======================================================================= HP OpenVMS Update kit for HP Integrity Servers and VMS84I_UPDATE-V0400 Release Notes ======================================================================= 1 KIT NAME: VMS84I_UPDATE-V0400 2 KIT DESCRIPTION: 2.1 Installation Rating: INSTALL_1 : To be installed by all customers for HP Integrity server blades and HP Integrity server systems. HP's support service is available only if INSTALL_1 is applied. 2.2 Reboot Requirement: Reboot Required. HP strongly recommends that a reboot is performed immediately after kit installation to avoid system instability. If you have other nodes in your OpenVMS cluster, they must also be rebooted in order to make use of the new image(s). If it is not possible or convenient to reboot the entire cluster at this time, a rolling reboot may be performed. 2.3 Version(s) of OpenVMS to which this kit may be applied: OpenVMS for Integrity Servers V8.4 2.4 New functionality or new hardware support provided: This update kit supports the 8 GB memory dual in-line memory module (DIMMs) on the Intel Itanium 9300 processor series based HP Integrity server blades. It includes BL860c i2, BL870c i2, and BL890c i2 server blades. These blades are supported on the HP c-class c3000 and c7000 enclosures. 3 KITS SUPERSEDED BY THIS KIT: - VMS84I_UPDATE-V0300 4 KIT DEPENDENCIES: 4.1 The following remedial kit(s), or later, must be installed BEFORE installation of this, or any required kit: - VMS84I_PCSI-V0100 4.2 In order to receive all the corrections listed in this kit, the following remedial kits, or later, should also be installed: - None 5 NEW FUNCTIONALITY AND/OR PROBLEMS ADDRESSED IN THE VMS84I_UPDATE-V0400 KIT 5.1 New functionality addressed in this kit 5.1.1 Adding /GPT qualifier to the BACKUP command The BACKUP command propagates the GPT.SYS file from the source disk to the destination disk during an image copy or restore operation. Earlier, there was no mechanism to control the propagation of GPT.SYS from the source disk to the destination disk. The BACKUP command is now modified and a new qualifier, /GPT is added to control the propagation of the GPT.SYS file from the source disk to the destination disk. The options are as follows: - The /GPT qualifier propagates the GPT.SYS file from the source to the destination. - The /GPT=NONE qualifier does not propagate the GPT.SYS file from the source to destination. - By default, the /GPT qualifier is enabled if the GPT.SYS file is located on the source disk. The /GPT and /GPT=NONE qualifiers controls the propagation of the GPT.SYS file from the source disk to destination disk if the GPT.SYS file exists in source disk. Note: If there is no GPT.SYS file on the source disk, the /GPT qualifier does not create a new GPT.SYS file in the destination disk. The /GPT and /GPT=NONE qualifiers can be used with the image restore or image copy operation only. Any attempt to use the /GPT and /GPT=NONE qualifiers with the save and non-image restore or non-image copy operation is ignored. 5.1.2 Increase in the maximum limit of Multi Block Count to 255 A Record Management Services (RMS) application specifies the number of blocks that it requires RMS to transfer with each I/O operation, via the Multi Block Count (MBC) field of the RAB structure, RAB$B_MBC. For more information on RAB$B_MBC, see Section 7.11, in the "OpenVMS Record Management Services Reference Manual." The maximum value of the MBC field is 127 blocks. The limit is now increased to 255 blocks. For applications that read or write sequential files, the increase in limit enhances the I/O performance by transferring more blocks in a single I/O operation. This change is applicable on all versions of OpenVMS V8.3 or later. Restriction ----------- If you intend to use a Multi Block Count (RAB$B_MBC) value of more than 127 blocks to open a file in shared mode in a VMS cluster, you must apply the latest RMS or SYS patches on all the nodes in the VMS cluster where the files will be shared. The following is a list of the latest RMS or SYS patches: 0 VMS83A_SYS-V1800 0 VMS83I_SYS-V1500 0 VMS831H1I_SYS-V1200 0 VMS84I_UPDATE-V0400 0 VMS84A_UPDATE-V0400 For OpenVMS versions before V8.3, MBC value greater than 127 is not supported. In mixed version clusters that include applications with OpenVMS versions before V8.3, do not open a file that has an MBC value greater than 127. Apply the following patches on OpenVMS versions before V8.3: 0 VMS732_RMS-V0600 0 VMS82A_RMS-V0200 0 VMS821I_RMS-V0500 5.1.3 Faster memory zeroing using IA64_STF_SPILL_F0 iMACRO The IA64_STF_SPILL_F0 is a new iMACRO built-in that allows faster memory zeroing by writing 16 bytes of memory at a time. In the earlier versions, only 8 bytes could be zeroed using the EVAX_STQ built-in. The IA64_STF_SPILL_F0 built-in provides significant performance benefits, especially when zeroing large areas of memory. Format: IA64_STF_SPILL_F0 addr, incr where, 'addr' is the address of the 16-byte memory region 'incr' is the value by which 'addr' is incremented; specifying 0 means no auto-increment. 5.1.4 Adding new option to PTHREAD_CONFIG to control the stack size of internal threads A new configuration parameter is added to PTHREAD_CONFIG to control the size of the stack for the exit-handling thread. For example, Define the following logical: $ DEFINE PTHREAD_CONFIG "stksize=exit=200000" This ensures the exit-handling thread's stack size to be at least to 200000 bytes. Images Affected: - [SYSLIB]PTHREAD$DBGSHR.EXE - [SYSLIB]PTHREAD$RTL.EXE 5.2 Problems addressed in this kit 5.2.1 Lookaside list corruption 5.2.1.1 Problem Description: Lookaside lists are maintained by the Distributed Computing Environment (DCE) for memory management. The variable that reflects the number of nodes in the list were corrupted. To prevent this, the value of the variable is populated by traversing the list and getting the actual count. Images Affected: - [SYSLIB]DCE$LIB_SHR.EXE Quix cases reporting this problem: QXCM1000207283, QXCM1000233849, QXCM1000310461 5.2.1.3 Work-arounds: None. 5.2.2 Wrong audit logs displayed when DECServer boots on LAN 5.2.2.1 Problem Description: When a DECServer is booted on a LAN from an OpenVMS system having SYS$ACM enabled loginout.exe (ACME login), the following incorrect audit message is displayed: - - Auditable event: Network login failure - Process name : MOM_EWA-1_1 Username : DECNET Image name : DKA100:[SYS0.SYSCOMMON.][SYSEXE]LOGINOUT.EXE Status : %ACME-E-TIMEOUT, requested operation has timed out Images Affected: - [SYSLIB]VMS$VMS_ACMESHR.EXE Quix cases reporting this problem: QXCM1000977223 5.2.2.2 Work-arounds: None. 5.2.3 Alignment faults in the audit server 5.2.3.1 Problem Description: The audit server generates alignment faults for certain operations on the Alpha and Integrity servers. This problem has been fixed. Images Affected: - [SYSEXE]AUDIT_SERVER.EXE - [SYSLIB]VMS$FORMAT_AUDIT_SYSTEM.EXE Quix cases reporting this problem: QXCM1000952940, QXCM1001008779 5.2.3.2 Work-arounds: None. 5.2.4 Alignment faults in SECURITY.EXE 5.2.4.1 Problem Description: SECURITY.EXE generated alignment faults during internal testing. This problem has been fixed. Images Affected: - [SYS$LDR]SECURITY.EXE - [SYS$LDR]SECURITY_MON.EXE Quix cases reporting this problem: None. 5.2.4.2 Work-arounds: None. 5.2.5 Adding the NFS or DFS device as a boot option using the boot_options.com command procedure might lead to system crash 5.2.5.1 Problem Description: On an Integrity server, OpenVMS might crash