|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
From: AIX Service Mail Server (aixserv
austin.ibm.com)Date: Tue Feb 20 2001 - 02:16:49 CST
APAR: IX82635 COMPID: 5765B7400 REL: 500
ABSTRACT: WHEN 'MQM' USERID IS NOT DEFINED, CRTMQM PRINTS A BROKEN ERROR M
PROBLEM DESCRIPTION:
When the 'mqm' userid is not defined on UNIX systems, crtmqm
ends with the ubiquitous 2195 return code and prints an FDC. It
also prints out a broken error message (AMQ6090) to the error
log. For a problem as simple as this, ie one which has a simple
answer within the user's range of control, we should print an
informative error message to the error log (if not to STDOUT).
At the moment, finding the problem involves walking code. This
should be a problem any user can handle.
LOCAL FIX:
Add the 'mqm' user to the system.
PROBLEM SUMMARY:
This problem has been fixed for MQSeries V5.0 and MQSeries V5.1.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
and the fix will be shipped in the following PTFs:
A) MQSeries for V5.0 CSD08
AIX U469658
HP-UX U469659
Sun Solaris U469671
B) MQSeries for V5.1 CSD06
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY07754 COMPID: 5765B7300 REL: 510
ABSTRACT: CRTMQM HANDLES FILESYSTEM MOUNTS INCORRECTLY
PROBLEM DESCRIPTION:
The 'crtmqm' program handles filesystems inconsistently.
That is, users may create filesystems beforehand for the queue
manager's data directories (/var/mqm/qmgrs/NAME), but not for
the queue manager's logs (/var/mqm/log/NAME). Instead, the
queue manager records an error if its logging directory already
exists, and crtmqm ends abnormally.
LOCAL FIX:
Create the queue manager first, then mount the filesystem and
copy over the queue manager logs.
PROBLEM SUMMARY:
This problem was caused by a difference in behaviour between the
routines for creating queue managers and for creating log files.
The routine for creating queue managers would create all of the
queue manager data tree under /var/mqm/XXX. The function would
either create this directory XXX or accept the fact that it
could be from a mounted file system. The way that it did this
is that if mkdir returned EEXIST, it then verified that the
directory was empty and if so it then continued by returning an
implicit success.
Conversely the routine responsible for creating the log files
had a different behaviour. If mkdir failed (for any reason
including the existence of the directory because it was mounted
or otherwise), then the attempt to create the log files would
fail.
This prevented the easy creation of a queue manager where the
logs were coming from a mounted file system.
The problem has been fixed by unifying the behaviour of these
two routines such that the routine for creating log files will
now perform the same empty directory check if it finds that the
directory, which it is trying to create, already exists.
PROBLEM CONCLUSION:
The change here was primarily to allow the creation of a queue
manager on a mounted file system. It should be noted that
mounted file systems when mounted at their top level often
include a lost+found entry. This entry will go against the rule
that the pre-existing directory entry should be empty and will
prevent the crtmqm writing log files into that mounted
directory. Therefore, it is wise to consider making a directory
below the root of the mounted file system and using that as the
mount point to /var/mqm/log The mounted file system must show
with owner group mqm:mqm otherwise it may not be modifiable by
chmod in the way that the new routine requires.
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
and the fix will be shipped in the following PTFs:
A) MQSeries for V5.0
AIX U469658
HP-UX U469659
Sun Solaris U469671
B) MQSeries for V5.1 CSD05 CSD06
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY08131 COMPID: 5765B7300 REL: 500
ABSTRACT: MQSERIES GENERATES PROBE ID VP011003 CONVERTING MQTMC2 MESSAGES
PROBLEM DESCRIPTION:
MQSeries may write an FDC with probe ID VP011003 when it tries
to convert an MQTMC2 trigger message. As it turns, the function
converting the trigger message expects only MQTMC type trigger
messages.
PROBLEM SUMMARY:
The probe id was produced by component vwb_trigger when it
encountered an MQTMC2 trigger message. This component only used
to recognise MQTMC trigger messages.
This problem has been fixed for MQSeries V5.0 and MQSeries V5.1.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
and the fix will be shipped in the following PTFs:
A) MQSeries for V5.0
OS/2 U200116
Windows NT U200117
AIX U469658
HP-UX U469659
Sun Solaris U469671
B) MQSeries for V5.1
OS/2 U200121
Windows NT U200122
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY08401 COMPID: 5765B7300 REL: 500
ABSTRACT: MQAIX RC2218 IS RECEIVED AFTER APPLYING CSD06 AND WHEN ATTEMPTIN
PROBLEM DESCRIPTION:
After applying CSD06 to MQS V5.0 on AIX, customer is getting
the above RC2218 when putting a message to the sender channel
from MVS to AIX. The message didn't transmit from MF to AIX and
ended up on the DLQ with MQI return code 2218, message too long
for channel. Channel definition on both MF and AIX showed
MAXMSGL=0. According to MQ manual, if this is coded, then
MAXMSGL is used which is coded on QM object. On AIX, QM MAXMSGL
was coded as 4M. Only after the receiving channel definition on
the AIX was altered to MAXMSGL = 4M would message flow.
LOCAL FIX:
Changing the MAXMSGL=4 M.
PROBLEM SUMMARY:
This problem has been fixed for MQSeries V5.0 and MQSeries V5.1.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
and the fix will be shipped in the following PTFs:
A) MQSeries for V5.0 CSD08
AIX U469658
B) MQSeries for V5.1 CSD06
AIX U471218
------
APAR: IY08576 COMPID: 5765B7300 REL: 510
ABSTRACT: CRTMQM CUTS AN FDC WITH PROBE XC037008 & ERROR XECP_E_PROC_LIMIT
PROBLEM DESCRIPTION:
crtmqm on UNIX platforms may end abnormally after writing an
FDC showing probe XC037008 from the function xcsExecProgram
with a major errorcode of xecP_E_PROC_LIMIT. This corresponds
to an errno of EAGAIN from fork(), which crtmqm calls to start
the logger process as part of formatting new log files for the
queue manager.
Typically EAGAIN means that the maximum number of processes
for a user has been reached, but this error occurs even when
there are no processes running under the 'mqm' user. The
error is correct, but it is caused by an odd historical quirk
in process accounting on UNIX whereby the fork will fail if
the number of processes running under root has already exceeded
the 'maxuproc' setting on the system.
LOCAL FIX:
Simply modify the maximum number of processes per user
(commonly labeled 'maxuproc') to a value greater than the
number of processes running under the 'root' ID. A code change
to MQSeries is required to obviate the need for this workaround.
PROBLEM SUMMARY:
This problem has been fixed for MQSeries V5.0 and MQSeries V5.1
by using setreuid on AIX in place of seteuid.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
------
APAR: IY08799 COMPID: 5765B7300 REL: 500
ABSTRACT: MQSERIES TRACING ON AIX SLOWS PERFORMANCE TOO MUCH
PROBLEM DESCRIPTION:
MQSeries' performance on AIX may be slowed somewhat even when
tracing is not active. On systems which are extremely heavily
loaded, this overhead can slow MQSeries processing noticeably,
although it should not cause any hard errors.
PROBLEM SUMMARY:
This problem has been fixed for MQSeries V5.0 and MQSeries V5.1
by using TRCGENT and TRCHKGT in place of trcgent and trcgkgt to
improve tracing performance on AIX. No change has been made to
the other unix platforms.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
and the fix will be shipped in the following PTFs:
A) MQSeries for V5.0
AIX U469658
B) MQSeries for V5.1
AIX U471218
------
APAR: IY08921 COMPID: 5765B7300 REL: 510
ABSTRACT: MQSERIES FUNCTIONS TRACE THE WRONG NAME
PROBLEM DESCRIPTION:
MQSeries functions may trace incorrectly in v5.1, pretending
to be a different function. For example, some traces show
MQOPEN calling zstMQDISC, when the next layer is zstMQOPEN.
This can be rather confusing when the error is not so obvious.
LOCAL FIX:
None. A correction to MQSeries is needed to avoid confusion.
PROBLEM SUMMARY:
This problem has been fixed for MQSeries V5.0 and MQSeries V5.1.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
and the fix will be shipped in the following PTFs:
A) MQSeries for V5.0 CSD09
OS/2 U200126
Windows NT U200127
AIX U472522
HP-UX U472523
Sun Solaris U472524
B) MQSeries for V5.1 CSD06
OS/2 U200121
Windows NT U200122
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY09040 COMPID: 5765D1400 REL: 220
ABSTRACT: SECD WILL CORE IN SYSLOG IF PRINCIPAL CONTAINS '%' CHAR
PROBLEM DESCRIPTION:
The security server will core in '_doprnt'
which is called by krb5 'syslog'. When an invalid
login occurs with a principal which contains
a '%' character, 'syslog_r' interpets the string
as a formatted string and therefore expects
an arguement. Since the '%' character is not
expected in the format string, no arguements
are passed into 'syslog_r' and hence the
core in '_doprnt'.
------
APAR: IY09122 COMPID: 5765D1400 REL: 220
ABSTRACT: DFS ASSERT CRASH IN 'RPC__DG_EXECUTE_CALL'
PROBLEM DESCRIPTION:
An assert occurs in DFS RPC in function
'rpc__dg_execute_call' if a packet sent by
a client contains an incorrect authenication
key.
------
APAR: IY09651 COMPID: 5765B7300 REL: 500
ABSTRACT: PING CHANNEL DOES NOT RETURN CHANNEL STATE TO RETRY
PROBLEM DESCRIPTION:
On MQ version 1.2 for MVS when the Sender channel is in RETRY
mode and PING CHANNEL is issued the Sender channel goes into
INACTIVE Mode and never returns to RETRY mode. Customer needs
the PING CHANNEL command to return the channel back to RETRY
Mode.
