Explaining WLAN Chips' Poor Linux Support

morcheeba writes "Kernel Traffic is reporting (mirror mirror list) that 'Some WLAN Chip Specs Secret To Protect Military Communications.' While this is stretching it a bit -- these radios are generally limited to a narrow frequency range and few modulation types -- software can cause illegal radio operation, especially when the laws vary by country. Is Linux support for 802.11g and Centrino chipsets going to be delayed by manufacturers afraid of FCC harassment? An interesting discussion on the future of Openness in radio chipsets." Interesting comments from Alan Cox in here about just how flexible some of these chips are.

Comments

I thought it was just about money

[Anonymous Coward]

If 95%+ of your market will be Windows users, it might make financial sense to just worry about that 95% and develop support for them.

Re:I thought it was just about money

[afidel]

It's really not that expensive, if you capture one large customer it's probably worth it. Cisco/Aironet has one coder who does the driver and support software for Linux and one primary tester who spends a lot of his time on testing the Linux stuff. These two together probably make somewhere in the $200K range when benifits are added in, not a whole lot when yearly sales are around $100 million. In addition there is a community developed driver made from resources that Cisco/Aironet made available to an outside developer.

Re:I thought it was just about money

[g4dget]

If you make it only possible to use your products on Windows, it isn't surprising that 95% of your market is on Windows.

If your "95%" figure is supposed to refer to the fraction of desktop users using Windows within the population, your number unsubstantiated and probably erroneous. Microsoft's market share is usually overestimated because many desktop uses of other operating systems aren't counted towards those other operating systems, many non-desktop uses of Windows are counted as Windows users, and many non-Windows users are counted as Windows users because they have a Windows license that they don't, or don't want to, use.

prism2 only!

[drwho]

I am not the only one unhappy with the very poor support of linux by many wifi chip vendors. Intersil seems to be the only one even close to being open, and you have to sign all sorts of agreements to get the specs (legally). My understanding of this is that a lot of the functions are being moved from the chip hardware to system software, so it's not just a device driver required to use these newer cards. Vendors don't want everyone to see their programming.

The prism2 were the first really popular wireless cards, partly because of low cost but also because of the ability to write drivers for them. I wish other manufacturers wouldn't be so reticent about their support. I actually prefer it if they keep the cards smart and the systems dumb, because it increases portability and compatibility. It probably adds to the cost though.

Time

[insecuritiez]

When I went wireless on my Linux laptop about a year ago the only card I could get supported was the Lucen Orinoco. It's a great card. But now, many months later there is the WLAN project and many many cards are supported. The future will be the same. More support is coming. The major problem I see though is the manufaturers. They write drivers for windows. Most of them arn't about to write them for linux. Meaning that every piece of hardware will be a few months behind while a group of dedicated programers do the dirty work for the company.

Re:Time

[insecuritiez]

Because the company gets respect and it's name out there amonge a group of power users and frequent buyers. Look at NVidia. They sure get a lot of coverage on Slashdot. Much of this is due to their great Linux drivers. If they made great cards that wern't supported or had poor/shotty support many fewer Slashdot readers would care one witt about what NVidia does. When Apple embraced open source all of a sudden they got a whole second army of geeks following. Why does it make sense to write drivers in support of linux? Because the 5% of the people that will use those products have a lot of influence in a much larger comunity. (IT, business settings)

Why not simply make it illegal to operate?

[Fefe]

I can buy a regular transmitter and modify it to transmit on a forbidden frequency. Does that mean we can't sell transmitters or books telling people how to build one? No!

I think this is a straw man argument.

Re:Why not simply make it illegal to operate?

[alienw]

Don't expect the manufacturer to help you modify it, though. If you want to do that, you would have to reverse-engineer the radio yourself. Then you can do whatever to it. Similarly, if you reverse-engineer a wifi driver, you can write your own. It's just that the manufacturer won't help you.

It revolves around the business model

[jj_johny]

I was looking into doing some WLAN engineering but found that it was too difficult to do without getting hold of the documents. Since these chips are used primarily in high volume applications - i.e. OEMs using 10 of thousands at a time - the manufacturers don't give out the specs to the public. First, they don't want to deal with questions from someone who is not paying them for their engineering support services. And second, they don't want the competition getting hold of the future direction in the product that may be hinted at in their technical documentation. All the manufacturers that I know of require you to sign an NDA to get a copy.

The only real leverage that anyone has is only buying products that have explicit Linux support from the OEM.

Re:Promotion?

