Re: FW: FYI on RS/6000 APARs

From: Saxon, Lamar (Lamar.Saxon@AMERICREDIT.COM)
Date: Thu Nov 14 2002 - 14:28:34 EST

Next message: Ron Schwingel: "Re: Newbie: routing trouble"
Previous message: Saxon, Lamar: "Re: user login problem;"
Maybe in reply to: Steve Argersinger: "FW: FYI on RS/6000 APARs"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

I think they are referring to the AIX Fixes mailing list.

Looks like the developers have finally gotten around to fixing a lot of the
open security holes with PTFs vs. eFixes.

Look through the attached message and scan for security to see them all.

Lamar

-----Original Message-----
From: Mary Deane [mailto:fclmmd@NERSP.NERDC.UFL.EDU]
Sent: Tuesday, November 12, 2002 9:05 PM
To: aix-l@Princeton.EDU
Subject: Re: FW: FYI on RS/6000 APARs

Steve,

I havn't seen anything about this either, and I'm supposed to get all
security announcements from CERT and IBM. Could you prod the folks doing
the "wellness study" to provide you with the specific information on which
"internet security groups" and if possible, the exact message? And if you
get it, could you share with the list?

We in the library world are only too used to getting problem reports
carefully sanitized of all useful diagnostic information...

Thanks,
Mary

Mary Deane
Unix Sys-admin
Florida Center for Library Automation

On Mon, 11 Nov 2002, Steve Argersinger wrote:

> Just the other day (late last week) I received this from some folks doing
a
> wellness study from us. In going through IBM website, I can't find
anything
> that has been put out there any more recent than 10/30. Does
anybody know
> what my e-mail below might be referring to and/or where I might be able to
> get these fixes.
>
> Steve Argersinger
> Ruan Transportation
>
>
>
> Received the following today from one of the Internet security groups:
>
> "AIX admins will be busy this week; IBM released a batch of new
> security-related APARs. A few new bugs will let local attackers gain root
> privileges."
>
> Might pass that along to whoever maintains the RS/6000's as these will
> likely begin to show up in our future vulnerability scans. Give them a
> heads-up if you would.
>

attached mail follows:

You are being sent this document because lsaxon@americredit.com
has requested a copy or has subscribed to the document named
"New_AIXV4_Fixes".
If you would like to be removed from this mailing list, send e-mail to
aixserv@austin.ibm.com with a subject of "unsubscribe New_AIXV4_Fixes", or
send a note to owner-aixserv@austin.ibm.com with your request.

APAR: IY22531 COMPID: 5765C3403 REL: 430
ABSTRACT: JAVA CORE DUMPS IN GBK LOCALE

PROBLEM DESCRIPTION:
Java application window may crash under GBK locale.

PROBLEM SUMMARY:
Java application window may crash under GBK locale.

PROBLEM CONCLUSION:
correct wrong memory operation

------

APAR: IY29477 COMPID: 5765C3403 REL: 430
ABSTRACT: IOCP COPIES OUT TOO MUCH DATA

PROBLEM DESCRIPTION:
The iocp GetQueuedCompletionStatus interface copies out
8 bytes of data when a 32 bit application uses the
interface while running on a 64 bit kernel. The extra
copied out data can corrupt adjacent variables.

PROBLEM SUMMARY:
The iocp GetQueuedCompletionStatus interface copies out
8 bytes of data when a 32 bit application uses the
interface while running on a 64 bit kernel. The extra
copied out data can corrupt adjacent variables.

PROBLEM CONCLUSION:
The solution is to copy out 4 bytes instead of 8 in this
situation.

------

APAR: IY30281 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSINTR ERRORS FOR DPMPA ADAPTER

PROBLEM DESCRIPTION:
The error shows SYSINTR errors for the interrupt level
corresponding to the dpmpa adapter.

PROBLEM SUMMARY:
LABEL: INTRPPC_ERR
IDENTIFIER: DADF69E4
Date/Time: Fri Mar 1 09:51:43
Sequence Number: 182693
Machine Id: 0002518F4C00
Node Id: a0101441s1
Class: H
Type: UNKN
Resource Name: SYSINTR
Resource Class: NONE
Resource Type: NONE
Location: NONE
Description
UNDETERMINED ERROR
Probable Causes
SYSTEM I/O BUS
SOFTWARE PROGRAM
ADAPTER
DEVICE
Recommended Actions
PERFORM PROBLEM DETERMINATION PROCEDURES
Detail Data
BUS NUMBER
0000 00C8
INTERRUPT LEVEL
0000 0025

PROBLEM CONCLUSION:
One last interrupt can hang in the queue without getting
processed before the port is taken down. Check for interrupts
and give them time to process before taking down the port.

------

APAR: IY30696 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSTEM CRASH IN TCP_UPSACKINFO()

PROBLEM DESCRIPTION:
Sytem crash in tcp_UpSACKInfo() if the "sack" network option
is set and number of retransimissions of TCP packets is more
across the network.

LOCAL FIX:
Disable "sack"

PROBLEM SUMMARY:
If sack network option is set, customer will experience
system crash if the no. of retransmissions of the packets
accross the network is more.

PROBLEM CONCLUSION:
Removing the unnecessary increment of both old and new
max_sack_blocks pointer inside the for loop in
tcp_UpSACKInfo().

------

APAR: IY30711 COMPID: 5765D5101 REL: 121
ABSTRACT: HACMP/ES SUPPORT MORE THAN 16 NETWORKS

PROBLEM DESCRIPTION:
HACMP/ES SUPPORT MORE THAN 16 NETWORKS

PROBLEM SUMMARY:
RSCT Topology Services can currently only monitor up to 16
Topology Services networks, which roughly translates into
a limit of 16 IP or non-IP adapters per node. If more
networks are specificed in an HACMP configuration then
these additional networks/adapters will be ignored
by Topology Services, and as result will appear as down to
HACMP. An AIX error log entry with the following LABEL
will be shown in this case: TS_OFF_LIMIT_ER.
The output of command
lssrc -ls topsvcs
will include the following line
>> Number of networks exceeds limit. Extra networks
ignored. <<
which indicates that the number of networks exceeds
Topology Services' limits.

PROBLEM CONCLUSION:
A change was made in RSCT Topology Services so that it can
now support additional networks. With the change, Topology
Services (and therefore HACMP/ES) should support around
40 adapters per cluster node.
In order to allow Topology Services to log its events
properly when running a configuration with large
numbers of adapters, it is advisable to
  1) Increase the number of lines logged by the Topology
     Services daemon, which can be done with the
     following smit sequence:
Cluster Configuration
Cluster Topology
   Configure Topology Services and Group Services
     Change / Show Topology and Group Services Configuration
       Topology Services log length (lines)
  2) Increase the space in the /var filesystem, which
     is where RSCT places its log data.
If a system with large numbers of adapters is being
upgraded with the new RSCT level, it is recommended
that the number of networks in the HACMP configuration be
reduced (until the number of IP or non-IP adapters falls
to 16 or below) before the new RSCT is applied. After
RSCT is applied in all the nodes of the cluster then
additional networks can be (re-)added. This
recommendation is only relevant if the RSCT updates are
applied in one node at a time while the cluster is kept
running on the remaining nodes.

------

APAR: IY31359 COMPID: 5765C3403 REL: 430
ABSTRACT: PACKETS DO NOT OBEY ROUTING TABLE IF IT HAS DUPLICATE ROUTES

PROBLEM DESCRIPTION:
Some packets do not obey the routing table when there are
duplicate entries for a particular route

PROBLEM CONCLUSION:
Assign "saved_match" the value of variable "match"

------

APAR: IY31636 COMPID: 5765C3403 REL: 433
ABSTRACT: FTP ERROR WHEN LOGGING INTO NFS MOUNTED HOME DIR WITH 750 PERM.

PROBLEM DESCRIPTION:
ftp displays error when logging to a m/c, whose home directory
is nfs mounted with the permission as 750.

PROBLEM SUMMARY:
If the home directory is nfs mounted with the file permission
750, customer will be getting the error message
"230 - No directory! loging in with home=/" while logging in.
But it places in the right directory, giving a wrong error
message.

PROBLEM CONCLUSION:
Modified the code, so that it will not give wrong error
message.

------

APAR: IY31641 COMPID: 5765C3403 REL: 433
ABSTRACT: M/C HANGS IF IT IS FLOODED WITH PACKETS WHOSE TH_FLAGS IS NULL.

PROBLEM DESCRIPTION:
System hangs if it is flooded with packets whose th_flags is
set to zero.

PROBLEM SUMMARY:
System might hang, if it receives packets with no flags set.
There might not be any problem with one packet. But system
hangs when it is flooded with packets.

PROBLEM CONCLUSION:
tcp_input0() didn't free the mbuf, if the flags were set to
zero. Modified the code to free the mbuf and send RST.

------

APAR: IY31825 COMPID: 5765C3403 REL: 430
ABSTRACT: DLOPEN/DLSYM DEAD LOCK

PROBLEM DESCRIPTION:
It has been found a deadlock situation can arise between
dlopen and dlsym (or any other dl function such as dlclose
or dlerror).
dl operations are serialized with a common pthread mutex
lock.
In this particular case, it was found one thread was loading
a C++ library using dlopen. This library had a static
constructor with pthread mutex lock within, which it was trying
to acquire while loading. Another thread has already acquired
the lock and was now trying to perform a dlsym operation.
This resulted in a deadlock.

PROBLEM SUMMARY:
It has been found a deadlock situation can arise between
dlopen and any other dl functions.
dl operations are serialized with a common pthread mutex
lock.
If dlopen tries to load C++ library which has static
constructors with mutex locks and these mutex locks have
been acquired by other threads prior to other dl operations
(such as dlclose, dlopen. dlsym), then there can be a
process level hang.

PROBLEM CONCLUSION:
Release the common pthread lock before calling C++ rte's
loadAndInit and reacquire it later. Also, since loadAndInit
is not threadsafe, protect it with another pthread mutex.

------

APAR: IY31934 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: AUTOFS EXECUTABLE MAPS ALLOWS ROOT ACCESS

PROBLEM DESCRIPTION:
A client configured with autofs and executable maps can be
exploited to allow commands to be run as root for any user with
access to the system.

LOCAL FIX:
Don't use executable maps

PROBLEM SUMMARY:
Change string handling around how the executable map is called
to avoid the exposure of this security hole.

PROBLEM CONCLUSION:
Change string handling around how the executable map is called
to avoid the exposure of this security hole.

------

APAR: IY31953 COMPID: 5765C3403 REL: 430
ABSTRACT: MOUNT USES EXCESSIVE CPU WITH NETGROUP EXPORTS

PROBLEM DESCRIPTION:
When large netgroups are used in the exports line for root
access of a filesystem, rpc.mountd might use excessive
amounts of CPU when servicing requests for that filesystem.

LOCAL FIX:
Remove the usage of netgroups in the root=-statement, use
a list of hostnames instead.

PROBLEM CONCLUSION:
Add a flag to disable netgroup resolution for root exports.

------

APAR: IY32039 COMPID: 5765C3403 REL: 430
ABSTRACT: GXT6000 GXT6500 SHADING RENDERED INCORRECTLY IN FLUENT APP.

PROBLEM DESCRIPTION:
On GXT6000P and GXT6500P, an OpenGL based application from
Fluent called gambit renders incorrectly in shaded mode. The
model incorrectly renders all white, when it should render grey
due to lighting effects.

PROBLEM SUMMARY:
On GXT6000P and GXT6500P, an OpenGL based application from
Fluent called gambit renders incorrectly in shaded mode. The
model incorrectly renders all white, when it should render grey
due to lighting effects.

PROBLEM CONCLUSION:
Correctly handle the case where the light position
is 0,0,0,0

------

APAR: IY32040 COMPID: 5765C3403 REL: 430
ABSTRACT: PS -O GIVES OUTPUT WITH LINES GREATER THAN 2048 CHARACTERS

PROBLEM DESCRIPTION:
The ps command with the -o option gives output greater than
2048 characters per line when piped or redirected to a file.
Other tools, like grep and vi that use this output will not
work properly.

PROBLEM SUMMARY:
ps command prints more than 2048 characters per line
with the -o option. This causes other commands like
grep and vi that use this output to fail.

PROBLEM CONCLUSION:
The defect was fixed with code chages to the ps command.

------

APAR: IY32108 COMPID: 5765C3403 REL: 433
ABSTRACT: FOR OPTION 81 IS NOT CORRECTLY SENT BACK TO CLIENT

PROBLEM DESCRIPTION:
When echoing back the option 81 to client
dhcpsd makes it FQDN (Fully Quantified
Domain Name) by ending the name with
"." but it is not checking whether
the client is already send FQDN.

PROBLEM SUMMARY:
The option 81 sent by the server has two "."
. This happens when the option 15 is not
set in the configuration file.

PROBLEM CONCLUSION:
The code has been changed to check for this
condition.

------

APAR: IY32163 COMPID: 5765C3403 REL: 430
ABSTRACT: MKSYSB TAKING MORE CPU DURING PROMPT FOR NEXT TAPE

PROBLEM DESCRIPTION:
When mksysb prompts for next tape volume to be
inserted, vmstat shows CPU idle time close to 0.

PROBLEM SUMMARY:
CPU idle time goes to almost zero, as mksysb loops in a script
waiting for next tape vol to be inserted.

PROBLEM CONCLUSION:
make mksysb script to sleep for sometime instead of looping
continuosly

------

APAR: IY32167 COMPID: 5765C3403 REL: 430
ABSTRACT: TOPAS DATA FIELDS OVERFLOW

PROBLEM DESCRIPTION:
Data value exceed the available space in the curses display.
This causes leftover numbers to remain on the screen.

PROBLEM CONCLUSION:
Coverts data values which exceed the fields to Kilo or Mega
and adds the appropriate 'K' or 'M' after the value.

------

APAR: IY32223 COMPID: 5765E5400 REL: 440
ABSTRACT: MIGRATION HAS->HAES: NO CLVER ENTRIES

PROBLEM DESCRIPTION:
When migration from HAS 4.4.1.x to HAES 4.4.1.x, the clver
entries
in /etc/services AND in /etc/inetd.conf do not get added
properly.
The user ends up with the HAS entries commented out, and the
HAES entries not present. Hence, no
verification/synchronization
is possible.
A workaround is to add these entries by hand, and restart the
inetd
subsystem.

PROBLEM CONCLUSION:
HAES installation will check for NxN Migration case and not
configure clver daemons. This configuration will be
performed later after HAS is automatically deinstalled.
During NxN migrations, the customer should ensure that the
latest available PTF of HAES is installed before
reintegrating the migrated node into the cluster.

------

APAR: IY32228 COMPID: 5765E5400 REL: 440
ABSTRACT: HAES: SELECTIVE FALLOVER DOES NOT OCCUR

PROBLEM DESCRIPTION:
selective fallover does not occur for application failure or
local network failure. The cluster correctly enques the
required rg_move event, but it does not take the event off
the queue and run it. After the failed network is reintegrated
into the cluster, the rg_move is run (but it is now no longer
necessary).

PROBLEM CONCLUSION:
Add the necessary checks to ensure that the rg_move event
will be run at the correct time.

------

APAR: IY32254 COMPID: 5765C3403 REL: 430
ABSTRACT: REALPATH ROUTINE SHOULD CHECK FOR EFFECTIVE ACCESS RIGHTS OF

PROBLEM DESCRIPTION:
realpath() should check for the effective access rights of the
file. If mode is 700 and effective ID is not the same as the
file's owner, realpath() should fail.

PROBLEM SUMMARY:
If mode is 700 and real UID is not same as file's owner, then
realpath fails with EACCES

PROBLEM CONCLUSION:
modified realpath() such that if mode is 700 and effective UID
is not same as owner id, then realpath() should fail with
EACCES.

------

APAR: IY32338 COMPID: 5765C3403 REL: 430
ABSTRACT: SUCCESSFUL FTP LOGIN DOES NOT RESET FAILED LOGIN COUNT

PROBLEM DESCRIPTION:
Failed login count gets incremented on unsuccessful login. But
on successful login, failed login count does not get reset to
zero. This behaviour of ftpd istotally different from that of
telnetd.

PROBLEM SUMMARY:
After successful ftp login the failed login count will not
reset.

PROBLEM CONCLUSION:
After a successful ftp login, the failed login count will get
reset to zero. And number of failed attempts since last
login, last unsuccessful login attempt and last successful
login attempt will be displayed as 230 reply messages
upon successful login to ftp.

------

APAR: IY32368 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: TOOLTALK SERVER : MULTIPLE ISSUES

PROBLEM DESCRIPTION:
an attacker can:
- Overwrite 4 bytes of memory in the running process
   with a zero (0x0L) value
- Remotely delete any file on the vulnerable host
- Locally create or overwrite any file on the vulnerable
   host with arbitrary contents.
- Remotely create arbitrary directory entries on the
   vulnerable host

PROBLEM CONCLUSION:
. perform proper range checks on the file descriptor used as
   an index into the table _tt_db_table, ie it must be between
>= 0 and < _TT_MAX_ISFD (0 and 127 inclusive)
. check to see if the log file is a symbolic link via lstat()
   don't open to write the transaction log if the log file is
   a symbolic link

------

APAR: IY32384 COMPID: 5765C3403 REL: 430
ABSTRACT: INCORRECT ERROR LOG ENTRY FOR CORE DUE TO SIGQUIT

PROBLEM DESCRIPTION:
Error log entry generated when a process is terminated
due to SIGQUIT is not filled in correctly. The entry
looks as follows:
Detail Data
SIGNAL NUMBER
           0
USER'S PROCESS ID:
           0
FILE SYSTEM SERIAL NUMBER
          -1
INODE NUMBER
          -1
PROGRAM NAME
ADDITIONAL INFORMATION
Unable to generate symptom string.

