GETDVI system services to build a list of devices that have a device class of DC$SCOM (synchronous communication device), a device characteristic of DEV$V_NET, a device status of UCB$V_ONLINE and UCB$V_ TEMPLATE, and a device name in the form _xxy0:. If If a device satisfies the device test but fails device initialization, an error is returned. The logical name DTSS$ETHERNET_DEVICE can be used to provide a list of devices that DTS should NOT use. All devices must be in the form _xxy0:. The string can contain spaces and other text which is ignored by DNS. For example, the following command tells DTS not to use the _EIA0: and _FWA0: devices. $ DEFINE/SYSTEM DTSS$ETHERNET_DEVICE "Don't use _EIA0: and _FWA0: " - Corrected a problem that caused a subsequent device to remain uninitialized if a lower device of the same type was offline or unavailable. 5 - Corrected a problem that caused programs using the JAVA runtime code to go into a COM state loop when the TDF change occurs. A TDF change notification event was added. Also, there are some VMS/THREADS corrections that are related to this problem. 4.7 SYS$LIBRARY:DTSS$SHR.EXE: Image Identification:"DTSS V2.1-0" 20-OCT-2005 01:51:22.06 ECO02: - Corrected a problem involved with API routine VMS_ LOCALTIME giving the wrong time, when the timezone was set to AUSTRALIA/WEST. 4.8 SYS$UPDATE:DTSS$TIMEZONE_RULES.DAT: ECO03: - The DTSS$TIMEZONE_RULES.DAT file has been modified to extend the Daylight Saving Time (DST) in accordance with the legislation passed by the US Congress, to extend the DST starting 2007. 4.9 SYS$LOADABLE_IMAGES:LES$LES_V30.EXE: Image Identification: "LES V3.0-2" 20-OCT-2005 01:55:42.53 ECO02: - This release fixes a system crash caused by the abnormal deletion of a CLNS CDB on a heavily loaded system. 4.10 SYS$SYSTEM:NET$ACP.EXE: Image Identification: "V7.3-2 ECO02" 20-OCT-2005 01:56:48.53 ECO01: - Include for compatibility with CDI$SHR. 6 4.11 SYS$MANAGER:NET$CONFIGURE.COM: ECO03: - This release fixes the problem with Menu Option 2 in the NET$CONFIGURE script not prompting for a Start of the network. - This release fixes the problem of NET$CONFIGURE input parsing error for Data link name and Routing circuit name. - In this release DECnet checks for the license keys before installation and configuration. If the license keys are not loaded the following message is displayed before starting DECnet - "Please Register/Load DECnet PAK's to start DECnet" 4.12 SYS$LOADABLE_IMAGES:NET$DRIVER.EXE: Image Identification: "X-4" 25-AUG-2006 21:47:48.32 ECO03: - This release fixes the system crash when a TYPE command is executed onto a NET device. The command, "TYPE//out=NETxyz:" will now report a 'FILNOACC' error. ECO02: - This release fixes INVEXCEPTN crashes in EXE_ STD$QIOACPPKT_C routine. The crash happened when the FAL process attempted a rundown which forced a cancel operation on the net devices to which FAL has assigned a channel. - Corrected to prevent a NET device going to ONLINE state when a SET DEVICE/RESET=ERROR command is attempted. ECO01: - Added spinlock synchronization to two routines responsible for determining that a NET device's connection count has gone to zero. Before the spinlock synchronization, each routine could independently detect the condition and attempt to close the association at the same time. This resulted in a hung NET device. 7 - Improved the reporting of connection timeouts in the transmit and receive queue processing code. The code now returns the traditional, more specific timeout error status SS$_LINKEXIT instead of the generic error status SS$_FILNOTACC. - Corrected a problem causing a process using the NET device to hang in the RWAST state. This problem was seen when multiple deassign calls were made for a connection instance. (For example, a remote system requesting a disconnect at the same time as the local system.) - Corrected a problem seen when a user program received notification of I/O completion using LOCAL event flags. Occasionally, the event flag was set before the IOSB and the data buffer were properly set. 4.13 SYS$SYSTEM:NET$MGMT.EXE: Image Identification: "V1.0" 20-OCT-2005 01:56:19.98 ECO01: - Modified internal tables to support more than 500 ports. Previously, the tables were hard-coded to support 500 ports; they are now allocated dynamically. 4.14 SYS$LIBRARY:NET$NISCS_LAA.EXE: Image Identification:"MOPLAA V3.0" 25-AUG-2006 21:47:26.90 ECO03: - This release fixes the cluster satellite boot failure using DECnet MOP if the system disk has HBMM enabled. 