SYS V5.0 ECO Kit for VSI OpenVMS Alpha V8.4-2L2

INSTALL_1: To be installed by all customers.

This installation rating serves as a guide to which customers should apply this remedial kit.

Reference the Disclaimer of Warranty and Limitation of Liability Statement.

A reboot is required after installing this kit.

VMS Software, Inc. strongly recommends that a reboot be performed immediately after kit installation to avoid system instability.

If you are installing this kit in an OpenVMS cluster, any systems that share the same system disk with the installing system must also be rebooted in order to make use of the new images.

VSI OpenVMS Alpha V8.4-2L2

Problem Description

Exception handling within the security sub-system may trigger an unexpected process crash under certain conditions.

This problem is corrected with this ECO kit.

Images and/or Files Affected

• [SYS$LDR]SECURITY.EXE
• [SYS$LDR]SECURITY.STB
• [SYS$LDR]SECURITY_MON.EXE
• [SYS$LDR]SECURITY_MON.STB

VSI Case Identifier

Jira BO-1871

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

A problem within pthread pshared object handling may cause a system crash with a footprint similar to the following:

Crash Time:        17-APR-2018 09:32:54.82
Bugcheck Type:     INVEXCEPTN, Exception while above ASTDEL
Node:              <nodename>
CPU Type:          HP rx6600  (1.59GHz/12.0MB)
VMS Version:       V8.4-2L1
Current Process:   JDBC_P4_00004DA
Current Image:     $1$DGA159:[SYS0.SYSCOMMON.RDB$JDBC.0703-5V0H8E]RDBJDBCEXEC73.EXE
Failing PC:        FFFFFFFF.805D13F0    EXE$PSHARED_MUTEX_UNBLOCK_C+003D0
Failing PS:        00000000.00000603
Module:            PROCESS_MANAGEMENT   (Link Date/Time: 27-JUL-2016 10:30:20.13)
Offset:            001028F0

The crucial information identifying this problem is the bugcheck type, INVEXCEPTN, and the failing PC, at some offset within a routine called EXE$PSHARED* such as EXE$PSHARED_MUTEX_UNBLOCK.

The crash is caused by a timing window which accidentally allows use of a recently deleted pshared mutex or condition value. Oracle JDBC is a known consumer of these object types and may experience the crash under certain loads. Other applications which use pshared objects have the same risk of a system crash.

This problem is corrected with this ECO kit.

Some additional sanity checking of pshared object types has also been added to prevent incorrect application calls which operate on the wrong object type.

Images and/or Files Affected

• [SYS$LDR]PROCESS_MANAGEMENT.EXE
• [SYS$LDR]PROCESS_MANAGEMENT.STB
• [SYS$LDR]PROCESS_MANAGEMENT_MON.EXE
• [SYS$LDR]PROCESS_MANAGEMENT_MON.STB

Quix and/or Bugzilla Cases Reporting This Problem

• VSI Bugzilla 1323

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

The OpenVMS system maintains a cache of unused KPBs (kernel process blocks) to avoid extra overhead when these structures are allocated. These structures are used within the executive and various device drivers, and reside in S0/S1 system virtual address space.

Due to a problem in checking the cache availability, some requests for KPBs would allocate new structures when cached unused KPBs were available. This could result in wasted S0/S1 space and wasted CPU overhead to allocate the new structures.

This problem is corrected with this ECO kit.

Images and/or Files Affected

• [SYS$LDR]SYSTEM_PRIMITIVES.EXE
• [SYS$LDR]SYSTEM_PRIMITIVES.STB
• [SYS$LDR]SYSTEM_PRIMITIVES_MIN.EXE
• [SYS$LDR]SYSTEM_PRIMITIVES_MIN.STB

Quix and/or Bugzilla Cases Reporting This Problem

• Internal Bugzilla 1414
• QXCM1001406945
• PTR 75-126-444

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

Under rare circumstances, the system may crash with an INVEXCEPTN bugcheck during RMS global buffer rundown. An incorrect register usage could overwrite a GSD (Global Section Descriptor) address, which would cause an access violation in MMG$FIND_GSTE.