LOCAL FIX:
None but please note that a core file is generated as
normal in the appropriate directory.

PROBLEM SUMMARY:
<< Replace these lines with text indicating >>
<< the symptom(s) surrounding this problem >>
Error log entry generated when a process is terminated
due to SIGQUIT is not filled in correctly. The entry
looks as follows:
Detail Data
SIGNAL NUMBER
           0
USER'S PROCESS ID:
           0
FILE SYSTEM SERIAL NUMBER
          -1
INODE NUMBER
          -1
PROGRAM NAME
ADDITIONAL INFORMATION
Unable to generate symptom string.

PROBLEM CONCLUSION:
Remove test for SIGQUIT when filling in the error log entry
for core dump.

------

APAR: IY32389 COMPID: 5765C3403 REL: 433
ABSTRACT: AIXTERM -E AIXTERM FAILS TO OPEN THE SECOND AIXTERM

PROBLEM DESCRIPTION:
When executing an aixterm from another aixterm using the -e
option, the second aixterm fails to open must of the time.

PROBLEM SUMMARY:
aixterm -e aixterm fails to open the second aixterm must of the
time

PROBLEM CONCLUSION:
Code change to remove an unnecessary exit.
The fix will be available in the next maintenance release.

------

APAR: IY32395 COMPID: 5765C3403 REL: 431
ABSTRACT: NEED FOR USER TUNEABLE FIN_WAIT_2 STATE TIMER

PROBLEM DESCRIPTION:
netstat -an will show lots of connections in FINWAIT2 state.
They will be hanging around for around 10 minutes.

PROBLEM CONCLUSION:
FINWAIT2 timer was earlier computed based on keepalive timer
and the number of probes. Now we will have a seperate timer
for FINWAIT2 which can be modified through "no" command.
The name of the tuneable is tcp_finwait2. The default value
will be 10 minutes.

------

APAR: IY32404 COMPID: 5765C3403 REL: 430
ABSTRACT: CSH COREDUMPS WHEN PDF FILES OPENED THROUGH NETSCAPE BROWSER

PROBLEM DESCRIPTION:
'csh' encounters segmentation fault when pdf files are
opened from netscape browser.

PROBLEM SUMMARY:
csh coredumps when a csh script is invoked for a pdf file
through netscape browser.

PROBLEM CONCLUSION:
The csh coredump occurs due to memory leak in csh.

------

APAR: IY32419 COMPID: 5765C3403 REL: 430
ABSTRACT: CORRECT README STATEMENT ON AIXTHREADS VARIABLES

PROBLEM DESCRIPTION:
bos README incorrectly states that the default value of the
pthreads library debug variables (AIXTHREAD_...) is ON, but
the default is now OFF.

PROBLEM SUMMARY:
bos README incorrectly states that the default value of the
pthreads library debug variables (AIXTHREAD_...) is ON, but
the default is now OFF.

PROBLEM CONCLUSION:
Changed README to state that the default value is OFF.

------

APAR: IY32448 COMPID: 5765C3403 REL: 430
ABSTRACT: PTRACE WATCHPOINT CAN LEAVE DABR SET OUTSIDE PROCESS

PROBLEM DESCRIPTION:
Using hardware watchpoints from user level
debuggers can cause a system crash on the
32bit kernel.

PROBLEM CONCLUSION:
Unset the DABR register when saving state
to fix the problem of the DABR remaining
set across certain context switches.

------

APAR: IY32456 COMPID: 5765C3403 REL: 430
ABSTRACT: GXT300P/GXT2000P COLOR PROBLEM WITH MANY INSTANCES OF APP

PROBLEM DESCRIPTION:
With GXT300P and GXT2000P, We have been
able to reproduce a visual problem related
to colors by minimizing about 7 MyoSPECT
applications. On the 7th one the color scales
start to fail by coming up wrong or not changing
when a new color scale is selected.
After reducing the number of open applications
to about 3 we can usually get the scales
back to normal. Customers have reported that
on occasion they require a system shutdown to
recover.

PROBLEM SUMMARY:
With GXT300P and GXT2000P, We have been
able to reproduce a visual problem related
to colors by minimizing about 7 MyoSPECT
applications. On the 7th one the color scales
start to fail by coming up wrong or not changing
when a new color scale is selected.
After reducing the number of open applications
to about 3 we can usually get the scales
back to normal. Customers have reported that
on occasion they require a system shutdown to
recover.

PROBLEM CONCLUSION:
Correct logic error in color map management, when a
color map is pre-empted by a new color map.

------

APAR: IY32485 COMPID: 5765C3403 REL: 430
ABSTRACT: SYMBOLS IN TERM.H NOT EXPORTED FROM LIBCURSES

PROBLEM DESCRIPTION:
Symbols like boolcodes, boolnames etc. which are seen
in term.h header file are not exported from the curses
library.

PROBLEM SUMMARY:
Some symbols that are in term.h are not exported
from the curses library.

PROBLEM CONCLUSION:
The symbols in term.h exported from curses library and
their declaration changed to const in term.h header file.

------

APAR: IY32505 COMPID: 5765E6400 REL: 220
ABSTRACT: JFS ERRORS DURING SITE ISOLATION OF REMOTE SITE

PROBLEM DESCRIPTION:
JFS ERRORS DURING SITE ISOLATION OF REMOTE SITE.

PROBLEM SUMMARY:
When a geo-mirroring network problem causes site isolation,
the HA event processing halts the nodes at one site, in
order to prevent data divergence which occurs when both
sites update the data without mirroring to each other.
If the network heals before this event processing completes,
a write request may go across the network after the gmddown
gets issued on the remote node but before any nodes are
halted. When this happens, the remote node returns an error
to the local node, and the local node fails the application
write, because it knows something is wrong. The problem is
that if the GeoMirror devices are being used to mirror a
JFS filesystem, and the application write happens to be
a write to the jfslog device, the write error to the jfslog
device sets off a chain reaction of JFS errors, which requires
the user to recover the filesystem. If the user panics and
reboots the local node, this causes the filesystem to
become corrupted, as JFS data that was stored in RAM and
needed to recover the filesystem gets lost after a reboot.
Note that this assumes the nodes at the remote site are
halted, and the nodes at the local site survive after the
site isolation processing completes.

PROBLEM CONCLUSION:
The solution is for the local node to avoid failing the
application write request by continuing to resend the
remote write request to the other site, even though it will
probably fail over and over again. This will cause the
application write to hang until the site isolation event
processing gets around to issuing a gmddown command on the
local node. The gmddown will break the write processing
out of the resend loop, and the filesystem and applications
running above it will continue running without any problems.

------

APAR: IY32556 COMPID: 5765E8200 REL: 230
ABSTRACT: MODS TRIGGERED 700 CRASH - A NODE OF HAGEO'S CLUSTER WILL HIT

PROBLEM DESCRIPTION:
Doing lots of Oracle record deletes with AIX xmalloc debug
turned on gives a 700 crash with following stack trace
pgfree+000220 ()
xmfree1+000320 ()
xmfree+0002FC ()
gmdpin:gmd_xmfree+000074 ()
gmdpin:gmd_gbuf_cleanup+000198 ()
gmdpin:gmd_gbuf_alloc+00009C ()
gmdpin:gmd_strategy+00032C ()
devstrat+000220 ()
aio.ext:arl_rdwr+0004B4 ()
aio.ext:ardwr+0005D0 ()
aio.ext:qlioreq+00012C ()
aio.ext:listio64+000194 ()

LOCAL FIX:
Turn off AIX kernel xmalloc debug setting .... crash scenario
should be much less likely without the extra mods checks

PROBLEM SUMMARY:
When debug xmalloc is enabled, the system may crash due to a
Program Interrupt (dump code 700) in the pgfree() function
with the following debug kernel error message:
Debug kernel error message: An internal error occurred during
VMM processing of a release request.
The running thread's stack traceback is similar to the
following:
pgfree+000204
xmfree1+000328
xmfree+00035C
gmd_xmfree+000060
gmd_gbuf_cleanup+000184
gmd_gbuf_alloc+000074
gmd_strategy+000318
gmd_intr_process+0001CC
procentry+000010

PROBLEM CONCLUSION:
There was a race condition which allowed one thread to pin
a memory buffer immediately before another thread tried to
free it. This is a programming error and causes the debug
kernel to crash the system. This problem has been fixed.

------

APAR: IY32557 COMPID: 5765C3403 REL: 430
ABSTRACT: CPIO REMOVES LINKS WHEN RUN SECOND TIME

PROBLEM DESCRIPTION:
The cpio when used to copy links works fine copying all the
files and links to target machine. But when the same command is
run second time, it removes the links .

PROBLEM SUMMARY:
cpio when used to second time to copy the files. It will
delet the links copied first time on target machine.

PROBLEM CONCLUSION:
cpio modified to see that last modified time is checked for
files only incase of ordinary dir entries but bot links.

------

APAR: IY32558 COMPID: 5765E8200 REL: 230
ABSTRACT: 300 CRASH WITH GMD WHEN WE SET STATE MAP SIZE TO 1024 AND 2048

PROBLEM DESCRIPTION:
Here are the stack in errpt.out.
Detail Data
DUMP STATUS
LED:300
csa:2ff3b400
[gmdpin:sm_region_value] c
[gmdpin:state_map_value] 2c
[gmdpin:state_map_value] 2c
[gmdpin:gmd_sync_build_krpc_req] 78
[gmdpin:gmd_sync_read] 350
[gmd:gmd_open] fc
devcopen 330
rdevopen b4
bdev_open 2c
spec_open e8
vnop_open 1c
openpnp 36c
openpath 9c
copen 16c
open 18
sys_call_ret 0
Symptom Data
REPORTABLE
1
INTERNAL ERROR
1
SYMPTOM CODE
PIDS/5765C3403 LVLS/430 PCSS/SPI1 MS/300 FLDS/[gmdpin:s
VALU/7cc430ae

LOCAL FIX:
To circumvent the "<state map size> * 512 * <region size>"
must be equal or greater than the GMD lv size. For example, if
the GMD lv size is 128 Gb, then the following values are needed:
<state map size> should be 8192
<region size> should be 32768 (the default).....
.........................
State Map represents a little bit more than 128 Gb.

PROBLEM SUMMARY:
When a statemap is too small for its GeoMirror device, the
node may crash with an error log entry which includes a stack
trace similar to the following:
LED:300
csa:2ff3b400
gmdpin:sm_region_value c
gmdpin:state_map_value 2c
gmdpin:state_map_value 2c
gmdpin:gmd_sync_build_krpc_req 78
gmdpin:gmd_sync_read 350
gmd:gmd_open fc
devcopen 330
rdevopen b4
bdev_open 2c
spec_open e8
vnop_open 1c
openpnp 36c
openpath 9c
copen 16c
open 18
sys_call_ret 0

PROBLEM CONCLUSION:
The cfggmd command tries to detect when a statemap is too
small for its GeoMirror device. The problem was that a
calculation which converts from 512 byte blocks to megabytes
was overflowing a 32 bit long variable. This problem has
been corrected.

------

APAR: IY32665 COMPID: 5765C3403 REL: 430
ABSTRACT: LAST COMMAND GIVES INCONSISTENT OUTPUT

PROBLEM DESCRIPTION:
The output of last command is inconsistent with that of
who command.

PROBLEM SUMMARY:
The output of last command is inconsistent with the output of
who command.

PROBLEM CONCLUSION:
The last command shows incorrect output. The user displayed
as active by who is shown as logged out by last.

------

APAR: IY32682 COMPID: 5765C3403 REL: 430
ABSTRACT: UNPREDICTABLE IOSTAT BEHAVIOUR WHEN IOSTAT PARAMETER IS SET

PROBLEM DESCRIPTION:
If the "iostat" paramters default value is false, and
several iostats are started simultaneously with "iostat"
parameter set to true, then iostat output may be faulty.

PROBLEM SUMMARY:
Incase, default value of iostat sys parameter is false,
and iostat command is used by several users simultaneously
with iostat parameter set to true, it may fail with error
"Disk statistics since boot not available".

PROBLEM CONCLUSION:
Query the ODM database properly for iostat value. i.e query
PdAt database if value not found in CuAt, and sync the
kernel value with that.

------

APAR: IY32690 COMPID: 5765C3403 REL: 430
ABSTRACT: MEMORY LEAKS IN SECURITY ROUTINES GETGRGID_R() AND LOGINSUCCESS

PROBLEM DESCRIPTION:
There are memory leaks in getgrgid_r() under
NIS environment. This is seen when a NIS
user makes connections to the informix
database, the memory of the authentication
process grows until no more connections are
possible. For a local or other users this
problem is not seen.
There is a memory leak in loginsuccess() call
for a local user. Since the leak is in the
internal library routines, this leak will be
seen while other routines like putuserattr()
etc. for a local user. This problem is seen in
the authentication process of the db2 server
which is growing in memory because of this
call, until it accepts no more connections.
This apar is recommended for any other memory
leak for a local user calling security routines.

LOCAL FIX:
Restart the server after it no longer accepts
any new connections.

PROBLEM SUMMARY:
    There are memory leaks in getgrgid_r() under
    NIS environment. This is seen when a NIS
    user makes connections to the informix
    database, the memory of the authentication
    process grows until no more connections are
    possible. For a local or other users this
    problem is not seen.
    There is a memory leak in loginsuccess() call
    for a local user. Since the leak is in the
    internal library routines, this leak will be
    seen while other routines like putuserattr()
    etc. for a local user. This problem is seen in
    the authentication process of the db2 server
    which is growing in memory because of this
    call, until it accepts no more connections.
    This apar is recommended for any other memory
    leak for a local user calling security routines.

PROBLEM CONCLUSION:
Memory freed is freed before the functions return.

------

APAR: IY32719 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: POTIENTIAL DNS RESOLVER BUFFER OVERFLOW

PROBLEM DESCRIPTION:
There is a potiential buffer overflow problem in DNS resolver
code. This will allow an attacker-controlled DNS server to
reply with a carefully crafted message to (for example) a
gethostbyname() call. This reply will trigger the
buffer overflow.

PROBLEM SUMMARY:
This apar is assocaited with defect 37378 and the
problem regarding the DNS resolver buffer overflow
is resolved.

PROBLEM CONCLUSION:
This apar is assocaited with defect 37378 and the
problem regarding the DNS resolver buffer overflow
is resolved.

------

APAR: IY32759 COMPID: 5765C3403 REL: 430
ABSTRACT: GXT3000P GLX_STEREO: RENDERING ARTIFACTS IN LEFT STEREO BUFFER

PROBLEM DESCRIPTION:
With GXT3000P adapter, rendering artifacts appear when using a
OpenGL application that use GLX_STEREO to render in stereo
mode, and renders to both buffers at once via GL_BACK.

PROBLEM SUMMARY:
With GXT3000P adapter, rendering artifacts appear when using a
OpenGL application that use GLX_STEREO to render in stereo
mode, and renders to both buffers at once via GL_BACK.

PROBLEM CONCLUSION:
Correct incorrect setting of blending flags for stereo case
with blending when drawing to GL_BACK, GL_FRONT and
GL_FRONT_BACK.

------

APAR: IY32775 COMPID: 5765C3403 REL: 430
ABSTRACT: AIX EXECUTABLES NOT STRIPPED

PROBLEM DESCRIPTION:
Some executables are not stripped when shipped so
their symbol information can be viewed.

LOCAL FIX:
Use the strip command to strip executables manually.

PROBLEM SUMMARY:
Commands tic, tput, captoinfo and infocmp are not
stripped of symbol information.

PROBLEM CONCLUSION:
Code modified to strip these commands.

TEMPORARY FIX:
Manually strip the executables with the
strip command.

------

APAR: IY32776 COMPID: 5765C3403 REL: 430
ABSTRACT: ERRORS WITH LIBC128.A LIBRARY OM 64 BIT MODE

PROBLEM DESCRIPTION:
The correct value of a 128 bit long double will not be printed
correctly when compiled in 64 bit mode. The value will be
different than in 32 bit mode because the program is not
branching to the libc128.a printf.

PROBLEM SUMMARY:
libc128.a library doesn't have a 64 bit shared
object and this causes different values to be
returned when 128 bit long double values
are printed.

PROBLEM CONCLUSION:
Required changes will be made to the libc128.a
library to have 64 bit shared objects.

------

APAR: IY32792 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: BUFFER OVERFLOW IN RPC.TTSERVERDB

PROBLEM DESCRIPTION:
a buffer overflow condition in the tooltalk database
server (rpc.ttdbserver) when TT_CREATE_FILE client call is
made with the inputed filename >1024 chars

PROBLEM CONCLUSION:
in _tt_create_file_1(), first check that the inputed filename
is not >1024 chars before proceeding

------

APAR: IY32799 COMPID: 5765C3403 REL: 430
ABSTRACT: PROBLEMS WITH MANY SEQUENTIAL CHFS -A SPLITCOPY COMMANDS

PROBLEM DESCRIPTION:
Some chfs -a splitcopy commands may fail when issued
sequentially on many file systems all using the same jfslog.

