APAR: IY15408 COMPID: 5765D5100 REL: 320
ABSTRACT: POWER3 NODE HANGS IN CSSDD3:CSS_DELAY

PROBLEM DESCRIPTION:
10 out of 44 375Mhz_Power3 nodes (all thin) hung
at the same time. The dumps we forced show all
the same stack trace:
MST STACK TRACE:
0x003ffeb0
  (excpt=00000000:00000000:00000000:00000000:00000000)
  (intpri=3)
 IAR: . cssdd3:css_delay +68 (01397b78):
              lwz r4,0x68(r1)
 LR: . cssdd3:css_delay +3c (01397b4c)
 003ffb40: . cssdd3:css_reset_adapter +e4 (01399944)
 003ffbb0: . fault_service_SP:fault_service_inotice +2c4
              (013e63a0)
 003ffcf0: . cssdd3:tb3mx_intr +304 (013a7394)
 003ffd80: .i_poll_soft+94 (0001c5f4)
 003ffde0: .i_softmod+140 (0001c030)
 003ffe70: flih_603_patch+cc (000280dc)

PROBLEM SUMMARY:
A machine with an SP switch adapter can experience a system
hang when a switch problem occurs that requires resetting
the
adapter.

PROBLEM CONCLUSION:
The SP switch device driver has been changed to prevent
a system hang problem that can occur during switch
recovery when the adapter is reset.

------

APAR: IY17203 COMPID: 5639I3500 REL: 310
ABSTRACT: DCEMAN PAGES DO NOT WORK DUE TO FILE NAME CHANGES

PROBLEM DESCRIPTION:
dceman pages do not work. The names were
changed from the 2.2 level to represent
what the contents of the file, but dceman.c
still has the old file names hardcoded in
it.

PROBLEM SUMMARY:
The customer will not be able to view the DCE 3.1 documentation
using dceman.

PROBLEM CONCLUSION:
The filenames used in the DCE 3.1 release changed from the
DCE 2.2 release. The dceman tool needed to be updated to
instead use the DCE 3.1 filename convention.

------

APAR: IY17660 COMPID: 5639I3400 REL: 310
ABSTRACT: DTS SYNCINTERVAL AND MAXDRIFTRATE NEED TO CHANGED.

PROBLEM DESCRIPTION:
DTS syncinterval and maxdriftrate were set in Solaris
DCE 3.1 from AIX DCE. They need to be changed to reflect
the pre 3.1 Solaris attributes instead.
  syncinterval +1-00:00:000I in 3.1
  syncinterval +0-00:02:000I in 2.0 (should be this)
  maxdriftrate 50000 nsec/sec in 3.1
  maxdriftrate 1000000 nsec/sec in 2.0 (should be this)

PROBLEM SUMMARY:
Once dce was ported to work on Solaris, the maxdriftrate
was left at the AIX default(50,000), but this was not what
the Solaris customers wanted (1,000,000).

PROBLEM CONCLUSION:
An ifdef in the function InitializeKernelParams in
utc_adjtime_user.c was corrected to only include AIX and
not Solaris.

------

APAR: IY17796 COMPID: 5765C3403 REL: 430
ABSTRACT: LOCALEDEF FAILS TO GENERATE 64BIT DOUBLEBYTE

PROBLEM DESCRIPTION:
localedef failed to generate 64 bit locale
when user defined 64 bit compiled methods
are given as input.

PROBLEM SUMMARY:
The localedef command fails to generate 64-bit locale
given user defined 64-bit compiled methods.

PROBLEM CONCLUSION:
The command fails because the loader can not load
XCOFF64 object file since localedef command is in XCOFF32
format.

------

APAR: IY18374 COMPID: 5639I3700 REL: 310
ABSTRACT: DFSBIND MEMORY LEAKS AND CORES

PROBLEM DESCRIPTION:
PTF 2 introduced memory leak wherein dfsbind will
grow and finally coredump when the process ulimit
is hit.

PROBLEM SUMMARY:
dfsbind grows bigger and could result in coredump when
it hit's the process limits.

PROBLEM CONCLUSION:
code is modified to free the appropriate cred structures
after the use.

------

APAR: IY19118 COMPID: 5639I3400 REL: 310
ABSTRACT: RUNNING PASSWD_EXPORT, LAST ENTRY EXCLUDED FROM GROUP FILE.

PROBLEM DESCRIPTION:
The last entry in the security registry will not be
properly exported.

PROBLEM CONCLUSION:
When the root account is skipped during password export,
add one to the index so that the last value in the registry
is completely exhausted.

------

APAR: IY20170 COMPID: 5765C3403 REL: 433
ABSTRACT: NAMED8 IXFR FILE SIZE PROBLEM

PROBLEM DESCRIPTION:
named8 IXFR file size is growing untill the filesystem is
becoming full '.'

PROBLEM SUMMARY:
named8 IXFR file size is growing untill the filesystem is
becoming full.

PROBLEM CONCLUSION:
IXFR files will be removed after the zone updates take place.

------

APAR: IY20290 COMPID: 5639I3700 REL: 310
ABSTRACT: UUID HASH ERRORS IN ERROR.LOG WHEN USING AUDIT

PROBLEM DESCRIPTION:
When using audit, the following may appear in the error.log:
2001-01-02-12:40:43.100-05:00I----- cdsd(6808) ERROR aud esl
esl_in_core.c 896 0x00001112 msgID=0x10323064 Cannot get the
hash value of the subject uuid: uuid_to_string_failed ;
status: -Að (B, -A (BH0K{

PROBLEM SUMMARY:
The following will occur in the error.log:
2000-12-18-11:26:16.186-05:00I----- cdsd(19030) ERROR aud
esl esl_in_core.c 896 0x00000f10 msgID=0x10323064 Cannot
get the hash value of the subject uuid:
uuid_to_string_failed ; status:

PROBLEM CONCLUSION:
In dce_aud_start, make sure the empty_pac has been
completely initialized (0'ed out), to ensure no garbage
exists in the uuid fields.
Also in aud_esl_entry_lookup, make sure the error_string
has been set before printing out error messages.

------

APAR: IY20386 COMPID: 5765C3403 REL: 430
ABSTRACT: __COMMITTED SYMBOL NOT EXPORTED

PROBLEM DESCRIPTION:
Customer cannot set breakpoint in __committed function.

PROBLEM CONCLUSION:
The __committed symbol is exported from the libpthreads
library.

------

APAR: IY20442 COMPID: 5639I3900 REL: 310
ABSTRACT: CONFIG ERROR: COULDN'T SET LOCALE PROPERLY

PROBLEM DESCRIPTION:
"couldn't set locale properly" error given during

PROBLEM SUMMARY:
"couldn't set locale correctly" messages while running DCE
configuration commands.

PROBLEM CONCLUSION:
The LC_MESSAGES variable is saved and set eo en_US before
the "id" command is run. It was restored incorrectly. That
has been fixed.

------

APAR: IY20523 COMPID: 5639I3700 REL: 310
ABSTRACT: CDS REMASTER FAILS WHEN OLD MASTER CDSD IS DOWN

PROBLEM DESCRIPTION:
Attempts to remaster via "cdscp set dir to new epoch", excluding
a dead master CDS server, fail in Solaris DCE 3.1 w/ message"Un
able to cummunicate with any CDS server". This functionality wor
ks in AIX DCE 3.1 (and worked in Solaris DCE 2.0).

PROBLEM SUMMARY:
Attempts to remaster via "cdscp set dir to new epoch",
excluding a dead master CDS server, fail in Solaris DCE 3.1
w/ message "Unable to communicate with any CDS server".

PROBLEM CONCLUSION:
Change injecting defect 6026 ifdef from IBM_DCE to AIX_PROD
since this defect has not appeared to have affected AIX in
the same way (i.e. the remastering won't work when the
master cdsd is down.)

------

APAR: IY20668 COMPID: 5639I3400 REL: 310
ABSTRACT: DUPLIATE SYMBOL NAME ERRORS WHEN LINKING LIBDCE AND KERBEROS

PROBLEM DESCRIPTION:
Customer is unable to link libdce and a vanilla Kerberos
library into the same application, because of duplicate symbol
names.

PROBLEM SUMMARY:
The DCE library in version 3.1 has numerous symbol
conflicts with the MIT Kerberos 5 libraries. This
prevents linking a binary with the both the DCE library
and the Kerberos libraries.

PROBLEM CONCLUSION:
The following are the functions
that needs to be obfuscated in export.h.
asn12krb5_buf
asn1_decode_addrtype
asn1_decode_ap_options
asn1_decode_authdata_elt
asn1_decode_charstring
asn1_decode_checksum
asn1_decode_cksumtype
asn1_decode_encrypted_data
asn1_decode_encryption_key
asn1_decode_enctype
asn1_decode_int32
asn1_decode_integer
asn1_decode_kdc_options
asn1_decode_kerberos_time
asn1_decode_krb5_flags
asn1_decode_krb_cred_info
asn1_decode_kvno
asn1_decode_last_req
asn1_decode_last_req_entry
asn1_decode_null
asn1_decode_octet
asn1_decode_octetstring
asn1_decode_principal_name
asn1_decode_printablestring
asn1_decode_realm
asn1_decode_sequence_of_enctype
asn1_decode_sequence_of_krb_cred_info
asn1_decode_sequence_of_ticket
asn1_decode_ticket
asn1_decode_ticket_flags
asn1_decode_ui_2
asn1_decode_ui_4
asn1_decode_unsigned_integer
asn1_encode_authorization_data
asn1_encode_charstring
asn1_encode_checksum
asn1_encode_encrypted_data
asn1_encode_encryption_key
asn1_encode_etype_info
asn1_encode_ia5string
asn1_encode_integer
asn1_encode_kdc_options
asn1_encode_kerberos_time
asn1_encode_krb5_authdata_elt
asn1_encode_krb5_flags
asn1_encode_krb_cred_info
asn1_encode_last_req
asn1_encode_last_req_entry
asn1_encode_null
asn1_encode_octetstring
asn1_encode_principal_name
asn1_encode_printablestring
asn1_encode_realm
asn1_encode_sequence_of_enctype
asn1_encode_sequence_of_krb_cred_info
asn1_encode_sequence_of_ticket
asn1_encode_ticket
asn1_encode_ticket_flags
asn1_encode_transited_encoding
asn1_encode_ui_4
asn1_encode_unsigned_integer
asn1_get_id
asn1_get_length
asn1_get_sequence
asn1_get_tag
asn1_make_etag
asn1_make_id
asn1_make_length
asn1_make_set
asn1_make_string
asn1_make_tag
asn1buf_create
asn1buf_destroy
asn1buf_ensure_space
asn1buf_remove_octet
asn1buf_remove_octetstring
asn1buf_size
asn1buf_sync
asn1buf_unparse
asn1buf_wrap_data
decode_krb5_ap_rep
decode_krb5_ap_rep_enc_part
decode_krb5_ap_req
decode_krb5_authdata
decode_krb5_authenticator
decode_krb5_cred
decode_krb5_enc_cred_part
decode_krb5_enc_data
decode_krb5_enc_kdc_rep_part
decode_krb5_enc_priv_part
decode_krb5_enc_tkt_part
decode_krb5_error
decode_krb5_etype_info
decode_krb5_kdc_req_body
decode_krb5_pa_enc_ts
decode_krb5_padata_sequence
decode_krb5_predicted_sam_response
decode_krb5_priv
decode_krb5_safe
decode_krb5_sam_challenge
decode_krb5_tgs_rep
decode_krb5_tgs_req
decode_krb5_ticket
encode_krb5_ap_rep
encode_krb5_ap_rep_enc_part
encode_krb5_ap_req
encode_krb5_authdata
encode_krb5_authenticator
encode_krb5_cred
encode_krb5_enc_cred_part
encode_krb5_enc_data
encode_krb5_enc_kdc_rep_part
encode_krb5_enc_priv_part
encode_krb5_enc_tkt_part
encode_krb5_error
encode_krb5_etype_info
encode_krb5_kdc_req_body
encode_krb5_pa_enc_ts
encode_krb5_padata_sequence
encode_krb5_predicted_sam_response
encode_krb5_priv
encode_krb5_safe
encode_krb5_sam_challenge
encode_krb5_tgs_rep
encode_krb5_tgs_req
encode_krb5_ticket
gmt_mktime
krb5_get_credentials_renew
krb5_get_credentials_validate
krb5_mk_1cred

------

APAR: IY20707 COMPID: 5765C3403 REL: 430
ABSTRACT: LOG FILE CORRUPTION OR CORE FILE DUE TO MEMORY LEAKS BECAUSE OF

PROBLEM DESCRIPTION:
The dhcpsd log file gets corrupted or the dhcpsd
core dumps. This happens because the system
receives INFORM packets from clients and when it
responds there is Memory leak. This causes
dhcpsd memory usage to reach Max of 256Meg over
a period of time (5 to 7 days), due to which
we get ENOMEM whenever we do malloc in the
code. This causes the log file corruption.

LOCAL FIX:
Write a cron job which will stop and start the
dhcpsd server every two days.

PROBLEM SUMMARY:
This apar fixes two problems
1) Memory leak due to response to INFORM Request
2) Memory corruptions because not locking the file
descriptor structure before logging the message
to catalog files.

PROBLEM CONCLUSION:
The problem can seen in many different forms
1) core dump which shows the core is happening in
malloc routines.
2) corrupted log files.
3) ps v <pid> output for dhcpsd is always growing
and the process core dump when the memory usage
reaches close to 256M.

TEMPORARY FIX:
Recycle dhcpsd every two to three days.

------

APAR: IY21009 COMPID: 5765D5100 REL: 320
ABSTRACT: NODE THRESHOLD FOR BAD PKTS NEEDS ADJUSTMENT

PROBLEM DESCRIPTION:
Nodes somtimes receive bad service packets and/or CRC errors
on incoming packets when a non-node link in the Col sw takes
an error (e.g., EDC). The thresholds for these errors (trans
prot -> perm port) are low, such that nodes are artificially
dropped from the switch when such a link event occurs. These
thresholds should be adjusted to make the nodes more resilient.

LOCAL FIX:
Eunfence the node after it has been dropped.

PROBLEM SUMMARY:
Increased thresholds for TBIC sequence errors and bad
service packets to avoid thresholding that will result in a
fenced node. Also disabled TBIC CRC errors to aviod
thresholding. CRC errors are always associated with bad
service packet and are therefore redundant.

PROBLEM CONCLUSION:
Bad service packets, TBIC sequence and CRC errors were
thresholding and causing the node to be fenced.

------

APAR: IY21315 COMPID: 5765C3403 REL: 433
ABSTRACT: MEMORY LEAK IN TEMP

PROBLEM DESCRIPTION:
Memory leak in temp.

PROBLEM SUMMARY:
temp keeps on growing on the customer machine

PROBLEM CONCLUSION:
check for the saioq* is NULL. if it is then only allocate
memory

------

APAR: IY21333 COMPID: 5765C3403 REL: 430
ABSTRACT: AWK GIVING LEX ERROR EVEN AFTER U474580 PTF

PROBLEM DESCRIPTION:
Even after installing U474580 PTF, awk
gives LEX error.

PROBLEM SUMMARY:
LEX error encountered with awk/nawk

PROBLEM CONCLUSION:
Makefile changed to refer to the latest lex command
while building awk command.

------

APAR: IY21382 COMPID: 5765C3403 REL: 430
ABSTRACT: SYNCVG -V DOESN'T SYNC ALL PARTITIONS

PROBLEM DESCRIPTION:
When using enhanced concurrent LVM, syncing a volume group
with a factor greater than 1 will cause the passive node to
fail to sync properly.

PROBLEM SUMMARY:
syncvg doesn't sync all partitions

PROBLEM CONCLUSION:
Add logic for factors greater than 1

------

APAR: IY21384 COMPID: 5765C3403 REL: 430
ABSTRACT: BAD SCSI DISK VARIES OFF VG WITH QUORUM OFF

PROBLEM DESCRIPTION:
Using enhanced concurrent LVM, a bad SCSI disk can force a VG
to vary off, even with quorum off.

PROBLEM SUMMARY:
Bad disk varies off VG with quorum off

PROBLEM CONCLUSION:
Fixed hd_sa_cont function

------

APAR: IY21385 COMPID: 5765C3403 REL: 430
ABSTRACT: VARYONVG -C SAYS VG NOT CONCURRENT CAPABLE WITH FAILED DISK

PROBLEM DESCRIPTION:
varyonvg -c says VG not concurrent capable with a failed disk

PROBLEM SUMMARY:
Varyonvg -c fails with bad disk

PROBLEM CONCLUSION:
Add check for failed disk

------

APAR: IY21386 COMPID: 5765C3403 REL: 430
ABSTRACT: SYNCVG INTERMITTENTLY HANGS

PROBLEM DESCRIPTION:
syncvg intermittently hangs

PROBLEM CONCLUSION:
Corrected getlvodm and putlvodm commands

------

APAR: IY21452 COMPID: 5765D5100 REL: 320
ABSTRACT: NODES CAN DROP OFF SP SWITCH 2 WHEN VFS MOUNT POINT BECOMES

PROBLEM DESCRIPTION:
Nodes can drop off SP Switch 2 when a VFS mount point becomes
unavailable and handleNodeReadStatusPacket() takes too long
due to popen() hanging on computing sum of topology file.

PROBLEM SUMMARY:
The sdrd and the switch processes such as the fault service
daemon can hang when a system() call is made and VFS mount
points are unavailable (e.g., an NFS server is not servicing
a mount to a the client which is seeing a hang).

PROBLEM CONCLUSION:
  The script, rc.switch, has been changed to ensure that
processes it forks (such as the fault service daemon) are
not succeptable to hanging on system calls when VFS mount
points
are unavailable. The SDR daemon, sdrd, has also been
changed to
avoid these hangs.

------

APAR: IY21511 COMPID: 5765C3403 REL: 430
ABSTRACT: CTL CODE CALLS XTRELEASEGC WITH BAD GC

PROBLEM DESCRIPTION:
TextOut.c in MOTIF2.1 has to check if the LayoutActive is set
or not before the allocation and the release of the BiDi added
Graphical Context (dc_save_gc and dc_imagegc).

PROBLEM SUMMARY:
CTL code calls XtReleaseGC with bad GC

------

APAR: IY21609 COMPID: 5765C3403 REL: 430
ABSTRACT: POTENTIAL RPC.YPASSWDD BUFFER OVERFLOW

PROBLEM DESCRIPTION:
There is risk for buffer overflows if bad data is sent to the
rpc.yppasswdd daemon. This could potentially be exploited to
gain unauthorized access to the system.

LOCAL FIX:
Protecting your network from the internet with a firewall
avoids the risk of someone outside of your network trying to
exploit this problem. However, best of all is to not run
the daemon unless the functionality provided by it is needed.

PROBLEM CONCLUSION:
Make sure that the string arguments passed in the RPC call are
not too long for the static string buffers used to hold them
during the processing. This requires changes in a handful of
code locations.

------

APAR: IY21753 COMPID: 5765D5100 REL: 320
ABSTRACT: SWITCH TABLE DOES NOT CLEAR

PROBLEM DESCRIPTION:
Switch table window is not released and cleared for follow
on jobs to use.

PROBLEM SUMMARY:
Switch table window is not released and cleared by the hung
task for follow on jobs to use.

PROBLEM CONCLUSION:
The switch table window will now free the resource that is
held by the hung task so that follow on jobs will be able to
start.

------

APAR: IY21818 COMPID: 5765D5100 REL: 320
ABSTRACT: LOADL GETS WRONG INFO ABOUT AVAILABLE SWITCH ADAPTER MEMORY

PROBLEM DESCRIPTION:
LOADL does not get back what it expects from an ioctl call for
the maximum amount of memory onthe switch adapter, less some
memory dedicated to non-loadl. It actially gets back the
current available amount of memory. As more jobs run on a node,
the less amount of adapter memory LOADL will see. It will reach
a point that it does not think a job could ever run on that node
, because enough memory would never be available. Node pool
decreases.

PROBLEM SUMMARY:
the ioctl to query the adapter resources was returning the
wrong data item for the amount of memory available on the
adapter.

PROBLEM CONCLUSION:
return the correct value

------

APAR: IY21998 COMPID: 5765C3403 REL: 430
ABSTRACT: LIB BIDI AIX 4.3 INCOMPATIBLE WITH AIX 4.2

PROBLEM DESCRIPTION:
Lib BiDi on AIX 4.3 gives different levels from that given by AI
X 4.2

PROBLEM SUMMARY:
LIB BiDi for AIX4.3 is not compatible with AIX4.2

------

APAR: IY22006 COMPID: 5639I3900 REL: 310
ABSTRACT: SEC_RGY_PGO_GET_MEMBERS() DOES NOT SET STATUS PROPERLY

PROBLEM DESCRIPTION:
DCE API call sec_rgy_pgo_get_members() does not set status
parameter to sec_rgy_no_more_entries when at the end of
the membership list for a group.

PROBLEM CONCLUSION:
rsdb_pgo_get_members_all() in rsdb_pgo.c line 1894:
if (member_cursor->handle > *number_members) {
shoud be:
if (member_cursor->handle >= *number_members) {
because the handle is EQUAL to number of numbers because
the handle is ZERO based index number.

------

APAR: IY22051 COMPID: 5639I3400 REL: 310
ABSTRACT: 0 LENGTH PE_SITE FILE WHEN / (ROOT) IS FULL

PROBLEM DESCRIPTION:
0 length pe_site file when / (root) is full

PROBLEM SUMMARY:
Symptom is that a full / (root) file system while DCE is
running could result in a 0-byte length
/opt/dcelocal/etc/security/pe_site file. When
DCE is restarted, the security server will fail to come
up, so DCE will not run until the pe_site file is
replaced with a good one.

PROBLEM CONCLUSION:
In DCED functions rebuild_pe_site() and rebuild_krb5_cnf(),
after the temporary file is believed to have been created
(because the fwrite() calls are returning '1' even though the
data is not committed to disk), the fclose() call must be
checked as follows:
    #if defined ( IBM_DCE ) if ( fclose (pesitefile) != 0 ) {
     DCE_SVC_DEBUG((dhd_svc_handle, dhd_s_secval, svc_c_debug1,
                     "File write error with tmp pe_site
                     file."));
     remove(tmp_pe_site_file); /* Erase tmp copy if it still
     exists*/ return 1; } #else (void)fclose (pesitefile);
    #endif
What happens is that the fclose() call performs an underlying
write() call, and this call will fail and the return code will
be propagated to fclose(), which will return EOF and set errno
to ENOSPC (28).

------

APAR: IY22186 COMPID: 5765C3403 REL: 430
ABSTRACT: IFCONFIG.AT.C DOES NOT DISPLAY THE ISNO

PROBLEM DESCRIPTION:
ifconfig.at.c does not display isno.

PROBLEM SUMMARY:
Run the ifconfig for atm interface. The output dosen't show
 the isno values
values.

PROBLEM CONCLUSION:
Variables added to structure. And changed the cfgif.at.c. So
that it picks the values from CuAt ODM, after rebooting.

------

APAR: IY22188 COMPID: 5765C3403 REL: 430
ABSTRACT: HANG IN IN_PCBBIND_NOLOCK

PROBLEM DESCRIPTION:
If an application tries to bind to a port, which is already in
use, the call returns without unlocking. When we try to do a
bind again to another port, we'll be waiting on a lock,
resulting in a deadlock.

PROBLEM SUMMARY:
If an application tries to bind to a port, which is already in
use, the call returns without unlocking. When we try to do a
bind again to another port, we'll be waiting on a lock,
resulting in a deadlock.

PROBLEM CONCLUSION:
Added INUDPHEAD_WRITE_UNLOCK before exiting the function.

------

APAR: IY22195 COMPID: 5639I3400 REL: 310
ABSTRACT: MEMORY CORRUPTION FOUND BY WEBSPHERE APPLICATION

PROBLEM DESCRIPTION:
Memory corruption found by Websphere application.

PROBLEM SUMMARY:
Websphere application core dumps with an agent not starting
using DCE 3.1 ptf set 4.

PROBLEM CONCLUSION:
In the routine sec_krb_gss_build_message,
increased the array size from 1024 to 2048
in src/security/utils/seca_gss.c file at line 560

------

APAR: IY22233 COMPID: 5697F6400 REL: 640
ABSTRACT: MESSAGE CENTER WEB INTERFACE ERROR 404

PROBLEM DESCRIPTION:
Message Center Web interface error 404
when going to Remote message settings from
Telephone interface settings (but from all other menus it is OK)
This is due to tellS.jsp going to remMs.jsp instead of

PROBLEM SUMMARY:
Error 404 when going from Telephone Interface
settings to Remote message settings.