occasionally with unexpected system service exception, if a Network File System (NFS) or Distributed File System (DFS) device is added as a boot option using boot_options.com. OpenVMS now displays an error message and exits when a NFS or DFS device is added as a boot option. Images Affected: - EFI$BCFG.EXE Quix cases reporting this problem: None. 5.2.5.2 Work-arounds: None. 5.2.6 An OpenVMS boot, after a failed InfoServer boot with an invalid InfoServer service name, displays error messages and configures only one CPU 5.2.6.1 Problem Description: OpenVMS boot with an invalid InfoServer service name as a parameter fails and control returns to the EFI shell. The subsequent boot displays error messages and configures only one CPU. This problem has been fixed. Images Affected: - VMS_LOADER.EFI Quix cases reporting this problem: None. 5.2.6.2 Work-arounds: None. 5.2.7 Accessing Digiport devices (TXD devices in OpenVMS) during boot might result in system crash 5.2.7.1 Problem Description: The system might crash when automated startup procedures that run during system startup attempt to access Digiport devices. This problem has been fixed. Images Affected: - SYS$DZCDRIVER.EXE Quix cases reporting this problem: QXCM1001041405 5.2.7.2 Work-arounds: None. 5.2.8 DNDRIVER improvements 5.2.8.1 Problem Description: The DNDRIVER has been enhanced to address the following issues: - The DNA device appears to be configured but fails with a fatal error or hangs when first mounted. - The initial mount fails after hot-plug (the second mount succeeds). - The DNA device appears in an online state even after devices are unplugged. - Occasional ACCVIO when a DNA device is removed while IO is active. The "USB SHOW DNA n" command added to SDA to aid debugging USB disk problems. Images Affected: - SYS$DNDRIVER.EXE - USB$SDA.EXE Quix cases reporting this problem: None. 5.2.8.2 Work-arounds: None. 5.2.9 Optimizations to SYSMAN IO AUTOCONFIGURE 5.2.9.1 Problem Description: Optimizations to SYSMAN IO AUTOCONFIGURE are made to speed up fiber channel device discovery. Images Affected: - IOGEN$FIBRE_CONFIG.EXE Quix cases reporting this problem: QXCM1001023718, QXCM1001039079 5.2.9.2 Work-arounds: None. 5.2.10 If the variable pool gets heavily fragmented, there are chances of incurring a CPUSPINWAIT bugcheck for the POOL spinlock 5.2.10.1 Problem Description: When lookaside lists are flushed to satisfy an allocation request, after a certain number of packets have been deallocated, it is checked if the POOL spinlock must be released. As the variable pool gets heavily fragmented, each such de-allocation of a packet takes long time thereby delaying the check for POOL spinlock. This leads to the CPU waiting for the POOL spinlock, to crash with the CPUSPINWAIT bugcheck. The count or the number of packets after which the check is done, is now reduced, thereby reducing the probability of a premature CPUSPINWAIT crash before the release of POOL spinlock is checked. Images Affected: - [SYS$LDR]SYSTEM_PRIMITIVES.EXE - [SYS$LDR]SYSTEM_PRIMITIVES.STB - [SYS$LDR]SYSTEM_PRIMITIVES_MIN.EXE - [SYS$LDR]SYSTEM_PRIMITIVES_MIN.STB Quix cases reporting this problem: QXCM1001025466 5.2.10.2 Work-arounds: None. 5.2.11 /RECORD_IONLY qualifier added with APPEND/NEW_VERSION 5.2.11.1 Problem Description: When COPY is used to concatenate multiple files, the output file might contain garbage data. This problem mostly occurs when the first source file organization is sequential, record format is fixed, and the last record size is less than Max Record Size (MRS). Images Affected: - [SYSEXE]COPY.EXE - [SYSUPD]COPY.CLD (update to DCLTABLES.EXE) Quix cases reporting this problem: QXCM1000952794 5.2.11.2 Work-arounds: None. 5.2.12 INSTALL LIST/GLOBAL shows negative value for section