Images and/or Files Affected

• [SYS$LDR]SYS$VM.EXE
• [SYS$LDR]SYS$VM.STB

Quix and/or Bugzilla Cases Reporting This Problem

• QXCM1001477127

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

Global section tracing for problem identification and analysis may be enabled by setting bit 7 in MMG_CTLFLAGS. This tracing should only be used for debugging purposes and is not ever enabled by default.

Should a failure to create a global section occur while the tracing is enabled, the system may crash with a SSRVEXCEPT bugcheck. This was due to an incorrectly initialized field if the failure path was taken.

This problem is corrected with this ECO kit.

Images and/or Files Affected

• [SYS$LDR]SYS$VM.EXE
• [SYS$LDR]SYS$VM.STB

Quix and/or Bugzilla Cases Reporting This Problem

• QXCM1001606217

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

In 2004, support was added for 64-bit value ranges in $CRMPSC. Due to an accidental re-insertion of the 32-bit equivalent instructions, $CRMPSC calls with descending address ranges would not map the address space correctly. This bug has been latent in all OpenVMS versions since that time.

As this problem was found during code inspection for X86 porting activity and never reported by a customer, it appears that descending address range use is essentially non-existent.

This problem is corrected with this ECO kit should any new application code ever use descending address ranges for $CRMPSC.

Images and/or Files Affected

• [SYS$LDR]SYS$VM.EXE
• [SYS$LDR]SYS$VM.STB

Quix and/or Bugzilla Cases Reporting This Problem

None.

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

OpenVMS has a per-process P1 pool ring buffer that can be used for tracking down P1 pool memory leaks. This is enabled when the SYSTEM_CHECK system parameter is set to 1. Packets which are allocated but not deallocated are moved to the top of the ring buffer to aid in analysis. The SYSTEM_CHECK debugging mechanism enables debug, trace, and sanity checks in many portions of the operating system, with a correspondingly higher overhead for normal system operation.

Enhancements have been made to allow a dynamic P1 pool history ring buffer without the extra overhead of SYSTEM_CHECK=1. This is enabled by setting bit 7 in the ACTIVE system parameter POOLCHECK. Processes created after the bit is set will have the pool history buffer mechanism in place.

For example, to enable the P1 pool history buffers, assuming that POOLCHECK is currently zero:

$ MCR SYSGEN
SYSGEN>  USE ACTIVE
SYSGEN>  SET POOLCHECK %X80
SYSGEN>  WRITE ACTIVE
SYSGEN>  EXIT

The ring buffer can be viewed with the SDA command:

SHOW PROCESS/POOL=P1/RING

The ring buffer can be disabled for future processes by clearing bit 7 in the ACTIVE system parameter POOLCHECK.

Images and/or Files Affected

• [SYS$LDR]PROCESS_MANAGEMENT.EXE
• [SYS$LDR]PROCESS_MANAGEMENT.STB
• [SYS$LDR]PROCESS_MANAGEMENT_MON.EXE
• [SYS$LDR]PROCESS_MANAGEMENT_MON.STB
• [SYS$LDR]SYSTEM_PRIMITIVES.EXE
• [SYS$LDR]SYSTEM_PRIMITIVES.STB
• [SYS$LDR]SYSTEM_PRIMITIVES_MIN.EXE
• [SYS$LDR]SYSTEM_PRIMITIVES_MIN.STB

Quix and/or Bugzilla Cases Reporting This Problem

• QXCM1001428470

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

Transient nonpaged pool expansion which nears the point of pool exhaustion could trigger CPUSPINWAIT, CLUEXIT, or other bugchecks while reclaiming memory from the lookaside lists, or because the variable nonpaged pool freelist has become very fragmented.

To alleviate this condition, the upper 1/16th of the nonpaged pool memory area (NPAGEVIR) is now reserved just for the variable nonpaged pool list. This allows quicker recovery from the transient usage spikes without massive lookaside list reclamation activity. This change will only affect systems whose nonpaged pool expands near to the NPAGEVIR maximum value, and thus will have no noticeable effect on most systems. Generally, systems should use AUTOGEN to set a reasonable minimum value for NPAGEDYN for expected pool use, and a larger reserve of potentially used memory for NPAGEVIR. A general rule of thumb is to have at least 20 to 25% extra virtual pool space available.

