Skip to contentSkip to conversation

Will there be Intel Wireless 3945ABG support?

DISCUSSION Joseph Michael Smidt 16 participants 24 messages Jul 11, 2006 – Jul 12, 2006

lore
Share & tools
lore
#0 lore raw

Will 2.6.18 or 2.6.19 support Intel Wireless 3945ABG? Please cc me since I am not subscribed. Thanks.

Joseph Smidt
------------------------------------------------------------------------
Joseph Smidt

422 Wymount
Provo, UT 84604
phone: 801-371-5564
email: [email protected]

#1 lore raw
Quoted 2 lines · Joseph Michael Smidt

Will 2.6.18 or 2.6.19 support Intel Wireless 3945ABG? Please cc me since I am not subscribed. Thanks.

There's a module available[1] for us but it's not merged in mainline yet.

1. http://ipw3945.sourceforge.net/

You could use it in meanwhile.

Signature · 8 lines
        O T A V I O    S A L V A D O R
---------------------------------------------
 E-mail: [email protected]      UIN: 5906116
 GNU/Linux User: 239058     GPG ID: 49A5F855
 Home Page: http://www.freedom.ind.br/otavio
---------------------------------------------
"Microsoft gives you Windows ... Linux gives
 you the whole house."
#2 lore raw
In reply to Joseph Michael Smidt
Quoted 1 line · Joseph Michael Smidt

Will 2.6.18 or 2.6.19 support Intel Wireless 3945ABG? Please cc me since I am not subscribed. Thanks.

It will not be in 2.6.18. Making 2.6.19 is not out of the question, but it may take some work.

Hth!

John

Signature · 2 lines
John W. Linville
[email protected]
#3 lore raw
Quoted 6 lines · John W. Linville
Quoted 2 lines · Joseph Michael Smidt

Will 2.6.18 or 2.6.19 support Intel Wireless 3945ABG? Please cc me since
I am not subscribed. Thanks.

It will not be in 2.6.18. Making 2.6.19 is not out of the question, but it may take some work.

Has some agreement been met regarding the mandatory use of the binary regulatory daemon? The webpage seems to suggest it is still necessary, and I'm sure that would disqualify merging the driver with Linux proper.

Signature · 5 lines
Cheers,
Alistair.

Final year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.
#4 lore raw
Quoted 9 lines · Alistair John Strachan
Quoted 5 lines · John W. Linville
Quoted 2 lines · Joseph Michael Smidt

Will 2.6.18 or 2.6.19 support Intel Wireless 3945ABG? Please cc me since
I am not subscribed. Thanks.

It will not be in 2.6.18. Making 2.6.19 is not out of the question, but it may take some work.

Has some agreement been met regarding the mandatory use of the binary regulatory daemon? The webpage seems to suggest it is still necessary, and I'm sure that would disqualify merging the driver with Linux proper.

Why not?
The whole point is running a system that you know you can support for many years, even without a vendor support...
And to have a system that you know exactly what running in it... Having a binary closed source violate this.

Also there is no good reason why supplying this daemon as closed source... All they wish is people don't mess with their frequencies, and sooner or later someone will...

Best Regards,
Alon Bar-Lev.

#5 lore raw

Ar Maw, 2006-07-11 am 21:25 +0300, ysgrifennodd Alon Bar-Lev:

Quoted 2 lines

And to have a system that you know exactly what running in it... Having a binary closed source violate this.

Also if the binary only chunk of code is neccessary to make the open source bit work then its a derivative work as I understand the situation, which makes it all rather questionable from a licensing perspsective.

Hopefully Intel will find a sensible solution to the problem or someone will just reverse engineer it away.

Alan

#6 lore raw

Le Tue, 11 Jul 2006 19:55:19 +0100, Alan Cox a écrit :

Quoted 4 lines

Ar Maw, 2006-07-11 am 21:25 +0300, ysgrifennodd Alon Bar-Lev:

Quoted 2 lines

And to have a system that you know exactly what running in it... Having a binary closed source violate this.

Quoted 8 lines

Also if the binary only chunk of code is neccessary to make the open source bit work then its a derivative work as I understand the situation, which makes it all rather questionable from a licensing perspsective.

Hopefully Intel will find a sensible solution to the problem or someone will just reverse engineer it away.

