OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
From: AIX Service Mail Server (aixservaustin.ibm.com)
Date: Tue Jul 10 2001 - 02:22:31 CDT

  • Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

    APAR: IY15631 COMPID: 5765C3403 REL: 430
    ABSTRACT: ADD VMTUNE OPTION FOR THE NUMBER OF FRAMESET PER MEMORY POOL

    PROBLEM DESCRIPTION:
    We need to add a vmtune option that will allow the
    customer to set the number of framsets per mempool.
    By setting it to one, it should fix the slowdowns
    on 604e-based SP systems.

    PROBLEM SUMMARY:
    Added vmtune -v option to specify the
    number of framesets per mempool.

    PROBLEM CONCLUSION:
    Added vmtune -v option to specify the
    number of framesets per mempool.

    ------

    APAR: IY16955 COMPID: 5765D5101 REL: 120
    ABSTRACT: RECEIVED A GROUP_PROCLAIM MESSAGE WITH A DIFFERENT INSTANCE

    PROBLEM DESCRIPTION:
    The customer had a condition where the instance number in the
    odm class HACMPtopsvcs was out of sync in the cluster. This
    prevented one of the nodes from joining the topsvcs group. The
    message was only found in the log. This error needs to be
    written to the errpt and given an action to fix it.

    PROBLEM SUMMARY:
    All the nodes in an HACMP cluster or in PSSP
    need to have the same version of the configuration.
    Nodes with different versions of the configuration
    will be unable to participate in the same adapter
    membership groups in Topology Services. In HACMP,
    if a node is down during a Topology DARE (that is,
    when the cluster topology is synchronized) then this
    node will have a different version of the configuration
    than the rest. If this node is later powered on, it
    will be unable to join the same adapter groups as
    the others. As a result, the nodes will tend to
    form a "virtual partition", with the nodes that
    were up during the Topology DARE in one side and
    the nodes that were down in another.
    The "virtual partitions" described above need to
    be avoided, since it will result in each side of
    the partition trying to acquire the same resources
    in the cluster. To prevent these kind of partitions from
    happenning, Topology Services will exit in the HACMP
    environment when the cluster starts-up in a node and then
    a peer message is immediately received that indicates that
    the sender has a different version of the configuration.
    This may indicate either of two things:
      - the cluster was started on a node during a Topology
        DARE, after the new configuration was propagated
        to the node but before all the others have been
        refreshed
      - the cluster was started on a node after nodes that
        were down during a Topology DARE are brought up
    With the current code, when Topology Services exits it
    does not clearly indicate the reason. An error log entry
    needs to be created when such a situation arises.
    The behavior of exiting when a different configuration
    instance number is detected on startup only occurs in
    an HACMP/ES environment, not in PSSP.

    PROBLEM CONCLUSION:
    When Topology Services exits due to configuration
    instance number mismatch, it will now create an AIX
    error log explaining why the daemon exited. The
    error log entry will also show the originator of
    the message with a different instance number.
    This should allow the user to determine the node
    that did not receive the latest instance of the
    configuration.

    ------

    APAR: IY17477 COMPID: 5765C3403 REL: 430
    ABSTRACT: FIND COMMAND WITH -NOGROUP FLAG LISTS FILES IN DCE GROUP

    PROBLEM DESCRIPTION:
    The 'find' command for the '-nouser' option is able to
    use the entry in /etc/security/user file to refer to the
    DCE and determine whether it is not a valid DCE user.
    However for the "-nogroup" option of the 'find' command,
    it simply checks whether the file belongs to a group or not
    in the /etc/group database. Apparently it is not DCE aware -
    i.e. for groups, the find cmd does not check the file
    /etc/security/user to determine whether it should refer
    to any other source (e.g. DCE) for information.

    PROBLEM SUMMARY:
    "find dir -nogroup" should list all the files in
    directory dir that DONOT belong to any of the groups.
    But on a dce environment, this command gives a long list of
    files that are present in a dce group. There is a difference
    in the behaviour of -nouser and -nogroup options for the find
    command on a DCE machine.

    PROBLEM CONCLUSION:
    The find code presently uses 'getgrent()' for the -nogroup
    option. getgrent() does not recognise a DCE group as it only
    searches for the group present in /etc/group. The -nogroup
    for find command is now implemented using getgrgid() which
    gets the DCE group for the gid mentioned as a parameter
    to it. find command now works similarly for both
    -nouser and -nogroup options.

    ------

    APAR: IY17556 COMPID: 5765C3403 REL: 430
    ABSTRACT: FTOK() GENERATES DUPLICATE KEYS

    PROBLEM DESCRIPTION:
    ftok() generates duplicate keys

    PROBLEM SUMMARY:
    ftok() subroutine generates duplicate
    keys occasionally, whereas ideally it
    should not.
    The logic for generating keys is changed.
    The new logic generates unique keys.

    PROBLEM CONCLUSION:
    The suggested new logic solves the problem.

    ------

    APAR: IY17709 COMPID: 5765C3403 REL: 430
    ABSTRACT: ATOF() PERFORMING FLOOR OPERATION IN FR_FR LOCALE

    PROBLEM DESCRIPTION:
    atof() returns wrong values in fr_FR locale.

    PROBLEM SUMMARY:
    atof() returns wrong values in fr_FR locale

    PROBLEM CONCLUSION:
    Code has been modified to solve this problem.

    ------

    APAR: IY17884 COMPID: 5765C3403 REL: 430
    ABSTRACT: GXT6000P HANGS WHEN RENDERING MULTI-VERTEX ARRAYS

    PROBLEM DESCRIPTION:
    OpenGL application on the GXT6000P may dump core or hang the
    graphics display when rendering multi-vertex arrays. This
    includes the OpenGL functions glMultiDrawArraysEXT,
    glMultiDrawElementsEXT, glMultiModeDrawArraysIBM, and
    glMultiModeDrawElementsIBM.
    This problem can only occur if GL_CULL_VERTEX_IBM (see
    GL_IBM_vertex_cull extension), a primitive has a 0 or less
    vertex count, or a primitive "mode" is invalid.

    PROBLEM CONCLUSION:
    Move userqueue assurance checks to the top of the primitive
    look so that check are made on the primitives rendered.
    Otherwise, without adequate assurance UQ buffers may overflow.

    ------

    APAR: IY17885 COMPID: 5765C3403 REL: 430
    ABSTRACT: EDGE FLAG DOES NOT WORK IN PVT_BUTTERFLY

    PROBLEM DESCRIPTION:
    OpenGL applications running on the GXT6000P will not get
    correct edge flags when rendering vertex arrays with edgeflags
    specified using the glEdgeFlagPointerListIBM function call.
    Using glEdgeFlagPointer or glEdgeFlagPointerEXT are ok.

    PROBLEM CONCLUSION:
    Reverse the setting of the slow and fast vertex arrays
    rendering flags. They were backwards.

    ------

    APAR: IY17916 COMPID: 5765C3403 REL: 430
    ABSTRACT: KSH BUILTIN TYPESET -F DISPLAYS FUNCTIONS PARTIALLY

    PROBLEM DESCRIPTION:
    ksh builtin "typeset -f" prints partially,
    the functions defined for the shell.

    PROBLEM SUMMARY:
    ksh builtin "typeset -f" prints partially
    the functions defined for that shell.

    PROBLEM CONCLUSION:
    Suitable code changes were made in ksh
    to take care of this problem.

    ------

    APAR: IY17962 COMPID: 5765D5100 REL: 320
    ABSTRACT: EUNFENCE ERRORS

    PROBLEM DESCRIPTION:
    Errors during Eunfence

    PROBLEM SUMMARY:
    On the SP Switch2 if an Eunfence operation fails (times
    out) because the node is not ready to be unfenced, the
    primary node will see two 'OFFLINE REQUEST' packets.
    These packets should not be sent because in this
    case the route to the target node is not valid.

    PROBLEM CONCLUSION:
    The FencePrimitive() function on the SP Switch2
    was checking for null routes. It has been changed
    to check for loopback routes to the primary.
    If the target route is a loopback route, the
    OFFLINE REQUEST will not be sent.

    ------

    APAR: IY18066 COMPID: 5765D5100 REL: 320
    ABSTRACT: CEDIAG:NEED CLEAN UP BETWEEN SDR AND ETC/HOSTS

    PROBLEM DESCRIPTION:
    cediag:need clean up between sdr and etc/hosts

    PROBLEM SUMMARY:
    A completed frame needs to have the name and ipaddr in the
    /etc/host file to avoid irrelevant error message in
    setup_serve

    PROBLEM CONCLUSION:
    The nodes in a Completed frame are added to /etc/hosts

    ------

    APAR: IY18067 COMPID: 5765D5100 REL: 320
    ABSTRACT: CEDIAG:DOCUMENTATION NEEDED FOR NODECOND S1TERM WINDOW

    PROBLEM DESCRIPTION:
    cediat:documentation needed for nodecond s1term window

    PROBLEM SUMMARY:
    The SDR and the FFDC may display error messages that are no
    relevant to running the cediags

    PROBLEM CONCLUSION:
    stderr was sent to /dev/null, and a message is displayed to
    ignore these 2 specific errors.

    ------

    APAR: IY18074 COMPID: 5765D5100 REL: 320
    ABSTRACT: CEDIAGS::FAILING DURING DIAGNOSTIC RESULTS

    PROBLEM DESCRIPTION:
    cediags::failing during diagnostic results

    PROBLEM SUMMARY:
    Some new adapters take up 2 lines when queried from lscfg.
    This caused a parsing error.

    PROBLEM CONCLUSION:
    The query was changed to be more specific about what it
    wanted which compensates for the longer output

    ------

    APAR: IY18075 COMPID: 5765D5100 REL: 320
    ABSTRACT: CEDIAGS:COLONY DIAGNOSTICS FAILURE

    PROBLEM DESCRIPTION:
    cediags:colony diagnostics failure

    PROBLEM SUMMARY:
    The switch code changed the output from colony to SPSwitch2,
    while cediags was still looking for colony.

    PROBLEM CONCLUSION:
    change the search from colony to Switch2

    ------

    APAR: IY18098 COMPID: 5765C3403 REL: 430
    ABSTRACT: PAX:DIRECTORY TRAVERSING ERROR

    PROBLEM DESCRIPTION:
    When the customer creates a file system in a
    directory and tries to run the following command,
            pax -wXvf /tmp/test.pax .
    He gets the error saying
            "./sub-dir:Directory already traversed."

    PROBLEM SUMMARY:
    When a subdirectory is on a different device from its
    parent and the command,
    pax -wXvf /tmp/test.pax .
    run in the parent directory results the following
    error:
     "./sub-dir:Directory already traversed."

    PROBLEM CONCLUSION:
    The code maintains a linked list data structure
    to keep track of already traversed directories.
    It stores only i-node number.When archiving, the
    code only checks for i-node number but not the
    device id.
    The problem is fixed by checking for device
    id and i-node number.

    ------

    APAR: IY18100 COMPID: 5765C3403 REL: 430
    ABSTRACT: NOHUP FAILS WITH BAD FILE DESCRIPTOR ERROR WITH PIPELINE

    PROBLEM DESCRIPTION:
    The nohup command fails on pipeline commands
    with error "Generated or received file descriptor
    that is invalid"

    PROBLEM SUMMARY:
    nohup command when executed with scripts
    containing pipeline commands fails
    with bad file descriptor error.
    This happens only on logout after running
    the script.

    ------

    APAR: IY18131 COMPID: 5765C3403 REL: 430
    ABSTRACT: PROBLEM WITH PAX RESTORE PERMISSION ON -P FLAG

    PROBLEM DESCRIPTION:
    If a directory is listed in the archive and the directory
    already exists at the time of extraction, then mkdir fails even
    with -p option on pax command which is meant for resetting
    permissions or groups.

    PROBLEM SUMMARY:
    If a directory is listed in the archive and the directory
    already exists at the time of extraction, then mkdir fails even
    with -p option on pax command which is meant for resetting
    permissions or owners.

    PROBLEM CONCLUSION:
    pax command will be modified to reset the permissions/owner of
    the file during extraction with the -p option even when a
    directory is already existing. And without the -p option the
    directory that is already existing will fail on overwrite and
    the operation will be continued as per POSIX standards
    conformance.

    ------

    APAR: IY18223 COMPID: 5765C3403 REL: 433
    ABSTRACT: READ/WRITE CONTENTION PROBLEM IN FTPD

    PROBLEM DESCRIPTION:
    read/write contention problem in ftpd.

    ------

    APAR: IY18265 COMPID: 5765C3403 REL: 433
    ABSTRACT: SECURITY: BUFFER OVERFLOW IN XNTPD

    PROBLEM DESCRIPTION:
    A buffer overflow vulnerability exists in (x)ntpd.

    PROBLEM SUMMARY:
    A buffer overflow vulnerability exists in (x)ntpd. A 128-byte
    string is not properly bounded in maximum size.

    PROBLEM CONCLUSION:
    Properly bound the 128-byte string.

    ------

    APAR: IY18305 COMPID: 5765C3403 REL: 430
    ABSTRACT: AWK -F SETTING IS IGNORED BY GETLINE

    PROBLEM DESCRIPTION:
    awk -F setting is ignored by getline

    ------

    APAR: IY18364 COMPID: 5765D5101 REL: 120
    ABSTRACT: CUSTOMER WANTS DUMPS IN /VAR/ADM/FFDC/DUMPS TO BE CLEANED UP BY

    PROBLEM DESCRIPTION:
    From the customer:
    Whenever the Topology Services daemon on remote nodes
    stopped an error log entry TS_NODEDOWN_EM is created on
    the neighbor nodes and a hats dump file is written to the
    /var/adm/ffdc/dumps on the nodes. These files fill up the
    /var on all nodes and never get cleaned up by hats!
    The /var inodes and filesystem will fill up which causes
    serious problems in the system operation.
    Clean the hats dumps when hats is stopped on the node.

    PROBLEM SUMMARY:
        Topology Services daemon generates an FFDC dump file to
    describe the remote node up/down event whenever a remote
    node up/down event is detected. These dump files are useful
    for debugging when Topology Services problems encounter.
    Nodes may be shutdown regularly for maintenance and cause
    other nodes to produce large number of FFDC dump files in
    large systems. The Topology Services should not generate
    the FFDC dump files in normal operations.

    PROBLEM CONCLUSION:
        The fix turns off the use of FFDC dump file in the
    Topology Services subsystem by default so that no FFDC dump
    files are generated in normal operations. When the system
    encounters problems and requires more debugging information,
    the Topology Services subsystem resumes writing FFDC dump
    files to produce the debugging information when the trace
    is turned on.

    ------

    APAR: IY18502 COMPID: 5765C3403 REL: 430
    ABSTRACT: SMITTY CHTZ SHOWS INCORRECT SDSD FOR EASTERN AUSTRALIA

    PROBLEM DESCRIPTION:
    smitty chtz shows M3.1.0 as the Stop Daylight Savings Day
    for Eastern Astrailian time zone(EET-10EETDT).
    The correct value is M3.5.0.