PROBLEM SUMMARY:
This problem has been fixed for MQSeries V5.0.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and the fix will be
shipped in the following PTFs:
A) MQSeries for V5.0 CSD08
OS/2 U200116
Windows NT U200117
AIX U469658
HP-UX U469659
Sun Solaris U469671
------
APAR: IY10592 COMPID: 5765B7300 REL: 500
ABSTRACT: ENHANCE MQSERIES FFSTS ON AIX TO INCLUDE THE PROGRAM NAME
PROBLEM DESCRIPTION:
MQSeries FFSTs on AIX do not print out the program name,
unlike FFSTs on most other platforms. This APAR will enhance
FFSTs on AIX by printing this information.
PROBLEM SUMMARY:
The new function has been provided for MQSeries V5.0 and V5.1.
PROBLEM CONCLUSION:
The new function has been provided and it will be shipped in CSD
#9 for MQSeries V5.0. It will be shipped in PTF U471218 for
MQSeries V5.1.
------
APAR: IY10596 COMPID: 5765B7300 REL: 510
ABSTRACT: CHANNEL FAILS TO START AFTER MIGRATING TO MQSERIES 5.1
PROBLEM DESCRIPTION:
Customer migrated their AIX MQSeries 5.0 qmgr (csd 5) to V5.1.
Before migrating they had working channels between AIX and
OS/390. After they migrated to 5.1 they were able to start the
sender from AIX to MVS but the channel from MVS to AIX gets
CSQX0206E with reason code 00000018. On AIX they get AMQ9002
and AMQ9001 saying the channel started normally and ended
normally.
The trace shows we negotiate and start the channel and do
indeed issue the AMQ9002 to say channel started. Later in the
trace it looks like we get some data from the Scratchpad then we
go on to ConvertSyncRecord which is failing with rc=10809902
which means rrcW_TRUNCATED.
This is happening when we are getting the old scratchpad
record to convert it to a syncq record. (SYSTEM.CHANNEL.SYNCQ)
This should not be a problem. However, the code is only looking
to see if there was a non zero return code and since it was
non-zero it fails the convert and stops the channel.
To get around the problem we asked the customer to delete the
receiver channel having the problem on AIX and redefine it to
remove the old scratchpad entry so we would not try to convert
it. This circumvented the problem.
LOCAL FIX:
DELETE AND REDEFINE THE CHANNEL.
PROBLEM CONCLUSION:
This problem has been fixed and the fix will be shipped in the
following PTFs:
A) MQSeries for V5.1
OS/2 U200121
Windows NT U200122
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY10830 COMPID: 5765B7300 REL: 500
ABSTRACT: TRIGGER MONITOR DOES NOT PASS MQTMC2 STRUCTURE IN SINGLE QUOTES
PROBLEM DESCRIPTION:
The MQSeries trigger monitor passes the MQTMC2 structure to
the triggered application in double quotes (ie " ") as opposed
to single quotes (ie ' '). This means that certain shell
constructs (for example, $$) may be interpreted rather than
being passed literally to the triggered application.
LOCAL FIX:
None. For those who wish to use constructs like $$, single
quotes are necessary to prevent premature interpretation of the
MQTMC contents.
PROBLEM SUMMARY:
This problem has been fixed for MQSeries V5.0 and MQSeries V5.1.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
and the fix will be shipped in the following PTFs:
A) MQSeries for V5.0
OS/2 U200116
Windows NT U200117
AIX U469658
HP-UX U469659
Sun Solaris U469671
B) MQSeries for V5.1
OS/2 U200121
Windows NT U200122
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY11729 COMPID: 5765D1400 REL: 220
ABSTRACT: USER LEVEL CLIENT RPC CONTACTING KRPC CAUSES A CANCEL PACKET
PROBLEM DESCRIPTION:
AFS adapt server shows a communication timeout
using the AFS 'dd' command. Executing an
iptrace shows a RPC cancel packet sent to
the server (the server is a DFS in this
particular case).
------
APAR: IY11864 COMPID: 5765D1400 REL: 220
ABSTRACT: SECD CORE DUMPING WHEN LOGIN_RESET IS RUN
PROBLEM DESCRIPTION:
Not able to handle invalid logins when an account
is disabled. Transarc has an login_reset utility
to reset the invalid logins. Right now the
utility cannot be run.
------
APAR: IY12600 COMPID: 5765D1000 REL: 220
ABSTRACT: DEADLOCK IN DCELFS.EXT
PROBLEM DESCRIPTION:
"In elbl_WaitForAnyTran () waiting on an empty tranAgeQ
could reasult in a deadlock on heavily loaded servers
if aggregate is being detached or a fileset operation is
waiting for all vnode operations to finish "
------
APAR: IY12665 COMPID: 5765D5101 REL: 111
ABSTRACT: THE NEW DOMAIN TERMINATED THEIR HAGS BECAUSE THE PREVIOUS
PROBLEM DESCRIPTION:
The new domain terminated their hags because the previous
nameserver came back and they think it is the better domain.
This should be the sequence of events that should occur when
cws has a network problem and it is the nameserver.
hags terminated on all the nodes is because of the inconsistency
of the node events in the hats layer. hats in the cws never knew
that he was detected as down by other nodes and he still thought
he is NS since he did not get update from hats about nodes statu
LOCAL FIX:
Re-establish the Domain by not node0
1>Kill subsystem on CWS hagsctrl -k, hatsctrl -k
2>Start subsystem on CWS hatsctrl -s, hagsctrl -s
PROBLEM SUMMARY:
If the Topology Services daemon is temporarily blocked,
or if a node's network adapter suffers a temporary glitch,
then this could result in a scenario where all other
nodes see this node as dead, while this node never sees
the remaining nodes die.
One case where this scenario could have adverse effects
occurs when the node in question is the Control Workstation
(which does not have a switch adapter) and the Group
Services's "Name Server" is located at that node.
In this case it is possible for this "node
event inconsistency" to result in the Group Services
daemon terminating in all the nodes, due to "domain
merging" or "non-stale proclaims".
There are several variants of the problem, according to
the location of the "Name Server" or Topology Services's
Group Leaders. In some instances, the Group Services
daemon terminates with a "Group services daemon received
a non-stale proclaim message" or "Group Services daemon
exit to re-join the domain" message in the AIX error log.
PROBLEM CONCLUSION:
The possibility of node event inconsistency caused
by daemon blockage or adapter/network glitch has been
eliminated with changes to the Topology Services
daemon. In case the Topology Services daemon is blocked
for a long period, the remote nodes will consider this
node dead, as before. With the changed code, this node
will also consider all the remote nodes as dead. This
will result in the Group Services daemon terminating
on this node (with a "merged domain"), but will avoid the
scenario where the Group Services daemon terminates
in a large number of nodes.
Eliminating the node event inconsistency also helps
preventing certain scenarios where the Group Services
domain becomes hung.
------
APAR: IY12773 COMPID: 576554300 REL: 240
ABSTRACT: MPI THREADED LIBRARY DOES NOT ALLOW USER TO HANDLE SIGNALS.
PROBLEM DESCRIPTION:
In PE version 2.4 threaded MPI library, the user can not control
SIGQUIT, SIGINT, etc. This capability is available in PE 3.1
but the customer insists that this be also available in PE 2.4.
Development in discussions with this customer has agreed to
provide this capability.
PROBLEM SUMMARY:
The relevant portion of Parallel Environment 3.1 code
code was retrofit to Parallel Environment 2.4 .
PROBLEM CONCLUSION:
Customer requested retrofit of Parallel Environment 3.1
capability to Parallel Environment 2.4,
specifically to register user signal handlers in
the threaded MPI library.
------
APAR: IY12788 COMPID: 5765D1400 REL: 220
ABSTRACT: FTS RELEASE -WAIT HANGS IF FTSERVER RPC FAILS
PROBLEM DESCRIPTION:
When using "fts release -wait", if one of the FTSERVER_ RPCs
fails, we still call fts_WaitForRepHostUpdates to do the -wait
part, because the anyRepChange flag is set. The wait, however,
is doomed to loop for a long time because the RPCs that trigger
the release did not succeed.
------
APAR: IY12909 COMPID: 5765D5100 REL: 311
ABSTRACT: CRASH CSS_KE_CLEAR_PG_TBL_ADDRS WITH 2 FAULT SERVICE DAEMONS
PROBLEM DESCRIPTION:
The switch fault service daemon can cause a node crash in
css_ke_clear_pg_tbl_addrs() if two fault service daemons are
running at the same time.
PROBLEM SUMMARY:
A node crash can occur if more than one fault service daemon
is started. There is a timing hole in which this can happen
if
more than one rc.switch is run at the same time.
PROBLEM CONCLUSION:
The script, rc.switch, has been changed to prevent more than
one fault_service_Worm_RTG's from being started, which
can cause a node to crash.
------
APAR: IY13007 COMPID: 5765B7300 REL: 510
ABSTRACT: MQSERIES WITH ORACLE 8.1.6 ON AIX HAS PROBLEMS WHEN THE
PROBLEM DESCRIPTION:
MQSeries with Oracle 8.1.6 on AIX has problems when the
XA switch file is unloaded .
PROBLEM CONCLUSION:
This problem has been fixed and the fix will be shipped in PTF
U471218 for CSD06.
------
APAR: IY13229 COMPID: 5765B7300 REL: 510
ABSTRACT: SESSIONS ARE TERMINATED AT THE TIMING OF 4 GIGA MM
PROBLEM DESCRIPTION:
all sessions are terminated at the timing of the timer at 4giga
mm.
PROBLEM SUMMARY:
MQSeries sessions went into indefinite wait. The problem was
caused by a flawed calculation in xcsQueryTime which was meant
to calculate the overflow when multiplying the number of seconds
since the epoch by 1000 to get the number of milliseconds. This
led xcsQueryTime to return inaccurate times when near the 4G
millisecond wrap time relative to the epoch. This in turn meant
that it would seem as if time had gone backwards and so the wait
interval was reset to 30 seconds despite any previous waiting
that had been performed. This continued and so the wait
interval never elapsed.
The problem has been fixed. The new calculation is done using
shifting operations and is more accurate.