In addition, new controls have been added to the POOLCHECK system parameter to aid in tracking down nonpaged pool corruption. If SYSTEM_CHECK is 0, and bit 6 is set in POOLCHECK (with some other bits in the low byte clear), then the second byte of POOLCHECK (bits 8-15) may be used to reflect a particular packet size. If this second byte is a value between 1 and 254, then only packets of that byte value multiplied by 64 will be included in the Nonpaged Pool History Ring-Buffer (thus corresponding to packets from 64 to 16256 bytes in size). This will reduce the "noise" of other packet activity in the ring buffer by including only packets of the size being investigated.

For example, if you are interested in packets of size 576 bytes, you could set POOLCHECK to:

Images and/or Files Affected

• [SYS$LDR]SYSTEM_PRIMITIVES.EXE
• [SYS$LDR]SYSTEM_PRIMITIVES.STB
• [SYS$LDR]SYSTEM_PRIMITIVES_MIN.EXE
• [SYS$LDR]SYSTEM_PRIMITIVES_MIN.STB

Quix and/or Bugzilla Cases Reporting This Problem

• QXCM1001380245
• QXCM1001399095
• PTR 75-74-320

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Workaround

Try to address whatever triggered the nonpaged pool expansion and exhaustion so the event does not recur, or increase NPAGEDYN and NPAGEVIR to larger values.

Problem Description

If an Alpha system has a COM1 device present in the hardware, an error message is generated during system boot:

%IOGEN-E-LOADERR, error configuring device SRA0, driver SYS$SRDRIVER.EXE
-RMS-E-FNF, file not found

This message would also appear during a SYSMAN IO AUTO command on such systems.

The sound driver, SYS$SRDRIVER.EXE, does not exist on Alpha. Entries for this driver in the SYS$CONFIG.DAT configuration file are only valid for IA64 systems. The spurious definitions have been removed and these error messages should no longer appear.

Images and/or Files Affected

• [SYSEXE]SYS$CONFIG.DAT

Quix and/or Bugzilla Cases Reporting This Problem

• Internal Bugzilla 862

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

The License Management Utility procedure SYS$UPDATE:VMSLICENSE.COM on Alpha systems defaults the producer field to DEC instead of VSI.

This was a simple oversight when this utility was rebranded for VSI during OpenVMS IA-64 V8.4-1H1 development.

This problem is corrected with this ECO kit.

Images and/or Files Affected

• [SYSMSG]VMSLICENSE_LANGUAGE.COM

Quix and/or Bugzilla Cases Reporting This Problem

• VSI Bugzilla 2583

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

The system may crash with a DELCONPFN bugcheck while deleting the process page table during process rundown.

This problem only occurs on systems with more than 128 GB of memory-resident global sections while deleting the process mapping to such global sections.

This problem is corrected with this ECO kit.

Images and/or Files Affected

• [SYS$LDR]SYS$VM.EXE
• [SYS$LDR]SYS$VM.STB

VSI Case Identifier

• QXCM1001722447

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

The $CRMPSC system service may incorrectly handle an overflow of the page count parameter. This issue is most likely encountered when the relative page argument (RELPAG) is used.

This problem is corrected with this ECO kit.

Images and/or Files Affected

• [SYS$LDR]SYS$VM.EXE
• [SYS$LDR]SYS$VM.STB

VSI Case Identifier

• Jira BO-653

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

The $CRMPSC system service may return an SS$_PAGOWNVIO error when encountering a range of demand zero pages to remap from previous calls. This is an obscure condition which most customers will never encounter.

This problem is corrected with this ECO kit.

Images and/or Files Affected

• [SYS$LDR]SYS$VM.EXE
• [SYS$LDR]SYS$VM.STB

VSI Case Identifier

None, found during internal testing.

