PK31696: SUBORDINATE ADDRESS SPACE ENDS IN 0C4 ABEND BUT MASTER ENDS RC=0

APAR status

• Closed as program error.

Error description

• Subordinate address space ends in 0C4 abend but master address
  space ends with RC=0
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: Users with IMS Database Recovery Facility    *
  *                 Version 3 Release 1 installed are affected.  *
  ****************************************************************
  * PROBLEM DESCRIPTION: Users may encounter an ABENDS0C4        *
  *                      in FRXICTL0 followed by a RC=0 in the   *
  *                      DRF Master Address Space during normal  *
  *                      termination.                            *
  *                                                              *
  *                      In addition, running DRF with           *
  *                      TAPECHK(Y) and CACHE(Y) results in a    *
  *                      ABEND40D in the master address space,   *
  *                      ABEND069 in the subordinate address     *
  *                      space, or a loop.                       *
  ****************************************************************
  * RECOMMENDATION: INSTALL CORRECTIVE SERVICE FOR APAR/PTF      *
  ****************************************************************
  During normal DRF termination, the Image Copy Restore
  Controller (RVIC) will enqueue a termination AWE notification
  to the Restore Thread (RVIR).  The IC Restore Controller will
  wait for the Restore Thread to complete termination processing
  and to return control back.  A timing issue exists where the
  Restore Thread completes termination and frees the AWE before
  the IC Restore Controller gets control back.  The IC Restore
  Controller attempts to access the AWE that is saved in R1 and
  finds it pointing at freed storage.  This is the cause of the
  ABENDS0C4.
  
  In addition, while using CACHE(Y) with TAPECHK(Y) DRF will
  attempt to allocate and unallocate the image copy data sets
  with VTS.  This proccess is faster than the allocate and
  unallocate done during recovery I/O.  This creates a timing
  issue with TAPECHK processing where it will attempt to make a
  tape device request before the CACHE process finishes
  unallocating a previous request.
  

Problem conclusion

• FRXICTL0 is change so that the AWE will not be freed upon
  return from FRXIRTH0.
  
  FRXIDYN0 and FRXMSTR0 are changed to wait for a tape device to
  be unallocated from a subordinate address space before allowing
  that address space to make another tape device request.
  

Temporary fix

Comments

• ×**** PE06/12/20 PTF IN ERROR. SEE APAR PK36561  FOR DESCRIPTION
  

APAR Information

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

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

  UK19166

Modules/Macros

• FRXICTL0 FRXIDYN0 FRXMSTR0
  

Fix information

• Fixed component name

  IMS DB RECOVERY

• Fixed component ID

  5655I4400

Applicable component levels

• R310 PSY UK19166

     UP06/10/28 P F610