PROBLEM CONCLUSION:
This problem has been fixed and the fix will be shipped in the
following PTFs for CSD06:
A) MQSeries for V5.1 CSD06
OS/2 U200121
Windows NT U200122
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY13399 COMPID: 5765B7300 REL: 510
ABSTRACT: MEMORY CORRUPTION IN QMANAGER
PROBLEM DESCRIPTION:
QMGR fails to start due to memory corruption . some of the
ProbeIDs seen were Probe Id :- AQ062000
Probe Id :- ZX000001
Probe Id :- KN002003
Probe Id :- AT013002
PROBLEM SUMMARY:
The performance event locking in kqiPutMsgSegments was
incorrect.
The problem has been fixed.
PROBLEM CONCLUSION:
This problem has been fixed and the fix will be shipped in the
following PTFs for CSD06:
A) MQSeries for V5.1
OS/2 U200121
Windows NT U200122
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY13426 COMPID: 5765D6100 REL: 210
ABSTRACT: KEY_COUNT GETTING QUITE LARGE WHEN THREAD OBJECT ALLOCATED
PROBLEM DESCRIPTION:
KEY_COUNT GETTING QUITE LARGE WHEN THREAD OBJECT ALLOCATED
PROBLEM SUMMARY:
This is a DCR APAR to change the design of the thread key
array so that the size of the array will shrink more
effeceintly when threads are deleted. The problem that
necessitated a change was that when the Schedd was up for
over a week the time it would take for jobs to start
running would increase. This is because the key count was
growing quite large. When a thread object is allocated
there is a loop that will get run "key_count" times. The
larger the key count the longer it would take threads to
get allocated thus it would take longer to start the job. A
code error was also found in the Schedd. When "llq -x -u
userlist" is run it increments the job reference count, but
never decrements it for jobs not belonging to users on the
userlist. This prevents jobs from having their thread
objects deleted which causes the thread key count to
continually increase.
PROBLEM CONCLUSION:
The code to reduce and reuse the thread keys has been
changed. The Schedd has been changed so that running "llq
-x -u userlist" does not leave job references to jobs not
belonging to user on the userlist. The result of these
changes will prevent jobs from taking longer to start after
LoadLeveler has been running for a while.
------
APAR: IY13534 COMPID: 5765C3403 REL: 430
ABSTRACT: IN NFSO, NFS_TCP_SOCKSIZE DOES NOT TUNE NFS TCP SOCKET SIZE
PROBLEM DESCRIPTION:
Even if the nfs_tcp_socksize is changed, the tcp socket buffer
size remains unaffected.
PROBLEM CONCLUSION:
The nfs_socksize is changed to nfs_tcp_socksize in the
clnt_tcp_init code.
------
APAR: IY13563 COMPID: 5765D1400 REL: 220
ABSTRACT: LOST I/O BUFFER DIAGNOSTICS
PROBLEM DESCRIPTION:
We believe that the VMM may have lost our buffer after we called
(*iodone)() or because it may have never handed us this buffer.
Based on this theory, the following changes were made in order t
narrow down the problem.
We have added some explicit settings of bp->b_error in parts of
code that manipulate the buffers for episode I/O. Each time we e
piece of the code that should be responsible for the dispositio
buffer for episode I/O we set a differnet innocuous error code i
bp->b_error field.
------
APAR: IY13753 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: FORMAT STRING VULNERABILITY IN LOCALE SUBSYSTEM
PROBLEM DESCRIPTION:
Format String Vulnerability in locale Subsystem
------
APAR: IY13888 COMPID: 5765C3403 REL: 430
ABSTRACT: PERFORMANCE IS NOT OPTIMIZED FOR HARDWARE MULTITHREADING
PROBLEM DESCRIPTION:
Performance is not optimized for hardware multithreading.
PROBLEM CONCLUSION:
Recognize in idle stealing code that under HMT cpus are paired.
------
APAR: IY13956 COMPID: 576565400 REL: 220
ABSTRACT: PTX SHOULD SHOW REASONABLE PROCESS NAMES
PROBLEM DESCRIPTION:
Current PTX agent displays the process names which is
truncated to 16 characters. And the displayed process
name is the name dierctly issued by the user on the
command line. This doesn't seems good all the time.
In Some case there might be symbolic links and the
process can change its name initially after the
execution starts. The ps -eaf command does gives more
reasonable process name. PTX agent should display
the same process name which is displayed by ps command.
PROBLEM SUMMARY:
Current PTX agent displays the process names which is
truncated to 16 characters. And the displayed process
name is the name dierctly issued by the user on the
command line. This doesn't seems good all the time.
In Some case there might be symbolic links and the
process can change its name initially after the
execution starts. The ps -eaf command does gives more
reasonable process name. PTX agent should display
the same process name which is displayed by ps command.
PROBLEM CONCLUSION:
Modified code to print appropriate process name.
------
APAR: IY14037 COMPID: 5765D5100 REL: 311
ABSTRACT: SYSCTL EFENCE FOR NON-ROOT USER DOESN'T WORK PROPERLY
PROBLEM DESCRIPTION:
When a authorized non-root user issues sysctl Estart, Efence, or
Eunfence, the command works if and only if there is a currrent
root.admin kerberos ticket. This should not be necessary
according to our documentation.
PROBLEM SUMMARY:
When an authorized non-root user uses sysctl to issue
Efence, Eunfence or Estart commands, the commands
work only when there is a current root.admin kerberos
ticket. The root.admin kerberos ticket should not be
necessary.
PROBLEM CONCLUSION:
The Efence, Eunfence and Estart commands were changed
to obtain a kerberos ticket before issuing any remote
commands.
------
APAR: IY14045 COMPID: 5765B7300 REL: 500
ABSTRACT: CHECK CHANNEL PROCESS WHEN ADDING/RETRIEVING CHANNEL STATUS
PROBLEM DESCRIPTION:
When an LU6.2 connection ends, SNA sends a SIGUSR1 to the
channel (SDR & RCVR) process. This causes the processes to end
without executing their atexit() function cccExitHandler, so
rrxReleaseRQMResource is not run and the conversation is not
freed. The result is that the channel status table contains an
invalid entry for the channel. This in turn can cause RCVR
channels not to start with an error "AMQ9514: Channel 'XXX' is
in use.", and SDR channels to display a status of RUNNING
instead of 'status not found'. When adding a status entry on
starting a chanel(rriAddStatusEntry), and retrieving a status
entry for any purpose (rrxAccessStatusEntry,
rrxQueryStatusEntry, rrxGetNextStatusEntry), check that the PID
in the status entry is still active. If not, reset the InUse
flag on the status entry.
LOCAL FIX:
none.
If possible, change to tcpip , but sometimes i might not be poss
ible.
PROBLEM SUMMARY:
This problem has been fixed for MQSeries V5.0 and MQSeries V5.1.
PROBLEM CONCLUSION:
This problem has been fixed in MQSeries V5.0 and MQSeries V5.1
and the fix will be shipped in the following PTFs:
A) MQSeries for V5.0 CSD08
OS/2 U200116
Windows NT U200117
AIX U469658
HP-UX U469659
Sun Solaris U469671
B) MQSeries for V5.1 CSD06
OS/2 U200121
Windows NT U200122
AIX U471218
HP-UX (V10) U471240
HP-UX (V11) U471444
Sun Solaris U471246
------
APAR: IY14215 COMPID: 576554300 REL: 240
ABSTRACT: MPI_ABORT DOES NOT PASS RETURN CODE
PROBLEM DESCRIPTION:
MPI_ABORT does not pass the return code back as the
MPI documentation states it should.
PROBLEM SUMMARY:
Problem was that MPI_Abort allows the caller to specify
a return code that is to become the job return code. PE
logic for generating return codes would deliver an rc
related to the mechanism used to bring down the job rather
than the rc the call specified.
PROBLEM CONCLUSION:
PE must take the return codes from a set of MPI tasks and
pick one return code to report for the job as a whole. This
is done by taking the MAX across the tasks and delivering
the low order 8 bits of this as job return code. When a job
is brought down by a user call to MPI_Abort, only rc
arguments to MPI_Abort calls should be input to the MAX.
That is now what happens. If no MPI_Abort call is involved
in ending the job, PE works as always. If 1 or more
MPI_Abort calls is made, PE will use the MAX accross these
as return code. (Note that if several tasks call MPI_Abort,
there may be only one that actually lasts long enough to do
so because when the first task reaches MPI_Abort, the job
starts going down and may end before other tasks reach the
call.
------
APAR: IY14578 COMPID: 5765D1400 REL: 220
ABSTRACT: RPC_RESTRICTED_PORTS IGNORED IF RPC_SUPPORTED_PROTSEQS SET
PROBLEM DESCRIPTION:
RPC_RESTRICTED_PORTS IGNORED IF RPC_SUPPORTED_PROTSEQS SET
------
APAR: IY14615 COMPID: 576552900 REL: 240
ABSTRACT: FSD <-> USER-SPACE APP HANDSHAKES CAN OVERLAP
PROBLEM DESCRIPTION:
If the timing is just right, there is the potential for two
consecutive fsd <-> user-space app handshakes to overlap. This
typically results in the app getting killed (SIGTERM+SIGKILL),
but the problem can manifest in other ways. One possible overlap
scenario (which was seen at a customer) is on a dist pt:
handshake 1 on DB-UPDATES
handshake 2 on LOAD-ROUTES
PROBLEM SUMMARY:
There is the potential for two consecutive fault service
daemon and user-space application handshakes to overlap.
This typically results in the application being killed.
It is also possible for the handshakes to overlap on a
distribution point:
handshake 1 on DB_UPDATES
handshake 2 on LOAD_ROUTES
This problem only occurs on the SP (TBS) Switch.
PROBLEM CONCLUSION:
A 1.25 second delay is added to the beginning of
quiesce_protocols() to ensure the user-space application
gets a chance to complete a previous handshake. The
user-space MPCI lib completes the handshake by polling a
kernel flag once per second. The addition of a 1.25
second delay will not cause artificial timeouts because
the minimum broadcast timeout is 28 seconds.
