IBM Support

VM66920: PEVM66907 Z/VM 7.4 CP FIX PACK - FEATURE 03 FIX 05

A fix is available

Subscribe

You can track all active APARs for this component.

 

APAR status

  • Closed as program error.

Error description

  • This is Feature Pack 03 Fix Pack 05 for z/VM 7.4.
    

Local fix

Problem summary

  • ****************************************************************
    * USERS AFFECTED: All users exploiting:                        *
    *                 - The z/VM Virtual Switch                    *
    *                 - EQDIO Virtual NICs                         *
    *                 - EQDIO adapters                             *
    *                 - QDIO adapters                              *
    ****************************************************************
    * PROBLEM DESCRIPTION:                                         *
    ****************************************************************
    * RECOMMENDATION: APPLY PTF                                    *
    ****************************************************************
    The following issues are addressed by this APAR:
    
    OSH SET VSWITCH SWITCHOVER loses connectivity when there are 2
    backup RDEVs
    ---------------------------------------------------------------
    When there are more than one backup RDEVs available in a
    traditional OSH VSWITCH, a SET VSWITCH SWITCHOVER command (with
    or without specifying an RDEV) appears successful, but
    connectivity is lost.
    
    The EQID of an EQDIO device is reset after a config off/on
    ----------------------------------------------------------
    When performing a config off/on of an EQDIO device associated
    with a VSWITCH or a LAG group, the EQID of the device is
    cleared.
    
    VSWITCH error recovery message indicates a "persistent
    error condition" which results in a recovery loop
    ------------------------------------------------------
    Persistent errors are meant to require operator intervention
    (SET VSWITCH CONNECT) when the problem has cleared. OSH VSWITCH
    support is still attempting to automatically recover devices
    that have encountered a persistent error, which leads to a
    recovery loop and many messages on the operator console.
    
    Additional monitor counters are needed for QDIO/EQDIO
    -----------------------------------------------------
    Monitor records do not currently provide sufficient counters for
    performing performance comparisons between QDIO and EQDIO
    devices.
    
    z/OS guest receives error code 03 when deactivating
    EQDIO virtual NICs
    ------------------------------------------------------------
    When a z/OS guest deactivates an EQDIO virtual NIC, it can
    receive message HCPSWU2831E indicating there was a controller
    error code 03 encountered. Aside from the message, there are no
    noticable symptoms.
    
    Eliminate unnecessary EQDIO instructions to improve efficiency
    --------------------------------------------------------------
    Performance analysis has shown that some relatively high cost
    EQDIO instructions are issued more frequently than required.
    
    Unexpected ISOLATE message when a switch port is enabled for LAG
    ----------------------------------------------------------------
    In a multi-port EQDIO LAG, if multiple ports are shutdown and
    then one is re-enabled, the VSWITCH support will encounter a
    message indicating a protocol violation because the EQDIO device
    is not in the correct state to issue an ISOLATE command. This
    occurs when the VSWITCH is not configured for isolation.
    
    LAG EQDIO VSWITCH does not attempt to initialize other devices
    if first device is disabled
    --------------------------------------------------------------
    When multiple EQDIO devices are configured in a link aggregation
    group, and one or more of those devices is offline when a
    VSWITCH using that group is brought online, device
    initialization will stop once it encounters the first disabled
    device. This means that in a LAG group with three devices where
    the first device is offline, no devices will complete
    initialization and there will be no connectivity.
    
    EQDIO VSWITCH loses connectivity for some guests after a
    hardware switch reboot
    --------------------------------------------------------
    When an exclusive LAG is set up that spans multiple hardware
    switches and one of those switches is rebooted, only partial
    connectivity is restored when the hardware switch comes back
    online. Some guests are no longer able to receive traffic.
    
    ABENDHTT001 HCPIVC+16E8 with 16 members in an IVL domain
    --------------------------------------------------------
    With 16 members in an IVL domain, it is possible to encounter an
    HTT001 abend.
    
    PEVM66907 HCP3034E when a user couples two NICs to the same
    VSWITCH
    -----------------------------------------------------------
    VM66907 addressed a problem where a user being relocated may
    encounter an error because the virtual port its NIC was
    connected to was already in use on the target system. With this
    fix applied, any user attempting to couple two NICs to the same
    VSWITCH will now receive an HCP3034E message and only one NIC
    will be able to be coupled.
    

Problem conclusion

  • As noted in details below, Firmware D61C Bundle 20 is required.
    
    OSH SET VSWITCH SWITCHOVER loses connectivity when there are 2
    backup RDEVs
    --------------------------------------------------------------
    In this scenario, the wrong control block was being stored into
    our routing entry so traffic was being routed to the wrong
    device. Code has been updated to select the appropriate control
    block for the routing table after the SWITCHOVER completes.
    
    The EQID of an EQDIO device is reset after a config off/on
    ----------------------------------------------------------
    During config off processing, the deletion of the VDEV block
    associated with the device hung.  This led to a new RDEV block
    being created when the device was configured back online.  The
    new RDEV block did not have the EQID associated with it.  Code
    has been changed to properly delete the VDEV block on config off
    such that the original RDEV (and EQID) are reused when the
    device is configured back online.
    
