PI11514: CICS TASKS HUNG ON SMLOCK AFTER TASK IS PURGED

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• CICS tasks hang in LMQUEUE suspends waiting for the SMLOCK.
  The owner of the SMLOCK is looping on it's L8 TCB in DFHSMGF.
  The range of the loop is very tight, DFHSMGF +x'10FC' to
  x'1100'. The problem is that the SCA +x'50' for task 1 is
  addressing task 2's SCE chain.
  .
  The problem happened after task 1 was purged while waiting for
  the SMLOCK. CICS trace shows the exception entry for DFHLMLM
  due to the purge:
  .
  LM 0047 LMLM *EXC* - TASK-REQUESTING-LOCK-HAS-BEEN-PURGED
    FUNCTION(LOCK) LOCK_TOKEN(18BA02B0) MODE(EXCLUSIVE)
  .
  The RET address for this trace entry shows that the caller of
  DFHLMLM is DFHSMMG (in the SMLOCK_GET proc). DFHSMMG issues the
  call to DFHLMLM to get the SMLOCK but does not check for a
  purged response, and neither does the caller of the SMLOCK_GET
  proc. So DFHSMMG continues on as if the call to get the lock
  was successful and the GETMAIN for task1 is not serialized
  allowing the SCEs for tasks 1 and 2 get entangled.
  .
  Additional Symptom(s) Search Keyword(s):
  KIXREVSCB ELEMCHAIN_SCAN_10SCA_ELEMHEAD SCA
  sce chain corrupt multiple tasks point to same SCE
  SCQ corruption  SCF subpool purge
  DFHPD0105  A loop has been detected. SCE at address
             already encountered
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All CICS Users                               *
  ****************************************************************
  * PROBLEM DESCRIPTION: Task suspended on SMLOCK can be purged  *
  *                      but response is not checked.            *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  A task that is waiting on SMLOCK can be force purged.  The
  response returned from Lock Manager is not correctly checked
  meaning that DFHSMMG / DFHSMMF will continue.  This can lead to
  two tasks manipulating storage and will lead to unpredictable
  results.
  

Problem conclusion

• DFHSMMG and DFHSMMF have been changed to correctly check the
  response from the LOCK calls and process accordingly.
  
  The CICS TS 4.2 Trace entries (SC34-7187-01) manual will be
  updated as follows:
  
  Chapter 43 is updated to add the following trace points:
  
  0C1C for Module DFHSMMG
  0C1D for Module DFHSMMG
  0C1E for Module DFHSMMG
  
  0D1A for Module DFHSMMF
  0D1B for Module DFHSMMF
  0D1C for Module DFHSMMF
  

Temporary fix

• FIX AVAILABLE BY PTF ONLY
  

Comments

APAR Information

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

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

  PI12864 UI17286 UI17287

Modules/Macros

• DFHAPDUF DFHAPSDF DFHPGDUF DFHSMAD  DFHSMAR  DFHSMCK  DFHSMDM
  DFHSMDUF DFHSMGF  DFHSMK7  DFHSMMCI DFHSMMC2 DFHSMMF  DFHSMMG
  DFHSMPP  DFHSMPQ  DFHSMSCP DFHSMSQ  DFHSMSR  DFHSMST  DFHSMSU
  DFHSMSVC DFHSMSY  DFHSMTRI DFHSMVN  DFHSMVP  DFHSMXDF DFHS2AD
  DFHS2AR  DFHS2GF  DFHS2MC2 DFHS2MF  DFHS2MG  DFHS2PP  DFHS2SQ
  DFHS2SR  DFHS7AD  DFHS7DM  DFHS7GF  EYUTNAGD
  

Fix information

• Fixed component name

  CICS TS Z/OS V4

• Fixed component ID

  5655S9700

Applicable component levels

• R70M PSY UI17287

     UP14/05/03 P F405

• R700 PSY UI17286

     UP14/05/03 P F405

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.