------
APAR: IY14663 COMPID: 5765D5100 REL: 311
ABSTRACT: PSSP INSTALL CHANGES /VAR/ADM PERMISSIONS
PROBLEM DESCRIPTION:
As part of the PSSP install procedure, the permissions on the
/var/adm directory are changed from root adm to bin bin. This
stops the adm user from writing to the directory, breaking
system accounting.
LOCAL FIX:
Adm user can be put into bin group to allow access to directory.
PROBLEM SUMMARY:
As part of the SP install procedure, the ownership of the
/var/adm directory is changed from root:adm to bin:bin.
The SP code should not modify the ownership of this
directory, as some customer applications are dependent on
the correct ownership.
PROBLEM CONCLUSION:
Modified the packaging files for ssp.basic and ssp.sysctl,
so that the ownership of /var/adm would return to root:adm.
------
APAR: IY14690 COMPID: 5765D1500 REL: 220
ABSTRACT: LOCKED FILESET MESSAGES TO BE PRINTED ON STDOUT INSTEAD OF
PROBLEM DESCRIPTION:
Locked fileset message to be printed on stdout instead of
stderr
------
APAR: IY14703 COMPID: 5765D1700 REL: 220
ABSTRACT: SECD. CORE DUMP
PROBLEM DESCRIPTION:
EPAC values which are not what they should be
(incorrect data, values out of range)that leads to
believe that the EPAC information is somehow being
corrupted before it is sent to the AIX DCE Security Servers.
------
APAR: IY14756 COMPID: 5765C3403 REL: 430
ABSTRACT: LOGGER COMMAND DUMPS CORE
PROBLEM DESCRIPTION:
When given some longer messages, the logger
command dumps core catching SIGSEGV signal.
This is observed in AIX 4.3.3
PROBLEM SUMMARY:
"logger" command dumps core when the input
argument contains the pattern "%<d>n" in it.
PROBLEM CONCLUSION:
"logger" command inturn calls syslog()
function for which only two arguments were
being passed. The problem is solved by
passing three parameters to syslog(), "%s"
being the format specifier, in printf() style.
------
APAR: IY14768 COMPID: 5765D1700 REL: 220
ABSTRACT: ADDITIONAL DEBUG INFORMATION IN RPC CODE TO TRACK THE SCTE
PROBLEM DESCRIPTION:
It was found that the 'refcnt' member of the 'key_info' struct
in the server connection table entry(scte) struct was zeroed
out resulting in the 'refcnt' value becoming zero which results
in a call to assert causing a panic, the debug messages that
have been added log serviceability at level 1, they track the
changes in the value of 'refcnt' in the 'key_info' struct and
finally also print the address of the 'key_info' struct.
------
APAR: IY14818 COMPID: 5765D1500 REL: 220
ABSTRACT: FILESET MOVE FAILS: FTSERVER_DELETETRANS FAILED TO END TRANS 24,
PROBLEM DESCRIPTION:
"fts move" fails. Performance is very slow, and the move tran
eventually fails as follows:
FTSERVER_DeleteTrans failed to End trans 24
Releasing token (2) for fileset 0,,25 from <cell>
Releasing token on file server...done
Freeing fx binding, princ = hosts/<host>
Error in move: Transaction doesn't exist
------
APAR: IY14876 COMPID: 5765C3403 REL: 433
ABSTRACT: DTPRINTINFO -ALL CLICK ON THE "+" SIGN IN FRONT OF ANY PRINTER
PROBLEM DESCRIPTION:
/usr/dt/bin/dtprintinfo -all.
Click on the "+" sign in front of any printer will always
display additionnal information for the last printer displayed
instead of displaying additionnal information for the selected
printer
PROBLEM SUMMARY:
When user clicks on the + button associated with each Queue
under dtprintinfo -all flag, the corresponding queue info
doesnot open. Instead the info for the last queue opens up.
PROBLEM CONCLUSION:
Memory allocation was done only for the 1st queue in the list
and it was deleted at the end, but the same location was used
by the subsequent queues in the list.
So memory was allocated each time for each queue properly.
------
APAR: IY14967 COMPID: 5765C3403 REL: 430
ABSTRACT: GXT6000P INDIVIDUAL PIXELS INCORRECTLY RENDERED LOOKS LIKE DUST
PROBLEM DESCRIPTION:
Dust particles are displayed in certain area of a model. Pixel
level holes or overpaints are displayed on irregular pixel
position for each redraw. It looks like a dust particle. It
emerges not on whole screen but on certain area of a model.
Recreating model:
11100PRAA000001,22L 1X-3 BLOCK COMP,CYLN PRA
AE0041.model
Recreating scenario:
2. Turn to SHD mode. 3. Rotate as illustrated in the file
"step1.gif". 4. Zoom in as illustrated in the file
"step2.gif". 5. Problem is the dust particles on the
screen that
you can see in "step2.gif".
PROBLEM SUMMARY:
Dust particles are displayed in certain area of a model. Pixel
level holes or overpaints are displayed on irregular pixel
position for each redraw. It looks like a dust particle. It
emerges not on whole screen but on certain area of a model.
PROBLEM CONCLUSION:
Work around of the "trimming precision" problem involves five
(5) passes through all concave polygons which have been
flagged as "maybe clipped"; ie, those concave polygons which
are neither trivially accepted nor trivially rejected. Three
stencil masks are created in the first three passes, and the
best two out of three stencil bits are used to render the
final image in two additional passes.
------
APAR: IY14977 COMPID: 576565400 REL: 220
ABSTRACT: WLM / SNMP ERROR IN XMSERVD
PROBLEM DESCRIPTION:
xmservd reports error in its log file when it
tries to add the WLM stats in the smux tree.
This leads to stop of xmservd when a kill -2 is
issued to generate a mib file.
PROBLEM CONCLUSION:
Fixed the ASSN number of WLM Stats
------
APAR: IY15019 COMPID: 5765D5100 REL: 311
ABSTRACT: SWITCH INFO IN SDR NODE CLASS NOT UPDATED WHEN UPGRADING FROM LO
PROBLEM DESCRIPTION:
When upgrading from a low-cost switch to a normal switch, the
switch information in the SDR node class (switch node number,
switch chip, and switch port) does not get updated as it should.
Low-cost switches use a different numbering scheme (namely they
always start at 0 and increment by 1). Running SDR-config -u
after the step to run hmreinit in the procedure for upgrading
the switch would solve this and is the workaround. Not sure if a
documentation change or code change is the way to go at this
point.
LOCAL FIX:
Run SDR_config -u
PROBLEM SUMMARY:
When upgrading from an SP Switch-8 to an SP Switch, the
switch data in the SDR Node class is not updated.
The Syspar_map is updated correctly, but the Node class
is not modified. As a result, the switch_node_number
is incorrect.
PROBLEM CONCLUSION:
Modified SDR_config to update the Node class when the
type of switch in the frame has changed. This will ensure
that the switch_node_numbers are correct.
------
APAR: IY15067 COMPID: 5765C3403 REL: 430
ABSTRACT: TRACEROUTE TO IPV6 DOESN'T WORK
PROBLEM DESCRIPTION:
IPv6 traceroute does not work when the target machine
is running AIX
PROBLEM SUMMARY:
Problem fixed by restoring fields in IPv6 header
before sending ICMP Error packet
------
APAR: IY15074 COMPID: 5765B8100 REL: 220
ABSTRACT: ATTEMPT TO DELETE APPPROFILE WITH ASSOCIATED MESSAGE IN DATABASE
PROBLEM DESCRIPTION:
When application profile with an existing voice message in the
GUI and database has the message deleted through the GUI (marked
for deletion in the database) so that the application profile
now appears to have no messages in the GUI, an attempt to delete
the application profile results in unhelpful error messages.
The error message is different through APPMAN and UPE. In UPE
error is "Unable to delete the data item. Database server unava-
ilable. Access to data including HELP system not possible at
this time.". APPMAN gives an SQL error (SQL0532N)
PROBLEM SUMMARY:
ATTEMPT TO DELETE APPPROFILE WITH ASSOCIATED
MESSAGE IN DATABASE WHICH IS MARKED FOR DELETION PRODUCES
UNHELPFUL ERROR MESSAGE.
PROBLEM CONCLUSION:
In Application profiles (UPE) the correct
error message was selected by updating the number in the
sysdef.h file.
Addionally the error message has been updated to say that
messages may be pending deletion.
In Appman an extra clause was put in to detect the particular
error case and replace the error message with the one used in
UPE.
------
APAR: IY15126 COMPID: 5765C3403 REL: 433
ABSTRACT: ROUTE ADD -EXTRA ARGUMENT OVERWRITES NETMASK
PROBLEM DESCRIPTION:
When the Customer uses route add with extra arguments,
it overwrites the netmask. No error message is issued
for the extra arguments.
------
APAR: IY15264 COMPID: 5765D6100 REL: 210
ABSTRACT: LLSUBMIT CORE DUMPS WHEN SUBMITTING NQS SCRIPT JOBS
PROBLEM DESCRIPTION:
llsubmit core dumps when submitting the following job :
$-lt 600
$-s /bin/csh
echo "LOADL_JOB_NAME" $LOADL_JOB_NAME
echo "LOADL_STEP_ID" $LOADL_STEP_ID
echo "SHELL" $SHELL
set -x
date
pwd
LOCAL FIX:
add $-o /PATH/
to the job command file
PROBLEM SUMMARY:
LLsubmit core dumps when submitting
NQS script jobs that do not specify a
-o for standard out.
PROBLEM CONCLUSION:
LLsubmit will now submit
NQS script jobs even if it do not
specify -o for standard out.
------
APAR: IY15309 COMPID: 5765C3403 REL: 430
ABSTRACT: INFOVISTA APPLICATION CRASHES WITH 888-102-700-0C0
PROBLEM DESCRIPTION:
Application InfoVista crashes with 888-102-700-0c0
IAR: .v_pfprot+154 (00071dac): tweqi r0,0x0
LR: .v_pfprot+130 (00071d88)
0044be18: .v_dpfget+264 (000a4728)
0044be78: ds_flih64+2ac (00147f64)
This problem can only occur in a threaded application running on
an MP machine.