PROBLEM CONCLUSION:
Modified mispelled link from remMs.jsp to
remMS.jsp in telIS.jsp

------

APAR: IY22236 COMPID: 5765C3403 REL: 430
ABSTRACT: LDR_CNTRL=LOADPUBLIC/LOADER DOMAIN BEHAVIOR HAS BEEN REMOVED

PROBLEM DESCRIPTION:
LDR_CNTRL=LOADPUBLIC/loader domain behavior has been removed

------

APAR: IY22283 COMPID: 5697F6400 REL: 630
ABSTRACT: MESSAGE CENTER FAX ID DISPLAYS BROOKTROUT.

PROBLEM DESCRIPTION:
When a fax machine calls Message Center, its display shows
the other fax machine id as Brooktrout Fax.
This is inappropriate - Message Center would be more
appropriate.

PROBLEM SUMMARY:
Fax machines display 'Brooktrout Fax' when
sending or receiving faxes to/from Message Center.

PROBLEM CONCLUSION:
Code fixed to output 'Message Center'
instead of 'Brooktrout Fax'.

------

APAR: IY22426 COMPID: 5765C3403 REL: 430
ABSTRACT: INFINITE LOOP IN KERNEL FROM CLOSE OF SOCKET

PROBLEM DESCRIPTION:
Infinite loop in kernel from close of socket

PROBLEM SUMMARY:
System hangs on soclose when user try to close a socket with
data on send Q

PROBLEM CONCLUSION:
Code is modified in tcp_input0() to handle this kind of
situation

------

APAR: IY22448 COMPID: 5765D5100 REL: 320
ABSTRACT: HSD ROUTINE CALLED INTERRUPTS DISABLED CRASH

PROBLEM DESCRIPTION:
Several years ago the vsd driver was changed from an interrupt
level driver to a process level mult-threaded driver. The hsd
driver was not changed from the interrupt level since no problem
had occurred. There is a case where the hsd driver, while holdi
an interrupt level lock, calls the vsd driver which in turn call
a system routine that cannot be called with interrupts disabled.
This is what cause the crash.

PROBLEM SUMMARY:
There is a case where the hsd driver, while holding
an interrupt level lock, calls the vsd driver which
in turn calls a system routine that cannot be called
with interrupts disabled. This is what cause the crash.

PROBLEM CONCLUSION:
The HSD device driver locking has been restructured.

------

APAR: IY22584 COMPID: 5765C3403 REL: 430
ABSTRACT: STATIC LIMIT OF 768 TCP FILEDESCRIPTORS FOR RPC MADE DYNAMIC

PROBLEM DESCRIPTION:
There is a static limitation of the number of file descriptors
available for RPC-based programs. If this limit is exceeded,
the application might coredump in makefd_xprt. Current limit is
1024, but since 256 is the first descriptor used, in practice
this sets a limit of 768 open TCP connections.

PROBLEM CONCLUSION:
Make sure that the __rpc_rlim_max() function uses the set
resource limits for files (+256 due to the offset) instead of
the static 1024 that used to be.

------

APAR: IY22732 COMPID: 5765D5100 REL: 320
ABSTRACT: FIRMWARE SUPPORT --SYSTEM MANAGEMENT CHANGES

PROBLEM DESCRIPTION:
pssp fibrechannel boot - system management changes

PROBLEM SUMMARY:
support for hard disk specifictaion in
phusical volume identifier (pvid) format is required.

PROBLEM CONCLUSION:
support for hard disk specification
in physical volume identifier (PVID) format has been added.

------

APAR: IY22890 COMPID: 5765C3403 REL: 430
ABSTRACT: REMOVE 1 SEC DELAY

PROBLEM DESCRIPTION:
rshd can delay while connecting stderr port.
Need to delay only when appropriate.

PROBLEM SUMMARY:
inetd slept for a second when not necessary.

PROBLEM CONCLUSION:
inetd no longer sleeps for one second except on error.

------

APAR: IY22911 COMPID: 5765C3403 REL: 430
ABSTRACT: TO REDEFINE PRN*,SCN* AS STRINGS WITHOUT % IN INTTYPES.H

PROBLEM DESCRIPTION:
The PRI* and SCN* macros would not work as documented by ISO
standards documents.

PROBLEM SUMMARY:
To redefine PRN*,SCN* as strings without % in inttypes.h

PROBLEM CONCLUSION:
PRN*,SCN* macros in inttypes.h were redefined as strings
without %sign.

------

APAR: IY23005 COMPID: 5765C3403 REL: 430
ABSTRACT: AUTOMOUNT MAP SEARCH ORDER CHANGED BETWEEN 4.3.2 AND 4.3.3

PROBLEM DESCRIPTION:
In AIX 4.3.2, NIS automount maps were searched before files.
This changed in 4.3.3, causing problems if there was a file in
/etc with the same name as the NIS automount map.

LOCAL FIX:
Remove the local file and the NIS map will be used instead.

PROBLEM CONCLUSION:
Move NIS to before files in the search order for autofs maps.

------

APAR: IY23007 COMPID: 5765C3403 REL: 430
ABSTRACT: YPMATCH COMMAND HANGS FOR KEY SIZE GREATER THAN 324 CHARACTERS

PROBLEM DESCRIPTION:
ypmatch command will hang for passwd map size more
than 324 charcters.

PROBLEM CONCLUSION:
Check for map_len in match_one() function for 324 charcters.
If the map size is more or equal than 324,pop the error
message

------

APAR: IY23029 COMPID: 5765B9501 REL: 330
ABSTRACT: ASSERT: SGIHOLD >= 1 + (THECLUSTERCFG->CCMGR == MYNODEADDR)

PROBLEM DESCRIPTION:
Assert: sgiHold >= 1 + (theClusterCfg->ccMgr == myNodeAddr)

PROBLEM SUMMARY:
GPFS self check logic found a lock conflict while running
with hierarchical storage enabled and TSM.

PROBLEM CONCLUSION:
 acquireAccessM() doesn't incremnt sgiHold anymore.
Caller needs to increment it before calling this function

------

APAR: IY23249 COMPID: 5765C3403 REL: 430
ABSTRACT: IBM GIGABIT ETHERNET-SX ADAPTER INCORRECT LINK STATUS FOLLOWING

PROBLEM DESCRIPTION:
Following system power cycle reboot, the IBM Gigabit
Ethernet-SX adapter shows an incorrect link status
via the entstat command. Specifically, the link
status will be indicated as "down" even if the link
is actually up and running.

LOCAL FIX:
A workaround is to unplug the cable from the adapter
for just a second or two, and plug it back in. The
adapter will then report the correct link status.

------

APAR: IY23266 COMPID: 5639I3700 REL: 310
ABSTRACT: DCECP PUTS TERMINATES SSH SESSION IN AIX4.3.3.10 AND ABOVE

PROBLEM DESCRIPTION:
ssh session is killed when DCE config scripts are used
from within it or when puts "" is run using dcecp tcl
interpreter

PROBLEM CONCLUSION:
Change in Tcl_PutsCmd function of file tclUnixAZ.c to avoid
sending just a NULL character to terminal (using write())
when user calls puts "". If user says puts "", write() is
not called at all.

------

APAR: IY23267 COMPID: 5639I3700 REL: 310
ABSTRACT: DCECP ADDS A NEWLINE CHARACTER BEFORE ITS TERMINATION.

PROBLEM DESCRIPTION:
A newline is created when a script is run which uses dcecp
as TCL interpreter as dcecp adds a newline before
termination.

PROBLEM CONCLUSION:
The fix to this problem will make the dcecp to terminate
in a normal fashion without adding any newline as it was
doing earlier.

------

APAR: IY23268 COMPID: 5639I3700 REL: 310
ABSTRACT: LM: FIX GENERAL SECURITY/RPC MEMORY LEAKS

PROBLEM DESCRIPTION:
For large cells, secd can grow 10-50M per day

PROBLEM CONCLUSION:
free malloc'ed structs in normal and error conditions
(see defect notes)

------

APAR: IY23269 COMPID: 5639I3700 REL: 310
ABSTRACT: CDS_GDAPOINTERS() DOES NOT GET CLEANED UP WHEN GDA IS UNCONFIGD

PROBLEM DESCRIPTION:
When GDA is unconfigured, the CDS_GDAPointers attribute does
not get updated to reflect that the GDA entry for the previous
configuration has gone away. The user may see multiple entries
associated with the CDS_GDAPointers attribute, including stale
entries associated with previous configuration of the GDA into
the cell. Cross-cell lookups may take longer due to these
stale entries.

PROBLEM CONCLUSION:
A new function, cleanup_parent() in gda_update_parent.c is
added to clean up the stale entry when gda is unconfigured or
stopped.

------

APAR: IY23271 COMPID: 5639I3700 REL: 310
ABSTRACT: DISABLE REPLAY CACHE IN THE SECD KDC

PROBLEM DESCRIPTION:
IBM Rochester saw performance bottleneck in their secd
due to I/O to the replay cache file.

PROBLEM CONCLUSION:
Disable the KDC replay cache in secd. The replay cache in
the secd provides no additional security and it can be a
(significant) performance issue.
Here's where we can disable the replay cache in
the process TGS request code path:
process_tgs_req() in src/security/krb5/kdc/do_tgs_req.c
|
+- kdc_process_tgs_req()
                    in src/security/krb5/kdc/kdc_util.c
   kdc_rcache is global, set up in rs_main.c
   |
   krb5_rd_req_decoded_anyflag(... kdc_rcache, ... )
   |
   +- krb5_rd_req_decoded_opt()
   /* only check rcache if sender has provided one---
      some services may not be able to use replay caches
      (such as datagram servers) */
   if (rcache)
   {
      ...
   }
The new Kerberos code passes in a NULL rcache for calling
krb5_rd_req_decoded_anyflag() and has other replay cache
calls like krb5_auth_con_setrcache() in
kdc_process_tgs_req(), which are disabled with the
"kdc_use_rcache" boolean added to the Kerberos code.
We should be safe passing in NULL as the rcache value to
krb5_rd_req_decoded_anyflag(), and we'll determine this
by a new environment variable
   KDC_USE_RCACHE=1
The default will be OFF for performance. The rcache file
will still be created and initialized, but the I/O
associates with TGS_REQ traffic should be minimized.

------

APAR: IY23273 COMPID: 5639I3700 REL: 310
ABSTRACT: 'REALMS' SECTION IN /ETC/KRB5.CONF BEING OVERWRITTEN

PROBLEM DESCRIPTION:
When /etc/krb5.conf is re-built, in the 'realms' section
any entries other than kdc will be erased.

PROBLEM CONCLUSION:
The parsing logic for the rebuild of /etc/krb5.conf has
been changed such that, in the 'realms' section, it retains
the entries other than kdc.

------

APAR: IY23274 COMPID: 5639I3700 REL: 310
ABSTRACT: TKT_FLG_PRE_AUTH WAS NOT SET IN THE TGT

PROBLEM DESCRIPTION:
Customer having a principal with the pre_auth_req ERA set,
and using MIT Kerberos V5 klist with -f option, is not
able to see TKT_FLG_PRE_AUTH flag in the klist output.

PROBLEM CONCLUSION:
The kdc checks for the existence of an instance of the ERA
"pre_auth_req" in the function build_princ_entry(). If it
is found, a flag is set so that functions called by
process_as_req() have access to this information. But there
was no check in process_as_req() to retrieve this info. The
problem was fixed by retrieving the same.

------

APAR: IY23275 COMPID: 5639I3700 REL: 310
ABSTRACT: LOCAL6.INFO LOGGING CAUSES FILESYSTEM TO FILL UP FASTER.

PROBLEM DESCRIPTION:
In aixdce22, the 'hosts/<hostname>/self' identity
authenticates about every 10 minutes. So, when the
local6.info logging is enabled using the AIX syslogd,
there is an increased number of entries in the log
files that causes the filesystems to fill up faster.

PROBLEM CONCLUSION:
Introduction of a new environment variable
DCE_SEC_KEYMGMT_WAKEUP_INTERVAL which controls how often a
client wakes up to check that the expiration policy has not
changed. In this approach, the default time will remain as
TEN_MINUTES. Only those users who want to enable the
local6.info logging facility and face the problem of
filesystem getting filled up faster can set the environment
variable to a higher value and solve the problem.

------

APAR: IY23276 COMPID: 5639I3700 REL: 310
ABSTRACT: RS_LOGIN_ACTIVITY_UPDATE RETURNS ERROR BREAKING LOGIN_RESET

PROBLEM DESCRIPTION:
On using login_reset to unlock the disabled account, the
customer will see the following error :-
Could not store new entry (status 0x39e4c5d0)
status 0x39e4c5d0 (???/zne)

PROBLEM CONCLUSION:
Clear the status in rs_login_activity_update() before we
call rsdb_name_canonicalize_and_convert(), so that we do
not check the junk value in status after return from the
function rsdb_name_canonicalize_and_convert().

------

APAR: IY23277 COMPID: 5639I3700 REL: 310
ABSTRACT: LOCAL UNCONFIG SEC_SRV -FORCE REMOVES FILES REQUIRED BY SEC_CL

PROBLEM DESCRIPTION:
If a registry designate is done to swap the roles of the
master security server and a replica, and then the old
master security server is unconfigured, any remaining DCE
daemons on the old Master security server machine will not
work properly.

PROBLEM CONCLUSION:
The unconfig code was modified to not delete necessary
files.

------

APAR: IY23278 COMPID: 5639I3700 REL: 310
ABSTRACT: SVT: SPLIT CONIFG IP ADDR CHG PE-SITE UPDATES TO OLD VALUES

PROBLEM DESCRIPTION:
The opt/dcelocal/etc/security/pe_site file is updated when
dced is restarted. This can cause problems when changing
the ip addresses of server machines. This can also cause
problems if there is a specific order in the pe_site file
for other reasons.

PROBLEM CONCLUSION:
The -n flag was added to dced to keep it from updating the
pe_site file when the daemon is restarted. Add this flag
to the dced line in the /opt/dcelocal/etc/cfgarg.dat file
if you do NOT want the pe_site file updated when dced is
started.

------

APAR: IY23279 COMPID: 5639I3700 REL: 310
ABSTRACT: INCORRECT FILE PERMISSIONS FOR PROTSEQS.RPC AND DCE_SMPL.RSP

PROBLEM DESCRIPTION:
Customer might have world writable protseqs.rpc file.

PROBLEM CONCLUSION:
Code has been modified to set proper permissions for
protseqs.rpc file on configuration.

------

APAR: IY23280 COMPID: 5639I3700 REL: 310
ABSTRACT: DCF_MESSAGE NOT IN GLOBAL STATEMENT IN GETREGISTRYINFO

PROBLEM DESCRIPTION:
Detailed messages about failures may not be displayed.

PROBLEM CONCLUSION:
Add DCF_MESSAGE to the global statements in each function.

------

APAR: IY23282 COMPID: 5639I3700 REL: 310
ABSTRACT: CAN'T RECONFIG PWDSTR AFTER LOCAL/ADMIN UNCONFIG

PROBLEM DESCRIPTION:
After running
unconfig.dce -config_type local pw_strength_srv
unconfig.dce -config_type admin -dce_host <dce_hostname>
        -host_id <ip_host_info> pw_strength_srv
an attempt to reconfigure this password strength server
will fail. The failure will indicate that the password
strength server principal is already in use.
The admin unconfig step did not unconfigure anything.
The program could not determine if there was a password
strength server on the remote machine. Once the password
strength server daemon is gone, the CDS Tower information
is no longer available. This is the only way to tie a
password strength server to a particular machine.

PROBLEM CONCLUSION:
The code was changed to take the password strength server
principal supplied on the command line and unconfigure that
password strength server for the remote machine. Care
should be used since the program can NOT verify that the
specified password strength server principal is in fact in
use on the remote machine.

------

APAR: IY23283 COMPID: 5639I3700 REL: 310
ABSTRACT: PDGQUERY GIVES NO OUTPUT ON AIX (SOLARIS OK)

PROBLEM DESCRIPTION:
The pdgquery command gives now output on AIX. Solaris works
fine.

PROBLEM CONCLUSION:
The nawk command handles differently on AIX and Solaris.
The nawk routine that pulls the message text from the txt
file was changed to not use "BEGIN" since it would not see
the message code that was being passed in.

------

APAR: IY23284 COMPID: 5639I3700 REL: 310
ABSTRACT: MERGE DCE_DCED_LISTENERS ENV VARIABLE FROM SOLARIS DCE 2.0

PROBLEM DESCRIPTION:
Some clients may deadlock using dced when there are not
enough listener threads to handle the amount of incoming
connections. This typically occurs on a heavily loaded
system or a very busy cell.

PROBLEM CONCLUSION:
Added code that was present in Solaris DCE 2.2 to allow
the customer to set the number of listener threads to some
number between 10 and 100 using the DCED_DCED_LISTENERS
ENVIRONMENT VARIABLE.

------

APAR: IY23353 COMPID: 5639I3700 REL: 310
ABSTRACT: "COULDN'T SET LOCALE CORRECTLY" MESSAGE

PROBLEM DESCRIPTION:
"couldn't set locale correctly" messages while running DCE
configuration commands.

PROBLEM CONCLUSION:
The LC_MESSAGES variable is saved and set eo en_US before
the "id" command is run. It was restored incorrectly. That
has been fixed.

------

APAR: IY23354 COMPID: 5639I3700 REL: 310
ABSTRACT: MORE ROBUST PE_SITE FILE HANDLING BY MIGRATE.DCE

PROBLEM DESCRIPTION:
When migrating a DCE Master Security server from a version
older than DCE 2.2 (AIX) or DCE 3.1 (Solaris), it may fail
becuase the migration code does not detect that the local
machine is the Master security server.

PROBLEM CONCLUSION:
The migration code is not handling all valid data formats
in the /opt/dcelocal/etc/security/pe_site file.

------

APAR: IY23356 COMPID: 5639I3700 REL: 310
ABSTRACT: SVT: LDAP, IPLANET WEB SERVER CANNOT START AFTER MKDCEWEB

PROBLEM DESCRIPTION:
Web server will not restart after mkdceweb is run on systems
that already have a LIBPATH defined in the netscape start
file.

PROBLEM CONCLUSION:
The code needs to be fixed so that it:
        doesn't update the file if /usr/lib is already in
                        the LIBPATH
        adds /usr/lib to an existing LIBPATH
        adds a new LIBPATH if one wasn't there
Currently, the code totally replaces a LIBPATH statement
with its own.
The unconfig code needs to be changed to not remove the
LIBPATH statement from the start file. The unconfig code
will no longer modify the start file at all. There
shouldn't be any problems with either leaving the LIBPATH
statement in the file or leaving /usr/lib in an existing
LIBPATH statement.

------

APAR: IY23360 COMPID: 5639I3700 REL: 310
ABSTRACT: CONFIG.DCE -NO_PESITE_UPDATE SHOULD SET DCED'S -N OPTION

PROBLEM DESCRIPTION:
dced will update the pe_site file even when the
-no_pesite_update flag was specified.

PROBLEM CONCLUSION:
When the -no_pesite_update flag is used, the -n flag will
be added to the dced line in /opt/dcelocal/etc/cfgarg.dat
file. If DCE is already configured, add the -n flag to the
dced line in /opt/dcelocal/etc/cfgarg.dat.

------

APAR: IY23361 COMPID: 5639I3700 REL: 310
ABSTRACT: DEADLOCK IN DCE LIBRARY

PROBLEM DESCRIPTION:
Server will hang and must be restarted.

PROBLEM CONCLUSION:
The condition wait routine reaquires the mutex, it is a
logic error for it to lock the mutex again. This can
cause the thread to deadlock waiting for itself to
release the lock.

------

APAR: IY23362 COMPID: 5765C3403 REL: 433
ABSTRACT: UNABLE TO SET MULTIPLE IF ATTRIBUTES WITH CHDEV

PROBLEM DESCRIPTION:
If one tries to set the ipaddress and an alias with one
chdev command, then only the alias is brought up. Mksysb/restore
has problems with this.

LOCAL FIX:
Use multiple chdev commands.

PROBLEM SUMMARY:
mksysb/restore doesnot bring up interface.

PROBLEM CONCLUSION:
If one restores a mksysb with an interface that has an alias
attribute, only the alias comes in the 'up' state.
Since chgif was never designed to call ifconfig multiple
times from one command line (unlike cfgif), I will change
chgif to be call ifconfig multiple times from one command
line.

TEMPORARY FIX:
reboot.

------

APAR: IY23373 COMPID: 5765C3403 REL: 430
ABSTRACT: CRASH IN VATTR_TO_NATTR

PROBLEM DESCRIPTION:
While running NFS version 2 with GPFS as an underlying
file system, NFS server crashes in vattr_to_nattr (called
from rfs_lookup).

LOCAL FIX:
Use NFS v3 instead of version 2.

PROBLEM CONCLUSION:
Make sure that the attribute field va_mask is set to AT_ALL
before passing arguement to VOP_LOOKUP.

------

APAR: IY23377 COMPID: 5765C3403 REL: 430
ABSTRACT: DBX THREAD-RELATED COMMANDS ARE SLOW

PROBLEM DESCRIPTION:
Thread-related commands' performance suffers as
the number of threads in debugged application
increases.

PROBLEM CONCLUSION:
Make more efficient use of the pthread debug library API,
especially the thread primitive iterators (e.g.
pthdb_pthread).

------

APAR: IY23422 COMPID: 5765C3403 REL: 430
ABSTRACT: IPV4 CLIENTS DO NOT WORK WITH IPV4/IPV6 SERVERS

PROBLEM DESCRIPTION:
The customer will not able to find the route to the
destination host. connect system call will fail with
can't find route to destination message.

PROBLEM CONCLUSION:
When connect system call fails for the first time
do a pure IPv4 lookup and do a connect again.

------

APAR: IY23439 COMPID: 5765C3403 REL: 433
ABSTRACT: SYSLOGD IS ROTATING FILES IN 4.3.3

PROBLEM DESCRIPTION:
Syslogd rotates files

PROBLEM SUMMARY:
syslogd rotating files in 4.3.3

------

APAR: IY23454 COMPID: 5765C3403 REL: 430
ABSTRACT: GXT4000P GREYED TEXT IN PULLDOWN NOT VISIBLE X APP.

PROBLEM DESCRIPTION:
Customer is using application LCABLE (by Mentor Graphics). On
GXT4000P, customer is seeing the following symptoms: - if a
new window opens the background color flashes between
  grey and turquoise. - the greyed text in pull-down menus is
not visible. The same problem was reported in 1999 for the
GXT2000P. this was solved with IY03590.

PROBLEM SUMMARY:
Customer is using application LCABLE (by Mentor Graphics). On
GXT4000P, customer is seeing the following symptoms: - if a
new window opens the background color flashes between
grey and turquoise. - the greyed text in pull-down menus is
not visible. The same problem was reported in 1999 for the
GXT2000P. this was solved with IY03590.

PROBLEM CONCLUSION:
Disable hardware stipple for all X stipples with height
greater than 16 pixels.

------

APAR: IY23490 COMPID: 5765C3403 REL: 430
ABSTRACT: PROBLEM RESTORING FILES USING MSMIT

PROBLEM DESCRIPTION:
Problem in restoring files using motif smit.

PROBLEM SUMMARY:
There is a problem in restoring files using
Motif SMIT.

------

APAR: IY23513 COMPID: 5765D9300 REL: 310
ABSTRACT: MPI_CART_CREAT() FAILS WITH BAD MEMORY READS USING ZEROFAULT.

PROBLEM DESCRIPTION:
The MPI function, MPI_Cart_create() fails with segmentation
faults and bad memory reads and other memory errors as shown
by the Zerofault tool. We have a recreatable test case
and Zerofault installed on node k35n07.ppd.pok.ibm.com.

PROBLEM SUMMARY:
Problem was solved so that data being passed in to the
helper function is correct, thus eliminating the bad
memory reads and resulting seg fault.

PROBLEM CONCLUSION:
One of the helper functions MPI_Cart_create routine calls
was passing data which was not correct. This resulted in
bad memory reads to appear in ZeroFault. Also it was
possible for a bad memory write to occur as a result, which
in turn could cause a seg fault.

------

APAR: IY23520 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSTEM CRASH IN IP_SETMOPTIONS()

PROBLEM DESCRIPTION:
System crash in ip_setmoptions while removing the gap
in the membership array.

PROBLEM SUMMARY:
System crash in ip_setmoptions() while removing the gap in
multicast group membership array.

PROBLEM CONCLUSION:
Code is modified to remove this gap before unlocking the
socket.

------

APAR: IY23575 COMPID: 5765C3403 REL: 430
ABSTRACT: SAR COMMAND OUTPUTS INCORRECT AVERAGE VALUES

PROBLEM DESCRIPTION:
The sar command outputs incorrect values for the average column
when three or more flags are used concurrently.