PROBLEM SUMMARY:
The chfs -a splitcopy command issues a call to
chlvcopy which results in a sync of the log logical
volume of the filesystem in question. This sync
is executed in the background, so the chfs command
may return before the sync of the log lv completes.
If another chfs -a splitcopy command is immediately
issued on a filesystem using the same jfslog, the
command may fail because the sync is still in
progress.

PROBLEM CONCLUSION:
The environment variable LVM_CHLVCOPY_FG_SYNC has
been added to chlvcopy. When this variable is set
to '1', the sync of the jfslog which occurs in
chlvcopy, called by chfs -a splitcopy, will run
in the foreground, not the background. Thus, the
chfs command will not return until this sync
completes if the variable has been set.

TEMPORARY FIX:
Insert sleep commands between calls to chfs -a
splitcopy.

------

APAR: IY32806 COMPID: 5765C3403 REL: 430
ABSTRACT: GETITIMER() CAN RETURN INCORRECT VALUES FOR EXPIRED TIMERS

PROBLEM DESCRIPTION:
getitimer() can return incorrect values for expired timers
when the system clock is set back after the timer has
already expired and is no longer active. Correct value
will be returned if the clock is set back before the timer
has expired.

LOCAL FIX:
None. Do not use getitimer() on timers that have already
expired.

PROBLEM CONCLUSION:
Check if the timer is still active before returning an time
remaining. Always return 0 if timer is not active.

------

APAR: IY32829 COMPID: 5765C3403 REL: 430
ABSTRACT: CORE DUMP WHEN STARTING X ON GXT130P

PROBLEM DESCRIPTION:
When the customer tries to start X windows via the xinit
command, he sometimes gets a core dump and X does not start.
This problem seems to only occur when the Tivoli Desktop
is started from xinitrc.

PROBLEM SUMMARY:
Attempting to start the Xserver on the GXT130P graphics
adapter can result in an X core dump. The core dump is more
likely to happen if an application which blits to the screen
is started as part of the X startup sequence.

PROBLEM CONCLUSION:
Initialize the Global Screen Pointer variable in the
IFB Declare Device Dependant variables macro.

------

APAR: IY32859 COMPID: 5765C3403 REL: 433
ABSTRACT: SYSTEM MAY CRASH WHEN BSDLOG() IS CALLED

PROBLEM DESCRIPTION:
NFS server may crash when bsdlog() is called. The stack trace
is similar to: trace -m:
IAR: .logbuf+2c (00012ba4): stbx r3,r6,r7 LR:
.bsdlog+a0 (000bba74)

PROBLEM CONCLUSION:
Lock was not held before calling logbuf. Lock it before
calling logbuf

------

APAR: IY32902 COMPID: 5765C3403 REL: 430
ABSTRACT: ERROR REPORT ENTRY APPEARS IF NO GRAPHICS ADAPTER INSTALLED

PROBLEM DESCRIPTION:
Following error appears in error log:
LABEL: GRAPHICS
IDENTIFIER: E85C5C4C
Date/Time: Wed Oct 17 13:30:57 CDT
Sequence Number: 915
Machine Id: 002008CA4C00
Node Id: brandylp2
Class: S
Type: PERM
Resource Name: STARTLFT
Description
SOFTWARE PROGRAM ERROR
Probable Causes
SOFTWARE PROGRAM
Failure Causes
SOFTWARE PROGRAM
        Recommended Actions
        IF PROBLEM CONTINUES TO OCCUR REPEATEDLY THEN
        DO THE FOLLOWING
        CONTACT APPROPRIATE SERVICE REPRESENTATIVE
        REPORT DETAILED DATADetail Data
DETECTED FAILED RC ERROR LOCATION
startlft 0 1204 6

LOCAL FIX:
Uninstall the devices.graphics.com package.

PROBLEM SUMMARY:
This problem is caused when devices.graphics.com is
installed on the system, but there is no graphics
adapter. An error report entry appears for "startlft",
and appears each time the machine is rebooted or cfgmgr
is run.

PROBLEM CONCLUSION:
The code which output the error report entry has been
removed. The customer will no longer see this error
report entry.

------

APAR: IY32956 COMPID: 5765C3403 REL: 433
ABSTRACT: CORRECT THE DOC. FOR HOST COMMAND TO SPECIFY THE CORRECT SYNTAX.

PROBLEM DESCRIPTION:
Correct the doc. of host command to specify correct syntax.

PROBLEM SUMMARY:
hostnew command doesn't display clear syntax.

PROBLEM CONCLUSION:
Modified the host2.msg file to show the correct syntax.

------

APAR: IY32964 COMPID: 5765D5100 REL: 320
ABSTRACT: SWTICH COULD NOT START DUE TO DIAG_FAIL AFTER NODE REBOOT

PROBLEM DESCRIPTION:

PROBLEM SUMMARY:
tbic_reg_test_pci was reading a value it had written before
the write had completed.

PROBLEM CONCLUSION:
tbic_reg_test_pci will allow time for the write to complete
before cehcking the value that is written.

------

APAR: IY32980 COMPID: 5765C3403 REL: 430
ABSTRACT: NISMKUSER MASTER LOOKUP

PROBLEM DESCRIPTION:
Duplicate users (by UID or username) can be created
causing authentication failures.

LOCAL FIX:
Check for UID and username prior to creation with
'niscat -M passwd.org_dir'.

PROBLEM CONCLUSION:
Require that all checks be made with the master only for user
information within the nismkuser command code.

------

APAR: IY33009 COMPID: 5765C3403 REL: 430
ABSTRACT: LOADER LEAKS FILE POINTERS

PROBLEM DESCRIPTION:
Unsuccesful load attempts may not properly
close shared libraries and object files.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* bos.mp 4.3.3.79
* bos.mp 4.3.3.80
* bos.mp 4.3.3.81
* bos.mp 4.3.3.82
* bos.mp 4.3.3.83
* bos.mp 4.3.3.84
* bos.up 4.3.3.79
* bos.up 4.3.3.80
* bos.up 4.3.3.81
* bos.up 4.3.3.82
* bos.up 4.3.3.83
* bos.up 4.3.3.84
***************************************************************
*PROBLEM DESCRIPTION: *
* Unsuccesful load attempts may not properly close
* shared libraries and object files. User may not be
* able to cleanly unmount filesystems.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY33009
***************************************************************

PROBLEM CONCLUSION:
Add code to appropriately close files opened during failed
loads.

------

APAR: IY33024 COMPID: 5765C3403 REL: 430
ABSTRACT: DB2START TRAPS DUE TO LOADBIND() PROBLEM.

PROBLEM DESCRIPTION:
Programs may abort due to bad addresses when accessing
deferred resolution symbols.

PROBLEM CONCLUSION:
Fix the conditional that is supposed to prevent multiple
relocations of the same address.

------

APAR: IY33028 COMPID: 5765C3403 REL: 430
ABSTRACT: RECREATEVG FAILS ON VG MADE WITH FLASHCOPY

PROBLEM DESCRIPTION:
After using flashcopy to copy half of mirrored disks in a
shark volume group, recreatevg fails to bring up the copy as
a volume group on a backup system.

PROBLEM SUMMARY:
After using flashcopy to copy half of mirrored disks in a
shark volume group, recreatevg fails to bring up the copy
as a volume group on a backup system.
The error reported is:
0516-013 varyonvg: The volume group cannot be varied on
because there are no good copies of the
descriptor area.

PROBLEM CONCLUSION:
Physical volume header handling code within
recreatevg has been modified to properly walk
through all pvs.

------

APAR: IY33068 COMPID: 5765C3403 REL: 430
ABSTRACT: FIX SNAP TO GATHER SSA INFO FOR MANY DEVICES

PROBLEM DESCRIPTION:
Snap doesn't correctly gather information if there are
many SSA devices.

PROBLEM CONCLUSION:
Allow many entries by not using shell variable.

------

APAR: IY33073 COMPID: 5765C3403 REL: 433
ABSTRACT: ADD SUPPORT FOR HACMP DATA IN SNAP

PROBLEM DESCRIPTION:
Add functionality to snap to collect HACMP information.

PROBLEM SUMMARY:
Add support for hacmp component in snap.

PROBLEM CONCLUSION:
-e option is added to snap command to collect
hacmp snap data into /tmp/ibmsupt/hacmp directory.
-m option is also added to snap command which can
be used in conjunction with -e option to pass nodenames
to snap command.

------

APAR: IY33145 COMPID: 5765C3403 REL: 430
ABSTRACT: SVMON -P AND -U FAILURES ON STRESSED SYSTEMS

PROBLEM DESCRIPTION:
svmon -P and -U do not return any output

PROBLEM SUMMARY:
svmon -P or -U options can fail when the process
table changes due to process exits during the
lookup of segment information.

PROBLEM CONCLUSION:
svmon updated to not report on processes that
exit during svmon collection phases.

------

APAR: IY33157 COMPID: 5765D5101 REL: 121
ABSTRACT: SUPPRESS ERROR MESSAGES FROM RSCT GROUP SERVICES APIS

PROBLEM DESCRIPTION:
When the RSCT group services API ha_gs_subscribe is called to
subscribe to a group that does not exist, group services writes
the following error messages to STDERR:
2520-014 Provider token 0 does not exist.
Subscribing to a non-existent group is not an error when the
intent is to check if the group exists. In this case the error
message is irrelevant and confusing if displayed on a terminal.
The group services APIs provide return values to the caller
to indicate the result, so writing to STDERR is unnecessary.

PROBLEM SUMMARY:
Group Services API prints an error message to stderr
and returns the corresponding error code.
The fix is to suppress the error message by default.

PROBLEM CONCLUSION:
Group Services API is currently printing error messages
to stderr, and also returning an error code
when API can not properly process the requests.
As such error messages are unnecessary and actually
troublesome,
they will be suppressed in default.
They can be enabled by setting an environment variable
before a Group Services API is called;
export HA_GS_APIMSG=1

------

APAR: IY33177 COMPID: 5765C3403 REL: 430
ABSTRACT: HACMP FAILOVER SLOW DUE TO LSVG -L PERFORMANCE

PROBLEM DESCRIPTION:
Performance of HACMP failover scripts is slow due to use of
lsvg -l calls on large volume groups.

PROBLEM SUMMARY:
HACMP failover commands both directly and indirectly
call lsvg -l. On very large volume groups, the
performance of this command may be quite slow and make
the failover process take too long.

PROBLEM CONCLUSION:
Performance improvement has been made to lsvg -l to
prevent unnecessary disk queries which were negatively
impacting the speed to the command.

------

APAR: IY33218 COMPID: 5765C3403 REL: 430
ABSTRACT: CLOSE AND UNCONFIG FOR RSDD_RSPC HAVE TIMING WINDOWS

PROBLEM DESCRIPTION:
When the unconfig is done, the call to i_clear is done
without any lock. When the close is done, the unpin and
call to i_clear both are done without any lock. This can
lead to duplicate calls to i_clear or unpin both of which
will assert the machine.

PROBLEM SUMMARY:
When the unconfig is done, the call to i_clear is done
without any lock. When the close is done, the unpin and
call to i_clear both are done without any lock. This can
lead to duplicate calls to i_clear or unpin both of which
will assert the machine.

PROBLEM CONCLUSION:
The solution is to add in a process level lock for each port
that is gained and released around the call to the code that
unpins the code and calls i_clear.

------

APAR: IY33231 COMPID: 5765C3403 REL: 430
ABSTRACT: CRONTAB COMMAND DOESN'T RETURN CORRECT ERROR CODE WHILE WRITING

PROBLEM DESCRIPTION:
crontab command doesnot check the proper return
status of writing to stdout. This returns wrong exit
status from crontab at situations when the file
system is full.

PROBLEM SUMMARY:
crontab command doesn't check the return status of
writing to stdout.

PROBLEM CONCLUSION:
Proper changes will be made to crontab to check
the return status of writing to stdout, to make
sure that correct exit status will be returned
by crontab.

------

APAR: IY33240 COMPID: 5765C3403 REL: 430
ABSTRACT: AUTOFS UNMOUNT PROCESS SPIN FIX

PROBLEM DESCRIPTION:
The automount 'aump' process begins consuming excessive
resources.

PROBLEM SUMMARY:
The aump kproc will consume excessive CPU resources. The
process can not be killed, therefore a reboot is necessary.

PROBLEM CONCLUSION:
The aump process will handle and clear signals that it
recieves and continue functioning normally.

------

APAR: IY33307 COMPID: 5765C3403 REL: 430
ABSTRACT: SIGNALS ARE LOST IN SPURIOUR WAKEUPS

PROBLEM DESCRIPTION:
A hang condition where all threads are now in sleep or wait.

PROBLEM CONCLUSION:
The fix here is to remove the case statement which
puts it back into the wait state - which then forces the
application to be in complete standard compliance for the
possibility of spurious wakeup.

------

APAR: IY33309 COMPID: 5765C3403 REL: 430
ABSTRACT: DBX MISHANDLES FORTRAN DYNAMIC ARRAYS

PROBLEM DESCRIPTION:
dbx does not correctly display dynamically
allocated arrays of character strings when
debugging fortran programs.

LOCAL FIX:
use the $vardim dbx variable to tell dbx
the size of each array element.

PROBLEM CONCLUSION:
Change the array descriptor parsing code to properly account
for the element size of dynamically allocated arrays of
multiple instance types (such as arrays of strings).

------

APAR: IY33369 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSTEM CRASH IN PHXENTDD

PROBLEM DESCRIPTION:
Customer will see the following stack trace in the system
dump:
    MST STACK TRACE: 0xf0005360
    (excpt=08004500:40000000:00000000:08004500:00000106)
    (intpri=3)
            IAR: . phxentdd:tx_complete +194
            (01591c10): lwz r0,0x0(r5)
            LR: . phxentdd:tx_complete +214 (01591c90)
            f0521a20: . phxentdd:phxent_output +578 (0159333c)
            f0521a90: . if_en:en_output +564 (015a3538)
            f0521b20: . netinet:ip_output +ae0 (051ae714)
            f0521c00: . netinet:tcp_output +1a20 (051c8434)
            f0521d10: . netinet:tcp_usrreq +a04 (051c62bc)
            f0521d80: .sosend+d44 (000d71bc) f0521ed0:
            .send+134 (000dd140) f0521f60: .sys_call_ret+0
            (00003a90)

PROBLEM SUMMARY:
System crash in phxentdd requires driver fix.

------

APAR: IY33416 COMPID: 5765D5101 REL: 121
ABSTRACT: 384 WAY D/S:HAGSGLSM FAILED WHEN COULD NOT OPEN NEW

PROBLEM DESCRIPTION:
384 way D/S:hagsglsm failed when could not open new log file

PROBLEM SUMMARY:
The current hags daemon will die when the log
directory is full so that a new log file cannot be
opened. Currently it will keep all possible
hags daemon and hagsglsm daemon log files if
the size doesn't exceed the limit. So in a system
with many nodes the log directory will become
full very quickly.

PROBLEM CONCLUSION:
The solution is that
1. no matter if the directory is full only keep
  3 different kinds of incarnation log files for both
  hags daemon and hagsglsm daemon.
2. If it cannot open a new log file the daemon
  will not die, it will mark it failed in opening
  a new log file and try again until a maximum
  number of times of trying is reached.

------

APAR: IY33436 COMPID: 5765C3403 REL: 430
ABSTRACT: FTP CLIENT ON PC DOESN'T WORK WITH AIX FTP SERVER

PROBLEM DESCRIPTION:
The ftp client on a PC aborts when doing a 'ls'.

PROBLEM SUMMARY:
The ftp client on a PC aborts when doing a 'ls'.

PROBLEM CONCLUSION:
For performance reasons, defect 323612 turned on Nagle
(remove TCP_NODELAY) from the data connection and turned off
Nagle (add TCP_NODELAY) to the control connection. It seems
certain PC's can't handle this. Therefore I will add the
'B' (for BULL) option to ftpd to get ftpd to use pre-323612
behaviour.

------

APAR: IY33441 COMPID: 5765C3403 REL: 433
ABSTRACT: RESTVG FAILS WITH TAPE BLOCK SIZE SET TO VARIABLE

PROBLEM DESCRIPTION:
Restoring a user vg using the restvg command
could fail if the block size of the device
is set to a variable size (0).
One error seen, indicates that the image.data file
does not exist but the error is misleading and is
caused by an unsuccessful restore of the image.dat
file. This is the error displayed:
0512-025 restvg: The image data file does not exist.
Restore of Volume Group canceled.
The error could also indicate an out of space condition
that is not real.

PROBLEM SUMMARY:
If a volume group backup is done to tape with variable block
size the restore (restvg) will fail.
0512-025 restvg: The image data file does not exist.
Restore of Volume Group canceled.

PROBLEM CONCLUSION:
The code has been modified to change
the block size to the largest
possible block size.

TEMPORARY FIX:
none, do not make backup with variable
block size.

------

APAR: IY33464 COMPID: 5765C3403 REL: 430
ABSTRACT: XSERVER HANG WHEN USING -BS X OPTION AND EASYCOPY

PROBLEM DESCRIPTION:
Xserver is started with -bs option, which can be done
via xinit -- -bs, or via modification of xserverrc
to add bs option. A screen capture utility called
easycopy causes the X server to hang when used to
do a screen grab of 3d application Catia.

