OA40883: APPLICATION ADDRESS SPACE HUNG AS TDUMP OR SVCDUMP WITH SDATA=GRSQ TARGET THIS ADDRESS SPACE. ISGDSDMP LOOP

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• When Application address space (i.e WAS) entountered unexpected
  condtion and TDUMP was requested, As SDATA=GRSQ was the dump
  option, SDUMP invokes GRS exit - ISGDSDMP to collect ENQ related
  contention information. ISGDSDMP uses a 4K buffer to communicate
  with GQSCAN service. As each buffer was returned (i.e. One
  contented resource and its asociated unit of works), ISGDSDMP
  invokes DUMP service to write out the buffer as one logical dump
  record. When number of unit of works under such contented
  resource can NOT be fit within the size of the buffer provided
  by the caller (i.e 4K buffer was provided by ISGDSDMP), GQSCAN
  will fill up the buffer with available data elements and
  returned with RC08 along with the TOKEN.
  
  Ideally, caller of GQSCAN can invoke GQSCAN with returned TOKEN
  and the buffer, so the buffer could be filled with data elements
  associated with next contented resource. Or caller can invoke
  GQSCAN with QUIT=YES along with the TOKEN  to stop this GQSCAN.
  Changes introduced in GQSCAN for R13, GQSCAN did not update the
  pointer for the next Contented resource when user provided
  buffer was NOT long enough to hold up all of the data elements
  associated with this resource. When the caller uses the TOKEN to
  acquire subsequence data from GQSCAN, GQSCAN keeps returning the
  same data elements as before along with RC08. ISGDSDMP keeps
  writing out the buffer then going back to GQSCAN. ISGDSDMP would
  NOT stop its GQSCAN invocation until dump dataset result of
  space in TDUMP or SDUMP reaches MAXSPACE in SVCDUMP.
  
  It is possible for other callers (besides ISGDSDMP) to encounter
  this issue!
  
  Impact
  
  High CPU until the loop is terminated.  System work may back up.
  
  Verification Steps
  
  If GRS ctrace options FLOW (FLOW1 or FLOW9) or REQUEST
  (REQUEST2) are active then ctrace entry 0000044 for entry point
  ISGQSCG will contain a RC 00000008 along with the ascb and tcb
  of the caller.
  
  If GRS ctrace is NOT active or GRS is not included in the dump,
  then the GQSCAN requests can be seen in the system trace table.
  Home will be the callers address space.
  
  If the caller is an application space (not ISGDSDMP) the GQSCANs
  maybe one after the other in the system trace table.  However,
  if the calling application is doing a lot of activity with the
  data return the loop may be difficult or impossible to spot.  In
  this scenario, the GQSCAN RC 8 can be gleaned from the calling
  applications diagnostic data.
  

Local fix

• There are two possible circumventions to bypass the issue but
  they both prevent the complete view of the ENQ contention in
  the environment.
  1) Using SETGRS GRSQ=LOCAL command to narrow the contention
     information to be collected on LOCAL system ONLY.
  .
  2) Using the following SLIP to ignore second and subsequence
     contented resource.
   Set up the following SLIP in IEASLPXQ as :
   SLIP SET,IF,A=REFBEFOR,L=(ISGDSDMP,976),ML=3000,ID=DMXQ,
     DATA=(10R,EQ,8,AND,15R,EQ,0),REFBEFOR=(15R,EQ,4),END
   Issue SET SLIP=XQ  to activate it or reactivate it after
   the SLIP was disabled after the matchlim reached.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of HBB7780 and up                  *
  ****************************************************************
  * PROBLEM DESCRIPTION: See Problem Summary                     *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  GQSCAN with WAITCNT= 1 or more, TOKEN, QUIT=NO, GENERIC,
  and SCOPE=ALL|SYSTEM|SYSTEMS
  incorrectly does not make progress if a RIB and its RIBEs
  (for a single resource) take up more space than is available
  in the GQSCAN Area. Until GQSCAN is invoked with a buffer
  large enough to contain all the data, GQSCAN incorrectly
  returns this same resource in the Area, and return code 8,
  each time it is invoked.
  
  ISGQUERY REQINFO=QSCAN, with non-specific searches with
  MINWAITERS=1 or more, SCANACTION=RESUME, similarly does not
  progress to the next potential matching resource
  if the ISGQUERY ANSAREA is too small to contain the RS and
  all of the RQs for a single resource.
  
  Also, on both types of scans specified above, matching
  Scope=Systems resources may not be returned.
  

Problem conclusion

• GQSCAN with WAITCNT=1 or more, TOKEN, QUIT=NO, GENERIC,
  and SCOPE=ALL|SYSTEM|SYSTEMS
  will progress to the next potential matching resource
  if the GQSCAN Area is too small to contain the RIB and
  all of the RIBEs for a single resource.
  
  ISGQUERY REQINFO=QSCAN, with non-specific searches with
  MINWAITERS=1 or more, SCANACTION=RESUME, has been similarly
  updated to progress to the next potential matching resource
  if the ISGQUERY ANSAREA is too small to contain the RS and
  all of the RQs for a single resource.
  
  On both types of scans specified above, matching
  Scope=Systems resources will be returned in the caller's area.
  

Temporary fix

• *********
  * HIPER *
  *********
  SEE ++APAR HA40883
  

Comments

APAR Information

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

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

  UA67484

Modules/Macros

• ISGQQSC  ISGQQSG  ISGQSC   ISGQSG
  

Fix information

• Fixed component name

  GRS

• Fixed component ID

  5752SCSDS

Applicable component levels

• R780 PSY UA67484

     UP12/12/27 P F212

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.