PROBLEM SUMMARY:
The fix will be available in the next full pubs release.or more
flags are used concurrently.

PROBLEM CONCLUSION:
The data structures which hold the data used in the average
calculations are now properly re-initialized after each pass,
resulting in correct average output.

------

APAR: IY23599 COMPID: 5765C3403 REL: 430
ABSTRACT: JFS ERRLOG FOR ASSERT IN PBITSET

PROBLEM DESCRIPTION:
Added error logging code to pbitset so that data will be
present after the system crash.

PROBLEM SUMMARY:
None -- the crash will still occur when there is a problem.
However, more data will be available for the developers to
help debug the problem.

------

APAR: IY23658 COMPID: 5765C3403 REL: 430
ABSTRACT: SHELL PROMPT GETS CORRUPTED IN EMACS MODE

PROBLEM DESCRIPTION:
The shell prompt gets corrupted in command line emacs
mode.

PROBLEM SUMMARY:
Prompt display gets corrupted when using command line emacs
mode.

PROBLEM CONCLUSION:
Suitable code changes are made so as to prevent the prompt
display from getting corrupted.

------

APAR: IY23674 COMPID: 5765C3403 REL: 430
ABSTRACT: SECURITY: POSSIBLE BUFFER OVERFLOW VULNERABILITY IN FTPD

PROBLEM DESCRIPTION:
There is a possible buffer overflow vulnerability in ftpd
which might elevate privileges of malicious user

PROBLEM CONCLUSION:
Reallocate memory, if it exceeds the allocated memory buffer

------

APAR: IY23676 COMPID: 5765C3403 REL: 430
ABSTRACT: ATM NETWORK SOCKETS FAIL TO CONNECT

PROBLEM DESCRIPTION:
with vci in disconnect mode and never go away.

PROBLEM CONCLUSION:
 extend the delay time between close/open to the adapter

------

APAR: IY23678 COMPID: 5765C3403 REL: 430
ABSTRACT: SYNC COMMANDS MAY BE REQUIRED

PROBLEM DESCRIPTION:
Driver to manage coherency domain between system and I/O space.

PROBLEM CONCLUSION:
sync work space before adapter acts on it.

------

APAR: IY23680 COMPID: 5765C3403 REL: 430
ABSTRACT: UNABLE TO ACTIVATE LARGE NUMBER OF VOLUME GROUPS

PROBLEM DESCRIPTION:
The customer will be unable to activate large numbers of
Volume Groups.

PROBLEM CONCLUSION:
ODM attribute "lg_term_dma" is changed to enable the user to
the ability to customize the size of the DMA memory pool
through the chdev command.

------

APAR: IY23683 COMPID: 5765C3403 REL: 430
ABSTRACT: DIAGS SHOW FRU NUMBER AS N/A FOR LOCATION CODE P1

PROBLEM DESCRIPTION:
Unable to retrieve the part number to replace.

PROBLEM CONCLUSION:
Whenever the Field Replacable Unit name is blank, try to use
the physical location code to find a part number. If the
physical location code is blank also, then no part number
can be obtained.

------

APAR: IY23685 COMPID: 5765C3403 REL: 430
ABSTRACT: FC - DLT TAPE DRIVE FAILS MKSYSB COMMAND.

PROBLEM DESCRIPTION:
When mutliple tar is created on NoRewind-on-close tape device,
subsequent tar command overwrites the previous tarfile.

PROBLEM CONCLUSION:
TestUnitReady is retried in case on UnitAttention, if
additional sense data show that Reset or "in transition to
ready" condition is present.

------

APAR: IY23693 COMPID: 5765C3403 REL: 430
ABSTRACT: MEMORY CORRUPTION WITH NOAC MOUNT OPTION AND FULL FILESYSTEM

PROBLEM DESCRIPTION:
If the "noac" mount option is used, and the filesystem being
written to gets full, then during particular circumstances
memory corruption might occur on the client, causing later
crashes and/or other problems.

LOCAL FIX:
Don't use the "noac" option, or make sure the filesystem
doesn't fill up.

PROBLEM CONCLUSION:
Check return value from writerp() call, if error break out of
the write function instead of continuing.

------

APAR: IY23696 COMPID: 5765D9300 REL: 310
ABSTRACT: USER SPECIFIED MPMD INCORRECTLY WHEN SPECIFYING A PROGRAM NAME

PROBLEM DESCRIPTION:
When a user incorrectly specified the MP_PGMMODEL to be MPMD
and a program name was specified then user received 0031-709
error. POE should either give a usage error or handle it
differently.

PROBLEM CONCLUSION:
poe will handle the case where user supplies a program name
and also specifies that the MP_PGMMODEL to be MPMD as
if it were SPMD. Therefore if a user enters
"poe hw -pgmmodel mpmd -procs 2" then poe will run the "hw"
program
as if spmd was specified. It will the run the "hw" program
twice.

------

APAR: IY23764 COMPID: 5765D2800 REL: 430
ABSTRACT: PORT SCAN CAUSES CLSTRMGR TO DIE - HACMP,HAES

PROBLEM DESCRIPTION:
The customer was using a port scanning program, nmap, to scan
all of the ports on the machine for security checking. The
particular scan being used caused the clstrmgr to die.

PROBLEM SUMMARY:
The customer was using a port scanning program, nmap, to scan
all of the ports on the machine for security checking. The
particular scan being used caused the clstrmgr to die.

PROBLEM CONCLUSION:
The clstrmgr was dying because the accept command was
failing. This command is normally accepting a connection
request from either clsmuxpd or cllockd and the command is
not expected to fail unless there is something wrong in the
system, so was being indicated as a system error. In the
case here of a port scan, using nmap, a packet containing
an RST (reset) bit is sent which clears the msg queue and
causes the accept to fail because of the empty queue.
In looking at this code, it seemed that aborting the
clstrmgr seemed too drastic for this case, as well
as any other reason for the accept to fail, so
the logic was changed in utils.c to just issue the
message indicating the accept failure and just
continue.

------

APAR: IY23767 COMPID: 5765E6110 REL: 220
ABSTRACT: REQUIRED MAINTENANCE UPDATE

PROBLEM DESCRIPTION:
required maintenance update

------

APAR: IY23768 COMPID: 5765E6110 REL: 110
ABSTRACT: REQUIRED MAINTENANCE UPDATE

PROBLEM DESCRIPTION:
required maintenance update

PROBLEM SUMMARY:
REQUIRED MAINTENANCE UPDATE

PROBLEM CONCLUSION:
REQUIRED MAINTENANCE UPDATE

------

APAR: IY23790 COMPID: 5765C3403 REL: 430
ABSTRACT: GETLVODM COMMANDS CAN HANG

PROBLEM DESCRIPTION:
Cust. may experience slowdown or hang of LVM commands.

PROBLEM CONCLUSION:
Removed unnecessary locking to reduce lock contention.

------

APAR: IY23798 COMPID: 576565400 REL: 220
ABSTRACT: SVMON DOESN'T DISPLAY INODE NUMBERS FOR LARGE FILES

PROBLEM DESCRIPTION:
svmon with -S or -P option doesn't display the file
inode number in the Description column when the file
is in Large File Enabled Filesystem.

PROBLEM CONCLUSION:
Modified the code such that when the descrption doesn't fit in
20 characters, then it is wrapped and displayed on next line.

------

APAR: IY23804 COMPID: 576565400 REL: 220
ABSTRACT: SPMI MEMORY LEAK PROBLEMS

PROBLEM DESCRIPTION:
Improper exit of any Suppliers leads to Shared memory
corruption and some times Segmentation Fault occurs.

PROBLEM CONCLUSION:
The code is fixed by adding separate code to determine
any Shared Memory Area corruption and correct it. And also
added code to prevent shared memory corruption due to
abrupt stop of PTX suppliers.

------

APAR: IY23845 COMPID: 569692600 REL: 110
ABSTRACT: SIMPLE_LOCK IN X25_READ AND XDH_DO_NOTIFY_READ_AVAILABLE OF

PROBLEM DESCRIPTION:
The system will hang when all the customer X.25 applications
are funneled on CPU 0 or the system is a UP system.

PROBLEM SUMMARY:
The system can hang when running COMIO X.25 applications.

PROBLEM CONCLUSION:
Change the simple_lock to disable_lock in x25_read and
xdh_do_notify_read_available.

------

APAR: IY23854 COMPID: 5765D2800 REL: 431
ABSTRACT: NODE_UP_REMOTE NEEDS TO STOP APPS IN REVERSE ORDER - HAS,HAES

PROBLEM DESCRIPTION:
A customer noticed that during a node_up_remote event on a node
which had resources for the remote node coming up, that it
stopped the applications in listed order, rather than in reverse
order. This caused a 'dependency' problem in the applications
thatdepended on the first listed applications to be up to stop
properly. However, with stopping the first listed applications,
the dependent applications failed to halt properly.

PROBLEM SUMMARY:
Customer noticed that application resources taken over by one
node, stopped the applications in listed order, when the
failed node returned to the cluster. This caused a
'dependency' problem with applications that depended on the
first listed applications to be active, so they they could
halt properly.
Meaning, applications started later in the application list
were dependent on the applications started sooner in the
application list to start, run, and halt properly.

PROBLEM CONCLUSION:
Code in HACMP/HAES 'node_up_remote.sh' event script changed
such that the applications to stop are listed and stopped in
reverse order.

------

APAR: IY23863 COMPID: 5765C3403 REL: 430
ABSTRACT: CHFS -A SPLITCOPY GIVES COPY INCORRECT LABEL

PROBLEM DESCRIPTION:
If the customer creates an online mirror backup with the
chfs -a splitcopy command, then exports and re-imports
the volume group, the backup has the wrong mount point
and cannot be umounted or removed (in fact, it has the
same mount point as the original).

PROBLEM SUMMARY:
See above ERROR DESCRIPTION.

PROBLEM CONCLUSION:
The splitcopy portion of chfs has been changed so
that the label field of the lvcb of an online
mirror backup copy stores the correct mount point
information.

TEMPORARY FIX:
Change the mount point information of the backup
in /etc/filesystems after the re-import to its
correct mount point.

------

APAR: IY23889 COMPID: 5765D6100 REL: 220
ABSTRACT: ON SWITCH TABLE ERROR, UNLOAD ST ONLY WHEN

PROBLEM DESCRIPTION:
When LoadL determines a load or Unload of a switch
table fails, the UNLOAD_ST_LL ioctl, unloading of the switch
table is always called. Want to change this to only be called
when the DRAIN_ON_SWITCH_TABLE_ERROR = false.

PROBLEM SUMMARY:
LoadLeveler needs to give the option of doing
a force unload switch table window if it determines
that it failed to unload the window.

PROBLEM CONCLUSION:
LoadLeveler now has the option of doing
a force unload switch table window if it determines
that an unload of the window had failed.
The option is if the DRAIN_ON_SWITCH_TABLE_ERROR = false,
then a force of unloading the switch table
will be effective.

------

APAR: IY23902 COMPID: 5765E6100 REL: 110
ABSTRACT: REQUIRED SERVICERM UPDATE

PROBLEM DESCRIPTION:
Required ServiceRM update.

PROBLEM SUMMARY:
REQUIRED SERVICERM UPDATE

PROBLEM CONCLUSION:
REQUIRED SERVICERM UPDATE

------

APAR: IY23906 COMPID: 5765C3403 REL: 430
ABSTRACT: AFTER STRIP -T DBX DOES NOT SHOW THE LINE NUMBER

PROBLEM DESCRIPTION:
Dbx doesnot show the line number information after strip
with "-t" option on a given object file.

PROBLEM SUMMARY:
Using dbx to debug an object file that has been
stripped with the "-t" option will cause dbx to
not be able to resolve line number information
properly.

PROBLEM CONCLUSION:
When using the "-t" option with the strip
command, we will no longer remove the ".bf"
and ".ef" symbols from the object file. This
will allow dbx to properly resolve line number
information.

------

APAR: IY23915 COMPID: 5639I3700 REL: 310
ABSTRACT: ULIMIT SUPPORT IN CONFIG COMMANDS

PROBLEM DESCRIPTION:
User commands can be run during the configuration commands,
by using the user_cmd.tcl script.
Unfortunately, there is no way to have an external command
run in the same process as the
configuration/start/stop/unconfiguration commands.
The ulimit command can not set the process limits for the
DCE daemons from within the user_cmd.tcl script.

PROBLEM CONCLUSION:
A command was added to dcecp called dcf_ulimit.
dcf_ulimit acts much like ulimit. It will take multiple
command line parameters. If a -?, -help, or help are
entered on the dcf_ulimit command, the following will be
displayed:
        dcf_ulimit -S|-H -f -c -t -d -s -m -n
The -a paremeter is not supported. If the user wants to
view the current ulimit settings, he can use the
"ulimit -a" command from within dcecp (or a tcl script).
The ulimit command in the operating system takes 64-bit
values. Because dcecp is a 32-bit program, it is restricted
by the size of the integer it must pass to the setrlimit
api, which is 2147483647. Also, although the ulimit command
takes the values in blocks and kbytes, the setrlimit api
only takes them in bytes. Because of this, we have to
multiply the value input by the user by 512 or 1024,
depending on the ulimit field being updated. Since I still
have the integer limit, the 2.1 billion limit is reduced.
Here are the maximum limits that dcf_ulimit will allow:
        -f FSIZE 4194303
        -c CORE 4194303
        -t CPU 2147483647
        -d DATA 2097151
        -s STACK 2097151
        -m RSS 2097151
        -n NOFILE 2147483647
The keyword "unlimited" is also allowed as a value for any
of the parameters.
The dcf_ulimit command should act the same way as the ulimit
command. If run by the root user, if the soft limit is set
higher than the max, the max will be reset. If the max is
set lower than the soft limit, the soft limit will be reset.
If the command is not being run by the root user, an error
message will be issued if the soft limit is set higher than
the max.

------

APAR: IY23916 COMPID: 5639I3700 REL: 310
ABSTRACT: DCED CORE DUMPS WHEN DCE IS STOPPED

PROBLEM DESCRIPTION:
1) configure a one machine cell
(rpc, sec_cl, sec_srv, cds_cl, cds_srv)
2) run stop.dce sec_srv
3) run stop.dce
At this point, dced core dumps.
There are two different cores observed
One is (lines are broken to fit)
Segmentation fault in longjmp._longjmp
           /usr/lib/libdcepthreads.a at 0xd12ab334
 ($t21)
0xd12ab334 (_longjmp+0xc) 90e10000 stw r7,0x0(r1)
(dbx) where
longjmp._longjmp() at 0xd12ab334
(dbx)
The other is (lines are broken to fit)
% dbx dced core
 using memory image in core
Segmentation fault in rpc__cn_sm_eval_event at 0xd108ed10
0xd108ed10 (rpc__cn_sm_eval_event+0x20) 93df000c stw
r30,0xc(r31)
(dbx) where
rpc__cn_sm_eval_event(??, ??, ??, ??) at 0xd108ed10
rpc__cn_call_transmit_fault(??, ??, ??) at 0xd1087a94
rpc_call_transmit_fault(??, ??, ??) at 0xd1084ac4
rpc_ss_send_server_exception_2(??, ??, ??, ??, ??) at
                          0xd10bbd9c
rpc_ssr_sstub(0x2029e280, 0x0, 0x2029e510, 0xf0131014,
                          0xd09ab968,
0x20204d90, 0x2029e500, 0x42000429) at 0xd10ba974
rpc__cn_call_executor(??, ??) at 0xd12516bc

PROBLEM CONCLUSION:
In src/admin/dced/server/oeops.c file in the function
old_ept_map function TRA/CATCH macro was introduced.
But in between the macros there is a return statement which
should not be done. That is the reason the thread context
is corrupted in the longjmp.

------

APAR: IY23917 COMPID: 5639I3700 REL: 310
ABSTRACT: DCE SERVERS DROP CORE IN TCP-RPC CODE

PROBLEM DESCRIPTION:
Bogus auth_level values in incoming packets can cause DCE
servers to drop core.
Two sample stacks are: (lines are broken to fit)
rpc__auth_inq_supported(??) at 0xd039f31c
rpc__cn_assoc_process_auth_tlr(??, ??, ??, ??, ??, ??,
                               ??, ??) at 0xd052ff94
do_assoc_req_action_rtn(??, ??, ??) at 0xd052de38
do_assoc_action_rtn(??, ??, ??) at 0xd052b898
rpc__cn_sm_eval_event(??, ??, ??, ??) at 0xd0378c40
receive_dispatch(??) at 0xd0532e58
rpc__cn_network_receiver(??) at 0xd0533b28
The other stack is
rpc__auth_rpc_prot_epv(authn_prot_id = 3735936685U,
                       rpc_prot_id = 0), at 0 xef2d99cc
rpc__cn_unpack_hdr(pkt_p = 0x115330), at 0xef3228c4
receive_dispatch(assoc = 0x119a28), line 1117 in "cnrcvr.c"
rpc__cn_network_receiver(assoc = 0x119a28),
                    line 570 in "cnrcvr.c"
dce_pthread_start(arg = 0x751b8),

PROBLEM CONCLUSION:
The packet data is now checked more carefully.

------

APAR: IY23928 COMPID: 5765C3403 REL: 430
ABSTRACT: LPSTAT -W -P <QNAME> SHOWS EXTRANEOUS OUTPUT FROM OTHER QUEUES

PROBLEM DESCRIPTION:
lpstat -W -p<qname> shows extraneous output from
other queues

LOCAL FIX:
omit the -W flag on lpstat, or use one of the
other queue status commands such as qchk.

PROBLEM SUMMARY:
lpstat -W -p <qname> shows extraneous output from other queues

PROBLEM CONCLUSION:
lpstat has been fixed so as to not show extraneous output

------

APAR: IY23931 COMPID: 5765C3403 REL: 430
ABSTRACT: PANIC FROM KRPC_DISPATCH()

PROBLEM DESCRIPTION:
Fix given: The socket was getting corrupted because of two
processes acting on it at a time.

PROBLEM SUMMARY:
System crashes while running nfs.

PROBLEM CONCLUSION:
Checking lock for the member socket before modifing the
socket structure contents.

------

APAR: IY24007 COMPID: 5765C3403 REL: 430
ABSTRACT: SETLOCALE UNSAFE IN MULTI-THREADED APPLICATIONS.

PROBLEM DESCRIPTION:
Possible core dumps may occur when using setlocale and other
APIs which access locale data in a multi-threaded environment.

PROBLEM SUMMARY:
possible coredump when running setlocale in
a multi-threaded environment.

PROBLEM CONCLUSION:
Suspended access to locale data while in setlocale.

------

APAR: IY24017 COMPID: 5765C3403 REL: 430
ABSTRACT: PORTMAP HANGS IN T_RCVDATA()

PROBLEM DESCRIPTION:
During rare circumstances, portmap might hang in a call to
t_rcvudata, causing it to go unresponsive.

LOCAL FIX:
Restart portmap (with the -w flag for warm start).

PROBLEM CONCLUSION:
Add O_NONBLOCK to the t_open calls, to make the reads
nonblocking.

------

APAR: IY24021 COMPID: 5765C3403 REL: 430
ABSTRACT: DSI IN CACHEFS WHILE POPULATING CACHE

PROBLEM DESCRIPTION:
A DSI might occur in cachefs if for some
reason the server file changes permission in the middle
of when the cache is being populated with that file, i.e.
when the file is being copied.

PROBLEM CONCLUSION:
Register an exception handler before touching the created
segment backing to the server-side file. This way, if we can't
page in the data we won't crash, instead we can abort
gracefully.

------

APAR: IY24033 COMPID: 5765C3403 REL: 430
ABSTRACT: GXT2000P/3000P OPENGL MIPMAPPED TEXTURE RENDERING FAILURE

PROBLEM DESCRIPTION:
OpenGL application does not render correctly on GXT2000P and
GXT3000P when using a 2048X1024 mipmapped texrure. No texture
at all is rendered.

PROBLEM SUMMARY:
OpenGL application does not render correctly on GXT2000P and
GXT3000P when using a 2048X1024 mipmapped texrure. No texture
at all is rendered.

PROBLEM CONCLUSION:
Correct textrure resource calculation in GL_PROXY_TEXTURE_2D
path.

------

APAR: IY24044 COMPID: 5639I3400 REL: 310
ABSTRACT: GDAD FAILS WHEN ANY FOREIGN CDS SERVERS ARE DOWN

PROBLEM DESCRIPTION:
When attempting to initiate cross-cell communication via
gdad, the attempt fails if ANY of the foreign cds servers
is down. "Down" in this case means that a cds server does not
respond on any of the interfaces for which it is
registered.
The error given is "software error detected in server".

LOCAL FIX:
Check all foreign cds servers and make sure they are up and
listening on all their registered interfaces. The other
workaround is to use the "cdscp define cached server" command
or the "dcecp cdscache create" command to indicate the location
of foreign cds servers.

PROBLEM SUMMARY:
When any replica is down the cross cell communication
doesnt work DNS recognizes text hiercal cell structures
when alias is used to make an envionrment similar to
hierichal cell

PROBLEM CONCLUSION:
Within bind_process() function, while going though our
initial list of cds servers (obtained from the dns recs),
we make sure they're actually up and running
(sockaddr_to_tower calls rpc_mgmt_is_server_listening).
The bug is that we completely break out of that loop if
we hit a downed server and incorrectly indicate that we
found *no* working cds servers.

------

APAR: IY24059 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY24081 COMPID: 5765C3403 REL: 430
ABSTRACT: SU DOESN'T INCREMENT UNSUCCESSFUL LOGIN COUNT ON FAILURE

PROBLEM DESCRIPTION:
SU does not increment the unsuccessful login
count value in the /etc/security/lastlog of the
user su'd on failure. login, telnet does the
same when failure occurs because of incorrect
password. SU needs to be enhanced to take this
into account.

PROBLEM SUMMARY:
SU does not increment the unsuccessful login
count value in the /etc/security/lastlog of the
user su'd on failure. login, telnet does the
same when failure occurs because of incorrect
password. SU needs to be enhanced to take this
into account.

PROBLEM CONCLUSION:
SU now increments the unsuccessful login count and resets
on sucessful login.

------

APAR: IY24086 COMPID: 5765B9501 REL: 320
ABSTRACT: ASSERT SUBROUTINE FAILED: WASASSIGNED BUFMGR.C, LINE 5285

PROBLEM DESCRIPTION:
unassignBuffer has to wait if the buffer is being unpinned by
some other thread.

PROBLEM SUMMARY:
GPFS self check logic asserted at bufmgr.C line 5285

PROBLEM CONCLUSION:
unassignBuffer has to wait if the buffer is being unpinned
by some other thread.

------

APAR: IY24093 COMPID: 5765C3403 REL: 430
ABSTRACT: SYNCVG ERROR WITH BAD BLOCK RELOCATION IN CONCURRENT MODE

PROBLEM DESCRIPTION:
Syncvg does not work properly with bad block relocation in
concurrent mode. Even with bad block relocation disabled
(according to lsvg), LVM was putting entries into the bad
block table on disks with bad blocks. When we tried writing
to these blocks, we returned EMEDIA errors, but didn't log any
errors because the bad block table indicated that the blocks
couldn't be written to, but also indicated that they hadn't
been relocated.

PROBLEM SUMMARY:
syncvg error with bad block relocation in concurrent VG

PROBLEM CONCLUSION:
LVM no longer writes to BB table in concurrent mode,
as per design.

------

APAR: IY24094 COMPID: 5765C3403 REL: 430
ABSTRACT: REDEFINEVG ERROR WITH CONCURRENT VGS

PROBLEM DESCRIPTION:
When redefinevg is run on a concurrent-capable volume group,
it turns it into a non-concurrent-capable volume group. This
is because there is no "Concurrent Capable" flag given
to putlvodm. Therefore, when the volume group is attempted to
be varied on in concurrent mode, an error message will be
returned.

PROBLEM SUMMARY:
redefinevg changes concurrent VGs to non-concurrent

PROBLEM CONCLUSION:
Add appropriate flags to putlvodm command

------

APAR: IY24120 COMPID: 5765D5100 REL: 320
ABSTRACT: CSS_COLONY:ADAPTRECV UPDATE SNAP ON PP ERRORS WITH -D FOR SRAM

PROBLEM DESCRIPTION:
css_colony:adaptrecv update snap on pp errors with -d for sram

PROBLEM SUMMARY:
Added new flag on invocation of snap to enable port
permanent erors to collect snap.

PROBLEM CONCLUSION:
Snap does not contain sram dump on port permanent errors
needed for debug.