Here is an excerpt from the system trace (with the real TIDs
replaced by 1 and 2, to correspond to description above
and with the timestamps omitted):
ID TID INFO
3DA 2 mmap addr=600D0000 len=1000 prot=0001 PROT_READ
flags=0101 MAP_SHARED MAP_FIXED fd=0005
106 1 dispatch: CPUID=0
(This is the thread that writes to page 0x600D0000.)
106 2 dispatch: CPUID=1
3DD 2 munmap addr=600D0000 len=1000
3DA 2 mmap addr=600D0000 len=1000 prot=0003 PROT_READ
PROT_WRITE flags=0101 MAP_SHARED MAP_FIXED fd=0005
PROBLEM SUMMARY:
Application crashes with 888-102-700-0c0
IAR: .v_pfprot+154 (00071dac): tweqi r0,0x0
LR: .v_pfprot+130 (00071d88)
0044be18: .v_dpfget+264 (000a4728)
0044be78: ds_flih64+2ac (00147f64)
This problem can only occur in a threaded application
running on an MP machine.
Here is an excerpt from the system trace (with the real TIDs
replaced by 1 and 2, to correspond to description above
and with the timestamps omitted):
ID TID INFO
3DA 2 mmap addr=600D0000 len=1000 prot=0001 PROT_READ
flags=0101 MAP_SHARED MAP_FIXED fd=0005
106 1 dispatch: CPUID=0
(This is the thread that writes to page 0x600D0000.)
106 2 dispatch: CPUID=1
3DD 2 munmap addr=600D0000 len=1000
3DA 2 mmap addr=600D0000 len=1000 prot=0003 PROT_READ
PROT_WRITE flags=0101 MAP_SHARED MAP_FIXED
fd=0005
Senario of cause of the crash:
Thread 1: Stores to write-protected page
(causes protection exception)
Thread 2: Calls munmap() on the referenced page
Thread 1: Runs page-fault handler. Acquires SCB lock to
perform processing.
Thread 2: Calls mmap() to remap the page with read-write
access. This system call spins, trying to acquire
the SCB lock.
Thread 1: Page-fault looks up mmap information for
referenced page, but it does not exist. An
assert() in the code crashes the system.
PROBLEM CONCLUSION:
The assert is invalid. The fix is to simply remove the assert
so that a user application will core dump if hit on such
timing windows.
------
APAR: IY15341 COMPID: 5765C3403 REL: 430
ABSTRACT: LSVG DUMPS CORE WITH VGS HAVING MANY LVS
PROBLEM DESCRIPTION:
lsvg is dumping core with VGs having
large number of LVs.
A pointer to logical view of lvs is
not freed properly which is causing
coredump, in the case of VGs with many
LVs and disks.
PROBLEM SUMMARY:
lsvg dumping core on vgs with lot
of lvs
PROBLEM CONCLUSION:
freeing a pointer to logview structure
within a loop instead of freeing it
after the loop fixes the problem.
with each pass of the loop more than 1 Mb
gets allocated. If you have a lot of lvs, lsvg
eventually ends up using all the memory inside
that segment.
------
APAR: IY15343 COMPID: 5765B9501 REL: 310
ABSTRACT: LOW LEVEL MIGRATION DEADLOCK AFTER DISK FAILURE
PROBLEM DESCRIPTION:
low level migration deadlock after disk failure
PROBLEM SUMMARY:
GPFS hung after failure of a disk drive
PROBLEM CONCLUSION:
Fix locking error after the failure of a set of disks.
------
APAR: IY15344 COMPID: 5765B9501 REL: 310
ABSTRACT: QUOTA: ASSERT OCCURRED AFTER TOKEN LOSS
PROBLEM DESCRIPTION:
QUOTA: assert occurred after token loss
PROBLEM SUMMARY:
GPFS terminated on a node with an assert in quota code.
PROBLEM CONCLUSION:
Correct locking error in quotas.
------
APAR: IY15518 COMPID: 5765C3403 REL: 430
ABSTRACT: USER INTERACTION PROBLEMS WITH LARGE TRACEBACKS
PROBLEM DESCRIPTION:
usability of where commands reduced in large programs
PROBLEM SUMMARY:
dbx improvement for handling large tracebacks.
PROBLEM CONCLUSION:
Done by adding serial # to each function in traceback.
------
APAR: IY15538 COMPID: 5765C3403 REL: 430
ABSTRACT: GRAPHIGS BASED APPLICATION HANGS GXT6000P
PROBLEM DESCRIPTION:
CATIA applicaton hangs with GXT6000P adapter when
zooming in on a particular model that uses a
styled line with many vertices.
Incorrect rendering of the lines also occurs.
PROBLEM SUMMARY:
GRAPHIGS BASED APPLICATION HANGS GXT6000P
PROBLEM CONCLUSION:
Order the PTF for this APAR when availalble.
------
APAR: IY15574 COMPID: 5765C3403 REL: 430
ABSTRACT: PAX DOESN'T WORK WITH OPTION -PE
PROBLEM DESCRIPTION:
A Usage error occurs when the pax command is used without
"-x pax" option and when "-p e" options are used.
Eg: pax -rf test_file.pax -p e test_file1.in
PROBLEM CONCLUSION:
When No "-x " option is used it defaults to tar archive
creation. In such a case, global headers have no meaning. In
this problem, "globexthdr" fields in the code are accessed
which should not be the case. So i out a check to see if
global headers globalexthdr or exthdr are not NULL.
------
APAR: IY15594 COMPID: 576554300 REL: 240
ABSTRACT: XPROFILER CORE DUMPS WHEN GENERATING FLAT PROFILES.
PROBLEM DESCRIPTION:
For some application codes xprofiler will core dump when trying
to generate a flat profile. In addition, the core file could
be very large.
LOCAL FIX:
A temporary workaround is to use gprof to generate a flat
profile.
PROBLEM SUMMARY:
there is a mis-calculation in the tool's memory management,
an extra bite is written that, in some cases, confuses the
memory management and causes the tool, Xprofiler, to fail.
PROBLEM CONCLUSION:
the mis-calculation in the tool's memory management is
corrected, which eliminate the possibility of any memory
overwriting, thus fixes the problem.
------
APAR: IY15597 COMPID: 5765C3403 REL: 430
ABSTRACT: CRASH IN SO_PIN_MORE_STACK CALLED FROM SOACCEPT+84
PROBLEM DESCRIPTION:
The system will crash with the stack trace showing
so_pin_more_stack and soaccept at the top.
PROBLEM CONCLUSION:
To take care of the race condtion between accept() and the
socket getting freed due to no response to all the keepalive
probes, increment the refcnt on the socket when the connection
is terminated so that the socket doesn't get freed while
accept() is working on it. accept() will do the freeing and
report ECONNABORTED error.
------
APAR: IY15605 COMPID: 5765C3403 REL: 430
ABSTRACT: MEMORY LEAK RESULTS ON SOME ZOOMED GLDRAWPIXEL CALLS
PROBLEM DESCRIPTION:
glDrawPixel calls on the GXT3000P may result in a memory leak
when performing a pixel zoom.
PROBLEM CONCLUSION:
Free all temprary working buffers when zooming pixel images.
------
APAR: IY15606 COMPID: 5765D5100 REL: 311
ABSTRACT: IF ADDING SWITCHED FRAME, FRESH INSTALLATION HAS SDR_CONFIG ERRS
PROBLEM DESCRIPTION:
In PSSP 3.1.1, SDR_config will display error messages on every
fresh installation when adding a switched frame. This problem
was fixed in Defect 63913; however, ONLY PSSP release levels at
Mohonk (3.2.0) or above were included in the fix. It must be
retrofixed at the 3.1.1 Release Level
LOCAL FIX:
In the fix for defect 63637, the Syspar_map was deleted
whenever aa frame with a switch was being added so that it could
be recreated. However, with a FRESH installation, there is NO
Syspar_map to delete, and this condition was not taken into
account. The fix for this defect involves checking to see if a
Syspar_map exists, and if so, delete it.
PROBLEM SUMMARY:
Issuing spframe to add a frame with a switch as part
of a new install, results in the following message:
SDR_config: 0016-724 SDR_config had some errors.
In addition, the following messages are written to the
SDR_config.log:
/usr/lpp/ssp/bin/SDRDeleteObjects: 0025-004 Item specified
for query, insertion or deletion was not found.
/usr/lpp/ssp/bin/SDRDeleteObjects Syspar_map
SDR_config: 0016-721 The above command failed with rc=4.
SDR_config: 0016-722 Failure while executing SDR command(s)
SDR_config: 0016-724 SDR_config had some errors.
The error messages are the result of SDR_config trying to
delete a non-existent Syspar_map. The frame is added
correctly and the Syspar_map is generated correctly, but
the messages should not be issued.
PROBLEM CONCLUSION:
When a frame with a switch is being added, SDR_config will
now check if a Syspar_map exists, and if so, delete it.
------
APAR: IY15611 COMPID: 576554300 REL: 240
ABSTRACT: TOTALVIEW DEBUGGER HANGS POE PROGRAMS
PROBLEM DESCRIPTION:
after the MPI_INIT is successful. Etnus has examined the
problem and believes the problem is not in totalview. The
problem occurs in both IP and US on the switch and on ethernet
IP. The primary data for this call is in PMR 23676 b487 if
this is not sufficient
LOCAL FIX:
Let MPI job start and run for a couple seconds before attaching
to the process, or get a level of AIX that does not interfear
with the original functionality.
PROBLEM SUMMARY:
At certain levels of AIX, totalview is hanging MPI programs
in MPI_INIT when the program is started under totalview.
PROBLEM CONCLUSION:
It was determined that AIX is changing the characterstics
of wait's action. A change in the operation of poe for a
previous APAR is now believed to make this wait
unneccessary. Customer testing on a failing level of AIX
proves that the fix eliminates the hang.
------
APAR: IY15613 COMPID: 5765C3403 REL: 430
ABSTRACT: BDIFF NOT FAILING WHEN /TMP FILESYSTEM IS FULL
PROBLEM DESCRIPTION:
bdiff was returning an exit code
of 0 when /tmp file system is full.
