Intel Releases Linux Driver For Centrino WLAN

Werner Heuser writes "Finally Intel has made their different announcements about Linux support for the WLAN part of the Centrino technology become true. Though not yet officially announced an Open-Source driver with included firmware is available at SourceForge. The driver is still experimental and supposed to work with 2.4 Kernels as well as with 2.6 ones." (See these previous stories for some background.)

Intel Feeling the Pressure?

This really feels like Intel's finally feeling its stranglehold on the industry wavering a little (given AMD's 64bit success). I'd like to believe that this is going to lead them to start treating us like customers, rather than prisoners. Certainly, this is a nice first step.

Re:Intel Feeling the Pressure?

This is no first step, it's the begining of the end of many steps Intel has taken forth with it's centrino *line.* The only remaining piece was the WLAN component they have already facilitated the release of the speed stepping and other integrated components.

Wintel isn't ALWAYS the badguy.

NOW, I can say THANK GOODNESS no more lockups in Fedora from DriverLoader BS, now my only question is how will they allow Linux users to flash their firmware when the manufacturers don't provide floppy drives on most of the Centrino lines.

Re:Intel Feeling the Pressure?

K3b allows you to emulate a floppy disk on CDROM by burning a DOS bootdisk image to CD. When it boots you have A:\ which is the contents of the image file, and if CDROM drivers were loaded you have another drive letter for the the rest of the CD contents. I flashed the BIOS on my laptop from a Linux only environment like this.

Re:Intel Feeling the Pressure?

The correct format for a CDROM where the first track is bootable (and presented by the BIOS) as a 1.44MB floppy is called "El Torito"

Boy oh boy, were those IBMer's wacky...

From ipw2100_main.c

Whew!

if (!((r <= w && (e < r || e >= w)) || (e < r && e >= w))) {
IPW2100_DEBUG_TX("exit - no processed packets ready to release.\n");
return 0;
}

Fortunately there's a little ASCII art right above it that helps explain what that if condition does:

/*
* Quick graphic to help you visualize the following
* if / else statement
*
* ===>| s---->|===============
* e>|
* | a | b | c | d | e | f | g | h | i | j | k | l
* r---->|
* w
*
* w - updated by driver
* r - updated by firmware
* s - start of oldest BD entry (txq->oldest)
* e - end of oldest BD entry
*
*/

Re:From ipw2100_main.c

Whatever happened to meaningful variables (which is taught at age 12 before you even touch a language)

Re:From ipw2100_main.c

Duh!
This is high performance code! Single-letter variable names execute more faster.
U R teh st00p3d.

Re:From ipw2100_main.c

Judging by the scope of these variables and the fact that they seem to be docummented right at the top, I don't think anyone could have an issue with that.

In fact, sometimes explaining what a variable means and then using just a one letter name is much more helpful than names like "thisOneINeedToDoThisBecauseOfThat".

Just think of the use of "i" in for loops, no one in the right set of mind would use something like "loopCounter".

It's a bit like in PDE theory, if you use t, then you don't have to bother specifying that t belongs to [0,T] and that it's time - everyone expects that.

Re:From ipw2100_main.c

Just think of the use of "i" in for loops, no one in the right set of mind would use something like "loopCounter"

Quite, but if you're choosing decent variable names, you would never think of chooseing loopCounter!

What are you counting? That's what the variable name should be.

Iterating over rows in a matrix (or whatever)? then the variable name should be 'row'! Not rowCount or RowNumber or count or r, simply 'row'.

Then row++ makes sense - next row.

NDISWRAPPER

Until these drivers stabilize you can use NDISWRAPPER [sourceforge.net].

This tool allows you to run the Windows driver for some wireless cards that have little or no Linux support.

Daniel

Re:NDISWRAPPER

Do not encourage the use of NDISWRAPPER! Someone will probably moderate this as Troll, but come on - we all know that having such a "fallback option" makes the hardware makers relax more when it comes to releasing natively running, opensource Linux drivers!

