PI38699: FIXED CSM GROWTH FOR CERTAIN TCP CONNECTIONS

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• .
  If a TCP socket is closed by the owning application when there
  is a large amount of data pending in the send buffer and the
  close action causes a significant portion of that data to be
  sent, it is possible for the set of packets to be split across
  two requests to the IP layer.  When this occurs the second
  request can become stranded, leaving it and the associated data
  in Fixed CSM buffers to remain allocated and unusable.  If a
  significant number of these events occur, it can lead to the
  Fixed CSM storage to become constrained (IVT5592I message); with
  the associated impacts to VTAM and TCPIP processing.
  
  Verification:
  
     A dump that includes common storage will have numerous
     x'8000' byte areas in SP241K6 with the eye-catcher SXPNSKMB
     at +18.  The older ones will have SKMBs with 0's at +4, +18
     will be an SKDB which will likely have the ALET of a CSM data
     space at +18, +28 will be x'00000001', and +4 will point to
     its owning SKMB which will have x'FFFFFFFF' at +14.
  
     If the dump includes TCPIP, an IP TCPIPCS xCB(* DATAQ)
     command (where x is T, U, and R) will probably not show any
     significant backlogs (some small amounts with time tags
     during the dump's collection are possible).
  
     A SYSTCPIP CTRACE with PFS and MESSAGE options enabled will
     show two (or more) 'Allocate MSGB/DATAB/data buffer' records
     with multiple 'Dupb Message Block' following them that all
     have the same DUCB as preceding 'Socket Close Entry' PFS
     record.  An IP SKMSG xxxxxxxx command using the address of
     the second MSGB allocated will show that first one with an
     IUDR and followed by several SKMBs that match the pattern
     above.
  Additional Keywords: 4KDS 64 4K dataspace 4K DSPACE64
  EZD1974E , compid B1 , SP231 key 6 IKTSH  SKMB  SKDB
  

Local fix

• .
  Monitor CSM usage using a DISPLAY NET,CSM command.  Recycle
  TCPIP to recover the stranded storage.  If the system has
  multiple stacks, use DISPLAY NET,CSM,OWNERID=ALL commands to
  determine which TCPIP instance has the growth.  If the system's
  real storage will support it, increasing the Fixed CSM limit can
  extend the period before the recycle is needed.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All users of the Communications Server for z/OS Version 2    *
  * Release 1 IP                                                 *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * An increase in CSM data space storage occurs.                *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Apply PTF                                                    *
  ****************************************************************
  An increase in CSM data space storage occurs due to some
  structures not getting put onto the queue which is used to send
  data. This causes those structures to never be freed.
  

Problem conclusion

• EZBTCSTR will be changed to ensure all structures which point to
  outbound data are put onto the queue to be sent out.
  

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:

  UI27273 UI27284 PI46872 PI62916

Modules/Macros

• EZBTCSTR
  

Fix information

• Fixed component name

  TCP/IP V3 MVS

• Fixed component ID

  5655HAL00

Applicable component levels

• R210 PSY UI27273

     UP15/06/24 P F506

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.