Asset ID:	1-71-1012587.1
Update Date:	2011-11-29
Keywords:

---

Solution Type  Technical Instruction Sure

Solution  1012587.1 :   A Workaround for CR 6293881 when forcing a core dump on Sun Fire[TM] V20z/V40z  


PreviouslyPublishedAs
217327

Oracle Confidential (PARTNER). Do not distribute to customers
Reason: Migrated distribution from Sun

Applies to:

Sun Fire V40z Server
Sun Fire V20z Server
All Platforms

Goal

Description

This document outlines a simple procedure on V20z/40z systems to allow a break sequence to be sent to the Solaris[TM] debugger via the serial over LAN feature whilst protecting the server against the consequences of a reboot of the Service Processor sending a break to Solaris[TM]

Solution

Steps to Follow

A Workaround for CR 6293881 when forcing a core dump on Sun Fire[TM] V20z/V40z

The procedure to force a dump on a hung x86 system is outlined in

Solaris[TM] Operating System: How to force a kernel core dump on an x86 or x64 system (Doc ID 1003085.1)

This involves booting Solaris with kadb (Solaris 9) or kmdb (Solaris 10) enabled and sending a break via the console port. If the Service Processor reboots on a Sun Fire[TM] V20z/V40z with the serial over LAN feature enabled then the serial connection is turned off and on which can be interpreted as a break signal by Solaris with kadb or kmdb enabled which will then drop the server into debug mode. This behaviour is considered normal and will not be changed. The workaround suggested by Newisys is not to use the Serial Over LAN feature and connect directly to the serial port rather than use the serial over LAN feature. The following workaround protects the server against interpreting a Service Processor reboot as a break and also preserves the ability to send a break sequence to the Solaris[TM] debugger via the Service Processor.

 1) Serial Over Lan must be enabled see How to setup platform serial console for Sun Fire[TM] V20z / Sun Fire[TM] V40z (Doc ID 1003869.1)

 2) Enable the alternate break sequence on the server. Use the command kbd -a alternate and to ensure the alternate break sequence is enabled over a reboot uncomment the line KEYBOARD_ABORT=alternate in /etc/default/kbd

3) System is booted with debugger enabled

4) The break sequence to use in the event of a hang is now ^ecl2. (instead of ^ecl1) (control-e typed together at one time, c, l, 2 which is a lowercase L followed by a "two" keystroke). Note on some older NSV/BIOS versions the break sequence is ^Exl2 control-shift-e together at one time, c, l, 2 which is a lowercase L followed by a "two" keystroke).

The ^Exl2 sequence will work on all versions of NSV/BIOS. This sequence typed in the console connection to the server on the SP will drop Solaris[TM] into the debugger prompt and a reboot of the Service Processor will not interpreted as a break sequence.

References and related documents.

Solaris[TM] Operating System: How to force a kernel core dump on an x86 or x64 system (Doc ID 1003085.1)
How to setup platform serial console for Sun Fire[TM] V20z / Sun Fire[TM] V40z (Doc ID 1003869.1)
<SUNBUG: 6293881> sp reboot sends break sequence to server.

Sun V20Z / V40z Current Product Issues (Doc ID 1340675.1)

Stinger, V20z, V40z, x86, kadb, kmdb, service processor, break sequence, 6293881
Previously Published As
82530



Attachments

This solution has no attachment