4.15 SYS$LOADABLE_IMAGES:NET$OSDRIVER.EXE: Image Identification: "X-4" 25-AUG-2006 21:48:03.60 ECO03: - This release fixes the system crash when a TYPE command is executed onto a NET device. The command, "TYPE//out=NETxyz:" will now report a 'FILNOACC' error. 8 ECO02: - Corrected a bug that was present in the redirect I/O completion code of DECnet-Plus V7.3 ECO04, V7.3-1 ECO03, V7.3-2 ECO01, and V8.2. This code is traversed when any OSI application uses the redirect functionality of OSAK. Various symptoms of this bug were occasionally observed on Alpha systems, but the symptoms occurred more frequently on I64. Such symptoms included: deaccess hangs of OS devices, corruption of the CPU-specific I/O postprocessing queue (whose queue headers reside at offsets CPU$L_PSFL and CPU$L_PSBL of the CPU database), freelist corruption, and a variety of other crashes when deallocated structures had been referenced. - This release fixes INVEXCEPTN crashes in EXE_ STD$QIOACPPKT_C routine. The crash happened when the FAL process attempted a rundown which forced a cancel operation on the net devices to which FAL has assigned a channel. ECO01: - Improved the reporting of connection timeouts in the transmit and receive queue processing code. The code now returns the traditional, more specific timeout error status SS$_LINKEXIT instead of the generic error status SS$_FILNOTACC. - Corrected a problem causing a process using the OS device to hang in the RWAST state. This problem was seen when multiple deassign calls were made for a connection instance. (For example, a remote system requesting a disconnect at the same time as the local system.) - Corrected a problem seen when a user program received notification of I/O completion using LOCAL event flags. Occasionally, the event flag was set before the IOSB and the data buffer were properly set. 9 4.16 SYS$LOADABLE_IMAGES:NET$OSVCM.EXE: Image Identification: "V7.3-2 ECO03" 25-AUG-2006 21:47:53.27 ECO03: - This release fixes the system crash when a TYPE command is executed onto a NET device. The command, "TYPE//out=NETxyz:" will now report a 'FILNOACC' error. ECO02: - This release has changes in NET$OSVCM which would fix the hung OS devices and process going to LEF state while attempting to deaccess the OS device. - Prevent a system crash of type "DECnet fatal error" in SESSION CONTROL due to flipping of the connection ID and transport association ID on a heavily loaded system. ECO01: - Corrected a possible double deallocation problem that occurred on incoming sessions where the remote session partner and the device driver simultaneously attempted to close the session. 4.17 SYS$LIBRARY:NET$PROCESS_EMAA.EXE: Image Identification: "V7.3-2 ECO02" 20-OCT-2005 01:51:38.54 ECO01: - Corrected a system crash caused by a problem with MCRP buffer allocation and initialization in the NET$EMAA_ Response routine. The buffer consists of a fixed length control block which includes a pointer to a variable length output string that is appended to the control block. The code was not properly initializing the output string pointer to null when the output string length was zero. - Corrected a system crash seen while managing a MOP entity. A block deallocation routine was running at an elevated IPL (IOLOCK8) when a page fault occurred. The elevated IPL is not required and was removed, thus allowing the page faulting to occur at a lowered IPL. 10 4.18 SYS$LOADABLE_IMAGES:NET$ROUTING_ES.EXE: Image Identification: "V7.3-2 ECO02" 20-OCT-2005 02:05:39.84 ECO01: - Added a check for the supplied MAC address pointer to the LAN circuit receive completion code. Previously, if the MAC address pointer was invalid a system crash occurred. Now, the VCRP containing the invalid MAC address pointer is simply discarded. - Ensure that an FDDI circuit is enabled before sending the last chance ES_HELLO (for all bcst circuits). - Corrected a system crash seen when disabling alias ports. The code was not handling the case where there were no ports to disable. - Corrected the code to allow a maximum of 32 routing circuits as described in the Software Product Description (SPD). Previously, the maximum was incorrectly limited to 16. - Corrected a system crash caused by improper integrity checking of incoming ES Hello messages. Also, if the integrity checking failed, a bugcheck was forced. This integrity check was corrected and the routine returns appropriate status instead of bugchecking. 4.19 SYS$LOADABLE_IMAGES:NET$ROUTING_IS.EXE: Image Identification: "V7.3-2 ECO02" 20-OCT-2005 01:55:39.24 ECO02: - This release corrects a problem in DECnet Host based Router forwarding process. The problem manifested into connectivity issues between Phase V and Phase IV end systems as DECnet HBR was forwarding the Phase IV packet on a Phase V adjacency in the Phase IV format itself. - This release corrects a problem in Host Based router's Phase IV designated router election algorithm. This will avoid a scenario where there is no Phase IV designated router in a LAN, which has more than 1 DECnet Host based routers. 11 - The release corrects a problem in DECnet Host Based routing update alogorithm. The problem was discovered when DECnet HBR was configured as a pure OSI router and Connected to CISCO routers. DECnet HBR was unable to connect via CISCO routers to the hosts connected to the other end of CISCO routers due to the problem in HBR update process. ECO01: - Corrected a problem seen when an alias port is disabled. Several minutes of address unreachable pdu discard events (with a discard reason of destination address unknown) were seen. The routing module now sends out an ES Hello message with the disabled alias NSAPs marked with a holding time of zero. This causes routers to immediately delete the adjacency for the alias. - Deleted unnecessary Phase IV parameter checks when in an OSI-only host-based routing environment. - Modified the output of the SHOW ROUTING ALL command to omit all Phase IV parameters in a pure OSI host-based routing environment. A further correction was required to handle all possible cases of the NET$ROUTING_ STARTUP.NCL script. 4.20 SYS$LOADABLE_IMAGES:NET$SESSION_CONTROL.EXE: Image Identification: "V7.3-2 ECO03" 25-AUG-2006 21:47:57.47 ECO03: - Corrected the problem of displaying junk characters in the create/delete audit messages with DECnet connections. - This release fixes a security problem wherein the correct password appears in the audit logfile, when a connection attempt is made with the right password after a "network breakin" event has occurred. - This release fixes the SSRVEXCEPT crash in AST_ COMPLETION routine within the NET$SESSION_CONTROL module. 12 ECO02: - Corrected a problem in NET$SESSION_CONTROL which resulted in DECnet Fatal error Bugchecks. The problem was observed on a busy system when shutting down network. This resulted in a window that could leave a connection in inconsistent state while another thread was tearing it down. - Corrected a problem in session layer to prevent system crash (INVEXCEPTN) while attempting to shutdown the network. - This release fixes CPUSPINWAIT crash with the CPUs getting into a contention over SCS and LOCKMGR spinlocks. This was due to the lowering of the IPL below 8 holding the SCS spinlock. ECO01: - Modified the towers created for IP addresses in the DECdns namespace to always use SC2. When support was added for storage of IP addresses, the Session Control value was forced to SC3. Also, Session Control now maintains tower and backtranslation softlinks for both SC2 and SC3. Previously, it maintained softlinks for SC3 only. See also the corrections for SYS$LIBRARY:CDI$SHR.EXE. - Corrected a rare SMP issue with the cleanup of orphan sessions. The code attempted to deliver a connect reject response to the caller before the transport takedown had completed. Because session control did not expect this type of event, a crash occurred. The code now waits for the transport takedown to complete. - Corrected a problem seen when a user program received notification of I/O completion using LOCAL event flags. Occasionally, the event flag was set before the IOSB and the data buffer were properly set. - Fix incoming alias table. All named objects were incorrectly assumed to be the TASK object. 4.21 SYS$STARTUP:NET$SHUTDOWN.COM: ECO03: - Corrected to change the order of shutdown of DECdns server and Clerk to prevent the system hang during the DECdns server shutdown while executing NET$SHUTDOWN. 