PROBLEM SUMMARY:
Xserver is started with -bs option, which can be done
via xinit -- -bs, or via modification of xserverrc
to add bs option. A screen capture utility called
easycopy causes the X server to hang when used to
do a screen grab of 3d application Catia.

PROBLEM CONCLUSION:
Change logic in mibstore.c to "punt" if depth of parent
window does not match.

------

APAR: IY33474 COMPID: 5765C3403 REL: 433
ABSTRACT: INCORRECT LC_NUMERIC ENTRY FOR SV_SE LOCALE

PROBLEM DESCRIPTION:
LC_NUMERIC entry for decimal_point has incorrect value for
sv_SE locale.

LOCAL FIX:
Customer can modify source file for sv_SE locale and generate
an updated locale.

PROBLEM CONCLUSION:
Correct the entry in LC_NUMERIC stanza for decimal_point from
fullstop to comma.

------

APAR: IY33491 COMPID: 5765C3403 REL: 430
ABSTRACT: PTY HAS DEADLOCK IN TELNETD SETUP

PROBLEM DESCRIPTION:
There is a deadlock that can happen when telnetd tries to
enter the special telnet mode of the pty. There is a sequence
of ioctls that must be sent down but if the queue has data
on them, the ioctls wait until they get to the top of the
queue. Yet the only process that can drain the queue is
telnetd. As a result, the ioctl's hang forever.

PROBLEM SUMMARY:
There is a deadlock that can happen when telnetd tries to
enter the special telnet mode of the pty. There is a sequence
of ioctls that must be sent down but if the queue has data
on them, the ioctls wait until they get to the top of the
queue. Yet the only process that can drain the queue is
telnetd. As a result, the ioctl's hang forever.

PROBLEM CONCLUSION:
The solution is ensure that the ioctl is process in a timely
manner.

------

APAR: IY33532 COMPID: 5765C3403 REL: 430
ABSTRACT: CNSDD PRODUCES UNUSABLE DATA STREAM

PROBLEM DESCRIPTION:
When an iocack message from the 7318 is split into seperate
messages and only the header has been received thus far,
then cnsdd produces an empty message block with the iocack
and send it up stream. This causes ctsdd to use data that
is not present and can cause data corruption or a panic.

PROBLEM SUMMARY:
When an iocack message from the 7318 is split into seperate
messages and only the header has been received thus far,
then cnsdd produces an empty message block with the iocack
and send it up stream. This causes ctsdd to use data that
is not present and can cause data corruption or a panic.

PROBLEM CONCLUSION:
The solution is to kick back to the calling routine with a
return code such that it will wait for the next packet to
arrive and then retry the disassembly of the messages into the
encapsulated message.

------

APAR: IY33559 COMPID: 5765C3403 REL: 430
ABSTRACT: IOCP CAN CAUSE CRASH

PROBLEM DESCRIPTION:
The system can crash in iocp in iocp_process_pending_io.
This is caused by a request being removed but still
being attached to a work list.

PROBLEM SUMMARY:
The system can crash in iocp in iocp_process_pending_io.
This is caused by a request being removed but still
being attached to a work list.

PROBLEM CONCLUSION:
The solution is to check the work queue to see if the request
is on there and remove it if it is found.

------

APAR: IY33604 COMPID: 5765C3403 REL: 430
ABSTRACT: GRAB PROBLEM IN MENUSHELL.C

PROBLEM DESCRIPTION:
focus grab problem with Motif 2.1 MenuShell

PROBLEM SUMMARY:
Motif 2.1 menus may have trouble grabbing keyboard or mouse
focus.

PROBLEM CONCLUSION:
Backport fix from AIX 5.0 (Motif 2.1.30).

------

APAR: IY33613 COMPID: 5765C3403 REL: 433
ABSTRACT: MODEL 850 FAILS TO CONFIGURE ISA IDE CDROM DRIVE

PROBLEM DESCRIPTION:
Customers will not be able to boot the latest AIX 4.3.3 from
the IDE cdrom on a 850/830 machine.

PROBLEM CONCLUSION:
a file that was linked incorrectly in the build tree. Fix the
link will solve the problem.

------

APAR: IY33619 COMPID: 5765C3403 REL: 430
ABSTRACT: XMSCALE VALUE INVISIBLE IN JA_JP

PROBLEM DESCRIPTION:
In multibyte locales, the slider value of an XmScale widget may
not be visible.

PROBLEM SUMMARY:
The value of an XmScale widget may not be visible in a
locale which uses multibyte characters.

PROBLEM CONCLUSION:
Motif 2.1 code fix.

------

APAR: IY33622 COMPID: 5765C3403 REL: 433
ABSTRACT: MOVE UPDATE_ALL TO RUN PRIOR TO CFGMGR TO AVOID CLONING ISSUES

PROBLEM DESCRIPTION:
After cloning a system backup, there may be missing
device support. The cfgmgr -i call is failing due to:
A base level fileset can not be installed as a requisite of
another fileset when a lower level of the requisite is already
installed.
Other symptoms may also occur.

LOCAL FIX:
Clone with media at the same level as mksysb.

PROBLEM SUMMARY:
After cloning a system backup, there may be missing
device support. The cfgmgr -i call is failing due to:
A base level fileset can not be installed as a requisite of
another fileset when a lower level of the requisite is already
installed
Other symptoms may also occur.

PROBLEM CONCLUSION:
The code has been changed to update the system
after the mksysb image has been restored, and
again at the end of all other installs
for the cloning of the mksysb.
This fix will only be available as a service fix.
In order to use this fix the user must use NIM
to install the client system and must update the
SPOT to contain this fix.

------

APAR: IY33623 COMPID: 5765C3403 REL: 433
ABSTRACT: MKSYSB FAILS TO INSTALL ON LARGER DISK

PROBLEM DESCRIPTION:
Cloning mksysb installs from a source system
with small disks to a target system with large
disks may fail.

LOCAL FIX:
install onto a smaller disk

PROBLEM SUMMARY:
Cloning mksysb installs from a source system
with small disks to a target system with large disks
may fail.

PROBLEM CONCLUSION:
The code has been changed to determine the TFACTOR
necessary to use the given PP size, and
use the provided PP size, only if this TFACTOR was
also given.

------

APAR: IY33662 COMPID: 5765C3403 REL: 430
ABSTRACT: PTY SERVICE ROUTINE INCORRECTLY DOES PUTQ ON HIGH PRI MESSAGE

PROBLEM DESCRIPTION:
The pty write service routines improperly check for canput
with a high priority message and put the message back
on the same queue if the canput returns false. With high
priority messages, the message must be forwarded no matter
what the flow control state is.
This can cause an infinate loop of streams calling the
service routine which can ultimately hang the system.

PROBLEM SUMMARY:
The pty write service routines improperly check for canput
with a high priority message and put the message back on the
same queue if the canput returns false. With high priority
messages, the message must be forwarded no matter what the
flow control state is.
This can cause an infinate loop of streams calling the service
routine which can ultimately hang the system.

PROBLEM CONCLUSION:
The solution is to forward all high priority messages.

------

APAR: IY33680 COMPID: 5765E5400 REL: 440
ABSTRACT: HAES: ENABLE LOG MESSAGES FOR NXN MIGRATIONS

PROBLEM DESCRIPTION:
During node by node migration, there is progress information
that can be logged which would help diagnose any failures.
This logging is turned off by default.

PROBLEM CONCLUSION:
Turn on the node by node logging information as a default.

------

APAR: IY33683 COMPID: 5765E5400 REL: 440
ABSTRACT: CORRECTLY RECOGNIZE FASTT DISKS

PROBLEM DESCRIPTION:
HACMP commands to reset reserves on a FAStT may not be
appropriate for OEM versions of this device, causing disk
takeover to fail.

PROBLEM CONCLUSION:
Check the lscfg information on the disk to see if it has the
IBM microcode. If not, treat it as a generic disk.

------

APAR: IY33708 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSLOGD DOES NOT SEND MESSAGES TO THE USER

PROBLEM DESCRIPTION:
Customer installed bos.net.tcp.client 4.3.3.83
on the system and paid attention that syslogd
not sending messages to the user as configured
but sending them to the file.

PROBLEM SUMMARY:
After ML10 syslogd messages are not getting printed on user
screen, even if syslogd is configured to do so.

PROBLEM CONCLUSION:
There was a string manipulation problem, because of which
the user part in /etc/syslog.conf was ignored. Fixed
the problem.

------

APAR: IY33746 COMPID: 5765C3403 REL: 430
ABSTRACT: CRASH IN IN_CKSUM() WHEN PMTU ENABLED

PROBLEM DESCRIPTION:
When PMTU is enabled, system can crash in in_cksum routine.

LOCAL FIX:
Disable pmtu discovery

PROBLEM CONCLUSION:
pmtuest is used to calculate the mbuf size required and this
value is protected by the IPPMTU_LOCK. We release it once in
the code and don't check the value of pmtuest after
acquiring the lock back. Verify the pmtuest value after
getting lock and exit if it is zero.

------

APAR: IY33766 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSTEM CRASH IN UNP_DISCONNECT()

PROBLEM DESCRIPTION:
System crash in unp_disconnect() due to mismatch in Unix domain
PCB pointers. When we call socket calls on unix sockets like
closing the socket or redeaing from the socket, we are
completely depending on the socket lock. We are acquiring this
socket lock in socket functions before calling uipc_usrreq().
Due to some unaviodable circumstances, we will release this
sockect lock in unp_detach() and unp_connect() and acquire it
back. Once we release the lock the socket is free to anyone and
they can acquire this lock and may alter the current values of
the socket resulting in system crash.

PROBLEM SUMMARY:
System crash in unp_disconnect() due to mismatch in Unix domain
PCB pointers. This problem occurs very rarely, when
applications call syslog extensively. Here is the stack trace
MST STACK TRACE: 0x2ff3b400
(excpt=00000000:0a000000:00000000:00000000:00000106)
(intpri=11) IAR: .simple_lock+14 (00009514): stwcx.
r6,r0,r3
         LR: .unp_disconnect+6c (000c84bc) 2ff3b0d0:
         .unp_detach+130 (000c82c8) 2ff3b120: .uipc_usrreq+1ac
         (000c7838) 2ff3b180: .soclose+540 (000d85cc) 2ff3b1f0:
         .soo_close+7c (000d8038) 2ff3b240: .closef+7c
         (001db048) 2ff3b290: .closefd+90 (001a4468) 2ff3b2f0:
         .fs_exit+54 (001e88b0) 2ff3b340: .kexitx+3f0
         (00041760) 2ff3b3c0: .kexit+7c (0006d1f8)

PROBLEM CONCLUSION:
unp_detach() calls bsdlog_unreg() to assign NULL to log_so
pointer for server related calles. We are releasing the
Socket lock during this time to avoid deadlock with
UNPMISC_LOCK() for both client and server. This fix will
release the lock only for the server and thus avoids
unnecessary release of socket lock during client calls.

------

APAR: IY33779 COMPID: 5765C3403 REL: 430
ABSTRACT: MAKE NOT GIVING ERROR WHEN SOURCE FILES REQUIRED FOR ARCHIVES

PROBLEM DESCRIPTION:
The make command does not give an error when a dependent
source file of an archive does not exist and the archive
cannot be built.

PROBLEM SUMMARY:
Make will not give error when a required source
file does not exist for an archive.

PROBLEM CONCLUSION:
The problem was fixed with code changes to
the make command.

------

APAR: IY33782 COMPID: 5765C3403 REL: 433
ABSTRACT: IFCONFIG GIVES ERROR MESSAGES

PROBLEM DESCRIPTION:
The Apar IY28977 intgroduced the ioctl SIOUPDROUTE. This apar
requires that bos.net.tcp.client, bos.adt.include and
bos.mp/up to be shipped together. But when customer downloads
bos.net.tcp.client 4.3.3.83, it does not prereq 4.3.3.83
levels of bos.adt.include and bos.mp/up. This may result in
hang with ifconfig command.

LOCAL FIX:
Install same levels of bos.adt.include and bos.mp/up along with
bos.net.tcp.client while applying apar IY28977

PROBLEM SUMMARY:
ifconfig gives error message after applying:
43: bos.net.tcp.client 4.3.3.83
51: bos.net.tcp.client 5.1.0.26
This is because changes due to IY28977 (43) or
IY30046 (51) shipped to multiple filesets, and the
required requisites weren't specified.
Workaround:
43: install all the fixes for IY28977 (bos.up
  4.3.3.83 or bos.mp 4.3.3.83 and bos.adt.include
  4.3.3.83)
51: install all the fixes for IY30046 (bos.up
  5.1.0.26 or bos.mp 5.1.0.26 or bos.mp64
  5.1.0.26 and bos.adt.include 5.1.0.26)

PROBLEM CONCLUSION:
Add the needed requisites to the next filesets.

------

APAR: IY33816 COMPID: 5765C3403 REL: 430
ABSTRACT: APAR_ABSTRACT: NODE UNRESPONSIVE AFTER X.25 ADAPTER

PROBLEM DESCRIPTION:
On an SNA END_NODE that has an active SNA link station
connected to an SNA NETWORK_NODE, which is running over QLLC
and an active X.25 network, the X.25 adapter is disconnected
from the X.25 network. Approximately 20 seconds later, the
node that was disconnected from the network becomes completely
non-responsive to terminal input.

PROBLEM CONCLUSION:
replace malloc's ipri save areas with stack vatriables.

------

APAR: IY33817 COMPID: 5765C3403 REL: 430
ABSTRACT: QLLC CRASH - QPM_EXCEPTION_FUNCTION+0002F4

PROBLEM DESCRIPTION:
Crash running CS/AIX V6 Router (64 bit) over X.25 network
qpm_exception_function+0002F4
xdh_pass_on_stat+000194
CioScheduleDone+00011C
CioScheduleRun+000524
CioScheduleDone+000240
cio_lrput+0007CC
csq_run+0002B0
csq_turnover+000278
csq_lateral+000104
scheduled_run+0000D0
Netintr+000140
netisr_thread+000014
threadentry+000014

PROBLEM CONCLUSION:
Add additional locking between X.25 Comio exception handler
and the QLLC kernel service process.

------

APAR: IY33843 COMPID: 5765E5400 REL: 450
ABSTRACT: REQUIRED CHANGES FOR HAGEO 2.4 SUPPORT

PROBLEM DESCRIPTION:
1. HACMP fails to recognize the site names
2. FS mount on remote site hangs
3. Error in cluster verification

PROBLEM CONCLUSION:
1. When a resource sync and verification is done after
migrating from HAGEO 2.3 to 2.4, filesystem and application
server errors will be generated due to site names in the
resource group node list and the absence of a GEOsite odm
class.
The verify routines will check the HACMP site class for a
match on the names in the RG nodelist and skip that
verification step if they are the same.
2. When processing a failover on a cluster with gmd's and
filesystems, the mount command will hang on the recovery
node.
mount command is being called before gmddown (which tells
the gmd's to stop mirroring). mount should be called
after gmddown.
3. clverify performs a filesystem consistency check by
testing for the availability of the filesystem on every
node in the resource group which contains the filesystem.
This check is not appropriate for filesystems which are
mounted over Geographic Mirror Devices (GMD's).

------

APAR: IY33855 COMPID: 5765D5100 REL: 320
ABSTRACT: RH3 MPV:MISSING PART NUMBER FOR TB3 PCI ADAPTER

PROBLEM DESCRIPTION:
rh3 mpv:missing part number for tb3 pci adapter

PROBLEM SUMMARY:
VPD for pci adapter is not available for diagnostic
controller to pick up.

PROBLEM CONCLUSION:
Get vpd data from adapter and save it in CuVPD during card
config. Now, when a problem is detected while running
diagnostics, Diagnostic Controller will get FRU from CuVPD
and report the problem.

------

APAR: IY33856 COMPID: 5765B9501 REL: 320
ABSTRACT: INVALIDATE DISKS WHEN MMADDISK FAILS

PROBLEM DESCRIPTION:
invalidate disks when mmaddisk fails

PROBLEM SUMMARY:
When adding new disks, it needs to add new segments to
allocation map. If there is no space left for the new
segments it stops. But after freeing space, the retry of the
mmadddisk says "Are you sure, these disks appear to be in
use", so need to use the -v no option on mmadddisk to
override the check.

PROBLEM CONCLUSION:
When completeDeleteDisks find disks that were being added
(but add failed), it needs to have changeDiskStates
invalidate the disk and SG descriptors when it changes the
state to BeingDeletedFromAllocMap so that the disks do not
look like they belong to a SG anymore.

------

APAR: IY33880 COMPID: 5765C3403 REL: 430
ABSTRACT: PCI 10/100 ETHERNET II CANNOT RECEIVE SHORT 802.3 PACKETS

PROBLEM DESCRIPTION:
Small receive packets are not accepted by the 10/100 ethernet
device driver when running IEEE 802.3.

PROBLEM CONCLUSION:
The driver code has been corrected to insure that the small
IEEE 802.3 packet is accepted and passed up.

