APAR: IY11750 COMPID: 5765D5101 REL: 111
ABSTRACT: ADAPTER FAILURES OR SWAPS ON WRONG NODE - HACMP/ES

PROBLEM DESCRIPTION:
A customer with gigabit ethernet adapters, pulled the cable
from the service adapter on one of the two machines and the
result was a swap_adapter event on both machines rather than
just the one where he had pulled the cable.

PROBLEM SUMMARY:
This problem happens when gigabit ethernet adapters
are used in an HACMP/ES cluster.
When adapter events (like pulling the adapter cable)
happen in one node, secondary and unexpected adapter
swap events may happen in another node.
The problem is more likely to occur in a 2-node HACMP/ES
cluster. If a standby adapter is brought down in one
of the nodes, the standby adapter in the other node
forms a "singleton adapter group", at the Topology
Services subsystem's level. This forces the subsystem
to use code that reads the adapter incoming packet counter
to determine whether the adapter is alive.
With gigabit ethernet adapters, the incoming packet
counter is not incremented in the same way as with
other kinds of adapters. As a result, the adapter is
sometimes considered dead (because the counter is sometimes
not incremented). Consequently, unexpected adapter swap
and "failed standby" events may be run on the node.

PROBLEM CONCLUSION:
The fix to the problem is a change to code in the
/usr/sbin/rsct/bin/netmonAdapterHealth program.
With the change, a longer interval is given between each
attempt at verifying whether the incoming packet
counter is being incremented. With the fix, the
chance that the counter will be incremented for a
working adapter will be close to 100%.

------

APAR: IY14068 COMPID: 576554802 REL: 110
ABSTRACT: ADD SUPPORT TO OPEN, CLOSE, READ, AND WRITE LINE SEQUENTIAL

PROBLEM DESCRIPTION:
AIX 4.2 introduced support for native files larger than 2GB. IBM
COBOL Set for AIX was developed on AIX 4.1.3 and doesn't support
the larger files. Customers have requested support. This APAR
adds support to open, close, read, and write line sequential
This support is only for ORGANIZATION IS LINE SEQUENTIAL.

LOCAL FIX:
Split large files into files smaller than 2GB and process the
smaller files.

TEMPORARY FIX:
Install, do not commit, Schwab.bff from
machine platypus.stl.ibm.com That machine is internal to IBM
developers. Customers need to create a PMR (or ETR), explain why
they wish an early shipment, understand that this early fix is
not fully tested, then queue that PMR to queue LNGWS,114 within
the U.S.A. or to WLNGWS,136 outside of the USA.

------

APAR: IY14763 COMPID: 5765D5100 REL: 311
ABSTRACT: THE CORE DUMP IS DUE TO A CODE BUG IN THE RECOVERY LOGIC.

PROBLEM DESCRIPTION:
== CORE_DUMP fault_service_Worm_RTG_SP as SIG# 11==
fault_service_Worm_RTG_SP caused segmentation vioration and
switch of Eprimary, Ebackup and some nodes were downed.
The worm caused the error while recovering the following error.
 -- fs_daemon_print.file --
// print_the_time: Time = Sun Nov 5 04:25:44 2000
    fs_daemon_fsm_main: got request, type = service-msg-received
    fs_daemon_fsm_main: packet Service Command (SC) = fa
    displayPacket: Packet type = TBS_SVC_CMD_ERROR_STATUS: //

PROBLEM SUMMARY:
A segment violation and core dump can occur during switch
recovery if the reinitNeighbors() function is invoked.
reinitNeighbors() is being called with the wrong
parameter. The function is called with the device id
(an uninitialized integer value), whereas it expects the
pointer to the device structure.

PROBLEM CONCLUSION:
Switch recovery code was changed to call
reinitNeighbors() with the pointer to the
device structure.

------

APAR: IY15401 COMPID: 5765D5101 REL: 111
ABSTRACT: HAGSD CORE DUMP HAS CORRUPTED STACK TRACE

PROBLEM DESCRIPTION:
Segmentation fault in SProviderMessgae.C causes corrupted stack
trace and hagsd to core.

PROBLEM SUMMARY:
The problem was due to a bug in a code
which should validate an array element
before using it since it may be reserved.

PROBLEM CONCLUSION:
After the fix of the problem to validate
the array element, the GSdaemon should not
core dumped.

