Centrino support

From http://news.com.com/2100-1006-993896.html:

Intel plans Linux support for Centrino

Intel is working on Linux support for Centrino, its package of chips for mobile computers with wireless networking abilities, but the company hasn't yet decided how or when to release it.

That was on March 24, 2003.

Well, that was almost 5 months ago. So I figured I'd ask if there's any progress -- so far the built-in wireless in my notebook still doesn't work with Linux and the machine is monstrously power-hungry because Linux doesn't scale the CPU frequency.

I know there are some Intel people on the list -- perhaps someone can comment?

--J.

Status:
Wlan - not supported

CPU - CPUfreq (-ac tree) and ACPI throttling work just fine. I've got my Pentium M running at 600 MHz when the Power Supply is plugged out.

Regards,
Martin List-Petersen
martin at list-petersen dot se
--
for ARTIFICIAL FLAVORING!!

Intel shows no inclination to release Centrino wireless drivers for Linux. There have been vague insinuations that this is due to excessive software controllability, but no public explanations have been given, beyond "we're not doing it at this moment".

If you want built-in wireless in the nearish term, you'll have to get a supported MiniPCI card and replace your Centrino card.

As far as CPU is concerned, if you're using recent 2.5 or 2.6 kernels, there's Pentium M support in cpufreq. Jeremy Fitzhardinge has written a userspace daemon that varies the Pentium M CPU frequency in response to load.

	<b

Got a list of supported good working cards?

Can you please point me to this daemon?

Regards

--
Christian Axelsson
[email protected]

On Fri, 15 Aug 2003 22:24:26 +0200 Christian Axelsson <[email protected]> wrote:

| Bryan O'Sullivan wrote:
| 
| >If you want built-in wireless in the nearish term, you'll have to get a
| >supported MiniPCI card and replace your Centrino card.
| >
| >  
| >
| Got a list of supported good working cards?

Mini-PCI or CardBus?
I think that some people just add a CardBus wireless card.

| >As far as CPU is concerned, if you're using recent 2.5 or 2.6 kernels,
| >there's Pentium M support in cpufreq.  Jeremy Fitzhardinge has written a
| >userspace daemon that varies the Pentium M CPU frequency in response to
| >load.
| >  
| >
| Can you please point me to this daemon?

http://www.goop.org/~jeremy/speedfreq/

--
~Randy

There's a Dell TrueMobile card that uses the Orinoco chipset. If you're feeling like life is too boring, there are cards based on the newer Intersil dual 802.11b/g chipsets available, too, and though I haven't checked into the shape of the drivers, I know they're under active development.

	<b

The Dell TrueMobile 1150 series are Agere/Orinoco/Hermes based (MiniPCI and PC-Card available). All other Dell TrueMobile cards are Broadcom based and have no Linux driver support either.

There are also MiniPCI, PC-Card, USB adapters with 802.11a/b/g and Linux drivers available: http://sf.net/projects/madwifi

These are based on the Ateros chipset, which also is around in some OEM products.

http://sf.net/projects/cpufreqd

Regards,
Martin List-Petersen
martin at list-petersen dot se
--
"An idealist is one who, on noticing that a rose smells better than a cabbage, concludes that it will also make better soup." - H.L. Mencken

That driver contains a binary-only part.

<quote>
The ath_hal module contains the Atheros Hardware Access Layer (HAL). This code manages much of the chip-specific operation of the driver. The HAL is provided in a binary-only form in order to comply with FCC regulations. In particular, a radio transmitter can only be operated at power levels and on frequency channels for which it is approved. The FCC requires that a software-defined radio cannot be configured by a user to operate outside the approved power levels and frequency channels. This makes it difficult to open-source code that enforces limits on the power levels, frequency channels and other parameters of the radio transmitter. See

http://ftp.fcc.gov/Bureaus/Engineering_Technology/Orders/2001/fcc01264.pdf

for the specific FCC regulation. Because the module is provided in a binary-only form it is marked "Proprietary"; this means when you load it you will see messages that your system is now "tainted". </quote>

US was a free country. It gets worse by the day.

vda

I have just read that FCC regulation.

It doesn't prevent open-sourcing the code. It does require that each software-defined radio must have some kind of authentication mechanism to ensure that unapproved software cannot be loaded on to the device.

It seems to me that distributing a binary, which is easily modified by users, (as proven by all the game patches and application cracks out there), does _not_ satisfy the FCC regulation.

The only way to satisfy the regulation is to have an authentication mechanism of some kind, so that the radio will not operate with unapproved software.

If Intel have not implemented an authentication mechanism, then they are not compliance with the FCC regulation as I read it because it won't be long before some enterprising user patches the firmware and makes the radio behave in an unapproved, possibly RF unsafe manner.

If Intel have implement such a mechanism, then regulation is no excuse for them to not release the source code.