------

APAR: IY24130 COMPID: 5639I3700 REL: 310
ABSTRACT: GDAD CORE DUMP IN L_DELETE()

PROBLEM DESCRIPTION:
gdad can core dump with the following stack:
dce_pthread_mutex_lock() at 0xef584624
write_lock() at 0x374cc
l_delete() at 0x2dbb0
process() at 0x30194
get_rpc_answer() at 0x31abc
cds_ReadAttribute() at 0x339b0
op21_ssr() at 0x42ad4
rpc__cn_call_executor() at 0xef515860
cthread_call_executor() at 0xef52b578
dce_pthread_start() at 0xef588448

PROBLEM CONCLUSION:
Within l_delete():
 ...
        else /* el->prev_el != NULL */
        {
            prev = el->prev_el;
            read_unlock(&el->rwl_link);
            /*
             * this write lock could fail because prev was
               just deleted
             */
            if((status = write_lock(&(prev2 =
               el->prev_el)->rwl_link)) != OK)
            {
 ...
The core is a result of el->prev_el being changed to NULL
when we drop the read lock on el. prev points to a valid
structure when we come into this else clause and that's
what we want the write lock on. prev2 is used to determine
if el->prev_el has changed in the short amount of time when
we drop the read lock and go for the write lock. (We check
further down to see if prev and prev2 are the same and if
not, an element was deleted from the list out from under us
when the lock was dropped.)
The fix is to simply set prev2 after the read_unlock and
call write_lock with prev (which we know is non-NULL):
            prev2 = el->prev_el;
            if((status = write_lock(&prev->rwl_link)) != OK)
This way we aren't running the risk of dereferencing a nil
pointer.

------

APAR: IY24133 COMPID: 5639I3700 REL: 310
ABSTRACT: PORTING DCEDEV DEFECTS TO DCE310 TO SUPPORT NS-4.X IN DCE310

PROBLEM DESCRIPTION:
When he runs mkdceweb for iplanet servers 4.1, the
configuration will not take place.

PROBLEM CONCLUSION:
Many source files were modified to make them compatible with
the new version of iplanet web servers

------

APAR: IY24134 COMPID: 5639I3700 REL: 310
ABSTRACT: NEED 2 VERSIONS OF LIBDCEWEB.A FOR NETSCAPE SERVER 3.X AND 4.X

PROBLEM DESCRIPTION:
We won't be able to ship libdceweb4.a to the customers
which is required for the support of IPlanet 4.X
webservers.

PROBLEM CONCLUSION:
Packaging code is added to ship libdceweb4.a in a ptf.

------

APAR: IY24143 COMPID: 5639I3700 REL: 310
ABSTRACT: ERROR IN FILES.TCL CAUSES PROBLEMS WITH -AUTOSTART

PROBLEM DESCRIPTION:
An error exists in /opt/dce/tcl/dcedcf/files.tcl that
will cause failures on startup if both the -autostart
and -clean_autostart flags are turned on simultaneously.
The code specifies that the links created in /etc/rc3.d
should be called S15-10dce-clean and S15-20dce; this
is to ensure that the clean function is carried out
before any attempt is made to start DCE.
However, a typo in files.tcl has left both files
with the same sequence number, e.g.:
S15-20dce
S15-20dce-clean
This causes both to be executed simultaneously, which
means that in most cases one will fail. Often this
results in DCE not starting properly at boot time.

PROBLEM CONCLUSION:
Correct files.tcl so that the
variable "R3_CLEANUP_DCE S15-20dce-clean" is changed
to "S15-10dce-clean" as specified elsewhere.

------

APAR: IY24144 COMPID: 5639I3700 REL: 310
ABSTRACT: IMPLICIT CHMOD IN '-AUTOSTART NO' CHANGES PERMS ON WRONG FILE

PROBLEM DESCRIPTION:
autostart -no leaves permissions incorrect on rc.dce

PROBLEM CONCLUSION:
Problem is that /opt/dce/tcl/dcedcf/cfg_utils.tcl
uses a routine called 'remove' to take links out
of /etc/rc3.d. The 2nd parameter in this routine
toggles FORCE y/n; if FORCE is y then we first do
a chmod rw-rw-rw- on the file to be removed. But
as we're working on a symlink, this changes the
permissions on the actual rc.dce instead.
The fix is either to change FORCE to FALSE or make
a change to the C++ code in EraseFile() to do better
checking before chmod'ing the file.

------

APAR: IY24162 COMPID: 5765C3403 REL: 430
ABSTRACT: SIGWAIT() HIGH-ORDER MASK BIT PROBLEM

PROBLEM DESCRIPTION:
An extra bit in the sigmask exposes a array overflow
when used with sigwait call. This causes the application
to segv. But since the signal is blocked, the application
just hangs.

PROBLEM CONCLUSION:
Add a check in sigwait to check this overflow and correct
it.

------

APAR: IY24165 COMPID: 5765B9501 REL: 320
ABSTRACT: KERNEL PANIC WHEN V_AUDIT FIELD HAD INVALID VALUE

PROBLEM DESCRIPTION:
KERNEL PANIC WHEN V_AUDIT FIELD HAD INVALID VALUE

PROBLEM SUMMARY:
Kernel panic in a system running AIX auditing.

PROBLEM CONCLUSION:
Vnode caching was using the v_audit field field of the vnode
as an aging value. If AIX auditing is turned on, the
counter is used as a pointer. Change to newer implementation
using counter in MMFSnode.

------

APAR: IY24174 COMPID: 5765D9300 REL: 310
ABSTRACT: MPI STRUCT TYPE NESTING PROBLEM.

PROBLEM DESCRIPTION:
The problem arises when the customer builds a type consisting
of one MPI struct type nested inside another. The type
decoder functions correctly decode the outer struct
type, but the inner, nested struct is not decoded
correctly. There is a test case that shows the problem
in /u/dprice/pmrs/pmr<thispmr> on any "pops" server such
as c676n01.ppd.pok.ibm.com.

PROBLEM SUMMARY:
Problem was solved so that nested datatypes are properly
decoded. Their internal data is filled in with the
correct information

PROBLEM CONCLUSION:
MPI was not decoding nested datatypes properly because the
internal data for those was not set. A nested datatype
would be a structure containing another structure as an
element.

------

APAR: IY24203 COMPID: 5765D5101 REL: 121
ABSTRACT: TOPSVCS NEEDS LONGER TIMEOUT VALUE WHEN NIM FAILS TO RESPOND TO

PROBLEM DESCRIPTION:
A DMS timeout occurred because the hatsd daemon became blocked
making several attempts to start and shutdown the NIM for inter-
face en0. This RAS APAR is being opened to increase the timeout
valuethat the hatsd waits for a response to an "Open Nim cmd"
before shutting down the NIM and restarting it.

PROBLEM SUMMARY:
The RSCT Topology Services subsystem comprises a
main daemon and a set of Network Interface Module
(NIM) processes. While the daemon is responsible
for running the distributed protocols, the NIM
process is used to communicate with each local
adapter being monitored.
At daemon initialization time, or when the adapter
is not configured at the expected address, the
daemon sends a command asking the NIM to start
monitoring the adapter. A response to the command
is expected from the NIM. The absence of such
response after a predetermined timeout value
is interpreted as a NIM misbehavior
and results in the daemon respawning a new NIM.
Under severe load conditions, the NIM may block
long enough that its response to the daemon's command
may be delayed past the daemon's threshold.
The actions involved in respawning a new NIM may
cause the daemon itself to be blocked, which could
result in the node being flagged as down. In an
HACMP/ES environment, this may provoke a node crash.

PROBLEM CONCLUSION:
The threshold used by the RSCT Topology Services
daemon to interpret the NIM process as
blocked/failed has been extended to account
for load-related blockage. With the fix, the
daemon should no longer respawn a new NIM when
the latter is blocked temporarily.

------

APAR: IY24205 COMPID: 5765D2800 REL: 431
ABSTRACT: RPC TIMEOUT MSGS ON CONSOLE AFTER NODE BOOT - HACMP,HAES

PROBLEM DESCRIPTION:
When the customer rebooted a node of an HA cluster which was
configured to export filesystems, rpc timeout and portmapper
error messages were received on the console.

LOCAL FIX:
Comment the start of rpc.statd from harc.net.

PROBLEM CONCLUSION:
These messages are evidently due to the rpc.statd finding
lock information and attempting to send out lock reclaim
requests but cannot reach the NFS server address, which is
the HA service address, because we are still on the boot
address at this time. Original implementation expected
that these lock reclaim requests would not be sent out at
this time, if the rpc.lockd was not running, but that is
evidently not the case. Since there is no apparent reason
to start the rpc.statd from harc.net, this line was
commented the same as the other NFS daemons.

------

APAR: IY24212 COMPID: 5765B8100 REL: 220
ABSTRACT: PROVIDE A NEW SERVICE UTILITY FUNCTION FOR REL 220

PROBLEM DESCRIPTION:
Provide extra function required for migration to future release

PROBLEM SUMMARY:
Provide extra function required for migration
to future release

PROBLEM CONCLUSION:
Function provided

------

APAR: IY24218 COMPID: 5765B9501 REL: 320
ABSTRACT: LX: GPFS ASSERT AFTER MMCRSNAPSHOT, METADATA-VFS.C:1074

PROBLEM DESCRIPTION:
LX: GPFS assert after mmcrsnapshot, metadata-vfs.C:1074

PROBLEM SUMMARY:
Code error found while reviewing code investigating an
unrelated problem.

PROBLEM CONCLUSION:
syncFS: when called during unmount must flush all files
where fileDirtyOrUncommitted() is true, not just files with
fileDirty() true.

------

APAR: IY24260 COMPID: 5765C3403 REL: 430
ABSTRACT: SHOWMOUNT: REMOVE 2048 ENTRY LIMIT

PROBLEM DESCRIPTION:
The showmount command could not display more than
2048 entries previously. This restriction has been
removed.

PROBLEM SUMMARY:
The showmount command could not display more than
2048 entries previously. This restriction has been
removed.

PROBLEM CONCLUSION:
The list is allocated dynamically to remove the 2048 entry
restriction.

------

APAR: IY24263 COMPID: 5639I3700 REL: 310
ABSTRACT: /OPT/DCE/PDG PERMISSIONS INCORRECT

PROBLEM DESCRIPTION:
The pdgquery command will fail when called by a non-root
user because it can not read the files in
/opt/dcelocal/pdg.

PROBLEM CONCLUSION:
Change the permissions on the files in /opt/dcelocal/pdg
to be 755.

------

APAR: IY24264 COMPID: 5639I3400 REL: 310
ABSTRACT: CELL PING -R REPORTS INCORRECT RESULT

PROBLEM DESCRIPTION:
In previous releases the command 'cell ping -replicas' returned
the success message "DCE servers available." In 3.1 some
message-handling code was altered and an incorrect message code
was pasted into this handler; it now reports "DCE clients
available" instead. This is due to an incorrectly-pasted error
code; whereas 'cell ping -clients' should return the message
noted above (e.g. 0x170e1e52), -replicas should use 0x170e1e53
instead.

PROBLEM SUMMARY:
In DCE 3.1 cell ping -replicas reports
'DCE clients available.' According to the documentation
it should report 'DCE servers available.'

PROBLEM CONCLUSION:
the messages in cell.dcp were changed to return
the correct error code.

------

APAR: IY24270 COMPID: 5765D2800 REL: 431
ABSTRACT: CSPOC:ADDING USER ID'S GREATER THAN 250000 REJECTED - HACMP,HAES

PROBLEM DESCRIPTION:
The customer was trying to add user id's of 700000 and above
through C-SPOC since these id's would match with their employee
numbers. This was rejected with message indicating that user
id's above 250000 are not supported.

LOCAL FIX:
Edit the cl_mkuser script to change the MAX_ID

PROBLEM SUMMARY:
The customer was trying to add user id's of 700000 and above
through C-SPOC since these id's would match with their employee
numbers. This was rejected with message indicating that user
id's above 250000 are not supported.

PROBLEM CONCLUSION:
Since the limit enforced appears to be an arbitrary limit
which does not seem to serve any good purpose, the limit
was changed to 2147483647 which is the largest value that
will not show as negative in lsuser output. The man page
for cl_mkuser was also updated to this value.

------

APAR: IY24305 COMPID: 5765C3403 REL: 430
ABSTRACT: XLIB CT CONVERTER PROBLEMS WHEN DIFFERENT LOCALES ARE USED

PROBLEM DESCRIPTION:
When the window manager is running in a locale different
than the applications, the Compound Text converter may have
trouble converting data. This can cause the window manager
to hang or crash in some instances.

PROBLEM CONCLUSION:
Fix the Compound Text converter to provide better error
recovery.

------

APAR: IY24310 COMPID: 569692600 REL: 110
ABSTRACT: AIXLINK/X.25 COMIO DRIVER FREES THE MBUF POINTER WHICH IS PASSE

PROBLEM DESCRIPTION:
When the customer runs QLLC over X.25 COMIO interface,
the system will crash.

PROBLEM CONCLUSION:
X.25 COMIO driver frees the mbuf pointer which passed in
from QLLC. The solution is not to free it.

------

APAR: IY24333 COMPID: 5765D9300 REL: 310
ABSTRACT: INTERMITTENT INTERNAL READ FAILED ERROR

PROBLEM DESCRIPTION:
Intermittent Internal read failed error

PROBLEM SUMMARY:
MPI-IO responder maintains file related info in a table.
After a file closed, the corresponding entry in the table
will be removed. Current responder code locates the entry to
be removed by checking file descriptor saved in each entry
of the table. However, there could be race condition for
file open and close if an MPI-IO responder handles MPI-IO
open and close command concurrently. File descriptor used by
the closed one can be assigned to the newly opened file if
the open happens after the close. That means sometimes the
table contains two entries having the same file descriptor,
and the first found one can be removed incorrectly.

PROBLEM CONCLUSION:
To avoid the problem, code changed so that the entry to be
removed is located by the entry address instead of the file
descriptor.

------

APAR: IY24338 COMPID: 5765C3403 REL: 430
ABSTRACT: COMPILE FAILURE WITH #DEFINE _XOPEN_SOURCE 500

PROBLEM DESCRIPTION:
C++ compile errors when _XOPEN_SOURCE symbol is
equal to 500 due to declaration errors in
/usr/include/wchar.h.

PROBLEM SUMMARY:
Compile errors in wchar.h with #define _XOPEN_SOURCE 500

PROBLEM CONCLUSION:
Add guard checking for _VA_LIST; use __va_list in declarations
of vfwprintf, vwprintf, and vswprintf if _VA_LIST is not
defined.

------

APAR: IY24378 COMPID: 5765C3403 REL: 430
ABSTRACT: LIBXT.A - SOME TRANSLATIONS HAVE STOPPED WORKING

PROBLEM DESCRIPTION:
Some translations have stopped working.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* X11.base.lib 4.3.3.51
* X11.base.lib 4.3.3.52
* X11.base.lib 4.3.3.75
* X11.base.lib 4.3.3.76
***************************************************************
*PROBLEM DESCRIPTION: *
* Some Translations in an Xt/Motif application may not
* work.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY24378
***************************************************************

PROBLEM CONCLUSION:
Fix a minor error in the translation manager code.

------

APAR: IY24387 COMPID: 5765C3403 REL: 433
ABSTRACT: TOOLTALK VULNERABILITY DEFECT 346271

PROBLEM DESCRIPTION:
This is to address the Security Vulnerability
in ToolTalk Server which was taken care in the
defect 346271.

PROBLEM SUMMARY:
Normal user can get privilaged access by exploiting some
Format String Vulnerabilities present in ttdbserverd.

PROBLEM CONCLUSION:
Changed the code to minimise Format String Vulnerability.

------

APAR: IY24415 COMPID: 5765C3403 REL: 430
ABSTRACT: ATM622 DROPS MILLIONS OF CELLS

PROBLEM DESCRIPTION:
We are seeing millions upon millions of dropped ATM cells
on our systems running with atm622 adapters displayed by
"atmstat"

PROBLEM CONCLUSION:
The cell drop doesn't seem to effect any
Tx or Rx from the adapter. This have been there from quit
some time. The Cell drop is a counter keep by the adapter
hardware
This problem is fixed in the driver code.

------

APAR: IY24431 COMPID: 5765C3403 REL: 430
ABSTRACT: ASSERT IN XMEMDMA64_PPC

PROBLEM DESCRIPTION:
There is an assert in xmemdma64_ppc which is supposed
to verify that the segment boundary we are crossing
is contiguous. With a 64bit application running on a
32bit kernel you could run into a case where we cross
from segment 0xF to 0x0 and they should be considered
contiguous, but the assert does not take that into
account.

PROBLEM CONCLUSION:
Modified the assert to tolerate this case.

------

APAR: IY24432 COMPID: 5639I3700 REL: 310
ABSTRACT: FIX INDENTATION, ADD COMMENTS TO BIND_PROCESS FOR READABILITY

PROBLEM DESCRIPTION:
The indentation is bad in the source code and no comments
at all.

PROBLEM CONCLUSION:
Fix the indentation and add comments.

------

APAR: IY24454 COMPID: 5765C3403 REL: 433
ABSTRACT: SYSTEM HANG DUE TO DEAD LOCK BETWEEN ROUTE LOCK AND IPPMTU LOCK

PROBLEM DESCRIPTION:
System hang due to dead lock between route lock and ippmtu lock
in ip_pmtu_timeout() in following scenario:
rn_walktree() acqires route lock while calling ip_pmtu_timeout()
and will try to acquire ippmtu lock in timeout fun. But this
lock is already acquired by another thread and that thread is
waiting for route lock.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* 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.mp 4.3.3.75
* bos.mp 4.3.3.76
* bos.mp 4.3.3.77
* bos.up 4.3.3.75
* bos.up 4.3.3.76
* bos.up 4.3.3.77
* bos.adt.include 4.3.3.75
* bos.adt.include 4.3.3.76
* bos.adt.include 4.3.3.77
***************************************************************
*PROBLEM DESCRIPTION: *
* The system may hang due to a deadlock between the
* ROUTE_LOCK and the IPPMTU_LOCK.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY24454
***************************************************************

PROBLEM CONCLUSION:
When doing PMTU processing from rn_walktree, instead of
actually doing the processing, which would require taking the
IPPMTU_LOCK, schedule a netisr to do the pmtu processing at
a later time when the ROUTE_LOCK is not held.

------

APAR: IY24466 COMPID: 5765C3403 REL: 433
ABSTRACT: SMITTY : BUG MODIFY INETD SERVICE OPTION

PROBLEM DESCRIPTION:
When going back to normal mode (no debug),
toggling the debug option from yes to no
t enough. Since the '-d' option also
appears in the full config file path
(last option in smitty).The later should
also be cleared manually.

PROBLEM SUMMARY:
Code checkedin. Hence closing the apar.

PROBLEM CONCLUSION:
Code checkedin. Hence closing the apar.

------

APAR: IY24471 COMPID: 5765C3403 REL: 430
ABSTRACT: SERIAL DISK FORMAT & CERTIFY FAIL

PROBLEM DESCRIPTION:
Attempting to use the format or certify service aid for
serial disks will cause an error message. The service
aid will not run for this class of disks.

PROBLEM CONCLUSION:
Use the Object Data Manager to store the correct executable
to use for formatting and certifying serial disks.

------

APAR: IY24478 COMPID: 5765C3403 REL: 433
ABSTRACT: IPSEC_LOGD DOESN'T START AFTER REBOOT

PROBLEM DESCRIPTION:
Sometimes, ipsec_logd cannot be started after system reboot.

PROBLEM CONCLUSION:
When starting ipsec_logd, the /etc/security/ipsec_logd.pid
file is checked to see whether it's obsolete. If it is,
the file need to be removed first.

------

APAR: IY24485 COMPID: 5639I3900 REL: 310
ABSTRACT: DCED CORE DUMP IN VFPRINTF

PROBLEM DESCRIPTION:
On Solaris, dced is core dumping with the following stack:
current thread: t61
=> 1 strlen() at 0xff136f54
   2 _doprnt() at 0xff1818f8
   3 vfprintf() at 0xff183720
   4 dce_svc__vfprintf() at 0xfeeb9f30
   5 dce_svc_printf_withfile() at 0xfeeba 8e0
   6 rca_site_resolve_update() at 0xfee48580
   7 _dce_exJACrdTWT() at 0xfee4a268
   8 _dce_RBcVlAWouE() at 0x77abc
   9 send_as_request() at 0xfed8601c
   10 _dce_aieYQgADmG() at 0xfed869c4
   11 _dce_FjYIdZiHTt() at 0xfee9c6fc
   12 sec_login_validate_identity() at 0xfee8c384
   13 sec_login_valid_from_keytable() at 0xfee8cde8
   14 keymgmt_handler() at 0x30fa8
   15 dce_pthread_start() at 0xfee276d0

PROBLEM SUMMARY:
On Solaris, dced is core dumping with the following stack:
(l61) terminated by signal SEGV (no mapping at the
                                  fault address)
(/opt/SUNWspro/bin/dbx) where
current thread: t61
=> 1 strlen(0x0, 0x0, 0x0, 0x7efefeff, 0x81010100,
         0xdb72c), at 0xff136f54
   2 _doprnt(0x39ee796, 0x0, 0x0, 0x0, 0x0, 0x0),
         at 0xff1818f8
   3 vfprintf(0xff1b9d24, 0x39ee778, 0xfdcf0620,
         0xff1bd5d4, 0xff1b9708, 0xff02d17c), at 0xff183720
   4 dce_svc__vfprintf(0x0, 0xa, 0x70, 0xfdcf0620,
         0xfef8770c, 0xff1b9d24), at 0xfeeb9f30
   5 dce_svc_printf_withfile(0x1, 0x84, 0x1, 0xff045904,
         0x1, 0x40), at 0xfeeba 8e0
   6 rca_site_resolve_update(0x3a26d70, 0x11ed18,
         0xfdcf0e80, 0xfdcf0f88, 0x0, 0x3507130),
         at 0xfee48580
   7 _dce_exJACrdTWT(0xfdcf0f88, 0x10d, 0xd0a550, 0x2000,
         0xfdcf0ffc, 0x3b14b78 ), at 0xfee4a268
   8 _dce_RBcVlAWouE(0x38a3120, 0xb90a8, 0x3a40930,
         0xfdcf10dc, 0xfee96e30, 0xf dcf13f0), at 0x77abc
   9 send_as_request(0xfdcf10dc, 0xfdcf10dc, 0x6a, 0x0,
         0x3bb94d4d, 0xfdcf10d4) , at 0xfed8601c
   10 _dce_aieYQgADmG(0x3a1ae44, 0xfdcf10dc, 0x0, 0x80ad4,
         0x39510c0, 0x3a1ae44 ), at 0xfed869c4
   11 _dce_FjYIdZiHTt(0x3ab41f0, 0x37616e0, 0x37616e0,
         0xfdcf1be8, 0x3ab41f0, 0 x0), at 0xfee9c6fc
   12 sec_login_validate_identity(0x3ab41f0, 0x3ab41f0,
         0x3ab41f0, 0x0, 0x0, 0x fdcf1bd4), at 0xfee8c384
   13 sec_login_valid_from_keytable(0x37616e0, 0x3ab41f0,
         0xfdcf1be8, 0xfdcf1be 8, 0xfdcf1cf4, 0xfdcf1bd8),
         at 0xfee8cde8
   14 keymgmt_handler(0xc11b0, 0x0, 0x6, 0xff24c9ac, 0x4,
         0x1fce40), at 0x30fa8
   15 dce_pthread_start(0x1fce40, 0xfefc2274, 0xfef8770c,
         0x1fce40, 0x1, 0xff24 c9ac), at 0xfee276d0

PROBLEM CONCLUSION:
This is a problem with NULL string pointers being passed
into vfprintf. Apparently AIX is more forgiving about
this than Solaris.
Within rca_site_resolve_update():
 ...
if (BAD_STATUS(status)) {
 rpc_binding_to_string_binding(tmp_rpch, &rpch_string, &xst);
 dce_svc_printf(SEC_RCA_OP_STATUS_MSG ,
                rpch_string,
                msgp=dce_sprintf(*status));
 ...
Both rpc_binding_to_string_binding and dce_sprintf involve
memory allocation to which rpch_string and msgp will point.
However, if for some reason that fails, or some other error
within rpc_binding_to_string_binding leaves
rpch_string NULL, we'll end up calling dce_svc_printf with
a NULL string pointer (which eventually cores in strlen).
The fix is to pass in a null string if either of these
pointers is NULL:
  msgp=dce_sprintf(*status);
  dce_svc_printf(SEC_RCA_OP_STATUS_MSG ,
  (rpch_string!=NULL)?rpch_string:(unsigned char *)"",
  (msgp!=NULL)?msgp:(unsigned char *)"");

------

APAR: IY24497 COMPID: 5765C3403 REL: 430
ABSTRACT: LARGE VALUE FOR LP0:LINE (LINES ON PAGE) CAUSES DUP PRINT

PROBLEM DESCRIPTION:
Large values for line (greater that 2051 lines on page) on a
native parallel printer may cause the driver job buffer to be
sent to the printer twice.

