OA63191: HUNG DIV I/O DUE TO ORPHAN RVR WITH OVERLAPPING RANGE

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• If an "orphan" RVR (RSM's representation of a div mapped
  range) from a prior job exists in an initiator
  address space, and a later DIV MAP causes a new RVR with a range
  that overlaps that of the orphan, a DIV I/O or segment/page
  fault on the overlapping range may hang indefinitely.
  When a VDAC ASSOCIATE is done on a range of storage, should part
  of that range reside in an invalid segment, that segment will
  simply be flag as "associate pending."  Any later process that
  causes that segment to be validated, will cause ASSOCIATE code
  to run, it will search the RVR chain for the first RVR it finds
  whose range includes the VSA(s) in that segment, and assign it
  to the newly created page table entries.  If the "orphan" RVR is
  found, it has no valid WCB pointer and as a result, any I/O for
  those pages will simply not be started with no notification.
  
  Note that this APAR does NOT address the presence of an "orphan"
  RVR, rather, it simply insures that no duplicate ranges can be
  created.
  
  KNOWN IMPACT:
  Hung address space.  In the reported case, the hung address
  space held a local lock resulting in a system outage.
  
  VERIFICATION STEPS:
  There will be two (or more) RVRs with overlapping ranges.
  There will be PCBs for the I/Os in question, however, the I/Os
  themselves will not have been started.  The VDI token will not
  point to a WCB, and PCBRVR will point to an "orphan" RVR.
  

Local fix

• BYPASS/CIRCUMVENTION:
  Recycling the initiator will remove the orphan RVR.
  
  RECOVERY ACTION:
  The hung job will need to be cancelled or forced.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * Users of HBB77B0 and up.                                     *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * HUNG DIV I/O DUE TO ORPHAN RVR WITH OVERLAPPING RANGE        *
  *                                                              *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  An RVR (RSM's representation of a div mapped range), 'orphaned'
  from a prior job, remains in an initiator address space.  A
  later DIV MAP attempt may create a new RVR, resulting in any of
  the following conditions:
  a range that overlaps the 'orphan' RVR
  a DIV I/O or segment/page fault on the overlapping range may
  hang indefinitely.
  

Problem conclusion

• The RSM Code for DIV map request is changed to force the check
  for an overlapping DIV MAP range with a previous DIV MAP request
  range and to fail the request if there is an overlap.  Also this
  APAR provides additional diagnostics to track a DIV UNMAP
  request to help in finding the root cause of the orphaned RVR
  problem.
  

Temporary fix

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

Comments

APAR Information

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

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

  UJ08700 UJ08701 UJ08702

Modules/Macros

• IAXKM    IAXKA    IAXKF    IAXKE    IAXKD    IAXKH
  

Fix information

• Fixed component name

  RSM - REAL STOR

• Fixed component ID

  5752SC1CR

Applicable component levels

• R7D0 PSY UJ08702

     UP22/07/27 P F207 ¢

• R7B0 PSY UJ08700

     UP22/07/27 P F207 ¢

• R7C0 PSY UJ08701

     UP22/07/27 P F207 ¢

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.