reference count 5.2.12.1 Problem Description: INSTALL LIST/GLOBAL displays a negative value for the section reference count when creating or mapping shared-memory global section with less region size. Images Affected: - SYS$VM.EXE - SYS$VM.STB Quix cases reporting this problem: None. 5.2.12.2 Work-arounds: None. 5.2.13 SDA$SET_PROCESS() issue 5.2.13.1 Problem Description: The system crashes when the target process that SDA$SET_PROCESS() tries to set is deleted while accessing the process memory of that target process. SDA is now modified to display the "BADPROC" error when such a scenario occurs. Images Affected: - [SYSLIB]SDA$SHARE.EXE - [SYS$LDR]SYS$BASE_IMAGE.EXE Quix cases reporting this problem: QXCM1001040338 5.2.13.2 Work-arounds: None. 5.2.14 SDA> CLUE CONFIG command displays incorrect "CPU Type" for Intel Itanium 9300 processor 5.2.14.1 Problem Description: The SDA> CLUE CONFIG command incorrectly displays ‘Quad-core’ instead of ‘Dual-core’ for the Dual Core Intel Itanium 9300 processors in the "CPU Type" field. Output from the earlier version: O/P on a Dual-Core Itanium 9300 chip, CPU Type Quad-Core Itanium (Intel Itanium 9300 Rev E0) O/P on a HPVM on Itanium 9300 chip, CPU Type Quad-Core Itanium (Intel Itanium 9300 Rev E0) The following output is displayed with this fix: O/P on a Dual-Core Itanium 9300 chip, CPU Type Dual-Core Itanium (Intel Itanium 9300 Rev E0) O/P on a HPVM on Itanium 9300 chip, CPU Type Intel Itanium 9300 Rev E0 Images Affected: - [SYSLIB]CLUE$SDA.EXE Quix cases reporting this problem: None. 5.2.14.2 Work-arounds: None. 5.2.15 Activating shareable images when installed with the /SHARED qualifier 5.2.15.1 Problem Description: Shareable images installed with the /SHARED qualifier that has only execute permission for World group results in NOPRIV error when the user belonging to World group tries to execute the shareable images. Images Affected: - [SYS$LDR]IMAGE_MANAGEMENT.EXE Quix cases reporting this problem: QXCM1001023910 5.2.15.2 Work-arounds: None. 5.2.16 Unable to install images with the /RESIDENT qualifier after increasing the GH regions 5.2.16.1 Problem Description: The Install utility displays an error message while installing images with the /RESIDENT qualifier from a user account, which does not have PFNMAP privilege. Images Affected: - [SYSEXE]INSTALL.EXE Quix cases reporting this problem: QXCM1000988667 5.2.16.2 Work-arounds: None. 5.2.17 Mounting and dismounting a particular volume at the same time can result in a system crash 5.2.17.1 Problem Description: When a volume is dismounted, there exists a timing window during which, an attempt to mount the same volume results in a system crash. This problem has been fixed. Images Affected: - F11BXQP.EXE - F11BXQP.STB Quix cases reporting this problem: QXCM1000956183 5.2.17.2 Work-arounds: None. 5.2.18 Issue with lbr$get_index affecting lbr$insert_key 5.2.18.1 Problem Description: When adding modules to a text library using calls to lbr$insert_key, the application might hang under the following conditions: - lbr$get_index is called on a library, which contains more than 21 modules. - Some of the modules have keys consisting of more than 16 characters. Under these conditions, if you call lbr$insert_key on another library, the application hangs. This problem has been fixed. Images Affected: - [SYSLIB]LBRSHR.EXE Quix cases reporting this problem: QXCM1000975651 5.2.18.2 Work-arounds: None. 5.2.19 SHOW ERROR/FULL exits with %SYSTEM-F-VASFULL 5.2.19.1 Problem Description: SHOW ERROR/FULL exits with the "%SYSTEM-F-VASFULL, virtual address space is full" error when the system has more than 700 devices. This problem has been fixed. Images Affected: - [SYSEXE]SHOW.EXE Quix cases reporting this problem: QXCM1001003793 5.2.19.2 Work-arounds: None. 5.2.20 Fix for stack pointer corruption when global buffer is used in a file 5.2.20.1 Problem Description: If the RMS global buffer is used in an RMS file, the system might bug check with INVEXCEPTN when a process using this file is abruptly stopped and another process uses that global buffer. This problem has been fixed. Images Affected: - RMS.EXE Quix cases reporting this problem: None. 5.2.20.2 Work-arounds: None. 5.2.21 Problem with the DIR/LINK command 5.2.21.1 Problem Description: If the file name is large, the DIR/LINK command does not display any space between linkcount value and the next file name. Images Affected: - DIRECTORY.EXE Quix cases reporting this problem: None. 5.2.21.2 Work-arounds: None. 5.2.22 XQP does not check if a file is created at LBN 0 5.2.22.1 Problem Description: Under certain circumstances, such as data structure corruption, XQP can create a file at LBN 0, which is reserved for INDEXF.SYS file. XQP does not perform a consistency check to verify if an LBN 0 is used to create files. This problem has been fixed. Images Affected: - F11BXQP.EXE - F11BXQP.STB Quix cases reporting this problem: None. 5.2.22.2 Work-arounds: None. 5.2.23 XFC uses a 32-bit variable to store 64-bit PFN's 5.2.23.1 Problem Description: XFC uses unsigned 32-bit integers to store PFN's. This can cause problems if the PFN's are more than 32-bits. This problem has been fixed. Images Affected: - SYS$XFCACHE.EXE Quix cases reporting this problem: None. 5.2.23.2 Work-arounds: None. 5.2.24 VMSINSTAL is unable to skip kit validation 5.2.24.1 Problem Description: While installing the software kits that are not signed by HPCSS, VMSINSTAL displays the following informational message and prompts: %VMSINSTAL-I-VALSIGN, Performing product kit validation of signed kits %VMSINSTAL-I-NOVALDONE, Product is not signed by HP Do you want to install this product [NO]? Validation can now be overridden by defining a job wide logical "VMI$VALIDATE_KIT" to "NO" $DEFINE/JOB VMI$VALIDATE_KIT NO When the validation is ignored, VALSIGN status logs as 'U'(Unsigned) in the VMSINSTAL history. Images Affected: - VMSINSTAL.COM Quix cases reporting this problem: None. 5.2.24.2 Work-arounds: None. 5.2.25 VMSINSTAL incorrectly handles CTRL/Z 5.2.25.1 Problem Description: While installing software kits that are not signed by HPCSS, VMSINSTAL prompts for user input during validation. For example: %VMSINSTAL-I-VALSIGN, Performing product kit validation of signed kits %VMSINSTAL-I-NOVALDONE, Product is not signed by HP Do you want to install this product [NO]? Press Ctrl/Z to continue kit installation instead of prompting for user input. Images Affected: - VMSINSTAL.COM - VMSINSTAL_LANGUAGE.COM Quix cases reporting this problem: None. 5.2.25.2 Work-arounds: None. 5.2.26 RMS incorrectly considers the non-directory files as directory on ODS-5 disk 5.2.26.1 Problem Description: When a non-directory file with the .DIR extension exists in a directory and if a recursive DIRECTORY command is issued on the directory, RMS incorrectly considers the non-directory file as a directory, thereby resulting in an RMS-E-ACC error. For example: $ CREATE TESTDIR:[FOO]TEST.DIR $ DIRECTORY TESTDIR:[FOO] Directory TESTDIR:[FOO] TEST.DIR;1 Total of 1 file. $ DIRECTORY TESTDIR:[FOO...] Directory TESTDIR:[FOO] %DIRECT-E-OPENIN, error opening TESTDIR:[FOO.TEST]*.