Re:NDISWRAPPER

But for them to relax more they'd have to be working on something in the first place, most of the hardware makers that are willing to support linux are gonna do it with the best drivers they can, not have people running their software in linux with some little hack. And the ones that don't support linux don't care that some little app lets people run windows drivers, they weren't going to support linux anyway it's not worth it for them. If NDISWRAPPER works then people should use it, I know I'd deffinitely use it if it supports my laptops network card (haven't been able to get this thing to work at all, some fairly old lucent technologies wireless card, I think there is support for some newer version of this card but not mine.) I'd use a newer card with linux support but the laptop itself doesn't support these (dunno why, tried some netgear card it didn't like that very much I think the PCMCIA slot in my laptop is 16 bit or something like that it's an old laptop.)

Re:NDISWRAPPER

You're both right! You should continue to use NDISWRAPPER, while complaining about it every chance you get.

This is a great sign

I can remember the day when I could only get spotty support for most hardware on my Linux box. Buying a computer was complicated if I wanted to use Linux -- I might only have one or two choices for a component (i.e. soundcard) because the drivers came from the community. This is a great sign, if Intel starts supporting all of their products under Linux, other vendors will follow suit, and it won't be long before you'll see Lindows boxes alongside the Macs at CompUSA!

Re:This is a great sign

This is a great sign, if Intel starts supporting all of their products under Linux, other vendors will follow suit, and it won't be long before you'll see Lindows boxes alongside the Macs at CompUSA!

Yeah I know pretty soon we might get some linux support from other companies! Like NVidia, 3Comm, Ceative Labs, ATI, Netgear, Linksys, man pretty soon I'm gonna be able to build a sweet linux computer!

*Looks at his own two linux computers*

Oh...

I'd actually be more excited about Intel's decision if they had any products I actually wanted. I don't know of any companies I'd buy from whose products don't work in linux one way or another. Sure some things might not work, but I haven't run into anything in the past 2-3 years that I couldn't get working in linux although setting up my ATI card was a real pain. There are even a few no name devices that I wouldn't expect to work, that just happened to have support since they use the same chipset as like 40 other no name devies.

SCO

Here goes my karma: Are you sure this code doesn't belong to SCO? I mean, we all know that all open source projects belong to them but we're hiding it. :)

Thanks, Intel...

I'm impressed. A real open-source driver from a major company...this shames the NVidias and the Lucents of the world who give stupid excuses for their closed-source drivers.

Looks like I'm going to be sniffing around for a refurbed IBM T41 ThinkPad with Centrino tech in the future.

Re:Thanks, Intel...

this shames the NVidias and the Lucents of the world who give stupid excuses for their closed-source drivers.

Better than no driver at all...

Re:Thanks, Intel...

Barely. I'd pine for the closed NVidia driver if I were a gamer (I still wouldn't use it, but I'd at least be tempted). For 2d stuff, XFree's 'nv' driver is fast, rock solid, and works out of the box - and it's Free.

Closed source drivers are evil, and are in fact what triggered RMS to begin the Free Software movement. They encourage complacency while giving nothing.

Re:Thanks, Intel...

Read the article.....it's not really there yet.

Re:Thanks, Intel...

Stupid excuses like "this cost us millions to produce, so we're not going to give the code away to you and our competitors, which would eventually cause us to lose so much revenue we'd not be able to make any more cards/drivers for you at all"?

Re:Thanks, Intel...

Intel makes $0 from the sale of their drivers. Also the marginal cost to produce a Linux driver when you already have 1) a Windows driver, and 2) a staff of Linux hackers is very much lower than "millions of dollars". However, the marginal cost of sales of Intel Centrino laptops to Linux users will be several hundred dollars each.

The economics are pretty simple. Probably some large client like Goldman Sachs or a similarly sized outfit wants to run Linux on laptops and told Intel to get their act together.

Re:Thanks, Intel...

this shames the NVidias and the Lucents of the world who give stupid excuses for their closed-source drivers.