------

APAR: IY15642 COMPID: 5765D5100 REL: 311
ABSTRACT: SP EXCLUSIVE ACCOUNTING CREATES NEGATIVE NUMBERS IN THE

PROBLEM DESCRIPTION:
when using sp exclusive accounting - the /var/adm/fee files
some times has negative numbers - which makes no sense.
This is caused whenan acctg files AIX fstat in acctexcl.c
returns an error which can not be duplicated. A suggested
solution is for us to handle the fstat error in acctexcl.c

PROBLEM SUMMARY:
When customers use excluse_node accounting,
there are instances when the /var/adm/fee
file has a job charge for a userid, that
is negative.
For example:
4077 jet 0 0 0 0 0 0 0 0 0 0 0 0 -961019740.000000 0 0 0
4077 jet 0 0 0 0 0 0 0 0 0 0 0 0 961020006.000000 0 0 0
This only occurs when:
a job is running using exclusive node accounting
              AND
nrunacct (SP accounting) program is run on the node
before the user's job is finished.
We attempt to calculate how much time userid's
have had excluse node usage thus far, even though
their job has not completed yet.
Acctexcl is called from nrunacct and tries to
access the input file statistical buffer (stat_buf).
However, were calling the function to access this
structure, fstat, with incorrect parameters, resulting
in 0 values in the stat_buf fields. Acctjob (also run
from nrunacct) uses this value, which is 0, to calculate
the total job charge fee for each user. Since we started
with 0, the algorithm returns a negative number, which
is incorrect. The code was changed to point to the
correct address for the stat_buf structure.

PROBLEM CONCLUSION:
nrunacct (SP accounting) program calls
executable acctexcl when SP exclusive
accounting is set to true.
If a job is still running when nrunacct
is run, we calculate how much time so far,
userid's have had exclusive node usage.
acctexcl is called from /usr/lpp/ssp/bin
nrunacct to keep track of the start and
end times of a job, per user.
In acctexcl we were using fstat to get
the address of the file statistical buffer
(stat_buf), passing it a file descriptor
pointer and the address for stat_buf.
Within the structure stat_buf, we were
using the field stat_buf.st_mtime, which
is the last data modification for a
specific userid. This field was always
0. This 0 value was then passed to
acctjob, which is also run in nrunacct.
Acctjob calculates the fee that should be
charged a userid for exclusive node usage.
This algorithm is based on subtracting
a value from the stat_buf.st_mtime value,
resulting in a negative number.
The code was changed to point correctly to
the real address of the stat_buf structure,
in which st_mtime value is correct.
OLD:
--> fstat(fd_acct, &stat_buf);
   stb.ac_btime = stat_buf.st_mtime;
NEW:
--> if (stat(argv 1 , &stat_buf) == 0) {
   stb.ac_btime = stat_buf.st_mtime;

------

APAR: IY15763 COMPID: 5765B8100 REL: 220
ABSTRACT: DIRECTTALK SUPPORT FOR SCI INTERFACE OF CISCO ICM V4.1

PROBLEM DESCRIPTION:
This APAR enables DirectTalk to use the Service Control Interfac
e of Cisco ICM V4.1.

PROBLEM SUMMARY:
This APAR enables DirectTalk to use the Service
 Control Interface of Cisco ICM V4.1

PROBLEM CONCLUSION:
Support added to allow use of Cisco ICM
V4.1

------

APAR: IY16041 COMPID: 5765D5100 REL: 311
ABSTRACT: NEED TO ENSURE TUNING.CUST IS ALWAYS RUN WHEN /ETC/RC.NET IS RUN

PROBLEM DESCRIPTION:
On a SP node, the network options are set by running
/tftpboot/tuning.cust. This script is called by /etc/rc.sp
when the node is booted. This overides any changes in
/etc/rc.net since rc.sp runs after rc.net. If a cfgmgr is
run, it will call /etc/rc.net. This means the "no" values
set in rc.net will now overide what was in tuning.cust.
To ensure tuning.cust is always dominant over rc.net on a
SP node, we need to make sure tuning.cust is always run
whenever rc.net is run.

LOCAL FIX:
If cfgmgr is run on, also run /tftpboot/tuning.cust if it
exists.

