VM65626: ABENDSXA004 WITH MANY STACKED CALLS TO HCPVAIEV

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• When the system is under extreme stress due to intense levels of
  HiperSockets network activity it is possible for device signals
  to be generated faster than some users can process them.  If
  this situation persists, free storage can be flooded with CPEBKs
  stacked for entry point HCPVAIEV to process the signals.
  

Local fix

• Apply PTF SXA004
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: Customers using HiperSockets Guest LAN for   *
  *                 connectivity may be able to recreate this    *
  *                 condition if network traffic is very high    *
  *                 and the system has limited CPU resources for *
  *                 the guest processing inbound network data.   *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  ****************************************************************
  * RECOMMENDATION: APPLY PTF                                    *
  ****************************************************************
  In an environment using HiperSockets Guest LAN connectivity for
  an unusually active network, the system failed with ABEND SXA004
  (out of storage frames).  The dump revealed that an unusual
  number of CP frames were filled with CPEBK structures
  representing stacked calls to HCPVAIEV.
  Each call to HCPVAIEV represents an update to a simulated
  HiperSockets input queue for a virtual NIC.  The dump also
  revealed that multiple calls were stacked for the same
  input queue (suggesting that senders were adding network
  data to one receiver faster than that guest could process
  the input queue).
  

Problem conclusion

• Research indicates that a receiver only needs one signal to
  initiate queue processing (no matter how many updates are made
  before it starts queue processing).  This led to a new strategy
  for handling the call to HCPVAIEV.
  Entry point HCPVAIEV is updated to reset a new flag (VDEVAIF1)
  when the receiver is ready to process the input queue.
  Callers of HCPVAIEV are updated to set (using TS serialization)
  the new flag (VDEVAIF1) before stacking a call to HCPVAIEV.
  If TS fails, there is no need to stack a redundant call to
  HCPVAIEV.  This eliminates the redundant CPEBK structures
  which could otherwise build up if multiple senders are feeding
  a single receiver faster than it can process the calls.
  It also saves CPU cycles which would be used in stacking and
  processing the unnecessary HCPVAIEV calls.
  

Temporary fix

• *********
  * HIPER *
  *********
  FOR RELEASE VM/ESA CP/ESA R540 :
  PREREQ: VM65283 VM65406 VM64651
  CO-REQ: NONE
  IF-REQ: NONE
  FOR RELEASE VM/ESA CP/ESA R620 :
  PREREQ: VM65042 VM65406
  CO-REQ: NONE
  IF-REQ: NONE
  FOR RELEASE VM/ESA CP/ESA R630 :
  PREREQ: VM65417
  CO-REQ: NONE
  IF-REQ: NONE
  

Comments

APAR Information

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

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

  UM34523 UM34524 UM34525

Modules/Macros

• HCPIQE   HCPIQR   HCPVAI   HCPVDEV  HCPVQA
  

Fix information

• Fixed component name

  VM CP

• Fixed component ID

  568411202

Applicable component levels

• R540 PSY UM34523

     UP15/01/26 P 1501 Ž

• R620 PSY UM34524

     UP15/01/27 P 1501 Ž

• R630 PSY UM34525

     UP15/01/26 P 1601 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.