APAR: IW00560 COMPID: 5765E8900 REL: 210
ABSTRACT: VARIOUS PAYMENT REGISTRY 2.1 PROBLEMS FOUND BY IBM CHINA.

PROBLEM DESCRIPTION:
This APAR describes the following issues:
1. Limiting the length of the password to 8 bytes. From this
   efix forward, the password for the users would not exceed
   8 bytes.
2. Simplified Chinese locale needs to be run under 1.1.8.12
   Java.rte level. This fixes a problem with the Modify Users
   function where the password field would remain grayed out
   inappropriately.
3. Searching by Date on the Approver (for user info) and in the
   Admin (for Audit info) in the Simplified Chinese locale is
   now functional.
4. Warning messages on incorrect password length have been added
   to make them more clear.
An efix is available from ecomww,109.

PROBLEM SUMMARY:
Improved error messages for password field data
. Fixed non-functional Modify User function in setadmin in Simpl
ified Chinese locale.

PROBLEM CONCLUSION:
In simplified Chinese locale:
1. Use jdk 1.1.8.12 or jdk 1.3
2. Date & Time locale changes are necessary to get those search
functions enabled in this locale.
3. Enhanced error messages are helpful when setting up password
and using the password

------

APAR: IW00562 COMPID: 5765E8900 REL: 210
ABSTRACT: PAYMENT REGISTRY 2.1 CRYPTO CARD D/T4758 ERROR RC=08 RC=90

PROBLEM DESCRIPTION:
After migrating the Payment Registry V1.2 to V2.1 is not
possible to get a merchant certificate. Using the tool
eecertreq to simulate a merchant requests. We get the policy
text, as expected, we say ok to the policy text but processing
the next step we get the error code 8/90. This error indicates
that the crypto card "denied access". The user exit
uses the role "setca" of the cryptocard , and this role is still
active,valid and has all rights needed.

LOCAL FIX:
Development is writing an efix that will allow the user exit to
establish its own session with the 4758, if the user exit wishes
to use the 4758 crypto card. The Payment Registry server
reestablishes the session with the 4758 once control is returned
from the user exit.

PROBLEM SUMMARY:
User exit using 4758 in 1.2 version of
Payment Registry, needs to now do an explicit, one time
logon in the User Exit to use 4758. Otherwise, it does not
work.

PROBLEM CONCLUSION:
In 2.1 Payment Registry, the logon
context provided does not cover the user exit anymore.
Hence an explicit logon to a separate profile is required
to use 4758 in the User Exit

------

APAR: IY15663 COMPID: 5765E2820 REL: 430
ABSTRACT: JAVA/API ERROR: EIB RESPONSE NOT RESET ON EXEC CICS LINK

PROBLEM DESCRIPTION:
CICS/JAVA error - An EXEC CICS LINK from non-JAVA program
to JAVA program does not properly handle the EIB
return code. If a JAVA program invokes an API, say
a NOTFND on an EXEC CICS READ , then the NOTFND
is propogated back on the call as the EIBRCODE in
the EXEC CICS LINK to the JAVA program.
Sample program locations are found on txlab15,
region TJEREG1. For my reference, modules affected
are pinca_java2.c. Also not pindc_exec.c. PinDC_DTCExec
properly sets EIB return codes. Because the
pinca_java2.c program (CRUISE name cics_api_exec_java)
does not reset the EIB before unloading, the
previous DTC exception gets thrown and returned in
eibresp.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS Users, Release 430
When using EXEC CICS LINK to run a Java program which does a
READ against a file that doesn't exist the NOTFND response is
incorrectly returned as being the response code for the LINK.
Problem Diagnosis After an EXEC CICS LINK command (in any
programming languages, not just Java) the values of the EIB are
not updated to the correct values for the EXEC CICS LINK, the
values
 of the last EXEC CICS command are returned instead. NOTE
 When using the EXEC CICS LINK command the EIB now contains
 values associated with the LINK command and not the values
 from the last EXEC CICS command executed in the linked to
 program. This change in behaviour may affect existing
 application code that has conditional logic based on the
 values found in the EIB, and these programs may have to be
 modified and recompiled.

PROBLEM CONCLUSION:
 TasPR_IRun has been modified to set the EIB after an EXEC CICS
 LINK.

------

APAR: IY18280 COMPID: 5765D5100 REL: 320
ABSTRACT: "TRY_TBIC_DEV" NOT INITIALIZED (BREAKS WRAPPED PORT CHECKING)

PROBLEM DESCRIPTION:
(In phase 1, and possibly other Worm paths), the field
"try_tbic_dev" is not initialized when the Worm is checking
a presumed wrapped port for a miswire (i.e., a node, versus a
wrap plug.) Due to this, the correct E/S response from the chip
is not recognized by ResponseWaitAndReceive(). There is also a
logging pbm (in this same code block) because the field
"att_dev_id" is not initialized to the current device;
specifically, the "route to" and "route from" messages to the
print file call out the wrong target device ID.

LOCAL FIX:
None. The effect of both of the these problems is minimal.
One is simply a logging problem. The other (wrap plug checking)
is not a concern (assuming the port is really wrapped, as
specified in the topology file), because the port will be
marked as wrapped when there is no response from the target
chip.

PROBLEM SUMMARY:
In phase 1 processing, during an Estart, the Worm for the
SP Switch2 can log incorrect messages. In an attempt to
identify all possible miswires, the Worm will try to query
a port that it believes to be wrapped. The response that
the Worm usually gets back is not handled properly.

PROBLEM CONCLUSION:
The Worm will now correctly handle an Error/Status packet
that comes back as a result of doing a query on a port that
is presumed to be wrapped.

------

APAR: IY20737 COMPID: 5639I0920 REL: 430
ABSTRACT: CMF TIMING DATA REPORTED ON NT IS MISSING OR INCORRECT

PROBLEM DESCRIPTION:
The formated output of the Monitoring data on Windows NT
indicates that several of the timing values are incorrect:
1.CICS_EMP_FID_TASK_UT EMP 8 /* CPU time used by task */
is always 0
2. CICS_EMP_FID_TASK_ST EMP 211 /* Task CPU tim */
is always equal to or greater than CICS_EMP_FID_TASK_ET
EMP 7 /*Elapsed time whilst task was executing */
3. CICS_EMP_FID_SUT_CICSSPACE EMP 216
/* System/User time spent in CICS space */
is always equal to or greater than
CICS_EMP_FID_ET_CICSSPACE EMP 96
/* Elapsed time in CICS space*/
4. CICS_EMP_FID_ET_TERM_IO EMP 9
 /* Elapsed time waiting for Terminal I/O */ is always equal to
CICS_EMP_FID_SUT_TERM_MGR EMP 16
/* System time spent in Terminal Manager.*/
When the monitoring code was originally written Windows did not
distinguish between the user and the system time and hence we
could get only one time. Therefore CICS_EMP_FID_TASK_UT = 0 and
CICS_EMP_FID_TASK_ST >= CICS_EMP_FID_TASK_ET (greater because we
get the system time after the elapsed time). Same reasoning for
other fields.
But, Windows NT does differentiate between elapsed time, user
time, and kernel time. Windows tools frequently reference
kernel time as "Privilege Time". This is shown in basic Windows
tools such as: 'pview', 'pstat', 'task manager', and
'performance monitor'.

PROBLEM SUMMARY:
USERS AFFECTED: TXSeries CICS Users on Windows platforms,
Release 430
CMF timing data to do with elapsed CPU, terminal I/O and real
time for a task is missing or incorrect. Problem Diagnosis The
code does not differentiate between user, system and elapsed
time and therefore the values produced for a tasks user space
CPU time and its system space CPU time are not correct, for
example. The same applies to other fields which measure the
user space and system space times, such as the time waiting for
terminal I/O and the total elapsed time.

------

APAR: IY22309 COMPID: 5639I0920 REL: 430
ABSTRACT: MESSAGE ERZ058415E GENERATED ON WINDOWS 2000 EVENT VIEWER LOG.

PROBLEM DESCRIPTION:
Message ERZ058415E generated on Windows 2000 event viewer log.
The error occurs when Windows Management Instrumentation is
used with TX Series. The error is written to the event viewer
application log. The error is generated from supos_ntsvc.c
within function SupOS_SvcCtrlHandler(DWORD dwCtrlCode). The
switch on case statement uses the dwCtrlCode and the default
is the generated message. From what the program shows it looks
like the only valid values are 128,129 and 130. We get a code
that takes the default path and puts out the error.

PROBLEM SUMMARY:
USERS AFFECTED: CICS Users on Windows Systems, Release 430
When using Netfinity's Director to monitor statistics the Event
log shows the message ERZ058415E/002: Service 'cics.TX01' must
be controlled using CICS commands or GUI. Problem Diagnosis
Netfinity Director uses the Service control code
SERVICE_CONTROL_INTERROGATE to make a service report its
current status information. CICS does not handle this service
code, so issued the ERZ058415/002 message.

PROBLEM CONCLUSION:
CICS now handles the service control code
SERVICE_CONTROL_INTERROGATE.

------

APAR: IY22369 COMPID: 5765E2820 REL: 430
ABSTRACT: WHEN REMOVING CICS REGION WITH CICSCP DESTROY,ALL PROGRAMS POINT

PROBLEM DESCRIPTION:
When trying to remove CICS region with cicscp destroy command, t
his process removes all of the programs pointed to by symlink th
at were in the region path. Not even an rm -rf would traverse do
wn the symlink and remove the underlying files.
Defect number 203557

PROBLEM SUMMARY:
USERS AFFECTED: TXSeries CICS Users on Unix platforms, Release
430
When removing a CICS region with the cicscp destroy command,
all programs are removed, even those that are physically in a
directory outside of the CICS region path which is itself
symlinked into the tree structure. Problem Diagnosis When
using the cicscp destroy command, it recurses into all the
subdirectorys within the region directory and removes the
files. This includes symbolic links to directories.

