RE: [ACPI] Metolious hardware-sensors-using-ACPI specs

RE: [ACPI] Metolious hardware-sensors-using-ACPI specs

[Grover, Andrew]

> From: Pavel Machek [mailto:pavel@suse.cz] 
> Ouch, I started implementing that hour ago... [Never mind, very little
> damage done so far].

Wow you work fast. ;-)

> But... Metolious sounds *needed*; how do you access voltage sensors
> without metolious, in a way that can coexist with ACPI thermal
> support?

(I think you mean thermal sensors)

A solution in search of a problem. I can say this because I helped define
it. :)

The machines that care about manageability (servers) appear to be entirely
disjoint from the ones that have thermal zones (and, servers use IPMI),
therefore thermal chip contention doesn't happen. And, Metolious required a
fair amount of AML code.

-- Andy

Re: [ACPI] Metolious hardware-sensors-using-ACPI specs

[Pavel Machek]

Hi!

> > But... Metolious sounds *needed*; how do you access voltage sensors
> > without metolious, in a way that can coexist with ACPI thermal
> > support?
> 
> (I think you mean thermal sensors)

No, I mean voltage sensors. They are on same smbus as thermal sensors,
yet their are not normally accessible using ACPI. I can talk to smbus
controller directly, but acpi may decide to read temperature while I'm
reading voltage, leading to armagedon.

How is that solved?

> A solution in search of a problem. I can say this because I helped define
> it. :)

Well, I thought that "simple" specs looks reasonable, I did not look
at "advanced" metolious. But it seemed sane (*).

> The machines that care about manageability (servers) appear to be entirely
> disjoint from the ones that have thermal zones (and, servers use IPMI),
> therefore thermal chip contention doesn't happen. And, Metolious required a
> fair amount of AML code.

I have seen desktop machine that can control CPU fan (altrough it is
slow/fast not on/off) and has thermal zone. It would be nice to be
able to check voltages in a safe way...
							Pavel

(*) comapred to ACPI2.0 ;-).
-- 
Casualities in World Trade Center: ~3k dead inside the building,
cryptography in U.S.A. and free speech in Czech Republic.

SMIC and IPMI (was: [ACPI] Metolious hardware-sensors-using-ACPI specs)

[Jirka Kosina]

On Mon, 16 Dec 2002, Grover, Andrew wrote:

> The machines that care about manageability (servers) appear to be entirely
> disjoint from the ones that have thermal zones (and, servers use IPMI),

Talking about IPMI - is there anyone working on SMIC interface to IPMI
driver written by Corey Minyard? (http://home.attbi.com/~minyard). He
isn't working on it personally as he told me, because he doesn't have such
HW.

I've just downloaded specification from intel and can start working on it
after I finish my other projects, but I am asking firstly, to avoid
implementing it if someone is currently doing it.

--
JiKos.

RE: [ACPI] Metolious hardware-sensors-using-ACPI specs

[Grover, Andrew]

> From: Pavel Machek [mailto:pavel@suse.cz] 
> Is it goign to be implemented in linux-acpi?
> 
> I took a look at specs at intel, and it has rather funny legaleese:

Wow, is that still on a website somewhere?

So as you may know from looking at the spec, Metolious was a spec that
defined a way for platforms to enumerate various motherboard sensors to the
OS, for manageability purposes.

It never took off, except for a couple companies that used the Windows
driver for other things because they didn't want to write a driver that
received ACPI device Notify()s.

The licensing may be weird, but given that there really is no point in
implementing it on Linux, does that really matter?

Regards -- Andy

Re: [ACPI] Metolious hardware-sensors-using-ACPI specs

[PSI-Systems]

> Wow, is that still on a website somewhere?
http://www.intel.com/design/servers/ipmi/metolious.htm
http://www.acpi.info/papers/Imple6/metolious.pdf

Re: [ACPI] Metolious hardware-sensors-using-ACPI specs

[Pavel Machek]

Hi!

> > Is it goign to be implemented in linux-acpi?
> > 
> > I took a look at specs at intel, and it has rather funny legaleese:
> 
> Wow, is that still on a website somewhere?

Yep, as someone already pointed out.

> So as you may know from looking at the spec, Metolious was a spec that
> defined a way for platforms to enumerate various motherboard sensors to the
> OS, for manageability purposes.
> 
> It never took off, except for a couple companies that used the Windows
> driver for other things because they didn't want to write a driver that
> received ACPI device Notify()s.

> The licensing may be weird, but given that there really is no point in
> implementing it on Linux, does that really matter?

Ouch, I started implementing that hour ago... [Never mind, very little
damage done so far].

But... Metolious sounds *needed*; how do you access voltage sensors
without metolious, in a way that can coexist with ACPI thermal
support?

								Pavel
-- 
Casualities in World Trade Center: ~3k dead inside the building,
cryptography in U.S.A. and free speech in Czech Republic.