------

APAR: IY33890 COMPID: 5765C3403 REL: 430
ABSTRACT: RAS IMPROVEMENT FOR IBM PCI 10/100 ETHERNET ADAPTER

PROBLEM DESCRIPTION:
In the start adapter function there is a 100ms delay. This
delay breaks customer's setsocket program with the option
IP_ADD_MEMBERSHIP

PROBLEM SUMMARY:
In the start adapter function there is a 100ms delay. This
delay breaks customer's setsocket program with the option
IP_ADD_MEMBERSHIP

PROBLEM CONCLUSION:
Introduce a ras_improvement attribute to allow the customer
reduce the delay time to 2ms

------

APAR: IY33920 COMPID: 5765C3403 REL: 430
ABSTRACT: XTERM/NUM LOCK ON AIXCLIENTS TO LINUX XSERVER

PROBLEM DESCRIPTION:
Customer says that when he exports DISPLAY from his AIX box to
Linux
system, it opens an xterm window, but when he tries to input
anything from the keyboard in that window, it doesn't accept
any text. It accepts
mouse movement. Even when he exports Netscape he gets the same
problem. He is experiencing a similar problem on a windows
box running exceed.

LOCAL FIX:
define num_lock to Mod2.

PROBLEM SUMMARY:
Customer says that when he exports DISPLAY from his AIX box to
Linux
system, it opens an xterm window, but when he tries to input
anything from the keyboard in that window, it doesn't accept
any text. It accepts
mouse movement. Even when he exports Netscape he gets the same
problem. He is experiencing a similar problem on a windows
box running exceed.

PROBLEM CONCLUSION:
define num_lock into Mod2

------

APAR: IY33926 COMPID: 5765D5101 REL: 121
ABSTRACT: HAGSGLSM DOES NOT CLOSE CSS DEVICE AFTER ATTEMPTING TO OPEN CAUS

PROBLEM DESCRIPTION:
hagsglsm opens css device to check it's status, however, it
does not close it. this can result in a problem if enough
opens have been issued causing it to hit the max opens that
can be at one time (this looks to be around 2000).
This can cause hagsglsm to hang and result in hags groups not
forming. Recycling hags/hagsglsm will solve the problem.

PROBLEM SUMMARY:
HAGSGLSM opens the switch device (e.g., /dev/css0)
to get the device number. However, HAGSGLSM mistakenly
did not close the file descriptor.
Because of that, the number of used file descriptors
keeps increasing whenever HAGSGLSM tries to open
a device, and eventually hits the max possible
number of file descriptors per process.

PROBLEM CONCLUSION:
HAGSGLSM opens the switch device (e.g., /dev/css0)
to get the device number. However, HAGSGLSM mistakenly
did not close the file descriptor.
The fix is to make sure to close the file descriptor
after opening a switch device, to keep the number
of open file descriptors low.

------

APAR: IY33934 COMPID: 5765E8200 REL: 230
ABSTRACT: ASYNC GMD FILESYSTEM NOT MOUNTED AFTER FAILOVER

PROBLEM DESCRIPTION:
filesystem created on async GMD not mounted after site failover

PROBLEM SUMMARY:
Asynchronous GeoMirror devices may be processed incorrectly
at the secondary site when multiple node_down events are
processed in quick succession. This problem affects only
HAGEO 2.3 configurations with HACMP Classic. This is not a
problem for configurations which use HACMP/ES.

PROBLEM CONCLUSION:
Some processing of asynchronous GeoMirror devices occurs
during the node_down_complete event. The problem is that
HACMP Classic processes node_down_complete events
asynchronously. If two nodes leave the HACMP cluster at the
same time, the node_down event processing for the second node
may execute while the node_down_complete event processing
for the first node is still running. This causes asynchronous
GeoMirror devices to be processed incorrectly. The solution
is to synchronize node_down and node_up event processing to
run after node_down_complete processing for the previous
node has completed.

------

APAR: IY33942 COMPID: 5765C3403 REL: 430
ABSTRACT: SORECEIVE() CAN LOSE DATA WHEN CALLED FROM KERNEL USERS

PROBLEM DESCRIPTION:
Encountered problem with XTI library
where t_rcv() function is missing packets.

PROBLEM SUMMARY:
Due to a race condition in soreceive(), protocols over xtiso
may see data/packet loss.

PROBLEM CONCLUSION:
*mp to point to &(*mp)->m_next, so that next time through the
loop we don't copy data into *mp. Thereby the original data
will be retained.

------

APAR: IY34004 COMPID: 5765C3403 REL: 430
ABSTRACT: CSH BUILTIN COMMAND JOBS GIVES UNEXPECTED OUTPUT WITH PIPES

PROBLEM DESCRIPTION:
The jobs builtin command in csh gives incorrect output
when piped to another process.

PROBLEM SUMMARY:
csh jobs builtin does not give any output when piped
to another process

PROBLEM CONCLUSION:
The jobs builtin function in csh does not give any output
when piped to another process

------

APAR: IY34018 COMPID: 5765C3403 REL: 430
ABSTRACT: CORE DUMP WHILE USING TELNET CLIENT WITH VT100 TERM

PROBLEM DESCRIPTION:
Core dump while using telnet client with VT100 terminal
due to buffer overflow in cmdprintf() in telnet.c.

PROBLEM SUMMARY:
Core dump while using telnet client with VT100 terminal
due to buffer overflow in cmdprintf() in telnet.c.

PROBLEM CONCLUSION:
So not set term to VT100.

------

APAR: IY34026 COMPID: 5765C3403 REL: 430
ABSTRACT: SEG FAULT IN RIGHTMOST+0X8 USING -BS X-SERVER OPTION

PROBLEM DESCRIPTION:
When backing store feature of the X server is
used (which can be enabled via -bs option on
xinit), a Seg Fault in rightmost+0x8 can
occur intermittently.

PROBLEM SUMMARY:
When backing store feature of the X server is
used (which can be enabled via -bs option on
xinit), a Seg Fault in rightmost+0x8 can
occur intermittently.

PROBLEM CONCLUSION:
Initialize pointers to null. In some error paths, pointers
that were never used previously where the targets of free's.

------

APAR: IY34044 COMPID: 5765D5100 REL: 320
ABSTRACT: POSSIBLE SP SYSTEM ATTACHMENT ADAPTER DATA ERROR

PROBLEM DESCRIPTION:
switches want to autojn on 3 nodes

PROBLEM SUMMARY:
It is possible for a data error to occur when
data is transferred between systems interconnected by an
SP Switch using the SP System Attachment Adapter
(SPSAA server feature code # 8396).

PROBLEM CONCLUSION:
The timing logic in the xilinx_file3pci that
is loaded on the SP System Attach Adapter (SPSAA, service
Feature code #8396) has been adjusted to address this
problem.

------

APAR: IY34056 COMPID: 5765E6110 REL: 220
ABSTRACT: SUPPRESS ERROR MESSAGES FROM RSCT GROUP SERVICES APIS

PROBLEM SUMMARY:
Group Services API prints an error message to stderr
and returns the corresponding error code.
The fix is to suppress the error message by default.

------

APAR: IY34074 COMPID: 5765E8200 REL: 230
ABSTRACT: GEO_CHFS: 0516-306, 0516-992, 0516-602, 0516-610, 506-390

PROBLEM DESCRIPTION:
IN hageo 2.3 environment issuing:
geo_chfs -a size=+<some number> receive many errors:
0516-306 getlvodm Unable to find <nodename> in Device config. db
0516-992 lsvg: Illegal combination of command line options.
expr 0402_050 syntax error
0516-610 extendlv : Number of logical partitions not entered

PROBLEM SUMMARY:
When a GMD is defined using the = metacharacter for
the logical volume device name of the remote node,
invocations of geo_chfs and geo_extendgmd will fail
with a series of syntax and failed command messages.

PROBLEM CONCLUSION:
geo_extendgmd has been modified to correctly process
a remote logical volume name that has been specified by "=".

------

APAR: IY34085 COMPID: 5765C3403 REL: 430
ABSTRACT: _KGRANTPT CREATES COPY OF CREDS, NOT JUST A REF

PROBLEM DESCRIPTION:
The _kgrantpt and and pty_setname routines make a copy of the
credentials structure unnecessarily.

PROBLEM CONCLUSION:
Remove the code which makes a copy of the credentials
structure.

------

APAR: IY34112 COMPID: 5765C3403 REL: 430
ABSTRACT: DEADLOCK - IP_FREEMOPTIONS CALLED WITH SOCKET_LOCK

PROBLEM DESCRIPTION:
Deadlock situation can happen and system will hang.
ip_freemoptions is called with socket_lock which in turn
calls in_delmulti with the socket_lock.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* bos.net.tcp.client 4.3.3.51cket as it was done before.
* bos.net.tcp.client 4.3.3.53
* bos.net.tcp.client 4.3.3.54
* bos.net.tcp.client 4.3.3.55
* bos.net.tcp.client 4.3.3.75
* bos.net.tcp.client 4.3.3.76
* bos.net.tcp.client 4.3.3.77
* bos.net.tcp.client 4.3.3.78
* bos.net.tcp.client 4.3.3.79
* bos.net.tcp.client 4.3.3.80
* bos.net.tcp.client 4.3.3.81
* bos.net.tcp.client 4.3.3.82
* bos.net.tcp.client 4.3.3.83
* bos.net.tcp.client 4.3.3.84
***************************************************************
*PROBLEM DESCRIPTION: *
* Unlock the socket before calling ip_freemoptions().
* Only thing ip_freemoptions does is call in_delmulti
* and in_delmulti should never be called with a socket
* lock. So, it is safe to unlock the socket as it was
* done before.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY34112
***************************************************************

PROBLEM CONCLUSION:
Unlock the socket before calling ip_freemoptions(). Only
thing ip_freemoptions does is call in_delmulti and
in_delmulti should never be called with a socket lock.
So, it is safe to unlock the socket as it was done before.

------

APAR: IY34169 COMPID: 5765C3403 REL: 430
ABSTRACT: CATIAV4 CRASHES THE SYSTEM WHEN CLOSING MODEL

PROBLEM DESCRIPTION:
Multi-threaded programs or libraries large enough to use the
loader overflow segment will crash the system.

PROBLEM CONCLUSION:
Loader environment initialized before processing the
pthreads stikylib request.

------

APAR: IY34194 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: POSSIBLE BUFFER OVERFLOW IN XDR CODE

PROBLEM DESCRIPTION:
Buffer overflow in the XDR (external data representation)
libraries. See CERT advisory CA-2002-25 for details.

PROBLEM SUMMARY:
Buffer overflow in the XDR (external data representation)
libraries. See CERT advisory CA-2002-25 for details.

PROBLEM CONCLUSION:
check the number of elements for an overflow condition.

------

APAR: IY34204 COMPID: 5765C3403 REL: 433
ABSTRACT: MEMORY LEAK IN FTPD CODE

PROBLEM DESCRIPTION:
Memory leak problem with ftpd

PROBLEM SUMMARY:
Memory leak problem with ftpd

PROBLEM CONCLUSION:
In ftpd code, we see some memory allocations being done, which
are not freed later. These memory allocations need to be
freed at the end of routines.

------

APAR: IY34224 COMPID: 5765C3403 REL: 430
ABSTRACT: ADD TRACE HOOKS TO PTY CODE

PROBLEM DESCRIPTION:
Add better trace hooks to the pty code.

PROBLEM CONCLUSION:
Added standard tty trace hooks.

------

APAR: IY34225 COMPID: 5765C3403 REL: 430
ABSTRACT: PCI 10/100 ETHERNET: ADDITIONAL CHOSEN_NETWORK_TYPE SUPPORT

PROBLEM DESCRIPTION:
NIM install support for new system firmware.

------

APAR: IY34249 COMPID: 5765C3403 REL: 433
ABSTRACT: PS REPORTS WRONG SZ/RSS VALUE

PROBLEM DESCRIPTION:
"ps" command some time display incorrect SZ/RSS
value. Above problem is because of wrong value
in the proc structure. "ps" commands reads proc
structure to report SZ/RSS value.

PROBLEM CONCLUSION:
The code which updates the proc structure was modified to
correct the problem

------

APAR: IY34255 COMPID: 5765E6110 REL: 220
ABSTRACT: HAGSGLSM DOES NOT CLOSE CSS DEVICE AFTER ATTEMPTING TO OPEN CAUS

------

APAR: IY34259 COMPID: 5765D5100 REL: 311
ABSTRACT: POSSIBLE SP SYSTEM ATTACHMENT ADAPTER DATA ERROR

PROBLEM DESCRIPTION:
Possible SP System Attachment Adapter Data Error

PROBLEM CONCLUSION:
The timing logic in the xilinx_file3pci
that is loaded on the SP System Attach Adapter (SPSAA,
service Feature code # 8396) has been adjusted to address
this problem.

------

APAR: IY34260 COMPID: 5765D5100 REL: 340
ABSTRACT: POSSIBLE SP SYSTEM ATTACHMENT ADAPTER DATA ERROR

PROBLEM DESCRIPTION:
Possible SP System Attachment Adapter Data Error

PROBLEM SUMMARY:
It is possible for a data error to occur when data
is transferred between systems interconnected by an
SP Switch using the SP System Attachment Adapter
(SPSAA server feature code # 8396).

PROBLEM CONCLUSION:
The timing logic in the xilinx_file3pci that is loaded
on to the SP System Attach Adapter (SPSAA, service
Feature code # 8396) has been adjusted to address this
problem.

------

APAR: IY34299 COMPID: 5765C3403 REL: 430
ABSTRACT: ICONV() CAN RETURN EBADF IN MULTITHREADED ENVIRONMENT

PROBLEM DESCRIPTION:
When using the iconv() subroutine in a multithreaded
application, the routine will intermittently
fail returning the return code EBADF.
This is an issue regarding timing and threads,
so it only occurs once in several thousand
attempts...

PROBLEM SUMMARY:
When using the iconv() subroutine in a multithreaded
application, the routine will intermittently
fail returning the return code EBADF.
This is an issue regarding timing and threads,
so it only occurs once in several thousand
attempts...

PROBLEM CONCLUSION:
Modify iconv library to use thread safe error
numbers, and recompile iconv converters in
thread safe fashion.

------

APAR: IY34307 COMPID: 5765C3403 REL: 433
ABSTRACT: LIBPERFSTAT DISK ERROR/MEMORY LEAK

PROBLEM DESCRIPTION:
ers of libperfstat.a disk API function will
experience a memory leak when consumers run
as root.

PROBLEM SUMMARY:
Users of the libperfstat disk
API may experience a memory
leak when consumers are run
as root

------

APAR: IY34344 COMPID: 5765B9501 REL: 320
ABSTRACT: NODE PANICKED WITH MMFLUSHSEG AND MMAP

PROBLEM DESCRIPTION:
node panicked with mmflushseg and mmap

PROBLEM SUMMARY:
GPFS panicked a node while running mmap'd
programs in a system with a heavy paging load.

PROBLEM CONCLUSION:
Don't use GP_TO_GNP to convert gnode
pointer to gpfsNode_t under disable_lock since the gnode might
be paged out.

------

APAR: IY34349 COMPID: 5765C3403 REL: 430
ABSTRACT: FIFO CAN LOSE DATA WHEN CLOSED

PROBLEM DESCRIPTION:
Writes of less than 32K to FIFO devices can be lost

PROBLEM SUMMARY:
FIFO devices can lose data when closed

PROBLEM CONCLUSION:
Force FIFO devices to wait for reads to
complete before closing

TEMPORARY FIX:
Use writes >32K

------

APAR: IY34370 COMPID: 5765C3403 REL: 430
ABSTRACT: PIODIGEST MIGHT FAIL WHEN CREATING POSTSCRIPT QUEUES

PROBLEM DESCRIPTION:
In rare instances where the file /usr/lib/ps/fontmap is
excessively large, piodigest might fail when a user is
adding postscript print queues.

LOCAL FIX:
Examine /usr/lib/ps/fontmap file to see if there
are font name repetitions.

PROBLEM SUMMARY:
In instances where the fontmap file is larger than expected
in /usr/lib/ps, piodigest might fail when creating new PS print
queues.

PROBLEM CONCLUSION:
piodigest has been fixed so it will be able to handle the
larger number of font definitions

------

APAR: IY34409 COMPID: 5765C3403 REL: 430
ABSTRACT: ICONV BUFFERING PROBLEMS WITH UTF-8 CONVERSIONS

PROBLEM DESCRIPTION:
When iconv() converts data from UTF-8 to UCS-2 and
doesn't have enough room in the output buffer, the
resulting position of the input buffer pointer is
incorrect ( i.e. advanced one character too far ).
As a result, other conversions that depend on this
conversion as an intermediate step, such as
iconv from UTF-8 to just about anything via
Universal_UCS_Conv converter, may produce incorrect
results with certain buffer sizes.

PROBLEM SUMMARY:
When iconv() converts data from UTF-8 to UCS-2 and
doesn't have enough room in the output buffer, the
resulting position of the input buffer pointer is
incorrect ( i.e. advanced one character too far ).
As a result, other conversions that depend on this
conversion as an intermediate step, such as
iconv from UTF-8 to just about anything via
Universal_UCS_Conv converter, may produce incorrect
results with certain buffer sizes.

PROBLEM CONCLUSION:
Modify UTF-8 to UCS-2 iconv() converter so that it doesn't
advance the input buffer pointer when the output buffer
is not large enough to contain the result.

------

APAR: IY34447 COMPID: 5765E5400 REL: 450
ABSTRACT: HACMP 4.5.0.2 MEDIA REFRESH BASE

PROBLEM DESCRIPTION:
HACMP 4.5.0.2 Media Rrefresh Base

PROBLEM SUMMARY:
HACMP 4.5.0.2 Media Refresh Base

PROBLEM CONCLUSION:
HACMP 4.5.0.2 Media Refresh Base

------

APAR: IY34489 COMPID: 5765C3403 REL: 433
ABSTRACT: VPATH_RESV_CFLICT ERROR WHEN CREATING MKSYSB ON ESS W/SDD

PROBLEM DESCRIPTION:
When creating a mksysb on ESS with SDD, the
command mkszfile will try to query disks
that belong to a VPATH. This will cause an
error to be logged into the error log.
VPATH_RESV_CFLICT error.