PROBLEM CONCLUSION:
cicscp will not remove files from a symbolic link.

------

APAR: IY22405 COMPID: 5765E2820 REL: 430
ABSTRACT: INVALID DEVICE FOR AUTOINSTALL DURING CICSADD/UPDATE/DELETE

PROBLEM DESCRIPTION:
Null device type and garbage device name for terminals are
created when a cicsadd, cicsupdate or cicsdelete are done to the
runtime database.
Recreation scenario.
1) create a region
2) start the region
3) try to update pd or td on runtime database using cicsadd
   command.
Observe the console log file, the device type will be a null
value but the device name will be specified by Junk characters.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS Users, Release 430
When using cicsadd, cicsdelete and cicsupdate with the -B or -R
flags, the following informational messages may be produced
with garbage characters: ERZ011004I, ERZ011010I or ERZ011012I.
Problem Diagnosis: When using cicsadd, cicsdelete and
cicsupdate with the -B or -R flags, a terminal with a null
device type is installed into the region which can result in
garbage characters being output in certain CICS messages.

PROBLEM CONCLUSION:
The TerSH_ThreadInit function has been corrected to initialise
a zero length string instead of accessing garbage data.

------

APAR: IY24160 COMPID: 5765E2811 REL: 430
ABSTRACT: UNABLE TO START 2ND SFS IF 7-CHAR OF THE SFS NAME IS SAME

PROBLEM DESCRIPTION:
The matching routine for the server name to get the server
bindings fails.
Routname to get the server bindings fails.
Routine AdmSF_StartSFS fails to pick the correct binding entry
when more than one exist with the same text string in first 7
characters of the server name.
Binding for server abcdefgh is delivered when look abcdefg
binding. This happens when abcdefg is before abcdefgh.
Message SFS_NO_SUCH_FILE_SYSTEM is displayed.

LOCAL FIX:
- Change the name of SFS server ( Don't use similar strings)
- Edit the Binding file manually
- Start the SFS servers as 'all' instead to star one of them.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS Users, Release 430
Unable to start the second SFS Server if 7-char of the SFS name
is same as first SFS. The selection of binding entry is
incorrect. Problem Diagnosis The matching routine for the
server name to get the server bindings, fails. Routname to get
the server bindings fails. The routine fails to pick the
correct binding entry when more than one exist with the same
text string in first 7 characters of the server name. Binding
for server abcdefgh is delivered when look abcdefg binding.
This happens when abcdefg is before abcdefgh. Message
SFS_NO_SUCH_FILE_SYSTEM is displayed.

PROBLEM CONCLUSION:
The code has been fixed to select the correct binding entry for
the matching SFS server

------

APAR: IY24360 COMPID: 5765E2920 REL: 430
ABSTRACT: TXSERIES 4.3 FOR SOLARIS -WHERE THE CEMT BEING ISSUED TO PLACE

PROBLEM DESCRIPTION:
In TXSeries 4.3 for solaris, CCIN transactions hang if a
terminal that had been scheduled for CCIN Uninstall is then
placed out of service via CEMT prior to the completion of the
CCIN Uninstall.
The problem lies in the order in which we uninstall a
connection.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS Users, Release 430
Placing a CICS Universal Client terminal out of service can
cause a hang if the connection that terminal is using is
being uninstalled.
Problem Diagnosis
The code to place a TCP/IP based terminal out of service calls
a generic communications layer function to send data back to
the terminal which, if the connection is already partially
uninstalled, will then try to acquire a new connection and
deadlock itself, thereby hanging the CEMT task that was
trying to set the terminal out of service.

PROBLEM CONCLUSION:
The ComCL_OOSTerminal function has been changed to check
whether the connection is partially uninstalled and abandons
the data send to the terminal if it is.

------

APAR: IY24619 COMPID: 5765D5100 REL: 320
ABSTRACT: LONG DEFAULT MSG TO FFDC_STACK_LOG MACRO WILL CORRUPT STACK

PROBLEM DESCRIPTION:
FFDC_STACK_LOG() macro doesn't accommodate a long default msg
string (argument). A default msg that is too long will cause the
the stack to be corrupted. The results are indeterminate.

PROBLEM SUMMARY:
Messages longer than 80 chars sent from the fault service
daemon are corrupting the FFDC stack.

PROBLEM CONCLUSION:
The function that creates the stack record for the fault
service daemon needs to have the extraneous "Msg not
found: " text removed allowing 15 additional characters
to be in the message. Then msgs 2510-606, 2510-203,
2510-816, 2510-817, and 2510-820 need to be shortened
so that they do not exceed the new limit of 98 characters.

------

APAR: IY24912 COMPID: 5765E2920 REL: 430
ABSTRACT: CICS REGION IS ABNORMAL TEMINATED WITH U1001 ABEND AND

PROBLEM DESCRIPTION:
Exception exc_e_illaddr in ComIP_CleanUp causing U1001 abend
and the termination of the cics region.
Defect : 203687

PROBLEM SUMMARY:
USERS AFFECTED:All TXSeries CICS Users, Release 430
The CICS region abnormally terminates with U1001 Abend and
EXC_E_ILLADDR while deallocating a Conversation ID (CONVID)
handle.
Problem Diagnosis
The handles associated with the convid needs to be validated
before destroying the convid.

PROBLEM CONCLUSION:
The handle associated with the convid is now validated before
destroying the convid.

------

APAR: IY26467 COMPID: 5639I0920 REL: 430
ABSTRACT: UNABLE TO SET EXIT 33 ON PROGRAM DEFINITION GUI PANEL

PROBLEM DESCRIPTION:
The "User exit number" field on the program definition GUI panel
does not include an entry for exit 33. So a program defined
for exit 33 appears as user exit "0-None", despite the fact that
the definition appears correctly in the database. If the user
updates the program definition using the GUI, the assocation
with exit 33 is lost.

LOCAL FIX:
Don't use the GUI to update / define the program definition.

PROBLEM SUMMARY:
USERS AFFECTED:TX Series CICS Users on Windows Systems, Release
430
When using TSConfig to configure user exit 33 (UE046033) the
program name will not be saved. Problem Diagnosis The TSConfig
tool is missing the definition for user exit 33, 52 and 53.

PROBLEM CONCLUSION:
User exits 33, 52 and 53 have been added to the TSConfig
resource definitions.

------

APAR: IY26663 COMPID: 5639I0920 REL: 430
ABSTRACT: FOR FILES WITH NO KEY, INQUIRE FILE RETURNS KEYLENGTH(-1),

PROBLEM DESCRIPTION:
EXEC CICS INQUIRE FILE returns -1 for Keyposition, KeyLength and
recordsize for non-keyed files. It should be returning 0 as per
CICS API documentation.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS user, Release 430
When using an EXEC CICS INQUIRE FILE with the RECORDSIZE or
KEYLENGTH options and the file is closed or remote, -1 is
returned. When using the KEYPOSITION option for a non-keyed
ESDS file -1 is also returned. The documentation does not
document -1 as a valid response. Problem Diagnosis The
expected behaviour is when using INQUIRE FILE with RECORDSIZE
or KEYLENGTH the value specified in the FD stanza should be
returned. When using KEYPOSITION, if the file is closed, or is
an ESDS file, 0 should be returned.

PROBLEM CONCLUSION:
The code has been modified so the correct values are returned

------

APAR: IY26724 COMPID: 5639I0920 REL: 430
ABSTRACT: EXIT 33 DOES NOT PROVIDE ENOUGH INFORMATION TO IDENTIFY FEPI CON

PROBLEM DESCRIPTION:
The exit 33 COMMAREA (cics_UE_046033_t) has a maximum resource
key length of 8 characters. This is not sufficient to determine
the key for the FEPI CONNECTION that has been installed
discarded.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
The exit33 buffer (cics_UE046033_t) had been set to a maximum
resource length of 8 bytes which would be insufficient to hold
the recently installed/discarded FEPI runtime information.
Problem Diagnosis a) FEPI: Previously the buffer (UE_Name) was
set to 8 bytes, and which would definitely truncate the 24 byte
connection information (8 bytes for node, 8 bytes for target
and 8 bytes for pool) b) NON-FEPI(Region runtime database):
Previous implementation of 8 bytes buffer UE_Name or 8 bytes
copy was not sufficient for XA, sfs, schema, listener and
gateway definitions.
NOTE The version number for UE046033 (Exit 33) has changed to
2. Programs written for this exit may have to be modified and
will have to be recompiled.

PROBLEM CONCLUSION:
The definition of exit 33 has been modified to increase the
amount of data returned.

------

APAR: IY27159 COMPID: 5639I0920 REL: 430
ABSTRACT: SET TRANSACTION TCLASS(0) SETS TCLASS TO 32

PROBLEM DESCRIPTION:
A SET TRANSACTION(xxxx) TCLASS(0) results in the target
transactions TCLASS being set to 32.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
SET TRANSACTION (tran name) TCLASS(0) returns a wrong value on
NT . If the TRANSACTION TCLASS is SET to a value 0 , then an
INQUIRE TRANSACTION returns a value 32 instead of 0 Problem
Diagnosis When Tclass value is set for a transaction, it is
stored as bit mask representation. This bit mask
representation of the value in NT is is wrong when the Tclass
is set to zero.

PROBLEM CONCLUSION:
The problem is resolved by having the SET operation store it as
0 if the value of TCLASS is zero and not in a bit mask
representation.

------

APAR: IY27201 COMPID: 5765E2820 REL: 430
ABSTRACT: CICS REGION ABENDS DURING AS INITIALIZATION IF LOG DAEMON IS NOT

PROBLEM DESCRIPTION:
During the initialization of a CICS Application Server, if the
log daemon is not available or ready the AS will abend with a
U5701 and the following message:
ERZ057002E/0135 01/10/02 04:18:04 LX01 : CICS internal error
unsuccessful condition '(LogInUse && !((*LNBlockPP)->Deletable))
for function DamJo_LogDaemon.
The cmvc defect for this apar is 203752.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
Region Abends with an ERZ057002E/0135: CICS internal error:
Unsuccessful condition '(LogInUse &&
!((*LNBlockPP)->Deletable))' for function 'DamJO_LogDaemon '
Problem Diagnosis In certain circumstances, for example a high
speed machine, the Log Daemon may not be initialised before the
CICS Application Server tries to use the log, and the error
message is issued.

PROBLEM CONCLUSION:
The problem is resolved by having the Log Daemon included in
the synchronization process, so as to have the Application
Manager to wait till Log Daemon comes up. And only that does
the Application Manager fork the application servers.

------

APAR: IY27252 COMPID: 5639I0920 REL: 430
ABSTRACT: INQ TASK(XXXX) REPORTS AN INCORRECT TRANCLASS IT TCLASS IS CHANG

PROBLEM DESCRIPTION:
After issuing a SET TRANSACTION TCLASS(xxxx) an INQ TASK(yyyy)
TRANCLASS against a TASK running the transaction changed by the
SET command (and started after the SET) causes an incorrect
TRANCLASS value to be returned.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
INQUIRE TASK returns DFHTCL16 for TRANCLASS after a SET
TRANSACTION TCLASS(5) command. Problem Diagnosis INQUIRE
TASK(xxxx) reports an incorrect TRANCLASS after a SET
TRANSACTION(yyyy) TCLASS(V) command, if the TCLASS value (V) is
nonzero value. TClass is stored in the bit mask representation
and INQ TASK(yyy) returns the TRANCLASS as the stored TCLASS as
a value.

PROBLEM CONCLUSION:
The functionality was changed so that the bitmask representation
is converted to a number before displaying it.

------

APAR: IY27444 COMPID: 5765D6100 REL: 220
ABSTRACT: SYNTAX ERR IN CONFIG CAUSES NEGOTIATOR/ALL DEAMONS DOWN

PROBLEM DESCRIPTION:
Syntax error in user defined keyword causes it to run into
Start expression corrupting it and corrupted data send to
Negotiator causing it to crash/all Deamons.

PROBLEM SUMMARY:
A missing closing parenthesis in an expression on one node
can bring down the negotiator.

PROBLEM CONCLUSION:
There are two routines that evaluate expressions. One of
them was fixed in the LoadL 2.2 GA code, and the other one
needs the same fix applied to it - implements a clean
error on a missing closing parenthesis.

------

APAR: IY27450 COMPID: 5639I0920 REL: 430
ABSTRACT: FEPI INQUIRE PROPERTYSET RETURNS FORMAT(DATASTREAM) FOR DEVICE

PROBLEM DESCRIPTION:
FEPI INQUIRE PROPERTYSET returns FORMAT(DATASTREAM) for
DEVICE(LUP) resources. Shouldn't this be FORMAT(NOTAPPLIC)
to match host CICS?

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS Users for Windows NT, Release
430
FEPI INQUIRE PROPERTYSET returns FORMAT(DATASTREAM) for
DEVICE(LUP) resources. This is not consistent with CICS/390
Problem Diagnosis TXSeries CICS supports only SLU P Devices.
Hence the FORMAT option is invalid for SLU P Devices in the
FEPI INSTALL PROPERTYSET. It should follow CICS/390 and
return a value NOTAPPLIC instead of DATA STREAM for LUP
devices.

PROBLEM CONCLUSION:
The changes have been incorporated by making it consistent as
on CICS/390.

------

APAR: IY27521 COMPID: 5765D5101 REL: 121
ABSTRACT: ADD HOSTNAME RESOLUTION CHECKING TO PHOENIX.SNAP

PROBLEM DESCRIPTION:
add hostname resolution checking to phoenix.snap

PROBLEM SUMMARY:
Add 2-way hostname resolution checking to phoenix.snap

PROBLEM CONCLUSION:
Added 2-way hostname resolution checking to phoenix.snap

------

APAR: IY27655 COMPID: 5765E2820 REL: 430
ABSTRACT: ERZ048008E/0303 U4802: FREEING REGION POOL STORAGE

PROBLEM DESCRIPTION:
A U4802 abend occurs because of an internal error in CICS. This
abend occurs when trying to free-up a non-allocated memory block

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
CICS region goes down with U4802 abend, while attempting to
free the storage allocated for an Event Problem Diagnosis When
the CICS region pool runs out of storage, the allocation of
storage for the Event fails. But this failure condition is not
handled. Hence on completion of tasks, it tries to free the
storage which is not allocated resulting in the abend U4802

PROBLEM CONCLUSION:
The problem is resolved by allocating the memory for the event
before the work item is scheduled so that the allocation
failure can be handled and task will be abended.

------

APAR: IY27807 COMPID: 5765D5100 REL: 320
ABSTRACT: THE PERSPECTIVE TABLE VIEW HAS A PROBLEM WITH SWITCH_RESPONDS

PROBLEM DESCRIPTION:
The table view has a problem with switch_responds on
switchless systems. Everything after (to the right)
of where the switch_responds is supposed to be displayed
does not get displayed either.Any columns after the one
where switch_responds WOULD be are blank. Any columns
after the one where switch_responds WOULD be are blank.

LOCAL FIX:
Modified .sphardwaremonitorTableView and put it on
/usr/lpp/ssp/perspectives/profiles/en_US. (current $LANG=en_US)
It swaps the position of switch_responds end Environment LED,
making switch_responds last so there is nothing to the right
of it for it to mess up.

PROBLEM SUMMARY:
The code was not properly handling a non-existent attribute
for example "switch responds" in a switchless system.
This was causing the attribute's column and all columns
to its right to not appear in the table.
Normally a non-existent attribute would not be present in
any profile. However, "switch responds" is a member of the
default table profile.

PROBLEM CONCLUSION:
The code has been modified to check for and skip
non-existent attributes when it creates the list
of columns for the table.

------

APAR: IY28080 COMPID: 5765E2820 REL: 430
ABSTRACT: PL/I TRANSACTIONS RUN FINE IF LEVEL OF PROTECTION AGAINST CORRUP

PROBLEM DESCRIPTION:
Customer has found a problem with PL/I where when the level of p
rotection against user corruption is set to none in the region
the transactions run normally. If it is set to normal, the trans
actions hang. Nothing is written to console.xxx file.

LOCAL FIX:
none available

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS Users of PL/I programs,
Release 430
CICS PL/I programs can cause a SIGSEGV resulting in an apparent
hang. Problem Diagnosis When the region SafetyLevel attribute
is set to normal to enable runtime memory protection, CICS PL/I
programs access an invalid memory address resulting in a
SIGSEGV which on most platforms then causes an apparent hang.

PROBLEM CONCLUSION:
PinCA_StartIBMPLI has been corrected by defect 203786 to not
attempt to access the EIB pointer held in the region pool after
the region pool has been detached by the runtime memory
protection code.

TEMPORARY FIX:
Turn off CICS runtime memory protection (set the region
SafetyLevel attribute to 'none').

------

APAR: IY28085 COMPID: 5765D5100 REL: 320
ABSTRACT: CSSADM CORE DUMP

PROBLEM DESCRIPTION:
In case of a clock problem on the switch the cssadm
recovery action may cause cssadm to dump core.
The srcmstr then starts a new cssadm which starts
over with the recovery (basicly running Eclock+Estart).
The cssadm dumps core after the Eclock and before the
Estart. As each new cssadm does an Eclock the switch
never gets up again.
The only work around is to stop cssadm, run Eclock -d
followed by an Estart.

PROBLEM SUMMARY:
Core dump in the Switch Admin Daemon (cssadm) when
global switch recovery is turned on and Eclock is run.
The core dump is due to an overflow of a buffer that
is used to generate shell commands; the core dump may
occur when using long hostnames. Also, on some cssadm
messages, there is a mismatch between the format
specification (%s) and the corresponding parameter
(which is defined as int). The result is that the data
(node number, for example) is not displayed.

PROBLEM CONCLUSION:
The buffer for the generated commands was increased
from 120 to 256 bytes. The message text for information
messages 168 and 169 was changed to display the node's
hostname so as to match the format specification (%s).

------

APAR: IY28191 COMPID: 5765D5100 REL: 311
ABSTRACT: ALLOWING NODES ON AND OFF THE SWITCH

PROBLEM DESCRIPTION:
allowing nodes on and off the switch

PROBLEM SUMMARY:
Additional support required to allow nodes
at PSSP 3.1.1 and PSSP 3.2 to not be on the SP Switch2.

PROBLEM CONCLUSION:
Provided new support to allow nodes at
PSSP 3.1.1 and PSSP 3.2 to not be on the SP Switch2

------

APAR: IY28192 COMPID: 5765D5100 REL: 320
ABSTRACT: ALLOWING NODES ON AND OFF THE SWITCH

PROBLEM DESCRIPTION:
allowing nodes on and off the switch

PROBLEM SUMMARY:
Additional suport required to allow nodes
at PSSP 3.1.1 and PSSP 3.2 to not be on the SP Switch2.

PROBLEM CONCLUSION:
Provided new support to allow nodes at
PSSP 3.1.1 and PSSP 3.2 to not be on the SP Switch2.

------

APAR: IY28333 COMPID: 5765D5100 REL: 320
ABSTRACT: NODE PANIC IN VSDD:SUSVSD+6CC, RETROFIT OF DT 79627 INTO

PROBLEM DESCRIPTION:
node crashes in vsdd:susvsd+6cc.
defect 79627 for pssp 3.4 describes that in detail.
fix must be retrofitted into 3.2.

PROBLEM SUMMARY:
Certain conditions would result in suspendvsd asserting.

PROBLEM CONCLUSION:
suspendvsd was changed to assert only when it finds a
request in any vsd queues. If no requests are found in any
queues after a timeout period, suspendvsd will now complete.
Also, suspendvsd will now only close the lvm device
if the file descriptor is not null.

------

APAR: IY28335 COMPID: 5639I0920 REL: 430
ABSTRACT: TXSERIES HANGS HANG AFTER A CUC TASK GETS CANCELLED

PROBLEM DESCRIPTION:
Due to abnormal termination of a CUC task, during the
FORCEPURGing of an application server, TXSERIES hangs after msgs
ERZ034114E/0731 Unsuccessful attempt to uninstall entry 'JRL'
from Runtime Database. Entry is marked as in use.
ERZ014016E/0036 Transaction 'CPMI', Abend 'A147', at '????'.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
In a XA registered environment, TXSeries CICS region hangs when
a task is cancelled. Problem Diagnosis Upon cancellation of a
task, TXSeries CICS initiates a XA FORCEPURGE operation to
terminate the transaction. A section of code in XA FORCEPURGE
had errorneously locked a mutex which was preventing TXSeries
CICS from initiating further transactions.

------

APAR: IY28342 COMPID: 5765E2820 REL: 430
ABSTRACT: START OF SFS FAILS WITH 'NO SUCH FILE SYSTEM'

PROBLEM DESCRIPTION:
Trying to start SFS and it fails with 'no such file system'.
Error messages received are:
erz038255e/0124
erz038208e/0188
erz105093e/0022
The error was caused by invalid entries in the server_bindings
which were left behind after a failed creation of a SFS server.

LOCAL FIX:
Place the entry for the SFS you are trying to start last
in the server_bindings.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
Invalid entries are left behind in the server_bindings file
after the failure in the creation of a SFS server.
Subsequently, this causes the failure while even starting a
valid SFS Server Problem Diagnosis When trying to create an SFS
server, if the creation is unsuccessful an entry is made in the
server_bindings file. The invalid entries needs to be cleaned
up from the server_bindings file when creation of a SFS server
fails

PROBLEM CONCLUSION:
The functionality has been rectified to allow putting the SFS
server entries into the binding file only after the successful
creation of the SFS server. This avoids putting entries into
the server_bindings file when the creation of a SFS server
fails.

------

APAR: IY28358 COMPID: 5765D5100 REL: 320
ABSTRACT: 2545-001 AND 2502-612 ERRORS FROM UPDAUTHFILES

PROBLEM DESCRIPTION:
Customers at ssp.basic 3.2.0.15 who do not have DCE enabled or
installed will see error messages when running updauthfiles
(called from spsetauth, spdelnode, spauthconfig, and setup_CWS).
These error messages may appear on the screen or in the console
log, as well as in /var/adm/SPlogs/auth_install/log:
   updauthfiles: 2545-001 Command "/bin/sptgtprin ssp/spbgroot
                 <host>" was unsuccessful. rc=1.
   sptgtprin: 2502-612 DCE is not running on this host.
In spite of these error messages, updauthfiles will still end
with rc=0, and there should be no effect on the system. The
messages can be ignored unless DCE is supposed to be active on
the system.

PROBLEM SUMMARY:
A block of code that was intended only to be run when dce
was enabled was being run all of the time because of a logic
error. In a non-dce environment the dce-specific code would
fail and print out the error messages, but otherwise
cause no adverse affects.

PROBLEM CONCLUSION:
The logic has been corrected so that the dce-specific
code is executed only in a dce environment.

------

APAR: IY28464 COMPID: 5765D5100 REL: 320
ABSTRACT: SINCE PTFSET 15 UPDAUTHFILES IS HANDLING UPPER-CASE HOSTNAMES

PROBLEM DESCRIPTION:
Customers at ssp.basic 3.2.0.15 with CWS hostname using
upper-case letters will find that hostname in their .klogin
file. For example: rcmd.B0D0200EB0D0200E instead of
rcmd.b0d0200eB0D0200E
This will no longer match the principal names used in
commands such as rcmdtgt:
   Ticket file: /tmp/tkt_rcmd12392
   Principal: rcmd.b0d0200eB0D0200E

LOCAL FIX:
Edit the .klogin file after any run of updauthfiles (it is
called by spsetauth, spdelnode, spauthconfig, and setup_CWS
-- note that setup_CWS is called by setup_server).

PROBLEM SUMMARY:
Code in updauthfiles was not translating upper-case letters
in the CWS hostname to lower-case when it compiled the rcmd
principal names for the .klogin file. A code change in
PTF15 exposed this problem by changing the rules for when
the .klogin gets regenerated.

PROBLEM CONCLUSION:
The code has been modified to translate the upper case
characters to lower case for the .klogin file.

------

APAR: IY28545 COMPID: 5765D6100 REL: 220
ABSTRACT: LLSUBMIT ERROR 2539-390 WHEN DCE_AUTHENTICATION_PAIR ENABLES

PROBLEM DESCRIPTION:
llsubmit error 2539-390 when dce_authentication_pair enabled

PROBLEM SUMMARY:
There is a potential of llsubmit giving back the
error 2539-390 when DCE_AUTHENTICATION_PAIR
is specified due to an uninitialized variable.

PROBLEM CONCLUSION:
The uninitialized variable for the DCE_AUTHENTICATION_PAIR
is initialized.

------

APAR: IY28546 COMPID: 5765D6100 REL: 220
ABSTRACT: LOADL_SCHEDD CORE DUMPS WITH AFS_GETNEWTOKEN PRG

PROBLEM DESCRIPTION:
loadl_schedd core dumps with afs_getnewtoken prg

PROBLEM SUMMARY:
LoadL_schedd will core dump when running
interactive jobs when the
AFS_GETNEWTOKEN keyword is set to
run a program.

PROBLEM CONCLUSION:
The schedd will not core dump
when running interactive jobs
when the AFS_GETNEWTOKEN is set to run
a program in the LoadL_config file.

------

APAR: IY28585 COMPID: 5765B9501 REL: 320
ABSTRACT: MMCHFS DOES NOT CHECK FOR EXISTANCE OF SPECIAL DEVICE FILE FOR

PROBLEM DESCRIPTION:
When using mmchfs to change nodeset for fileystem there is no
check to see if the device special file for the gpfs filesystem
already exists on the nodes in the target nodeset. This becomes
a problem when the customer has a jfs filesystem that has a
device special file with the same name. The mmchfs command
completes successfully; however, if the mmchfs command is used
to bring the gpfs filesystem back to the original nodeset the
special device file that belongs to the jfs filesystem is
removed.

PROBLEM SUMMARY:
When using mmchfs to change nodeset for fileystem there is
no check to see if the device special file for the gpfs
filesystem already exists on the nodes in the target
nodeset. This becomes a problem when the customer has a jfs
filesystem that has a device special file with the same
name. The mmchfs command completes successfully; however, if
the mmchfs command is used to bring the gpfs filesystem back
to the original nodeset the special device file that belongs
to the jfs filesystem is removed.

PROBLEM CONCLUSION:
Do not remove /dev entries if they were not created by GPFS

------

APAR: IY28716 COMPID: 5765D5101 REL: 121
ABSTRACT: ADAPTER DETECTED DEAD TOO FAST

PROBLEM DESCRIPTION:
adapter detected dead too fast

PROBLEM SUMMARY:
RSCT Topology Services has tunable values (which in
HACMP/ES are externalized via the "Configure Network
Modules" SMIT panel) that control how fast it will
detect a node or adapter as dead.
In steady state -- in a cluster with 2 nodes or more --
the actual detection time for a failed node or adapter is
indeed close to the value specified in the tunable values.
However, under some circumstances, like when only a single
node is up or when an adapter has just undergone IP
Address Takeover, an adapter may be detected as dead in
much less time than what is dictated by the tunable
values. This has been seen to cause "false adapter down"
events in HACMP, mainly after IP Address Takeover
operations -- such as when the adapter acquires a service
address -- when the node's ARP cache is flushed by HACMP.
In HACMP, the problem usually manifests itself as an
unexpected "adapter down" event (could be an "adapter
swap") right after cluster initialization, when the
adapter acquires the service address, or right after
an IP Address Takeover operation.

PROBLEM CONCLUSION:
RSCT Topology Services has been changed to wait until the
time specified by its tunable values to declare a local
adapter as down. With the fix, even if the adapter is not
part of a multi-adapter group (such as when there is only
one node in the cluster which is up or right after IP
Address Takeover), the subsystem will not flag the adapter
as down too quickly.
The fix should allow Topology Services to avoid marking an
adapter as down when it just suffers a quick outage.

------

APAR: IY28732 COMPID: 5765D6100 REL: 220
ABSTRACT: LOADL_STARTER HANGING WITH DEFUNCT CHILD PROCESS

PROBLEM DESCRIPTION:
LoadL_starter hanging with defunct child process.
This hangs the whole job. Sending a SIGKILL to
the LoadL_starter puts the job into the VACATED
state and it will be restarted. Sometimes the
job may run then.

PROBLEM SUMMARY:
A child defunct process would be
produced if the SA_RESETHAND flag is set
and LoadLeveler had inherited that environment.

PROBLEM CONCLUSION:
LoadLeveler would always disable the SA_RESETHAND
flag so no defunct child process would
be produced.

------

APAR: IY28831 COMPID: 5765D5100 REL: 320
ABSTRACT: VSDDIAG ENDS IN ERROR CALLING SYSCTL. RETROFIT OF DEFECT 70685

PROBLEM DESCRIPTION:
vsddiag ends in error calling sysctl. Defect 70685for pssp
3.3 and 3.4 describes this in detail.
running /usr/lpp/csd/bin/vsddiag
results in:
vsddiag: 0034-610 Fails to call sysctl sysctl_vsdl1diag.
sysctl: 2501-004 Unknown host

LOCAL FIX:
Link csd_rm02 versions of vsdl1diag2.perl and dovsdl1diag.perl
files to csd_rmoh.

PROBLEM SUMMARY:
A change to the format of /usr/lpp/csd/vsdfiles/VSD_ipaddr
caused the parsing in dovsdl1diag.perl to fail.

PROBLEM CONCLUSION:
dovsdl1diag.perl was modified to parse the data in
/usr/lpp/csd/vsdfiles/VSD_ipaddr correctly. Also, the
processing in vsdl1diag2.perl to convert hostnames to
numbers was modified to remove leading blanks from
<HOSTNAME> before splitting.

------

APAR: IY28835 COMPID: 5765D5100 REL: 320
ABSTRACT: EXTRANEOUS CHARACTERS ON COLONY DIAG MENU

PROBLEM DESCRIPTION:
extaneous characters on colony diag menu

PROBLEM SUMMARY:
An adapter diags screen had extraneous characters displayed.
In previous releases of AIX and PSSP the positions occuppied
by these characters was automatically cleared, now we must
explicitly clear them.

PROBLEM CONCLUSION:
The diag code had to be changed to clear the extraneous
characters.

------

APAR: IY29056 COMPID: 5639I0920 REL: 430
ABSTRACT: UNEXPECTED SIGNAL CAUGHT ERROR WHEN COMPILING PLI PGMS

PROBLEM DESCRIPTION:
Compiling PL/I Programs gives different errors in WinNT and Win2
K. In both the cases, compiling fails.
In Windows NT, the following error is reported:
ERZ058405E/0161: The program 'pli' failed to execute, errno is '
2'.
SYMPTOMS = PIDS/5724NT620 LVLS/500 PTFS/(#)supos, 21:58:42, Sep
21 2001 RIDS/SupOS_SignalHandler LINE/109 MS/058916 MSN/1 SRC/11
PRCS/0 AB/ PID/1248 TID/0 TIME/010926154038 Eastern Daylight Tim
e. ERZ058916E/0001: Unexpected signal 11 caught
In WinNT, compilation and linking succeeds, but the PL/I transac
tion is getting abended.
ERZ058002E/0189 09/26/01 15:47:56 DEA2AD01 KG3R: Unable to find
entry point forprogram 'c:\st\AD\bin\spi\tcse.ibmpli'; error 127
ERZ058004E/0191 09/26/01 15:47:56 DEA2AD01 KG3R: Unsuccessful
unload of program with handle 004200C6C; error 126
ERZ014016E/0028 09/26/01 15:47:56 DEA2AD01 KG3R: Transaction 'tc
se', Abend 'APCT', at 'KG3R'

LOCAL FIX:
There are two problems. First one, the entry-point for PL/I prog
rams being wrong. The entry-point should be LICICS instead of PL
ICICS. Changing the entry-point fixes the first problem. The sec
ond problem is addressed as a separate CMVC defect.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
Unable to work with PL/I programs on either Windows 2000 or
Windows NT. On Windows 2000, the incorrect entry point of the
program causes the complilation/linking to fail while on
Windows NT, the execution of the transaction fails due to
failure during unloading of the program Problem Diagnosis There
are two parts to the problem - - One being, on Windows 2000,
the entry point of the program is wrong.
  Dumpbin on the <application-program>.exp shows plicics as the
  exported function but dumpbin on <appl_prog>.ibmpli shows
  licics as the exported function. - Other part of the problem
being, on Windows NT, the unload of the program
  fails because CICS finds that the memory has been corrupted.

PROBLEM CONCLUSION:
The entry point problem is rectified by having "plicics" in the
def file corresponding to which the dll file shows "licics" as
the exported function. The unload problem due to the memory
corruption is resolved by passing the parameters to the right
registers.

------

APAR: IY29110 COMPID: 5765E2820 REL: 430
ABSTRACT: ERROR CREATING SWITCHLOAD FILE FOR DB2 1PC

PROBLEM DESCRIPTION:
Error creating switchload file for DB2 1PC

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
Error creating switchload file for DB2 single phase commit.
Problem Diagnosis The error in creating the switchload file for
DB2 single phase commit This problem does not occur if you
modify the db21pc.mk file. Prior to executing the db2
precompiler, a db2 connect to database must be done and the C
compiler must be executed in a new line. It turns out that the
NT version of this makefile had all of the correct DB2 connect
statements, so the UNIX version must be made consistent with
the NT version.

PROBLEM CONCLUSION:
The problem has been rectified by modifying the db21pc.mk file
with the correct bind and connect statements.

------

APAR: IY29113 COMPID: 5639I0940 REL: 430
ABSTRACT: FEPI COMMANDS WOULD NOT COME UP PROPERLY WITHOUT THE CICSLU

PROBLEM DESCRIPTION:
A cicsas process cannot come up wihout
the cicslu (LU0 Listener) initialized. If this
happens, the cicsas would not be able to run
FEPI commands properly, as the FEPI run time
tables would not be initialized.

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
FEPI commands would not come up properly without the cicslu
(LU0 Listener) being properly initialised Problem Diagnosis
Currently there is no check provided to ensure that cicslu is
properly initialised before the application server comes up.
Without this, the cicsas would not be able to run FEPI commands
properly, as the FEPI run time tables would not be
initialized.

PROBLEM CONCLUSION:
The problem has been resolved by enforcing a check to ensure
that the cicslu comes up properly before the application
servers comes up.

------

APAR: IY29114 COMPID: 5639I0940 REL: 430
ABSTRACT: FEPI PERFORMANCE ENHANCEMENTS FOR MUTEX HOUSEKEEPING,EXCEPTION

PROBLEM DESCRIPTION:
FEPI Performance Enhancements for mutex housekeeping,exception
catches and connection retry logic

PROBLEM SUMMARY:
USERS AFFECTED: All TXSeries CICS users, Release 430
FEPI: Performance Enhancements
Problem Diagnosis
This consists of numerous enhancements to the mutex
housekeeping, exception catches, and connection retry logic.

PROBLEM CONCLUSION:
FEPI Performance enhancements are incorporated.

------

APAR: IY29121 COMPID: 5765B9501 REL: 320
ABSTRACT: GPFS WITH QUOTA ON, IN_DOUBT GROWS TOO QUICKLY ON LARGE SYSTEMS

PROBLEM DESCRIPTION:
On large systems where GPFS quota subsystem is turned on,
the per user in_doubt value grows unchecked.
The reclaiming of unused in_doubt does not bring down its value
to an acceptable level.

LOCAL FIX:
Running mmcheckquota on a GPFS file system that is offline will
reclaim all of the in_doubt . Running mmcheckquota on a mounted
GPFS file system will reduce the in_doubt size but will not
fully reclaim it.

PROBLEM SUMMARY:
The in-doubt value for GPFS files systems using quotas does
not get reclaimed after a period of inactivity.

PROBLEM CONCLUSION:
Provide a way to automatically reclaim unused client
shares after a period of allocation/deallocation inactivity
of the user on a client node, so that the overall inDoubt
for a user does not grow "unboundedly".

------

APAR: IY29156 COMPID: 5765B9500 REL: 130
ABSTRACT: MMCHFS FAILS WITH _MOUNT_CHECK_ONLY ERROR

PROBLEM DESCRIPTION:
mmchfs fails with _MOUNT_CHECK_ONLY_error

PROBLEM SUMMARY:
can not reassign a file system from a nodeset
where all nodes are unavailable.

PROBLEM CONCLUSION:
when moving a file system there is no need
for the daemon to be running anywhere in the source nodeset.

------

APAR: IY29231 COMPID: 5765D9300 REL: 310
ABSTRACT: MPI TASK DUMPS CORE IN COL_READPKT

PROBLEM DESCRIPTION:
Signal handling (non-threaded) allows an mpi write
to be re-entered by the same task causing tasks of
a POE job to hang or dump core. The stack trace
of a core dumping task is:
Segmentation fault in col_readpkt at 0xd0505d88
0xd0505d88 (col_readpkt+0x12fc) cbea0008 lfd fr31,0x8(r10)
(dbx) t
col_readpkt() at 0xd0505d88
kickpipes() at 0xd04d9870
mpci_recv(??, ??, ??, ??, ??, ??, ??, ??) at 0xd04f6a70
barrier_shft_b(??) at 0xd05920e8
_mpi_barrier(??, ??, ??) at 0xd0591e4c
MPI__Barrier(??) at 0xd059105c
mpi__barrier(??, ??) at 0xd011c3ec
gather_field() at 0x101d3d48
pp_output_slice() at 0x101d31f4
pp_output() at 0x101c9e24
pp_makegribs() at 0x101b97b4
pporg() at 0x101ab6a8
progorg() at 0x100c502c
gmeorg() at 0x1000079c

PROBLEM SUMMARY:
Running a non-threaded user space ( mpi signal handling
library ) program, a thread was re-entering writepkt driven
by a signal and causing the program to core dump with
various errors.

PROBLEM CONCLUSION:
Running a non-threaded ( signal handling mpi ) user space
program the thread that is writing will not reenter the
write routine.