Quit bitching. IICRC, NVidia doesn't own all the code in their drivers and anyway, why should they be forced to disclose stuff they consider a trade secret? They provide solid, working drivers for an OS used by like 1% of the desktop market. That's pretty impressive, imo.

Open Source Driver + Firmware

Is this a full driver or is the firmware a subtle way of making a closed-source driver?

(Honest question)

Re:Open Source Driver + Firmware

Given the supposed lack of foresight in their hardware design that most wlan vendors have taken recently (using basically 'soft wlan' cards), it is probably more akin to a 'partially closed driver', in that you probably won't have access to the channel frequencies, adding new network modes (master, monitor, etc). HOWEVER given that, it should allow future patching to the kernel side of the driver to support whatever future interface changes happen to ensure the card won't suddenly become useless.

IMHO, this is what all wlan dealers should be doing... if you can't give direct access to the hardware due to possible legal/FCC constraints, then you should have firmware to handle the interfacing so that you can at least release firmware interface specs, and hopefully be able to cut down on cross development costs by having your firmware patches enhance both linux and windows functionality while stomping out mutual bugs.

Re:Open Source Driver + Firmware

to answer my own question (partly):
"As the firmware is licensed under a restricted use license, it can not be included within the kernel sources. To enable the IPW2100 you will need a firmware image to load into the wireless NIC's processors." From http://ipw2100.sourceforge.net/README.ipw2100 [sourceforge.net].
And look at the firmware license [sourceforge.net]!

Re:Open Source Driver + Firmware

Is this a full driver or is the firmware a subtle way of making a closed-source driver?

That's a rather more complicated question than you might think.

The way most wireless cards work is that there's some radio hardware, hooked up to a microprocessor on the card that handles the low-level 802.11 frames, and some host software that talks to the microprocessor.

The microprocessor --- which tends to be an embedded ARM, these days --- runs a tiny nearly-an-operating-system out of flash or RAM. If RAM, then you need to download the microprocessor's code when you power up the card. That's the firmware.

This has a number of advantages: it means that the crucial, real-time processing is done with a custom processor that doesn't have to worry about running user code; it means that the vendor can change the hardware without having to change the driver, because the driver's just talking to a well-defined interface provided by the microprocessor; and it means that it's much easier to make cross-platform drivers.

It also means that the vendor can hide stuff in the firmware that they really, really don't want the user to play with. Such as the power, channel and timing settings that are mandated by the FCC.

I don't know if there are any wireless vendors out there who actually release source code to their firmware. (I'd be interested to find out if there are.) Which means that the answer to your question is both yes and no: the firmware's not open source, but the driver is.

Re:Open Source Driver + Firmware

Not really. Most hardware nowadays contains firmware (modern wireless cards are often just ARM cores attached to a radio transmitter), but in many cases it's in ROM or flash and you've never noticed. Older wireless cards with entirely open drivers (such as the orinocos) had similar quantities of firmware, but the cards shipped with it in flash. Requiring it to be loaded by the OS makes hardware implementation slightly easier, and you can upgrade the firmware along with the drivers without involving potentially risky reflashing.

Would you consider Linux closed-source because on most hardware it requires a closed-source BIOS or firmware in order to boot?

(Yes, I know about LinuxBIOS. It supports a subset of x86 hardware)

No WEP

WEP currently no support
Notice how WEP support is not yet done [sourceforge.net].

Hardly Intel...

I fail to see how "Finally Intel has made their different announcements about Linux support for the WLAN part of the Centrino technology become true."

when the SourceForge web site clearly states in the first paragraph.

"This project was created to enable support for the Intel PRO/Wireless 2100 (IPW2100) mini PCI adapter. This project is intended to be a community effort as much as is possible given some working constraints (mainly, no HW documentation is available)"

Sounds like Intel haven't helped at all and some enterprising folks have done their own. Kudos to them, shame on Intel.

And shame on Werner and Timothy for getting basic cursory facts right. Unless of course the SF website is failing to give credit to Intel.