The command is not running to its
completion and eventhen it is not returning
correct exit status.Also, the command
is not giving any error message to
that effect.
PROBLEM SUMMARY:
Bdiff returns 0 when the command fails because /tmp
file system is full.It's also not returning any error message
when the command fails.
PROBLEM CONCLUSION:
Code was fixed to generate error message and
exiting with proper error code.
------
APAR: IY15614 COMPID: 5765C3403 REL: 430
ABSTRACT: /ETC/SHUTDOWN BLOCKED IN UMOUNT WHEN AUTOMOUNT FILES ARE USED
PROBLEM DESCRIPTION:
/etc/shutdown command is blocked while
umounting 'automounted' file systems
when LIBPATH or NLSPATH is set to the
automounted file systems.
PROBLEM SUMMARY:
/etc/shutdown blocked in umount when automounted
filesystems are used, and LIBPATH or NLSPATH is
set to those filesystems.
PROBLEM CONCLUSION:
Unsetting LIBPATH and NLSPATH inside the
/etc/shutdonw script will solve the problem.
------
APAR: IY15615 COMPID: 5765C3403 REL: 430
ABSTRACT: NICE VALUE FOR CRON JOBS CANNOT BE REDUCED BELOW A DEFAULT
PROBLEM DESCRIPTION:
The nice value for the cron jobs could not
be reduced below the default set value of 20.
The nice value for all the cron jobs is defined
in the file /var/adm/cron/queuedefs.
To increase the priority the entry in this
queuedefs file should be entered as a negative.
Cron is not able to take negative values as
its nice value.
PROBLEM SUMMARY:
Nice value of cron jobs cannot be reduced below
the default value
PROBLEM CONCLUSION:
The /var/adm/cron/queuedefs file takes the entries for
the cron jobs, like the nice value, max. number of jobs
in the queue.
The negative entries in the queuedefs file for the nice
value is not recognised by the cron daemon.
This problem was solved and now cron jobs can take nice
values less than the default set value.
------
APAR: IY15618 COMPID: 5765C3403 REL: 430
ABSTRACT: LSPV NOT CONSISTENT IN DISPLAYING 32 DIGIT PVIDS
PROBLEM DESCRIPTION:
From AIX 4.3.3-ML 06 onwards lspv command
is not reporting the PVIDS conssitently in
the same manner.
Incase, if the last 16 digits of the PVID
are zeroes, then lspv <noarguments> is
displaying only first 16 digits as the PVID.
lspv <pvname> is not doing so. It is
diplaying the whole 32 digits as PVID
even if last 16 digits are zero.
PROBLEM SUMMARY:
If the last 16 digits of pvid are zeroes,
then lspv should be consistent in returning
the first 16 digits only as pvid
PROBLEM CONCLUSION:
lspv <withno pvname> is already working as intended
code changed to see that lspv <pvname> also works in the
same manner
------
APAR: IY15628 COMPID: 5765C3403 REL: 430
ABSTRACT: BLIST COUNT INCORRECT FOR D_MAP_LIST
PROBLEM DESCRIPTION:
DMA NO_SUCC AIX error logs and device driver failures
PROBLEM SUMMARY:
DMA NO_SUCC AIX error logs and device driver failures.
PROBLEM CONCLUSION:
Keep the vlist and blist counts the same when doing DMA lists
------
APAR: IY15661 COMPID: 5765C3403 REL: 430
ABSTRACT: HANG DURING SYSTEM TO SCREEN BLIT ON GXT6000P
PROBLEM DESCRIPTION:
An OpenGL application can hang when it is executing a large
system to screen blit to the GXT6000P graphics adapter.
PROBLEM CONCLUSION:
Reserve two words at the end of every user queue for the noop
command that follows the blit data.
------
APAR: IY15664 COMPID: 5765B8100 REL: 220
ABSTRACT: WAITEVENT NOT SETTING SYSVARS CORRECTLY FOR JAVA APPS
PROBLEM DESCRIPTION:
When using a Java Applications, a State Table WaitEvent does
not set the System Variables in the same manner as it would if
originally called from a StateTable based application.
PROBLEM SUMMARY:
WAITEVENT NOT SETTING SYSVARS CORRECTLY FOR
JAVA APPS
PROBLEM CONCLUSION:
Instead of passing a local variable to
vp_get_server_event, the global variable CHP_server_event_data
is passed thus ensuring the system variable SV232, 233, 234
are set.
------
APAR: IY15673 COMPID: 5765D5100 REL: 311
ABSTRACT: SPFRAME'S LIMIT NOT BEING ADHERED TO EXCEEDING LIMIT OF 128
PROBLEM DESCRIPTION:
The spframe command is allowing more entries than the maximum
limit of 128.
PROBLEM SUMMARY:
spframe should only support frame numbers between 1 and 128.
It is not currently preventing larger frame numbers from
being entered. A flag which is used to prevent this was
being checked incorrectly.
PROBLEM CONCLUSION:
Corrected the frame number checking within spframe so that
only frames 1 through 128 can be entered.
------
APAR: IY15682 COMPID: 5765C3403 REL: 430
ABSTRACT: INSTFIX: 0503-005 ON 432 PTFS
PROBLEM DESCRIPTION:
Users installing by fix can get 0503-005 instfix:
The format of the toc file is invalid on some 432
ptfs.
LOCAL FIX:
Use installp/smitty install_all or edit .toc
PROBLEM SUMMARY:
installing by fix can get 0503-005 instfix:
The format of the toc file is invalid on some
432 ptfs
PROBLEM CONCLUSION:
instfix will ignore invalid line in .toc
TEMPORARY FIX:
edit .toc, insall using different method
------
APAR: IY15686 COMPID: 576565400 REL: 220
ABSTRACT: AGENT DOESN'T SHOW DISKSTATS WHEN IOSTAT FLAG IS NOT SET
PROBLEM DESCRIPTION:
when iostat flag is turned off , ptx agent
does not display disk stats. (only zeros are displayed)
------
APAR: IY15701 COMPID: 5765C3403 REL: 430
ABSTRACT: SMOOTH POINTS DRAWN TOO BIG
PROBLEM DESCRIPTION:
OpenGL applications on the GXT6000P draw large non-smooth
points when the point size is set to a value larger than 1.0
and smooth points are enabled. According to the queriable
smooth point size range, the GXT6000P smooth point size should
be clamped to the range 1.0 to 1.0.
PROBLEM CONCLUSION:
Correct OpenGL SW to correctly clamp the smooth point size to
the implentation dependent range of 1.0 to 1.0.
------
APAR: IY15720 COMPID: 5765C3403 REL: 430
ABSTRACT: FIND -FSTYPE JFS SEARCHES NFS MOUNTS AND HITS PERFORMANCE
PROBLEM DESCRIPTION:
The find command with -fstype option on "jfs" file
systems takes long time in searching. The output
of file command however gives required results.
The performance hit on this command option is due to
the reason that the command searches into "nfs" mounts
also.
PROBLEM SUMMARY:
The find command with -fstype jfs option takes very long time
to give the desired output..
PROBLEM CONCLUSION:
The source code has been modified accordingly so as to avoid
find searching into nfs mount for -fstype jfs option.
------
APAR: IY15765 COMPID: 5765C3403 REL: 430
ABSTRACT: KERNEL & FSCK REPORT INCOR ERRORS FOR ALMOST FULL BIG FILE FS
PROBLEM DESCRIPTION:
If small files are added to a big file enabled filesystem which
is almost full incorrect error messages can be reported in the
system error log and fsck output. The small files could be
reported as corrupt when they are actually ok.
LOCAL FIX:
An optional fix would be to make sure big file filesystem
ALWAYS has 128k free space at all times.
or
to make sure you never create a file smaller then 32k
in the big file fileystem.
PROBLEM SUMMARY:
If small files are added to a big file enabled filesystem which
is almost full incorrect error messages can be reported in the
system error log and fsck output. The small files could be
reported as corrupt when they are actually ok.
PROBLEM CONCLUSION:
Correct the BADBLK calculation for disk addresses.
------
APAR: IY15773 COMPID: 576565400 REL: 220
ABSTRACT: %COMP FREEZES IN 100
PROBLEM DESCRIPTION:
The %comp value of Mem/Real statistics freezes at 100.
PROBLEM CONCLUSION:
Made modifications in the code to make all the VMM related
statistics to get VMM updates simultaneously.
------
APAR: IY15795 COMPID: 5765B9500 REL: 120
ABSTRACT: MMCONFIG FAILING:CREATES 0 LENGTH FILE MMFDRFS IN SDR
PROBLEM DESCRIPTION:
mmconfig failing: creates 0 length file mmsdrfs in SDR
PROBLEM SUMMARY:
GPFS did not correctly handle a zero length
SDR file.
PROBLEM CONCLUSION:
Add some messages to validate the SDR file.
------
APAR: IY15828 COMPID: 576565400 REL: 220
ABSTRACT: FDPR: LARGE 64-BIT FILES CREATE BACK EXECUTABLES
PROBLEM DESCRIPTION:
there is a problem with the -regs_release optimization
which has the potential for certain large 64-bit applications
and their executables to be corrupted.
PROBLEM CONCLUSION:
FDPR updates for very large 64-bit applications
------
APAR: IY15833 COMPID: 576565400 REL: 220
ABSTRACT: SSVMON INTERNAL ERROR: 0X0007 : GET_NUMPINNED_VSID: OCCURENCE=1
PROBLEM DESCRIPTION:
svmon reportsinternal error, get_numpinned_vsid
PROBLEM CONCLUSION:
corrections needed to support large systems/workloads
------
APAR: IY15848 COMPID: 5765C3403 REL: 430
ABSTRACT: DBX THREAD LIST SCANNING CAUSES CORRUPTION
PROBLEM DESCRIPTION:
pthdebug scan of pthread causes contention
------
APAR: IY15890 COMPID: 5765C3403 REL: 430
ABSTRACT: NEEDS SUPPORT OF HYPHEN IN SHARE NAME
PROBLEM DESCRIPTION:
cannot add file/printer shares with non-alphanumeric characters
in printer queue name that supported by the AIX
PROBLEM CONCLUSION:
allow printer queue name's charaters that supported by AIX
------
APAR: IY15897 COMPID: 5765C3403 REL: 430
ABSTRACT: APAR IY14484 DIDN'T CLEAN UP ALL THE DCE CREDENTIAL FILES
PROBLEM DESCRIPTION:
Apar IY14484 made it possible to clean up the dce credential
files in /var/dce/security/creds after a ftp session has
ended. However, it didn't take into account someone using
dce (integrated login) and not gssapi.