PROBLEM SUMMARY:
On an SP node, the network options are set by running
/tftpboot/tuning.cust. This script is called by /etc/rc.sp
when the node is booted. This overides any changes in
/etc/rc.net since rc.sp runs after rc.net. If cfgmgr is
run, it will call /etc/rc.net. This means the "no" values
set in rc.net will now overide what was in tuning.cust.

PROBLEM CONCLUSION:
/etc/rc.sp has been modified to add code to /etc/rc.net.
If /etc/rc.net is invoked at a time other then when the
node is rebooted, (determined by the fact that
/usr/sbin/srcmstr is not an active process),
it will now call /tftpboot/tuning.cust if it exists.

------

APAR: IY16125 COMPID: 5765D5100 REL: 311
ABSTRACT: ADD -P FLAG TO PTPECTRL USAGE STATEMENT

PROBLEM DESCRIPTION:
A new flag (-p Set Poll Count) was added to the ptpectrl command
While the -p flag is documneted in the ptpectrl man page, it
does not appear in the usage statement that is displayed when
you run ptepctrl -h. Customers would not know this feature is
available to them.

PROBLEM SUMMARY:
The -p flag (Set Poll Count) is documented in the ptpectrl
man page, but does not appear in the usage statement that
is displayed when you run ptpectrl -h.

PROBLEM CONCLUSION:
The -p flag (set poll count) will be displayed when you
run "ptpectrl -h", or when you enter an invalid flag,
for example "ptpectrl -a -x".

------

APAR: IY16179 COMPID: 5765D5100 REL: 311
ABSTRACT: SMITTY AND PRESPECTIVES DO NOT SUPPORT PARTITION SIZE MORE THAN

PROBLEM DESCRIPTION:
smitty and prespectives do not support partition size more than
256 MB when user tries to create VSD.

PROBLEM SUMMARY:
The createvsd command supports physical partition sizes of
512 and 1024 mb, but the smit interface was never updated
to accept these values. Perspectives, which invokes smit
config_data has the same problem.

PROBLEM CONCLUSION:
The smit panel "Create a Virtual Shared Disk" has been
expanded to allow physical partition sizes of 512 and 1024
megabytes.

------

APAR: IY16249 COMPID: 5765D5100 REL: 311
ABSTRACT: SPDELFRAM FAILS TO DELETE INFORMATION

PROBLEM DESCRIPTION:
spdelfram fails to delete frame information because spdelnode
passes a non-zero return code.

LOCAL FIX:
The work around for this problem is to issue spdelfram a second
time.

PROBLEM SUMMARY:
When issuing spdelfram to delete a frame that has just been
created, spdelnode fails because it is trying to remove
node information that has not been entered yet.
Because the nodes have not yet been configured, spdelnode
issues error messages when it attempts to delete the NIM
resources for the node as well the Adapter class
information for the node in the SDR. As a result, the
spdelfram command fails.

PROBLEM CONCLUSION:
spdelnode was modified to verify that a node has NIM
resources prior to attempting to remove them. It will also
verify that Adapter information exists in the SDR for a
node prior to attempting to delete it. As a result,
spdelnode will no longer issue error message when invoked
for a node that has not been configured. This will allow
spdelfram to delete a frame that has just been created.

------

APAR: IY16289 COMPID: 5765B9501 REL: 310
ABSTRACT: MNMSG DEADLOCK

PROBLEM DESCRIPTION:
mnMsg deadlock - known problem in 1.3. Parts of 298678 and
319508 from 1.3 may need to be retrofitted to 1.2.

LOCAL FIX:
The deadlock can be broken by recyclong the node(s) that show
hits on the following:
mmfsadm dump waiters | egrep 'Msg handler mnMsg.*change_lock_sha
rm waiting to set lx-lock-held flag'

PROBLEM SUMMARY:
Deadlock between two applications
using the fsync call against the same file

PROBLEM CONCLUSION:
Correct locking within the fsync call

------

APAR: IY16305 COMPID: 5765D5100 REL: 311
ABSTRACT: DIAG -C -D CSS0 BRINGS UP PROBLEM DETERMINATION SCREEN

PROBLEM DESCRIPTION:
executing diag -c -d css0 should not take the user to ELA screen
s, it appears css0 diagnostic method does not use the -c flag
anymore.
I think we want to check DA_CONSOLE_TRUE before calling ela_run
damode bits runing diags -c:
DA_CONSOLE_FALSE 0x00080000
da mode bits when running diags without the -c flag:
DA_CONSOLE_TRUE 0x00040000

PROBLEM SUMMARY:
When running diags -c -d css0, the diag method was calling
diagrpt without first checking for a console causing the
user to be prompted.