PROBLEM CONCLUSION:
Accept impossibly large values.

------

APAR: IY24501 COMPID: 5765D5100 REL: 320
ABSTRACT: MICROCODE DEAD ERROR RUNNING USER SPACE JOBS

PROBLEM DESCRIPTION:
Running user space job in a double/single SP-Switch2
environment a node in the job pool dropped off of the plane1
network. The node had the following errpt entry:
LABEL: CS_ADAPT_MC_ER
IDENTIFIER: FF355378
/var/adm/SPlogs/css1/adapter.log reported: Microcode Dead.

PROBLEM SUMMARY:
Without this fix microcode will hang when a double adapter
window is opened with a user buffer size that is not a
multiple of 4k.

PROBLEM CONCLUSION:
The code was changed to allocate memory on 4k boundaries
even when the user buffer size is not a multiple of 4k.

------

APAR: IY24510 COMPID: 5765D5100 REL: 320
ABSTRACT: IP INTERFACE DMA MISMATCH BETWEEN CORSAIR AND COLONY

PROBLEM DESCRIPTION:
interface dma mismatch between corsair and colony

PROBLEM SUMMARY:
SP_Switch2_PCI_Adapter h/w limitations require that DMAs
happen on word boundaries with lengths of intermediate
packets within a large message being word multiples. The
SP_Switch2_PCI_Adapter IP interface driver is designed to
handle this. Since this was not an issue on the
SP_Switch2_Adapter, the SP_Switch2_Adapter IP interface
driver was not designed to conform to this restriction. In
order for SP_Switch2_Adapter and SP_Switch2_PCI_Adapter to
coexist on the same network it will be necessary to
incorporate this change in the SP_Switch2_Adapter IP
interface driver.

PROBLEM CONCLUSION:
This is a design issue for SP_Switch2_Adapter. For
SP_Switch2_PCI_Adapter compatibility the SP_Switch2_Adapter
IP interface driver will need to ensure all packets, that
are part of a datagram, are word aligned (address & length).

------

APAR: IY24511 COMPID: 5765D5100 REL: 320
ABSTRACT: LAPI VECTOR FUNCTIONS NEED TO CHECK HAL'S RC

PROBLEM DESCRIPTION:
lapi vector functions need to check hal's rc

PROBLEM SUMMARY:
Possible data-mismatch when using the new LAPI vector
functions. This will only happen on SP_Switch2 system when
the adapter is automatically recovered without the protocols
knowing the recovery has taken place.

PROBLEM CONCLUSION:
The fix is to verify all return codes and not assume the
read of the packet is successful.

------

APAR: IY24512 COMPID: 5765D5100 REL: 320
ABSTRACT: CA & PA ERRORS IN CSS1 CAUSE ERR & SNAP ON CSS0

PROBLEM DESCRIPTION:
ca & pa errors on css1 cause err & snap on css0

PROBLEM SUMMARY:
css.snap was getting info for all adapters even if called
specifically for only one.

PROBLEM CONCLUSION:
css.snap's internal call to colread was changed to specify
which adapter to read.

------

APAR: IY24513 COMPID: 5765B9501 REL: 320
ABSTRACT: STRIPE GROUP PANIC AFTER UNMOUNT/MOUNT OF A FS AFTER SNAPSHOP

PROBLEM DESCRIPTION:
stripe group panic after unmount/mount of a fs after snapshop

PROBLEM SUMMARY:
An unmount of a file system followed by an immediate remount
failed when there were mmapped files active at the time of
the unmount.

PROBLEM CONCLUSION:
Fix error in unmount cleanup that resulted in incorrect
structures being around at the next mount if files had been
mmapped

------

APAR: IY24514 COMPID: 5765B9501 REL: 330
ABSTRACT: STRIPE GROUP PANIC AFTER UNMOUNT/MOUNT OF A FS AFTER SNAPSHOP

PROBLEM DESCRIPTION:
stripe group panic after unmount/mount of a fs after snapshop

PROBLEM SUMMARY:
An unmount of a file system followed by an immediate remount
failed when there were mmapped files active at the time of
the unmount.

PROBLEM CONCLUSION:
Fix error in unmount cleanup that resulted in incorrect
structures being around at the next mount if files had been
mmapped

------

APAR: IY24516 COMPID: 5765B9501 REL: 320
ABSTRACT: MOUNT RETURNS EBUSY BECAUSE LOG MARKED LOGNEEDSRECOVERY

PROBLEM DESCRIPTION:
mount returns ebusy because log marked logneedsrecovery

PROBLEM SUMMARY:
GPFS file system refused to mount on some nodes after some
disk connectivity errors.

PROBLEM CONCLUSION:
If openForAppend fails, panic the filesystem so that the FS
manager will check the assigned logs for recovery and mark
them logAvailable again.

------

APAR: IY24517 COMPID: 5765B9501 REL: 330
ABSTRACT: MOUNT RETURNS EBUSY BECAUSE LOG MARKED LOGNEEDSRECOVERY

PROBLEM DESCRIPTION:
mount returns ebusy because log marked logneedsrecovery

PROBLEM SUMMARY:
GPFS file system refused to mount on some nodes after some
disk connectivity errors.

PROBLEM CONCLUSION:
If openForAppend fails, panic the filesystem so that the FS
manager will check the assigned logs for recovery and mark
them logAvailable again.

------

APAR: IY24518 COMPID: 5765B9501 REL: 320
ABSTRACT: GPFS 1.5 SP:WRITEWHAT != UNKNOWNBLOCK REPUPDATE.C, LINE 1941

PROBLEM DESCRIPTION:
gpfs 1.5 sp: writewhat != unknownblock repupdate.c, line 1941

PROBLEM SUMMARY:
GPFS self check logic detected an error at repUpdate.C, line
1941.

PROBLEM CONCLUSION:
In UpdateLogger::begineWrite(), when updating replicated
data in place, must synchronize with spoolDoneRecord()
before spooling a begin-replicated-write log record.

------

APAR: IY24519 COMPID: 5765B9501 REL: 330
ABSTRACT: GPFS 1.5 SP:WRITEWHAT != UNKNOWNBLOCK REPUPDATE.C, LINE 1941

PROBLEM DESCRIPTION:
gpfs 1.5 sp: writewhat != unknownblock repupdate.c, line 1941

PROBLEM SUMMARY:
GPFS self check logic detected an error at repUpdate.C, line
1941.

PROBLEM CONCLUSION:
In UpdateLogger::begineWrite(), when updating replicated
data in place, must synchronize with spoolDoneRecord()
before spooling a begin-replicated-write log record.

------

APAR: IY24532 COMPID: 5765C3403 REL: 430
ABSTRACT: UNRESOLVED SYMBOLS

PROBLEM DESCRIPTION:
The symbols which_Rhook, which_Qhook, which_vc,
which_s, which_hostalias and which_connected were
removed from libc.a and now the customer's binary
won't run as it cannot find these symbols.
The sysmbols though exported should never be used.
They are not in a header file, nor are they
documented user interface.

PROBLEM SUMMARY:
The customer's application will fail returning:
ENOEXEC: unresolved symbols or invalid XCOFF header
Then it will the following symobols.
which_Rhook, which_Qhook, which_vc, which_s,
which_hostalias and which_connected.

PROBLEM CONCLUSION:
Exported all the symbols back to libinet.exp and
libinet64.exp. And defined empty dummy routine for each of
the symbols we exported.

------

APAR: IY24548 COMPID: 5765C3403 REL: 430
ABSTRACT: DBX FAILS WHEN RUNNING COMPLEX C++ PROGRAMS

PROBLEM DESCRIPTION:
dbx fails with SIGSEGV or SIGILL when running a
program with objects complex enough to generate
debug stabstrings > 5000 bytes in length.

PROBLEM SUMMARY:
dbx will crash with an illegal instruction or segmentation
violation error when debugging a program that has complex
objects or types.

PROBLEM CONCLUSION:
Modify data structures in stabstring module to handle entries
greater than 5000 bytes.

------

APAR: IY24557 COMPID: 5765D5100 REL: 320
ABSTRACT: COLONY:AUTOUNFENCE FLOW NEEDS TO HANDLE ASYNC.E/S PKTS BETTER

PROBLEM DESCRIPTION:
colony:autounfence flow needs to handle async.e/s pkts better

PROBLEM SUMMARY:
While testing a node for autounfence readiness, the link to
the
node is enabled in the INIT packet sent to the switch chip
attached to the node. In case the link has an error, an
asynchronous Error Status packet will be sent to the
primary.
Depending on the order in which the asynchronous and
response-type
Error Status packets arrive at the primary, different
problems
occur. One undesirable effect is that the the scan code can
bail
out leaving the link enabled. Switch error recovery is
expected
to handle the asynchronous packet. As soon as switch error
recovery resets the error, the next one is generated as fast
as it was reset. This can cause the primary to be flooded
with
errors from the same port.

PROBLEM CONCLUSION:
The fix is to change the autounfence code so that before
bailing out when we get an async packet after enabling the
link of the node to unfence, we will disable the open link
before returning.

------

APAR: IY24558 COMPID: 5765D5100 REL: 320
ABSTRACT: DMA INTERRUPT ERROR

PROBLEM DESCRIPTION:
dma interrupt error

PROBLEM SUMMARY:
When running on a Regatta node extra care must be taken to
execute a sync instruction before DMA operations. Otherwise
the DMA operations could pick up the wrong information,
resulting in duplicate packets or corrupted data.

PROBLEM CONCLUSION:
Sync operations were added to code that does DMA operations.

------

APAR: IY24559 COMPID: 5765D5100 REL: 320
ABSTRACT: MPCI SHOULD CHECK HAL RC FOR READPKT

PROBLEM DESCRIPTION:
mpci should check hal rc for readpkt

PROBLEM SUMMARY:
Due to a timing condition a read of a packet from the switch
may fail after a peek into that data has completed
successfully. In addition the read may return different
data than the peek did, causing data corruption.

PROBLEM CONCLUSION:
The return codes are checked and care is taken to avoid bad
return codes when possible.

------

APAR: IY24560 COMPID: 5765B9501 REL: 320
ABSTRACT: ASSERT IN ACL GARBAGE COLLECTOR

PROBLEM DESCRIPTION:
assert in acl garbage collector

PROBLEM SUMMARY:
GPFS self check logic detected an error in acl.C.

PROBLEM CONCLUSION:
Acquire flush flag after BR lock in acl garbage collector.

------

APAR: IY24561 COMPID: 5765B9501 REL: 330
ABSTRACT: ASSERT IN ACL GARBAGE COLLECTOR

PROBLEM DESCRIPTION:
assert in acl garbage collector

PROBLEM SUMMARY:
GPFS self check logic detected an error in acl.C.

PROBLEM CONCLUSION:
Acquire flush flag after BR lock in acl garbage collector.

------

APAR: IY24563 COMPID: 576565400 REL: 220
ABSTRACT: SYSCALLS FAILED TO TRACE ALL THE SYSTEM CALLS

PROBLEM DESCRIPTION:
syscalls failed to trace all the system calls which is
called between syscalls -start and syscalls -stop commands.

PROBLEM CONCLUSION:
Fixed the code to make syscalls properly trace all the
system calls

------

APAR: IY24566 COMPID: 5765C3403 REL: 430
ABSTRACT: SETPCRED FAILS AS GETPRCRED REMOVES DOESN'T PUT '=' AFTER

PROBLEM DESCRIPTION:
The getpcred call for a NISPLUS user doesn't put
'=' after GROUPS keyword while returning the
keyword value pairs of the form <key>=<value>.
setprcred fails when it uses the output from
getpcred to set the process credentials. Due to
this some applications like printing for a
NISPLUS user fails. This may not only be limited
to NISPLUS but can extend to other users as well.

PROBLEM SUMMARY:
The getpcred call for a NISPLUS user doesn't put
'=' after GROUPS keyword while returning the
keyword value pairs of the form <key>=<value>.
setprcred fails when it uses the output from
getpcred to set the process credentials. Due to
this some applications like printing for a
NISPLUS user fails. This may not only be limited
to NISPLUS but can extend to other users as well.

PROBLEM CONCLUSION:
If ',' is there then only NULL it. Code is added for
incorporating this logic.

------

APAR: IY24568 COMPID: 5765C3403 REL: 430
ABSTRACT: GETPWUID IN 64BIT APPLICATION FAILS WHILE USING INDEX FILES

PROBLEM DESCRIPTION:
The getpwuid related calls like getpwuid(),
getpwuid_r() in a 64bit application will fail
while resolving the 'id' from an index file.
The 32bit calls work fine.
The workaround for this is to have /etc/passwd file
used instead of the index file for id resolution.
This can be done by 'touch /etc/passwd' making this
file current. If there are large number of users
then this will cause a performance problem as the
library functions read each line from the file, but
this will resolve the getpwuid() problem.
The apar IY22458 didn't fix the problem. Installing
that apar will not cause any other problems also other
than the problem being unresolved. This apar is opened
to take care of the problem.

PROBLEM CONCLUSION:
The file_info_arG array assigned values has KEYT_ULONG for
key id. This is changed to KEYT_INT. Also there is an
occurance of KEYT_ULONG in the code which is changed
to KEYT_INT.

------

APAR: IY24575 COMPID: 5765D2800 REL: 431
ABSTRACT: CLUSTER NODE VG DEFINITIONS ARE LOST - HACMP,HAES

PROBLEM DESCRIPTION:
The customer had a hardware problem which caused his ssa disks
to be in a defined state rather than available on one node.
When the other node then fell over to this node, the vgs were
exported and the following importvg failed so the vg definitions
were completely lost from the node.

PROBLEM SUMMARY:
The customer had a hardware problem which caused his ssa disks
to be in a defined state rather than available on one node.
When the other node then fell over to this node, the vgs were
exported and the following importvg failed so the vg
definitions were completely lost from the node.

PROBLEM CONCLUSION:
The code in clvaryonvg was changed to check the return
code from the initial varyonvg attempt and if that return
code was NOQUORUM, to exit at that point instead of
attempting exporvg/importvg.

------

APAR: IY24596 COMPID: 5765C3403 REL: 430
ABSTRACT: BUFFER OVERFLOW IN LIBDTSVC.A

PROBLEM DESCRIPTION:
Buffer Overflow problem in the
libDtSvc.a CDE library.

PROBLEM SUMMARY:
Invoke "dtprintinfo -session <really_long_Filename>
It dumps core.
This happens for dtterm also.

PROBLEM CONCLUSION:
In the libDtSvc.a library a check for the maximum filename
length was done.

------

APAR: IY24597 COMPID: 5765C3403 REL: 430
ABSTRACT: NISPLUS PASSWORD EXPIRATION PROBLEMS WITH NON-AIX SERVER

PROBLEM DESCRIPTION:
For non-AIX NISPLUS servers the expiration fields might contain
"-1" to indicate no expiration. However, this is misinterpreted
by the AIX client and the user will not be allowed to log in.

LOCAL FIX:
Modify the shadow field to be empty (::) instead if -1 (:-1:).

PROBLEM CONCLUSION:
Treat both empty field and "-1" as never expiring passwords.

------

APAR: IY24618 COMPID: 5765C3403 REL: 430
ABSTRACT: BINLD BREAKS THE IPFORWARDING CAPABILITY OF THE ROUTER

PROBLEM DESCRIPTION:
binld breaks the ipforwarding capability of a router.
Once binld is started in a router (by startsrc -s
binld) the router no longet forward the IP packets.
This problem exists till the next reboot.

PROBLEM SUMMARY:
Customer will see IP forwarding getting affected at the
gateway

PROBLEM CONCLUSION:
The lines which set the option IP_DHCPMODE were removed

------

APAR: IY24680 COMPID: 5765C3403 REL: 430
ABSTRACT: TRAP IN BACKTRACKING STATE

PROBLEM DESCRIPTION:
Crash happens in xmemunpin.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* bos.mp 4.3.3.75
* bos.mp 4.3.3.76
* bos.mp 4.3.3.77
* bos.up 4.3.3.75
* bos.up 4.3.3.76
* bos.up 4.3.3.77
***************************************************************
*PROBLEM DESCRIPTION: *
* Crash happens in xmemunpin.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY24680
***************************************************************

PROBLEM CONCLUSION:
Save temporary segment registers so they can be reused.

------

APAR: IY24685 COMPID: 5765C3403 REL: 430
ABSTRACT: HDISK ERROR LOGGED AFTER A REMOVE OPERATION(HOT PLUG TASK)

PROBLEM DESCRIPTION:
Error is logged after a remove operation from the
Diagnostics Hot Plug Task

PROBLEM CONCLUSION:
Change of program flow

------

APAR: IY24692 COMPID: 5765C3403 REL: 430
ABSTRACT: DSI SYSTEM CRASH IN FINDFREE() WITH 0C20 IN LEDS

PROBLEM DESCRIPTION:
System crashed with DSI

PROBLEM CONCLUSION:
grow fixed stack window

------

APAR: IY24699 COMPID: 5765C3403 REL: 433
ABSTRACT: ICMP PACKETS NOT SEEN IN TCPDUMP AFTER SWITCHING ADAPTER IN NETI

PROBLEM DESCRIPTION:
icmp packets not seen in tcpdump after switching of adapters in
netif_backup mode since a watchdog timer was not started.

PROBLEM SUMMARY:
icmp packets not seen in tcpdump after switching of adapters
in netif_backup mode since a watchdog timer was not started.

PROBLEM CONCLUSION:
 Timer is restarted in netif_backup

------

APAR: IY24721 COMPID: 5765C3403 REL: 430
ABSTRACT: MAN COMMAND DOESN'T SHOW SOME JA_JP CHARACTERS

PROBLEM DESCRIPTION:
more command does not display some JA_JP characters.

PROBLEM SUMMARY:
Though the problem was reported with the man command,
the actual problem is with the more command.
It it the more command which fails to display some
Ja_JP characters.

PROBLEM CONCLUSION:
more commands code was changed to fix the problem.

------

APAR: IY24766 COMPID: 5765C3403 REL: 430
ABSTRACT: PTRACE PTT_WRITE_SPRS ON 64BIT DEBUGGEE CLOBBERS SF BIT IN MSR

PROBLEM DESCRIPTION:
Debugging a 64bit threaded executable may
crash the system.

PROBLEM CONCLUSION:
Fix ptrace PTT_WRITE_SPRS to not clobber
the SF bit in the msr.

------

APAR: IY24774 COMPID: 5765C3403 REL: 430
ABSTRACT: FC TAPEDRIVER REPORTS UNEXPECTED ERRONO.

PROBLEM DESCRIPTION:
Same errcode (ENXIO) seen if EOM or EOT check condition
and on failed I/O commands due to disconnect.
Also see (EIO) on close(),if EOM/EOT check condition reported
and res_supported attribute is NOT set.
Close() after Diag open, will report unknown errorno.

PROBLEM CONCLUSION:
Correct errno reporting in close() routine.
Initialize retcode for error variable.
Correct errno reporting in iodone() routine.
Process ENXIO correctly for certain types of error conditions.

------

APAR: IY24801 COMPID: 5765C3403 REL: 430
ABSTRACT: RPC.STATD CORE DUMPS

PROBLEM DESCRIPTION:
rpc.statd might coredump due to initialized pointer
dereference.

PROBLEM SUMMARY:
***************************************************************
*USERS AFFECTED: *
* All users with the following filesets at these levels *
* bos.net.nfs.client 4.3.3.75
* bos.net.nfs.client 4.3.3.76
* bos.net.nfs.client 4.3.3.77
***************************************************************
*PROBLEM DESCRIPTION: *
* rpc.statd might coredump due to initialized pointer
* dereference. There's also a small memory leak in the
* daemon.
***************************************************************
*RECOMMENDATION: *
* Apply apar IY24801
***************************************************************

PROBLEM CONCLUSION:
Correct initialization of structures. Fix memory leak.

------

APAR: IY24802 COMPID: 5765C3403 REL: 430
ABSTRACT: INFINIE LEASE NOT GETTING IMPLEMENTED PROPERLY

PROBLEM DESCRIPTION:
When the config file has been configured
for giving infinite lease it doesnot work
as per design ie when the garbage handler
thread runs. It put all the INFINITE leased
addresses to EXPIRIED.

PROBLEM CONCLUSION:
Changing function expire_leased_addrs and expire_reserverd_
addrs. Also changes to save_clientrecords.

------

APAR: IY24813 COMPID: 5765B9501 REL: 330
ABSTRACT: LONG WAIT FOR SG RECOVERY

PROBLEM DESCRIPTION:
long wait for sg recovery

PROBLEM SUMMARY:
GPFS code looped causing access to one file system to hang
under extreme loads

PROBLEM CONCLUSION:
Correct an uninitialized variable in the log sequence number
functions.

------

APAR: IY24814 COMPID: 5765B9501 REL: 320
ABSTRACT: LONG WAIT FOR SG RECOVERY

PROBLEM DESCRIPTION:
long wait for sg recovery

PROBLEM SUMMARY:
GPFS code looped causing access to one file system to hang
under extreme loads

PROBLEM CONCLUSION:
Correct an uninitialized variable in the log sequence number
functions.

------

APAR: IY24815 COMPID: 5765B9501 REL: 320
ABSTRACT: GPFS ASSERT- STP->CURRENTRANGES->NRANGES !=0 TOKENCLASS.C

PROBLEM DESCRIPTION:
gpfs assert -stp->current ranges ->nranges !=0 tokenclass.c

PROBLEM SUMMARY:
GPFS self check logic asserted a failure at tokenclass.C,
line 3899.

PROBLEM CONCLUSION:
Correct an incorrect assert.

------

APAR: IY24816 COMPID: 5765B9501 REL: 330
ABSTRACT: GPFS ASSERT- STP->CURRENTRANGES->NRANGES !=0 TOKENCLASS.C

PROBLEM DESCRIPTION:
gpfs assert -stp->current ranges ->nranges !=0 tokenclass.c

PROBLEM SUMMARY:
GPFS self check logic asserted a failure at tokenclass.C,
line 3899.

PROBLEM CONCLUSION:
Correct an incorrect assert.

------

APAR: IY24817 COMPID: 5765B9501 REL: 320
ABSTRACT: IMPROVE DEBUGGING CLIENT TOKEN SEQUENCE NUMBERS

PROBLEM DESCRIPTION:
improve debugging client token sequence numbers

PROBLEM SUMMARY:
Additional debug information in service tools

PROBLEM CONCLUSION:
Additional debug information in service tools

------

APAR: IY24818 COMPID: 5765B9501 REL: 330
ABSTRACT: MMFS CAUSES PANIC AFTER GPFS ASSERTS

PROBLEM DESCRIPTION:
mmfs causes panic after gpfs asserts

PROBLEM SUMMARY:
A node panic'd when a file system mount failed due to a disk
error while doing a remount while restarting GPFS on the
node.

PROBLEM CONCLUSION:
If a filesystem gets force unmounted while the daemon is
trying to remount it during daemon restart, do not find it
in getVfsByMountid.

------

APAR: IY24819 COMPID: 5765B9501 REL: 330
ABSTRACT: GPFSASERT AFTER CHANGING DISK FAILURE GROUP DURING RESTRIPING

PROBLEM DESCRIPTION:
gpfs assert after changing disk failure group during restriping

PROBLEM SUMMARY:
GPFS self check logic failed running restripe after running
mmchdisk.

PROBLEM CONCLUSION:
Fix error handling in FileAllocator::realloc.
If an error is returned, be sure to dealloc any disk
addresses that were allocated.

------

APAR: IY24820 COMPID: 5765B9501 REL: 330
ABSTRACT: PUB AND MANPAGE ERROR FOR MMDELCLUSTER

PROBLEM DESCRIPTION:
pub and manpage error for mmdelcluster

PROBLEM SUMMARY:
The man page for mmdelcluster has the following incorrect
example:
mmdelcluster -n k145n08,k145n09,k145n10

PROBLEM CONCLUSION:
delete the -n from the example to make it
mmdelcluster k145n08,k145n09,k145n10

------

APAR: IY24825 COMPID: 5765C3403 REL: 430
ABSTRACT: LRU REDPAGE DMSTIMEOUT

PROBLEM DESCRIPTION:
Excessive scan rate for the lrud when
wlm memory management is turned on, where
the lrud is searching for "RED" pages (ie
pages belonging to segments in a class
over its limit). This could trigger a
dms_timeout or a performance degradation.
There is a possibility that before the lrud
meets its goal enough pages have been stolen
from the class to bring it back below its
limit, so lrud wouldn't find the "RED" pages
to steal that it is searching for.

PROBLEM SUMMARY:
1. Use a bigger red zone in scb_color, the macro
   used by LRU to determine the color of the pages.
   This makes it less likely that the color will
   change while the LRU is running.
2. Fix wlm_search_color to return WHITE when there
   are no more RED pages when called with LRU_BUCKW
   and LRU_MAXBUCKW.

PROBLEM CONCLUSION:
1. Use a bigger red zone in scb_color, the macro
   used by LRU to determine the color of the pages.
   This makes it less likely that the color will
   change while the LRU is running.
2. Fix wlm_search_color to return WHITE when there
   are no more RED pages when called with LRU_BUCKW
   and LRU_MAXBUCKW.

TEMPORARY FIX:
turn off wlm memory management

------

APAR: IY24892 COMPID: 5765D5100 REL: 320
ABSTRACT: SWITCH-ATTACHED SERVER DROPS OFF SWITCH ON TB3PCI DMA ERROR.

PROBLEM DESCRIPTION:
SP switch-attached servers may detect a false DMA error which
will result in the machine loosing switch_responds. In
addition to dropping off the switch, this TB3PCI DMA error
condition may cause a system hang if the fix for IY15408 is not
applied.

PROBLEM SUMMARY:
SP switch attached servers can drop off the switch due
to a false DMA error.

PROBLEM CONCLUSION:
The switch xilinx file for SP attached servers
(xilinx_file3pci) has been changed to avoid the problem
of a false DMA error.

------

APAR: IY24909 COMPID: 5765B9501 REL: 330
ABSTRACT: GETSUBBLOCKSPERFILEBLOCK - GPFS FAILING MMTRACE SHOWS

PROBLEM DESCRIPTION:
gpfs failure in getSubblocksPerFileBlock. Aix kernel trace
shows:
MMFS ERRLOG: Error logger dataStr threadId 133991 Failure at lin
e 1311 in file /project//sprelmo2/build/rmo2s06a/export/power/us
r/include/mmfs/OpenFile.h metadata.getSubblocksPerFileBlock
() != -1)

PROBLEM SUMMARY:
GPFS self check logic asserted at OpenFile.h line 1311

PROBLEM CONCLUSION:
Strengthen synchronization when trying to operate on a
recently deleted inode.

------

APAR: IY24910 COMPID: 5765C3403 REL: 430
ABSTRACT: WEBSM FAILS TO CORRECTLY DISPLAY LATEST FILESET INSTALLED

PROBLEM DESCRIPTION:
When displaying installed software using websm, applied updates
with a fix level of two digits for which th first digit is a
number that is lower than other applied or committed levels
will cause the wrong fileset to be listed as the highest level
installed.

LOCAL FIX:
Use lslpp to display information about installed software or
collapse the software trees in websm to reveal all the
different levels for a particular fileset. When the problem is
present the fileset with the highest modification level will
show after collapsing the tree.

PROBLEM SUMMARY:
Software updates were being listed out of order (fix - wise) in
the wsm 97 install application. Changes were made to the code
to sort in a way that greater fix numbers are shown first.

PROBLEM CONCLUSION:
The application should now display the highest level of a
fileset installed as the user would expect.

TEMPORARY FIX:
Collapsing the trees on websm for a particular fileset, reveals
all the levels installed. It would be up to the user to
notice the highest. Alternatively, lslpp can be used to
display the same information.

------

APAR: IY24925 COMPID: 5697F6400 REL: 630
ABSTRACT: AUTOMATIC MESSAGE DELETION STOPPED WORKING

PROBLEM DESCRIPTION:
The function of deleting read or unread messages stopped
working after Sept.

PROBLEM SUMMARY:
Messages are no longer automatically deleted
On systems where this is essential, filesystems holding voice
data and databases start to fill up

------

APAR: IY24927 COMPID: 5765C3403 REL: 430
ABSTRACT: ASSERT IN PANIC TRAP CALLED FROM LOADER CODE

PROBLEM DESCRIPTION:
assert in panic trap called from loader code

PROBLEM SUMMARY:
System performance will go down and some time it crashes
the system and dump will show assert in panic trap called
from loader code.

PROBLEM CONCLUSION:
Problem caused by a memory leak is corrected to properly free
the memory when the loader finishes with it.

------

APAR: IY24937 COMPID: 569692600 REL: 110
ABSTRACT: X.21 HAS PROBLEM DETECTING THE DISCONNECTION OF THE CABLES

PROBLEM DESCRIPTION:
When x.21 cables are disconnected from
the modem, the information is not passed
to the frame layer, believing that it is still
connected.

PROBLEM SUMMARY:
Customer disconnect the x.21 cable and the disconnection
is not being reported to the application

PROBLEM CONCLUSION:
Artic x.21 circuit depends %100 on the clock. The clock is
provided externaly. If the cable is disconnected, the clock
is lost, and the X21MODE will hold the last signal for a while
and will read it again. After the last read, it will pass the
information to the dusscc. This cause that when cable is
disconnected, the last state of the I signal is the actual
state, so after the X21MODE read it again, it will find the
signal on. This is done for noise protection, but it is
causing problems to customers that needs to remove the cables.
To solve the problem, the X21MODE will be removed so the
state of the signal will be report to the duscc inmediately.

------

APAR: IY24938 COMPID: 5765C3403 REL: 430
ABSTRACT: OUT OF MEMORY CAUSESON SEGMENT 13 CAUSES THE SYSTEM TO CRASH

PROBLEM DESCRIPTION:
Out of memory on segment 13 causes the system to crash.

PROBLEM SUMMARY:
Shared library text segment full can cause system crash
when system call table is allocated.

PROBLEM CONCLUSION:
Garbage collect unused libraries when out-of-memory condition
is detected while allocating the syscall table.

TEMPORARY FIX:
Frequent invokations of the 'slibclean' command as root.

------

APAR: IY24939 COMPID: 576565400 REL: 220
ABSTRACT: SVMON DOESN'T SHOW CLIENT SEGMENT FOR NFS FILES

PROBLEM DESCRIPTION:
svmon -P does not report client segment for the
NFS files.

PROBLEM SUMMARY:
Customer will not be able to see client segments in the svmon
report for NFS files.

PROBLEM CONCLUSION:
Fixed the code by adding a adding a zero check condition for
gnode.gn_seg

------

APAR: IY24946 COMPID: 5765C3403 REL: 430
ABSTRACT: SAR -D DOES NOT GIVE PROPER OUTPUT

PROBLEM DESCRIPTION:
This APAR addresses the following problem.
"sar -d" does not show the system activity in the expected
record.

PROBLEM SUMMARY:
sar -d does not show the system activity in the immediate
next record if i/o activity that has been undergone on the
system prior to that.

------

APAR: IY24950 COMPID: 5765B9501 REL: 330
ABSTRACT: EXECUTABLE COMPLETELY MEMORY MAPPED WON'E BE TERMINATED ON UNMOU

PROBLEM DESCRIPTION:
If a program is completely mapped in memory and the filesystem
unmounts, when the filesystem comes back and you try to run the
program again it dies with Bus error (coredump).

LOCAL FIX:
make a copy of the executable and run that instead

PROBLEM SUMMARY:
A program which is mapped from GPFS core dumps when trying
to reexecute the program after a forced file system unmount
while it was mapped.

PROBLEM CONCLUSION:
Invalidate memory mapped pages after remount.

------

APAR: IY24954 COMPID: 5765B9501 REL: 330
ABSTRACT: MMCHATTR FAILS TO SET NUMBER OF REPLICAS ON DIRECTORIES

PROBLEM DESCRIPTION:
mmchattr fails in setting -m 2 on directories,
even if max number of replicas has been set to 2.
        error message is : Open failed:
Cannot write to a directory.
however, if you create the directory after changing the
filesys with mmchfs <fsname> -m 2 , the replication factor
is set to 2.

PROBLEM SUMMARY:
mmchattr would not change the replication on a directory

PROBLEM CONCLUSION:
Correct mmchattr logic.

------

APAR: IY24965 COMPID: 5765C3403 REL: 433
ABSTRACT: CFGIPSEC NEED PARSE & EXEC NO CMDS BEFORE LOADING NETINET

PROBLEM DESCRIPTION:
If ipsec devices are available on the system, loadtime "no"
configurables in /etc/rc.net are not used to set network
parameters internally, instead the default values will be
used.

PROBLEM CONCLUSION:
cfgipsec will parse the /etc/rc.net file and execute all "no"
command lines before loading inetnet.

------

APAR: IY24967 COMPID: 5765C3403 REL: 430
ABSTRACT: MKLV WON'T ALLOW BB RELOCATION WITH STRIPING ON BIGVGS

PROBLEM DESCRIPTION:
mklv will not create an lv with both striping and bb
relocation on bigVGs; bb relocation will be set off.

PROBLEM CONCLUSION:
mklv command now properly allows striping and bb
relocation on bigVGs.

TEMPORARY FIX:
Use chlv to enable bb relocation after creating
the lv with striping (using mklv) on a big VG.

------

APAR: IY24977 COMPID: 576565400 REL: 220
ABSTRACT: SVMON -P OUTPUT INCORRECT WHEN TWO PROCESSES ATTACHED THE SAME

PROBLEM DESCRIPTION:
When two processes attached the same shared memory to
different addresses, the "Description" and "Esid" for
the shared memory become blank and "-" respectively.

PROBLEM CONCLUSION:
corrected the code such that segment info for every process
is updated properly.

------

APAR: IY24983 COMPID: 5765C3403 REL: 430
ABSTRACT: BACKPORT LIBPERFSTAT

PROBLEM DESCRIPTION:
No perfstat Library API support on AIX 433

PROBLEM CONCLUSION:
Port perfstat Library API back to AIX 433

------

APAR: IY24997 COMPID: 5765C3403 REL: 430
ABSTRACT: FIXES FOR 0401/4400 AND LUN 0 CONFIG ISSUES

PROBLEM DESCRIPTION:
Newly installed systems may show the first hdisk as DEFINED
rather than AVAILABLE. Systems may have I/O errors during I/O
with 4400 or 0401 ASC/ASCQ's in the detail data.

PROBLEM CONCLUSION:
Retry open operation when configuring hdisks. Allow 0401 and
4400 errors to trigger failover to the alternate controller.

------

APAR: IY25003 COMPID: 5765C3403 REL: 430
ABSTRACT: ADD MORE OUTPUT TO VMTUNE -A, BACKPORT -T,-A FLAGS

PROBLEM DESCRIPTION:
add more output to vmtune -a, and backport
-t and -A options from AIX Version 5.
-a displays current statistic counters
-t tunes the maxclient parameter, where the
   argument given is the desired maxclient
   value expressed as a percentage of real
   memory (ex. 80)
-A display mempool information

PROBLEM SUMMARY:
add more output to vmtune -a, and backport
-t and -A options from AIX Version 5.
    -a displays current statistic counters
    -t tunes the maxclient parameter, where the
       argument given is the desired maxclient
       value expressed as a percentage of real
       memory (ex. 80)
    -A display mempool information

PROBLEM CONCLUSION:
add more output to vmtune -a, and backport
-t and -A options from AIX Version 5.
    -a displays current statistic counters
    -t tunes the maxclient parameter, where the
       argument given is the desired maxclient
       value expressed as a percentage of real
       memory (ex. 80)
    -A display mempool information

------

APAR: IY25022 COMPID: 5765C3403 REL: 430
ABSTRACT: FIND REPORTS WRONG INODES WHEN USED WITH -DEPTH OPTION

PROBLEM DESCRIPTION:
find command when used with -depth option in certain cases
reports wrong inode values for certain dir entries

PROBLEM SUMMARY:
find command when used with -depth option in
certain cases reports wrong inode values for
certain dir entries.

PROBLEM CONCLUSION:
find performs a recursive search into the directories
when -depth is specified.Once the leaf is reached
it goes to previous directory and displays all
directory entries in that. Currently it is going
back to previous dir even when filename is "." Code
modified to see that chdir is not done in this
particular case.

------

APAR: IY25032 COMPID: 5765C3403 REL: 430
ABSTRACT: BOOTEXPAND FAILS WITH 8MB MEMORY MODULES IN SOME MCA MACHINES

PROBLEM DESCRIPTION:
185 is displayed on the LED display during the boot process
on certain MCA models that have 8MB dimms installed.

LOCAL FIX:
A workaround is to use dimms larger than 8MB.

PROBLEM SUMMARY:
The fix for /usr/sbin/bootexpand has been submitted.

PROBLEM CONCLUSION:
The fix for /usr/sbin/bootexpand has been submitted.

------

APAR: IY25035 COMPID: 5765B9501 REL: 320
ABSTRACT: GETSUBBLOCKSPERFILEBLOCK - GPFS FAILING MMTRACE SHOWS

PROBLEM DESCRIPTION:
gpfs failure in getSubblocksPerFileBlock. Aix kernel trace
shows:
MMFS ERRLOG: Error logger dataStr threadId 133991 Failure at lin
e 1311 in file /project//sprelmo2/build/rmo2s06a/export/power/us
r/include/mmfs/OpenFile.h metadata.getSubblocksPerFileBlock
() != -1)

PROBLEM SUMMARY:
GPFS self check logic asserted at OpenFile.h line 1311

PROBLEM CONCLUSION:
Strengthen synchronization when trying to operate on a
recently deleted inode.

------

APAR: IY25052 COMPID: 569692600 REL: 110
ABSTRACT: L2 DOWN AND UP FOR EVER, MICROCODE PROBLEM

PROBLEM DESCRIPTION:
Frame layer begins going up and down.
Other site in the nework did not receive any frame.
L2 up and down, SABMs, UA.

PROBLEM SUMMARY:
After an unexpected interrupt, x.25 frame layer start going
up and down for ever. This is cause by the transmitter that
stop.

PROBLEM CONCLUSION:
Modify the microcode to send a Transmit start of message every
time that we get into this loop.

------

APAR: IY25060 COMPID: 5697F6400 REL: 640
ABSTRACT: MESSAGE CENTER FAX ID DISPLAYS BROOKTROUT.

PROBLEM DESCRIPTION:
When a fax machine calls Message Center, its display shows
the other fax machine id as Brooktrout Fax.
This is inappropriate - Message Center would be more
appropriate.

PROBLEM SUMMARY:
Fax machines display 'Brooktrout Fax' when
sending or receiving faxes to/from Message Center.

PROBLEM CONCLUSION:
Code fixed to output 'Message Center'
instead of 'Brooktrout Fax'.

------

APAR: IY25063 COMPID: 5697F6400 REL: 640
ABSTRACT: AUTOMATIC MESSAGE DELETION STOPPED WORKING

PROBLEM DESCRIPTION:
The function of deleting read or unread messages stopped
working after Sept.

PROBLEM SUMMARY:
Messages are no longer automatically deleted
On systems where this is essential, filesystems holding voice
data and databases start to fill up

------

APAR: IY25064 COMPID: 5765C3403 REL: 430
ABSTRACT: SORT COMMAND REPORT ERRORA LINE OF THE INPUT FILE CONTAINS MORE

PROBLEM DESCRIPTION:
sort command report error"A line of the input file contains
more than 20480 characters".

PROBLEM SUMMARY:
sort command reports the error a line of the
input file contains more that 20480 characters.

PROBLEM CONCLUSION:
sort command's code modified to fix the problem.

------

APAR: IY25096 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSTEM HANG RUNNING TCP/IP STRESS TESTS

PROBLEM DESCRIPTION:
System hung due to IPPMTU_LOCK being held and never released.

PROBLEM CONCLUSION:
Make sure that the IPPMTU_LOCK is not left locked

------

APAR: IY25098 COMPID: 5765C3403 REL: 430
ABSTRACT: POSSIBLE SYSTEM CRASH WITH PMTU DISCOVERY ON

PROBLEM DESCRIPTION:
System may crash with memory corruption when PMTU discovery
is on. This can happen if the timeout function is called
before an estimate is established.

PROBLEM CONCLUSION:
Make sure that the current estimate is valid before using it
in case timeout occurs before the estimate is established.

------

APAR: IY25099 COMPID: 5765C3403 REL: 430
ABSTRACT: LIBXT.A - SOME TRANSLATIONS HAVE STOPPED WORKING

PROBLEM DESCRIPTION:
Some Translations in an Xt/Motif application may not work.

PROBLEM CONCLUSION:
Fix a minor error in the translation manager code.

------

APAR: IY25100 COMPID: 5765C3403 REL: 430
ABSTRACT: CATIA: WHEN IN SHD MODE F5 CONCAVE POLY PROB

PROBLEM DESCRIPTION:
The customer will see extraneous lines within concave
polygons when going a rotation after switching from SHD mode
to HRD mode using the F5 key instead of the menu.

PROBLEM CONCLUSION:
This phenomenon was occuring in our overflow code for
polygons. We were not setting a flag correctly to indicate
that we were in hollow mode down this particular code path.
We were not checking for the hollow flag and therefore
using incorrect breakup code.

------

APAR: IY25102 COMPID: 5765C3403 REL: 430
ABSTRACT: NOT HANDLING USER PERMISSIONS PROPERLY

PROBLEM DESCRIPTION:
If a custom tries to run any applications that call PKCS11
commands (for example ikeyman) as a non-root user then any
PKCS11 certificates, keys or data that is created by this
application may not be stored.

PROBLEM CONCLUSION:
Code fix to add check for user to be part of pkcs11 group.

------

APAR: IY25103 COMPID: 5765C3403 REL: 430
ABSTRACT: ADD 64BIT CAPABILITIES TO PKCS11

PROBLEM DESCRIPTION:
Calls to PKCS11 API will not work

PROBLEM CONCLUSION:
code fix to create 64bit versions of the PKCS11 library

------

APAR: IY25104 COMPID: 5765C3403 REL: 430
ABSTRACT: MULTI-THREADED PKCS11 APPLICATION CAN CORE DUMP

PROBLEM DESCRIPTION:
While running a multi-threaded that utilizes the PKCS11
library, it is possible that the library will cause a core dump

PROBLEM CONCLUSION:
Reduce the number of times an object is loaded by
updating and checking the counters that track the
object's revision.

------

APAR: IY25108 COMPID: 5765C3403 REL: 430
ABSTRACT: NOT DUMPING/LOGGING DATA FOR DEFERRED FATALITY JFS CORRUPTION

PROBLEM DESCRIPTION:
Certain types of systemn crashes are difficult to diagnose.

PROBLEM CONCLUSION:
Provides better debug information for analysis.

------

APAR: IY25109 COMPID: 5765C3403 REL: 430
ABSTRACT: DTLOGIN RESOURCE CUSTOMIZATION PROBLEM

PROBLEM DESCRIPTION:
Customized dtlogin screen will not appear after
the reboot.

PROBLEM CONCLUSION:
Reload the resources after file systems are mounted.

------

APAR: IY25118 COMPID: 5765C3403 REL: 430
ABSTRACT: "IMPROVE IDE CDROM ERROR HANDLING ON BAD CDS"

PROBLEM DESCRIPTION:
In this case, 'bad' CD includes invalid formatted CD, severely
scratched data CD, or incompatible CD. For example, performing
the following on an audio CD, 'mount -v cdrfs -r /dev/cd0
/mnt', will cause the IDE interface to hang. The process may
not return for an extended period of time. Although this does
not affect other parts of the system's functionality, it does
require a reboot to recover the IDE interface.

PROBLEM CONCLUSION:
The above problem is associated with the DMA data transfer, as
the DMA engine has trouble handling the error situation while
maintaining high speed data transfer. This new driver will give
the user an option of turning off DMA data transfer to obtain
better error handling capability. For better performance, both
in data tranfer rate and CPU utilization, the IDE driver will
be installed with DMA enabled. However, if the user desires
more predictable error handling, the DMA feature may be turned
off by executing the following system commands: "rmdev -l cd0;
chdev -l ide0 -a enable_dma=no;mkdev -l cd0".

------

APAR: IY25119 COMPID: 5765C3403 REL: 430
ABSTRACT: WEBSM SUBSYSTEMS APPLET AND PC CLIENT NOT WORKING

PROBLEM DESCRIPTION:
When trying to display the src subsystems using the websm gui
in pc client mode or applet mode no information is displayed.

PROBLEM CONCLUSION:
Initialize allocated memory

------

APAR: IY25136 COMPID: 5765D5100 REL: 320
ABSTRACT: NEW OPTION ST_FORCE_WINDOW FOR SWTBL_CLEAN_TABLE API

PROBLEM DESCRIPTION:
new option st_force_window for swtbl_clean_table api

PROBLEM SUMMARY:
The swtbl_clean_table API needs a new option
to force unloading a window that is hung.

PROBLEM CONCLUSION:
The swtbl_clean_table API now has a new option
to force unloading a window that is hung
called ST_FORCE_WINDOW.

------

APAR: IY25139 COMPID: 5765D2800 REL: 430
ABSTRACT: LOGREDO NOT RUN ON ALL JFSLOGS - HACMP,HAES

PROBLEM DESCRIPTION:
In a cluster configuration consisting of multiple volume
groups included in a single resource group, it was discovered
that the logredo command was run only against the jfslogs of
the last volume group in get_disk_vg_fs.

PROBLEM CONCLUSION:
The get_disk_vg_fs script was changed so that logredo is
now run against all of the jfslogs of all volume groups.

------

APAR: IY25144 COMPID: 5765D5101 REL: 121
ABSTRACT: INCORRECT GRPGLSM'S CSS (0!1) MEMBERSHIP

PROBLEM DESCRIPTION:
incorrect grpglsm's css(0!1) membership

PROBLEM SUMMARY:
Due to the incorrect evaluation of adapter status
in the HAGS switch support daemon (HAGSGLSM),
the switch memberships report the incorrect
result at certain situations.
For example, if two IP aliases are configured
for a switch adapter (e.g., css0), and the first
IP interface is down, the cssMembership group
may not include that adapter even if the second
IP interface is up.
With this fix, a switch adapter will be always
a member of the associated switch membership
as long as one of IP aliases is up.

PROBLEM CONCLUSION:
The globalized switch memberships (cssMembership and
css0Membership)
with alias network configuration in HACMP should be
based on one of adapter status between aliases.
It is because two aliases are mapped to one physical
adapter.

------

APAR: IY25148 COMPID: 5765C3403 REL: 430
ABSTRACT: NETSTAT -ZM DOESN'T ZERO OUT THE EXTENDED NET STATISTICS (I.E.

PROBLEM DESCRIPTION:
 netstat -Zm doesn't zero out the extended net statistics.

PROBLEM CONCLUSION:
Passing all the parameters of tbl_kmemstats to zeroit function.

------

APAR: IY25155 COMPID: 5765B9501 REL: 320
ABSTRACT: ONLINE MMFSCK CAN CORRUPT FILESYSTEM

PROBLEM DESCRIPTION:
online mmfsck can corrupt filesystem

PROBLEM SUMMARY:
Potential incorrect calculation of free blocks
mmfsck

PROBLEM CONCLUSION:
fix a problem where online fsck will
potentially assume an incorrect nsubblocks for an address
in a file currently being allocated. this latter problem
has very bad consequences where we could delete allocations
in the bitmap that are actually in use.

------

APAR: IY25156 COMPID: 5765B9501 REL: 330
ABSTRACT: ONLINE MMFSCK CAN CORRUPT FILESYSTEM

PROBLEM DESCRIPTION:
online mmfsck can corrupt filesystem

PROBLEM SUMMARY:
potential incorrect calculation of free
blocks in mmfsck

PROBLEM CONCLUSION:
fix a problem where online fsck will
potentially assume an incorrect nsubblocks for an address
in a file currently being allocated. This latter problem has
very bad consequences where we could delete allocations in the
bitmap that are actually in use.

------

APAR: IY25157 COMPID: 5765C3403 REL: 430
ABSTRACT: PROBLEMS WITH REORGVG AND STRICTNESS

PROBLEM DESCRIPTION:
Customer will experience failures with reorgvg when running
on volume groups which contain logical volumes which are
mirrored in non-strict or super-strict mode.

PROBLEM CONCLUSION:
Call allocp with correct strictness flags within reorgvg.

------

APAR: IY25158 COMPID: 5765C3403 REL: 430
ABSTRACT: DUMPS MAY NOT BE FULLY COPIED.

PROBLEM DESCRIPTION:
Last bytes from the dump copied to a file from a logical volume
might be missing because of the header added by LVM.

PROBLEM CONCLUSION:
Skip the LVM header before copying the dump.

------

APAR: IY25159 COMPID: 5765C3403 REL: 430
ABSTRACT: ADD DUMP -TV DATA TO LVM SNAP

PROBLEM DESCRIPTION:
customer have an extra step to send LVM data.
Add more data collection commands(LVM) to snap.

PROBLEM CONCLUSION:
Add more data collection commands(LVM) to snap.

------

APAR: IY25160 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSINTR ERRORS ARE OCCURING W/64 BIT OGL TESTCASES

PROBLEM DESCRIPTION:
SYSINTR error logs due to bogus interrupt.

