IV47347: VARIOUS I/O AND LINK ISSUES IN PUREFLEX/BLADE SYSTEMS

APAR status

• Closed as program error.

Error description

• - Linkup between the 8Gb QLogic FC adapter and the 16Gb Brocade
  Flex switch module may fail when the Brocade switch is set to
  a speed setting other than 'auto'.
  - Link bounces on the internal or external ports of the 16Gb
  Brocade AG switch may cause path failures or very long path
  failover times.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:
  * Systems running the 5300-12 Technology Level with
  * devices.pci.77102224.com below the 5.3.12.9 level.
  ****************************************************************
  * PROBLEM DESCRIPTION:
  * Customers using PureFlex or Power Blade systems with
  * Access Gateway / Transparent mode switches may see
  * any number of the following issues:
  * linkup failures, devices not configuring, long IO delays,
  * long path failover times, hangs in the fcstat application
  * during link failures, link errors, and timeouts logged in the
  * error log for the fcs devices
  *
  * This is due to a couple issues seen with linkup and with
  * bouncing the link between the Top of Rack (TOR) and chassis
  * switches.
  *
  * Any customer using these switches should apply this fix to the
  * machines connected to the switches.  This applies to both AIX
  * and VIOS partitions.  Some of these issues are especially
  * likely when attached to a 16Gb Access Gateway switch.
  ****************************************************************
  * RECOMMENDATION:
  * Install APAR IV47347.
  * Prior to fix availability, an interim fix is available from
  * ftp://public.dhe.ibm.com/aix/ifixes/iv47347/
  ****************************************************************
  

Problem conclusion

• - Modify linkup auto-negotiation logic to use adapter NVRAM
  hints on the best topology settings for linkup operations.
  - Modify FC protocol driver async event assumptions to avoid
  waiting too long for an adapter RESUME event that will never
  arrive.
  - Modify adapter driver to abort timers waiting for events
  after back-to-back link bounces.
  - Improve handling of FLOGI timeout when trying to enable the
  adapter firmware virtual port (VP) associated with the physical
  connection.
  - Fix race condition between sleep and wakeup of ioctl calls
  for NDD_CHBA_DATA and NDD_RNID_DATA, both associated with
  fcstat.  This may cause fcstat to hang.
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

APAR Information

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

  IV47037

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

Fix information

• Fixed component name

  AIX 5.3 EXT SUP

• Fixed component ID

  5765G0399

Applicable component levels

• R53E PSY

     UP