Re:Hardly Intel...

My first reaction was the same as yours... But if you browse at the end of the page and hover your mouse on the maintainer name, you'll see that he has an Intel e-mail address.

So yes Intel is, kind of, supporting Linux driver for the Centrino chip as the pay the guy...

However, I don't beleive this is a priority for them. If it was so, they would have released something that is fully functional... What it seems to me is that they are paying one guy to do it and hope the OS community will jump in and help them out! I don't see any real corporate backing behind this project.

Bzzt. Wrong. Look who's doing this.

Read the copyright on the source code, and look at the contact info posted on the sf site. It's intel. (Hint: "Copyright 2003 - 2004 Intel Corporation" and the contact is jketreno AT linux.intel.com)

Just because they aren't loudly tooting their own horn by splashing "intel" all over the sf.net website doesn't mean they're not helping/having their people do the work. What you saw simply means they haven't been able to work out how to get the HW docs out the door to the community, and are being candid about this in the first sentence of their page.

And shame on you for making bad assumptions about helpful people, and unfairly criticizing an accurate news article.

I suppose I may have been trolled here, and I hate to bite, but this needs to be corrected :-)

if only....

broadcom would follow intel's lead and release a linux driver. while driverloader and ndiswrapper work, it would be nice to see the hardware vendor stop making crappy excuses (fcc regulations other stupid ones) about releasing a linux driver.

I don't get it

The contact email address is ipw2100<dash>admin<at>linux<dot>intel<dot>co m, the readme says it's copyright intel, but the home page says

This project is intended to be a community effort as much as is possible given some working constraints (mainly, no HW documentation is available)

So intel is not releasing the necessary documentation for the hardware, but has set up a SF project for "The Open Source Community" to figure it out?

Re:I don't get it

Maybe they are truly in touch with open source projects everywhere, and when it comes to documentation simply said "screw it, someone else will write it!"

Re:I don't get it

They probably can't release the documentation for some reason, however as long as there are a number of intel people on the project _with_ access to the documentation this isn't as huge a problem as it would otherwise be.

This allows the community to help stear the portions of the code that don't require the documentation and to help them properly tie the driver into Linux.

As long as the code isn't a complete mess it will also be possible to get some understanding of the workings of the chip from the code.

I agree that it is not ideal, however it's better than a binary-only driver.

Hooray!

Maybe now I'll reconsider buying that Thinkpad over a Powerbook for a split second.

Don't bother yet, its not finished

TODO

- long/short preamble support
- enhance wireless extension support
- adhoc
- encryption (WEP)
- continue to add support for addtional SW RF kill switch implementations
- "shared" authentication
- transmit power control
- power states support (ACPI)

Yes you read that right. So is there anything this driver does do?

After promising and promising to support Linux we get this. A crappy not finished driver. I suppose I'm supposed to be happy that Intel finally started to work on this after like what, a year after we should have had support? Sorry Intel but screw off. I already bought a PCMCIA Wireless NIC. And I'm sure as heck not going to replace it with you crappy nic and unfinished drivers. Thanks for nothing. Next notebook I buy is going to be AMD powered.

Re:Don't bother yet, its not finished

and that's why it hasn't been announced apart to a list mainly inhabited by developers

Uh, this is people getting fed up

And building their own driver.

Like the eepro100 driver from before? Or those Texas Instruments wireless chipsets in the DLink 650+? And a whole mess of other drivers for other devices from hardware companies that won't release technical specifications. Heck, are Broadcom 11g drivers out yet?

Good news, but...

...I'm curious why it took so long for this to finally happen. Intel knew, for a long time, that there was extensive interest.

The Centrino is a good chipset, and Centrino-based laptops are fairly popular. Even without the wireless support, I've been happy using a Linux-based Centrino laptop for the last six months. The lack of wireless access was the one thing that had been sticking in my craw.

Now, I'll be able to unequivocally recommend these laptops to friends who use Linux. This will mean more sales for Intel. This, I would think, would be considered a Good Thing (tm). So why the wait?