LOCAL FIX:
Ignore the error.

PROBLEM SUMMARY:
When mkszfile runs on systems with ESS
running SDD, errors are generated because mkszfile
queries these disks.
This causes reports and callouts to investigate
a possible error condition.

PROBLEM CONCLUSION:
The code was changed to not query volume groups
that are not autovaryon. These kinds of
volume groups fall into this state. The reason
we are querying volume groups is so that we can
autovaryon at boot time after install, therefore
there is no reason to query volume groups
that are not autovaryon.

TEMPORARY FIX:
None, just ignore messages as there is not
a problem, it is just informational.

------

APAR: IY34493 COMPID: 5765E88AP REL: 130
ABSTRACT: CSM FOR AIX INITIAL SUPPORT

PROBLEM DESCRIPTION:
csm for aix initial support

------

APAR: IY34576 COMPID: 5765C3403 REL: 433
ABSTRACT: ADDITIONAL CHOSEN_NETWORK_TYPE SUPPORT FOR GIGABIT ETHERNET

PROBLEM DESCRIPTION:
NIM installs fail on Gigabit ethernet when
latest system firmware is installed.

LOCAL FIX:
Go back to previous version of system firmware.

PROBLEM SUMMARY:
System firmware was changed to support additional
chosen_network_type fields for bootable LAN
adapters. This change also requires an updated
cfggxent config method for the gigabit ethernet
adapter device driver. Without this updated
config method, the NIM installs may fail.

PROBLEM CONCLUSION:
An updated cfggxent config method was created
which supports the new chosen_network_type
fields generated by the new system firmware.

TEMPORARY FIX:
Do not update systems to the latest level of
system firmware.

------

APAR: IY34577 COMPID: 5765C3403 REL: 430
ABSTRACT: PCI 10/100 ETHERNET: ADDITIONAL CHOSEN_NETWORK_TYPE SUPPORT

PROBLEM DESCRIPTION:
NIM install support for new system firmware.

------

APAR: IY34590 COMPID: 5765C3403 REL: 430
ABSTRACT: BROKEN FILESETS FOUND RUNNING SMITTY REJECT "ALL" UNABLE TO REJ

PROBLEM DESCRIPTION:
Attempting to reject applied filesets on a system restored
from a tape created with mksysb, results in broken filesets
due to an error of the form:
3001-008 /dev/rhdisk0 appears in the entries for files
/dev/ipldevice and /dev/rhdisk0.

PROBLEM CONCLUSION:
In a tcb enabled system that has been restored from mksysb,
the result of rejecting filesets could lead to multiple
filesets left in a BROKEN state.
To determine if a system could have this problem the user can
run the command: tcbck -n /dev/ipldevice
If the result from that command is ANYTHING OTHER than
3001-047 There is no matching file entry for /dev/ipldevice.
then the problem could happen.
The PTF delivered by this APAR should be applied and
committed to ensure that it is not accidentally rejected.
If this happens, the problem could show up again.

TEMPORARY FIX:
For a system that already has the problem, the user could
try to reinstall the broken filesets with the options of
APPLY, COMMIT, NOSAVE (not saving old data).

------

APAR: IY34591 COMPID: 5765C3403 REL: 430
ABSTRACT: INETD.CONF CHANGES IF SOME SCRIPT CALLS RC.NET DURING REBOOT

PROBLEM DESCRIPTION:
If some script calls rc.net during reboots, inetd.conf
file will get modified. Even though authentication methods
were not modified, if a script calls rc.net internally,
the 'shell' and 'login' lines of /etc/inetd.conf file will
get uncommented.

LOCAL FIX:
Comment out 'shell' and 'login' each time a script calls
rc.net.

PROBLEM CONCLUSION:
While reconfiguring inet0 device (from cfginet), if
authm has only the default value (std AIX), do
not call chauthent.

TEMPORARY FIX:
Comment out shell and login lines of inetd.conf file
manually each time after reboot

------

APAR: IY34595 COMPID: 5765C3403 REL: 430
ABSTRACT: FILE ACCESSES HANG IF AUTOMOUNTD IS KILLED

PROBLEM DESCRIPTION:
File accesses hang if automountd is killed or cored

PROBLEM CONCLUSION:
if autofs gets an error while trying to talk to autofs daemon,
it'll check to see if the daemon that it's talking to die and
there is a new daemon running. If there is then it'll retry
with a new connection to the new daemon else it'll print out
a message and fall back to the old behavior.

------

APAR: IY34597 COMPID: 5765C3403 REL: 430
ABSTRACT: KERNEL CRASH DURING CONFIG OF IBM CRYPTOGRAPHIC ACCELERATOR

PROBLEM DESCRIPTION:
Kernel crash during configuration of the IBM Cryptographic
Accelerator (devices.pci.1410e601). The kernel thread stack
at the time of the error shows the following:
00003EC0 .sc_debug_fixup+000030 ()
0000379C sc_trace_end_point+00000C ()
10001088 query_vpd+000090 (??, ??, ??, ??)
1000079C main+00049C (??, ??, ??)
10000188 __start+000088 ()

PROBLEM CONCLUSION:
Install the latest update to the IBM Cryptographic Accelerator
( devices.pci.1410e601.rte ) device driver.

------

APAR: IY34598 COMPID: 5765C3403 REL: 430
ABSTRACT: DISABLE GB ETHERNET CHECKSUM OFFLOAD FOR IMPROVED PERFORMANCE

PROBLEM DESCRIPTION:
Customer may experience degraded throughput since checksum
offload requires adapter resources.

PROBLEM CONCLUSION:
Introduce an ODM attribute, checksum_offload, which allows
the user to completely disable checksum offload on the
adapter.

------

APAR: IY34601 COMPID: 5765C3403 REL: 430
ABSTRACT: TIMING BUG CAUSES MACHINE TO CRASH WITH IFCONFIG DETACH

PROBLEM DESCRIPTION:
The tfree is issued without prior checkx for a pending timer
call. This could cause the machine to crash with
ifconfig detach.

PROBLEM CONCLUSION:
Add code to stop the timer object to make sure that there is
no timer call pending before free it.

------

APAR: IY34607 COMPID: 5765C3403 REL: 430
ABSTRACT: ROUTING TABLE NEEDS TO BE UPDATED, WHEN AN IP ADDRESS IS DELET

PROBLEM DESCRIPTION:
When we change/delete an ip address the routing table will
not be updated to reflect the changed address.

PROBLEM SUMMARY:
When we change/delete an ip address the routing table will
not be updated to reflect the changed address.

PROBLEM CONCLUSION:
Added a function to update the routing table, when an ip
address is changed/deleted.

------

APAR: IY34613 COMPID: 5765C3403 REL: 430
ABSTRACT: LABELS NOT CORRECT ON DESCRIPTION OF ETHERCHANNEL ON WEBSM

PROBLEM DESCRIPTION:
Mislabelled fields in the menus.

PROBLEM CONCLUSION:
Change index numbers in the file (adapter.pseudo.ibm_ech.add)
that populates PdAt for
uniquetype = "adapter/pseudo/ibm_ech"
nls_index = 70 & nls_index = 71
Swapping the index numbers for the two stanzas results
in swapping the labels on the two mislabelled fields.

------

APAR: IY34617 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: INSECURE TEMPORARY FILE CREATED BY DUMP_SMUTIL.SH

PROBLEM DESCRIPTION:
Insecure temporary file created.

PROBLEM CONCLUSION:
Prevent symlink exposure.

------

APAR: IY34620 COMPID: 5765D9300 REL: 310
ABSTRACT: _GETODMNN MAY CORRUPT MEMORY

PROBLEM DESCRIPTION:
_getodmnn may corrupt memory

PROBLEM SUMMARY:
During MPI_Init(), odm_set_path() is called for gathering
info from ODM database. The returned memory pointer is kept
in a variable for memory releasing. The variable was not
initialized and the validity of its value was not checked
before it being used in memory freeing. If ODM function
calls fail for some reason, invalid memory pointer could be
used in the memory freeing and cause memory corruption.

PROBLEM CONCLUSION:
Initialize the variable to NULL and check whether memory is
allocated before it is freed.

------

APAR: IY34624 COMPID: 5765C3403 REL: 430
ABSTRACT: ATM CONNECTION WON'T RECOVER AFTER CABLE RECONNECTED

PROBLEM DESCRIPTION:
Signalling does not restart when the fiber is disconnected
and reconnected from the card.

PROBLEM CONCLUSION:
The signaling daemon is ignore the request from call manger
on the restart. Process the request being sent by call
manager.

------

APAR: IY34626 COMPID: 5765C3403 REL: 430
ABSTRACT: ADD NEW IOCTL IN ATMCM

PROBLEM DESCRIPTION:
The ATM signalling code needs to know the exact number
of vci that the adapter supports.

PROBLEM CONCLUSION:
Add a new ioctl in atmcm.

------

APAR: IY34627 COMPID: 5765C3403 REL: 430
ABSTRACT: ATMSVCDS CHILDREN REPEATEDLY CORE DUMP DURING STARTUP.

PROBLEM DESCRIPTION:
atmsvcd's children repeatedly core dump because the value
returned for max_calls is not stored in the ILMI.cfg
structure.

PROBLEM CONCLUSION:
Store the value returned for max_calls in the ILMI.cfg
structure.

------

APAR: IY34628 COMPID: 5765C3403 REL: 430
ABSTRACT: DSIS WITH CRASHDUMPS SIZE ZERO OR NO CRASHDUMP TAKEN

PROBLEM DESCRIPTION:
It is possible for the system to not get a dump if
we get multiple exceptions nearly simultaneously,
on CPUs disabled to INTMAX.

PROBLEM CONCLUSION:
Don't grab the rtas_lock if dumping from the pal.

------

APAR: IY34630 COMPID: 5765C3403 REL: 430
ABSTRACT: CHMOD WITH G+T AND O+T DOES NOT RETURN RC>0

PROBLEM DESCRIPTION:
The chmod command when used with "o+t" and "g+t"
options, does not set the sticky bit, but returns
rc = 0. Instead, the rc must be greater than 0.

PROBLEM CONCLUSION:
Suitable code changes are done so that chmod
with "g+t" and "o+t" returns rc > 0.

------

APAR: IY34632 COMPID: 5765C3403 REL: 433
ABSTRACT: MKCD CREATES CD'S OF APPROX 32MB ONLY

PROBLEM DESCRIPTION:
mkcd only puts 32 MB of data on a CD before going to the
next volume.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* bos.sysmgt.sysbr 4.3.3.80ackup.
* bos.sysmgt.sysbr 4.3.3.81
* bos.sysmgt.sysbr 4.3.3.82
***************************************************************
*PROBLEM DESCRIPTION: *
* Errors in the calculations cause mkcd to only create
* volumes of about 32MB. This could result in many
* volumes (CD's) being required for a system backup.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY34632
***************************************************************

PROBLEM CONCLUSION:
The code has been modified to correct this
problem. The calculations have been corrected
to keep a running total of filesystem sizes.

------

APAR: IY34633 COMPID: 5765C3403 REL: 430
ABSTRACT: A MISSING PV MAY CAUSE AN LVM WRITE REQUEST TO HANG

PROBLEM DESCRIPTION:
If a logical volume has a missing PV, a write request could
cause a system crash.

PROBLEM CONCLUSION:
Check for missing PV's before a write is issued.

------

APAR: IY34634 COMPID: 5765C3403 REL: 430
ABSTRACT: INFINITE LOOP IN SYSLOGD

PROBLEM DESCRIPTION:
When the IP address is specified as remote host
for syslog forwarding, the requested syslog data
is not sent to the IP addressd host, but it is
put in the local syslog file repeated forever.
This problem is not found in case of being
specified the host name instead of IP address.

LOCAL FIX:
Use hostname instead of IP address for syslog
data forwarding.

PROBLEM SUMMARY:
When the IP address is specified as remote host for syslog
forwarding, the requested syslog data is not sent to the
IP addressd host, but it is put in the local syslog file
repeated forever.

PROBLEM CONCLUSION:
Corrected the IP address problem by zeroing out the proper
field of filed structure before it is assigned the IP addres.

------

APAR: IY34635 COMPID: 5765C3403 REL: 430
ABSTRACT: CORRUPTED CORE FILE FOR 64BIT APP ON 32BIT KERNEL

PROBLEM DESCRIPTION:
corrupted core file for 64bit app on 32bit kernel.

PROBLEM CONCLUSION:
unusable core file for 64bit app on 32bit kernel.

------

APAR: IY34637 COMPID: 5765C3403 REL: 430
ABSTRACT: SVMON -P DISPLAYS WRONG CMD NAME

PROBLEM DESCRIPTION:
Bad command name written on svmon -P reports

PROBLEM SUMMARY:
Bad command name written on svmon -P reports

PROBLEM CONCLUSION:
increase the size of the buffer passed through to getargs

------

APAR: IY34644 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: POSSIBLE DNS RESOLVER BUFFER OVERFLOW

PROBLEM DESCRIPTION:
Possible DNS resolver code buffer overflow problem.
If DNS resolver code gets a carefully crafted message from
DNS server. It's possible to trigger the buffer overflow.

PROBLEM CONCLUSION:
Need to update the "buflen" in DNS resolver code correctly
so that it will prevent the possible buffer overflow from
happening.

------

APAR: IY34645 COMPID: 5765C3403 REL: 430
ABSTRACT: POSSIBLE SYSTEM CRASH ON 32-BIT KERNEL

PROBLEM DESCRIPTION:
unlikely possibility of crash when sopyin_sigctx called for
32-bit processes

PROBLEM SUMMARY:
unlikely possibility of crash when sopyin_sigctx called for
32-bit processes

PROBLEM CONCLUSION:
call vm_att to ensure addressability of uthread

------

APAR: IY34647 COMPID: 5765C3403 REL: 430
ABSTRACT: APPLICATIONS CORE DUMP AFTER DETACHING FROM DBX

PROBLEM DESCRIPTION:
Applications coredump after detching from dbx.

LOCAL FIX:
export AIXTHREAD_SCOPE=S

PROBLEM SUMMARY:
Applications coredump after detaching from dbx.

PROBLEM CONCLUSION:
Application coredump after detaching from dbx.
This only happens in M:N mode.

------

APAR: IY34655 COMPID: 5765C3403 REL: 430
ABSTRACT: /USR/BIN/ADMIN CORE DUMPS WHEN GIVEN EXTREMELY LONG ARGUMENT

PROBLEM DESCRIPTION:
admin command dumps core when passed extremely long
argument.

PROBLEM CONCLUSION:
buffer overflow taken care to resolve core dump.

------

APAR: IY34662 COMPID: 5765E8200 REL: 240
ABSTRACT: MAINTENANCE UPGRADE

PROBLEM DESCRIPTION:
Maintenance Upgrade

------

APAR: IY34669 COMPID: 5765C3403 REL: 430
ABSTRACT: IFCONFIG ACCEPT WRONG NETMASKS

PROBLEM DESCRIPTION:
ifconfig accepts invalid ip addresses and netmask as input.
ie., it accepts addresses/netmask > 255.

PROBLEM CONCLUSION:
ifconfig was calling getnetbyname() which was using
inet_network() which was not checking the validity of the ip
address or netmask. Modified ifconfig to check the validity
of the ip address by using inet_addr() subroutine.

------

APAR: IY34670 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: BUFFER OVERFLOW IN NSLOOKUP

PROBLEM DESCRIPTION:
Possible buffer overflow in nslookup.

PROBLEM SUMMARY:
Possible buffer overflow in nslookup.

PROBLEM CONCLUSION:
use snprintf to replace sprintf

------

APAR: IY34671 COMPID: 5765C3403 REL: 430
ABSTRACT: INETD MAY CORE DUMP WHEN STARTED

PROBLEM DESCRIPTION:
Core dump when inetd starts.

PROBLEM SUMMARY:
Core dump when inetd starts.

PROBLEM CONCLUSION:
change sprintf to snprintf.

------

APAR: IY34676 COMPID: 5765C3403 REL: 430
ABSTRACT: PASSWD COREDUMPS WHEN PASSWORD HISTORY IS ENABLED IN LDAP

PROBLEM DESCRIPTION:
The passwd command core dumps after 8 times
of changing password for a ldap user when
the password history is enabled by setting
attribute histsize = 8 in LDAP. For local
users it may work fine.

LOCAL FIX:
Do not enable password history.

PROBLEM SUMMARY:
The passwd command core dumps after 8 times
of changing password for a ldap user when
the password history is enabled by setting
attribute histsize = 8 in LDAP. For local
users it may work fine.

PROBLEM CONCLUSION:
A memory write beyond the allocated amount is causing the
segmentation fault. This is removed.

------

APAR: IY34678 COMPID: 5765C3403 REL: 430
ABSTRACT: RPC.YPPASSWDD MAY COREDUMP W/ ALTERNATE PASSWD FILE

PROBLEM DESCRIPTION:
/usr/lib/netsvc/yp/rpc.yppasswdd may coredump if an alternate
password file is being used for NIS authentication and the
file /var/yp/passwd.ptmp already exists. The corefile would
have the following stacktrace:
(dbx) where
fclose_unlocked()
fclose()
changepasswd()
svc_getreqset()
svc_run()
main()

PROBLEM SUMMARY:
Fix cleanup code to ensure fclose is no longer called on
an invalid file pointer.