------

APAR: IY29236 COMPID: 5765D9300 REL: 310
ABSTRACT: POE MAY NOT HANDLE MULTIPLE GMON.OUT FILES ON A NODE CORRECTLY

PROBLEM DESCRIPTION:
When customers compile a program for profiling using the -pg
option using POE, executing the program will sometimes not
create all or full gmon.out files when there are more than
a few tasks on a node. Sometimes the customer will see
messages like the following :
ATTENTION: 0031-662 Node 1 did not send PROFILE_DONE, sent
msgtype 15
ATTENTION: 0031-679 Profiling may not have completed on node 1

PROBLEM SUMMARY:
When customers compile a program for profiling using the -pg
option using POE, executing the program will sometimes not
create all or full gmon.out files when there are more than
a few tasks on a node. Sometimes the customer will see
messages like the following :
create all or full gmon.out files when there are more than
a few tasks on a node. Sometimes the customer will see
messages like the following :
ATTENTION: 0031-662 Node 1 did not send PROFILE_DONE, sent
msgtype 15
ATTENTION: 0031-679 Profiling may not have completed on node 1

------

APAR: IY29248 COMPID: 5765B9501 REL: 320
ABSTRACT: HUNG THREAD IN 'WAIT FOR BUFFER PIN OR UNPIN'

PROBLEM DESCRIPTION:
Hung thread in 'wait for buffer pin or unpin'
Before a buffer address range can be split, the caller must wait
for pinning or unpinning to finish, so that the second buffer
range description does not inherit the pinning state.

PROBLEM SUMMARY:
GPFS deadlock trying to pin a buffer for I/O.

PROBLEM CONCLUSION:
Before a buffer address range can be split, the caller must
wait for pinning or unpinning to finish, so that the second
buffer range description does not inherit the pinning state.

------

APAR: IY29304 COMPID: 5765B9501 REL: 330
ABSTRACT: LS FROM NFS3 CLIENT IS SLOW

PROBLEM DESCRIPTION:
ls command from NFS3 client is slow.
Inode prefetching was not happening because NFS does not have
an openInstance. If there are no instances, start prefetch
anyway if a readdir operationoccurred recently.

PROBLEM SUMMARY:
Customer reported ls -l performance was very slow when done
from an NFSV3 client.

PROBLEM CONCLUSION:
Inode prefetching was not happening because NFS does not
have an openInstance. If there are no instances, start
prefetch anyway if a readdir operation occurred recently

------

APAR: IY29328 COMPID: 5765B8100 REL: 220
ABSTRACT: CADENCE HUP ON ISDN IN GETDTMF CAUSES CHP TO HANG.

PROBLEM DESCRIPTION:
If Cadence Hangup detection is enabled on an ISDN line, and we
receive a Cadence tone whilst blocking for DTMF, the CHP process
will hang indefinitely.Any associates resources (including
DTBE apps) will also remain active, and unusable.

LOCAL FIX:
Check the Signalling Definitions for each trunk, and remove
Cadence Hangup detection if possible.

PROBLEM SUMMARY:
If Cadence Hangup detection is enabled on an
ISDN line, and we receive a Cadence tone whilst blocking for
DTMF, the CHP process will hang indefinitely.Any associates
resources (including DTBE apps) will also remain active,
and unusable.

PROBLEM CONCLUSION:
Make sure that anyone sleeping on the
channel is woken up before reseting the sleep words during
a near end CCS hangup.

------

APAR: IY29362 COMPID: 5765D6100 REL: 220
ABSTRACT: PASSDCE AND AUTHDCE DO NOT WORK

PROBLEM DESCRIPTION:
passdce and authdce do not work

PROBLEM SUMMARY:
When an installation supplies its own
DCE_AUTHENTICATION_PAIR, the user sees the following
erroneous error message displayed from llsubmit
llsubmit: Unable to read acknowledgement from process pipe,
read returned 0.

PROBLEM CONCLUSION:
The problem is caused because LoadLeveler is expecting to
read an integer from the child process running the first
executable in the DCE_AUTHENTICATION_PAIR. The integer is
only written by the child process when the IBM supplied
executables are specified. If an installation provides its
own pair, they do not write the integer, and LoadLeveler
produces the error message. The solution is to change
LoadLeveler to accept not receiving the integer from the
child process.

------

APAR: IY29363 COMPID: 5765D5100 REL: 320
ABSTRACT: CFGCOL & CFGCOR DON'T UNREGISTER W/ CSS PDD IN SOME EXIT PATHS

PROBLEM DESCRIPTION:
cfgcol and cfgcor have exit paths where they do not de-register
with the CSS pseudo device driver (pdd) for Communication Matrix
(CM) updates. This leaves a stale entry in the pdd data struct
that maintains the list of clients registered for CM updates.
The results are indeterminate, but one possible result is a node
crash.

PROBLEM SUMMARY:
The Colony and Corsair configuration programs (cfgcol
and cfgcor) have error exit paths where they do not
unregister the adapter with the CSS pseudo device driver
(pdd) table. This situation could arise, for example,
if post diagnostics fail. This may leave a stale entry
in the pdd data structure that maintains the list of
clients registered for Communication Matrix (CM) updates.
The results are indeterminate, but one possible result
is a node crash.

PROBLEM CONCLUSION:
If there are errors in the configuration program
(cfgcol or cfgcor) of the Colony or Corsair adapter
after the device is registered with the CSS pseudo device
driver, the adapter will be unregistered before the
configuration program exits.

------

APAR: IY29411 COMPID: 5765D5100 REL: 320
ABSTRACT: PSSP_SCRIPT VERIFY_QUORUM FUNCTION BREAKS WHEN LVSG IS EXECUTED

PROBLEM DESCRIPTION:
pssp_script verify_quorum function breaks when lvsg is executed

PROBLEM SUMMARY:
If users at ssp.basic 3.3.0.2 or higher, install or
migrate a node when the selected volume group is not rootvg,
pssp_script will fail with a messages that the lsvg command
failed. pssp_script was invoking volume group commands
with the name of the selected volume group when it should
have been specifying rootvg.

PROBLEM CONCLUSION:
pssp_script was modifed to use rootvg as the name of the
volume group, instead of using the name of the selected
volume group.

------

APAR: IY29414 COMPID: 5765B9501 REL: 330
ABSTRACT: READ USING DIRECTIO RETURNS BAD DATA

PROBLEM DESCRIPTION:
If direct I/O is used when create an Oracle database the GPFS
file system corrupted.

PROBLEM SUMMARY:
Incorrect data provided to application using asynchronous
I/O in conjunction with O_DIRECT application.

PROBLEM CONCLUSION:
Restore original segment registers before attaching user
buffer for direct I/O. Otherwise, if direct I/O is used
with asynch I/O it will use the wrong segment.

------

APAR: IY29415 COMPID: 5765B9501 REL: 320
ABSTRACT: FILESYSTEM HUNG AND LONG WAITERS

PROBLEM DESCRIPTION:
filesystem hung and long waiters

PROBLEM SUMMARY:
GPFS Deadlock on P690 with AIX 5.1.

PROBLEM CONCLUSION:
If a connection is dropped while an incomplete message was
in progress from the node, the hasData flag must be cleared
in receiveMsg, or it will loop forever.

------

APAR: IY29416 COMPID: 5765B9501 REL: 330
ABSTRACT: FILESYSTEM HUNG AND LONG WAITERS

PROBLEM DESCRIPTION:
filesystem hung and long waiters

PROBLEM SUMMARY:
GPFS Deadlock on P690 with AIX 5.1.

PROBLEM CONCLUSION:
If a connection is dropped while an incomplete message was
in progress from the node, the hasData flag must be cleared
in receiveMsg, or it will loop forever.

------

APAR: IY29430 COMPID: 5765D5100 REL: 320
ABSTRACT: SP SWITCH 2 DAEMON HAS INVALID DEPENDENCY ON SWITCH NODE NUM 0.

PROBLEM DESCRIPTION:
The SP Switch 2 fault-service daemon has a dependency on the
existence of switch node number 0. (Switch node number 0 doesn't
have to be initialized on the switch, but it must exist in the
SDR.) This is an invalid dependency because a customer can
remove the node that is assigned switch node number 0.

LOCAL FIX:
Add the node that is assigned switch node number 0 back to
the system.

PROBLEM SUMMARY:
On an SP_Switch2, if a customer deletes a node that has
been assigned switch node number 0, Estart will fail.
You will see the following message in the flt file on
the primary node:
CSswitchInit: CSworm_bfs_phase1() failed with rc=-51

PROBLEM CONCLUSION:
During phase 1 switch initialization, an assumption was
made that there always was a node that was assigned
switch node number 0 (zero). This is not a valid
assumption. Phase 1 initialization has been changed to
use the primary node's switch node number, instead of
the value 0, as the starting point for exploring the
network.

------

APAR: IY29471 COMPID: 5765B9501 REL: 330
ABSTRACT: GPFS WITH QUOTA ON, IN_DOUBT GROWS TOO QUICKLY ON LARGE SYSTEMS

PROBLEM DESCRIPTION:
On large systems where GPFS quota subsystem is turned on,
the per user in_doubt value grows unchecked.
The reclaiming of unused in_doubt does not bring down its value
to an acceptable level.

LOCAL FIX:
Running mmcheckquota on a GPFS file system that is offline will
reclaim all of the in_doubt . Running mmcheckquota on a mounted
GPFS file system will reduce the in_doubt size but will not
fully reclaim it.

PROBLEM SUMMARY:
the in-doubt value for gpfs files systems
using quotas does not gte reclained after a period of
inactifity.

PROBLEM CONCLUSION:
Provide a way to automatically reclaim
unused client shares after a period of allocation/deallocation
inactivity of the user on a client node, so that the overall
indoubt for a user does not grow "unboundedly".