Wireless extensions

Anyone know if this implementation uses wireless extensions? Will these drivers use iwconfig and the rest of wireless tool or will you have to use some proprietary intel (probably binary only) tools? If it doesnt use wireless extensions, all of the neat scripts that come with stock distributions (debian, redhat, etc) wont work without some modification.

Open Source??

If it's like many "open source with firmware" drivers, it's probably a lot like this:


unsigned char firmware[] = { 0x22, 0x45, ...[many thousands of bytes].... };

void driver(void)
{
run_firmware(firmware);
}


Uh, yea, I'd consider that open source all right...

Re:Open Source??

Keyspan [keyspan.com] USB to serial converters are like this as well. This sparked a lot of debate on lkml on whether the firmware, clearly not open source, could be included in the kernel driver code. The upshot of that lengthy discussion [debian.org] was that yes, firmware can be bundled in the kernel code since it's not actually run by the host processor that's running the kernel.

*BSD Driver?

Thats really nice the released for linux, but how about us FBSD folks.. or are we out of luck on this one...

not excited

Only a year after there was an official announcement for linux drivers, which was later recanted, intel releases incomplete drivers. I'm sorry but Intel could of handled this situation a LOT better. I feel really sorry for the people who have had a "centrino" laptop for the past year and a useless wifi card.

When I purchased my X31 from IBM a year ago, instead of going for a wireless option, I bought the machine "wireless rdy" and put in my own linux compatible prism2 minipci card, purchased off ebay. Because of this incident, I will certainly stay away from purchasing any item from intel where linux support is promised in the near future.

Hopefully companies like Intel will start to realize that Desktop Linux is here and people who are decision makers & influencer's in IT make up a significant portion of the desktop linux populous.

Open source?

More like an open-source interface to a closed-source firmware.
You still have to go here [sf.net], agree to a EULA and download a binary image to be able to use this module (I found it humorous that Intel's download site admonished me for using Firefox on linux, and suggested I upgrade to IE6 or NS6).
You use the driver by doing:
modprobe ipw2100 firmware=/usr/share/firmware/ipw2100-1.0.fw
where ipw2100-1.0.fw is the current binary firmware image.

Re:Open source?

Pray tell, on which non-x86 arch are you going to use Intel Centrino drivers for the INTEL Centrino processor on the INTEL board?

The problem I have with Centrinos WLAN...

The problem I have with Centrinos WLAN is the fact that due to its embedded nature, I've yet to see any with antennas. I'd like to put a gain on my WiFi laptop around campus as the AP are far and wide. Why don't they make a standard jack for an antenna or something? It'd be quite useful other than being limited to "the AP inside the same room" sort of thing.

One word.

Progress.

Inspiring

Wow now this is interesting. Intel have decided that they're going to call on the massive contingent of Open Source developers for their Linux driver. This benefits them because they don't have to hire programmers or support the drivers and can outsource it to the community at large who will maintain it. Under license, they still own the drivers, but the public who want the drivers are developing them effectively for Intel.

Has there been an instance of this before? OEMs don't usually endorse open source dev projects for their hardware, but if more OEMs did do this then we'd see a huge amount of support for devices that are currently not compatible with various operating systems, and an extended development life for drivers for aging hardware. Obviously this method applies to not only "classic" open source OS's such as Linux, but Windows as well. After all, the OEM isn't selling the drivers, it's selling the hardware and firmware.

I'd love to see more manufacturers posting the source to their drivers and the developer documentation to allow their drivers to be improved and worked on by the willing open source community at large, while the OEM maintains endorsement and ownership of the developed software. It seems to me this method harms nobody and benefits all.

standards?

/etc/firmware/ipw2100-1.0.fw

wow, so now firmware is considered a config file that belongs in /etc. Im glad to see people are following the LSB.

Opensource Firmware

From what I'm reading on other posts it seems like the open source part is just the interface to the firmware. Is it possible to write a totally new open source firmware that would