PROBLEM CONCLUSION:
Create new function cleanup_cred_file() to clean up the
remaining credential files at session end.
------
APAR: IY15902 COMPID: 5765C3403 REL: 430
ABSTRACT: CAN'T PICK SOFAS
PROBLEM DESCRIPTION:
Can't pick sofas.
PROBLEM CONCLUSION:
surface_properties was not being set in gP_PROCESS_SOFA.
------
APAR: IY15911 COMPID: 5765D5100 REL: 311
ABSTRACT: FSD <-> USER-SPACE APP HANDSHAKES CAN OVERLAP
PROBLEM DESCRIPTION:
If the timing is just right, there is the potential for two
consecutive fsd <-> user-space app handshakes to overlap. This
typically results in the app getting killed (SIGTERM+SIGKILL),
but the problem can manifest in other ways. One possible overlap
scenario (which was seen at a customer) is on a dist pt:
handshake 1 on DB-UPDATES
handshake 2 on LOAD-ROUTES
PROBLEM SUMMARY:
There is the potential for two consecutive fault service
daemon and user-space application handshakes to overlap.
This typically results in the application being killed.
It is also possible for the handshakes to overlap on a
distribution point:
handshake 1 on DB_UPDATES
handshake 2 on LOAD_ROUTES
This problem only occurs on the SP (TBS) Switch.
PROBLEM CONCLUSION:
A 1.25 second delay is added to the beginning of
quiesce_protocols() to ensure the user-space application
gets a chance to complete a previous handshake. The
user-space MPCI lib completes the handshake by polling a
kernel flag once per second. The addition of a 1.25
second delay will not cause artificial timeouts because
the minimum broadcast timeout is 28 seconds.
------
APAR: IY15918 COMPID: 5765C3403 REL: 430
ABSTRACT: PTYDD FAILS IN TSTART
PROBLEM DESCRIPTION:
The pty code calls tstart from redo_allocb with a timer
struct that has not been set up.
PROBLEM CONCLUSION:
The solution is to set up the timer structure before calling
tstart.
------
APAR: IY15929 COMPID: 5765C3403 REL: 430
ABSTRACT: GRAPHICS ERROR IN ERROR REPORT WHEN STARTING XTERM
PROBLEM DESCRIPTION:
If you start xwindows with the xinit command, then trying
to start an xterm from .xinitrc or .mwmrc file will
cause a Software error in the errorlog:
LABEL: GRAPHICS
Class: S
Type: PERM
Resource Name: LFTDD
PROBLEM CONCLUSION:
Modified xterm to use setupterm, instead of tgetent.
------
APAR: IY15930 COMPID: 5765C3403 REL: 430
ABSTRACT: TITLE BAR OF AIX IM SUBWINDOW SHOULD NOT APPEAR IN 64BIT APPS
PROBLEM DESCRIPTION:
Some input method popups have title bars when they shouldn't,
when the application is compiled in 64-bit mode.
PROBLEM CONCLUSION:
Correct a data type error in the code that sets the
WM property on the popup window.
------
APAR: IY15946 COMPID: 5765C3403 REL: 430
ABSTRACT: CATIA CONFERENCING FUNCTION ABENDS WITH A LARGE MODEL
PROBLEM DESCRIPTION:
The CATIA V4 conferencing function can abend with a large
model due to graPHIGS writing past the end of a buffer.
PROBLEM CONCLUSION:
Change the structure archive buffer allocation unit size to
an integer multiple of a doubleword size (8 bytes).
------
APAR: IY15951 COMPID: 5765C3403 REL: 430
ABSTRACT: POSSIBLE RACE CONDITION IN NFS_ASYNC_READDIR
PROBLEM DESCRIPTION:
In rare occations, it might happen that the threads waiting on
nfs_async_readdir wake up right after after the
async_rddir_queue.async_rddir_wait CV is signalled, execute and
release the vnode before nfs_async_readdir has a chance to put
a VNOP_HOLD() on it.
PROBLEM CONCLUSION:
Move the VNOP_HOLD() call to before the cv_signal() of
async_rddir_queue.async_rddir_wait in nfs_async_readdir.
------
APAR: IY15986 COMPID: 5765B8100 REL: 220
ABSTRACT: BASE CHANGES FOR MESSAGECENTER
PROBLEM DESCRIPTION:
DirectTalk base changes for MessageCenter
PROBLEM CONCLUSION:
Changes added for MessageCenter
------
APAR: IY15996 COMPID: 5765C3403 REL: 430
ABSTRACT: ATM 622 DSI WHEN RUN COMMAND CHDEV
PROBLEM DESCRIPTION:
system DSI when run rmdev, mkdev or chdev.
PROBLEM CONCLUSION:
only free the buffer own by the driver
------
APAR: IY16033 COMPID: 5765C3403 REL: 430
ABSTRACT: LOOK-AND-FEEL DIFFERS BETWEEN SCRIPTS
PROBLEM DESCRIPTION:
The look and feel differs on some of the SNAPP gui panels
because the OK and CANCEL buttons are not placed in the
same places.
PROBLEM CONCLUSION:
Put the OK and CANCEL buttons in the same place for each panel.
------
APAR: IY16035 COMPID: 5765C3403 REL: 430
ABSTRACT: SNAPPD DAEMON LOSES SECOND PING OUTPUT SOMETIMES
PROBLEM DESCRIPTION:
The ping test used to test the interface configured with
the SNAPP application may hang.
PROBLEM CONCLUSION:
Add a sleep to the perl script, so that the snappd daemon
has enough time to send the output before it gets wiped out.
------
APAR: IY16036 COMPID: 5765C3403 REL: 430
ABSTRACT: SET PAGING SPACE ERROR MESSAGE NOT DISPLAYED
PROBLEM DESCRIPTION:
Customer would not see the results of a failed set paging space
command
PROBLEM CONCLUSION:
Removed the command from the perl script
------
APAR: IY16039 COMPID: 576552900 REL: 240
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: IY16050 COMPID: 576552900 REL: 240
ABSTRACT: NODE CRASH ON DETACH OF CSS0 INTERFACE
PROBLEM DESCRIPTION:
A node crash can occur when detaching the css0 interface; eg:
/usr/lpp/ssp/css/ifconfig css0 detach
LOCAL FIX:
A delete of the interface may suffice instead of detaching the
interface.
PROBLEM SUMMARY:
A node crash can occur when the css0 interface
is detached as follows:
"/usr/lpp/ssp/css/ifconfig css0 detach"
PROBLEM CONCLUSION:
The switch IP kernel extension, if_ls, has been
changed to prevent a node crash when detaching
the css0 interface.
------
APAR: IY16053 COMPID: 5765D1700 REL: 220
ABSTRACT: MEMORY LEAK IN ASN1 CODE
PROBLEM DESCRIPTION:
secd leaks memory.
------
APAR: IY16055 COMPID: 5765D5100 REL: 320
ABSTRACT: NODE CRASH ON DETACH OF CSS0 INTERFACE
PROBLEM DESCRIPTION:
A node crash can occur when detaching the css0 interface; eg:
/usr/lpp/ssp/css/ifconfig css0 detach
LOCAL FIX:
A delete of the interface may suffice instead of detaching the
interface.
PROBLEM SUMMARY:
A node crash can occur when the css0 interface
is detached as follows:
"/usr/lpp/ssp/css/ifconfig css0 detach"
PROBLEM CONCLUSION:
The switch IP kernel extension, if_ls, has been
changed to prevent a node crash when detaching
the css0 interface.
------
APAR: IY16065 COMPID: 5765C3403 REL: 430
ABSTRACT: ANYNET SOCKETS-OVER-SNA, DO A 'TALK ROOTSOCKSNA'. IT FAILS.
PROBLEM DESCRIPTION:
AnyNet Sockets-over-SNA, Do a 'talk rootsocksna'. It fails.
PROBLEM CONCLUSION:
Export the udp_hash_chn_lock variable; so that code can be
inserted in the driver to access pcb hash table organization
------
APAR: IY16077 COMPID: 5765C3403 REL: 430
ABSTRACT: KSH SCRIPT CORE DUMPS WITH COPROCESS
PROBLEM DESCRIPTION:
Script using coprocess can core dump.
------
APAR: IY16105 COMPID: 5765C3403 REL: 430
ABSTRACT: NFS UID MISMATCH POSSIBLE ON CREATE
PROBLEM DESCRIPTION:
If an NFS client is using uid/gid mapping
with a filesystem that has been exported
with ANON. Files could be created with the
wrong uid.
PROBLEM SUMMARY:
If an NFS client is using uid/gid mapping
with a filesystem that has been exported
with ANON. Files could be created with the
wrong uid.
PROBLEM CONCLUSION:
check the UID and GID of creds against vnode attributes.
------
APAR: IY16106 COMPID: 5765C3403 REL: 430
ABSTRACT: PROBLEM WITH CPU ALLOCATION (PMR 79915,519)
PROBLEM DESCRIPTION:
When running CPU bound threads on an SMP, with WLM the CPU
utilization between the classes some times cannot be correctly
regulated by WLM and some classes get more CPU than they
should based to the shares set up by the system administrator.
PROBLEM CONCLUSION:
Improve the multiple runqueue load balancing and the
interaction between WLM and the AIX scheduler.