-- Jamie

Bryan> Intel shows no inclination to release Centrino wireless drivers Bryan> for Linux. There have been vague insinuations that this is due Bryan> to excessive software controllability, but no public Bryan> explanations have been given, beyond "we're not doing it at this Bryan> moment".

Bryan> If you want built-in wireless in the nearish term, you'll have Bryan> to get a supported MiniPCI card and replace your Centrino card.

That's what I find extremely annoying. Especially in the context of Intel's trumpeted announcements about support for Linux (see the URL in my previous E-mail). I mean, you either support Linux, or you don't. Intel announced that support is coming and then hasn't delivered it.

This is offtopic on this list, but frankly, I'm surprised why RedHat (or any other Linux company for that matter) hasn't filed an unfair competition suit yet. Intel's approach basically favors Microsoft over other companies by giving them a year or so headway before anybody else has a chance of getting the hardware supported. That surely sounds like an unfair practice to me.

Bryan> As far as CPU is concerned, if you're using recent 2.5 or 2.6 Bryan> kernels, there's Pentium M support in cpufreq. Jeremy Bryan> Fitzhardinge has written a userspace daemon that varies the Bryan> Pentium M CPU frequency in response to load.

I keep dreaming about the day when I'll be able to have a modern laptop with a stable Linux kernel. As for now, it has taken me (on one of my laptops) about 1.5 years to get to a point where 2.4 works, most of my hardware works, and software suspend (pretty much a requirement for laptops) works. I'm not about to give that up easily, so I'm not that eager to jump to 2.5/2.6.

Question time:

1. Will cpufreq make it into the standard 2.4 kernels?
2. If not, will Alan incorporate swsusp into -ac kernels? (given that -ac kernels seem to have cpufreq included)
3. Where does one get 2.4 cpufreq?

thanks,
--J.

Can't say that's been my experience. I bought a new Thinkpad X31 the other day, and it's already running 2.6.0-test3. Suspend works, all's happy.

Highly unlikely.

There are snapshots available at www.linux.org.uk (hint: Google for "cpufreq"), or in -ac. See the cpufreq mailing list archives for a bunch of daemons that control this stuff in various semi-cooked ways.

	<b

Bryan> Can't say that's been my experience. I bought a new Thinkpad Bryan> X31 the other day, and it's already running 2.6.0-test3. Bryan> Suspend works, all's happy.

Lucky you! That's because almost two years (or so) of heavy work by many people went into this. Also, you're probably lucky with your Thinkpad.

When I bought this laptop (a Sharp Mebius PC-MT1-H5) back in Sept 2001, there weren't many machines on the market that had only ACPI (and no APM). And Linux ACPI wasn't in a very sane state back then, not to mention swsusp. It took almost two years for the software to mature, and only recently did I get a stable machine that I can work on for a month without rebooting (suspending/resuming several times a day).

I've just gotten a Centrino-based Toshiba Dynabook SS S7/290LNKW, and the story continues -- I've already hit at least two ACPI bugs, while swsusp problems seem to have been ironed out thanks to hard work by Nigel Cunningham. And of course, the built-in wireless card does not work. My guess is another 6 months (if not more) until Linux works on it.

--J.

We asked Marcelo if he wanted it for 2.4.23, and he didn't object, so time permitting, it could show up there next time.

Very unlikely.

Normally http://www.codemonkey.org.uk/projects/cpufreq
However the box hosting that website dies within a few hours if I restart apache (Its incredibly underpowered for the load it gets), so plans are afoot to move it to something that can handle it. Should be back up (for good this time hopefully) within a week or so.

Dave

 Dave Jones     http://www.codemonkey.org.uk

In time maybe - up to Marcelo.

Not in its current form. To do the job well swsuspend needs the kernel device model.

I've tried it a few times on 2.6.0-test3 right after the system comes up and I log in to text mode, and the sucker spits out exactly one line of text (something like "Suspending processes:") and hangs with the cursor right after the semicolon. After about a minute I hit ctrl-scroll lock, but everything except the suspend thread (which was waiting in some kind of yield for other threads to do something) scrolled off the top and I don't know how to cursor up to see it (or get it to not scroll off).

I'd be happy to test someting better. APM suspend doesn't work on my new thinkpad, so it's swsusp or nothing. Currently, it's "shutdown -h now", which has some obvious deficiencies...

Rob

I thought that this line of argument was due to FCC regulations. That is, software settings would allow the hardware to violate frequency or strength-of-signal limitations set by government regulations. This is only from memory, so feel free to correct.

-Brandon

I think I have read in an earlier thread something the like.
But I cannot understand how this can be logically linked to releasing docs. If all companies would follow this thought e.g. Siemens would never have released the docs for ISDN chipsets and therefore no ISDN drivers would be in the kernel. I'd rather say someone with money is afraid ...
Mobile equipment like laptops is a booming market and it all turns around shares...
There is another point about this topic. How does M$ (with their centrino drivers) guarantee that no user tries to switch on the addtional ETSI frequencies? I doubt this is possible at all. Of course you can simply ignore that there are ETSI frequencies which basically means to ignore the overwhelming part of europe and their respective regulation.
Some political explosives are in this thread ...