    PROBLEM SUMMARY:
    SMITTY CHTZ SHOWS INCORRECT SDSD FOR EASTERN AUSTRALIA

    PROBLEM CONCLUSION:
    necessary changes are made in the ODM files.

    ------

    APAR: IY18513 COMPID: 5765C3403 REL: 430
    ABSTRACT: PTY CAN WASTE CPU TIME DURING SLAVE CLOSE

    PROBLEM DESCRIPTION:
    When a pts closes and there is output on the ptc, the thread
    closing the pts will wake itself up along with the gil thread
    in an endless loop. This consumes cpu time for no benifit.

    PROBLEM CONCLUSION:
    The solution is to remove two unneeded qenables from the
    code.

    ------

    APAR: IY18543 COMPID: 5765C3403 REL: 430
    ABSTRACT: PROBLEM WITH SMITTY CHTZ_USER MENU WHEN SOME FIELD ARE LEFT

    PROBLEM DESCRIPTION:
            When the user time zone is set(smitty chtz_user)
    with Start and Stop Daylight Savings Day fields blank,
    the values in the Start and Stop Daylight Savings Time
    fields appear in the Start and Stop Daylight Savings Day
    fields when smitty chtz_user command is run in a new session.

    PROBLEM SUMMARY:
    Problem with smitty chtz_user menu when some fields are left
    blank.

    PROBLEM CONCLUSION:
    necessary changes are made in the chtz.add file.

    ------

    APAR: IY18609 COMPID: 5765C3403 REL: 430
    ABSTRACT: FUSER WAS NOT DETECTING MMAP FILES

    PROBLEM DESCRIPTION:
    This apar has two fixes one to fix the -x option which was
    not detecting all the processes using the shared libraries
    and it also adds a improved function for detecting the
    mmap files.

    PROBLEM SUMMARY:
    fuser does not detect certain files (like memory mapped files,
    shared libraries) in the memory even if they are open and
    being used.

    PROBLEM CONCLUSION:
    fuser is fixed in the -x option to detect shared libraries
    and mmap files. The pid's of the mmap files will be appended
    with "m" to identify it is a mmap file.

    ------

    APAR: IY18629 COMPID: 5765C3403 REL: 430
    ABSTRACT: W COMMAND FORMATTING PROBLEMS WITH 8 CHARACTER USERNAMES

    PROBLEM DESCRIPTION:
    When running the w command in certain languages
    hu_HU for example, and when there is a user name
    that is exactly 8 characters long - the username
    displayed may print extra garbage characters.

    PROBLEM CONCLUSION:
    Modify the w command to only print the user name and not
    the extra characters.

    ------

    APAR: IY18700 COMPID: 5765D5100 REL: 320
    ABSTRACT: MKNIMINT BROKEN WHEN NETMASK SET IN HEX. THIS CAUSES MKNIMINT

    PROBLEM DESCRIPTION:
    mknimint calls the "ipfill" subroutine on the netmask of
    each adapter on the cws. ipfill expects that netmask to be
    in dotted decimal form, e.g. 255.255.255.0, but mknimint
    gets the netmask from the output of lsattr, which may
    report the netmask in hex, e.g. 0xffffff00, if the
    interface was originally configured that way.
    This causes mknimint to exit with a confusing message:
    "incorrect number of tuples in IP address"

    LOCAL FIX:
    Change the netmask from hex format to decimal.

    PROBLEM SUMMARY:
    Node installation fails if any octet of an IP address or of
    a mask are given in octal or hexadecimal. The errors issued
    from mknimint are "0016-202 The IP address tuple is not
    numeric". and "0016-201 There is an incorrect number of
    tuples in the IP address."

    PROBLEM CONCLUSION:
    The ipfill routine is part of the install_lib.pl module, and
    is invoked by a number of installation scripts to create a
    "normalized" IP address. It has been corrected to properly
    convert hexadecimal or octal octets to decimal before
    normalizing them.

    ------

    APAR: IY18709 COMPID: 5765D6100 REL: 220
    ABSTRACT: LLQ -S <JOBID> COMMAND GENERATES A CORE

    PROBLEM DESCRIPTION:
    LLQ -s <jobid> command generates a core

    PROBLEM SUMMARY:
    LoadLeveler llq -s jobid would sometimes core dump.

    PROBLEM CONCLUSION:
    LoadLeveler llq -s jobid will now execute
    successfully without core dumping.

    ------

    APAR: IY18724 COMPID: 5765D5101 REL: 120
    ABSTRACT: TOPSVCS DOES NOT RECOGNIZE 2ND RS232 NET IN 2 NODE CLUSTER

    PROBLEM DESCRIPTION:
        This problem occurs because a two-node cluster was
    treated as a special case and was assumed to have one non-IP
    network per non-IP adapter type only. Using multiple non-IP
    networks to avoid "false down" problem has become popular
    in many HACMP clusters. Assuming a two-node cluster has
    one non-IP network is no longer true. RSCT Topology Services
    should remove the assumption and treat two-node clusters the
    same as larger clusters.
       This fix, however, changes the way the Topology Services
    configure the network topology. This fix must be installed
    on both nodes before starting the cluster to avoid
    inconsistency for Topology Services. ALSO: an alternative is
    to remove all the non-IP networks in a two-node cluster --
    via a Topology DARE -- before applying the fix, and only
    re-add the non-IP networks after the fix has been installed
    in the entire cluster.

    PROBLEM CONCLUSION:
        This fix removes the assumption that a two- node cluster
    has only one non-IP network of the same type and treats
    two-node clusters the same way as larger clusters.

    ------

    APAR: IY18730 COMPID: 5765C3403 REL: 430
    ABSTRACT: DTLOGIN USER DOES NOT HAVE NLSPATH

    PROBLEM DESCRIPTION:
    When logged into CDE, NLSPATH is NULL.

    PROBLEM CONCLUSION:
    Handle _DtEnvControl function to reset the variable.

    ------

    APAR: IY18732 COMPID: 5765C3403 REL: 433
    ABSTRACT: FTPD LOGS DLOPEN ERROR

    PROBLEM DESCRIPTION:
    A fix was put into ftpd (IY15897) to cleanup the dce
    credential files of an integrated login dce user. This fix
    was not correct for a non-dce user or root dce user.

    PROBLEM CONCLUSION:
    In dologout, make correct check for KRB5CCNAME and also
    check to see if user has root credentials (which should
    never be removed).

    ------

    APAR: IY18739 COMPID: 5765D6100 REL: 220
    ABSTRACT: LLSUMMARY -L VS LLSUMMARY -L -X NOT SHOW ALL NODES USED

    PROBLEM DESCRIPTION:
    llsummary -l only shows one set of nodes being used which
    happen to have the same number of tasks.

    PROBLEM SUMMARY:
    For parallel jobs, llsummary may show incorrect values for
    Min Processors, Max Processors, Allocated Host Count, and
    Allocated Host List. This happens when the job command
    file contained a total_tasks value that was not an even
    multiple of the node value.

    PROBLEM CONCLUSION:
    llsummary was changed to loop to find the correct values
    for Min Processors, Max Processors, Allocated Host Count,
    and the Allocated Host List.

    ------

    APAR: IY18746 COMPID: 5765C3403 REL: 433
    ABSTRACT: LED 581 WITH BOS.NET.TCP.CLIENT 4.3.3.29

    PROBLEM DESCRIPTION:
    led 581 with bos.net.tcp.client 4.3.3.29

    PROBLEM SUMMARY:
    When /etc/rc.net script is run from the command line, led
    581 is set.

    PROBLEM CONCLUSION:
    The problem was that the phase(runtime and bootphase) was not
    set properly and the leds were not reset when exiting from
    cfgech.

    ------

    APAR: IY18769 COMPID: 5765D5100 REL: 320
    ABSTRACT: VSD WOULDN'T CONFIGURE ON SOME NODES DUE TO MAJOR NUMBER

    PROBLEM DESCRIPTION:
    The VSD configuration routines will add defensive code to check
    for major number mismatches and "remjor" the vsd if necessary.
    Under some unknown (very rare) circumstances the VSD major
    number is getting reassigned in the Object Data Manager(ODM).
    This causes a mismatch between the major numbers that have been
    assigned to existing shared devices and the device driver major
    in the ODM. As a result, the VSD device driver will not load and
    and vsds can not be configured.
    You will get an error message in the consol log like:

    PROBLEM SUMMARY:
    Under some unknown (very rare) circumstances the VSD
    major number is getting reassigned in the Object
    Data Manager (ODM). This causes a mismatch between
    the major numbers that have been assigned to existing
    shared devices and the device driver major number in
    the ODM. As a result, the VSD device driver will
    not load and vsds can not be configured.
    You will get an error message in the consol log like:
     0034-036 special device file /dev/rvsdname already exists
     and does not have vsd major number of XX

    PROBLEM CONCLUSION:
    The VSD configuration routines will add defensive
    code to check for major number mismatches and
    "remajor" the vsds if necessary.

    ------

    APAR: IY18778 COMPID: 5765C3403 REL: 430
    ABSTRACT: PROBLEMS HANDLING NATIONAL CHARACTERS IN ISO8859-2 LOCALES

    PROBLEM DESCRIPTION:
    After applying apar IY11380 (correct use of
    MB_CUR_MAX ) certain AIX commands have problems
    processing national characters from ISO8859-2 and
    ISO8859-5 codesets correctly.
    Specifically, the tr command may corrupt national
    characters embedded in the datastream which were
    not specifically included in the transformation...
    Other commands such as the more command may
    also be effected...

    PROBLEM SUMMARY:
    National characters from ISO8859-2 and ISO8859-5 based
    locales may be corrupted when using commands that handle
    character data such as tr or more.

    PROBLEM CONCLUSION:
    Modify the putwc() function in libc.a to not assume that
    wchar_t value and mb value when MB_CUR_MAX = 1.
    Since wchar_t values are stored in Unicode, this assumption
    does not hold true in certain locales.

    ------

    APAR: IY18809 COMPID: 5765C3403 REL: 430
    ABSTRACT: THE X-SEVER MAY CORE DUMP OR LEAK MEMORY.

    PROBLEM DESCRIPTION:
    The X-Server may core dump or leak memory.

    PROBLEM CONCLUSION:
    Assign pGC->ops->devPrivate.val to 0 so the pGC->ops static
    structure in sonValidateGC will not be freed.

    ------

    APAR: IY18811 COMPID: 5765C3403 REL: 430
    ABSTRACT: INCORRECT RENDERING ON VERTEX COLORED POLYGONS

    PROBLEM DESCRIPTION:
    Vertex colored polygons may not appear as expected.

    PROBLEM CONCLUSION:
    Force vertex colored polygons down the original code path.

    ------

    APAR: IY18813 COMPID: 5765C3403 REL: 430
    ABSTRACT: MOTIF UNABLE TO DISPLAY TO DUAL-HEAD SUN CDE

    PROBLEM DESCRIPTION:
    A Motif program running on AIX will not be able to display to
    a dual-head Sun machine running Sun's CDE.

    PROBLEM CONCLUSION:
    Implement fix accomodating changes to Sun's dtsession.

    ------

    APAR: IY18925 COMPID: 5765C3403 REL: 430
    ABSTRACT: GRAPHIGS APP. LEAVES DEFUNCT PROCESS WITH AFM_CHILD_NUC=1

    PROBLEM DESCRIPTION:
    Customers graPHIGS based application can have
    multiple workstations opened simultaneously,
    and may dynamically open and close
    workstations. If AFM_CHILD_NUC=1 is used,
    this can result in defunct process(es).
    Although these defunct processes are in no way
    useing any CPU or memory resources, they can
    be disconcerting.

    LOCAL FIX:
    Don't use AFM_CHILD_NUC=1.

    PROBLEM SUMMARY:
    Customers graPHIGS based application can have
    multiple workstations opened simultaneously,
    and may dynamically open and close
    workstations. If AFM_CHILD_NUC=1 is used,
    this can result in defunct process(es).
    Although these defunct processes are in no way
    useing any CPU or memory resources, they can
    be disconcerting.

    PROBLEM CONCLUSION:
    As part of Close Connection, if a child nucleus is being
    used, invoke wait to acknowledge termination of child process.

    ------

    APAR: IY18930 COMPID: 5765D5101 REL: 120
    ABSTRACT: HATS ON STARTUP SHOULD MAKE SURE /ETC/SERVICES AND SYSPAR_PORTS

    PROBLEM DESCRIPTION:
    On startup hats should check that the port in /etc/services
    matches that in the SDR Syspar_ports class. If it does not, it
    should log errors indicating such and exit.

    LOCAL FIX:
    Change SDR and /etc/services so the hats port number matches

    PROBLEM SUMMARY:
    If the hats (topology services) port specified in
    /etc/services does not match that in the SDR Syspar_ports
    class, upon startup hats will use the entry in
    /etc/services and ignore the SDR. But if the subsystem is
    restored (syspar_ctrl -R hats), /etc/services will be
    changed to match the SDR. Thus you can get into the
    situation where hats on different nodes are using different
    ports, in which case they will form different groups. This
    should not be allowed to happen.

    PROBLEM CONCLUSION:
    A check has been added to the hats startup script. If the
    hats entry in /etc/services does not match the SDR, hats
    will exit and log appropriate error messages.

    ------

    APAR: IY18937 COMPID: 5765C3403 REL: 430
    ABSTRACT: NO VALUES SHOWN IN DEVICE PROPERTIES (TOK0)

    PROBLEM DESCRIPTION:
    Token ring attributes in WebSM are not translated into any
    foreign language.

    PROBLEM CONCLUSION:
    Added NLS_VALUES to the ODM with the translatable values.

    ------

    APAR: IY18978 COMPID: 5765C3403 REL: 430
    ABSTRACT: TABLET USED AS DEVICE POINTER IN 1600X1200 RESOLUTION NOT

    PROBLEM DESCRIPTION:
    When the screen resolution is set to 1600X1200, and the
    tablet is used as the pointing device (mouse), the tablet
    can NOT be used to access all portions of the screen.

    PROBLEM CONCLUSION:
    Modify initialization logic in loadddx for sonora and mtn
    family of adapter to get the true screen resulution.

    ------

    APAR: IY18981 COMPID: 5765C3403 REL: 430
    ABSTRACT: GIGABIT ETHERNET ADAPTER COMES UP IN DEFINED STATE

    PROBLEM DESCRIPTION:
    During adapter config of the Gigabit Ethernet Fiber or UTP
    adapter, the config method will fail with the following
    error message:
    Method error (/usr/lib/methods/cfggxent):
            0514-050 Cannot perform the requested function
                     because the specified device was not
                     detected.

    PROBLEM CONCLUSION:
    Replace io_delay() system call with delay() system call
    in adapter initialization routine.

    ------

    APAR: IY18983 COMPID: 5765C3403 REL: 430
    ABSTRACT: TOTAL_LEN CALCULATED INCORRECTLY IN SAVE_PUBLIC_TOKEN_OBJECT

    PROBLEM DESCRIPTION:
    Loading a previously saved public token object can fail.

    PROBLEM CONCLUSION:
    Corrected the calculation of total length when saving a public
    token object.

    ------

    APAR: IY18985 COMPID: 5765C3403 REL: 430
    ABSTRACT: POSSIBLE CORE DUMP IF LOADING STDLL FAILS

    PROBLEM DESCRIPTION:
    If the STDLL doesn't load successfully, an application that
    tries to use PKCS #11 functions could core dump.

    PROBLEM CONCLUSION:
    The function list pointer is always checked to make sure it's
    valid before any PKCS #11 functions can be called.

    ------

    APAR: IY18988 COMPID: 5765C3403 REL: 430
    ABSTRACT: XETHCHAN_TRACE_LOCK IS KILLING ETHERCHANNEL PERFORMANCE

    PROBLEM DESCRIPTION:
    netif_backup introduces an unneeded lock into etherchannel
    that significantly slows etherchannel down

    PROBLEM SUMMARY:
    ***************************************************************
    *USERS AFFECTED: *
    * All users with the following filesets at these levels *
    * devices.common.IBM.ethernet.rte 4.3.3.51
    ***************************************************************
    *PROBLEM DESCRIPTION: *
    * netif_backup introduces an unneeded lock into
    * etherchannel that significantly slows etherchannel
    * down
    ***************************************************************
    *RECOMMENDATION: *
    * Apply apar IY18988
    ***************************************************************

    PROBLEM CONCLUSION:
    remove the lock from the transmit path for none netif_backup
    etherchannel configurations

    ------

    APAR: IY18989 COMPID: 5765C3403 REL: 430
    ABSTRACT: NETWORK TIME PROTOCOL: MULTICASTING DOSEN'T SEEM TO BE

    PROBLEM DESCRIPTION:
    When Network Time Protocol multicast is done nothing happens
    i.e it doesn't ntp multicast

    PROBLEM CONCLUSION:
    The new code actually sendsto 224.0.1.1 ( with the correct MAC
    for the mcast group ) on each interface.

    ------

    APAR: IY18992 COMPID: 5765C3403 REL: 430
    ABSTRACT: DES KEYS SHOULD BE GENERATED RANDOMLY AND HAVE CORRECT PARITY

    PROBLEM DESCRIPTION:
    DES and 3DES keys generated by C_GenerateKey using ICA are not
    sufficiently random and don't have parity bits set to conform
    to the NIST standard.

    PROBLEM CONCLUSION:
    Filled key with random data and implemented parity-fixing code.

    ------

    APAR: IY18993 COMPID: 5765C3403 REL: 430
    ABSTRACT: ATM WORKAROUNDS FOR UNIQUE PCI BRIDGES

    PROBLEM DESCRIPTION:
    need to recognize new model architecture for OEM

    PROBLEM CONCLUSION:
    change config method to recognize OEM model architecture

    ------

    APAR: IY18996 COMPID: 5765C3403 REL: 430
    ABSTRACT: LSSRC -LS NAMED: MISSING LINES FROM OUTPUT WHEN USING NAMED8

    PROBLEM DESCRIPTION:
    Fixes the problem of every seventh line of the expected
    output of "lssrc -ls named" (where named8 is being used)
    being skipped.

    PROBLEM SUMMARY:
    Loop that consolidates status information was fixed so it
    is exited after and not just before processing the last
    item of each set of 7 elements.

    ------

    APAR: IY19006 COMPID: 5765C3403 REL: 430
    ABSTRACT: GXT6000P HANGS RUNNING CATIA SHD WITH PTF U474642

    PROBLEM DESCRIPTION:
    With GXT6000P adapter with PTF U474642 installed, Catia
    application hangs with a specific customer model in
    SHD mode when being rotated. Actually graPHIGS is in
    a tight loop, which makes it appear that X is hung.

    PROBLEM SUMMARY:
    ***************************************************************
    *USERS AFFECTED: *
    * All users with the following filesets at these levels *el
    * PEX_PHIGS.dev.pci.14106e01 4.3.3.51 on irregular pixel
    * PEX_PHIGS.dev.pci.14106e01 4.3.3.53e a dust particle. It
    * PEX_PHIGS.dev.pci.14107001 4.3.3.51ain area of a model.
    * PEX_PHIGS.dev.pci.14107001 4.3.3.53lowly, and a hang
    ***************************************************************
    *PROBLEM DESCRIPTION: *
    * The following problem occurs on GXT6000P when doing
    * clipped polygons: Dust particles are displayed in
    * certain area of a model. Pixel level holes or
    * overpaints are displayed on irregular pixel position
    * for each redraw. It looks like a dust particle. It
    * emerges not on whole screen but on certain area of a
    * model. The original fix for this problem ran slowly,
    * and a hang existed in certain situations.
    ***************************************************************
    *RECOMMENDATION: *
    * Apply apar IY19006
    ***************************************************************

    PROBLEM CONCLUSION:
    Utilize a different set of hardware commands when rendering
    clipped polygons on a GXT6000P.

    ------

    APAR: IY19008 COMPID: 5765C3403 REL: 430
    ABSTRACT: CHECK_UNIX MIGHT FAIL

    PROBLEM DESCRIPTION:
    Check_unix may fail and thereby snap will
    fail to collect dump.

    LOCAL FIX:
    use dd to collect dump.

    PROBLEM CONCLUSION:
    Make sure malloc does not fail with error.

    ------

    APAR: IY19032 COMPID: 5765D5100 REL: 320
    ABSTRACT: EUNFENCE TIMEOUT

    PROBLEM DESCRIPTION:
    Eunfence timeout colony large system

    PROBLEM SUMMARY:
    Timeouts insufficient for Eunfences on large systems.
    causes an Estart to be redriven.

    PROBLEM CONCLUSION:
    Base timeouts in broadcast routines were modified.

    ------

    APAR: IY19069 COMPID: 5765D5100 REL: 320
    ABSTRACT: DCE KEYFILE DIRECTORIES HAVE PERMISSION 777

    PROBLEM DESCRIPTION:
    The installation of PSSP 3.2 with DCE 3.1
    authentification creates keyfile directories
    /spdata/sys1/keyfiles/ssp,
    /spdata/sys1/keyfiles/rsct,
    /spdata/sys1/keyfiles/ppe, mmfs, LoadL,
    and all their subdirectories with permission 0777.
    Thus, any nonprevileged user on an SP node can erase the
    keyfiles, thereby causing failure of SP trusted services.

    LOCAL FIX:
    Workaround is to change the dir to 755 or 700.

    PROBLEM SUMMARY:
    create_keyfiles was creating directories with incorrect
    permissions.

    PROBLEM CONCLUSION:
    Modified create_keyfiles to create directories with correct
    permissions.

    ------

    APAR: IY19087 COMPID: 5765D5100 REL: 320
    ABSTRACT: SVC_WINDOW_FULL THRESHOLD TO LOW

    PROBLEM DESCRIPTION:
    Node received an extremely large number of
    transient errors in a short period of time.
    This caused the service window to fill and
    the svc_window_full threshold to be exceeded
    resulting in the node dropping off the switch.

    PROBLEM SUMMARY:
    Service window full conditions raised by the microcode were
    resulting in taking the node off the switch.

    PROBLEM CONCLUSION:
    Increase the threshold on servive window full adapter errors
    to 7000 in 1 minute. This error will never threshold and
    take the node off the switch.

    ------

    APAR: IY19094 COMPID: 5765C3403 REL: 430
    ABSTRACT: FIX DEADLOCK CONDITION IN OBJ_MGR.C

    PROBLEM DESCRIPTION:
    Application using object management functions may lock up.

    PROBLEM CONCLUSION:
    Several calls to unlock global_shm->mutex were added along
    error paths in obj_mgr.c to avoid the possibility of deadlock.

    ------

    APAR: IY19112 COMPID: 5765C3403 REL: 430
    ABSTRACT: IP FRAGMENTS GO OUT WRONG ADAPTER OF ETHERCHANNEL

    PROBLEM DESCRIPTION:

    PROBLEM SUMMARY:
    The first IP fragment of a multi-fragment IP packet will use
    destination MAC address for the hash to select the outgoing
    adapter. Instead of using the destination IP address.

    PROBLEM CONCLUSION:
    In IP output processing, when fragmenting a packet, leave
    room for the link-level header in the mbuf.This will ensure
    that Etherchannel sees the the IP header in the first mbuf.

    ------

    APAR: IY19192 COMPID: 576565400 REL: 220
    ABSTRACT: FILEMON STOPAS, DISPLAYS TRC_READ ERRORS DURING HEAVY IO

    PROBLEM DESCRIPTION:
    When the system has a heavy IO or load, if filemon
    is started at this time, then both the offline and
    realtime filemon produces lot of trc_read and trc_lseek
    errors and also it abruptly stops its execution after
    reporting a trc_vfs_ino error stating not a OPEN, OPENX
    or CREAT SVC.

    PROBLEM CONCLUSION:
    Fixed the code to take into account missing SVCs due to
    heavy IO. This will fix the trc_lseek and trc_read errors
    The trc_vfs_ino error is fixed by adding READ and WRITE svcs
    inaddition to OPEN, OPENX and CREATE SVCs in trc_vfs_ino
    error check, hence filemon will stop only if the SVCs are
    not OPEN, OPENX, CREAT, READ and WRITE.

    ------

    APAR: IY19195 COMPID: 5765C3403 REL: 430
    ABSTRACT: INCORRECT ICONV() CONVERSIONS ISO8859-5 TO IBM-866

    PROBLEM DESCRIPTION:
    When using the iconv command or iconv() subroutine
    to convert data from ISO8859-5 to IBM-866 codeset,
    some codepoints do not convert correctly to their
    IBM-866 counterparts.

    PROBLEM CONCLUSION:
    Modify the conversion table ISO8859-5_IBM-866_src to produce
    correct conversion results according to the standard

    ------

    APAR: IY19196 COMPID: 5765C3403 REL: 430
    ABSTRACT: MON-CXMA DOES NOT REPORT TTY NAME CORRECTLY

    PROBLEM DESCRIPTION:
    The mon-cxma program does not report the tty name correctly on
    large systems with multiple buses.

    PROBLEM CONCLUSION:
    The solution is to do the odm searchs better so that the
    proper device is found based upon the parent and location
    code.

    ------

    APAR: IY19209 COMPID: 5765D5100 REL: 320
    ABSTRACT: 384WAY:ADAPTER CONFIGURATION FAILED FOR CSSN IN RC.SWITCH

    PROBLEM DESCRIPTION:
    384way:adapter configuration for cssn in rc.switch

    PROBLEM SUMMARY:
    The SP Switch2 adapter can encounter "A colony PLL lost
    lock" condition as part of its normal running condition. The
    requirement to ignore this error was not forwarded to the
    adapter diagnostics team.

    PROBLEM CONCLUSION:
    The register_test failed when a value of 0x00000008 00000000
    was seen in the NBA LHS Error Accumulator Reg. This
    indicates that "A colony PLL lost lock." condition occurred.
    The test has been changed to ignore this error.

    ------

    APAR: IY19218 COMPID: 5765C3403 REL: 430
    ABSTRACT: CRASH IN SO_PIN_MORE_STACK CALLED WITHIN UDP_INPUT0

    PROBLEM DESCRIPTION:
    System crashes in so_pin_more_stack called within udp_input0
    routine, when "stopsrc -g nfs" command is run.

    PROBLEM CONCLUSION:
    The fix to the problem is two parts :
    udp_usrreq.c :
    increment the ref. count of the member socket that gets
    chosen to handle this connection. This needs to be done
    holding the inpcb hash chain lock.
    socket.c :
    In soclose(), move the section of the code that issues soclose
    on the member sockets of the socket-group into the end of the
    routine after PRU_DETACH.

    ------

    APAR: IY19220 COMPID: 5765C3403 REL: 433
    ABSTRACT: CRASH IN UNP_EXTERNALIZE FUNCTION

    PROBLEM DESCRIPTION:
    crash in unp_externalize function.

    PROBLEM SUMMARY:
    System crashes in unp_externalize routine
    while attempting to clean up the sockets.

    PROBLEM CONCLUSION:
    The crash is because invalid socket pointer
    gets passed as argument to closef. The fix
    requires a variable to be reinitialized
    before closef is called.

    ------

    APAR: IY19226 COMPID: 5765C3403 REL: 430
    ABSTRACT: NON BLOCK ALIGNED TRANSFERS TO RAW CD/DVD MAY CAUSE PANIC

    PROBLEM DESCRIPTION:
    Machine may panic when running io to raw cd/dvb device.
    This may only happen on non block aligned transfers that are
    multiples of 512.

    PROBLEM CONCLUSION:
    Initialization of all buf structures that are created by the
    disk driver for this situation.

    ------

    APAR: IY19247 COMPID: 5765C3403 REL: 430
    ABSTRACT: IPREPORT DOES NOT DECODE THE AAL MODE FIELD IN UNI30 MESSAGES

    PROBLEM DESCRIPTION:
    ipreport will generate a line like the following:
    Q2931: = <<Length unusually small>>

    PROBLEM CONCLUSION:
    Add a case entry to decode the AAL Parameter.

    ------

    APAR: IY19259 COMPID: 5765C3403 REL: 430
    ABSTRACT: WINDOW MANAGER CTWM CAUSES X WINDOWS TO CORE DUMP

    PROBLEM DESCRIPTION:
    The X-Server may core dump when the mouse is moved while
    the display is being rendered.

    PROBLEM SUMMARY:
    ***************************************************************
    *USERS AFFECTED: *
    * All users with the following filesets at these levels *
    * X11.base.rte 4.3.3.50
    * X11.base.rte 4.3.3.51
    * X11.base.rte 4.3.3.53
    ***************************************************************
    *PROBLEM DESCRIPTION: *
    * The X-Server may core dump when the mouse is moved
    * while the display is being rendered.
    ***************************************************************
    *RECOMMENDATION: *
    * Apply apar IY19259
    ***************************************************************

    PROBLEM CONCLUSION:
    Change the WheelMouseID to be different than the FastCursorID.

    ------

    APAR: IY19289 COMPID: 5765D6100 REL: 220
    ABSTRACT: LLCTL PURGESCHEDD HANGS AND FAILS TO REMOVE JOBS

    PROBLEM DESCRIPTION:
    llctl purgeschedd hangs and fails to remove jobs

    PROBLEM SUMMARY:
    When attempting to purge jobs scheduled from a schedd node
    which has been taken down, the llctl purgeschedd command
    does not completely remove jobs. After invoking the command
    the jobs will still appear to be running when an llq command
    is issued.

    PROBLEM CONCLUSION:
    The fix for this problem is to correct the transaction
    protocol
    between the llctl purgeschedd command and the central
    manager to
    allow notification that jobs have been removed.

    ------

    APAR: IY19293 COMPID: 5765C3403 REL: 430
    ABSTRACT: SMIT PARSING PROBLEMS IN ISO8859-15 LOCALES

    PROBLEM DESCRIPTION:
    Running certain dialogs such as smit
    'Add a TTY' may fail in ISO8859-15 based
    locales.

    PROBLEM CONCLUSION:
    Modify the regular expression routines in libc.a to correctly
    parse the data.

    ------

    APAR: IY19305 COMPID: 5765C3403 REL: 430
    ABSTRACT: INVALID ENTRY IN /ETC/ENVIRONMENT

    PROBLEM DESCRIPTION:
    Using chlang command to change the LANG environment
    creates an invalid entry "export LC_MESSAGES" in
    /etc/environment file.

    PROBLEM SUMMARY:
    When the command 'chlang -m Ja_JP:en_US ja_JP' is run to
    change the locale environment, an invalid entry,
    'export LC_MESSAGES' is created in /etc/environment file.

    PROBLEM CONCLUSION:
    Necessary code changes are made in the chlang script.

    ------

    APAR: IY19320 COMPID: 5765C3403 REL: 433
    ABSTRACT: TMD COREDUMP WITH IRE CLIENT

    PROBLEM DESCRIPTION:
    tmd coredump, the errorlog has following error:
        PROGRAM NAME
        tmd
        ADDITIONAL INFORMATION
        -04--80-8-80- EC
        -04--80-8-80- D4
         1170
        process_m 1C
        process_m 28
        ProcessDa 104
        _pthread_ D4

    PROBLEM CONCLUSION:
    Correctly set the timer for phase1 tunnel lifetime expiration
    in tmd will fix the problem.

    ------

    APAR: IY19322 COMPID: 5765B9501 REL: 320
    ABSTRACT: PANIC IN VFS_UNHOLD USING SOFT MOUNT

    PROBLEM DESCRIPTION:
    PANIC IN VFS_UNHOLD USING SOFT MOUNT

    PROBLEM SUMMARY:
    A node paniced in VFS_UNHOLD.

    PROBLEM CONCLUSION:
    Do not steal unused cached vnodes for a
    soft mount.

    ------

    APAR: IY19334 COMPID: 5765C3403 REL: 430
    ABSTRACT: GXT6000P DSI

    PROBLEM DESCRIPTION:
    While running the OGL stress test suite on Sphinx 2 with 64
    bit kernel, the GXT6000P device driver could cause the data
    storage interrupt.

    PROBLEM CONCLUSION:
    A device lock is also needed to synchronize the dirty_tids
    list in the remove_texture_thread routine with the rest
    places in the MP environment.

    ------

    APAR: IY19362 COMPID: 5765B9501 REL: 330
    ABSTRACT: PANIC IN VFS_UNHOLD USING SOFT MOUNT

    PROBLEM DESCRIPTION:
    PANIC IN VFS_UNHOLD USING SOFT MOUNT

    PROBLEM SUMMARY:
    A node paniced in VFS_UNHOLD.

    PROBLEM CONCLUSION:
    Do not steal unused cached vnodes for a
    soft mount.

    ------

    APAR: IY19372 COMPID: 5765C3403 REL: 430
    ABSTRACT: 'RMDEV -DL' RETURNS EBUSY

    PROBLEM DESCRIPTION:
    If you varyon a volume group in concurrent
    mode (using 'varyonvg -c <vgname>') then
    after you varyoffvg you will not be able to
    'rmdev -dl <hdiskname>' any of the disks in
    that volume group. The rmdev returns a
    message stating that the device is busy.
    Rebooting the system clears this state and
    should allow the rmdev to succeed.

    PROBLEM SUMMARY:
    ***************************************************************
    *USERS AFFECTED: *
    * All users with the following filesets at these levels *
    * bos.rte.lvm 4.3.3.26.
    * bos.rte.lvm 4.3.3.27
    * bos.rte.lvm 4.3.3.28
    * bos.rte.lvm 4.3.3.29
    * bos.rte.lvm 4.3.3.50
    * bos.rte.lvm 4.3.3.52
    ***************************************************************
    *PROBLEM DESCRIPTION: *
    * If you varyon a volume group in concurrent mode (using
    * 'varyonvg -c <vgname>') then after you varyoffvg you
    * will not be able to 'rmdev -dl <hdiskname>' any of the
    * disks in that volume group. The rmdev returns a
    * message stating that the device is busy. Rebooting
    * the system clears this state and should allow the
    * rmdev to succeed.
    ***************************************************************
    *RECOMMENDATION: *
    * Apply apar IY19372
    ***************************************************************

    PROBLEM CONCLUSION:
    The 'varyonvg -c <vgname>' path was mistakely
    doing two opens of the disk when it should
    only be doing one. Modified the code so that
    we only do one open.

    TEMPORARY FIX:
    A reboot of the system will clear this condition
    and allow you to successfully 'rmdev -dl' the
    disk.

    ------

    APAR: IY19377 COMPID: 5765C3403 REL: 430
    ABSTRACT: INTERMITTENT CORE DUMP IN CLOSEANINST()

    PROBLEM DESCRIPTION:
    Server core dump in closing AnInstance

    PROBLEM SUMMARY:
    Server core dump

    PROBLEM CONCLUSION:
    Code fix in closing session's instance.

    ------

    APAR: IY19383 COMPID: 5765B9501 REL: 320
    ABSTRACT: MMFSD HANGS ON SHUTDOWN

    PROBLEM DESCRIPTION:
    When shutting down GPFS sometimes mmfsd does not exit but is
    left hanging

    PROBLEM SUMMARY:
    GPFS termination deadlocked because of a lock conflict in
    AIX process termination logic freeing memory.

    PROBLEM CONCLUSION:
    Directly call the _exit routine instead of using the libc
    version.

    ------

    APAR: IY19392 COMPID: 5765B9501 REL: 320
    ABSTRACT: PFS DOESN'T HANDLE LOCKF WITH MAXIMUM FILE SIZE (OR FTLOCK)

    PROBLEM DESCRIPTION:
    Using the following program
     include fcntl.h
     include stdio.h
     include sys/utsname.h
    main()
    struct utsname name;
    FILe *myfile;
    int fd;
    char *filename="/ptmp/polo";
    uname(&name);
    if((fd=open(filename,O_WRONLY]O_CREAT]O_APPEND,0666))==-1) retur
    n;
    else chmod(filename,0666);
    myfile=fdopen(fd,"a");
    if(lockf(fd,FTLOCK,0)==0)
    fprintf(myfile,"%s n",name.nodename);
    sleep(20);
    lockf(fd,F_ULOCK,0);
    else printf("LOCKED n");
    fclose(myfile);
    close(fd);
    /ptmp is a GPFS filesystem.
    1/ if you run that code on 2 nodes within 20 seconds and /ptmp/p
    olo does not exist, it will write the nodename in /ptmp/polo on
    the first node you run it on and display LOCKED on the other nod
    e. This is the correct behavior.
    2/ if you run 2 instances of that code within 20 seconds on a si
    ngle node one will append nodename to /ptmp/polo the other will
    display LOCKED. This again is the correct behavior.
    3/ if /ptmp/polo already exists and you run that code on two
    different nodes within 20 seconds, they will not be aware of eac
    h other's lock. Both instances will append nodename to /ptmp/pol
    o. This is incorrect.
    4/ if you change the code slightly replacing the setting of the
    lock by :
    struct flock slock;
    slock.l_type=F_WRLCK;
    slock.l_whence=SEEK_SET;
    slock.l_start=0;
    slock.l_len=0;
    slock.l_sysid=0;
    slock.l_pid=0;
    slock.l_vfs=0;
    fcntl(fd,F_SETLK.&slock);
    and the unsetting of the lock by :
    slock.l_type=F_UNLCK;
    fcntl(fd,F_SETLK,&slock);
    Then everything runs fine.
    More surprising, the lockf version detects a lock set by the fcn
    tl version, and the fcntl version detects a lock set by the lock
    f version.

    PROBLEM SUMMARY:
    GPFS sometimes not honoring lockf calls.

    PROBLEM CONCLUSION:
    revoke handler was not traslating a request for range.end
    of "maximum possible file size" to an l_len of zero when
    checking for
    conflicting local lock

    ------

    APAR: IY19399 COMPID: 5765C3403 REL: 430
    ABSTRACT: CRASH WHEN USING SEND_FILE

    PROBLEM DESCRIPTION:
    System crashes when using send_file or other applications that
    use NBC private segments.

    PROBLEM CONCLUSION:
    Don't unpin a unpinned memory and don't pin an already pinned
    memory. When reusing memory from the freelist maintained by
    NBC for memory in private segments, reset the m_next to
    NULL. Otherwise this memory was getting pointed to in 2 lists
    and this was resulting in double freeing of memory. There
    were some uninitialized variables that was causing crash too.

    ------

    APAR: IY19400 COMPID: 5765C3403 REL: 430
    ABSTRACT: LP'S NOT REPORTED ON MON-CXMA

    PROBLEM DESCRIPTION:
    mon-cxma does not work on lp devices.

    PROBLEM CONCLUSION:
    There are some ODM searches that restrict the output to only
    tty's. These searches were widened to include printers.

    ------

    APAR: IY19408 COMPID: 5765C3403 REL: 430
    ABSTRACT: SLOW PERFORMANCE WHEN XCOPY FROM WIN9X TO FC

    PROBLEM DESCRIPTION:
    Slow perfromance when xcopy from win9x to FC

    PROBLEM SUMMARY:
    Slow performance when xcopy from win9x to FC

    PROBLEM CONCLUSION:
    Fix code so clients request bigger receive
    packet size.

    ------

    APAR: IY19443 COMPID: 5765C3403 REL: 430
    ABSTRACT: GXT6000P LINE SEGMENTS MISSING IN HRD MODE

    PROBLEM DESCRIPTION:
    With GXT6000P graphics Accelerator, on some
    models in Catia in HRD mode, some line
    segments may be incompletly rendered.
    This occurs after applying APAR 17172.

    PROBLEM SUMMARY:
    Fast concave going down the slow path.

    PROBLEM CONCLUSION:
    Sent fast concave down fast path.

    ------

    APAR: IY19449 COMPID: 5765C3403 REL: 430
    ABSTRACT: BOOT HANG AT LED 555 WHILE RESTORING FROM DVD

    PROBLEM DESCRIPTION:
    When a physical volume is associated with two volume groups
    (error scenario), the /etc/filesystems stanzas for one of
    the volume groups will be deleted during importvg -L.

    PROBLEM CONCLUSION:
    When the -L option is used with importvg, do not
    issue the imfs -x command.

    ------

    APAR: IY19450 COMPID: 5765C3403 REL: 430
    ABSTRACT: UNABLE TO VARYON MULTIPLE VGS WITH 128-BIT VGIDS

    PROBLEM DESCRIPTION:
    Customer will see error message:
    0516-004 varyonvg: The mapped file is currently being used
            by another process.
    during system IPL.

    PROBLEM CONCLUSION:
    Use all the characters of the Volume Group ID to generate the
    name of the mapped file.

    ------

    APAR: IY19473 COMPID: 5765C3403 REL: 430
    ABSTRACT: NURB Z BUFFERING DISABLED WITH FRAME BUFFER COMPARISON

    PROBLEM DESCRIPTION:
    Primitives will render inside out.

    PROBLEM CONCLUSION:
    Reverse the order of primitives for Tstrip and Qmesh.

    ------

    APAR: IY19474 COMPID: 5765C3403 REL: 430
    ABSTRACT: ATTRIBUTE STATE CAN GET CONFUSED

    PROBLEM DESCRIPTION:
    Anti aliased lines will ocassionally fail.

    PROBLEM CONCLUSION:
    Force the validation code to always validate
    when enabling anti aliased lines.

    ------

    APAR: IY19475 COMPID: 5765C3403 REL: 430
    ABSTRACT: PATTERN LEAVES EXCESS OUTSIDE POLY 3

    PROBLEM DESCRIPTION:
    The customer may see "trash" outside a primitive if the
    preceding primitive had a patterned interior.

    PROBLEM CONCLUSION:
    The macro that does our extent calculation was truncating
    decimal values without rounding up to account for the decimal.
    Once the decimal is taken into account and the rounding is
    done correctly, the "trash" no longer occurs.

    ------

    APAR: IY19476 COMPID: 5765D6100 REL: 220
    ABSTRACT: 384WAY: COL D/S:SWITCH RESOURCE TABLE COULD NOT BE LOADED

    PROBLEM DESCRIPTION:
    384 way: Col D/S: Switch resource table could not be loaded

    PROBLEM SUMMARY:
    When the virtual device csss is requested in a LoadLeveler
    network statement, LoadLeveler did not always detect the
    sucessful load of the switch table for the device properly

    PROBLEM CONCLUSION:
    LoadLeveler was modified to properly detect the successful
    load of the switch table associated with the csss device

    ------

    APAR: IY19478 COMPID: 5765D5100 REL: 320
    ABSTRACT: EM.DEFAULT LOG HAS WARNING MSG CONCERNING LIBSDR.A BEING LINKED

    PROBLEM DESCRIPTION:
    In /var/ha/log/em.default log the following warning msg appears
    Error!!! The SDR has detected that multiple copies of itself
    have been linked into this program. This happens if a library or
    the binary has been linked with a static copy of libSDR.a and
    shared libraries are used that reference the shared SDR library,
    libSDR.a. This program must be re-linked!
    The way EM uses the SDR library was changed in 3.1.1 causing
    this problem.

    PROBLEM SUMMARY:
    As a result of the process flow in haemd, when a call is
    made to SDROpenSession, the following messages are
    erroneously written to the em.default log in /var/ha/log/:
    Error!!! The SDR has detected that
    multiple copies of itself have been linked into this
    program.
    This happens if a library or the binary has been linked with
    a static copy of libSDR.a and shared libraries are used that
    reference the shared SDR library, libSDRs.a. This program
    must be re-linked!
    There is actually not an error condition in this situation,
    but SDRCloseSession has failed to clear an environment
    variable which causes the message to be written to the log.

    PROBLEM CONCLUSION:
    SDRCloseSession was modified to clear the environment
    variable which is used to track if multiple copies of the
    SDR have been linked into a program.

    ------

    APAR: IY19522 COMPID: 5765C3403 REL: 430
    ABSTRACT: CAPABILITY TO QUEUE OUTGOING IP PACKETS THROUGH FIREWALL HOOKS

    PROBLEM DESCRIPTION:
    There is no provision in 4.3 firewall hooks interface for
    the external function to queue the outgoing packets.

    PROBLEM CONCLUSION:
    The interface is enhanced so that a NULL return value of the
    mbuf will indicate to IP that the called function has decided
    to handle the packet itself. We don't have to do any more
    processing on that packet.

    ------

    APAR: IY19525 COMPID: 5765C3403 REL: 430
    ABSTRACT: ERROR ID MESSAGE FOR CFDDI_PORT NOT COMPLETE

    PROBLEM DESCRIPTION:
    Error Report Entry for CFDDI_PORT does not display the
    detail data, probable cause or Failure reason information.

    PROBLEM CONCLUSION:
    Rebuilt the driver to ensure that a new Error ID is generated.

    ------

    APAR: IY19526 COMPID: 5765C3403 REL: 430
    ABSTRACT: NEW VERSION OF PROE (REGISTERED TM OF PTC) HANGS ON GXT3000P

    PROBLEM DESCRIPTION:
    Applications can hang the GXT3000P graphics adapter if they
    render to a window that is positioned near the extents of
    X's signed short address space -32786, 32767 .

    PROBLEM CONCLUSION:
    Restrict win_origin to the range -32000 to 28000. This will
    generally guarantee that a OpenGL viewport will not attempt
    to render outside the -32768, 32767 coordinate range.

    ------

    APAR: IY19533 COMPID: 5765C3403 REL: 430
    ABSTRACT: SIMPLIFIED CHINESE DOES NOT WORK IN GRAPHIGS SINCE AIX433

    PROBLEM DESCRIPTION:
    The Simplified Chinese input method is not working correctly
    in graPHIGS in releases AIX433 and later.

    PROBLEM SUMMARY:
    Input method indicator text doesn't change. Pre-edit location
    is wrong. Attempts at input result in core dumps.

    PROBLEM CONCLUSION:
    Replace the code that generates the indicator text. Check
    the current input state to avoid or make up for out-of-order
    input method callbacks.

    ------

    APAR: IY19548 COMPID: 5765B9501 REL: 330
    ABSTRACT: PFS DOESN'T HANDLE LOCKF WITH MAXIMUM FILE SIZE (OR FTLOCK)

    PROBLEM DESCRIPTION:
    Using the following program
     include fcntl.h
     include stdio.h
     include sys/utsname.h
    main()
    struct utsname name;
    FILe *myfile;
    int fd;
    char *filename="/ptmp/polo";
    uname(&name);
    if((fd=open(filename,O_WRONLY]O_CREAT]O_APPEND,0666))==-1) retur
    n;
    else chmod(filename,0666);
    myfile=fdopen(fd,"a");
    if(lockf(fd,FTLOCK,0)==0)
    fprintf(myfile,"%s n",name.nodename);
    sleep(20);
    lockf(fd,F_ULOCK,0);
    else printf("LOCKED n");
    fclose(myfile);
    close(fd);
    /ptmp is a GPFS filesystem.
    1/ if you run that code on 2 nodes within 20 seconds and /ptmp/p
    olo does not exist, it will write the nodename in /ptmp/polo on
    the first node you run it on and display LOCKED on the other nod
    e. This is the correct behavior.
    2/ if you run 2 instances of that code within 20 seconds on a si
    ngle node one will append nodename to /ptmp/polo the other will
    display LOCKED. This again is the correct behavior.
    3/ if /ptmp/polo already exists and you run that code on two
    different nodes within 20 seconds, they will not be aware of eac
    h other's lock. Both instances will append nodename to /ptmp/pol
    o. This is incorrect.
    4/ if you change the code slightly replacing the setting of the
    lock by :
    struct flock slock;
    slock.l_type=F_WRLCK;
    slock.l_whence=SEEK_SET;
    slock.l_start=0;
    slock.l_len=0;
    slock.l_sysid=0;
    slock.l_pid=0;
    slock.l_vfs=0;
    fcntl(fd,F_SETLK.&slock);
    and the unsetting of the lock by :
    slock.l_type=F_UNLCK;
    fcntl(fd,F_SETLK,&slock);
    Then everything runs fine.
    More surprising, the lockf version detects a lock set by the fcn
    tl version, and the fcntl version detects a lock set by the lock
    f version.

    PROBLEM SUMMARY:
    GPFS sometimes not honoring lockf calls.

    PROBLEM CONCLUSION:
    revoke handler was not traslating a request for range.end
    of "maximum possible file size" to an l_len of zero when
    checking for
    conflicting local lock

    ------

    APAR: IY19549 COMPID: 5765B9501 REL: 330
    ABSTRACT: MMFSD HANGS ON SHUTDOWN

    PROBLEM DESCRIPTION:
    When shutting down GPFS sometimes mmfsd does not exit but is
    left hanging

    PROBLEM SUMMARY:
    GPFS termination deadlocked because of a lock conflict in
    AIX process termination logic freeing memory.

    PROBLEM CONCLUSION:
    Directly call the _exit routine instead of using the libc
    version.

    ------

    APAR: IY19554 COMPID: 5765C3403 REL: 430
    ABSTRACT: NIM SPOT CREATE NOT INSTALLING ALL FIXES IN LPP_SOURCE

    PROBLEM DESCRIPTION:
    When creating a NIM SPOT, not all of the bos.rte.* updates
    in the lpp_source are being installed.

    PROBLEM CONCLUSION:
    The search pattern in c_instspot was changed to include
    all bos.rte.* updates during the SPOT creation.

    ------

    APAR: IY19556 COMPID: 5765C3403 REL: 430
    ABSTRACT: OPENGL TESTCASE ABENDS WHEN CALLING GLCOLOR4UBVERTEX2FSUN

    PROBLEM DESCRIPTION:
    If an application calls any of the gl*SUN subroutines in
    either a run-time linker environment or when glXGetProcAddrARB
    has been used, then an abend can occur.

    PROBLEM CONCLUSION:
    When the GLX problems with the run-time linker were first
    found (APAR IX89740), all of the indirect context versions of
    the OpenGL subroutines had a check added to route the code
    path to the direct context version of the subroutines in
    a number of different situations. When the gl*SUN subroutines
    were added, these checks were not added to the indirect
    version of the subroutines. Instead, these subroutines call
    other OpenGL subroutines to have the work done. This became
    a problem when the front end of the direct context versions
    of the subroutines were rewritten in assembler, because the
    assembly code made some assumptions about which routine would
    be calling them which are no longer valid. These assumptions
    lead to the TOC register being corrupted, which eventually
    causes the abend. The fix for this problem is to add the
    missing checks in the indirect versions of the gl*SUN
    subroutines to avoid this situation.

    ------

    APAR: IY19561 COMPID: 5765D6100 REL: 220
    ABSTRACT: XLOADL: JOBS, MACHINES NOT ALIGNED WITH HEADERS

    PROBLEM DESCRIPTION:
    xloadl: jobs, machines not aligned with headers

    PROBLEM SUMMARY:
    A fix has been made to prevent other X Resources from over
    riding the fontList used for the LoadLeveler GUI.

    PROBLEM CONCLUSION:
    The Xloadl and Xloadl_so resource files have been changed so
    that they specify the resource name of the application and
    take precedence over other resources.

    ------

    APAR: IY19562 COMPID: 5765B9501 REL: 320
    ABSTRACT: ASSERT IOSTATE == BUFFERNOIO BUFDESC.C, LINE 1557

    PROBLEM DESCRIPTION:
    assert ioState == BufferNoIO bufdesc.C, line 1557

    PROBLEM SUMMARY:
    GPFS self check logic terminated GPFS with an error in
    bufdesc.C line 1557

    PROBLEM CONCLUSION:
    In exit handling for writeBuffer, turn the buffer ioState
    back to BufferNoIO
    before calling endWrite which turns off the writeInProgress
    flag
    and wake up other threads that will assert that ioState ==
    BufferNoIO.

    ------

    APAR: IY19563 COMPID: 5765B9501 REL: 330
    ABSTRACT: ASSERT IOSTATE == BUFFERNOIO BUFDESC.C, LINE 1557

    PROBLEM DESCRIPTION:
    assert ioState == BufferNoIO bufdesc.C, line 1557

    PROBLEM SUMMARY:
    GPFS self check logic terminated GPFS with an error in
    bufdesc.C line 1557

    PROBLEM CONCLUSION:
    In exit handling for writeBuffer, turn the buffer ioState
    back to BufferNoIO
    before calling endWrite which turns off the writeInProgress
    flag
    and wake up other threads that will assert that ioState ==
    BufferNoIO.

    ------

    APAR: IY19564 COMPID: 5765B9500 REL: 130
    ABSTRACT: MMCRFS CORRUPTS MMSDRFS FILE IF V3 FILE EXISTS

    PROBLEM DESCRIPTION:
    mmcrfs corrupts mmsdrfs file if v3 file exists

    PROBLEM SUMMARY:
    mmcrfs fails if you have a file named v3 in
    the working directory.

    PROBLEM CONCLUSION:
    don't use the current working directory for
    a temporary file.

    ------

    APAR: IY19565 COMPID: 5765B9500 REL: 140
    ABSTRACT: MMCRFS CORRUPTS MMSDRFS FILE IF V3 FILE EXISTS

    PROBLEM DESCRIPTION:
    mmcrfs corrupts mmsdrfs file if v3 file exists

    PROBLEM SUMMARY:
    mmcrfs fails if you have a file named v3
    in the working directory.

    PROBLEM CONCLUSION:
    don't use the current working directory
    for a temporary file.

    ------

    APAR: IY19566 COMPID: 5765B9501 REL: 320
    ABSTRACT: LX: GPFS DAEMON CORE DUMPED (SIG 11) ON NODE C5N42 DURING DUMP

    PROBLEM DESCRIPTION:
    LX: GPFS daemon core dumped (SIG 11) on node c5n42 during dump

    PROBLEM SUMMARY:
    Clean up a message

    PROBLEM CONCLUSION:
    use objtype_tostring (not objtype_id2name) in a dprintf.

    ------

    APAR: IY19567 COMPID: 5765B9501 REL: 330
    ABSTRACT: LX: GPFS DAEMON CORE DUMPED (SIG 11) ON NODE C5N42 DURING DUMP

    PROBLEM DESCRIPTION:
    LX: GPFS daemon core dumped (SIG 11) on node c5n42 during dump

    PROBLEM SUMMARY:
    Clean up a message

    PROBLEM CONCLUSION:
    use objtype_tostring (not objtype_id2name) in a dprintf.

    ------

    APAR: IY19568 COMPID: 5765B9501 REL: 320
    ABSTRACT: NO ADMIN SLEEP VARIABLES, TRCFMT QUOTA

    PROBLEM DESCRIPTION:
    No admin sleep variables, trcfmt quota

    PROBLEM SUMMARY:
    Fix some trace formatting

    ------

    APAR: IY19569 COMPID: 5765B9501 REL: 330
    ABSTRACT: NO ADMIN SLEEP VARIABLES, TRCFMT QUOTA

    PROBLEM DESCRIPTION:
    No admin sleep variables, trcfmt quota

    PROBLEM SUMMARY:
    Fix some trace formatting

    ------

    APAR: IY19570 COMPID: 5765B9501 REL: 320
    ABSTRACT: DEADLOCK BETWEEN MMFSCK AND MMCHECKQUOTA

    PROBLEM DESCRIPTION:
    Deadlock between mmfsck and mmcheckquota

    PROBLEM SUMMARY:
    Running mmfsck followed by mmcheckquota without allowing the
    mmfsck to complete caused a deadlock.

    PROBLEM CONCLUSION:
    Do not hold MgrMutex in permission check since this could
    block
    the thread waiting for conflict command to finish.

    ------

    APAR: IY19571 COMPID: 5765B9501 REL: 330
    ABSTRACT: DEADLOCK BETWEEN MMFSCK AND MMCHECKQUOTA

    PROBLEM DESCRIPTION:
    Deadlock between mmfsck and mmcheckquota

    PROBLEM SUMMARY:
    Running mmfsck followed by mmcheckquota without allowing the
    mmfsck to complete caused a deadlock.

    PROBLEM CONCLUSION:
    Do not hold MgrMutex in permission check since this could
    block
    the thread waiting for conflict command to finish.

    ------

    APAR: IY19572 COMPID: 5765B9501 REL: 320
    ABSTRACT: LONG WAITERS QUOTA ENTRY REASON 'WAIT FOR GETSHAREISACTIVE

    PROBLEM DESCRIPTION:
    Long waiters Quota entry reason 'wait for getShareIsActive

    PROBLEM SUMMARY:
    Deadlock running stress load on a filesystem with quotas
    enabled.

    PROBLEM CONCLUSION:
    Fix serialization error in quota handling

    ------

    APAR: IY19573 COMPID: 5765B9501 REL: 330
    ABSTRACT: LONG WAITERS QUOTA ENTRY REASON 'WAIT FOR GETSHAREISACTIVE

    PROBLEM DESCRIPTION:
    Long waiters Quota entry reason 'wait for getShareIsActive

    PROBLEM SUMMARY:
    Deadlock running stress load on a filesystem with quotas
    enabled.

    PROBLEM CONCLUSION:
    Fix serialization error in quota handling

    ------

    APAR: IY19574 COMPID: 5765B9500 REL: 130
    ABSTRACT: MMREPQUOTA TRIES TO REPORT ON FILE SYSTEMS IN A DIFFERENT NODE

    PROBLEM DESCRIPTION:
    mmrepquota tries to report on file systems in a different node

    PROBLEM SUMMARY:
    mmrepquota -a -u -g issues error messages when
    it tries to report on file systems that belong to a different
    nodeset from the nodeset that the command is issued from.

    PROBLEM CONCLUSION:
    correct internal handling of this command
    from another nodeset.

    ------

    APAR: IY19575 COMPID: 5765B9500 REL: 140
    ABSTRACT: MMREPQUOTA TRIES TO REPORT ON FILE SYSTEMS IN A DIFFERENT NODE

    PROBLEM DESCRIPTION:
    mmrepquota tries to report on file systems in a different node

    PROBLEM SUMMARY:
    mmrepquota -a -u -g issues error messages
    when it tries to report on file systems that belong to a
    different nodeset from the nodeset that the command is issued
    from.

    PROBLEM CONCLUSION:
    correct internal handling of this command
    from another nodeset.

    ------

    APAR: IY19576 COMPID: 5765B9501 REL: 330
    ABSTRACT: FENCING IN SINGLENODEQUORUM STATE DOESN'T OPERATE AS DESIGNED

    PROBLEM DESCRIPTION:
    Fencing in SingleNodeQuorum State doesn't operate as designed

    PROBLEM SUMMARY:
    In two node quorum systems with HACMP, fencing was not
    occurring if no log replay was required.

    PROBLEM CONCLUSION:
    Always fence out the failed node in log recovery code.

    ------

    APAR: IY19578 COMPID: 5765B9501 REL: 330
    ABSTRACT: GPFS SP 19000 SYNC HANDLER WAIT FOR INODEFLUSHFLAG

    PROBLEM DESCRIPTION:
    GPFS SP 19000 sync handler wait for inodeFlushFlag

    PROBLEM SUMMARY:
    Potential deadlock discovered in development

    PROBLEM CONCLUSION:
    Correct order of getting the ServerAlmMutex and the
    SGDescMutex in
    AllocMgrServer::setAllocDisksBitmap
    FileAllocator::allocOneBlock
    FileAllocator::alloc RoundRobinStripeMethod::excludeDisk

    ------

    APAR: IY19579 COMPID: 5765B9501 REL: 330
    ABSTRACT: UMASK NOT HONORED THROUGH NFS

    PROBLEM DESCRIPTION:
    umask not honored through nfs

    PROBLEM SUMMARY:
    umask not being honored from nfs.

    PROBLEM CONCLUSION:
    let lfs and nfs set the file or directory mo
    de masked with the umask.

    ------

    APAR: IY19580 COMPID: 5765B9501 REL: 330
    ABSTRACT: GPFS 1.4 SP DEADLOCK ON C271N05 DURING MOUNT

    PROBLEM DESCRIPTION:
    GPFS 1.4 SP deadlock on c271n05 during mount

    PROBLEM SUMMARY:
    Deadlock in mount running GPFS 1.4

    PROBLEM CONCLUSION:
    Release the SGDescMutex while calling
    get_free_info_block_counts to
    prevent deadlock with AlmServerMutex.

    ------

    APAR: IY19582 COMPID: 5765C3403 REL: 430
    ABSTRACT: GETHOSTBYADDR RETURNS INCORRECTLY WHEN AN IPADDRESS IN

    PROBLEM DESCRIPTION:
    when try to resolve the loopback address 127.0.0.1
    by using /etc/hosts, it returns incorrect address.
    For example:
    export NSORDER=local
    host 127.0.0.1
    loopback is 127.0.0.0 Aliases: localhost
    Basically if there is one number contain 0 in the
    ipaddress, any number after 0 will be returned as 0.

    LOCAL FIX:
    trying to resolve the loopback from dns before through
    /etc/hosts file by create the /etc/netsvc.conf file and
    contain the following entry:
    hosts=bind,nis,nisp,local

    PROBLEM SUMMARY:
    ***************************************************************
    *USERS AFFECTED: *
    * All users with the following filesets at these levels *
    * bos.adt.prof 4.3.3.53
    * bos.rte.libc 4.3.3.53**************************************
    ***************************************************************
    *PROBLEM DESCRIPTION: *
    * when try to resolve the loopback address 127.0.0.1 by
    * example: ==> export NSORDER=local ==> host 127.0.0.1
    * loopback is 127.0.0.0 Aliases: localhost Basically if
    * using /etc/hosts, it returns incorrect address. For
    * example: ==> export NSORDER=local ==> host 127.0.0.1
    * loopback is 127.0.0.0 Aliases: localhost Basically if*******
    * there is one number contain 0 in the ipaddress, any *
    * number after 0 will be returned as 0.
    ***************************************************************
    *RECOMMENDATION: *
    * Apply apar IY19582
    ***************************************************************

    PROBLEM CONCLUSION:
    The problem is fixed by making sure
    the ipaddress is correctly copied
    if one of the number is 0.

    TEMPORARY FIX:
    edit /etc/netsvc.conf file, add:
    hosts=bind,nis,nisp,local

    ------

    APAR: IY19601 COMPID: 5765D5100 REL: 320
    ABSTRACT: WRITE-FILE 0716 SHOULD BE INCLUDED IN UCODE DIR

    PROBLEM DESCRIPTION:
    write-file 0716 should be included in ucode dir

    PROBLEM SUMMARY:
    The work around is to manually power off / on the swtich

    PROBLEM CONCLUSION:
    On some colony switches, after a power cycle, the chips
    would report inconsistant BIST values. If this is detected
    by the supervisor, up to 3 additonal reset cyles will be
    attempted.

    ------

    APAR: IY19603 COMPID: 5765B9501 REL: 320
    ABSTRACT: AIX 51 STRESSED GPFS SYSTEMS COMPLAIN ABOUT 'TOO BIG'

    PROBLEM DESCRIPTION:
    AIX 51 stressed gpfs systems complain about 'too big'

    PROBLEM SUMMARY:
    339359 fixes a memory leak in the token manager. When some
    node asked another node(s) to downgrade its tokens,
    sometimes the second node would relinquish the token
    entirely (down to NL). When this relinquish was processed on
    the TM, it just set the client holder structure to say the
    node had an NL token, but should have deleted the structure.
    Eventually, there would be no clients with any non-NL token
    for a particular inode, but the token structure was never
    reclaimed in the TM because of these NL client structures.
    Eventually the TM would notice it was running out of TM
    space and try to get nodes to reinquish their tokens, but
    the unused ones would still lay around because no client
    node knew about these previously relinquished token
    structures. ENOMEM would eventually be returned from TM
    requests and odd things would happen to applications.

    PROBLEM CONCLUSION:
    A memory leak in Token Manager code caused application
    failues with error "ENOMEM".

    ------

    APAR: IY19604 COMPID: 5765B9501 REL: 320
    ABSTRACT: ASSERT SPRUNG IN_CPY->COPYSETS J .NODE == NODE, FILE HANDLEREQ

    PROBLEM DESCRIPTION:
    Assert sprung in_cpy->Copysets j .Node == node, file HandleReq

    PROBLEM SUMMARY:
    The assertion was incorrect and was replaced by a test.

    PROBLEM CONCLUSION:
    In Token manager code when token manager was trying to
    reclaim storage, there was a race in clearing up byte range
    tokens which showed that the assert that went off was an
    incorrect one.

    ------

    APAR: IY19605 COMPID: 5765B9501 REL: 330
    ABSTRACT: ASSERT SPRUNG IN_CPY->COPYSETS J .NODE == NODE, FILE HANDLEREQ

    PROBLEM DESCRIPTION:
    Assert sprung in_cpy->Copysets j .Node == node, file HandleReq

    PROBLEM SUMMARY:
    The assertion was incorrect and was replaced by a test.

    PROBLEM CONCLUSION:
    In Token manager code when token manager was trying to
    reclaim storage, there was a race in clearing up byte range
    tokens which showed that the assert that went off was an
    incorrect one.

    ------

    APAR: IY19607 COMPID: 5765B9501 REL: 330
    ABSTRACT: AIX 51 STRESSED GPFS SYSTEMS COMPLAIN ABOUT 'TOO BIG'

    PROBLEM DESCRIPTION:
    AIX 51 stressed gpfs systems complain about 'too big'

    PROBLEM SUMMARY:
    339359 fixes a memory leak in the token manager. When some
    node asked another node(s) to downgrade its tokens,
    sometimes the second node would relinquish the token
    entirely (down to NL). When this relinquish was processed on
    the TM, it just set the client holder structure to say the
    node had an NL token, but should have deleted the structure.
    Eventually, there would be no clients with any non-NL token
    for a particular inode, but the token structure was never
    reclaimed in the TM because of these NL client structures.
    Eventually the TM would notice it was running out of TM
    space and try to get nodes to reinquish their tokens, but
    the unused ones would still lay around because no client
    node knew about these previously relinquished token
    structures. ENOMEM would eventually be returned from TM
    requests and odd things would happen to applications.

    PROBLEM CONCLUSION:
    A memory leak in Token Manager code caused application
    failues with error "ENOMEM".

    ------

    APAR: IY19611 COMPID: 5765C3403 REL: 430
    ABSTRACT: PRESERVE DOSATTS W/ ACL-INHERITANCE

    PROBLEM DESCRIPTION:
    With acl_inheritance=1 and dosattrmapping=1,
    DOS file-attributes were not being properly copied
    when copying files from PC to the cifsServer.
    Instead, the R-A-S-H DOSattributes were being cleared
    on the target (copied) file. (Note: XCOPY clears
    the ReadOnly attribute by default, and both XCOPY
    and WinExplorer set the Archive bit by default.)

    LOCAL FIX:
    Manually set the desired DosAttributes after copying
    the file, from the PC.

    PROBLEM SUMMARY:
    This fix will be available on CIFS version 2.1.1.x and CIFS
    3.1.0.x

    PROBLEM CONCLUSION:
    This fix is for both CIFS 2.1.1.x and 3.1.0.x

    TEMPORARY FIX:
    efix available from L3CIFS,195.

    ------

    APAR: IY19630 COMPID: 5765C3403 REL: 430
    ABSTRACT: PKCS11 RSA OPS WITH LEEDSLITE FAILURES

    PROBLEM DESCRIPTION:
    PKCS11 RSA ops with leedslite device driver fail

    PROBLEM CONCLUSION:
    fix removed defines in ica_api.h

    ------

    APAR: IY19632 COMPID: 5765C3403 REL: 430
    ABSTRACT: RUN MULTIPLE RESTVG/SAVEVG COMMANDS AT THE SAME TIME

    PROBLEM DESCRIPTION:
    User cannot execute multiple savevg/restvg commands at the
    same time because files used for processing have the same
    name and thus each command overwrites each others files.

    PROBLEM SUMMARY:
    Could not run multiple savevg/restvg at the same time because
    they used generic processing file names that would overwrite
    each other in the process of saving or restoring a user
    volume group.

    PROBLEM CONCLUSION:
    Needed to make the processing files unique to each execution
    of restvg/savevg so that the files did not get overwritten.

    TEMPORARY FIX:
    Do not run more than one savevg or restvg at the same time.

    ------

    APAR: IY19643 COMPID: 5765D5100 REL: 320
    ABSTRACT: 45800001 CHECKSTOPS

    PROBLEM DESCRIPTION:
    45800001 checkstops

    PROBLEM SUMMARY:
    Reading sram on TBIC data parity errors via the snap was
    causing the node to chkstp.

    PROBLEM CONCLUSION:
    Toke out sram reads done from the snap on TBIC data parity
    errors

    ------

    APAR: IY19666 COMPID: 5765D5101 REL: 120
    ABSTRACT: HATS: RACE CONDITION IN MAIN THREAD OF HATS

    PROBLEM DESCRIPTION:
    This is a race condition between the main thread in HATS which
    is trying to shut down the DMS timer, and the HB thread, which
    is trying to reset the timer as usual. Because of the weird way
    in which dms_stop() and dms_free() call each other recursively,
    the race condition resulted (apparently) in dms_free()
    trying to free an already-freed timer.

    PROBLEM SUMMARY:
    A race condition in the RSCT Topology Services daemon,
    when invoking the "Dead Man Switch" kernel extension,
    sometimes causes the AIX operating system to crash.
    The problem occurs when shutting down the "Dead Man
    Switch" facility, which happens when the daemon is
    about to terminate.
    The stack trace of the AIX crash often includes the
    dms_free() function.
    The "Dead Man Switch" kernel extension is only used
    in an HACMP/ES environment, not in PSSP.

    PROBLEM CONCLUSION:
    A fix was introduced into the Topology Services daemon
    and into the "Dead Man Switch" kernel extension to prevent
    the race condition from happening. With the fix,
    AIX should no longer crash when Topology Services is
    terminated in an HACMP/ES environment.
    Important: the node needs to be rebooted in order
    to pick up the updated version of the kernel
    extension.

    ------

    APAR: IY19747 COMPID: 5765C3403 REL: 430
    ABSTRACT: LINUX: ADD SYMLINKS TO /USR/BIN FROM /USR/OPT/PERL5/BIN

    PROBLEM DESCRIPTION:
    Need to create sym links for the executables in
    /usr/opt/perl5/bin to /usr/bin

    PROBLEM CONCLUSION:
    Add sym links for the executables in /usr/opt/perl5/bin
    to /usr/bin for perl.rte.

    ------

    APAR: IY19772 COMPID: 5765B9501 REL: 320
    ABSTRACT: READ/WRITE NOT RETURNING EIO ON BROKEN ADDRS.

    PROBLEM DESCRIPTION:
    Read/Write not returning EIO on broken addrs

    PROBLEM SUMMARY:
    GPFS returned a block of zeros and a correct return code
    when encountering certain media errors. It should return an
    I/O error and no data.

    PROBLEM CONCLUSION:
    Correct handling of broken disk addresse

    ------

    APAR: IY19778 COMPID: 5765C3403 REL: 430
    ABSTRACT: JFSERRLOGGING CODE SHOULD CORRECTLY HANDLE MULTI-SEG .INDIRECT

    PROBLEM DESCRIPTION:
    In some cases of file system corruption, information needed
    to analyze the system crash is not included in the dump.

    PROBLEM CONCLUSION:
    Enable JFS Smart Dump to include extended .indirect segments.

    ------

    APAR: IY19799 COMPID: 5765C3403 REL: 430
    ABSTRACT: CCET: FREEIBLK() BACKT FAULT ON MOUNT INODE

    PROBLEM DESCRIPTION:
    System crash in v_jfscorruption() via v_freeiblk() or
    v_allociblk(), due to an apparent attempt to free an already
    free .indirect page. Free chain will be intact and all
    .indirect structures will appear consistent.

    PROBLEM SUMMARY:
    ***************************************************************
    *USERS AFFECTED: *
    * All users with the following filesets at these levels *
    * bos.mp 4.3.3.29s will appear consistent.
    * bos.mp 4.3.3.30
    * bos.mp 4.3.3.31
    * bos.mp 4.3.3.32
    * bos.mp 4.3.3.50
    * bos.mp 4.3.3.51
    * bos.mp 4.3.3.53
    * bos.up 4.3.3.29
    * bos.up 4.3.3.30
    * bos.up 4.3.3.31
    * bos.up 4.3.3.32
    * bos.up 4.3.3.50
    * bos.up 4.3.3.51
    * bos.up 4.3.3.53
    ***************************************************************
    *PROBLEM DESCRIPTION: *
    * System crash in v_jfscorruption() via v_freeiblk() or
    * v_allociblk(), due to an apparent attempt to free an
    * already free .indirect page. Free chain will be intact
    * and all
    * .indirect structures will appear consistent.
    ***************************************************************
    *RECOMMENDATION: *
    * Apply apar IY19799
    ***************************************************************

    PROBLEM CONCLUSION:
    Touch in the mount inode at the beginning of the functions
    above, in order to prevent a backtrack pagefault after
    updating the .indirect free list.

    ------

    APAR: IY19802 COMPID: 5765D5100 REL: 320
    ABSTRACT: UCODE REASSEMBLY TIMEOUT TOO SHORT FOR SOME ENVIRONMENTS

    PROBLEM DESCRIPTION:
    In some environments, the default TB3MX ucode reassembly
    timeout (24 msec) is too short, causing datagrams to be
    artificially dropped. Testing at one such location suggests
    that the timeout should be increased to 96 msec.

    PROBLEM SUMMARY:
    Microcode reassembly timeout algorithm was too aggressive,
    causing messages to be dropped prematurely.

    PROBLEM CONCLUSION:
    Extend the timeout.

    ------

    APAR: IY19852 COMPID: 5765D6100 REL: 220
    ABSTRACT: STARTUP FAILURE WITH BOS.MP.4.3.2.X AND 4.3.3.51 INSTALLED

    PROBLEM DESCRIPTION:
    LoadL fails to start with bos.mp 4.3.2.x and bos.mp
    4.3.3.51 kernel installed.

    LOCAL FIX:
    Force overwrite install of bos.mp.4.3.3.0 and reinstall
    bos.mp.4.3.3.51 on top to clean up ODM

    PROBLEM SUMMARY:
    If a LoadLeveler system, that is at 4.3.3 for bos.mp or
    bos.up, got there by upgrading from 4.3.2, the LoadLeveler
    Startd Daemon will use the wrong algorithm for trying to
    determine the amount of virtual memory that exists on the
    machine. The result of that is that LoadLeveler will
    probably think that the amount of virtual memory on the
    machine is 0 and the Negotiator will, therefore, not assign
    any jobs to run on that machine.

    PROBLEM CONCLUSION:
    LoadLeveler generates an lslpp system call in the Startd
    Daemon and greps for a release string in the results. The
    lslpp command was changed to use the -L option so that it
    will see only the current level, in the lslpp output,
    instead of the history of all the levels that have been
    installed on the node.

    ------

    APAR: IY19880 COMPID: 5765C3403 REL: 430
    ABSTRACT: ALLOW TCP_MAXSEG OPTION TO SET MSS

    PROBLEM DESCRIPTION:
    IP packets will get fragmented if you add new headers to the
    TCP packet.

    PROBLEM CONCLUSION:
    Allow MSS (TCP Maximum Segment Size) to modified through the
    TCP_MAXSEG socket option.

    ------

    APAR: IY19928 COMPID: 5765C3403 REL: 430
    ABSTRACT: NEED A "CIFSSNAP" SCRIPT FOR TESTCASE-DATA

    PROBLEM DESCRIPTION:
    n/a (New Feature) adding a new cifs command to gather service
     data.

    PROBLEM CONCLUSION:
    added new files need for the cifsSnap command.

    ------

    APAR: IY19929 COMPID: 5765C3403 REL: 430
    ABSTRACT: ERROR HANDLING FOR 4 NEW JAPANESE CHARACTERS

    PROBLEM DESCRIPTION:
    Server does not check if the entered value for double_byte_char
    is invalid.

    PROBLEM CONCLUSION:
    Error handling is done by checking if the entered parameter
    value contains the right characters both at the server start
    and also while using net command

    ------

    APAR: IY19930 COMPID: 5765C3403 REL: 430
    ABSTRACT: UNBALACENCED CPU UTILICATION OF NICE'D PROCESSES

    PROBLEM DESCRIPTION:
    A cpu may be idle at the same time another cpu has runnable
    threads waiting.

    PROBLEM CONCLUSION:
    Modify idle load balancing so that a non-empty run queue
    can never carry a zero weight.

    ------

    APAR: IY19936 COMPID: 5765D5100 REL: 320
    ABSTRACT: 384WAY:COLONY DIAGS FAIL

    PROBLEM DESCRIPTION:
    384way: colony diags fail

    PROBLEM SUMMARY:
    The SP Switch2 adapter has a NBA_LHS_ACCUMULATOR_REG_ADDR
    register which accumulates errors (indications that an error
    has occurred). The adapter diagnostics did not account for a
    bit being turned on, that should not cause the diags to
    fail. By simply masking off the bit, the problem is solved.

    PROBLEM CONCLUSION:
    The register test was modified to mask off a bit that can
    turn ON for no reason (no consequence) in the
    NBA_LHS_ACCUMULATOR_REG_ADDR register.

    ------

    APAR: IY19938 COMPID: 5765D5100 REL: 320
    ABSTRACT: DIAGS CABLE TEST NEED TO WAIT ON BACKGROUND S1TERM

    PROBLEM DESCRIPTION:
    diags cable test need to wait on background s1term

    PROBLEM SUMMARY:
    The cable_test tool uses s1term sessions from the CWS to
    read and write registers on the switch. The procedure which
    initiated the s1terms was not using the 'wait' function to
    verify that the s1term closed properly. The tool would just
    sleep for a second, and assume that everything was OK and
    that the next s1term could be opened. However, if the s1term
    did not close and the cable_test issued another OPEN, the
    hardware monitor would kill the process trying to open the
    s1term. This resulted in the cable_test being stopped.

    PROBLEM CONCLUSION:
    The change introduced by this Defect was to simply to 'wait'
    on the process id which is the s1term session. The tool used
    to sleep for a second, and assume the s1term process was
    complete.

    ------

    APAR: IY19991 COMPID: 5765B9501 REL: 320
    ABSTRACT: MMDF SHOW AND MARK SUSPENDED DISKS

    PROBLEM DESCRIPTION:
    mmdf show and mark suspended disks

    PROBLEM SUMMARY:
    Have mmdf without -q option show how much space is free on a
    disk that is suspended

    ------

    APAR: IY19992 COMPID: 5765B9501 REL: 330
    ABSTRACT: MMDF SHOW AND MARK SUSPENDED DISKS

    PROBLEM DESCRIPTION:
    mmdf show and mark suspended disks

    PROBLEM SUMMARY:
    Have mmdf without -q option show how much space is free on a
    disk that is suspended

    ------

    APAR: IY19993 COMPID: 5765B9501 REL: 320
    ABSTRACT: GPFS SP: DIRTYINDBUFS == 0 S/SYNC.C, LINE 2903

    PROBLEM DESCRIPTION:
    GPFS SP: dirtyIndBufs == 0 s/sync.C, line 2903

    PROBLEM SUMMARY:
    GPFS self check logic detected an error at sync.C line 2903.

    PROBLEM CONCLUSION:
    When creating a directory or symlink, keep the byte-range
    lock until finished initializing the inode.

    ------

    APAR: IY19994 COMPID: 5765B9501 REL: 330
    ABSTRACT: GPFS SP: DIRTYINDBUFS == 0 S/SYNC.C, LINE 2903

    PROBLEM DESCRIPTION:
    GPFS SP: dirtyIndBufs == 0 s/sync.C, line 2903

    PROBLEM SUMMARY:
    GPFS self check logic detected an error at sync.C line 2903.

    PROBLEM CONCLUSION:
    When creating a directory or symlink, keep the byte-range
    lock until finished initializing the inode.

    ------

    APAR: IY19995 COMPID: 5765B9501 REL: 320
    ABSTRACT: HANGS IN FCNTL LOCKING DUE TO SYNCHRONIZATION PROBLEMS

    PROBLEM DESCRIPTION:
    Hangs in fcntl locking due to synchronization problems

    PROBLEM SUMMARY:
    Deadlock between multiple applications using fcntl locks

    PROBLEM CONCLUSION:
    Fix hangs in fcntl locking due to synchronization problems
    caused by the
    removal of the vnode lock. Use the revoke lock to prevent a
    sleep element
    from being queued by a locking thread at the same time that
    an unlock
    thread is scanning for sleep elements to post.

    ------

    APAR: IY19996 COMPID: 5765B9501 REL: 330
    ABSTRACT: HANGS IN FCNTL LOCKING DUE TO SYNCHRONIZATION PROBLEMS

    PROBLEM DESCRIPTION:
    Hangs in fcntl locking due to synchronization problems

    PROBLEM SUMMARY:
    Deadlock between multiple applications using fcntl locks

    PROBLEM CONCLUSION:
    Fix hangs in fcntl locking due to synchronization problems
    caused by the
    removal of the vnode lock. Use the revoke lock to prevent a
    sleep element
    from being queued by a locking thread at the same time that
    an unlock
    thread is scanning for sleep elements to post.

    ------

    APAR: IY19997 COMPID: 5765B9501 REL: 320
    ABSTRACT: QUOTA PERFORMANCE

    PROBLEM DESCRIPTION:
    quota performance

    PROBLEM SUMMARY:
    Performance improvements in quota handling

    ------

    APAR: IY19998 COMPID: 5765B9501 REL: 330
    ABSTRACT: QUOTA PERFORMANCE

    PROBLEM DESCRIPTION:
    quota performance

    PROBLEM SUMMARY:
    Performance improvements in quota handling

    ------

    APAR: IY19999 COMPID: 5765B9501 REL: 320
    ABSTRACT: TOKEN MGR KEEPING TOKENS WITH ONLY NL HOLDERS

    PROBLEM DESCRIPTION:
    Token mgr keeping tokens with only NL holders

    PROBLEM SUMMARY:
    Token manager was retaining unnecessary storage when locks
    were freed

    PROBLEM CONCLUSION:
    Release storage for tokens in NL state.

    ------

    APAR: IY20000 COMPID: 5765B9501 REL: 330
    ABSTRACT: TOKEN MGR KEEPING TOKENS WITH ONLY NL HOLDERS

    PROBLEM DESCRIPTION:
    Token mgr keeping tokens with only NL holders

    PROBLEM SUMMARY:
    Token manager was retaining unnecessary storage when locks
    were freed

    PROBLEM CONCLUSION:
    Release storage for tokens in NL state.

    ------

    APAR: IY20001 COMPID: 5765B9501 REL: 330
    ABSTRACT: READ/WRITE NOT RETURNING EIO ON BROKEN ADDRS.

    PROBLEM DESCRIPTION:
    Read/Write not returning EIO on broken addrs

    PROBLEM SUMMARY:
    GPFS returned a block of zeros and a correct return code
    when encountering certain media errors. It should return an
    I/O error and no data.

    PROBLEM CONCLUSION:
    Correct handling of broken disk addresse

    ------

    APAR: IY20086 COMPID: 5765D5100 REL: 320
    ABSTRACT: CORRECT CALL TO HAL_STATS IN MPCI

    PROBLEM DESCRIPTION:
    Correct call to hal_stats in MPCI

    PROBLEM SUMMARY:
    Window packet and byte counts from estat were not being
    updated correctly by MPCI.

    PROBLEM CONCLUSION:
    MPCI was calling hal_stats incorrectly. Call has been
    fixed.

    ------

    APAR: IY20121 COMPID: 5765D5100 REL: 320
    ABSTRACT: ADAPTREC SNAL DOES NOT ACCEPT -A -P FLAGS ON 1 PLANE

    PROBLEM DESCRIPTION:
    adaptrec snal does not accept -a -p flags on 1 plane

    PROBLEM SUMMARY:
    This APAR forces default behavior based on the
    version/release level from the odm.

    PROBLEM CONCLUSION:
    This APAR forces default behavior based on the
    version/release level from the odm.

    ------

    APAR: IY20257 COMPID: 5765D5100 REL: 320
    ABSTRACT: SP ATTACH OF MODEL 6H0

    PROBLEM DESCRIPTION:
    sp attach of model 6H0

    PROBLEM SUMMARY:
    Support required for H60

    PROBLEM CONCLUSION:
    Added support for H60

    ------

    APAR: IY20341 COMPID: 5765E6100 REL: 510
    ABSTRACT: INCORRECT DFA READS ON THE GXT120P/130P GRAPHIC ADAPTERS.

    PROBLEM DESCRIPTION:
    Incorrect DFA reads on the GXT120P/130P graphic adapters.

    PROBLEM CONCLUSION:
    Read a word of data using a DFA read after all raster
    operations.

    ------

    APAR: IY20372 COMPID: 5765E6100 REL: 510
    ABSTRACT: 64BIT ZH_CN SUPPORT

    PROBLEM DESCRIPTION:
    Can not input chinese in 64bit applications

    PROBLEM CONCLUSION:
    Add new code for supporting 64bit zh_CN local

    ------

    APAR: IY20373 COMPID: 5765E6100 REL: 510
    ABSTRACT: 64BIT ZH_CN SUPPORT

    PROBLEM DESCRIPTION:
    can not input CHinese in 64bit application

    PROBLEM CONCLUSION:
    add new code to support 64bit ZH_CN

    ------

    APAR: IY20374 COMPID: 5765E6100 REL: 510
    ABSTRACT: IME (ABC, NM) PROBLEM IN 64BIT

    PROBLEM DESCRIPTION:
    NM ABC UNIVERSAL input method do not work correctly

    PROBLEM CONCLUSION:
    change ime_nm.c ime_abc.c files ...

    ------

    APAR: IY20403 COMPID: 5765B9501 REL: 330
    ABSTRACT: ASSERT (QUOTAFLAGS & 0X0004)== SPRUNG FILE QUOCLT.C LINE 851

    PROBLEM DESCRIPTION:
    assert (quotaflags & 0x0004)==0 sprung file quoclt.x line 851

    PROBLEM SUMMARY:
    The problem (code assert) occurs when a datashipping
    application needs to expand the last block, before entering
    data shipping mode, on a quota enabled file system. The fix
    added to the code is to expand the block as root, so that
    allocation is not enforced in this case.

    PROBLEM CONCLUSION:
    When datashipping has to expand the last block, do it as if
    it is root for purposes of quota.

    ------

    APAR: IY20405 COMPID: 5765B9501 REL: 330
    ABSTRACT: LOCKCTL NOT CHECKING FOR NFS3 MOUNTS

    PROBLEM DESCRIPTION:
    lockctl not checking for nfs3 mounts

    PROBLEM SUMMARY:
    Development discovered error in NFSV3 export with
    applications using fcntl locks

    PROBLEM CONCLUSION:
    Check for NFS or NFS3 in lockctl paths.

    ------

    APAR: IY20406 COMPID: 5765B9501 REL: 320
    ABSTRACT: BETTER REPORTING OF BROKEN ADDRS IN FSCK

    PROBLEM DESCRIPTION:
    better reporting of broken addrs in fsck

    PROBLEM SUMMARY:
    This defect improves fsck to better handle file systems with
    lots of broken files. It increases the number of broken
    files being fixed in each fsck run.

    PROBLEM CONCLUSION:
    The command mmfsck was changed to allow bigger number of
    files with broken addresses as well as to report the number
    of files with broken addresses.

    ------

    APAR: IY20407 COMPID: 5765B9501 REL: 330
    ABSTRACT: INCREASE NUMBER OF POSSIBLE FSCK PROBLEMS DIAGNOSED

    PROBLEM DESCRIPTION:
    increase number of possible fsck problems diagnosed.

    PROBLEM SUMMARY:
    Improve the number of problems handled by mmfsck.

    PROBLEM CONCLUSION:
    Increased the number of possible problems that can be
    recorded to a full blocks worth (8191)

    ------

    APAR: IY20408 COMPID: 5765B9501 REL: 330
    ABSTRACT: ALLOC MGR SERVER STRUCT BEING REINITED TOO OFTEN

    PROBLEM DESCRIPTION:
    alloc mgr server struct being reinited too often

    PROBLEM SUMMARY:
    Alloc mgr server struct being reinited too often.
    Change allocation manager to only reset data structure on
    disk state changes and node failures.

