PK52492: GET/RELEASE BUFFERS FROM WRONG BUFFER POOL 07/09/26 PTF PECHANGE

APAR status

• Closed as program error.

Error description

• Get/Release buffers from wrong buffer pools
  .
  This APAR PE's APAR PK24575 whose PTF is UK19165
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of IMS DRF version 3               *
  *                 release 1 might be affected.                 *
  ****************************************************************
  * PROBLEM DESCRIPTION: There are two issues in this APAR.      *
  *                      First, the recover process gets into    *
  *                      a loop while reading logs. Second,      *
  *                      error handling while getting a buffer   *
  *                      during retrieve data processing caused  *
  *                      data corruption.  The buffer error      *
  *                      also caused an abend U385-A.            *
  *                      This is a PE APAR of PK24575.           *
  ****************************************************************
  * RECOMMENDATION: INSTALL CORRECTIVE SERVICE FOR APAR/PTF      *
  ****************************************************************
  There are two loops during the recover process:
  
  The loop in FRXUORM0 was caused by the logic failing to properly
  maintain the free chain of RVUR blocks; in some cases the
  storage for an RVUR block was released, but the free chain was
  not updated to reflect this release.  This resulted in the free
  chain having a pointer to previously released storage. the chain
  had incorrect pointers that resulted in an endless loop.
  
  The loop in FRXRBUF0 was created when a log record of length 0
  was read from an incorrectly allocated log file.  The log file
  had incorrectly been allocated as having unformatted, 0-length
  records. Logic in FRXRBUF0 depended on the length of the log
  record to index through a block of log records kept in memory.
  Because the index was incremented by 0 the end of the block of
  log records was never reached.
  
  The data corruption and Abend U385-A are caused when a process
  in DRF obtains a buffer and saves it for future processing.  The
  buffer is filled with log records and sent to the UOR manager.
  The pointer to the buffer is not cleared when the buffer is
  sent.  The next set of log records is copied into the same
  buffer that is enqueued to the UOR manager.  This happens over
  and over again until the end of data is reached.  The Abend
  U385-A happens when the earlier buffer data is overlayed and
  detected by the UOR manager.  The data corruption occurs when
  the overlay is within a log record and not detected by the
  UOR manager.  The later case sends the overlayed log data to
  later processing which writes the bad data to the output
  database data set.
  

Problem conclusion

• AIDS: RIDS/UTIL RIDS/DBS DBS/UTIL
    GEN:
  
  *** END IMS KEYWORDS ***
  The following modules have been modified to fix the problems:
  
  FRXUORM0: The logic was changed to properly maintain the free
            chain of RVUR blocks.
  FRXRBUF0 and FRXRLRA0: changed to detect 0-length records and
            terminate with a message that includes the name of the
            log dataset that was incorrectly allocated.
  FRXQBUF0: Was modified to get a new buffer each time when we
            retrieve a new block of data to fix the data
            corruption and ABENDU385-A.
  FRXHBUF0: The definition of two buffer pools was modified to
            avoid buffer size mismatches as a potential problem.
            The buffer size was also corrected from a standard
            buffer size to the size of the buffer that we are
            actually getting.
  FRXLMRG0,FRXICTL0,FRXPDSR0,FRXPDSS0,FRXPSDR0,FRXPSDS0:
            These modules are modified to make sure that the
            buffers will be returned to the right buffer pools.
  FRXIOCN0: The field of input record was initialized to NIL to
            avoid the situation of presenting I/O error message
            when no error has occurred.
  FRXCAMG0: A new message was added for detecting 0-length
            records.
  FRXRBUF0 and FRXCAMG0:  MSGFRD2892I is modified to add a new
            reason:  'INVALID REC LENGTH' to the message.
  
  The User's Guide and Reference of IBM IMS Database Recovery
  Facility for z/OS has been changed (SC18-9407-02).
  
  FRD2892I needs to be modified:
  
  Reason    Indicates the reason for the error, and is one of
            the following:
            . INVALID TIME STAMP
            . INVALID REC LENGTH
  
  
  After existing Explanation for 'invalid time stamp', add the
  following:
  
  INVALID REC LENGTH   An invalid record length is detected when
            processing a record while reading from a log data set.
            The record length is zero.  This can occur when the
            log data set is created with unformatted data set
            characteristic.  DRF terminates early without
            completing the recovery.
  
  After existing User response for 'invalid time stamp', add the
  following:
  
  User response:   Examine the log data set
            identified in the message and repair it to the
            log data set characteristics which the IMS product
            supports.
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

APAR Information

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

  PK52261

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

  UK30638

Modules/Macros

• FRXBDMG0 FRXCAMG0 FRXHBUF0 FRXICTL0 FRXLMRG0
  FRXPDSR0 FRXPDSS0 FRXPSDR0 FRXPSDS0 FRXQBIN0 FRXQBUF0 FRXRBUF0
  FRXRLRA0 FRXRVGB  FRXUORM0
  

Fix information

• Fixed component name

  IMS DB RECOVERY

• Fixed component ID

  5655I4400

Applicable component levels

• R310 PSY UK30638

     UP07/10/27 P F710 «