PROBLEM CONCLUSION:
The problem was due to pre-fetched and pre-handled interrupts
were mis-informed in the interrupt handler which caused bogus
error logs. To fix the problem, a better tracking of the
pre-processed interrupt was implemented.

------

APAR: IY25163 COMPID: 5765C3403 REL: 430
ABSTRACT: DMA PAGE MAPPING CALCULATION ERROR FOR FC ADAPTER

PROBLEM DESCRIPTION:
I/O may be slow to respond and perhaps fail due to
misallocation of DMA resources when allocating DMA pages in
segment 15 with address space 0xf000_0000<->0xffff_ffff.

PROBLEM CONCLUSION:
Create temporary space to hold the page offset from buf
address and start calculation from page 0.

------

APAR: IY25164 COMPID: 5765C3403 REL: 430
ABSTRACT: TARGET MODE SCSI: SYNC COMMAND REQUIRED

PROBLEM DESCRIPTION:
PowerPC architecture defines a storage model for ordering of
storage accesses that allows for improved performance. However,
this model requires that software explicitly order accesses to
storage that is shared by threads or IO devices. Otherwise
there exists the possibility for threads or IO devices to read
stale data. Newer and faster systems that use this storage
model extensively can surface these problems.
After review of the SCSI adapter target-mode driver code, we
have identified a few areas that need some ordering to prevent
such problems.

PROBLEM CONCLUSION:
Target specific areas in the SCSI adapter device driver that
require storage access to be ordered. This fix completes the
necessary changes in the driver for ordered storage access.

------

APAR: IY25165 COMPID: 5765B9501 REL: 320
ABSTRACT: MMCHATTR FAILS TO SET NUMBER OF REPLICAS ON DIRECTORIES

PROBLEM DESCRIPTION:
mmchattr fails in setting -m 2 on directories,
even if max number of replicas has been set to 2.
        error message is : Open failed:
Cannot write to a directory.
however, if you create the directory after changing the
filesys with mmchfs <fsname> -m 2 , the replication factor
is set to 2.

PROBLEM SUMMARY:
mmchattr would not change the replication on a directory

PROBLEM CONCLUSION:
Correct mmchattr logic.

------

APAR: IY25166 COMPID: 5765C3403 REL: 430
ABSTRACT: MP ATTRIBUTES STILL GET CONFUSED

PROBLEM DESCRIPTION:
On a MP system facet normal vertex color
Tstrips will not always render properely.

PROBLEM CONCLUSION:
Force the facet normal vertex color
Tstrip renderer to always validate
the color material.

------

APAR: IY25169 COMPID: 5765C3403 REL: 430
ABSTRACT: CANNOT INPUT NATIONAL CHARACTERS WHEN USING ALTERNATE KEYBOARD

PROBLEM DESCRIPTION:
The customer could not use alternate keyboard
to input national characters.

PROBLEM CONCLUSION:
create the link for altim

------

APAR: IY25170 COMPID: 5765D2800 REL: 430
ABSTRACT: CLSMUXPD USES INCORRECT SUBSYSTEM NAME WHEN CALLING SRCSTAT

PROBLEM DESCRIPTION:
The clsmuxpd uses the srcstat subroutine to know the status
of clinfo on all nodes. If the clsmuxpd call to srcstat fails
the remote request the status is set to unknown. The name
used in the getStat call in refresh.c is "clinfo", it should
be "clinfoES".

PROBLEM SUMMARY:
The "snmpinfo -m dump -o /usr/es/sbin/cluster/hacmp.defs
clinfo" command always returns status for clinfo as unknown(8).

PROBLEM CONCLUSION:
Use subsystem name "clinfoES" in the getStat calls in
refresh.c.

------

APAR: IY25171 COMPID: 5765D2800 REL: 430
ABSTRACT: THE SNMPINFO COMMAND SHOWS INCORRECT MIB VALUE FOR THE CLSMUXPD.

PROBLEM DESCRIPTION:
The following command will show -1 in the clsmuxpd MIB
information and the version is also incorrect.
snmpinfo -v -m dump -o /usr/sbin/cluster/hacmp.defs clsmuxpd
clsmuxpdGets.4294967295 = 90996
clsmuxpdGetNexts.4294967295 = 969
clsmuxpdSets.4294967295 = 0
clsmuxpdTraps.4294967295 = 24
clsmuxpdErrors.4294967295 = 2
clsmuxpdVersion.4294967295 = "4.3"

PROBLEM CONCLUSION:
Changed the o_clsmuxpd function to start loop at 0.
Changed #define VERSION for correct version.

------

APAR: IY25176 COMPID: 5765C3403 REL: 430
ABSTRACT: ZAPDBREPLAY CORES ON ZOOMDINO BINARY TRACE

PROBLEM DESCRIPTION:
ZAPdbReplay does not recognize a valid GLX attribute and
coredumps while trying to replay a binary trace file.

PROBLEM CONCLUSION:
Modify parser to recognize GLX_TRANSPARENT_INDEX_EXT
attribute in binary trace header file.

------

APAR: IY25178 COMPID: 5765C3403 REL: 430
ABSTRACT: ZAPDBREPLAY CORES ON 64 BIT TRACES

PROBLEM DESCRIPTION:
64-bit version of ZAPdb will do core dump when trying to
replay a binary trace file.

PROBLEM CONCLUSION:
Added new code to support capture and replay of binary trace
files within 64-bit version of ZAPdb. A new script file and
64-bit executable have been added for performing replays.

------

APAR: IY25181 COMPID: 5765C3403 REL: 430
ABSTRACT: SMITTY WRONGLY DISPLAY CHANGED PRIMARY KEYBOARD

PROBLEM DESCRIPTION:
smitty could not display the primary keyboard correctly
when it has been changed from ISO8859-1 to others and it
always be set back to ISO8859-1

PROBLEM CONCLUSION:
make suitable changes to sm_mle.add file

------

APAR: IY25183 COMPID: 5765C3403 REL: 430
ABSTRACT: PCI LATENCY TIMER SETTING CAUSES DEGRADATION ON ADSM BACKUPS

PROBLEM DESCRIPTION:
For customers using the JBOD SCSI 825A and 875 adapters, when
moving from AIX 4.2 to AIX 4.3.3 or AIX version 5, throughput
degradation may be noticed on the following systems: F50, H50,
150 and 260.

PROBLEM CONCLUSION:
Created a special configuration attribute to work around the
throughput degradation. This change will not have any affect
on systems other than F50, H50, 150 or 260. Moreover, the
attribute is only for the JBOD SCSI 825A amd 875 adapters.

------

APAR: IY25184 COMPID: 5765C3403 REL: 430
ABSTRACT: DIAGNOSTIC TRACE FILES ARE APPENDED INSTEAD OF BEING REWRITTEN

PROBLEM DESCRIPTION:
Turning on diagnostic trace will cause files to be appended
rather than overwritten. This will eventually fill up the
/tmp directory.

PROBLEM CONCLUSION:
Any time a trace file is opened for writing, first check if
the file already exists. If it does, then unlink the file.

------

APAR: IY25185 COMPID: 5765C3403 REL: 430
ABSTRACT: SLOW SCROLL ON JAVA2 SHOWS GARBAGE WITH GXT135P

PROBLEM DESCRIPTION:
Java swingset two may leave garbage on the screen running on
a GXT135P.

PROBLEM CONCLUSION:
Added a wait for adapter flush after the screen to screen
copy code.

------

APAR: IY25186 COMPID: 5765C3403 REL: 430
ABSTRACT: DRV-07 PERFORMANCE DOWN ON GXT4000P ON 51B_0133A

PROBLEM DESCRIPTION:
When running OpenGL programs that do a lot of display list
initializations, a performance degredation can occur.

PROBLEM CONCLUSION:
When creating the octagonal bounding box for a display list,
a bounding box is created to for the display list. The
original code was not sufficient to create an adequate
bounding box. This was fixed in a prior APAR. However,
the code path for building the bounding box was increased
such that it took four times as long to build the bounding
box. The routine that handles this is being changed to
eliminate some of the redundency that was being performed
to shorten the code path.

------

APAR: IY25187 COMPID: 5765C3403 REL: 430
ABSTRACT: ATM: 155 CARD REPORTED AS BEING 622 CARD BY NETSTAT -V

PROBLEM DESCRIPTION:
ATM PCI 155mbps adapter: output of netstat -v indicates
that it is 622mbps adapter.

PROBLEM CONCLUSION:
Some adapter code was changed to correct this problem.

------

APAR: IY25188 COMPID: 5765C3403 REL: 430
ABSTRACT: PCIDMA UNDETERMINED ERROR ENTRIES IN THE AIX ERROR LOG

PROBLEM DESCRIPTION:
PCIDMA UNDETERMINED ERROR entries in the AIX error log
appeared during bootup

PROBLEM CONCLUSION:
Changing the error log routine to correct this problem.

------

APAR: IY25189 COMPID: 5765C3403 REL: 430
ABSTRACT: SYSTEM HANG RUNNING CLEAN_NET

PROBLEM DESCRIPTION:
The driver was not able to close when some of the application
is still using the buffer from the driver memory pool
This cause the system to hang.

PROBLEM CONCLUSION:
Some driver code was added to handle this situation

------

APAR: IY25190 COMPID: 5765C3403 REL: 430
ABSTRACT: ATM ILMI POSSIBLE DOUBLE FREE OF MEMORY.

PROBLEM DESCRIPTION:
Atmsvcd may core dump.

PROBLEM CONCLUSION:
Fix daemon so that it does try and release memory that
has already been freed.

------

APAR: IY25191 COMPID: 5765C3403 REL: 430
ABSTRACT: SVC RELEASED WHEN A BLLI IE IS INCLUDED IN THE CONNECT

PROBLEM DESCRIPTION:
SVC released after being established with the cause value:
100(Invalid Information Element Contents)

PROBLEM CONCLUSION:
Change atmsvcd so it fills in the AAL type fields regardless
if the fields changed or not.

------

APAR: IY25192 COMPID: 5765C3403 REL: 430
ABSTRACT: GARBAGE COLLECTION DOES NOT FUNCTION ON UP KERNEL

PROBLEM DESCRIPTION:
On the UP kernel, garbage collection of kernel memory is not
present. While this normally will have no effect, it could
lead to xmalloc() failures after a sizable number of
allocations with none of a certain size.

PROBLEM CONCLUSION:
Start the garbage collection kernel process on the UP kernel.

------

APAR: IY25193 COMPID: 5765C3403 REL: 430
ABSTRACT: TRCRPT SEEMS TO BE INCORRECTLY/HALF DECODING TRACE FILES

PROBLEM DESCRIPTION:
Using $BREAK or $SKIP, or a string error, in a function
causes the report to be unreliable.

PROBLEM CONCLUSION:
Reset the function level when starting to format a hook.

------

APAR: IY25194 COMPID: 5765C3403 REL: 430
ABSTRACT: X25 QLLC - CRASH AT QLM_DELETE_STATION+00008C

PROBLEM DESCRIPTION:
pvthread+085500 STACK:
qlm_delete_station+00008C (0000000B, 01D100CC, D97430C8 ?? )
qlm_receive_data+00048C (??, ??, ??)
qpm_interrupt_handler+000240 (??, ??, ??)
procentry+000010 (??, ??, ??, ??)

PROBLEM CONCLUSION:
Insure that a link station is not doubly deleted when the
X.25 comio device is continually responding with EIO.

------

APAR: IY25195 COMPID: 5765C3403 REL: 430
ABSTRACT: IME (ZH_TW) RELATED PROGRAMS COREDUMP WITH MALLOC DEBUG

PROBLEM DESCRIPTION:
IME (zh_TW) related programs coredump with malloc debug
enabled.

PROBLEM CONCLUSION:
Change yacc program in zh_TW

------

APAR: IY25196 COMPID: 5765C3403 REL: 430
ABSTRACT: NEW WRAP PLUG FOR ATM FIBER ADAPTER DIAGNOSTICS

PROBLEM DESCRIPTION:
On an ATM Fiber adapter, if wrap plug 21H3547 is attached to
the adapter on a boot, a wire fault error may be logged in
the error log.

PROBLEM CONCLUSION:
New wrap plug 04P9438 to be used. Diagnostic code modified
to support new wrap plug.

------

APAR: IY25198 COMPID: 5765C3403 REL: 430
ABSTRACT: PROBLEMS WITH ARABIC LAYOUT ENGINE

PROBLEM DESCRIPTION:
Bidirectional Hebrew does not work in a Motif 1.2 TextField.

PROBLEM CONCLUSION:
Motif 1.2 bidi code fix.

------

APAR: IY25199 COMPID: 5765C3403 REL: 430
ABSTRACT: 622 ATM ADAPTER STOP SENDING

PROBLEM DESCRIPTION:
622 atm or refresh atm adapter failed to send

PROBLEM CONCLUSION:
change the handle allocation to round robin to minimize the
chance of reuseing the same handle immediately.

------

APAR: IY25200 COMPID: 5765C3403 REL: 430
ABSTRACT: 622 ATM ADAPTER STOP SENDING

PROBLEM DESCRIPTION:
622 atm or refresh atm adapter failed to send

PROBLEM CONCLUSION:
change the handle allocation to round robin to minimize the
chance of reuseing the same handle immediately.

------

APAR: IY25201 COMPID: 5765C3403 REL: 430
ABSTRACT: DSI WHEN ISSUING IOCTL COMMANDS TO AN INOPERABLE SCSI ADAPTER

PROBLEM DESCRIPTION:
Customers may experience a crash due to a data storage
interrupt. The problem is associated with the SCSI adapter
device driver error recovery process. If the driver marks the
adapter is in a permanent error state, any IOCTL calls to the
driver (i.e. those made by config manager or third party
applications/software) could result in this DSI.

PROBLEM CONCLUSION:
Close the window where this can occur by adding checks in the
driver code that will sleep on error recovery process
completing prior to freeing memory.

------

APAR: IY25203 COMPID: 5765C3403 REL: 430
ABSTRACT: FLOPPY DISKETTE DIAGNOSTIC APPLICATION CORE DUMPS

PROBLEM DESCRIPTION:
The Floppy Diskette Diagnostic Application may core dump
under varied use.

PROBLEM CONCLUSION:
Initialize file descriptor to an invalid value.

------

APAR: IY25207 COMPID: 5765C3403 REL: 430
ABSTRACT: NOT FORWARDING TCPIP TO FORMERLY INVALID NET ADDRESSES

PROBLEM DESCRIPTION:
A system may accept packets that are not addressed to its IP
address. This will happen only if interfaces have been
detached since the last reboot.

PROBLEM CONCLUSION:
When an interface is detached, make sure all broadcast
addresses associated with the interface are removed from
the in_addr_hash_table so the system no longer accepts
packets for those addresses.

------

APAR: IY25208 COMPID: 5765C3403 REL: 430
ABSTRACT: CHANGES TO SUPPORT FUTURE GRAPHICS ADAPTERS

PROBLEM DESCRIPTION:
Changes to support future graphics adapters.

------

APAR: IY25210 COMPID: 5765C3403 REL: 430
ABSTRACT: GREP ERRORS IN BI_MAIN

PROBLEM DESCRIPTION:
Due to incorrect grep statements, the ordering of disks
in the bosinst.data file may not be optimal.

PROBLEM CONCLUSION:
Correct the grep statements (2 instances) to have a : (colon)
after the disk name.

------

APAR: IY25211 COMPID: 5765C3403 REL: 430
ABSTRACT: LIBX11 CORE DUMPING DOWN IN XCLOSEIM

PROBLEM DESCRIPTION:
In rare cases, applications may core dump in XCloseIM.

PROBLEM CONCLUSION:
Add a simple fix to check for a null pointer.

------

APAR: IY25212 COMPID: 5765C3403 REL: 430
ABSTRACT: ETC_IMAGE RENDERS UPSIDE DOWN ON GXT800M WITH XDWA.

PROBLEM DESCRIPTION:
With a graPHIGS based application and a GXT800M
adapter, blits can be rendered incorrectly if
zoomed or scaled. Image may render inverted or
not rendered at all.

PROBLEM CONCLUSION:
modified the pixmap algorithm to handle
larger pixmaps

------

APAR: IY25218 COMPID: 5765C3403 REL: 430
ABSTRACT: 64BIT APP DOESN'T WORK ON RAMDISK IN 32BIT KERNEL

PROBLEM DESCRIPTION:
Ramdisk doesn't work on 64-bit apps in 32bit kernel

PROBLEM CONCLUSION:
Ramdisk driver must indicate that it is capable of
supporting 64bit applications in both AIX 4.3 and 5.1
Setting DEV_64BIT bit for the device before calling
devswadd().

------

APAR: IY25230 COMPID: 5765C3403 REL: 430
ABSTRACT: UNABLE TO LOAD CATIAV5

PROBLEM DESCRIPTION:
Sometimes, programs that depend on a library that depends on
a system call that is in a kernel extension will dump core
when that system call is invoked, if the main program is not
also dependent on that kernel extension.

PROBLEM CONCLUSION:
Scan dependent libraries for newer /unix dependencies when
satisfying dependencies prior to loading

------

APAR: IY25231 COMPID: 5765C3403 REL: 430
ABSTRACT: MODULE TYPE "UR" SHOULD ONLY BE CHECKED FOR MAIN PROGRAM

PROBLEM DESCRIPTION:
procp->p_flag will get set incorrectly.

PROBLEM CONCLUSION:
check/update p_flag based on main executables' modtype and
will not do this for dependent modules at all.

------

APAR: IY25244 COMPID: 5765B9501 REL: 330
ABSTRACT: NON-IBM NFS CLIENT LOOPING DOING READDIR AFTER ALL ENTRIES READ

PROBLEM DESCRIPTION:
When AIX NFS code does an rfs3_readdirplus, it calls the vnode
operation VNOP_READDIR_EOFP. For filesystems that have the flag
GFS_VERSION421 set, this will result in a call to the
VN_READDIR_EOFP function for that filesystem.
Passed into readdir_eofp is a vnode pointer, a uio structure, a
pointer to an integer (eofp) and a credential pointer. The
readdir_eofp function is supposed to mark if end-of-directory
has been reached, by setting the integer pointed to by "eofp"
to a non-0 value.
If the GFS flag GFS_VERSION421 is not set, the lfs code will try
to be clever and guess if EOF has been reached or not -- if the
underlying readdir function doesn't return an error, but didn't
read anything, it assumes it reached end-of-directory.
However, if GFS_VERSION421 is set, the NFS code checks to see if
any data has been read. If it hasn't, and end-of-file has not
been reached (iseof is 0), then it assumes that no entries could
have been read because they didn't fit in the buffer, not that
there's nothing to read.
GPFS does not seem to be setting this eofp pointer properly,
which is resulting in "NFS3ERR_TOOSMALL" errors that cause
some user applications to fail.

PROBLEM SUMMARY:
NFS clients using the GFS_VERSION421 flag can cause a loop
when reading directories in GPFS.

PROBLEM CONCLUSION:
Set the EOF flag correctly in the VNOP_READDIR_EOFP call.

------

APAR: IY25253 COMPID: 5765C3403 REL: 430
ABSTRACT: CSH IS COREDUMPING

PROBLEM DESCRIPTION:
Continuous growth in csh's process size, finally resulting
in core dump.

PROBLEM CONCLUSION:
Fixing unfreed memory references.

------

APAR: IY25254 COMPID: 5765C3403 REL: 430
ABSTRACT: CSH DUMPS CORE FOR BUILTIN ECHO COMMAND.

PROBLEM DESCRIPTION:
csh when run on a simple shell script fails due to memory
leak.

------

APAR: IY25256 COMPID: 5765C3403 REL: 430
ABSTRACT: 64-BIT BINARY TRACE FILES WON'T REPLAY IN SOME CASES

PROBLEM DESCRIPTION:
ZAPdbReplay fails in 64-bit mode if multiple traces and
replays are attempted within the same ZAPdb session.
ZAPdbReplay causes seg. fault or runs out of memory in certain
cases.

PROBLEM CONCLUSION:
Added fix for multiple trace and replay in 64-bit mode. Added
dynamic memory allocation for replay tool. Fixed bug which
corrupted binary trace file in certain circumstances.

------

APAR: IY25270 COMPID: 5765C3403 REL: 430
ABSTRACT: RAID DIAGNOSTICS DISPLAYS MENU GOAL FOLLOWING SRN

PROBLEM DESCRIPTION:
- Running Service Aids to Analyze Internal Log has user input
  problems.
- Diagnostics application does not callout an SRN when a
  SCSI_ARRAY_ERR9 is logged in the error log.
- RAID Diagnostics displays a menu goal follow by an SRN.
- Diagnostics ELA does not start when SCSI_ARRAY_ERR9 is
  logged into the errorlog.

PROBLEM CONCLUSION:
- Fixed the date input for the Analyze Internal Log Service
  Aid.
- Added support to callout SRN when a SCSI_ARRAY_ERR9 is
  logged along with some detail data.
- Modified the code to callout an SRN or Menugoal but not
  both.
- Add support to run Diagnostics when SCSI_ARRAY_ERR9 is
  logged.

------

APAR: IY25274 COMPID: 5765D5100 REL: 340
ABSTRACT: REQUIRED FIXES FOR PSSP R340

PROBLEM DESCRIPTION:
required fixes for pssp R340

------

APAR: IY25275 COMPID: 5765E6900 REL: 310
ABSTRACT: REQUIRED FIXES FOR LOADL R310

PROBLEM DESCRIPTION:
required fixes for LoadL R310

PROBLEM SUMMARY:
required fixes for loadl R310

PROBLEM CONCLUSION:
required fixes for loadl R310

------

APAR: IY25276 COMPID: 5765D9300 REL: 320
ABSTRACT: REQUIRED FIXES FOR PE R320

PROBLEM DESCRIPTION:
required fixes for pe R320

------

APAR: IY25277 COMPID: 5765B9500 REL: 150
ABSTRACT: REQUIRED FIXES FOR GPFS R150

PROBLEM DESCRIPTION:
required fixes for gpfs r150

PROBLEM SUMMARY:
required fixes for gpfs R150

------

APAR: IY25278 COMPID: 5765B9501 REL: 340
ABSTRACT: REQUIRED FIXES FOR GPFS R340

PROBLEM DESCRIPTION:
required fixes for gpfs 340

PROBLEM CONCLUSION:
required fixes for gpfs R340

------

APAR: IY25282 COMPID: 5765C3403 REL: 430
ABSTRACT: JAVA CORE DUMPS AFTER INSTALLING IY22549

PROBLEM DESCRIPTION:
After installing APAR IY22549 (bos.mp or bos.up 4.3.3.77), Java
1.2.2 or 1.3.0 may intermittently core dump.

PROBLEM SUMMARY:
****************************************************************
* USERS AFFECTED:
* Users of Java 1.2.2 or 1.3.0 on AIX 4.3 with the bos.mp or
* bos.up filesets at the 4.3.3.77 level.
****************************************************************
* PROBLEM DESCRIPTION:
* AIX kernel pthreads changes expose Java 1.2.2 and 1.3.0 to
* intermittent core dumps.
****************************************************************
* RECOMMENDATION:
* Reject the bos.mp or bos.up 4.3.3.77 updates until a fix for
* this problem is available.
****************************************************************

PROBLEM CONCLUSION:
Investigating changes to resolve core dump issues.

------

APAR: IY25287 COMPID: 5765C3403 REL: 430
ABSTRACT: PROCESS OR SYSTEM HANG USING WLM MEMORY HARD MAX LIMITS

PROBLEM DESCRIPTION:
Process or system hang when using WLM memory hard max limits.

PROBLEM CONCLUSION:
Allow kernel-mode faults to get a page even if the faulter is
a member of a class above its hard memory max.

------

APAR: IY25290 COMPID: 5765C3403 REL: 430
ABSTRACT: RPC.STATD QUICK MEMORY/THREAD GROWTH

PROBLEM DESCRIPTION:
Under certain circumstances, rpc.statd will spawn off a number
of threads. These are not detached, and never joined, so they
will stay around in the terminated state, causing what might be
seen as a memory leak.

PROBLEM CONCLUSION:
Detach the threads as they are created to avoid them to stay
around in the terminated state.

------

APAR: IY25315 COMPID: 5765C3403 REL: 430
ABSTRACT: RENAME A FILE/FOLDER TO DIFFERENT CASE FAILS

PROBLEM DESCRIPTION:
Unable to move a file/folder name to same name but different
case on FastConnect.

PROBLEM CONCLUSION:
Fixed the file/folder move functionality.

------

APAR: IY25316 COMPID: 5765C3403 REL: 430
ABSTRACT: FILES WITH .BMP CANNOT BE OPENED FROM CLIENTS

PROBLEM DESCRIPTION:
edit the file with .bmp extension and it fails to edit.