------

APAR: IY29473 COMPID: 5765B9501 REL: 330
ABSTRACT: KERNEL PANIC DUE TO ASSERT IN SHHASHV.C LINE 1127:LM_HAVE != NL

PROBLEM DESCRIPTION:
If a file was changing from datashipping to non-datashipping
state in the middle of a read/write request. The file was left
in an unlocked state when returning from dsRdWr, and kSFSRead
was asserting when trying to upgrade the lock.
So gpfsperf with the -ds option is just cleaning up the
datashipping state from all the nodes when the next job starts
reading the same file. This read hits a timing window where it
gets into this retry-after-DS-just-turned-off code and forgot
to relock the file.
The result of all this is a kernel panic which causes the node
to crash with this type of traceback:
   mmfs:DoPanic__FPcT1iN23T1+0000EC
   mmfs:logAssertFailed+0000F0
   mmfs:change_lock_vfs__5LkObjFP8CacheObjQ2_5LkObj12LockMode
        EnumT2i+000C
   mmfs:upgradeFileLock__FP15KernelOperationP8OpenFile7
        FileUIDPQ2_5LkOb

PROBLEM SUMMARY:
GPFS self check logic detected an error ShHashV.C, line 1127
while running jobs using the MPI-Io library.

PROBLEM CONCLUSION:
If a file was changing from datashipping to non-datashipping
state in the middle of a read/write request. The file was
left in an unlocked state when returning from dsRdWr and
kSFSRead was asserting when trying to upgrade the lock. Fix
is to relock the file before returning EAGAIN.

------

APAR: IY29474 COMPID: 5765B9501 REL: 330
ABSTRACT: BAD DATA WITH AIO DIRECTION NEW ALLOCATIONS

PROBLEM DESCRIPTION:
bad data with aio direction new allocations

PROBLEM SUMMARY:
Under certain timing conditions using direct
IO and AIO, an incorrect block of data was written.

PROBLEM CONCLUSION:
If direct I/O write had to bail out after
writing part of the date (because a new data block had to be
allocated), it was not adjusting the file offset correctly.
This could cause the remaining data be written at the wrong
offset, over-writing the first part of the data.

------

APAR: IY29490 COMPID: 5765D5101 REL: 121
ABSTRACT: HAGSGLSM HANG PROBLEM WITH CSSRAWMEMBERSHIP

PROBLEM DESCRIPTION:
hagsglsm hang problem with cssrawmembership

PROBLEM SUMMARY:
HAGSGLSM might be hung either in lssrc command
or in a function to handle the cssMembership,
due to the misbehavior of signal handling.
The following fixes will be done.
1) SIGDANGER should be catched from signal thread.
2) Too heavy operation including many locks and waitpid
  should be removed to outside the signal thread.

PROBLEM CONCLUSION:
After the fix, lssrc should not be hung
and cssMembership on all nodes should be consistent.

------

APAR: IY29527 COMPID: 5765D5101 REL: 121
ABSTRACT: S/S FOLLOWING SWITCH FAULT, NODES DID NOT REGAIN CSSMEMBERSHIP

PROBLEM DESCRIPTION:
s/s following switch fault, nodes did not regain cssmembership

PROBLEM SUMMARY:
HAGSGLSM currently tries to open the devices
/dev/css0 or /dev/css1 and HAGSGLSM halts
to proceed to monitor their status (from
Topology Services) if open() == -1.
However, in certain situations (e.g., temporary
device driver problem), this kind of error may be
only temporary. So, open()==-1 test is not reliable.
The fix is to use "the existence" of devices /dev/css0 or
css
to determine whether HAGSGLSM should try to monitor or not.

PROBLEM CONCLUSION:
HAGSGLSM will see the existence of /dev/css0 or /dev/css0
(rather than checking the error code of open())
to determine whether it should monitor the switch
adapter status to form the css{0|1}Membership
groups.

------

APAR: IY29545 COMPID: 5765B9501 REL: 330
ABSTRACT: MMCHECKQUOTA PRODUCES NEGATIVE NUMBERS

PROBLEM DESCRIPTION:
mmcheckquota sometimes produces negative numbers when GPFS is
under heavy load.

PROBLEM SUMMARY:
mmquotacheck sometimes produces negative numbers for disk
usage.

PROBLEM CONCLUSION:
Do not update server's shadow entries at ComputeShare and
Relinquish routines since the quota usage and quota share
accounting in this case is done through regular quota
entries.

------

APAR: IY29618 COMPID: 5765D5101 REL: 121
ABSTRACT: IX:CT FFDC CAN'T TAKE 30 SEC TO TIMEOUT IF DNS NOT PRESENT

PROBLEM DESCRIPTION:
ix:ct ffdc can't take 30 sec to timeout if dns not present

PROBLEM SUMMARY:
When a system running the RSCT software is configured to
make use of a domain name server (DNS) for host name
resolution, and when the DNS service is either unavailable
or unreachable, failure would occur within the RSCT
software. Specifically, the RMC daemon would hang during
a restart operation, causing it to reject local or remote
connections, making it impossible to control system
resources through RMC. Also, the cthats script would hang
while rebuilding the machine.lst file.
The failure was tracked to the libct_ffdc library routine
fc_init(), which is attempting to resolve an IP address
into a host name. With DNS unavailable, the name
resolution interfaces would hang for 30 seconds or more.
causing the processes that use this interface -- the RMC
daemon and the cthats script -- to hang as well.
This placed a single point of failure within RSCT on the
domain name server.

PROBLEM CONCLUSION:
The failure occurs in the libct_ffdc library routine
fc_init(). This routine attempted to obtain a host name
for the IP addresss that it selected to identify the local
host, in order to avoid using a private interface. Once
the verification was done, the host name that was obtain was
discarded, not used anywhere else in the library.
The underlying logic was not correct, since the absence of a
host name assignment to a valid IP address is not an
indication as to whether the IP address is private or not.
Therefore, the attempt to retrieve a host name is
unnecessary, and could only serve to cause the failure
described in this APAR.
The attempt to retrieve the host name has been removed from
the routine.

------

APAR: IY29681 COMPID: 5765B9501 REL: 320
ABSTRACT: RECOVERY FAILED AFTER NODE FAILURE DURING RESTRIPE

PROBLEM DESCRIPTION:
recovery failed after node failure during restripe

PROBLEM SUMMARY:
Recovery of the GPFS file system failed after a node failure
while running the mmrestripe command.

PROBLEM CONCLUSION:
When moving indirect blocks for restripe, must spool a done
record before changing the disk address. Otherwise, log
recovery might apply the updates after the indirect block
has been re-used

------

APAR: IY29682 COMPID: 5765B9501 REL: 330
ABSTRACT: RECOVERY FAILED AFTER NODE FAILURE DURING RESTRIPE

PROBLEM DESCRIPTION:
recovery failed after node failure during restripe

PROBLEM SUMMARY:
Recovery of the GPFS file system failed after a node failure
while running the mmrestripe command.

PROBLEM CONCLUSION:
When moving indirect blocks for restripe, must spool a done
record before changing the disk address. Otherwise, log
recovery might apply the updates after the indirect block
has been re-used

------

APAR: IY29759 COMPID: 5765D5100 REL: 311
ABSTRACT: SYSLOGD NOT RESTARTED BY CLEANUP.LOGS.NODES

PROBLEM DESCRIPTION:
The syslogm routine in logmgt.cmds, called by cleanup.logs.nodes
(via psyslclr) stops syslogd after trimming a log. If trimming
another log results in an error, the routine exits without
restarting syslogd.

PROBLEM SUMMARY:
When psyslclr is invoked to trim multiple logs and it
successfully trims the first log, but does not have enough
space to trim subsequent logs, syslogd is stopped
but not restarted.

PROBLEM CONCLUSION:
Modified code in logmgt.cmds so that if syslogd is stopped,
it is always restarted.

------

APAR: IY29775 COMPID: 5765D5100 REL: 320
ABSTRACT: 64BIT:CSS0 DIAGS FAILS WITH TB3PCI

PROBLEM DESCRIPTION:
64bit:css0 diags fails with tb3pci

PROBLEM SUMMARY:
css0 diags may fail with TB3PCI on LPAR node.
The problem was caused when a previously undefined constant
was given a definition in system header files. A conditional
compilation was then reversed.

PROBLEM CONCLUSION:
The solution was to rename the constant so that it is again
undefined.

------

APAR: IY29830 COMPID: 5765D5100 REL: 320
ABSTRACT: SETUP_SERVER RETURNS 1 WHEN SPNET_ENX EXISTS ALREADY

PROBLEM DESCRIPTION:
setup_server returns 1 when spnet_enx exists aleady

PROBLEM SUMMARY:
On a CWS, after changing the IP address of an external
ethernet adapter that is not part of the SP LAN,
setup_server exits with a return code of 1. Before failing
though, setup_server displays a message letting the user
know there was a problem defining the associated NIM
spnet_enX resource (where enX is the en number of the
external ethernet with the changed IP address). The
following is an error message put out by mknimint ... it
is these errors which cause setup_server to truly fail:
0042-001 nim: processing error encountered on "master":
0042-032 m_mknet: object name must be unique and
                 "spnet_enX" already exists
mknimint: 0016-286 The "nim -o define" command had a
         problem defining spnet_enX on <master_hostname>
         with a return code value of 1.

PROBLEM CONCLUSION:
The new code within /usr/lpp/ssp/bin/mknimint does a check
to ensure that the interface associated with the network
being changed is not referenced by any client other than the
master. If the master is the only machine which contains the
changed network then the new code will blank out the nim
'if' install interface and cabletype in order to remove the
spnet_enX network. Once the network associated with the
external ethernet adapter (not part of the SP LAN) is
removed, the code will drop down into the 'regular' routine
of defining/redefining the network to nim.