Well openbsd guys already reversed engineer it :
http://kerneltrap.org/node/6650

#7 lore raw
Quoted 11 lines · Matthieu CASTET
Quoted 9 lines · Alan Cox

Ar Maw, 2006-07-11 am 21:25 +0300, ysgrifennodd Alon Bar-Lev:

Quoted 2 lines

And to have a system that you know exactly what running in it... Having a binary closed source violate this.

Also if the binary only chunk of code is neccessary to make the open source bit work then its a derivative work as I understand the situation, which makes it all rather questionable from a licensing perspsective.

Quoted 5 lines
Quoted 3 lines

Hopefully Intel will find a sensible solution to the problem or someone will just reverse engineer it away.

Well openbsd guys already reversed engineer it :
http://kerneltrap.org/node/6650

[quote]

Porting wpi

When asked the likelihood that the wpi driver would be ported to Linux, Damien thought that this was unlikely. "I doubt the Linux community will ever leverage this work since Intel is developing a 802.11 layer for the Linux kernel," he said, "Linux kernel developers probably can't afford to upset Intel."

[/quote]

Huh ?

Signature · 1 line
What this world needs is a good five-dollar plasma weapon.
#8 lore raw
In reply to Alan Cox
Quoted 2 lines · Alan Cox

Hopefully Intel will find a sensible solution to the problem or someone will just reverse engineer it away.

The ipw3945_daemon.h file includes a pretty full description of what the daemon has to do, and all the structures have nice friendly names. There's also a changelog of the protocol. It shouldn't take someone long.

Signature · 1 line
Matthew Garrett | [email protected]
#9 lore raw
Quoted 7 lines · Matthew Garrett
Quoted 2 lines · Alan Cox

Hopefully Intel will find a sensible solution to the problem or someone will just reverse engineer it away.

The ipw3945_daemon.h file includes a pretty full description of what the daemon has to do, and all the structures have nice friendly names. There's also a changelog of the protocol. It shouldn't take someone long.

It's already been done, too.

-hpa

#10 lore raw
Quoted 8 lines · H. Peter Anvin
Quoted 5 lines · Matthew Garrett

The ipw3945_daemon.h file includes a pretty full description of what the daemon has to do, and all the structures have nice friendly names. There's also a changelog of the protocol. It shouldn't take someone long.

It's already been done, too.

For Linux, too? The OpenBSD approach seems to have been to integrate it into the driver, which is fine for them but would probably make it more awkward for us to keep things synced with the Intel code.

Signature · 1 line
Matthew Garrett | [email protected]
#11 lore raw
In reply to Alan Cox
Quoted 7 lines · Alan Cox

Ar Maw, 2006-07-11 am 21:25 +0300, ysgrifennodd Alon Bar-Lev:

Quoted 2 lines

And to have a system that you know exactly what running in it... Having a binary closed source violate this.

Also if the binary only chunk of code is neccessary to make the open source bit work then its a derivative work as I understand the situation,

Following that logic one must draw the conclusion that the firmware that runs on a scsi controller is derived from the driver provided with the kernel that communicates with it.

The obvious distinction between scsi firmware and the regulatory daemon blob being discussed here is that the regulatory daemon runs on the host vs. an adapter. However, if you consider the communication interface between the kernel and the user space daemon to be analogous to the communication interface between the kernel driver and the firmware that runs on an adapter, then the distinction of running on the host is irrelevant.

Quoted 2 lines

which makes it all rather questionable from a licensing perspsective.

There are no questions from a licensing standpoint.

The regulatory daemon is in no way derived from any GPL work, and the GPL driver is provided with everything it needs to talk to the interfaces defined through the ipw3945_daemon.h header file.

Quoted 2 lines

Hopefully Intel will find a sensible solution to the problem or someone will just reverse engineer it away.

Invariably someone will make such a piece of code available on Linux.

To that end I would encourage anyone that may be interested in using such a piece of code to read the regulatory notice packaged with our drivers, and linked for your reference here[1].