PROBLEM CONCLUSION:
The diag method first checks for a console before running
diagrpt.

------

APAR: IY16490 COMPID: 5765B9501 REL: 310
ABSTRACT: GPFS HANG ON META-NODE

PROBLEM DESCRIPTION:
gpfs hang on meta-node

PROBLEM CONCLUSION:
When getting the mnode token, use the LO_NOLX flag to avoid
deadlock when the current lkobj status is invalid.
Also, do not init the mnode lkobj if it has already been
initialized
by some other thread.

------

APAR: IY16543 COMPID: 5765B9501 REL: 310
ABSTRACT: FILE SYSTEM CANNOT BE REBALANCED

PROBLEM DESCRIPTION:
file system cannot be rebalanced

PROBLEM CONCLUSION:
ReuseBitmap caused bad endMask to be generated for last word
of the map
when the number of bits was exactly a multiple of 32.
Since the disk allocation code used this function
extensively, if there was
a multiple of 32 disks in the filesystem strange things
happened.

------

APAR: IY16559 COMPID: 5765B9501 REL: 310
ABSTRACT: FIX FOR IY16289 INCOMPLETE

PROBLEM DESCRIPTION:
IY16289 did not account for some other Release 3 design changes,
so the message handler thread needed to reacquire the lock with
a stronger mode to delete the invalid object. The thread
deadlocked with itself since it already held a RO lock, then
asked for an LX lock.

PROBLEM SUMMARY:
Follow-on to IY16289

------

APAR: IY16710 COMPID: 5765D5100 REL: 311
ABSTRACT: VSD/RVSD ENHANCEMENTS

PROBLEM DESCRIPTION:
VSD/RVSD support for the Subsystem
Device Driver (SDD) for the Enterprise Storage Server (ESS)

PROBLEM SUMMARY:
VSD/RVSD enhancements

PROBLEM CONCLUSION:
VSD/RVSD enhancements

------

APAR: IY16789 COMPID: 5765D6100 REL: 210
ABSTRACT: CRXLF90 GIVES: LD: 0711-317 ERROR: UNDEFINED SYMBOL: AIO_READ64

PROBLEM DESCRIPTION:
crxlf90 gives:
ld: 0711-317 ERROR: Undefined symbol: aio_read64
ld: 0711-317 ERROR: Undefined symbol: aio_write64
ld: 0711-317 ERROR: Undefined symbol: aio_suspend64
ld: 0711-317 ERROR: Undefined symbol: aio_cancel64
ld: 0711-317 ERROR: Undefined symbol: .einfo

LOCAL FIX:
at line 88 of the script crxlf90
add -l xlf90 before the \

PROBLEM SUMMARY:
The crxlf90 compilation script is missing a reference
to the xlf90 library.

PROBLEM CONCLUSION:
-lxlf90 can be added to the script so that customers do
not have to include this change as part of the
customization required for this script to run.

------

APAR: IY16980 COMPID: 5765B9501 REL: 310
ABSTRACT: DEADLOCK IN RENAME ON MULTIPLE NODES

PROBLEM DESCRIPTION:
Deadlock in rename on multiple nodes

PROBLEM SUMMARY:
GPFS deadlock when running multiple renames.

PROBLEM CONCLUSION:
GPFS locking error is rename

------

APAR: IY17356 COMPID: 5765C3403 REL: 433
ABSTRACT: AIX 4330-08 RECOMMENDED MAINTENANCE LEVEL

PROBLEM DESCRIPTION:
This APAR delivers the 4330-08 Recommended Maintenance Level for
AIX 4.3.3. This package applies to AIX 4.3.3 only, and should
be ordered specifying a service level of 4330 or higher.
This is a packaging APAR only. It will not appear in the list
of APARs on the SMIT "Update Software by Fix (APAR)" panel, nor
will the 'instfix' command show this APAR as being installed
after the updates delivered by this package are installed.
To install all updates from this package that apply to installed
filesets on your system, use the command:
  smit update_all
Installing all updates is recommended. However, you can install
selected updates from this package using the command:
  smit update_by_fix
Once installed, you can determine if all fileset updates for the
4330-08 level are installed using the following command. The
'oslevel' command will still indicate AIX 4.3.3.0.
  instfix -ik 4330-08_AIX_ML
If the output indicates that not all filesets are found, you
can determine which filesets are not updated using the command:
  instfix -ciqk 4330-08_AIX_ML | grep ":-:"

PROBLEM SUMMARY:
Packaging only.