PROBLEM CONCLUSION:
Fix cleanup code to ensure fclose is no longer called on
an invalid file pointer.

------

APAR: IY34682 COMPID: 5765C3403 REL: 430
ABSTRACT: GRANTPT DOESN'T RETURN -1 ON FAILURE

PROBLEM DESCRIPTION:
The grantpt library routine does not return -1 upon failure.
Instead, it returns the error code.

PROBLEM CONCLUSION:
We changed the routine to return -1 upon failure, and to set
errno to the appropriate value.

------

APAR: IY34699 COMPID: 5697E3000 REL: 230
ABSTRACT: FIXES OF WNN7 FOR AIX 520

PROBLEM DESCRIPTION:
Japan kit V2.3 September fix package.

LOCAL FIX:
Change jkit.Wnn7.base.

PROBLEM SUMMARY:
This apar includes fixes of some problems.

PROBLEM CONCLUSION:
All problems we found were fixed.

------

APAR: IY34707 COMPID: 5765E6110 REL: 110
ABSTRACT: CSM FOR AIX INITIAL CLIENT SUPPORT ON AIX 5.1.F

PROBLEM DESCRIPTION:
csm for aix initial client support on aix 5.1.f

PROBLEM SUMMARY:
CSM for AIX Initial Client Support on
AIX 5.1.F

------

APAR: IY34724 COMPID: 5765F07AP REL: 230
ABSTRACT: RSCT FOR CSM FOR AIX INITIAL SUPPORT

PROBLEM DESCRIPTION:
rsct for csm for aix initial support

------

APAR: IY34725 COMPID: 5765E6110 REL: 221
ABSTRACT: RSCT FOR CSM FOR AIX INITIAL CLIENT SUPPORT ON AIX 5.1.F

PROBLEM DESCRIPTION:
rsct for csm for aix initial client support on aix 5.1.f

------

APAR: IY34727 COMPID: 5765E8200 REL: 240
ABSTRACT: 300 CRASH WITH GMD WHEN WE SET STATE MAP SIZE TO 1024 AND 2048

------

APAR: IY34728 COMPID: 5765E8200 REL: 240
ABSTRACT: SYNC COMMAND HANGS

PROBLEM DESCRIPTION:
Application writes to GeoMirror devices may hang on
a node which leaves the HACMP cluster by way of
forced node down.

PROBLEM SUMMARY:
Application writes to GeoMirror devices may hang on
a node which leaves the HACMP cluster by way of
forced node down.

PROBLEM CONCLUSION:
The forced node down event processing has been changed so
that the remote peer nodes can correctly determine the
GeoMirror device status on the node which has left the
HACMP cluster by way of forced node down.

------

APAR: IY34733 COMPID: 5765C3403 REL: 430
ABSTRACT: PHASE 1 PROPOSAL SPONTANEOUSLY CHANGING TO DEFAULT

PROBLEM DESCRIPTION:
The customer will not be able to modify the properties
of phase1 tunnel without viewing the proposals page

PROBLEM CONCLUSION:
The above problemm has been fixed by getting the policy name
and assigning the appropriate proposal to it.

------

APAR: IY34737 COMPID: 5765C3403 REL: 430
ABSTRACT: WEBSM CRASHES WHEN YOU ENTER NEWLINE AS PART OF TEXT

PROBLEM DESCRIPTION:
Websm crashes when newline characters are entered

PROBLEM SUMMARY:
Websm crashes when newline characters are entered

PROBLEM CONCLUSION:
Changed the code to not accept control characters and accept
backslashes as text.

------

APAR: IY34738 COMPID: 5765C3403 REL: 430
ABSTRACT: VMTUNE FAILS WITH MISSING SYMBOL __PTA_BALANCE_THRESH

PROBLEM DESCRIPTION:
The vmtune command fails to load due to missing symbol
__pta_balance_thresh.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* bos.adt.samples 4.3.3.77
***************************************************************
*PROBLEM DESCRIPTION: *
* The vmtune command fails to load due to missing symbol
* __pta_balance_thresh.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY34738
***************************************************************

PROBLEM CONCLUSION:
Add a requisite between bos.adt.samples and bos.mp/bos.up at
the 4.3.3.84 level.

------

APAR: IY34742 COMPID: 5765C3403 REL: 430
ABSTRACT: INCLUDE HOST NAME IN SUBJECT OF DIAGELA MAIL

PROBLEM DESCRIPTION:
diagela mail messages do not identify the sending system,
and some customers can not determine the sender from
the sendmail header.

PROBLEM CONCLUSION:
Include the host name that sent the diagela mail message
in the subject line of the message.

------

APAR: IY34743 COMPID: 5765C3403 REL: 430
ABSTRACT: PERIODIC DIAGNOSTICS IS NOT TESTING THE PROCESSORS

PROBLEM DESCRIPTION:
Periodic Diagnostics only does error log analysis on
processors. The notifications to the console from Periodic
Diagnostics and Automatic Error Log Analysis display only
the Service Request Number.

PROBLEM CONCLUSION:
The diagela script that manages Periodic Diagnostics will
change to invoke the processor diagnostic application
without the ELA flag.
The Periodic Diagnostics and Automatic Error Log Analysis
notification messages has been changed to include the SRN
error text along with the SRN number.
The floating point processor failure messages is changed to:
815-100: A floating point instruction produced a result that
is not correct. The processor should be replaced as soon as
possible. Verify any critical data calculated using the
floating point function.
This updates filesets bos.diag.rte, bos.diag.com and
devices.common.base.

------

APAR: IY34759 COMPID: 5765E7200 REL: 310
ABSTRACT: CHECK HANDLE TABLE TO PREVENT SYSTEM CALLING SETLOCAL

PROBLEM DESCRIPTION:
When handle table is empty, it throw an exception, which call
-ing un-thread-safe setlocal causing server coredump

PROBLEM SUMMARY:
When handle table is empty, it throw an exception, which call
-ing un-thread-safe setlocal causing server coredump

PROBLEM CONCLUSION:
check the handle table before accessing it

------

APAR: IY34761 COMPID: 5765E7200 REL: 310
ABSTRACT: OPEN A FILE (W/ MSWORD) FROM TWO CLIENTS TAKES LONG TIME

PROBLEM DESCRIPTION:
Some windows clients expect exact request size

PROBLEM SUMMARY:
Some windows clients expect exact request size

PROBLEM CONCLUSION:
re-caculate the size before sending it

------

APAR: IY34762 COMPID: 5765E7200 REL: 310
ABSTRACT: CIFSSERVER CORE DUMPS BY NET SESSION W/ LOT OF FILES OPENED

PROBLEM DESCRIPTION:
net session command core dump when listing more than 500
user sessions

PROBLEM SUMMARY:
net session command core dump when listing more than 500
user sessions

PROBLEM CONCLUSION:
matching data buffer size between the server and net command
to handle large number of user session information.

------

APAR: IY34770 COMPID: 5765C3403 REL: 430
ABSTRACT: CAN NOT RUN 'RMGROUP -R LDAP' TO REMOVE A LDAP GROUP

PROBLEM DESCRIPTION:
Administer can not remove a group defined on the ldap server
with the rmuser command, and he would get the following
error message:
3004-692 Error changing "users" to "".

PROBLEM CONCLUSION:
Modify code in ldap_delgroups.c, to verify the existence of
a group in ldap server as well as in the cache before deleting
it. Previously, it only checks the cache.

------

APAR: IY34772 COMPID: 5765C3403 REL: 430
ABSTRACT: CAN NOT CHANGE USER'S PGRP TO SYSTEM IN LDAP MODULE

PROBLEM DESCRIPTION:
Even if a ldap user's group list has system as one of the
groups, when the ldap user trying to change the pgrp to the
system group with the command
"chuser -R LDAP pgrp=system foo", he will get an error:
3004-692 Error changing "pgrp" to "system": Value is invalid.

PROBLEM CONCLUSION:
When converting string user ID to numerical ID, do not treat
as an error if the converted value is 0. System group's gid
is 0.

------

APAR: IY34773 COMPID: 5765C3403 REL: 430
ABSTRACT: 517 OR 553 LED HANG MINUTES ON SYNCVG ROOTVG IF TCB INSTALLED

PROBLEM DESCRIPTION:
0517 LED Hang on syncvg rootvg if tcb installed.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* devices.pci.14108c00.rte 4.3.3.52
* bos.rte.tty 4.3.3.79
* bos.sysmgt.serv_aid 4.3.3.82
***************************************************************
*PROBLEM DESCRIPTION: *
* 0517 LED Hang on syncvg rootvg if tcb installed.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY34773
***************************************************************

PROBLEM CONCLUSION:
Use tcbck to replace mksysck system call.

------

APAR: IY34774 COMPID: 5765C3403 REL: 430
ABSTRACT: X MAY NOT HANDLE PROTOCOL ERRORS WITH BIGREQUESTS CORRECTLY

PROBLEM DESCRIPTION:
X may not handle protocol errors with BigRequests correctly.

PROBLEM CONCLUSION:
Change StandardReadRequestFromClient so that oci->bufptr is
pointing to the next request after a BadLength protocol error
is sent to the client.

------

APAR: IY34775 COMPID: 5765C3403 REL: 430
ABSTRACT: PAX: LEAKS FILE DESCRIPTORS

PROBLEM DESCRIPTION:
pax reports error "error opening directory".

PROBLEM CONCLUSION:
pax code would be fixed to resolve the problem.

------

APAR: IY34777 COMPID: 5765C3403 REL: 430
ABSTRACT: CHANGE SMIT PASSWORD SETTING PANEL TO USE PWDADM

PROBLEM DESCRIPTION:
Members of security group can not change password for users
using smit.

PROBLEM CONCLUSION:
smit pannel was changed to call pwdadm when a member of
security group is changing passwd for another user.

------

APAR: IY34778 COMPID: 5765C3403 REL: 430
ABSTRACT: CALLOC GIVES SEGFAULT IF SIZE > 4GB

PROBLEM DESCRIPTION:
In 64-bit mode, memset behaves incorrectly, possibly causing
coredumps

PROBLEM CONCLUSION:
fixed instances where 64 bit register values were getting
clobbered.

------

APAR: IY34779 COMPID: 5765C3403 REL: 430
ABSTRACT: SHIP A DEBUG VERSION OF SMITTY

PROBLEM DESCRIPTION:
Smitty core dumps alot.

PROBLEM CONCLUSION:
Provide an official debug version of smitty to assist in
debuging.

------

APAR: IY34780 COMPID: 5765C3403 REL: 430
ABSTRACT: UNINITIALIZED LOCKWORD IN VSSCANF

PROBLEM DESCRIPTION:
uninitialized lockword in vsscanf() may cause core
dump on multithreaded applications.

PROBLEM CONCLUSION:
memset(0) the FILE structure in curses routines to avoid core
dump.

------

APAR: IY34781 COMPID: 5765C3403 REL: 430
ABSTRACT: LAZY LOADING WITH 8-CHARACTER SYMBOL NAMES FAILS

PROBLEM DESCRIPTION:
Calling lazily loaded routines that have
8-character names may fail.

PROBLEM CONCLUSION:
Change librtl's symbol lookup code to accommodate symbols
that are not nul-terminated.

------

APAR: IY34782 COMPID: 5765C3403 REL: 430
ABSTRACT: MALLOCDEBUG REPORT_ALLOCATIONS REPORTS INCORRECTLY

PROBLEM DESCRIPTION:
The wrong address will be displayed with the function name
and the function name may not always be displayed when using
MALLOCDEBUG's report_allocations option.

PROBLEM CONCLUSION:
Dereferenced the pointer to the return address to display the
return address instead of the address that points to the
address. Decode the traceback table to correctly output
the function name.

------

APAR: IY34783 COMPID: 5765C3403 REL: 430
ABSTRACT: DLPAR: DSI DURING MOVE CPU OPERATION

PROBLEM DESCRIPTION:
When the cpu_vector grows, it is not zeroed out. This creates
fields that appear to be initialized but are not. When those
fields are used, the system can crash.

PROBLEM CONCLUSION:
The solution is to zero out the new array when it is
allocated.

------

APAR: IY34784 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSTEM HANG OR CRASH DURING CPU ALLOCATION

PROBLEM DESCRIPTION:
System using 128-port asynchronous adapter can hang or
crash during CPU allocation.

PROBLEM CONCLUSION:
The cpu_vector was being reallocated without any type of
locking. The solution was to stop all the pollers and timers,
reallocate the array, and then start things back up. This was
chosen rather than a general locking scheme to save overhead
during normal operation.

------

APAR: IY34785 COMPID: 5765C3403 REL: 430
ABSTRACT: SUMAC.APT: ITI TSTART+000044 SCXMA_START_TIME+000050

PROBLEM DESCRIPTION:
The timer for the poller was started for each adapter instead
of just the first adapter on that cpu.

------

APAR: IY34786 COMPID: 5765C3403 REL: 430
ABSTRACT: WCRTOMB INCORRECTLY HANDLES THE FIRST ARGUMENT BEING NULL

PROBLEM DESCRIPTION:
when the first parameter to wcrtomb is NULL, the routine
simply returns 0, rather than exhibit standards compliant
behaviour

PROBLEM CONCLUSION:
When the first parameter to wcrtomb is NULL, act as if
wcrtomb(<internal buffer>, L'0', state) is called.

------

APAR: IY34787 COMPID: 5765C3403 REL: 430
ABSTRACT: THREAD-SAFING SETLOCALE AND NLS ROUTINES

PROBLEM DESCRIPTION:
Application coredumps, while doing NLS routines, with typical
stack traces of
ptrgl._ptrgl() at 0xd016ef90
mbtowc.mbtowc(??, ??, ??) at 0xd0170350
__mbstowcs_std(??, ??, ??, ??, ??, ??) at 0xd002cbd4
mbstowcs.mbstowcs(??, ??, ??) at 0xd0198e84
Reloading the locale in use causing this problem.

PROBLEM CONCLUSION:
Avoid reloading the locale.

------

APAR: IY34788 COMPID: 5765C3403 REL: 430
ABSTRACT: THREAD-SAFING SETLOCALE AND NLS ROUTINES

PROBLEM CONCLUSION:
Avoid reloading the locale.

------

APAR: IY34789 COMPID: 5765C3403 REL: 430
ABSTRACT: ATM ADAPTER UNABLE TO RECOVER FROM EEH ERROR

PROBLEM DESCRIPTION:
622Mbps ATM PCI Adapter fails to recover from an Enhanced
Error Handling (EEH) error, continuously logging
"ADAPTER ERROR" and "PROBLEM RESOLVED" error log entries.

PROBLEM SUMMARY:
622Mbps ATM PCI Adapter fails to recover from an Enhanced
Error Handling (EEH) error, continuously logging
"ADAPTER ERROR" and "PROBLEM RESOLVED" error log entries.

PROBLEM CONCLUSION:
Correctly reset the adapter statistic structure in the
device driver.

------

APAR: IY34790 COMPID: 5765C3403 REL: 430
ABSTRACT: IMPROVE HACMP FAILOVER USING FIBRE CHANNEL ADAPTER

PROBLEM DESCRIPTION:
There's a delay in opening the diskhead driver when multiple
ioctl's are stressing the adapter driver due to lock
contention. This can cause long failover times using HACMP
on Fibre Channel adapter.

PROBLEM SUMMARY:
There's a delay in opening the diskhead driver when multiple
ioctl's are stressing the adapter driver due to lock
contention. This can cause long failover times using HACMP
on Fibre Channel adapter.

PROBLEM CONCLUSION:
Release adapter driver's global lock to allows other processes
to go through in openning other adapters.

------

APAR: IY34791 COMPID: 5765C3403 REL: 430
ABSTRACT: POTENTIAL LOCK CONTENTION WHEN OPENNING MULTIPLE FSCSI DRIVER

PROBLEM DESCRIPTION:
Openning of multiple fscsi adapters may result in lock
contention when an adapter is going through error recovery
while many ioctls are pending.

PROBLEM CONCLUSION:
While multiple ioctls are pending, the locking strategy is
changed to allow openning of other adapters to go through.

------

APAR: IY34792 COMPID: 5765C3403 REL: 430
ABSTRACT: USING C API NAMES FOR GETCERTIFICATE AND VALIDATECERTIFICATE

PROBLEM DESCRIPTION:
The CPSD daemon is not started and the customer will not be
able to use authentication using certificates

PROBLEM CONCLUSION:
The code has been changed to use the new APIs provided by GSKIT

------

APAR: IY34793 COMPID: 5765C3403 REL: 430
ABSTRACT: COREDUMP ON LSVPD COMMAND WITH ULTRA-2 SCSI I/O CONTROLLER

PROBLEM DESCRIPTION:
A coredump on the lsvpd command can occur if a PCI based SCSI
Ultra-2 adapter (type 4-R, FC 5702) is installed in the
system.

PROBLEM CONCLUSION:
The coredump occurs because the end tag for the VPD is not
found when lsvpd scans the device VPD data on the system. The
end tag is not found because the adapter's config method and
driver do not write the end tag to the ODM. The solution is
to modify the VPD extraction routine to read all of the
adapter VPD data (including the end tag) and write it out to
ODM.