    PROBLEM CONCLUSION:
    Changed allocation manager logic. Only reset the free info
    per disk data on disk state changes and node failures. The
    resetAllocDisksBitmap routine was causing this to be reset
    everytime a new fileAllocator was built.

    ------

    APAR: IY20409 COMPID: 5765D9300 REL: 310
    ABSTRACT: CORRECT CALL TO HAL_STATS IN MPCI

    PROBLEM DESCRIPTION:
    Correct call to hal_stats in MPCI

    PROBLEM SUMMARY:
    Window packet and byte counts from estat were not being
    updated correctly by MPCI.

    PROBLEM CONCLUSION:
    MPCI was calling hal_stats incorrectly. Call has been
    fixed.

    ------

    APAR: IY20410 COMPID: 5765B9501 REL: 330
    ABSTRACT: LEAK BROADCASTING FS DESC UPDATE

    PROBLEM DESCRIPTION:
    leak broadcasting fs desc update

    PROBLEM SUMMARY:
    Fix a storage leak in certain file system configuration
    commands.

    PROBLEM CONCLUSION:
    Fix a storage leak in certain file system configuration
    commands.

    ------

    APAR: IY20411 COMPID: 5765B8100 REL: 220
    ABSTRACT: CORE DUMP IF STOPWORD IS SPOKEN AS LAST WORD OF UTTERANCE

    PROBLEM DESCRIPTION:
    Core dump if stopword is spoken as last word of utterance

    PROBLEM SUMMARY:
    Core dump if stopword is spoken as last word of
     utterance

    PROBLEM CONCLUSION:
    If RecoComplete message is received when
    state=stopping then results are validated and returned to DT

    ------

    APAR: IY20412 COMPID: 5765B9500 REL: 140
    ABSTRACT: MMLSMGR SHOWS INCORRECT FS MGR

    PROBLEM DESCRIPTION:
    mmlsmgr shows incorrect fs mgr

    PROBLEM SUMMARY:
    mmlsmgr shows a stripe group manager for a
    filesystem which has been moved to another nodeset.

    PROBLEM CONCLUSION:
    Correct error in invocation of the change
    manager routine to force the manager to resign.

    ------

    APAR: IY20414 COMPID: 5765B9500 REL: 140
    ABSTRACT: LEFT OVER MOUNT POINT AFTER MMCHFS -C

    PROBLEM DESCRIPTION:
    left over mount point after mmchfs -c

    PROBLEM SUMMARY:
    the mount point and the /etc/filesystems
    stanza are not always removed from the old nodes after
    mmchfs -C.

    PROBLEM CONCLUSION:
    Cleanup the /etc/filesystems stanza.

    ------

    APAR: IY20416 COMPID: 5765B9501 REL: 320
    ABSTRACT: MMAUTOLOAD WILL NOT START MMFSD EVENTHOUGH AUTOLOAD=YES

    PROBLEM DESCRIPTION:
    mmautoload will not startsrc -s mmfs because the following
    test made in the script is nether true :
    checkGpfsEnvOutput=$($mmremote checkGpfsEnv mmautoload) 2>/dev/n
    ull
    rc=$?
    if $rc -eq 0 && -n $checkGpfsEnvOutput
    This is because mmremote checkGpfsEnv mmautoload output is
    always empty.

    LOCAL FIX:
    startsrc -s mmfs manually

    PROBLEM SUMMARY:
    GPFS autoload on system reboots was not working

    PROBLEM CONCLUSION:
    Correct autoload processing.

    ------

    APAR: IY20418 COMPID: 5765B9501 REL: 320
    ABSTRACT: INCORRECT LOG RECOVERY RANGE TEST

    PROBLEM DESCRIPTION:
    incorrect log recovery range test

    PROBLEM SUMMARY:
    Segmentation violating in mounting a file system after a
    disk error which resulted in an unreadable disk sector

    PROBLEM CONCLUSION:
    importDA of an exported broken diskaddr was not handling the
    signed
    value -1 correctly when it used nValidDiskAddrs

    ------

    APAR: IY20419 COMPID: 5765B9501 REL: 330
    ABSTRACT: INCORRECT LOG RECOVERY RANGE TEST

    PROBLEM DESCRIPTION:
    incorrect log recovery range test

    PROBLEM SUMMARY:
    Segmentation violating in mounting a file system after a
    disk error which resulted in an unreadable disk sector

    PROBLEM CONCLUSION:
    importDA of an exported broken diskaddr was not handling the
    signed
    value -1 correctly when it used nValidDiskAddrs

    ------

    APAR: IY20445 COMPID: 5765E5300 REL: 120
    ABSTRACT: PROBLEMS WITH SKIPPING TO NEXT PHRASE (DT TTS)

    PROBLEM DESCRIPTION:
    Problems with skipping to next phrase (DT TTS)

    PROBLEM CONCLUSION:
    return_code was overridden if stop_reason==
    DTMF||VOICE||HUP||EVENT

    ------

    APAR: IY20464 COMPID: 5765B9501 REL: 330
    ABSTRACT: MMFSCK STATS ABOUT SUSPENDED DISK ADDRESSES

    PROBLEM DESCRIPTION:
    mmfsck stats about suspended disk addresses

    PROBLEM SUMMARY:
    more data ouput by mmfsck when dealing with
    failed disks

    ------

    APAR: IY20465 COMPID: 5765B9501 REL: 330
    ABSTRACT: INODE REUSE PERFORMANCE, FSYNC PERFORMANCE

    PROBLEM DESCRIPTION:
    inode reuse performance, fsync performance

    PROBLEM SUMMARY:
    performance optimizations with large numbers
    of files.

    ------

    APAR: IY20509 COMPID: 5765B8100 REL: 220
    ABSTRACT: DIRECTTALK ISDN RLT DISCONNECT MSG ERROR. ISDN.COM PART

    PROBLEM DESCRIPTION:
    ISDN in directtalk RLT disconnect msg error.

    PROBLEM SUMMARY:
    directtalk ISDN RLT disconnect msg error.

    PROBLEM CONCLUSION:
    Fixed RLT protocol functions.
    1)Incorrectly sending of Disconnect messages.
    Changed callcontrol terminate call function.
    2)Copy the "Original Called PN" IE when present in bound setup t
    to the outbound setup. Added "Original Called PN" to transfer
    object in the SIT table, and updated ISDNDL3_DMSNAT and
    ISDNDL3_DMS250 to support "Original Called PN" IE in bound.

    ------

    APAR: IY20513 COMPID: 5765B9501 REL: 320
    ABSTRACT: IF MMDELDISK STOPPED EARLY, TRASHED FILESYSTEM

    PROBLEM DESCRIPTION:
    if mmdeldisk stopped early, trashed filesystem

    PROBLEM SUMMARY:
    ctl-c on mmdeldisk may corrupt some files.

    PROBLEM CONCLUSION:
    Correct error handling when mmdeldisk is interrupted by a
    user termination so that correct state is maintained.

    ------

    APAR: IY20528 COMPID: 5765B8100 REL: 220
    ABSTRACT: DIRECTTALK ISDN RLT DISCONNECT MSG ERROR. ISDN.EURO PART

    PROBLEM DESCRIPTION:
    ISDN in directtalk RLT disconnect msg error.

    PROBLEM SUMMARY:
    directtalk ISDN RLT disconnect msg error.

    PROBLEM CONCLUSION:
    Fixed RLT protocol functions.
    1)Incorrectly sending of Disconnect messages.
    Changed callcontrol terminate call function.
    2)Copy the "Original Called PN" IE when present in bound setup t
    to the outbound setup. Added "Original Called PN" to transfer
    object in the SIT table, and updated ISDNDL3_DMSNAT and
    ISDNDL3_DMS250 to support "Original Called PN" IE in bound.

    ------

    APAR: IY20537 COMPID: 5765B9501 REL: 330
    ABSTRACT: IF MMDELDISK STOPPED EARLY, TRASHED FILESYSTEM

    PROBLEM DESCRIPTION:
    if mmdeldisk stopped early, trashed filesystem

    PROBLEM SUMMARY:
    ctl-c on mmdeldisk may corrupt some files.

    PROBLEM CONCLUSION:
    Correct error handling when mmdeldisk is interrupted by a
    user termination so that correct state is maintained.

    ------

    APAR: IY20569 COMPID: 5765B8100 REL: 220
    ABSTRACT: DIRECTTALK ISDN RLT DISCONNECT MSG ERROR. ISDN.DMS PART

    PROBLEM DESCRIPTION:
    ISDN in directtalk RLT disconnect msg error.

    PROBLEM SUMMARY:
    directtalk ISDN RLT disconnect msg error.

    PROBLEM CONCLUSION:
    Fixed RLT protocol functions.
    1)Incorrectly sending of Disconnect messages.
    Changed callcontrol terminate call function.
    2)Copy the "Original Called PN" IE when present in bound setup t
    to the outbound setup. Added "Original Called PN" to transfer
    object in the SIT table, and updated ISDNDL3_DMSNAT and
    ISDNDL3_DMS250 to support "Original Called PN" IE in bound.

    ------

    APAR: IY20570 COMPID: 5765B8100 REL: 220
    ABSTRACT: DIRECTTALK ISDN RLT DISCONNECT MSG ERROR. ISDN.ATT PART

    PROBLEM DESCRIPTION:
    ISDN in directtalk RLT disconnect msg error.

    PROBLEM SUMMARY:
    directtalk ISDN RLT disconnect msg error.

    PROBLEM CONCLUSION:
    Fixed RLT protocol functions.
    1)Incorrectly sending of Disconnect messages.
    Changed callcontrol terminate call function.
    2)Copy the "Original Called PN" IE when present in bound setup t
    to the outbound setup. Added "Original Called PN" to transfer
    object in the SIT table, and updated ISDNDL3_DMSNAT and
    ISDNDL3_DMS250 to support "Original Called PN" IE in bound.

    ------

    APAR: IY20571 COMPID: 5765B8100 REL: 220
    ABSTRACT: ENHACEMENTS TO CALLING NUMBER FOR FUTURE FEATURE

    PROBLEM DESCRIPTION:
    Enhancements to calling number for future feature

    PROBLEM SUMMARY:
    enhancements to calling number for future
    feature

    ------

    APAR: IY20588 COMPID: 5648C9802 REL: 430
    ABSTRACT: SDK 1.3.0 PTF 7 : CA130-20010615

    PROBLEM DESCRIPTION:
    Fixes since PTF 6 (ca130-20010516) :
    (Note: The descriptions here have been truncated.)
    +--------+------+---------------------------------------------+
    |20010522|32211 |NullPointerException with GridBagLayout |
    +--------+------+---------------------------------------------+
    |20010522|32301 |BigInteger does not serialize properly |
    +--------+------+---------------------------------------------+
    |20010522|32412 |Overline input as other char in JTextArea |
    +--------+------+---------------------------------------------+
    |20010523|32448 |DateFormat.parse fails in arabic |
    +--------+------+---------------------------------------------+
    |20010524|32488 |Re-doing changes from 28050 (overwritten by 3|
    +--------+------+---------------------------------------------+
    |20010525|32372 |Incorrect Russian month names in long date fo|
    +--------+------+---------------------------------------------+
    |20010530|32390 |Missing reference to ZipFile in ZipFileInputS|
    +--------+------+---------------------------------------------+
    |20010530|32580 |MediaTracker image load failure |
    +--------+------+---------------------------------------------+
    |20010530|32602 |Ldap exOp support encodes tags wrong. |
    +--------+------+---------------------------------------------+
    |20010530|32671 |JVM hangs |
    +--------+------+---------------------------------------------+
    |20010531|29246 |Missing error message on invalid device |
    +--------+------+---------------------------------------------+
    |20010615|29377 |Javacore fix #2 - native stack backtrace fix |
    +--------+------+---------------------------------------------+
    |20010615|32242 |activeCount is not accurate |
    +--------+------+---------------------------------------------+
    |20010615|33071 |Incorrect Apperance of AWT Checkboxes |
    +--------+------+---------------------------------------------+
    |20010615|33132 |UpFolder Button on JFileChooser initially ina|
    +--------+------+---------------------------------------------+

    ------

    APAR: IY20675 COMPID: 5765D5100 REL: 320
    ABSTRACT: 45800001 CHECKSTOP DURING BOOT

    PROBLEM DESCRIPTION:
    Checkstop 45800001 during boot with SP-Switch2 adapters. The
    checkstop is very intermittent and occurs when running diags
    during boot against the css adapter.

    PROBLEM SUMMARY:
    Nodes intermittently checkstop during boot calling out the
    SP_Switch2 adapter. All error register bits are turned on.
    Adapter diagnostics inadvertently enabled the Phase-Lock-
    Loop (PLL) checkstop error.

    PROBLEM CONCLUSION:
    CSS diagnostics will not enable the Phase-Lock-Loop
    (PLL) checkstop.

    ------

    APAR: IY20735 COMPID: 5765B9501 REL: 320
    ABSTRACT: ENXIO FROM MAPPED FILE REQUESTS

    PROBLEM DESCRIPTION:
    ENXIO error being returned to mmap calls under heavy mmap

    PROBLEM SUMMARY:
    ENXIO error being returned to mmap calls under heavy mmap
    loads.

    PROBLEM CONCLUSION:
    Don't reuse a small buffer if it is memory mapped.

    ------

    APAR: IY20749 COMPID: 5765B9501 REL: 330
    ABSTRACT: ENXIO FROM MAPPED FILE REQUESTS

    PROBLEM DESCRIPTION:
    ENXIO error being returned to mmap calls under heavy mmap

    PROBLEM SUMMARY:
    ENXIO error being returned to mmap calls under heavy mmap
    loads.

    PROBLEM CONCLUSION:
    Don't reuse a small buffer if it is memory mapped.

    ------

    APAR: IY20946 COMPID: 5765D5100 REL: 320
    ABSTRACT: LATEST PSSP 3.2.0 FIXES AS OF JULY 2001

    PROBLEM DESCRIPTION:
    This is the lastest PSSP ptf as of July 2001.
    Order this apar to get all of the ptfs as of July 2001.

    PROBLEM SUMMARY:
    This is a packaging apar for PSSP 3.2.0 fixes
    as of July 2001.

    PROBLEM CONCLUSION:
    This is a packaging apar for PSSP 3.2.0
    fixes as of July 2001.

    ------

    APAR: IY20985 COMPID: 5765B8100 REL: 220
    ABSTRACT: CORRECT JAVA - CHP SOCKET PROBLEM

    PROBLEM DESCRIPTION:
    Java - CHP sockect problem

    PROBLEM SUMMARY:
    Java - CHP socket problem

    PROBLEM CONCLUSION:
    Correct code to ensure compatibility between
     CHP and Java beans

    ------