Regards,
Stephan

Stephan> On Fri, 15 Aug 2003 16:46:19 -0400

Stephan> I think I have read in an earlier thread something the like. Stephan> But I cannot understand how this can be logically linked to Stephan> releasing docs. If all companies would follow this thought Stephan> e.g. Siemens would never have released the docs for ISDN Stephan> chipsets and therefore no ISDN drivers would be in the Stephan> kernel. I'd rather say someone with money is afraid ...

Yes, that sounds rather ridiculous. Sooner or later someone is going to reverse-engineer the thing, so not releasing drivers or specs just delays this moment. If there's a manager at Intel that thinks this way, he doesn't understand much about security.

[...]

Stephan> Some political explosives are in this thread ...

Politically, this sounds to me like an antitrust case against Intel and Microsoft handed on a platter. A major hardware manufacturer releases information and software only to one dominating software company, while locking out the others.

Unfortunately, we live in a world of compromises. If I were RedHat, I'd stay real quiet and play nice in order to get Intel's cooperation on servers (which brings revenue) instead of fighting for laptops (which is a niche market for Linux, in a chicken-and-egg sort of way).

It's rather sad.

--J.

Let's be fair to Intel for a moment.

With hardware radios, anyone can open it up, fiddle with electronics, and make it do something illegal, possibly dangerous. Manufacturers aren't required to make them impregnable!

All manufacturers have to do is not put any knobs on the front which can make the radio do unapproved things.

Folk are allowed to fiddle with radio electronics, with care, as long as they get themselves a radio license and stick to the rules.

They can even sell an altered device, if they take it through the FCC approval process.

With a software radio, it's analagous. The manufacturer doesn't have to make it _impossible_ to reprogram, they just have to make it hard enough that ordinary users won't do it.

Releasing the source code may or may not result in ordinary users reprograming their radios in harmful ways. Though, you can imagine people would circulate patches to boost the power in no time.

At least there is some hope for the expert hobbyist: they _can_ reverse engineer the device. It is good that this is possible.

If Intel build a crypto-based authentication mechanism into their software radios, then there is no hope for the amateur radio hobbyist to fiddle with their radios.

So, really, Intel has done the amateur radio community a favour by leaving the tantalising possibility of reverse engineering the driver, compared with a DRM solution which totally prevents even expert tinkering.

-- Jamie

On Sad, 2003-08-16 at 20:58, Jan Rychter wrote

Vendors are beginning to make commitments to Linux support on some laptops, in addition Intel and ATI in paticular have been doing nice things to the laptop chipset market by drastically shrinking the number of variants. Centrino in some ways is a help as is the ATI+ALI setup used by all the nice mobile Athlon laptops - now when we get one working the rest tend to just follow.

Alan

As though one couldn't do all this with other wireless hw...

Tomas Szepe <[email protected]>

This does not in any way prevent them from releasing a driver, open or closed source. It merely makes tampering with the driver illegal.

And it doesn't prevent them from merely releasing programming specs if they're too cheap to make a driver either. Of course whoever programs the driver will then have to get approved
by the bureaucracy himself, or leave it to some linux vendor before it can be (legally) used.

No driver _and_ no specs means they aren't supporting linux at the moment.

Helge Hafting

Brandon> I thought that this line of argument was due to FCC Brandon> regulations. That is, software settings would allow the Brandon> hardware to violate frequency or strength-of-signal Brandon> limitations set by government regulations. This is only Brandon> from memory, so feel free to correct.

Helge> This does not in any way prevent them from releasing a driver, Helge> open or closed source. It merely makes tampering with the Helge> driver illegal.

Unless of course it can be programmed to listen in on the AMPS cell phone band(s). In that case it could arguably be a felony for them to release the programming info.... (Stupid law, but it has been enforced rather obsessively....)

-JimC

How is that different from selling a radio which can be retuned by opening it up and changing some resistors and capacitors?

-- Jamie

arguably a patch is much easier to implement and distribute where as changing some resistors and capacitors would take more technical knowledge, although I don't know where the stands on that.

CPU frequency scaling is supported now at least. (Though you'll need -ac for 2.4, or 2.6). Wireless is still unsupported AFAIK.

Dave

 Dave Jones     http://www.codemonkey.org.uk

I have made a patch set for my laptop which you may like to try. It mainly contains stuff from -aa and -ac kernels and various other pieces.

See http://www.sonarnerd.net/projects/linux/

With this and my backport of latest XFree86 ATI drivers I have Linux running nicely on my Compaq Evo N1015v and N1020v laptops. At least it _subjectively_ warms up less and consumes less power under Linux compared to Windows XP.

Jussi Laako <[email protected]>