Well I am here reading the latest kernel news and something strikes me as strange. Intel has agreed to help with drivers for linux but yet none of the wireless products have been added to the linux kernel. Were still stuck with the same old drivers in the wireless tree. Its almost as if the person who is maintaining that part of the kernel has given up trying to add new things and has moved on and forgot about this. I know that there has been lost of work being done trying to improve the sound for intel HD audio but it seams wireless has gotten a cold shoulder in recent days. please post your comments or reasons to why this is the case.
I apologize to anyone if this seams trivial but if you have ever tried installing or getting intel wireless working without it being included in your distribution then you know what I am talking about when you say it would be desirable to see it ported to the kernel instead.

Does intel wireless have open source drivers?
Have the drivers been submitted to the kernel project?

eg
http://kerneltrap.org/node/6270
the ipw3945 driver constains a binary blob (i.e. it is not open source).

Can you find an intel wireless driver which does not need the blob?

There are actually some big changes afoot in the kernel, but at the moment they are all contained in John Linville's branch. Quite awhile ago, Devicescape released their stack code under GPL, and the kernel wireless team decided to ditch SoftMAC and move everything to Devicescape since Devicescape supported more functions and was further along in development. So a lot of the driver development you've been hearing about is being done on the Devicescape stack, and until that gets merged into the main branch (which could be any version now) a lot of these drivers won't be released.

If you're adventurous, you could try creating a kernel off of John Linville's branch. You just need to have a working knowledge of git to patch the kernel source.

I thought that this was just a microcode image and not a running damon on the system?

I've noticed this as well. The Intel wireless drivers supplied in the recent kernels are old and broken in some cases. In the meantime, until this better kernel version comes out, you can replace the old driver modules with the ones here:

http://ipw3945.sourceforge.net/
http://ipw2200.sourceforge.net/
http://ieee80211.sourceforge.net/

The last one not being an Intel driver, but a necessary requirement nonetheless.

Regards,
Sam

Unfortunately, the binary blob is Intel's way of meeting their FCC requirements regarding output power and frequency. I suppose microcode/firmware would be another way to go, but they cannot (by federal law in the US) release any tools that would allow someone to modify the frequency or output power of their wireless card.

Yeah its unfortunate that this is the case but intel has made cards that were previously unusable usable for the few that can muddle though there walkthough on there site. It badly written and seams to be from the point of view of somone wth a specific distro and not general build instructions. One of the key things it misses is what you need to enable in the kernel in order to get TKIP encryption to work. I was stuck with a working card unable to get my encryption working with anything and found later that you need to enable something in the kernel to make it work. Now that its working it works great and actually works a little better then the windows driver as the light on the laptop actually gives you the status of the connection rather then just a pretty blue light that never turns off.

I would be REALLY hesitant to suggest replacing the ieee80211 subsytem. I know the Intel driver directions say to do so, but I personally think that they ought to be flogged for that. You need to check that you have a recent version of the subsytem and any even fairly recent kernel should be able to handle the ipw drivers without replacing the ieee80211 subsystem.

There are many threads around here of people borking their kernel pretty seriously be doing this.

I emailed the person at intel that is supposed to be maintaining the build instructions for at least the ipw3945 drivers and he said that the 2.6.20 kernel has a working sub system that can be used with the drivers.

Really? Interesting to hear this. I've replaced these modules on several systems, along with the intel drivers, and have achieved stabler networking and wireless. Should I be watching for a certain kind of damage or "borking" that these modules can do? Or does that damage occur only on a specialized setup?

I can see how improperly replacing the modules could cause problems. It's not a very familiar install procedure. But when I installed them the right way, everything became seamless, and so far there have been no errors.

Wouldn't it also help that they're a higher version than what's in the kernel?

Sam

You're right that it is very possible to replace the subsystem successfully. However, at least in my opinion, it is something that shouldn't be done unless you really need to and have a good grasp of patching and compiling kernels. I've seen several threads (granted, they were from new users and most of them were Fedora to boot) where the only way out was a complete re-install because they had messed up the kernel. They could boot, but their network services were completely screwed and they were too new to Linux to back out of what they had done, even with hand-holding.
One of the reasons I settled on Slackware as my distro is that I thoroughly believe that newer does not necessarily equate to better. The question really is does the existing kernel have a 802.11 subsytem capable of supporting the driver? If yes, then tinkering with the subsystem probably isn't a good idea, even if something newer is out there. If no, then you obviously have to replace the subsytem.

My beef with the Intel drivers site is that the present replacing the 802.11 subsytem as a MUST rather than a MAYBE. Given that wireless in Linux is a gigantic pain in the butt, and that it is one of the first serious problems that a new Linux user is likely to grapple with, I think their advice is irresponsible.

OK, thanks for the response. So then the 80211 subsystem currently in the kernel is sufficient for the latest released intel drivers. I suppose it makes sense to me now when I look at the latest 80211 downloads.

Now that I look at their download site again, their changelog does recommend that if anyone's using a modern kernel, to stick with the ieee80211 that comes with it, and not to bother replacing the subsystem. They just need to update their install README.

Sam

Very good information to know. I too have been replacing the ieee stack for my ipw2200. I am glad to know that I do not need to do this anymore as I always compile an up to date kernel. Thanks, good thread.

Yeah the main page for these drivers are a bit dated from Intel.