kern/169475: IPMI MAX_TIMEOUT is too short

[Garrett Wollman]

>Number:         169475
>Category:       kern
>Synopsis:       IPMI MAX_TIMEOUT is too short
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Wed Jun 27 04:50:10 UTC 2012
>Closed-Date:
>Last-Modified:
>Originator:     Garrett Wollman
>Release:        FreeBSD 9.0-RELEASE-p3 amd64
>Organization:
Maintainer of {cvsup3,ftp5}.freebsd.org
>Environment:
System: FreeBSD xyz.csail.mit.edu 9.0-RELEASE-p3 FreeBSD 9.0-RELEASE-p3 #2 r237283M: Wed Jun 20 12:23:12 EDT 2012 wollman at xyz.csail.mit.edu:/usr/obj/usr/src/sys/CSAIL amd64

The machine is a Quanta QSSC-S99Q 2U fileserver (specifically,
ftp5.freebsd.org).  We have two other identical machines, soon to be
six.

>Description:

sys/dev/ipmi/ipmivars.h contains the following definition:

#define MAX_TIMEOUT 3 * hz

This determines how long ipmi(4) is willing to wait in its attach
routine for the management device to identify itself.  Unfortunately,
it is too short.  The IPMI specification recommends a timeout not less
than five seconds, and in practice on this platform it takes between
five and ten seconds.

>How-To-Repeat:

Boot FreeBSD on a Quanta QSSC-S99Q server.  Sometimes ipmi will
attach; most of the time it will time out.

>Fix:

Increase the timeout to ten seconds.  If this is felt to be too long,
make it a tunable.
>Release-Note:
>Audit-Trail:
>Unformatted: