Patches 111336-04 and 112489-02 May Cause "Arbstop" on E10K Domains

---

Impact

An E10K running SSP (System Service Processor) 3.4 or SSP 3.5 may encounter a global "Arbstop" (arbitration stop) on one or more system domains, resulting in the affected domains becoming completely unresponsive.

"Arbstop" stands for "arbitration stop". It normally occurs only when the E10K hardware detects a fatal error. In such cases, no more grants are provided on any of the buses. The affected domains are completely dead until POST ("Power On Self Test") is run again.

Contributing Factors

This issue can occur in the following releases:

SPARC Platform

• SSP 3.4 (for Solaris 2.6, 7 and 8) with patch 111336-04 and without patch 111336-05
• SSP 3.5 (for Solaris 7 and 8) with patch 112489-02 and without patch 112489-03

The described issue may only occur if the following conditions are met and the following steps are performed (exactly in this order) during system administration:

1. Running "check_host " reports all domain(s) as DOWN.
2. Running "check_host -b" reports at least one domain as ALIVE.
3. The "setfailover -t cb force" command is issued.
4. The domain is rebooted.
5. The spare control board (not providing the clock) is powered off.

At this point the domain(s) will "Arbstop".

See the man pages for: check_host(1M) and setfailover(1M) for more information.

Symptoms

An E10K system domain stops with an "Arbstop" while a spare control board (not providing the clock) is powered off.

Workaround

If you cannot upgrade to the current patch level, please backout patch 111336-04 (SSP 3.4) or patch 112489-02 (SSP 3.5).

If you are using either patch 111336-04 (SSP 3.4), or patch 112489-02 (SSP 3.5), do not issue the command "setfailover -t cb force" while all domains are down.

Note: If you need to move the clock and jtag to the spare control board, verify that "check_host -b" returns "HOST down" for ALL domains before initiating the control board failover.

Resolution

This issue is addressed in the following releases:

SPARC

• SSP 3.4 (for Solaris 2.6, 7 and 8) with patch 111336-05 or later
• SSP 3.5 (for Solaris 7 and 8) with patch 112489-03 or later

Modification History


Date: 30-APR-2003

• State: Resolved
• Updated Contributing Factors, Relief/Workaround and Resolution sections

Date: 06-MAY-2003

• Re-opened as the issue is not Resolved
• Updated Contributing Factors, Relief/Workaround and Resolution sections

Date: 18-AUG-2003

• State: Resolved, patches released
• Updated Contributing Factors and Resolution sections

Attachments

This solution has no attachment