------

APAR: IY29855 COMPID: 5765D5100 REL: 320
ABSTRACT: CSS:CFGMGR:051-604 CANNOT ACCESS THE CUDV

PROBLEM DESCRIPTION:
css:cfgmgr:0514-604 cannot access the cudv

PROBLEM SUMMARY:
In non-verbose mode, some messages are printed during
adapter configuration. Also, when configuration fails, some
error messages are printed to stdout instead of stderr.

PROBLEM CONCLUSION:
Made code changes to cfgcol.c so that no message will be
printed to stdout during cfgcol in non-verbose mode.

------

APAR: IY29913 COMPID: 5765B9500 REL: 140
ABSTRACT: MMSTARTUP -W FAILS WHEN THERE IS A SPACE AFTER THE NODE

PROBLEM DESCRIPTION:
mmstartup -w fails when there is a space after the node

PROBLEM SUMMARY:
gpfs not handling a blank after a nodename
in the node file associated with mmstartup -w.

PROBLEM CONCLUSION:
remove leading and trailing white space
around hostnames.

------

APAR: IY29923 COMPID: 5765D5101 REL: 121
ABSTRACT: C209F3N16:HAGSGLSM DAEMON HUNG DURING SIGNAL HANDLING

PROBLEM DESCRIPTION:
c209f3n16:hagsglsm daemon hung during signal handling

PROBLEM SUMMARY:
SIGCHLD signal is currently handled by the main thread
which also runs Group Services dispatch loop.
Some cases when a callback or SRC request is arrived,
HAGSGLSM calls the trace logging routine. Inside the
routine, there is a chance to receive a SIGCHLD signal
which invokes the signal handler. The signal handler
will try to call the same trace routine and may cause
the thread blocked due to the recursive locking.

PROBLEM CONCLUSION:
SIGCHLD signal will also be handled by the separate
signal handler. As the result, all asynchronous signals
now will be handled by a thread, and no nested
mutex lockings will be permitted.

------

APAR: IY29991 COMPID: 5765E2820 REL: 430
ABSTRACT: THIS IS THE REFERENCE APAR FOR TXSERIES 4.3, PTF6 ON AIX

PROBLEM DESCRIPTION:
This is the Reference APAR for TXseries 4.3, PTF6 on AIX. This
PTF contains the following APARs :
IY15663, IY22369, IY22405, IY24160, IY24360, IY24912, IY26663,
IY26724, IY27159, IY27201, IY27252, IY27450, IY27655, IY28080,
IY28335, IY28342, IY29056, IY29110, IY29113, IY29114.

PROBLEM SUMMARY:
USERS AFFECTED: CICS users, Release 430 for AIX

------

APAR: IY30061 COMPID: 5765D5100 REL: 320
ABSTRACT: SWITCH CLOCK NEEDS TO BLOCK SIGNALS DURING INITIALIZATION

PROBLEM DESCRIPTION:
switch clock needs to block signals during initialization

PROBLEM SUMMARY:
the switch clock API listener thread was not blocking
signals, and was taking delivery of signals intended to
signal other threads in the MPI job.

PROBLEM CONCLUSION:
block signals in switch clock API listener thread.

------

APAR: IY30154 COMPID: 5765B9501 REL: 320
ABSTRACT: FSSTRUCT 111 DIRECTORY ERROR

PROBLEM DESCRIPTION:
fsstruct 111 directory error

PROBLEM SUMMARY:
Cached data block of deleted directory could cause FSSTRUCT
on a newly created file.

PROBLEM CONCLUSION:
Before using the new file data block as a directory block,
compare the generation number in gnode with the generation
number of the new file.

------

APAR: IY30155 COMPID: 5765B9501 REL: 330
ABSTRACT: FSSTRUCT 111 DIRECTORY ERROR

PROBLEM DESCRIPTION:
fsstruct 111 directory error

PROBLEM SUMMARY:
Cached data block of deleted directory could cause FSSTRUCT
on a newly created file.

PROBLEM CONCLUSION:
Before using the new file data block as a directory block,
compare the generation number in gnode with the generation
number of the new file.

------

APAR: IY30194 COMPID: 5765D5100 REL: 311
ABSTRACT: UPDSUPPWD SHOULD ONLY UPDATE THE SUPMAN PASSWORK IF CHANGED

PROBLEM DESCRIPTION:
updsuppwd should only update the supman password if changed

PROBLEM SUMMARY:
The updsuppwd routine will compare the checksum files of the
current password and the previous password. If the checksum
file match then the password will not be transferred over
the s1term and updated on the node.

PROBLEM CONCLUSION:
The updsuppwd routine must be changed to not request the
password for the supman id over the s1term if the password
has not changed.

------

APAR: IY30226 COMPID: 5765B9501 REL: 330
ABSTRACT: MMCRLV FAILS YET RETURNS ERROR CODE OF 0

PROBLEM DESCRIPTION:
mmcrlv fails yet returns error code of 0

PROBLEM SUMMARY:
mmcrlv and mmcrvsd updated to ensure zero return code on a
failure related to hdisk already part of a VG.

PROBLEM CONCLUSION:
mmcrlv and mmcrvsd: fix bug in which conditional call to
unlockSDR() clobbered rc

------

APAR: IY30244 COMPID: 5765D5100 REL: 311
ABSTRACT: KFSERVER_TIMEOUT DEFAULT SHOULD BE 200

PROBLEM DESCRIPTION:
kfserver_timeout default should be 200

PROBLEM SUMMARY:
The value of the kfserver_timeout attribute in the SP class
is currently set to 600. There is no longer a reason for
the value to be this high. It should be lowered to 200.

PROBLEM CONCLUSION:
The value of the kfserver_timeout attribute in the SP class
has been lowered to 200.

------

APAR: IY30310 COMPID: 5765B9501 REL: 320
ABSTRACT: HANG: MMAP FLUSH BUFFER, VMM IOWAIT

PROBLEM DESCRIPTION:
hang: mmap flush buffer, VMM iowait
When findOrCreate has to go to the daemon to determine the gnode
type the gnode lock must be dropped or a deadlock can occur
involving the gnode lock and hash table lock.

LOCAL FIX:
The transition flag will stop the gnode from being used. Other
threads that find transition set now must wait untill the flag
is cleared, either by the gnode being deleted, or by the gnType
finally being set.

PROBLEM SUMMARY:
Deadlock condition could occur while mmap flush buffer

PROBLEM CONCLUSION:
When findOrCreate has to go to the daemon to determine the
gnode type the gnode lock must be dropped or a deadlock can
occur involving the gnode lock and hash table lock.

------

APAR: IY30340 COMPID: 5765B9501 REL: 330
ABSTRACT: HANG: MMAP FLUSH BUFFER, VMM IOWAIT

PROBLEM DESCRIPTION:
hang: mmap flush buffer, VMM iowait
When findOrCreate has to go to the daemon to determine the gnode
type the gnode lock must be dropped or a deadlock can occur
involving the gnode lock and hash table lock.

LOCAL FIX:
The transition flag will stop the gnode from being used. Other
threads that find transition set now must wait untill the flag
is cleared, either by the gnode being deleted, or by the gnType
finally being set.

PROBLEM SUMMARY:
Deadlock condition could occur while mmap flush buffer

PROBLEM CONCLUSION:
When findOrCreate has to go to the daemon to determine the
gnode type the gnode lock must be dropped or a deadlock can
occur involving the gnode lock and hash table lock.

------

APAR: IY30724 COMPID: 5765B9501 REL: 320
ABSTRACT: PROBLEMS WITH NOSUID FLAG IN GPFS FILESYSTEMS

PROBLEM DESCRIPTION:
problems with nosuid flag in gpfs filesystems

PROBLEM SUMMARY:
Security Problem.

PROBLEM CONCLUSION:
Security Problem Resolved.

------

APAR: IY30878 COMPID: 5765D5100 REL: 311
ABSTRACT: LATEST PSSP 3.1.1 FIXES AS OF MAY 2002.

PROBLEM DESCRIPTION:
This is the lastest PSSP ptf as of May 2002.
Order this apar to get all of the ptfs as of May 2002.

PROBLEM SUMMARY:
This is the latest PSSP ptf as of May 2002

PROBLEM CONCLUSION:
This is the latest PSSP ptf as of May
2002.

------

APAR: IY30928 COMPID: 5765B8100 REL: 230
ABSTRACT: CADENCE HUP ON ISDN IN GETDTMF CAUSES CHP TO HANG.

PROBLEM DESCRIPTION:
If Cadence Hangup detection is enabled on an ISDN line, and we
receive a Cadence tone whilst blocking for DTMF, the CHP process
will hang indefinitely.Any associates resources (including
DTBE apps) will also remain active, and unusable.

LOCAL FIX:
Check the Signalling Definitions for each trunk, and remove
Cadence Hangup detection if possible.

PROBLEM SUMMARY:
If Cadence Hangup detection is enabled on an
ISDN line, and we receive a Cadence tone whilst blocking for
DTMF, the CHP process will hang indefinitely.Any associates
resources (including DTBE apps) will also remain active,
and unusable.

PROBLEM CONCLUSION:
Make sure that anyone sleeping on the
channel is woken up before reseting the sleep words during
a near end CCS hangup.

------

APAR: IY31040 COMPID: 5765D5100 REL: 320
ABSTRACT: LATEST PSSP 3.2.0 FIXES AS OF MAY 2002

PROBLEM DESCRIPTION:
This is the latest PSSP ptf as of March 2002.
Order this apar to get all of the ptfs as of March 2002.

PROBLEM SUMMARY:
This is a packaging apar for PSSP 3.2.0 fixes
as of May 2002

------

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