------
APAR: IY16108 COMPID: 5765C3403 REL: 430
ABSTRACT: STOP LOGGING NON-ERRORS FOR ONLINE MIRROR COPY
PROBLEM DESCRIPTION:
Stale pps will be logged when doing online lvm mirror copies
PROBLEM CONCLUSION:
Stop logging STALEPP error if the pb_sched field is
hd_stalepp_now_end
------
APAR: IY16109 COMPID: 5765C3403 REL: 430
ABSTRACT: BACKPORT SUPPORT OF MAP FILES WITH STRIPED LOGICAL VOLUMES
PROBLEM DESCRIPTION:
Provide map file support for making and extending striped
logical volumes.
------
APAR: IY16115 COMPID: 5765C3403 REL: 430
ABSTRACT: XLIB: IMFILTER REFERENCES XIC AFTER IT IS DESTROYED.
PROBLEM DESCRIPTION:
It is possible in rare occasions for an InputMethodFilter to
get registered, but not unregistered (even when the XIC that
it uses has been destroyed). If this filter is used it may
cause a crash.
PROBLEM CONCLUSION:
Do a better job of scrubbing the list of registered IMFilters
when XDestroyIC is called.
------
APAR: IY16132 COMPID: 5765C3403 REL: 430
ABSTRACT: ADD LINK SPEED INFORMATION TO DETAIL DATA
PROBLEM DESCRIPTION:
Update FC adapter error log details.
------
APAR: IY16142 COMPID: 5765C3403 REL: 430
ABSTRACT: ADD EUROPEAN ISO8859-15 ALT KEYBOARD SYMLINKS
PROBLEM DESCRIPTION:
Xwindows will not recognize an iso8859-15 alt keyboard map.
PROBLEM CONCLUSION:
Create symlinks from the ISO15 alt map to the ISO1 alt map.
------
APAR: IY16147 COMPID: 5765C3403 REL: 430
ABSTRACT: ADD SUBTYPE FOR SCSD CD-ROM DEVICE
PROBLEM DESCRIPTION:
Add new ODM attribute to differentiate SCSD CD-ROMS
------
APAR: IY16148 COMPID: 5765C3403 REL: 430
ABSTRACT: FUTURE HARDWARE SUPPORT
PROBLEM DESCRIPTION:
Future hardware support
------
APAR: IY16158 COMPID: 5765C3403 REL: 430
ABSTRACT: ROOTVG ONLY INSTALLS ON ONE DISK WHEN THREE ARE SELECTED
PROBLEM DESCRIPTION:
On reformated disks (PVID is 0's), if multiple disks are
selected for BOS Install, only one will be used.
PROBLEM CONCLUSION:
Change check for short PVID's to check for long and short
PVIDs.
------
APAR: IY16169 COMPID: 5765B8100 REL: 220
ABSTRACT: 3270 DOESNOT RESTART CORRECTLY AFTER SNA ERRORS.
PROBLEM DESCRIPTION:
The 3270 sessions do not re-enable correctly after an SNA host
error has occurred, and the SNA link has recycled. Existing
sessions are left active, or are not restarted after SNA
recovery
PROBLEM SUMMARY:
3270 DOESNOT RESTART CORRECTLY AFTER SNA
ERRORS.
PROBLEM CONCLUSION:
The e32pool recovery code did not call the
command if a failureoccurred during a refresh script.
------
APAR: IY16175 COMPID: 5765B8100 REL: 220
ABSTRACT: MESSAGE WAITING INDICATORS DO NOT GET TURNED ON.
PROBLEM DESCRIPTION:
Message waiting indicators do not get turned on.
PROBLEM SUMMARY:
Message waiting indicators do not get turned on
------
APAR: IY16177 COMPID: 5765C3403 REL: 430
ABSTRACT: PICK ECHO GONE WILD
PROBLEM DESCRIPTION:
se_qm3_setup.c was calling a class2 renderer
instead of the pick echo EchoTriangle()
renderer.
PROBLEM CONCLUSION:
Check if the traditional renderer should be
called.
------
APAR: IY16178 COMPID: 576552900 REL: 240
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: IY16182 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: BUFFER OVERFLOW IN BIND8
PROBLEM DESCRIPTION:
It is possible to overflow a buffer handling TSIG signed
queries, thereby obtaining access to the system.
PROBLEM SUMMARY:
The bind8 bugs have been fixed.
PROBLEM CONCLUSION:
These bugs were due to buffer overflow.
------
APAR: IY16214 COMPID: 5765C3403 REL: 430
ABSTRACT: BUFFER OVERFLOW AND FORMAT STRING VULNERABILITIES IN BIND 4.X
PROBLEM DESCRIPTION:
It is possible to overflow the buffer.
PROBLEM SUMMARY:
The DNS bugs have been fixed.
PROBLEM CONCLUSION:
This was bug due to buffer overflows.
------
APAR: IY16265 COMPID: 5765B9501 REL: 310
ABSTRACT: GPFS FILE CORRUPTED WHEN UPDATING FILE NEAR END OF FILE LIKE WHE
PROBLEM DESCRIPTION:
A file can get corrupted on a GPFS file system when the
file is updated in small amounts near the end of the file.
This can typically be seen when using the 'ar' command to
update an archive file.
PROBLEM SUMMARY:
File data was incorrect in a case where a small file was
updated with a small write.
PROBLEM CONCLUSION:
Correct data movement to copy old fragment when creating a
larger fragment
------
APAR: IY16271 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: INFOLEAK IN NUMEROUS VERSIONS OF NAMED4 AND NAMED8
PROBLEM DESCRIPTION:
It is possible to construct a inverse query that allows the
stack to be read remotely exposing environment variables.
PROBLEM CONCLUSION:
Fixing the code which handles inverse queries.
------
APAR: IY16333 COMPID: 5765C3403 REL: 430
ABSTRACT: PROBLEMS WITH MESSAGES AND SET-ID ROOT PROGRAMS
PROBLEM DESCRIPTION:
The NLSPATH variable for "root" is improperly reset on login.
It is inherited from /etc/environment and this value is lost.
PROBLEM CONCLUSION:
Do not relinquish all root privileges during login.
------
APAR: IY16370 COMPID: 5765C3403 REL: 430
ABSTRACT: DATA CORRUPTION IN LIBNSL
PROBLEM DESCRIPTION:
Routines to encode certain RPC packets can cause problems.
------
APAR: IY16396 COMPID: 5765C3403 REL: 430
ABSTRACT: NFS STRESS TEST FAILS
PROBLEM DESCRIPTION:
The changes made in APAR IY14031 cause problems, in this
case surfacing in the nfs stress tests. It seems like
files are not removed properly and stray .nfs files are
left lying around.
PROBLEM CONCLUSION:
Back out the changes made in apar IY14031. This will
reintroduce the problem that IY14031 was supposed to fix.
------
APAR: IY16459 COMPID: 576554300 REL: 240
ABSTRACT: MPI_ALLREDUCE PERFORMANCE IMPROVMENT WHEN TASK NUMBER IS A POWER
PROBLEM DESCRIPTION:
MPI_Allreduce algorithm needs a perfomance improvement when the
number of tasks is a power of 2.
PROBLEM SUMMARY:
MPI_ALLREDUCE tries to select the best one of a set of
algorithms. In the days of 1task/node, a special case
algorithm for power of 2 tasks was used. This algorithm
turns out to be rather poor with several tasks per node.
The result was noticably poorer performance for 64 tasks
than for 63 or 65 (likewise for other powers of 2).
PROBLEM CONCLUSION:
The selection logic in MPI_ALLREDUCE has been changed so if
there is more than one task per node, even power of 2 task
Allreduces passes up the the special case algorithm.
Running exactly 1 task / node will still use the special
case algorithm.
------
APAR: IY16583 COMPID: 576552900 REL: 240
ABSTRACT: LATEST PSSP 2.4 FIXES AS OF FEBRUARY 2001.
PROBLEM DESCRIPTION:
Latest PSSP 2.4 fixes as of February 2001.
PROBLEM SUMMARY:
This is a packaging apar for PSSP 2.4 fixes as
of February 2001.
PROBLEM CONCLUSION:
This is a packaging apar for PSSP 2.4 fixes
as of February 2001.
------
APAR: IY16599 COMPID: 5765D1700 REL: 220
ABSTRACT: AIXDCE22 MAINTENANCE PACKAGE FOR PTF SET 9
PROBLEM DESCRIPTION:
THIS APAR CREATED FOR MAINTENANCE PACKAGE OF PTF SET 9
aixdce22 Maintenance Package for PTF SET 9
PROBLEM SUMMARY:
AIXDCE22 MAINTENANCE PACKAGE FOR PTF SET 9
PROBLEM CONCLUSION:
AIXDCE22 MAINTENANCE PACKAGE FOR PTF SET 9
------
APAR: IY16719 COMPID: 5765D2800 REL: 430
ABSTRACT: LATEST HACMP FOR AIX R431 UPDATES AS OF FEBRUARY 2001
PROBLEM DESCRIPTION:
Order this APAR to receive the latest HACMP for AIX 4.3.1
fixes available as of February 2001.
This is a packaging APAR only. It will not appear in the
list of APARs on the SMIT "Install Fileset Update by Fix"
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
PROBLEM CONCLUSION:
Packaging only
------
APAR: IY16723 COMPID: 5765C3403 REL: 433
ABSTRACT: LATEST AIX 4.3.3 UPDATES AS OF FEBRUARY 2001
PROBLEM DESCRIPTION:
Order this APAR to receive the latest AIX 4.3.3 updates
available as of February 2001. These updates are applicable to
AIX 4.3.3 with the 4330-06 Recommended Maintenance Level, and
should be ordered using service level 433006.
This is not a recommended preventive service package and could
contain unresolved PTFs in error (PEs). This is simply a
packaging of the latest available updates.
This is is packaging APAR only. It will not appear in the list
of APARs on the SMIT "Update Software by Fix (APAR)" panel, nor
will the the 'instfix' command show this APAR as being installed
after the updates delivered by this package are installed.
To install selected updates from this package, use the command:
smit update_by_fix
To install all updates from this package that apply to installed
filesets on your system, use the command:
smit update_all
PROBLEM SUMMARY:
Packaging only.
PROBLEM CONCLUSION:
Packaging only.
------
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]