PROBLEM CONCLUSION:
enable the sharing for .bmp file.

------

APAR: IY25317 COMPID: 5765C3403 REL: 430
ABSTRACT: SELECT FILE PROPERTIES CHANGES DATE TO 2497

PROBLEM DESCRIPTION:
Timestamp is wrong

PROBLEM CONCLUSION:
Checking if all the timestamp bits are set or not

------

APAR: IY25322 COMPID: 5765D5101 REL: 121
ABSTRACT: HAGS COLONY D/S SUPPORT FOR HACMP

PROBLEM DESCRIPTION:
hags colony d/s support for hacmp

PROBLEM SUMMARY:
group services provides a set of adaptre member
ships where each adapter group represents a set of working same
type of adapters. For example, enMembership will report any
ethernet adapters which are monitored by Topology Services.
Similarly csddRawMembership is for the switch adapters.
HACMP/ES is the main user of those adapter groups. When the new
SP switch sybsystem provides more that one adapter per node,
Group Services mistakenly represents them by two separate
groups; cssRawMembership and css1RawMembership which causes
a concern in HACMP/ES because HACMP/ES expects that they
should be in a group.

PROBLEM CONCLUSION:
This fix will merge two cssRawMembership
groups into one cssRawMembership so that MACMP/ES can
subscribe only csdRarMembership to monitor the all switch
adapters.
Other globalized switch groups (cssMembership and css1Membership
) will remain the same.

------

APAR: IY25364 COMPID: 5765B9501 REL: 320
ABSTRACT: NON-IBM NFS CLIENT LOOPING DOING READDIR AFTER ALL ENTRIES READ

PROBLEM DESCRIPTION:
When AIX NFS code does an rfs3_readdirplus, it calls the vnode
operation VNOP_READDIR_EOFP. For filesystems that have the flag
GFS_VERSION421 set, this will result in a call to the
VN_READDIR_EOFP function for that filesystem.
Passed into readdir_eofp is a vnode pointer, a uio structure, a
pointer to an integer (eofp) and a credential pointer. The
readdir_eofp function is supposed to mark if end-of-directory
has been reached, by setting the integer pointed to by "eofp"
to a non-0 value.
If the GFS flag GFS_VERSION421 is not set, the lfs code will try
to be clever and guess if EOF has been reached or not -- if the
underlying readdir function doesn't return an error, but didn't
read anything, it assumes it reached end-of-directory.
However, if GFS_VERSION421 is set, the NFS code checks to see if
any data has been read. If it hasn't, and end-of-file has not
been reached (iseof is 0), then it assumes that no entries could
have been read because they didn't fit in the buffer, not that
there's nothing to read.
GPFS does not seem to be setting this eofp pointer properly,
which is resulting in "NFS3ERR_TOOSMALL" errors that cause
some user applications to fail.

PROBLEM SUMMARY:
NFS clients using the GFS_VERSION421 flag can cause a loop
when reading directories in GPFS.

PROBLEM CONCLUSION:
Set the EOF flag correctly in the VNOP_READDIR_EOFP call.

------

APAR: IY25365 COMPID: 5765E0200 REL: 711
ABSTRACT: WRONG OUTPUT AFTER COMPILATION WITH -O

PROBLEM DESCRIPTION:
Wrong output value after compilation with -O

PROBLEM SUMMARY:
Incorrect results with -O.

PROBLEM CONCLUSION:
This APAR has been fixed in the xlf ptf with
 version stamp 7.1.1.1.

------

APAR: IY25395 COMPID: 5765E0200 REL: 711
ABSTRACT: CORRUPTION FO NESTED SUBROUTINE PARAMETER

PROBLEM DESCRIPTION:

PROBLEM SUMMARY:
Corruption of a derived type internal subroutin
e argument.

PROBLEM CONCLUSION:
This APAR has been fixed in the xlf ptf with
 version stamp 7.1.1.1.

------

APAR: IY25398 COMPID: 5765B9501 REL: 320
ABSTRACT: SIGSEGV IN MERGE_FOR_STEAL

PROBLEM DESCRIPTION:
sigsegv in merge_for_steal

PROBLEM SUMMARY:
GPFS took a segmentation violation in merge_for_steal.

PROBLEM CONCLUSION:
Since the beingStolen flag cannot be set when querying
whether a buffer is stealable, the paMutex cannot be
released while the type specific steal callback routine is
called

------

APAR: IY25399 COMPID: 5765B9501 REL: 330
ABSTRACT: SIGSEGV IN MERGE_FOR_STEAL

PROBLEM DESCRIPTION:
sigsegv in merge_for_steal

PROBLEM SUMMARY:
GPFS took a segmentation violation in merge_for_steal.

PROBLEM CONCLUSION:
Since the beingStolen flag cannot be set when querying
whether a buffer is stealable, the paMutex cannot be
released while the type specific steal callback routine is
called

------

APAR: IY25400 COMPID: 5765B9501 REL: 320
ABSTRACT: 128WAY ASSERT:POS.DATABLOCKNUM==BLOCKTOEXPAND, /MNODE.C

PROBLEM DESCRIPTION:
128way assert:pos.datablocknum==blocktoexpand,/mnode.c

PROBLEM SUMMARY:
GPFS self check logic asserted at mnode.C line 3224

PROBLEM CONCLUSION:
Do expand from end of fragment instead from filesize in the
case that the fragment is shorter than the filesize.

------

APAR: IY25401 COMPID: 5765B9501 REL: 330
ABSTRACT: 128WAY ASSERT:POS.DATABLOCKNUM==BLOCKTOEXPAND, /MNODE.C

PROBLEM DESCRIPTION:
128way assert:pos.datablocknum==blocktoexpand,/mnode.c

PROBLEM SUMMARY:
GPFS self check logic asserted at mnode.C line 3224

PROBLEM CONCLUSION:
Do expand from end of fragment instead from filesize in the
case that the fragment is shorter than the filesize.

------

APAR: IY25411 COMPID: 5765E0200 REL: 711
ABSTRACT: ENDLESS COMPILATION WITH XLF 7.1.0.2

PROBLEM DESCRIPTION:
endless compilation. no way to see the end of the compile of a
3kbytes fortran code with 64bit compiler option. ok if reverting
 to 32. compiler looks taking more and more memory little by
little.

PROBLEM SUMMARY:
Compilation hung with -O3 -q64.

PROBLEM CONCLUSION:
This APAR has been fixed in the xlf ptf with
 version stamp 7.1.1.1.

------

APAR: IY25432 COMPID: 5765E6110 REL: 220
ABSTRACT: HAGS COLONY D/S SUPPORT FOR HACMP

PROBLEM DESCRIPTION:
hags colony d/s support for hacmp

PROBLEM SUMMARY:
group services provides a set of adaptre member
ships where each adapter group represents a set of working same
type of adapters. For example, enMembership will report any
ethernet adapters which are monitored by Topology Services.
Similarly csddRawMembership is for the switch adapters.
HACMP/ES is the main user of those adapter groups. When the new
SP switch sybsystem provides more that one adapter per node,
Group Services mistakenly represents them by two separate
groups; cssRawMembership and css1RawMembership which causes
a concern in HACMP/ES because HACMP/ES expects that they
should be in a group.

PROBLEM CONCLUSION:
This fix will merge two cssRawMembership
groups into one cssRawMembership so that MACMP/ES can
subscribe only csdRarMembership to monitor the all switch
adapters.
Other globalized switch groups (cssMembership and css1Membership
) will remain the same.

------

APAR: IY25433 COMPID: 5765E6110 REL: 220
ABSTRACT: INCORRECT GRPGLSM'S CSS (0!1) MEMBERSHIP

PROBLEM DESCRIPTION:
incorrect grpglsm's css(0!1) membership

PROBLEM SUMMARY:
Due to the incorrect evaluation of adapter status
in the HAGS switch support daemon (HAGSGLSM),
the switch memberships report the incorrect
result at certain situations.
For example, if two IP aliases are configured
for a switch adapter (e.g., css0), and the first
IP interface is down, the cssMembership group
may not include that adapter even if the second
IP interface is up.
With this fix, a switch adapter will be always
a member of the associated switch membership
as long as one of IP aliases is up.

PROBLEM CONCLUSION:
The globalized switch memberships (cssMembership and
css0Membership)
with alias network configuration in HACMP should be
based on one of adapter status between aliases.
It is because two aliases are mapped to one physical
adapter.

------

APAR: IY25434 COMPID: 5765D5100 REL: 320
ABSTRACT: REACQUIRE WIN LOCK AFTER CLOSE_WIN IN UNLOAD_ST

PROBLEM DESCRIPTION:
reacquire win lock ater close_win in unload-st

PROBLEM SUMMARY:
Need to reacquire win lock after close_win in UNLOAD_ST
otherwise might fail in multiple lock frees.

PROBLEM CONCLUSION:
Have reacquire win lock after close_win in UNLOAD_ST
so it would not fail in multiple lock frees.

------

APAR: IY25435 COMPID: 5765D5100 REL: 320
ABSTRACT: LOADL RUNNING GANG DOES NOT HANDLE US WINDOW UNLOAD FAILURE

PROBLEM DESCRIPTION:
loadl running gang does not handle us window unload failure

PROBLEM SUMMARY:
The new UNLOAD_ST_LL IOCTL did not exist
in the SP switch environment.

PROBLEM CONCLUSION:
The new UNLOAD_ST_LL IOCTL now exists
in the SP switch environment.

------

APAR: IY25438 COMPID: 5765B9501 REL: 330
ABSTRACT: NODE PANICKED WITH MMFLUSHSEG AND MMAP IN STACK

PROBLEM DESCRIPTION:
node panicked with mmflushseg and mmap in stack

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_d under disable_lock since the gnode might be
paged out.

------

APAR: IY25450 COMPID: 5765B9501 REL: 330
ABSTRACT: MMFSCK STUCK WITH ASSERTS

PROBLEM DESCRIPTION:
mmfsck stuck with asserts

PROBLEM SUMMARY:
gpfs self check logic failed running fsck

PROBLEM CONCLUSION:
ninodes is not necessarily a multiply
of 32 and therefore fsck must handle condition

------

APAR: IY25452 COMPID: 5765B9501 REL: 330
ABSTRACT: ASSERT !"KERNEL REQUESTED TO DIE" GPFSOPS.C RC 0

PROBLEM DESCRIPTION:
assert!"kernel requested to die" gpfsops.c rc 0

PROBLEM SUMMARY:
gpfs panic'd a node while recovering another
node failure under extreme stress loads

PROBLEM CONCLUSION:
handle the rc==0 that may result from
kxretryrecover following a node failure

------

APAR: IY25453 COMPID: 5765B9501 REL: 330
ABSTRACT: ADD CHECKS OF SHARED SEGMENT STATUS DURING CLEANUP

PROBLEM DESCRIPTION:
add checks of shared segment status during cleanup

PROBLEM SUMMARY:
additional debug information collected

PROBLEM CONCLUSION:
additional debug information collected

------

APAR: IY25454 COMPID: 5765B9501 REL: 330
ABSTRACT: LONG WAITERS:FILE BLOCK WRITE 'WAIT FOR BUFFER FOR FETCH'

PROBLEM DESCRIPTION:
long waiters:file block write 'wait for buffer for fetch'

PROBLEM SUMMARY:
gpfs recovery went into a loop after
handling a number of disk failures.

PROBLEM CONCLUSION:
fix wait loop to check for forcable
unmounted fs.

------

APAR: IY25455 COMPID: 5765B9501 REL: 330
ABSTRACT: SIGSERV IN ROUNDROBINSTRIPEMETHOD::RESELECTDISK

PROBLEM DESCRIPTION:
sigsegv in roundrobinstripemethod::reselectdisk

PROBLEM SUMMARY:
Segmentation violation in GPFS daemon

PROBLEM CONCLUSION:
Correct use of uninitialized disk index

------

APAR: IY25456 COMPID: 5765B9501 REL: 330
ABSTRACT: RUNTSCHECKQUOTA:REPLY.ERR=22 ON DISK FULL FILESYSTEM

PROBLEM DESCRIPTION:
runtscheckquota:reply.err=22 on disk full filesystem

PROBLEM SUMMARY:
Invalid argument error returned from mmcheckquota when run
on an almost full file system.

PROBLEM CONCLUSION:
 Do not terminate checkquota command if received reply.err
22 on quotaMsgRelinquish message reply.

------

APAR: IY25457 COMPID: 5765B9501 REL: 330
ABSTRACT: LOGSHUTDOWN SYNCHRONIZATION

PROBLEM DESCRIPTION:
logshutdown synchronization

PROBLEM CONCLUSION:
Additional debug information

------

APAR: IY25458 COMPID: 5765B9501 REL: 330
ABSTRACT: 128WAY:FILE SYSTEM MANAGER ASSERTS DURING ONLINE FSCK

PROBLEM DESCRIPTION:
128way:file system manager asserts during online fsck

PROBLEM SUMMARY:
GPFS self check logic detected an error at
dSynch.C, line 1454

PROBLEM CONCLUSION:
Ensure StopAllocMsg is not incorrectly left on at a node,
only to have that node become the stripe group manager
and attempt to honor it.

------

APAR: IY25474 COMPID: 5765B9501 REL: 330
ABSTRACT: HUNG THREAD IN 'WAIT FOR BUFFER PIN OR UNPIN'

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

PROBLEM SUMMARY:
Deadlock when running GPFS understress.

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

------

APAR: IY25482 COMPID: 5765B8100 REL: 220
ABSTRACT: CASEACTION CANNOT COMPARE TO NULL

PROBLEM DESCRIPTION:
Using ASCII State Table import, the CaseAction statement
is not imported correctly if the compare is to NULL

PROBLEM SUMMARY:
Using ASCII State Table import, the CaseAction
statement is not imported correctly if the compare is to NULL

PROBLEM CONCLUSION:
The problem was fixed by added some lines to
the STEP routines that checks for the presence of a null in
state table import - and if it finds one - then it changes it
to be have double quotes instead.

------

APAR: IY25488 COMPID: 5765B9501 REL: 330
ABSTRACT: HOLDDAEMONSEG IN KXRECLOCKRESET

PROBLEM DESCRIPTION:
holddaemonseg in kxreclockreset

PROBLEM SUMMARY:
incorrect lock sequence found during
code review.

PROBLEM CONCLUSION:
need holddaemonseg in kxreclockreset

------

APAR: IY25493 COMPID: 5765B9500 REL: 130
ABSTRACT: SWITCH TABLE DOES NOT CLEAR

PROBLEM DESCRIPTION:
Switch table window is not released and cleared for follow
on jobs to use.

PROBLEM SUMMARY:
Switch table window is not released and cleared by the hung
task for follow on jobs to use.

PROBLEM CONCLUSION:
The switch table window will now free the resource that is
held by the hung task so that follow on jobs will be able to
start.

------

APAR: IY25693 COMPID: 5765B8100 REL: 220
ABSTRACT: ADD CONNECT_ACK MESSAGE ISDN TO ALLOW CONNECTIVITY TO SOME QSIG

PROBLEM DESCRIPTION:
Add connect_ack message to ISDN to provide symmetrical support
to Euro ISDN. This will allow connectivity to some qsig
switches.

PROBLEM SUMMARY:
Add connect_ack message to ISDN to provide
symmetrical support to Euro ISDN. This will allow
connectivity to some qsig switches.

PROBLEM CONCLUSION:
Changed logic in plygsxs1.c to correctly
check incoming cause codes

------

APAR: IY25710 COMPID: 5765E0300 REL: 711
ABSTRACT: INTERNAL APAR FOR XLF RUNTIME PTF 7.1.1.1.

PROBLEM DESCRIPTION:
Internal APAR for XLF Runtime PTF 7.1.1.1.

PROBLEM CONCLUSION:
This APAR has been fixed in the xlf ptf with
 version stamp 7.1.1.1.

------

APAR: IY25711 COMPID: 5765E0300 REL: 711
ABSTRACT: INTERNAL APAR FOR XLF RUNTIME FOR AIX51 PTF 7.1.1.1.

PROBLEM DESCRIPTION:
Internal APAR for XLF Runtime for AIX51 PTF 7.1.1.1

PROBLEM SUMMARY:
Internal APAR for XLF Runtime for AIX51 PTF 7.1
.1.1.

PROBLEM CONCLUSION:
This APAR has been fixed in the xlf ptf with
 version stamp 7.1.1.1.

------

APAR: IY25712 COMPID: 5765E0300 REL: 711
ABSTRACT: INTERNAL APAR FOR XLF RUNTIME FOR AIX43 PTF 7.1.1.1.

PROBLEM DESCRIPTION:
Internal APAR for XLF Runtime for AIX43 PTF 7.1.1.1.

PROBLEM SUMMARY:
Internal APAR for XLF Runtime for AIX43 PTF 7.1
.1.1.

PROBLEM CONCLUSION:
This APAR has been fixed in the xlf ptf with
 version stamp 7.1.1.1.

------

APAR: IY25714 COMPID: 5765E0300 REL: 711
ABSTRACT: INTERNAL APAR FOR XLF RUNTIME FOR AIX41 PTF 7.1.1.1.

PROBLEM DESCRIPTION:
Internal APAR for XLF Runtime for AIX41 PTF 7.1.1.1.

PROBLEM SUMMARY:
Internal APAR for XLF Runtime for AIX41 PTF 7.1
.1.1.

PROBLEM CONCLUSION:
This APAR has been fixed in the xlf ptf with
 version stamp 7.1.1.1.

------

APAR: IY25807 COMPID: 5765C3403 REL: 430
ABSTRACT: MULTIPLE ARRAY CONFIG CHANGES TO SINGLE DISK

PROBLEM DESCRIPTION:
After a redistribute LUNs operation, or other path switch
event in Dual Active, a LUN may experience multiple config
change logged events, then fail with an I/O error.

PROBLEM CONCLUSION:
Double-check for pending switch events in fcparray_start

------

APAR: IY25988 COMPID: 5765D5100 REL: 320
ABSTRACT: LATEST PSSP 3.2.0 FIXES AS OF NOVEMBER 2001

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

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

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

------

APAR: IY25991 COMPID: 5765B8100 REL: 220
ABSTRACT: DIRTALK 3270 SESSIONS WITHOUT INITIAL SCRIPT WON'T GO READY

PROBLEM DESCRIPTION:
3270 Sessions without an initial script won't go ready.

PROBLEM SUMMARY:
3270 Sessions without an initial script won't
go ready.

PROBLEM CONCLUSION:
Removed check for disconnect_3270_ps failing
 release_3270_session

------

APAR: IY25997 COMPID: 5765B8100 REL: 220
ABSTRACT: IBM_TROMBONE DROPS INCORRECT CALL UNDER LOAD

PROBLEM DESCRIPTION:
Under load the IBM_Trombone custom server can drop the
wrong call causing many IBM_Trombone_Custom_Server errors.

PROBLEM SUMMARY:
Under load the IBM_Trombone custom server can
drop the wrong call causing many IBM_Trombone_Custom_Server
errors.

PROBLEM CONCLUSION:
Changed IBM_Trombone_CS_function.c to only
request the outbound state table to close down if the Close
Host Server Link came from the incoming call state table.

------

APAR: IY26014 COMPID: 5765D5100 REL: 340
ABSTRACT: LATEST PSSP 3.4.0 FIXES AS OF NOVEMBER 2001

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

PROBLEM SUMMARY:
This is a packaging apar for PSSP 3.4.0 fixes
as of November 2001.

PROBLEM CONCLUSION:
This is a packaging apar for PSSP 3.4.0
fixes as of November 2001.

------

APAR: IY26024 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK 3270 V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY26058 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK SS7_ISUP V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY26059 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK ISDN.COM V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY26060 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK ISDN.ATT V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY26061 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK ISDN.DMS100 V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY26062 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK ISDN.EURO V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY26063 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK ISDN.INS1500 V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY26064 COMPID: 5765B8100 REL: 230
ABSTRACT: INITIAL UPDATE TO DIRECTTALK BROOKTROUT V2.3 FOR GA CODE

PROBLEM DESCRIPTION:
This APAR is a collection of fixes to update the gold code at GA
time.

PROBLEM SUMMARY:
Initial update to Gold Code

------

APAR: IY26069 COMPID: 5765B8100 REL: 220
ABSTRACT: CALL DURATION IN CDR.LOG IS SOMETIMES INCORRECT

PROBLEM DESCRIPTION:
Sometimes the call duration shown in the cdr.log is incorrect.

PROBLEM SUMMARY:
Sometimes the call duration shown in the
cdr.log is incorrect.

PROBLEM CONCLUSION:
Correct method of calculating call duration

------

APAR: IY26166 COMPID: 5765B9501 REL: 330
ABSTRACT: THIRD PARTY SOFTWARE (CLEARCASE) NON-STANDARD USE OF THE VNOP-

PROBLEM DESCRIPTION:
A Third party program called VNOP_OPEN without passing a valid
pointer (vinfo). This led to a node crash due to GPFS kernel
panic in GPFS_V_OPEN.

LOCAL FIX:
Allow open without a vinfo pointer.

PROBLEM SUMMARY:
A third party distributed file system called a GPFS entry
point with an invalid pointer resulting in a GPFS panic.

PROBLEM CONCLUSION:
Allow open without a vinfo pointer. If this occurs, close
the file after successfully opening it.

------

APAR: IY26212 COMPID: 5648C9802 REL: 430
ABSTRACT: SDK 1.3.0 PTF 8A : CA130-20010925MQ

PROBLEM DESCRIPTION:
Fixes since PTF 8 (ca130-20010925) :
(Note: The descriptions here have been truncated.)
+----------------+-----+--------------------------------------+
|ca130-20010925mq|38712|JVM garbage-collects live objects |
+----------------+-----+--------------------------------------+

------

APAR: IY26222 COMPID: 5697F6400 REL: 640
ABSTRACT: REPACKED MESSAGE CENTER WEB INTERFACE ERROR 404

PROBLEM DESCRIPTION:
Message Center Web interface error 404
when going to Remote message settings from
Telephone interface settings (but from all other menus it is OK)
This is due to tellS.jsp going to remMs.jsp instead of

PROBLEM SUMMARY:
Error 404 when going from Telephone Interface
settings to Remote message settings.

PROBLEM CONCLUSION:
Modified mispelled link from remMs.jsp to
remMS.jsp in telIS.jsp

------

APAR: IY26233 COMPID: 5765B9501 REL: 330
ABSTRACT: NODE PANICKED:KSYNCH.C:1252

PROBLEM DESCRIPTION:
node panicked:ksynch.c:1252

PROBLEM SUMMARY:
GPFS panic'd a node under a stress load.

PROBLEM CONCLUSION:
Be sure to check return code from acquireCacheObjMutex so
that exit does not try and release a mutex it does not hold.

------

APAR: IY26234 COMPID: 5765B9501 REL: 330
ABSTRACT: IMPROVE MULTITHREAD RANDOM IO PERFORMANCE

PROBLEM DESCRIPTION:
improve multithread random io performance

PROBLEM SUMMARY:
Random I/O performance improvement

PROBLEM CONCLUSION:
Drop vinfo lock during fetch for random reads/writes.
Improves parallelism when multiple threads share the same
file handle.

------

APAR: IY26237 COMPID: 5765B9501 REL: 330
ABSTRACT: LONG WAITERS - 'WAITING FOR THE FLUSH FLAG'

PROBLEM DESCRIPTION:
long waiters-'waiting for the flush flag'

PROBLEM SUMMARY:
Deadlock in GPFS under heavy load.

PROBLEM CONCLUSION:
StealOpenFile needs to make a better test for
fileDirtyOrUncommitted when stealing to reduce the cache
down to minimum size.

------

APAR: IY26242 COMPID: 5639I3700 REL: 310
ABSTRACT: DCE310 MAINTENANCE PACKAGE FOR PTF SET 5

PROBLEM DESCRIPTION:
THIS APAR CREATED FOR MAINTENANCE PACKAGE OF PTF SET 5
dce310 Maintenance Package for PTF SET 5

PROBLEM SUMMARY:
AIXDCE310 MAINTENANCE PACKAGE FOR PTF SET 5

PROBLEM CONCLUSION:
AIXDCE310 MAINTENANCE PACKAGE FOR PTF SET 5

------

APAR: PQ54924 COMPID: 5765C4100 REL: 230
ABSTRACT: POSSIBLE WRONG ANSWERS WITH PZGELS WHEN M<N AND TRANS = C.

PROBLEM DESCRIPTION:
PZGELS may produce wrong answers when M<N and TRANS=C.

LOCAL FIX:
Install PTF.

PROBLEM SUMMARY:
PZGELS may produce wrong answers when M<N and
TRANS = C.

PROBLEM CONCLUSION:
PZGELS fixed internally.

------

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