PM37396: QDIO ACCELERATOR CONTINUES PROCESSING AFTER STORAGE SHORTAGE IS DETECTED

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• The QDIO Accelerator function that is accelerating Sysplex
  Distributed traffic with an associated VIPAROUTE or traffic
  with divergent outbound interfaces may require the use of
  pre-staged storage.  If the pre-staged storage becomes
  exhausted the accelerating logic forces any additional
  packets presented in the inbound buffer up through the normal
  TCP/IP inbound path.  A portion of the accelerating code
  path is being erroneously executed after the decision to
  force the normal inbound path has been made.  The errant
  accelerating code traverses the remaining inbound packets
  and there is a potential loss of data that may occur during
  this traversal process if a specific data pattern exists.
  If the packet(s) with lost data requires IP forwarding by
  the normal inbound path the lost data will be detected by
  the destination TCP/IP, resulting in the packet(s) being
  discarded.  If the packet(s) with lost data has a final
  destination of the local TCP/IP stack the lost data will
  not be detected and the packet(s) payload will be presented
  to the socket connection.
  The pre-staged storage is a special pool of storage that
  is made available for disabled code paths.  When the pool
  becomes exhausted an enabled processes is started to
  replenish the pool.  The pool management provides no
  external notification of the state of the pool.  When the
  pool becomes exhausted there are no messages issued or any
  other form of external notification.
  

Local fix

• Disable the QDIO Accelerator function using the IP Sys Admin
  Command, VARY TCPIP,,OBEYFILE.
  
  1. Create a data set that contains the following TCP/IP profile
     configuration statement.
  
     IPCONFIG NOQDIOACCELERATOR
  
  2. Issue the OBEYFILE command to dynamically change the TCP/IP
     configuration without stopping and restarting the TCP/IP
     address space.
  
     V TCPIP,tcpproc,O,datasetname
  
  3. Update the existing TCP/IP profile configuration data set to
     disable QDIO Accelerator in the event TCP/IP is stopped and
     restarted.
  
       Change QDIOACCELerator to NOQDIOACCELerator and delete
       the optional QDIOPriority if present.
     or
       Delete QDIOACCELerator and delete the optional
       QDIOPriority if present.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of the IBM Communications Server   *
  *                 for z/OS Version 1 Releases 11, 12 and 13    *
  *                 IP: QDIO Accelerator and Hipersockets        *
  *                 Accelerator                                  *
  ****************************************************************
  * PROBLEM DESCRIPTION: Potential loss of data following        *
  *                      exhaustion of pre-staged storage        *
  *                      for QDIO accelerator.                   *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  A pre-staged storage pool used by the QDIO/HiperSockets
  accelerator became constrained.  An asynchronous storage
  management process, running as an enabled thread, was
  started to replenish this pool.  During the time interval
  the storage pool was exhausted, the accelerator exit
  performed incorrect processing on some remaining inbound
  packets that held a specific data pattern.  The error
  in processing resulted in the subsequent loss of data.
  
  The problem is possible for configurations using:
  
   - QDIO Accelerator with Sysplex Distributer
   - QDIO Accelerator with non-Sysplex Distributer
   - HiperSockets Accelerator (i.e.IQDIOROUTING)
  
  The problem does not occur when the Accelerator function
  is disabled.
  +-------------------------------------------------------------+
  + Please check our Communications Server for OS/390 homepages +
  + for common networking tips and fixes.  The URL for these    +
  + homepages can be found in Informational APAR II11334.       +
  +-------------------------------------------------------------+
  

Problem conclusion

• Module EZBIFQDR (the Accelerator Routing Exit) has been
  amended to correctly process inbound packets following
  exhaustion of the pre-staged storage pool.
  
  * Cross Reference between External and Internal Names
  

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:

  UK66995 106PC1Ÿ 106PC1Ÿ UK66996 106PC1Ÿ 106PC1Ÿ 106PC1Ÿ UK66997
  106PC1Ÿ 106PC1Ÿ 106PC1Ÿ

Modules/Macros

• EZBIFQDR
  

Fix information

• Fixed component name

  TCP/IP V3 MVS

• Fixed component ID

  5655HAL00

Applicable component levels

• R1B0 PSY UK66995

     UP11/05/04 P F105 Ž

• R1C0 PSY UK66996

     UP11/05/18 P F105 Ž

• R1D0 PSY UK66997

     UP11/05/19 P F105 Ž

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.