13 4.22 SYS$LOADABLE_IMAGES:NET$TPCONS.EXE: Image Identification: "V7.3-2 ECO03" 25-AUG-2006 21:47:58.38 ECO03: - This release fixes the DECnet over IP connection issue due to a race condition between the DECnet startup and the PWIP driver initialization. ECO02: - This release fixes system crash with DECnet fatal error during normal shutdown. This was due a race condition between DECnet and TCP/IP PWIP driver while attempting to shutdown. ECO01: - Corrected a forced crashdump caused during network shutdown. The Deactivate TPCB routine was incorrectly called before the connection reached the UNBND state. This occurred because the calling code had attempted to UNBIND the PWIP driver at the same time that the PWIP driver was shutting down. This resulted in an ERROR_ACK to the UNBIND request which was not properly handled. - Modified the code to correctly respond to an ERROR_ ACK response from the PWIP driver in response to a BIND request. - Modified the code to correctly set up local IP address and TCP port fields. See also NET$TRANSPORT_OSI for more information about how this change allowed the correct local RFC1006 and remote RFC1006 IP address and port number characteristics to be displayed. - Modified the code to correctly handle a control block inconsistency seen during processing of an incoming disconnect when the local system is also initiating a disconnect (rather than force a system crash). - Removed sanity checking in DeallocateTPCB that was no longer compatible with newer HP TCP/IP Services for OpenVMS releases. 14 4.23 SYS$LOADABLE_IMAGES:NET$TRANSPORT_NSP.EXE: Image Identification: "V7.3-2 ECO02" 20-OCT-2005 01:56:17.50 ECO02: - Corrected a code path used when the transmit window approached zero. In particular, this problem was seen when the RMS NETWORK_BLOCK_COUNT qualifier was set to 127 in an attempt to improve performance. Because this caused the transmit window to approach zero frequently, the incorrect code actually caused decreased performance. - This release has changes in NET$TRANSPORT_NSP.EXE which will result in all pending I/Os to complete with SS$_ABORT status when a cancel is issued prior to this, pending I/Os would complete with SS$_LINKDISCON status after a CANCEL. - This release has changes in NET$TRANSPORT_NSP.EXE which will prevent crashes (of type NETNOSTATE) due to faulty frame received during connect initiate attempt. ECO01: - Corrected a system crash caused by a fork to 0 seen during retransmissions on high traffic LAN connections. The problem occurred when the LAN driver was completing the initial transmission request while NSP had already timed out and was trying to retransmit the data. A LAN driver change for fastpath support allowed both processes to modify the scratch area (which contains the fork address) of the transmit VCRP at the same time. - Corrected a problem resulting in poor throughput when NSP was using piggyback ACKs. The code now sends a standalone ACK if no data segment is queued for transmission. 15 4.24 SYS$LOADABLE_IMAGES:NET$TRANSPORT_OSI.EXE: Image Identification: "V7.3-2 ECO03" 25-AUG-2006 21:47:48.34 ECO03: - This release fixes a system crash that is caused by accessing the fields of an allocated but non-activated port. Such a scenario occurs when a system receives TPDU with both source and destination reference fields set to zero. The TPDU validation routine now checks for the source and destination reference fields and if both are zero, it discards the TPDU as "Unknown Reference". - This release fixes the erroneous "unknown-ref" OPCOM messages log, when a Class 0 Data Transfer Transport Port Data Unit (DT-TPDU) is received. ECO01: - Corrected a system crash caused by a fork to 0 seen during retransmissions on high traffic LAN connections. The problem occurred when the LAN driver was completing the initial transmission request while NSP had already timed out and was trying to retransmit the data. A LAN driver change for fastpath support allowed both processes to modify the scratch area (which contains the fork address) of the transmit VCRP at the same time. - Corrected a session hang problem seen when a disconnect was received from the remote system and was being processed at the same time that a request to deaccess the NET channel came from the local system. - Corrected a double deallocation of the NCCB seen during the disabling of the OSI transport when DECnet is shutting down. - Modified the code to correctly display the local RFC1006 and remote RFC1006 IP address and port number characteristics instead of "" and "0" respectively. - Ensure that all routines calling a NCCB queue manipulation routine have IOLOCK8. 16 4.25 SYS$LOADABLE_IMAGES:SYS$NETWORK_SERVICES.EXE: Image Identification: "V7.3-2 ECO03" 25-AUG-2006 22:49:46.05 ECO03: - Corrected to prevent the occurrence of "%DECnet-W-NOREAD, could not read SYS$SYSROOT:[SYSEXE]NET$CONFIG.DAT" warning message while attempting to read the NET$CONFIG.DAT file during a system reboot. This problem occurs with larger disks where the cluster_size parameter is set to a value greater than 512. ECO02: - Fixed NET$ACP process going into a tight CPU bound loop due to a stack overflow. - This release fixes a system crash when the CTF tracepoints are being disabled. ECO01: - Corrected a system crash seen when CTF tracing was active during a shutdown of the OSI transport. During transport shutdown, the common code for handling trace terminations failed to verify the existence of the trace information buffer before using a pointer within the buffer. - Update ECO number in the DECNET_VERSION item code to "00051001". 