------

APAR: IY34794 COMPID: 5765C3403 REL: 430
ABSTRACT: ENSURE LOAD/STORE ORDER IN SCSI RAID DRIVER

PROBLEM DESCRIPTION:
Unpredictable symptoms can occur if loads/stores are not
performed in the correct order.

PROBLEM CONCLUSION:
Call an io_flush to ensure data syncronized.

------

APAR: IY34795 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSTEM CRASH RUNNING PCI 2-PORT MULTIPROTOCOL ADAPTER

PROBLEM DESCRIPTION:
pvthread+034300 STACK:
0003E6FC i_clear+0000C4 (0000000000000000 ?? )
0471FC38 free_adapter_services+000070 (??)
04722FE4 dpmp_close+00011C (??)
0013A0A8 ns_free+000058 (??)
00139908 ndd_usrreq+000160 (??, ??, ??, ??, ??)
00165D6C soclose2+0005F8 (??, ??)
0016BC78 soclose+000010 (??)
0016BB20 soo_close+0000C8 (??)
006EF8C0 closef+0000A4 (??)
0068D42C closefd+0000D8 (??, ??)
0068D284 close+000284 (??)
00003798 sc_msr_2_point+000028 ()
10000ED8 main+00074C (??, ??)
10000188 __start+000088 ()

PROBLEM SUMMARY:
pvthread+034300 STACK:
0003E6FC i_clear+0000C4 (0000000000000000 ?? )
0471FC38 free_adapter_services+000070 (??)
04722FE4 dpmp_close+00011C (??)
0013A0A8 ns_free+000058 (??)
00139908 ndd_usrreq+000160 (??, ??, ??, ??, ??)
00165D6C soclose2+0005F8 (??, ??)
0016BC78 soclose+000010 (??)
0016BB20 soo_close+0000C8 (??)
006EF8C0 closef+0000A4 (??)
0068D42C closefd+0000D8 (??, ??)
0068D284 close+000284 (??)
00003798 sc_msr_2_point+000028 ()
10000ED8 main+00074C (??, ??)
10000188 __start+000088 ()

PROBLEM CONCLUSION:
More serialization is required to keep dpmp_close from
closing before dpmp_open completes.

------

APAR: IY34796 COMPID: 5765C3403 REL: 430
ABSTRACT: TMD DOES NOT ACCEPT 0.0.0.0 AS A VALID NETMASK.

PROBLEM DESCRIPTION:
If a phase 2 ID payload specifies a netmask of 0.0.0.0, the
isakmp negotiation will succeed, but then tmd will throw an
error when trying to create the tunnel

PROBLEM CONCLUSION:
Make tmd recognize 32 bits of 0 as a valid netmask

------

APAR: IY34797 COMPID: 5765C3403 REL: 430
ABSTRACT: SECLDAPCLNTD THREAD RUNNING IN A INFINIT LOOP:

PROBLEM DESCRIPTION:
On SMP machines the secldapclntd consumes an increasing amount
of CPU time which decreases the overall system performance.
Analysis revealed that secldapclntdwas running in an infinite
loop of a socket read systemcall.
Each read returned immediately and there's no delay before the
next same call.

PROBLEM CONCLUSION:
The socket read system call return value is now checked for
0 to quit the indefinite for loop.

------

APAR: IY34798 COMPID: 5765C3403 REL: 430
ABSTRACT: SMITTY CHTZ APPENDING 2 COMMAS AFTER TZ

PROBLEM DESCRIPTION:
The problem can be reporduced by the following steps:
1. Run smitty chtz. 2. Enter the sub menu where daylight
savings exist. 3. All fields in the menu are blanked except
the time zone field. 4. Apply the settings by pressing enter.
On rebooting the system type "echo $TZ". The output shows 2
commas appended to the TZ.

PROBLEM SUMMARY:
****************************************************************
* CUSTOMERS AFFECTED:
* Systems running AIX 4.3 with bos.rte.date between the 4.3.3.51
* and 4.3.3.76 level.
****************************************************************
* PROBLEM DESCRIPTION:
* The TZ (timezone) variable is not set correctly using
* 'smit chtz'. This will cause the clock not to be set
* correctly during daylight savings time changes.
****************************************************************
* RECOMMENDATION:
* Use the 'chtz' command from the command line to set the TZ
* variable.
* The system must be rebooted for the change to completely take
* effect.
****************************************************************

PROBLEM CONCLUSION:
The script that smitty executes was changed
to fix this bug.

TEMPORARY FIX:
*********
* HIPER *
*********

------

APAR: IY34799 COMPID: 5765C3403 REL: 430
ABSTRACT: COMMANDS STRERR AND STRACE CORE DUMP OR WRITE BAD STRINGS

PROBLEM DESCRIPTION:
strlog() won't accept format %s, %x, %X, %u, %l, %o.

PROBLEM SUMMARY:
strlog() did not accept %s format.

PROBLEM CONCLUSION:
Change codes to let strlog() accept %s format. At same time,
avoid stack corrupting, stack overflow, un-expected interrupt
problems showed in 51 APAR: IY26801 & 43 APARs: IY32017 &
IY32016

------

APAR: IY34800 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: VULNERABILITY IN YPSERV

PROBLEM DESCRIPTION:
Possible security hole in ypserv.

PROBLEM SUMMARY:
Possible security hole in ypserv.

PROBLEM CONCLUSION:
Fix security problem.

------

APAR: IY34801 COMPID: 5765C3403 REL: 430
ABSTRACT: GXT6000 GXT6500 SHADING RENDERED INCORRECTLY IN FLUENT APP.

PROBLEM CONCLUSION:
Correctly handle the case where the light position
is 0,0,0,0 in all paths.

------

APAR: IY34803 COMPID: 5724C3505 REL: 310
ABSTRACT: ERROR.SEMANTIC ERRORS ON DEFINED VARIABLES.

PROBLEM DESCRIPTION:
An error.semantic variable not defined error is raised within
a subdialog. The reported problem error is defined in the same
page, and should be within the document`s scope.

LOCAL FIX:
Moving the variables to a different component on the same page
may resolve this problem.

PROBLEM SUMMARY:
An error.semantic variable not defined error is
raised within a subdialog. The reported problem error is
defined in the same page, and should be within the document

PROBLEM CONCLUSION:
Certain rules apply to the re-initialisation
and resetting of documents when control moves from one document
to another. These rules are defined in the VoiceXML spec
but have not been implemented correctly. The fix was to
implement the rules correctly.

------

APAR: IY34814 COMPID: 5765C3403 REL: 430
ABSTRACT: A 64BIT APPLICATION SEG FAULTS IN PASSWDEXPIRED()

PROBLEM DESCRIPTION:
A 64bit application calling AIX security
subroutine passwdexpired() will segfault
under certain conditions. This is the stack trace
from a DB2 application:
(dbx) where
allocate_topusr_val(??, ??, ??, ??) at 0x900000000094480
db_usrattr.getuserattr(0xfffffff, 0xfffff5f0, 0xfffffff,
                       0xfffff580) at 0x900000000096218
_getregistry(??) at 0x90000000008e5e8
passwdexpired(??, ??) at 0x9000000000fb934
unnamed block $b1, line 1871 in "sqlosecu.C"
sqlo_get_userpw(userpwinfo = 0x0ffffffffffff8ac,
                pSqloUbi = 0x00000001000029c8),
                             line 1871 in "sqlosecu.C"
main(??, ??) at 0x10000233c
This occurs only in the 64bit applications and 32bit
applications work fine. This will be seen in bos.rte.libc
filesets 4.3.3.81 or above in AIX 4.3 and 5.1.0.25 or
above in AIX 5.1.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* bos.adt.prof 4.3.3.80
* bos.adt.prof 4.3.3.81
* bos.adt.prof 4.3.3.82?, ??, ??) at 0x900000000094480
* bos.adt.prof 4.3.3.83fffffff, 0xfffff5f0, 0xfffffff,
* bos.adt.prof 4.3.3.84fffff580) at 0x900000000096218
_getregistry(??) at 0x90000000008e5e8
passwdexpired(??, ??) at 0x9000000000fb934
* bos.adt.prof 4.3.3.85871 in "sqlosecu.C"
* bos.rte.libc 4.3.3.81o = 0x0ffffffffffff8ac,
* bos.rte.libc 4.3.3.82= 0x00000001000029c8),
* bos.rte.libc 4.3.3.83 line 1871 in "sqlosecu.C"
* bos.rte.libc 4.3.3.843c
* bos.rte.libc 4.3.3.85
* bos.rte.libc 4.3.3.864bit applications and 32bit
***************************************************************
filesets 4.3.3.81 or above in AIX 4.3 and 5.1.0.25 or
above in AIX 5.1.
*PROBLEM DESCRIPTION: *
* A 64bit application calling AIX security subroutine
* passwdexpired() will segfault under certain
* conditions. This is the stack trace from a DB2
* application: (dbx) where allocate_topusr_val(??, ??,
* ??, ??) at 0x900000000094480
* db_usrattr.getuserattr(0xfffffff, 0xfffff5f0,
* 0xfffffff,
* 0xfffff580) at
* 0x900000000096218
* _getregistry(??) at 0x90000000008e5e8
* passwdexpired(??, ??) at 0x9000000000fb934 unnamed
* block $b1, line 1871 in "sqlosecu.C"
* sqlo_get_userpw(userpwinfo = 0x0ffffffffffff8ac,
* pSqloUbi = 0x00000001000029c8),
* line 1871 in "sqlosecu.C"
* main(??, ??) at 0x10000233c
*
* This occurs only in the 64bit applications and 32bit
* applications work fine. This will be seen in
* bos.rte.libc filesets 4.3.3.81 or above in AIX 4.3 and
* 5.1.0.25 or above in AIX 5.1.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY34814
***************************************************************

PROBLEM CONCLUSION:
Declare the prototype for a function call.

------

APAR: IY34844 COMPID: 5765C3403 REL: 430
ABSTRACT: CD-ROM FIRMWARE CHANGES IDENTIFICATION STRINGS OF DRIVE

PROBLEM DESCRIPTION:
Commands such as lsdev -Cc adapter
return device description as:
Other IDE CD-ROM Drive

PROBLEM CONCLUSION:
Add model_map attributes to .add files to match on more than
one ASCII string or substring. This handles more than one
production level of drive/firmware.

------

APAR: IY34888 COMPID: 5765C3403 REL: 430
ABSTRACT: GPROF GIVES BAD OUTPUT WITH EXP FILES IN LIBRARIES

PROBLEM DESCRIPTION:
While running gprof on some application which uses a
library which contains a exp file, the output may
not be correct.

PROBLEM SUMMARY:
The gprof command may give wrong output if
thre is a non-xcoff file, for example an
export file, in a library archive.

PROBLEM CONCLUSION:
The bug was fixed with code changes to
the gprof command.

------

APAR: IY35000 COMPID: 5765E8200 REL: 230
ABSTRACT: GMDSIZING DOES NOT WORK WITH VPATHS

PROBLEM DESCRIPTION:
If customer tries to issue the gmdsizing command on a system
where vpaths are defined, the command fails with the following
error:
vpath<#> not found in /dev/kmem; unable to read disk status from
kernel

PROBLEM SUMMARY:
Running gmdsizing with Shark disks and SDD installed. The
pipt code expects hdisks instead of vpaths

PROBLEM CONCLUSION:
Convert the vpaths to hdisks

------

APAR: IY35013 COMPID: 5765C3403 REL: 430
ABSTRACT: NEW TIMING REQUESTED FOR STEREO (1280X1024@96HZ)

PROBLEM DESCRIPTION:
GXT4500P/GXT6500P: new refresh rate added (1280x1024@96Hz).

PROBLEM CONCLUSION:
Add new timing to display timings table in device driver.

------

APAR: IY35021 COMPID: 5765E7200 REL: 310
ABSTRACT: SUPPORT FOR CIFS 3.1.2

PROBLEM DESCRIPTION:
Need an update path from CIFS 3.1.1 to CIFS 3.1.2

------

APAR: IY35046 COMPID: 5765E7200 REL: 310
ABSTRACT: NLS SUPPORT FOR CIFS 3.1.2

PROBLEM DESCRIPTION:
Need an NLS update path from CIFS 3.1.0 to CIFS 3.1.2

PROBLEM SUMMARY:
Need an NLS update path from CIFS 3.1.0 to CIFS 3.1.2
This APAR ships the following filesets, which are install
images rather than updates.
cifs.msg.Ja_JP
cifs.msg.ZH_CN
cifs.msg.Zh_CN
cifs.msg.Zh_TW
cifs.msg.ca_ES
cifs.msg.cs_CZ
cifs.msg.de_DE
cifs.msg.es_ES
cifs.msg.fr_FR
cifs.msg.hu_HU
cifs.msg.it_IT
cifs.msg.ja_JP
cifs.msg.ko_KR
cifs.msg.pl_PL
cifs.msg.pt_BR
cifs.msg.ru_RU
cifs.msg.sk_SK
cifs.msg.zh_CN
cifs.msg.zh_TW
This APAR 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 apply to installed
filesets on your system, use the command:
smit update_all

PROBLEM CONCLUSION:
Provide PTFs to update NLS from CIFS 3.1.0 to CIFS 3.1.2

------

APAR: IY35052 COMPID: 5765E8500 REL: 200
ABSTRACT: ERROR WHEN STARTING X25MON

PROBLEM DESCRIPTION:
When 'x25mon -fpct -n sx25a#' is used with the latest level
of the X.25 filesets including the 2.0.1.3 level
of sx25.rte, the following error is returned:
>>x25mon -fpct -n sx25a0
x25mon: 2200-016 Frame error receiving acknowledgement,
errno = 12

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* sx25.rte 2.0.1.3 IY32126 including the
* sx25.server 2.0.1.1rte:
* sx25.adt.rte 2.0.1.1
***************************************************************
*PROBLEM DESCRIPTION: *
>>x25mon -fpct -n sx25a0
x25mon: 2200-016 Frame error receiving acknowledgement,
* After updating to the latest X.25 filesets: for
* V2.0.1 with APAR IY32741 including the 2.0.1.3 level
* of sx25.rte; for V1.1.5 with APAR IY32126 includinghen
* the 1.1.5.24 level of sx25.rte:
* thod error: /usr/lib/method/cfgfx25 error
* various errors related to the frame layer may be
* received. One example for x25mon is: >>x25mon -fpct0x5
* -n sx25a0 x25mon: 2200-016 Frame error receiving
* acknowledgement,
* errno = 12
*
* Another example is that of an error message received
* when configuring a port: method error:
* /usr/lib/method/cfgfx25 error 0514-048 error
* downloading microcode or software cfgfx25 load frame
* profile failed fctl.dl_primitive = 0x5
***************************************************************
*RECOMMENDATION: *
* Apply apar IY35052
***************************************************************

PROBLEM CONCLUSION:
Verify that all executables changed due to:
IY32741 for V2.0.1
IY32126 for V1.1.5
are shipped.

------

APAR: IY35093 COMPID: 5765C3403 REL: 430
ABSTRACT: MAKE FAIL ON UNWANTED LIBRARY

PROBLEM DESCRIPTION:
In the makefile, if some archive member has an extension of
more than 1 character, make will give an error.

PROBLEM CONCLUSION:
Code modified so that make handles archive members with
extension of more than 1 character.

------

APAR: IY35163 COMPID: 5765C3403 REL: 430
ABSTRACT: 64-BIT THREADED PROGRAMS CRASH IN __LOADX

PROBLEM DESCRIPTION:
Users who install bos.rte.libpthreads >= 4.3.3.78 and
bos.rte.libc >= 4.3.3.84 without installing bos.64bit >=
4.3.3.77 experience 64-bit threaded application crashes at
program startup or when the pthreads library is loaded:
ptrgl_64._ptrgl() at 0x9000000000a0b2c
__loadx(0x0, 0x0, 0x0, 0x200ff718, 0x9001000,
0xa00af448, 0xc000000, 0xc000000) at x900000000095368
pthread_init() at 0x9000000000068e8
call_pthread_init() at 0x1001ea900
__threads_init.__threads_init() at 0x1000003ac
Environments without 64-bit applications are not affected.

PROBLEM CONCLUSION:
Add requisites from bos.rte.libc and bos.adt.prof to
bos.64bit 4.3.3.77.

------

APAR: IY35177 COMPID: 5765C3403 REL: 430
ABSTRACT: CMD_ELEM LEAK IN EFSC_START_PENDING_CMDS() ROUTINE.

PROBLEM DESCRIPTION:
The cmd_elem leak may strain on memory resource.

PROBLEM CONCLUSION:
Free the command element was just allocated, before return.

------

APAR: IY35184 COMPID: 5724C3505 REL: 310
ABSTRACT: GEOTEL CUSTOM SERVER CORE DUMPS DUE TO SIGCHILD

PROBLEM DESCRIPTION:
Sometimes a sigchild signal received by the GeoTel Custom Server
causes it to core dump.

PROBLEM SUMMARY:
GEOTEL CUSTOM SERVER CORE DUMPS DUE TO SIGCHILD

PROBLEM CONCLUSION:
Change socket read code such that if an
error occurs we do not memcpy 4GB of memory. Add code to
remember registration of ECC variables to stop the same
variable being registered twice.

------

APAR: IY35226 COMPID: 5724C3505 REL: 310
ABSTRACT: ISDN D CHANNEL BACKUP FEATURE