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

A synchronization issue when updating a structure that maintains the reference count for a global section could result in the reference count becoming incorrect. If this should happen, the reference count could become either too low or too high.

If the reference count becomes too low, this can lead to SECREFNEG system crashes. If the reference count becomes too high, then a global section may not get deleted when there are no longer any processes referencing the global section.

This patch corrects the synchronization issue when manipulating the global section reference count.

Images and/or Files Affected

• [SYS$LDR]SYS$VM.EXE
• [SYS$LDR]SYS$VM.STB
• [SYS$LDR]SYSTEM_PRIMITIVES.EXE
• [SYS$LDR]SYSTEM_PRIMITIVES.STB
• [SYS$LDR]SYSTEM_PRIMITIVES_MIN.EXE
• [SYS$LDR]SYSTEM_PRIMITIVES_MIN.STB

VSI Case Identifiers

• Jira BO-707
• Netsuite NS2318

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

The SYS V3.0 ECO kit included several synchronization fixes which prevented potential system crashes. However, the behavior of how these changes interacted together had an inadvertent side effect.

An optimization for handling multiple page address ranges during the creation or mapping of global sections was effectively disabled. Applications with a significant reliance on this global section behavior could experience a noticeable performance degradation.

The affected behavior has been adjusted to have the required synchronization but also re-enable the performance optimization. With the change in this kit, the performance of the various $CRMPSC services has been restored to the same level as before installing the SYS V3.0 ECO kit.

Images and/or Files Affected

• [SYS$LDR]SYS$VM.EXE
• [SYS$LDR]SYS$VM.STB

VSI Case Identifiers

• Jira BO-1649
• Netsuite NS6705

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

Problem Description

Attempting to analyze a crash dump file with ANALYZE/CRASH may fail with an SDA-E-UNEXPECTED message similar to:

%SDA-E-UNEXPECTED, unexpected "Unknown" LMB in file #1 at VBN 01400051

This condition occurs if a memory section written to a dump file was incompletely saved, for example if the space reserved for saving the crash dump file was too small or if writing the dump file was interrupted.

Rather than exit from ANALYZE/CRASH, SDA will now issue the message to describe that some of the dump file cannot be interpreted, but then continue execution to allow analysis of already successfully read portions of the dump file.

Images and/or Files Affected

• [SYSLIB]SDA$SHARE.EXE

VSI Case Identifiers

• Jira KT-194
• Netsuite NS5745

Release Version of VSI OpenVMS That Will Contain This Change

The next VSI OpenVMS Alpha release after V8.4-2L2.

This kit is provided for download within a ZIP archive container file.

The kit files may be extracted on any system with UNZIP and copied to your OpenVMS system, or extracted on your OpenVMS system directly.

Assuming you have created an UNZIP symbol to invoke the UNZIP image, you can invoke UNZIP to unpack the kit on OpenVMS using the command:

$ UNZIP VMS842L2A_SYS-V0500

This will extract the installable PCSI product kit file and its associated signed manifest (_VNC file), used for kit validation during PRODUCT commands.

VSI strongly recommends always using the manifest to validate the kit content during any PRODUCT commands. This will occur automatically if the files are both contained in the same directory.

Install this kit with the POLYCENTER Software Installation Utility by logging into the SYSTEM account and typing the following command at the DCL prompt:

$ PRODUCT INSTALL VMS842L2A_SYS [/SOURCE=location_of_kit]

The kit location may be a tape drive, CD/DVD, or a disk directory that contains the kit. The /SOURCE qualifier is not needed if the PRODUCT INSTALL command is executed from the same directory as the kit location.

This kit requires the use of /RECOVERY_MODE and /SAVE_RECOVERY_DATA and will automatically set them; they do not need to be present on the command line.

The release notes for any kit may be extracted prior to kit installation using the PRODUCT EXTRACT RELEASE_NOTES command.

User-selectable options for installation behavior and scripting are available in this kit, refer to Appendix A, "User-Selectable Control Options and Scripting Considerations" for further details.

Additional help on installing PCSI kits can be found by typing HELP PRODUCT INSTALL at the system prompt.