4.26 New Trace Images Included in SYS$LIBRARY for Compatibility: The following CTF trace images are included: ECO03: CTF$ROUTING_ANALYZE.EXE "V1.0" 25-AUG-2006 21:48:02.31 ECO02: 17 CTF$DNA_ANALYZE.EXE "X-7" 20-OCT-2005 01:56:47.80 CTF$ESEVENT_ANALYZE.EXE "V7.3-2 ECO02" 20-OCT-2005 02:05:37.28 CTF$NSPTP_ANALYZE.EXE "X-2" 20-OCT-2005 01:56:24.49 CTF$OSITP_ANALYZE.EXE "I-10" 20-OCT-2005 01:56:45.39 CTF$OSVCM_ANALYZE.EXE "X-2" 20-OCT-2005 01:56:37.61 CTF$SCL_ANALYZE.EXE "X-10" 20-OCT-2005 01:57:23.64 CTF$TPCONS_ANALYZE.EXE "I-5" 20-OCT-2005 01:57:02.17 CTF$VOTS_ANALYZE.EXE "X-7" 20-OCT-2005 01:57:01.61 18 5 Configuring Link State Routing Nodes Using the ISIS$CONFIGURE Procedure This ECO release includes support for the new ISIS$CONFIGURE procedure available in the HP DECnet-Plus for OpenVMS Version 8.2 release. The NET$CONFIGURE procedure supports configuring host- based routing nodes that use the routing vector routing algorithm. If you want to create a host-based routing node that uses the link state routing algorithm, you must use the ISIS$CONFIGURE procedure. If you want to use the ISIS$CONFIGURE procedure, see Appendix D in the new HP DECnet-Plus for OpenVMS Installation and Configuration manual. This manual is available in the DECnet documentation area of the OpenVMS documentation web site (http://www.hp.com/go/openvms/doc/). 6 CTF Utility Now Requires Process Privileges The CTF utility has always required that the invoking process have certain rights identifiers in order to use the START and STOP commands to control the collection of trace records. These rights identifiers are described in Chapter 3 of the DECnet/OSI for VMS CTF Use manual. Now, in addition to one or more of the required rights identifiers, the invoking process must have the following process privileges before invoking CTF for any purpose: ALTPRI, BYPASS, CMKRNL, NETMBX, PRMGBL, PRMMBX, PSWAPM, SYSGBL, SYSLCK, SYSNAM, TMPMBX, and WORLD If you invoke CTF from a process without the required process privileges, the utility exits with an error message indicating the first of the required privileges not granted. 7 DECdns Update 19 7.1 Files Patched or Replaced: SYS$COMMON:[SYSEXE]DNS$ADVER.EXE "V7.3-2 ECO03" 25-AUG- 2006 21:47:54.84 SYS$COMMON:[SYSEXE]DNS$SERVER.EXE "V7.3-2 ECO02" 20-OCT- 2005 01:57:49.11 7.2 Problems Corrected in DECdns Images ECO03: - This release fixes the DECnet startup problems occurring due to DNS clerk startup failure. Better startup handling for DNS clerk is added. ECO01: - Corrected a problem with Domain proxies brought about by using Session Control V3 (SC3) with RFC1006 connections. The problem was corrected by reverting IP address towers back to SC2 (see SYS$LOADABLE_ IMAGES:NET$SESSION_CONTROL.EXE corrections). CDI is now used to look up the node name for RFC1006 connections. This can cause rejected connections if the backtranslation entries in DECdns are incorrect. - Removed the static device tables formerly used determine the devices used by DNS. Now, DNS uses the $DEVSCAN and $GETDVI system services to build a list of devices that have a device class of DC$SCOM (synchronous communication device), a device characteristic of DEV$V_NET, a device status of UCB$V_ONLINE and UCB$V_ TEMPLATE, and a device name in the form _xxy0:. If If a device satisfies the device test but fails device initialization, an error is returned. The logical name DNS$ETHERNET_DEVICE can be used to provide a list of devices that DNS should NOT use. All devices must be in the form _xxy0:. The string can contain spaces and other text which is ignored by DNS. For example, the following command tells DNS not to use the _EIA0: and _FWA0: devices. $ DEFINE/SYSTEM DNS$ETHERNET_DEVICE "Don't use _EIA0: and _FWA0: " - Corrected a problem that caused a subsequent device to remain uninitialized if a lower device of the same type was offline or unavailable. 