[ivan256]

How will advocates convince home users (a large chunk of that 95%)

Home users are not the issue here (and not the largest part of the 95% anyways). Businesses that may have, internally, 99% of their hardware running windows and 1% and growing running something else may choose another manufacturer's product because it supports 100% of thir machines. All you need to loose is one big sale and it would have been worthwhile to pay a developer a few thousand dollars to write a Linux driver. Are you saying companies should give up on trying to maximize profits, and start shooting for 95% instead?

So very true...

[GarfBond]

I'm going to give up mod points here just to chime in, since this is an issue I care about.

Anyone looking to get good wireless card support (802.11b) should buy one with a prism2 chip or an Orinoco. I know many that have had good luck with these cards, and I know for a fact that the Orinoco cards are essentially plug and play in linux. Do NOT buy the TI chipsets (sometimes marketed as 22mbps 802.11b+) or the Broadcom chipsets; word on the street (heh) is that these companies have been less than forthcoming with specs so people can write proper drivers for them.

It's too bad that this is the sad state of wireless support in linux, that we must be at the manufacturer's mercy to get our hardware working properly. I've been waiting for 2 years now to get my USB wireless card (oh yeah, avoid those too if you can) working in linux, and it's all because the company doesn't care.

Re:eh

[Anonymous Coward]

Yours is not a 802.11g, is it?

Re:eh

[Eneff]

Oh wait, I have an idea!

Let's require everyone to spend a week learning fundamentals of Unix and 500 bucks and another week putting together OpenBSD-compliant parts (not to mention the day or two researching what exactly will and will not work on BSD) just to give up because they can't figure out why the fuck EverQuest won't work! (Adknowledged, the real problem is that most consumer programs were developed with a direct connection and a modem in mind. Guess what? That's irrelevant.)

Yeah, sounds REAL swell.

Yes, Those Netgear and LinkSys boxes are really routers. They really work. Until you can tell me how to get a wireless solution up and running with BSD for an hour's work and under 150 dollars, your argument is irrelevant.

Exactly

[FreeLinux]

It's all about the money. People seem to be unwilling to accept that Linux is such a tiny segment of users, that it is simply not worth the effort for a hardware manufacturer to develop a Linux driver, let alone multiple drivers for Red Hat, Mandrake, SuSE, Gentoo, Debian, etc. with kernel versions 2.4.1, 2.4.8, 2.4.12, 2.4.18, 2.4.21 etc.

These same poeple also refuse to see that the chip manufacturers make their money by selling the chips and the software that drives them. Releasing the chip specs, in many manufacturer's minds, gives their competition an advatage because said competition will have access to the specs too. Theoretically, the competition could then build a similar chip for less or even a better chip without as much expenditure for R&D.

Regardless of the truth on this matter there is simply no economic incentive for the manufacturer to release the specs. It has nothing to do with military secrets or national security or anything else, it is simply a matter of economics. But, all is not lost. As Linux continues to grow and its "market share" increases the economic incentive will increase for manufacturers to develop the drivers or release the specs.

Re:eh

[Anonymous Coward]

buy OpenBSD CD = U$S 40 + Shipping.
Follow fold out instructions on old pentium 120 ~ 1h.
Edit /etc/sysctl.conf
net.inet.ip.forwarding=1
net.inet6.ip6.forwarding=1

Edit /etc/rc.conf
change "PF=NO" to "PF=YES"

Edit /etc/pf.conf

$ext_if="rl0"
$int_if="rl1"

nat on $ext_if from $int_if to all -> ($ext_if)

block in on $ext_if all
pass out quick on $ext_if proto {tcp,udp,icmp} all keep state

#reboot

Listo!
Under U$S 50.
Under 1h 30m.
Only required knowledge: vi

Re:eh

[Saint Stephen]

If you believe all systems are crackable, how about this theory:

The problem of breeding humans who do not wish to commit havoc is of equal difficulty to the problem of creating uncrackable systems. (I.E., both are nearly impossible.)

Therefore, it is equally useful to spend one's time trying to convince people to "just not be an asshole" as it is securing systems. This cuts both ways -- privacy and security. I once made this statement in defense of things like Hailstorm -- the fact that the government/organizations have all my personal data does not in and of itself harm me, only the intentions of those that hold that data.

For example, your employer typically knows your Checking account routing #s, SS#s, heath data, but doesn't abuse the knowledge, because they have a self-serving interest not to. Why not imagine that situation everywhere -- everybody knows everything, nobody abuses knowledge. This position is no more bullshit than trying to secure everything, both are nearly impossible!!!

[grin]

Re:Secret because of the military?

[Anonymous Coward]

The whole reason is there in the phrase "shot down". Even if the enemy can't decipher your transmission, its mere presence lets them:

Know that you're around,

Determine where you are (direction finding and triangulation),

Know when you're up to something (increased traffic density)
It's a "compromising emanation". "Your son died because the enemy had a WLAN card and a Pringles can."