    VSWITCH error recovery message indicates a "persistent error
    condition" which results in a recovery loop
    ------------------------------------------------------------
    Code has been changed to skip enabling the recovery timer for
    devices thathave encountered a persistent error condition.
    
    Additional monitor counters are needed for QDIO/EQDIO
    -----------------------------------------------------
    A new monitor record (MRIODQEQ - D6 R55) has been defined to
    collect EQDIO queue based data.  The user based counters already
    in MRUSEACT are also appropriately populated with this fix.
    
    z/OS guest receives error code 03 when deactivating EQDIO
    virtual NICs
    ---------------------------------------------------------
    The error code 03 was indicating that EQDIO simulation was
    processing an unexpected adapter command. This occurred because
    the code for cleaning up mulitcast addresses associated with the
    virtual NIC was assuming it was operating in layer 3 mode.  Code
    has been updated to validate the transport mode the virtual
    adapter is operating in when cleaning up multicast addresses,
    and respond appropriately.
    
    Eliminate unnecessary EQDIO instructions to improve efficiency
    --------------------------------------------------------------
    The manner in which z/VM tracks queue indicies for EQDIO
    adapters has been changed to eliminate the vast majority of the
    queries that are issued to the adapter for ensuring indicies
    stay in sync. This greatly improves the efficiency with which
    the EQDIO support operates.
    
    Unexpected ISOLATE message when a switch port is enabled for LAG
    ----------------------------------------------------------------
    Code has been updated to not send an ISOLATE command to the
    EQDIO adapter when neither ISOLATE nor VEPA is enabled.
    
    LAG EQDIO VSWITCH does not attempt to initialize other devices
    if first device is disabled
    --------------------------------------------------------------
    EQDIO initialzation logic has been updated to react
    appropriately when it encounters a device offline. Port group
    initialization will no longer halt when the first device is
    offline.
    
    EQDIO VSWITCH loses connectivity for some guests after a
    hardware switch reboot
    --------------------------------------------------------
    This problem is addressed by a mixture of firmware and z/VM
    updates. For z/VM, initialization of link aggregation groups is
    now done in multiple steps. This is because gratuitious ARP
    packets that are used to announce MAC address changes when
    takeover events occur may be lost if they are sent before a link
    is fully LACP enabled. z/VM will now initially enable the EQDIO
    data plane for a LAG port just enough to flow LACP PDUs (but not
    guest data).  Once the port becomes LACP enabled, the data plane
    will then be enabled for guest data as well and initialization
    will complete with full connectivity.
    Firmware D61C Bundle 20 is required.
    
    ABENDHTT001 HCPIVC+16E8 with 16 members in an IVL domain
    --------------------------------------------------------
    This abend occurs when the IVL control block is allocated near
    the end of a 4K page. When sending a broadcast packet, the
    control loop stepping through the IVL domain was using an
    improper index which was causing it to address memory outside of
    the current page. Code has been updated to fix the indexing when
    stepping through the IVL members.
    
    PEVM66907 HCP3034E when a user couples two NICs to the same
    VSWITCH
    -----------------------------------------------------------
    Code has been updated to fix an error with VM66907 that was
    improperly discerning the difference between a transient port
    number and a reserved port number.
    
    The following publication is updated as part of this APAR:
    z/VM Performance:
    MRIODQEQ is added to the list of monitor records
    D6 R55 EQDIO Queue-based counters
    

Temporary fix

Comments

APAR Information

  • APAR number

    VM66920

  • Reported component name

    VM CP CP

  • Reported component ID

    568411202

  • Reported release

    740

  • Status

    CLOSED PER

  • PE

    NoPE

  • HIPER

    NoHIPER

  • Special Attention

    NoSpecatt / Xsystem

  • Submitted date

    2026-02-13

  • Closed date

    2026-02-24

  • Last modified date

    2026-03-24

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

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

    UM90590

Modules/Macros

  • CPLOAD   HCPCPE   HCPDTS   HCPEQL   HCPEQS   HCPIOT   HCPIPN
    HCPIQC   HCPIQG   HCPIVC   HCPIVL   HCPLLT   HCPMDLAT HCPMES
    HCPMESB  HCPMNO   HCPMONEQ HCPMXD   HCPMXD$  HCPNDF   HCPNICBK
    HCPNIDBK HCPNVM   HCPPRV   HCPQDIO  HCPRP    HCPSWA   HCPSWB
    HCPSWU   HCPVLC   HCPVLE   HCPVLG   HCPVLI   HCPVLM   HCPVQE
    HCPVQF   HCPVQG   HCPVQH   HCPVQS   MRIODQEQ MRUSEACT 7VMCPR40
    

Publications Referenced
SC24630174    

Fix information

  • Fixed component name

    VM CP CP

  • Fixed component ID

    568411202

Applicable component levels

  • R740 PSY UM90590

       UP26/02/26 P 7404

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.

[{"Business Unit":{"code":"BU029","label":"Software"},"Product":{"code":"SG27M","label":"APARs - z\/VM Environment"},"Platform":[{"code":"PF025","label":"Platform Independent"}],"Version":"740","Line of Business":{"code":"LOB16","label":"Mainframe HW"}}]

Document Information

Modified date:
24 March 2026