20 - Corrected a problem that could cause more than three copies of the DNS$CHFAIL.LOG to be present on the system. This condition occurred when the file version of DNS$CHFAIL.LOG cycled through the upper version boundary (that is, cycling from version 32767 back to version 1). The code now correctly handles this situation and maintains three copies of the file. File versions are rolled over when the file version reaches 32000. For example, if versions 31998, 31999, and 32000 are present, the code renumbers the files to version 1, 2, and 3, respectively. - Corrected a problem that could cause multiple copies of the file used to maintain the current cache file version (SYS$SYSTEM:DNS$CACHE.VERSION) to be present on the system. This condition occurred when the file version of DNS$CACHE.VERSION cycled through the upper version boundary (that is, cycling from version 32767 back to version 1). The code now correctly handles this situation and maintains a single version of the file. 8 Restriction When Using Applications Supported Using Cluster Aliases Due to a design restriction, incoming ALIAS connections are forwarded to cluster members somewhat differently depending upon whether the SESSION CONTROL APPLICATION entity's ADDRESS set contains a NUMBER parameter (equivalent to Phase IV's connect by object number) or a TASK parameter (equivalent to Phase IV's connect by object name). For an incoming connect by object number, each ALIAS entity has access to a cluster-wide object number table that specifies which object numbers have been defined for each cluster member and the status of their INCOMING ALIAS characteristic. The result is that the connection forwarding mechanism works as expected, that is, connections are forwarded or not forwarded to cluster members correctly depending upon whether the application exists on the member and depending on the setting of the INCOMING ALIAS parameter. 21 However, the mechanism for mapping incoming connections by object name using this cluster-wide object number table does not exist and the result is that connections are forwarded to a given cluster member regardless of whether the application exists on a given member and regardless of the setting of the INCOMING ALIAS parameter on a given member. If you wish to restrict access to a network application supported using a cluster alias (either by not having the application present on a member or by setting the INCOMING ALIAS characteristic to FALSE on a member), you must create the SESSION CONTROL APPLICATION by supplying the NUMBER keyword for the ADDRESS characteristic (that is, supplying the object number), so that connection forwarding works as expected. 9 Tuning to Support the Larger CML BYTLM Value The default BYTLM value for CML has been increased from 10,000 to 64,000. This correction was done to prevent NCL from returning the NCL-E-OVERFLOWBYTLM error message when large amounts of data were needed from CML. If you have ever experienced this problem, or if your MAXBUF value is already set to 64,000 and you have received NCL-E- NORESOURCEAVAIL error messages, then you should make the changes described in this section. To fully take advantage of the increased BYTLM value, do the following: 1. Increase the value of MAXBUF to 64,000. 2. Raise the BYTLM quota on any account that might issue NCL commands requiring the increased resources to at least 65,000. Be sure to do this to at least the SYSTEM, CML$SERVER, and MIRRO$SERVER accounts. For example, to modify the CML$SERVER account, use the following command: $ run sys$system:authorize UAF> modify cml$server/bytlm=65000 3. Check for the existence of the logicals NET$ACP_ BYTLM, NET$EVD_BYTLIM, and NET$MOP_BYTLIM. If any of these logicals are defined and they are set to a value lower than 65,000, modify their entries in 22 SYS$MANAGER:NET$LOGICALS.COM to raise their value to at least 65,000. 4. Check for the existence of the NET$LESACP_BUFFER_ LIMIT logical. If this logical is defined and it is set to a value lower than 65,000, modify its entry in SYS$STARTUP:NET$LES_STARTUP.COM to raise the value to at least 65,000. 5. Check for the existence of the NET$ADVER_BUFFER_LIMIT logical. If this logical is defined and it is set to a value lower than 300,000, modify its entry in SYS$STARTUP:SYLOGICALS.COM to raise the value to at least 300,000. 6. Reboot the system to make all the preceding changes effective. If, after applying the measures described in the previous list, you now receive NCL-E-NORESOURCEAVAIL errors instead of NCL-E-OVERFLOWBYTLIM errors, this generally indicates that your process BYTLM quota is not set high enough. Set the quota to a value slightly higher than the MAXBUF value. Setting process BYTLM quotas to 65,000 (as specified in the previous list) should be sufficient to avoid this problem. 23