PM85798: LOOP IN DFHAPEX OFFSET 8B0 TO 8C4 DURING TRAN DUMP PROCESSING

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• After DB2 abends, some CICS regions don't reconnect to DB2,
  but instead go into a loop. System trace shows a tight loop
  in module DFHAPEX offset x'8B0' to x'8C4' (UK76215). The
  loop occurs becaue the PLCB pointer for a GLUE program points
  to itself at +x'10' (field PLCB_PREV ).
  .
  The looping task was being abended with AD3K (an abort request
  to DB2, issued during the second phase of syncpoint failed).
  During transaction dump processing, the GLUE program was
  invoked at exit point XDUOUT. The GLUE did a WAIT_MVS request
  that timed_out. After the time-out, DFHSUEX tried to do a
  CHANGE_MODE request back to the L8 TCB and this fails since the
  L8 TCB has now abended. Since the CHANGE_MODE fails, DFHSUEX
  does not do the DFHPGEXI CLEAR_EXIT request that would remove
  the PLCB level for the GLUE program. After putting out the
  DFHAP0002 message, DFHAPEX returns a DISASTER response back to
  DFHDUXW. The XDUOUT program is now called again and this call
  returns without problems. But now the DFHPGEXI CLEAR_EXIT call
  in DFHSUEX sets the PLCB_PREV to point to itself, and then the
  next call to the XDUOUT program ends up looping in DFHAPEX.
  .
  Additional Symptom(s) Search Keyword(s):
  KIXREVSVR  CHANGE-MODE-FAILED loops
  DFHAP0002 A severe error (code X'D52F') has occurred in
   module DFHSUEX
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All CICS Users.                              *
  ****************************************************************
  * PROBLEM DESCRIPTION: Loop in DFHAPEX from offset x'8B0' to   *
  *                      x'8C4' during transaction dump          *
  *                      processing.                             *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  A user CICS region has an XDUOUT exit enabled. A task encounters
  a DB2 SQL problem and a transaction dump is requested. DFHSUEX
  is called to invoke XDUOUT and it does a change mode from its
  open TCB to QR to call this XDUOUT. The PLCB for this GLUE is
  chained to the PTA_PLCB_HEAD. After return from the call to the
  GLUE, it does a change mode from QR to the origin open TCB.
  However, the open TCB has been killed by DB2 and the task then
  encounters a change mode error. DFHSUEX is exited without
  clearing the PLCB from the PTA_PLCB_HEAD. This causes the loop
  in DFHAPEX and the system gets hung.
  Keywords: SUEX DFHUEHC DFHUEH APEX PTA LOOPING
  

Problem conclusion

• DFHSUEX has been changed to unstack the PLCB if it encounters
  the change mode error.
  DFHDUXD and DFHTRXDF have been changed to make the JOBSTEP TCB
  as the owner of the TRT trace buffer.
  

Temporary fix

• *********
              * HIPER *
              *********
  FIX AVAILABLE BY PTF ONLY
  

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

  PM84024

• APAR is sysrouted TO one or more of the following:

  UK96353

Modules/Macros

• DFHDUXD  DFHSUEX  DFHTRXDF DFHUEH
  

Fix information

• Fixed component name

  CICS TS Z/OS V5

• Fixed component ID

  5655Y0400

Applicable component levels

• R800 PSY UK96353

     UP13/08/09 P F308

Fix is available

• Select the PTF appropriate for your component level. You will be required to sign in. Distribution on physical media is not available in all countries.