PK33467: DFHLM0001 ABEND 0C4/AKEA AT OFFSET X'1512' IN DFHLMLM. AN 0C6 MAY ALSO OCCUR X'FFFF' IN DFHLMLM DUE TO WILD BRANCH.

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• DFHLMLM program checks with an abend 0C4 or 0C6 due to the
  kernel stack registers for the running task being corrupt.  In
  the 0C6 case, register 5 was bad, and in the 0C4 case, register
  13 was bad.  The invalid registers are captured in the dump
  registers, but may be different in the kernel stack at the time
  of the dump due to DFHMEME being in the stack due to the error.
  A CPIH pipeline transaction has recently suspended in DFHSMGF
  trying to get the SMLOCK (smlock_get) and when it wakes up and
  resumes, the registers are bad in the stack and the program
  check happens.  The registers were good at the time of the
  DSSR suspend, because DFHLMLM would have program checked
  earlier if they had been bad.  So the overlay of registers
  happens somewhere after the CPIH task suspends and when it
  runs again.  The value that is overlaying the stack is also
  in a pipeline control block called the Service Handler
  Element (SNE) at offset x'40'.  And at offset x'44' into the
  SNE (Service Node Element, also referred to as the Service
  Handler Element) it is pointing into the kernel stack that
  gets corrupted.  Note that there are other CPIH transactions
  running at the time, and some are on L8 (open) TCBs.
  This APAR is being taken for further investigation of the
  problem by the CICS Change Team.
  .
  Additional Keywords: ABENDS0C4 ABEND0C4 ABENDS0C6 ABEND0C6
  DFHPISNE DFHPISN2 LMQUICK1 PI_GENRL subpool
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All CICS users.                              *
  ****************************************************************
  * PROBLEM DESCRIPTION: MSGDFHLM0001 An abend (code 0C4/AKEA)   *
  *                      has occurred in module DFHLMLM after    *
  *                      processing an inbound web services      *
  *                      request.                                *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  Multiple web service requests for the same provider pipeline are
  executing at the same time.  Each one has its own L8 TCB.
  Eventually they get to the soap node in the pipeline and DFHPISN
  is called.  DFHPISN needs to match up <headerprogram> elements
  in the pipeline configuration file with the actual SOAP header
  elements in the incoming message.
  
  The configuration file was processed at PIPELINE install time
  and a chain of Header Program Element (HPE) control blocks were
  created.  The chain is anchored off the associated Service Node
  Element (SNE).
  
  To browse the chain of HPE's DFHPISN creates a local iterator
  and updates the list of iterators contained in the SNE.  Once
  the header processing is complete DFHPISN removes its iterator
  from the list.  Both of these actions are not threadsafe.
  When multiple tasks execute this same code under different L8
  TCBs the list of iterators (at offset x'40' and x'44' in the
  SNE) can become corrupted and an address of a local iterator
  (in a task's stack storage) can remain in the list after the
  associated task has ended.  When the list is manipulated
  again, by a new task adding an iterator, the stack storage
  will get corrupted.  The effect of this will vary depending
  on what the task currently owning that storage is doing.  In
  the cases seen so far register save areas were overlaid leading
  to abends in DFHLMLM.
  
  Additional keywords:
  S0C4  S0C6  unpredictable
  

Problem conclusion

• DFHPISN has been changed to obtain the exclusive PILOCK around
  the HOP_StartIterator and HOP_EndIterator calls used to process
  the header program elements.
  

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:

  UK21470

Modules/Macros

•    DESPISH  DESPISN  DFHPISN  DFHPISN1 DFHPISN2
  

Fix information

• Fixed component name

  CICSTS 3.1 Z/OS

• Fixed component ID

  5655M1500

Applicable component levels

• R400 PSY UK21470

     UP07/01/26 P F701

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.