------

APAR: IY17479 COMPID: 5765D5100 REL: 311
ABSTRACT: KERBEROS 4 SECURITY PROBLEM

PROBLEM DESCRIPTION:
kerberos 4 security problem

PROBLEM SUMMARY:
Security Issue.

------

APAR: IY17492 COMPID: 5765B9501 REL: 310
ABSTRACT: IN-MEMORY BUFFER ACCOUNTING INCORRECT FOR EXPANDED FRAGMENT.

PROBLEM DESCRIPTION:
In-memory buffer accounting incorrect for expanded fragment.

PROBLEM SUMMARY:
GPFS self check logic stopped GPFS because of incorrect
accounting for end of file in certain situations where the
buffer containing the file is stolen

PROBLEM CONCLUSION:
Correct buffer management logic

------

APAR: IY17496 COMPID: 5765D5101 REL: 111
ABSTRACT: KERBEROS 4 SECURITY PROBLEM

PROBLEM DESCRIPTION:
kerberos 4 security problem

PROBLEM SUMMARY:
Security Issue.

------

APAR: IY17741 COMPID: 5765C3403 REL: 430
ABSTRACT: SRN 66E-301 TESTING DVD IN GOOD MACHINE PATH

PROBLEM DESCRIPTION:
diagnostics returns SRN (Unit Attention) on Good Machine Path

PROBLEM CONCLUSION:
small code modification

------

APAR: IY17742 COMPID: 5765C3403 REL: 430
ABSTRACT: 803 SRN WITH DEVICE BUSY: DOESN'T RECOGNIZE INCORRECT MEDIA

PROBLEM DESCRIPTION:
Confusion on Diagnostic Selection Menu for DVD-RAM Drive.

PROBLEM CONCLUSION:
Small code changes

------

APAR: IY17748 COMPID: 5765C3407 REL: 210
ABSTRACT: V2R1 PTF-12

PROBLEM DESCRIPTION:
Added filesets for AIX V2R1 NSM PTF12

------

APAR: IY17749 COMPID: 5765C3407 REL: 210
ABSTRACT: V2R1 PTF-12 DUMMY

PROBLEM DESCRIPTION:
Added filesets for AIX V2R1 NSM PTF12

------

APAR: IY17750 COMPID: 576554801 REL: 110
ABSTRACT: ADD SUPPORT TO OPEN, CLOSE, READ, AND WRITE LINE SEQUENTIAL

PROBLEM DESCRIPTION:
AIX 4.2 introduced support for native files larger than 2GB. IBM
COBOL Set for AIX was developed on AIX 4.1.3 and doesn't support
the larger files. Customers have requested support. This APAR
adds support to open, close, read, and write line sequential
This support is only for ORGANIZATION IS LINE SEQUENTIAL.

LOCAL FIX:
Split large files into files smaller than 2GB and process the
smaller files.

TEMPORARY FIX:
Install, do not commit, Schwab.bff from
machine platypus.stl.ibm.com That machine is internal to IBM
developers. Customers need to create a PMR (or ETR), explain why
they wish an early shipment, understand that this early fix is
not fully tested, then queue that PMR to queue LNGWS,114 within
the U.S.A. or to WLNGWS,136 outside of the USA.

------

APAR: IY17785 COMPID: 5765C3403 REL: 430
ABSTRACT: FC BOOT INSTALLS WITH/WITHOUT EMC SYMMETRIX FILESET

PROBLEM DESCRIPTION:
When booting from FC Symmetrix, adding EMC software after
installing system may make system unbootable.

PROBLEM CONCLUSION:
Fix issue with migrating from generic OS FC disk to
more specific FC disk type.

------

APAR: IY17870 COMPID: 5765D5100 REL: 311
ABSTRACT: LATEST PSSP 3.1.1 FIXES AS OF MARCH 2001

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

PROBLEM SUMMARY:
This is a packaging apar for PSSP 3.1.1 fixes
as of March 2001.

PROBLEM CONCLUSION:
This is a packaging apar for PSSP 3.1.1
fixes as of March 2001.

------

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