*;* as input -RMS-E-ACC, ACP file access failed -SYSTEM-F-FILNOTACC, file not accessed on channel TEST.DIR;1 This problem has been fixed. Images Affected: - [SYS$LDR]RMS.EXE Quix cases reporting this problem: None. 5.2.26.2 Work-arounds: None. 5.2.27 BACKUP-E-VBNMISSING Error message 5.2.27.1 Problem Description: BACKUP reports the following error messages during the physical copy and restore operations, where the destination disk is larger in capacity than the source disk. BACKUP reports the VBNMISSING error even though the BACKUP operation is successful. %BACKUP-W-PHYDDEVLG, device $100$dkc0: larger than input %BACKUP-E-VBNMISSING, $100$DKC0: has missing blocks 35565008 through 286749487 This problem has been fixed. BACKUP is now modified to report informational messages (shown below) during physical restore or copy operation where the destination disk is of larger capacity than the source disk. %BACKUP-I-PHYDEVLG, device $9$LDA3: larger than input %BACKUP-I-UNUSEDBLKS, $9$LDA3: has uninitialized blocks 40000 through 49999, this may result in the uninitialized blocks being unusable Images Affected: - BACKUPSHR.EXE - SYSMGTMSG.EXE Quix cases reporting this problem: QXCM1001044942 5.2.27.2 Work-arounds: None. 5.2.28 The ICC$DISCONNECT call from the USER mode code can result in an INVEXCEPTN system crash in IPC_SERVICES 5.2.28.1 Problem Description: When Inter Process Communication (IPC) is in the process of replying to a message and also receives a disconnect request from the other end simultaneously, the LCB gets de-allocated by the disconnect request and tries to access the stale LCB to serve the reply resulting in the INVEXCEPTN system crash. Images Affected: - [SYS$LDR]SYS$IPC_SERVICES.EXE - [SYS$LDR]SYS$IPC_SERVICES.STB Quix cases reporting this problem: QXCM1000997287 5.2.28.2 Work-arounds: None. 5.2.29 When using the SHOW CLUSTER/OUTPUT qualifier, the file name gets truncated to 63 characters 5.2.29.1 Problem Description: In the [CLIUTL] modules, the size of the output buffer for SHOW CLUSTER/OUTPUT is 63, which means that the complete file name including the path should not exceed 63 characters in length. Example from the previous version: $ SHOW CLUSTER/OUTPUT=SYS$SYSDEVICE:[LOGIN.TOPSBRT.SYSTEM_REPORT.TMP] SYSTEM_REPORT.TMP_00000028 returns SYSTEM_REPORT.TM;1 in the directory Structure. The fixed image displays the following output for the SHOW CLUSTER command: Example: $ SHOW CLUSTER/OUTPUT=SYS$SYSROOT:[SYSMGR.SHUBH.CLUSTER_PTR_FOR_ANALYZSIS] THIS_IS_THE_SHOW_CLUSTER_QUIX_PROBLEM_INVESTIGATION.LIS $ dir Directory SYS$SYSROOT:[SYSMGR.SHUBH.CLUSTER_PTR_FOR_ANALYZSIS] THIS_IS_THE_SHOW_CLUSTER_QUIX_PROBLEM_INVESTIGATION.LIS;1 Images Affected: - [SYSEXE]SHWCLSTR.EXE Quix cases reporting this problem: QXCM1001030551 5.2.29.2 Work-arounds: None. 5.2.30 Alignment faults in cluster components 5.2.30.1 Problem Description: Alignment faults are seen in the cluster components such as PEDriver, Cluster Wide Process Services (CWPS), Connection Manager, and Lock Manager. This problem has been fixed. Images Affected: - [SYS$LDR]SYS$PEDRIVER.EXE - [SYS$LDR]SYS$PEDRIVER.STB - [SYS$LDR]SYS$PEDRIVER_MON.EXE - [SYS$LDR]SYS$PEDRIVER_MON.STB - [SYS$LDR]PROCESS_MANAGEMENT.EXE - [SYS$LDR]PROCESS_MANAGEMENT.STB - [SYS$LDR]PROCESS_MANAGEMENT_MON.EXE - [SYS$LDR]PROCESS_MANAGEMENT_MON.STB - [SYS$LDR]LOCKING.EXE - [SYS$LDR]LOCKING.STB - [SYS$LDR]SYS$CLUSTER.EXE - [SYS$LDR]SYS$CLUSTER.STB - [SYS$LDR]SYS$CLUSTER_MON.EXE - [SYS$LDR]SYS$CLUSTER_MON.STB Quix cases reporting this problem: None. 5.2.30.2 Work-arounds: None. 5.2.31 Alignment faults in shadowing commands 5.2.31.1 Problem Description: Alignment faults are seen in the shadowing commands. This problem has been fixed. Images Affected: - [SYS$LDR]IO_ROUTINES.EXE - [SYS$LDR]IO_ROUTINES.STB - [SYS$LDR]IO_ROUTINES_MON.EXE - [SYS$LDR]IO_ROUTINES_MON.STB Quix cases reporting this problem: None. 5.2.31.2 Work-arounds: None. 5.2.32 The avoid ICC application hangs in a cluster while shutting down a multithreaded process 5.2.32.1 Problem Description: When a multithreaded process that uses Intra Cluster Communication/Inter Process Communication (ICC/IPC) tries to shutdown, the process waits