PM81536: z/TPFDF should ensure that the correct file ID exists in pool records before releasing those records.

APAR status

• Closed as program error.

Error description

• See Problem Summary.
  

Local fix

• na
  

Problem summary

• APAR NUMBER:  PM81536
  PRODUCT:  z/TPFDF
  FUNCTIONAL AREA:  TPFDF CENTRAL DATABASE ROUTINES
  SHIPPED IN PUT:  10
  
  ABSTRACT:
  z/TPFDF should ensure that the correct file ID exists in pool
  records before releasing those records. Also, when a DBRST is
  issued after an LLR was deleted from the prime block, multiple
  releases of the LLR or a DB0168 may occur.
  
  PACKAGE CONTENTS:
  Source Segments:
  (C) tpfdf/include/i_dfhd.h
  (C) tpfdf/rt/uas0.asm
  (C) tpfdf/rt/ub90.asm
  (C) tpfdf/rt/ufka00.cpp
  (C) tpfdf/rt/ufka03.cpp
  (C) tpfdf/rt/uftk.cpp
  
  Object Only Binaries:
  None.
  
  Configuration Independent Binaries:
  None.
  
  Support Files:
  None.
  
  OTHER BINARIES TO BUILD: YES
  (C) <sys>/lib/libUTDF.so
  (C) <sys>/load/UFKA.so
  (C) <sys>/load/UFTH.so
  (C) <sys>/load/UTDF.so
  (C) <sys>/obj/uas0.o
  (C) <sys>/obj/ub90.o
  (C) <sys>/obj/ufka00.o
  (C) <sys>/obj/ufka03.o
  (C) <sys>/obj/uftk.o
  
  COMMENTS:
  (1) z/TPFDF should ensure the correct file ID exists in a pool
  record before it releases that pool file address. In most parts
  of z/TPFDF, the block has already been accessed and its ID has
  been verified before z/TPFDF issues a RELFC on its file
  address. Therefore, an additional ID check on the RELFC API is
  not needed in those cases. However, there are some routines in
  large logical record (LLR) and B+Tree processing where file
  addresses are released without accessing them first. These
  routines should be enhanced to validate the file ID before
  releasing the records in case those records are in use by
  another database.
  
  (2) Furthermore, a DB0100 system error is generated when an I/O
  error occurs when z/TPFDF attempts to retrieve a record.
  However, the online message does not indicate which I/O error
  occurred. It would be helpful if the online message indicated
  which specific I/O error occurred.
  
  (3) In addition, if there was an LLR in the prime block that
  had previously been deleted and a DBRST is done, the blocks in
  the LLR could be released more than once. This is because DBRST
  copies a possible user header part of the "from" block to the
  "target" block. An incorrect length was used in the
  calculation, so the z/TPFDF header component of the standard
  header is copied to the target block, potentially overwriting
  the next available byte (NAB) value. As a result, a possibly
  inactive LLR is activated again and either generates a DB0168
  if an attempt is made to re-access it, or reults in multiple
  releases if the LLR is re-released.
  

Problem conclusion

• SOLUTION:
  (1) Segments ufka00.cpp and ufka03.cpp have been updated so
  that LLR release processing issues a find and wait on LLIBs and
  LLDBs to ensure the file ID is valid before releasing the
  blocks either directly in function UFKA03_release_blocks() or
  asynchronously in function
  UFKA00_release_all_LLDBs_from_this_LLIB(). If an error is
  encountered, informational dump OPR-DB0189 is issued and the
  block is not released. The dump is issued only once per LLIB,
  but the loop to release additional file addresses continues.
  
  Segment uftk.cpp has been updated so that B+Tree processing
  issues a find and wait on B+Tree nodes to ensure the file ID is
  valid before releasing the blocks. If an error is encountered,
  informational dump OPR-DB0189 is issued and the block is not
  released.
  
  Header i_dfhd.h has been updated to contain a common macro to
  generate the DB0189 system error that is based on the DB0100
  system error.
  
  (2) Segment ub90.asm has been updated to add information to the
  DB0100 system error such that its content matches the DB0189
  system error.
  
  (3) Segment uas0.asm has been updated to use the correct
  displacements and lengths (taking into account both the z/TPF
  and z/TPFDF headers) in the calculation of the user header in
  the copy routines.
  
  COREQS: NO
  None.
  
  MIGRATION CONSIDERATIONS: YES
  Functional, automation, and operation changes:
  New system error: DB0189
  Modified system error: DB0100
  
  
  
  BUILD COMMANDS AND INSTRUCTIONS: YES
  #maketpf commands for linux
  maketpf -f UTDF uas0.o ub90.o
  maketpf -f UFKA ufka00.o ufka03.o
  maketpf -f UFTH uftk.o
  maketpf UTDF link
  maketpf UFKA link
  maketpf UFTH link
  
  UPDATED INFORMATION UNITS: YES
  z/TPFDF Messages (System Error, Online, Offline)
  
  See your IBM representative if you need additional information.
  
  DOWNLOAD INSTRUCTIONS:
  http://www.ibm.com/software/htp/tpf/maint/maintztpf.html
  
  APAR URL:
  http://www.ibm.com/software/htp/tpf/ztpfmaint/put10/PM81536.htm
  

Temporary fix

Comments

APAR Information

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

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

Publications Referenced

SK2T8062

Fix information

• Fixed component name

  ZTPFDF

• Fixed component ID

  5748F1501

Applicable component levels

• R110 PSY

     UP