[coreboot] [blubkin at vmware.com: RE: [Openipmi-developer] [Discuss] [PATCH] ipmi: use round_jiffies on timers to reduce timer overhead/wakeups]

[Peter Stuge]

Interesting notes from LKML and openipmi. An estimate was that kipmid
hogs 10W from a 70W server.


//Peter

----- Forwarded message from Bela Lubkin <blubkin at vmware.com> -----

From: Bela Lubkin <blubkin at vmware.com>
To: 'Corey Minyard' <minyard at acm.org>,
	Randy Dunlap <randy.dunlap at oracle.com>
CC: "discuss at LessWatts.org" <discuss at LessWatts.org>,
	"Kok, Auke" <auke-jan.h.kok at intel.com>,
	Arjan van de Ven <arjan at linux.intel.com>,
	"openipmi-developer at lists.sourceforge.net"
	<openipmi-developer at lists.sourceforge.net>, lkml
	<linux-kernel at vger.kernel.org>
Subject: RE: [Openipmi-developer] [Discuss] [PATCH] ipmi: use round_jiffies
	on timers to reduce timer overhead/wakeups
Date: Wed, 21 Oct 2009 16:46:41 -0700

Corey Minyard & Randy Dunlap wrote:

Randy>> From what I recall (probably 2 years ago), [older] ipmi hardware
Randy>> does not generate event interrupts, so it has to be polled.
Randy>>
Randy>> Corey, can you elaborate on this?

Corey> Certainly.  Yes, some (probably most) IPMI hardware does not use
Corey> interrupts, and unfortunately, it's not just older machines.
Corey> The driver used to poll more slowly, but in many cases the
Corey> performance was unacceptable.
Corey>
Corey> kipmid is only started if the hardware doesn't support
Corey> interrupts, so only users with sub-standard hardware have to
Corey> suffer with this problem.

Regrettably, of the "big three" in the "PC Server" world, only HP's iLO2
BMC supports interrupts.  Dell's DRAC4 & 5 don't, IBM's ASM, RSA, etc.
don't.  Also (at least out of a sample of one) SuperMicro also doesn't
have an interrupt.

They also have all settled on the KCS interface, which dribbles one
character through per non-interrupt.  So sad.  Dell's DRAC3 had a BT BMC
which transferred whole IPMI packets via DMA _and_ had an interrupt.
HP's ancient SMIC equipment also had an interrupt (but that's also
char-by-char, and their current KCS has an interrupt, so at least they
haven't regressed).  I've guessed that some chip vendor must have
come out with a Really Cheep KCS implementation and drove every other
implementation out of the market.  :-(

I've heard rumors that some current Sun hardware has BT.

>Bela<--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

----- End forwarded message -----