in the LEF state forever. As part of IPC$UNLINK, it queues an Asynchronous System Traps (AST) by calling SCH$QAST, which returns the SS$_NONEXPR (no process exist) error if the process on the other node is already terminated. This problem is fixed by handling the incorrect status. Images Affected: - [SYS$LDR]SYS$IPC_SERVICES.EXE Quix cases reporting this problem: QXCM1001028956 5.2.32.2 Work-arounds: None. 5.2.33 The duplicate "Shadow Server" process causes SHADDETINCON bugcheck 5.2.33.1 Problem Description: Executing the SYS$SYSTEM:STARTUP SHADOW command starts a duplicate shadow server process, which causes SHADDETINCON bugcheck during shutdown. Images Affected: - [SYS$STARTUP]VMS$CONFIG-050_SHADOW_SERVER.COM Quix cases reporting this problem: QXCM1001017963 5.2.33.2 Work-arounds: None. 5.2.34 SHADOW_SERVER exception handler issues 5.2.34.1 Problem Description: The SHADOW_SERVER exception handler issues bugcheck in response to any signal delivered to the process. Images Affected: - [SYSEXE]SHADOW_SERVER.EXE Quix cases reporting this problem: QXCM1001057993 5.2.34.2 Work-arounds: None. 5.2.35 The SHOW ERROR command problem 5.2.35.1 Problem Description: When a shadowset member is removed from a shadowset because of an error, the UCB device error count is not incremented. As a result, even if the error is logged, the SHOW ERROR command fails to recognize the error and the error log might not be reviewed. Images Affected: - [SYS$LDR]SYS$SHDRIVER.EXE Quix cases reporting this problem: QXCM1000974273 5.2.35.2 Work-arounds: None. 5.2.36 The SET SHADOW/SITE problem 5.2.36.1 Problem Description: SET SHADOW/SITE accepts any value, except 0 as the site ID. This problem is fixed to accept any value between 0 and 255. 0 indicates that there is no SITE value. Images Affected: - [SYSEXE]SETSHOSHADOW.EXE Quix cases reporting this problem: None. 5.2.36.2 Work-arounds: None. 5.2.37 Discrepancies in the NPAGERAD and NPAGEDYN values on Integrity servers and Integrity blades with RAD support 5.2.37.1 Problem Description: When support for Resource Affinity Domain (RAD) and per-RAD non-paged pool are enabled on Integrity cell-based systems and BL8x0c i2 server blades, the distribution of non-paged pool over the base RAD and the remaining RADs does not comply with the documentation mentioned under NPAGERAD in the "HP OpenVMS System Management Utilities Reference Manual: M-Z" for V8.4. Note: The value of 10944731 in the RAD_SUPPORT parameter enables support for RAD and per-RAD non-paged pool. As mentioned in the guide: (a) By default, when NPAGERAD is 0, the system must compute NPAGERAD using the following formula: NPAGERAD= NPAGEDYN * (1- Base RAD memory/ Total memory) Currently, the system does not calculate the value required for NPAGERAD. The NPAGERAD value remains as 0 and per-RAD non-paged pool is not enabled. (b) When NPAGERAD is non-zero, the system must distribute the non-paged pool over the base RAD and the remaining RADs such that NPAGEDYN corresponds to the total amount of non-paged pool in the system. The NPAGERAD corresponds to the amount of non-paged pool from the RADs other than the base RAD. For example, if the system is configured with: NPAGEDYN = 6291456 bytes and NPAGERAD = 2097152 bytes, OpenVMS must allocate a total of approximately 6,291,456 bytes of non-paged pool. Of this amount, the system must divide 2,097,152 bytes among the RADs that are not the base RAD. However, OpenVMS divides 2,097,152 bytes among the RADs that are not the base RAD and assigns 6,291,456 bytes to the base RAD. Hence, in this scenario the system has a total of 8,388,608 (NPAGEDYN+NPAGERAD) bytes. Images Affected: - [SYSEXE]SYSBOOT.EXE Quix cases reporting this problem: None. 5.2.37.2 Work-arounds: None. 5.2.38 Improved performance in handling RECNOTFOU exceptions 5.2.38.1 Problem Description: The logical definition method has been implemented because modifying the application to call the DBASIC$IO_NO_SIGNAL routine before the GET or FIND statement requires considerable programming effort. This allows a user program to avoid an expected I/O exception, RMS$_RNF by defining the DBASIC$IO_NOSIGNAL logical to 1. Defining the DBASIC$IO_NOSIGNAL logical in the process table to 1 will amount to not signal any errors for any RECNOTFOU error in a user program. Explicit call to the DBASIC$IO_NO_SIGNAL routine is not required with this implementation. However, if you want to conditionally handle the RMS$_RNF condition for certain channels (as exception handling will be disabled for RMS$_RNF), you need to modify the application to make a call to RMSSTATUS for the required channel. Images Affected: - [SYSLIB]DEC$BASRTL.EXE Quix cases reporting this problem: QXCM1000982041 5.2.38.2 Work-arounds: None. 5.2.39 Shutdown hangs if it is either affinitized to non-primary CPU or if primary CPU is hard offloaded 5.2.39.1 Problem Description: Shutdown stops all the secondary CPUs. Hence, if the primary CPU is hard offloaded, or if the process that shuts down the system is affinitized to a non-primary CPU, the process will be in the RWCAP state and the shutdown procedure hangs. Images Affected: - SHUTDOWN.COM Quix cases reporting this problem: None. 5.2.39.2 Work-arounds: None. 5.2.40 Device added during run time are discovered but not reported with the /LOG qualifier in the OpenVMS guest environment 5.2.40.1 Problem Description: Earlier, on Integrity VM, $MC SYSMAN IO AUTOCONFIGURE /LOG discovered the newly added device at run time but does not report them. This is because the routine to log such information is not called after the driver is loaded successfully. Images Affected: - [SYSLIB]IOGEN$AVIO_CONFIG.EXE - [SYSLIB]IOGEN$SCSI_CONFIG.EXE Quix cases reporting this problem: None. 5.2.40.2 Work-arounds: None. 5.2.41 The LNM tracing code issue 5.2.41.1 Problem Description: Logical name tracing does not work whenever the logical name string address belongs to 64-bit address space. This is because the code assumed it to be a 32-bit address and was accessing an incorrect address. The LNM tracing code is now modified to handle logical name addresses in 64-bit address space. Images Affected: - [SYS$LDR]LNM$DEBUG.EXE - [SYS$LDR]LOGICAL_NAMES.EXE - [SYSLIB]LNM$SDA.EXE Quix cases reporting this problem: QXCM1001016858 5.2.41.2 Work-arounds: None. 5.2.42 AUTOGEN behavior on a 2 TiB target disk 5.2.42.1 Problem Description: While installing OpenVMS V8.4 on a target disk (system disk) of capacity 2 TiB, the system automatically runs AUTOGEN to evaluate your hardware configuration and estimate typical workloads. AUTOGEN tries to set system parameters, the size of the page, swap, and dump files, during which the system might hang because AUTOGEN cannot handle disks of large sizes. Images Affected: - [SYSUPD]AUTOGEN.COM Quix cases reporting this problem: None. 5.2.42.2 Work-arounds: None. 5.2.43 BL890c i2 fails to boot in Mostly NUMA configuration with maximum memory 5.2.43.1 Problem Description: On BL890c i2 server blade with all the 96 memory slots populated with 8 GB memory dual in-line memory module (DIMMs) (96 * 8GB = 768 GB) and with the system memory interleaved configuration being Mostly NUMA, the system fails to boot. OpenVMS currently supports a maximum of 736 GB in this configuration. In this scenario, HP recommends you to change the system memory interleaved configuration to any of the following four options available for OpenVMS to successfully boot the server: - MaxNUMA - Balanced - MostlyUMA - MaxUMA This can be done at the EFI shell by invoking the "memconfig -mi