James

1. http://bughost.org/ipw3945/NOTICE

Quoted 2 lines

Alan

#12 lore raw

From: James Ketrenos <[email protected]>
Date: Tue, 11 Jul 2006 17:35:48 -0700

Quoted 7 lines

The obvious distinction between scsi firmware and the regulatory daemon blob being discussed here is that the regulatory daemon runs on the host vs. an adapter. However, if you consider the communication interface between the kernel and the user space daemon to be analogous to the communication interface between the kernel driver and the firmware that runs on an adapter, then the distinction of running on the host is irrelevant.

The core issue is whether the userland blob could stand alone and is something that could exist independant of Linux and the kernel side GPL'd driver.

Difficult areas arise when you design a set of interfaces specifically to talk to the binary blob, and which exist for no other purpose and do not provide some well defined "standard" API such as the SCSI command set, as an example. It is just a backdoor into the binary blob, and therefore this could make it more likely to be considered a derivative work.

Firmware that runs on the card is also a tricky area, and not everyone agrees on that issue as well.

Unfortunately, none of this is fun stuff :-/

#13 lore raw
In reply to James Ketrenos

Ar Maw, 2006-07-11 am 17:35 -0700, ysgrifennodd James Ketrenos:

Quoted 3 lines

The obvious distinction between scsi firmware and the regulatory daemon blob being discussed here is that the regulatory daemon runs on the host vs. an adapter.

"It isn't a pirate copy of the movie because my copy is on VHS and theirs is on DVD"

Derivative works are not really a question of where something is, think about a distributed computation, or a tool which partly compiles a program into VHDL for high performance execution.

I'm not sure this is a useful kernel list argument anyway since I am sure Intel legal will have considered the question and reached their own conclusions and also vendors and Intel will continue beating each other up until a neat solution is found.

Quoted 1 line

There are no questions from a licensing standpoint.

For Intel sure, if it owns all the bits then it can do what it likes with its bits.

Quoted 3 lines

To that end I would encourage anyone that may be interested in using such a piece of code to read the regulatory notice packaged with our drivers, and linked for your reference here[1].

Oh I understand *why* the issue arises, and as a licensed radio amateur I'm quite well aware of the concerns. I'm also permitted to use 2.4Ghz at higher power for amateur purposes.

Alan

#14 lore raw
In reply to James Ketrenos
Quoted 3 lines · James Ketrenos

The regulatory daemon is in no way derived from any GPL work, and the GPL driver is provided with everything it needs to talk to the interfaces defined through the ipw3945_daemon.h header file.

You are kidding, right?!?!
If you cannot operate the device using GPLed software!

What use is that software if after using all the GPLed available drivers for the device it still does not work, and need a properiteary modules in order to complete it? The derived work is taking something that *WORKS* and making some modifications... In this case you don't even have something that *WORKS*.

Best Regards,
Alon Bar-Lev.

#15 lore raw
In reply to James Ketrenos
Quoted 4 lines

Also if the binary only chunk of code is neccessary to make the open source bit work then its a derivative work as I understand the situation,

The issue is not whether the userspace daemon is a derivative work of any
GPL'd work and therefore must be covered. The issue is whether the kernel driver and the userspace program are one work or two. If neither is any use without the other, and the two were designed together, it seems implausible to argue that they are two works.

A boundary that consists of an API that allows the work on either side to
be interchanged with others and the other work still operate substantially the same can certainly divide two works. That's why a C program that uses the standard library can be a separate work from the standard library. Any C program works with that library, and the library works with any C program.

A boundary custom designed for these two programs, and which is not
intended to allow either program to be replaced with another, would not seem to do the job.

DS

#16 lore raw
In reply to Alon Bar-Lev
Quoted 19 lines · Alon Bar-Lev
Quoted 11 lines · Alistair John Strachan
Quoted 6 lines · John W. Linville
Quoted 2 lines · Joseph Michael Smidt

Will 2.6.18 or 2.6.19 support Intel Wireless 3945ABG? Please cc me
since I am not subscribed. Thanks.

It will not be in 2.6.18. Making 2.6.19 is not out of the question, but it may take some work.

Has some agreement been met regarding the mandatory use of the binary regulatory daemon? The webpage seems to suggest it is still necessary, and I'm sure that would disqualify merging the driver with Linux proper.

Why not?
The whole point is running a system that you know you can support for many years, even without a vendor support...
And to have a system that you know exactly what running in it... Having a binary closed source violate this.

Quoted 3 lines

Also there is no good reason why supplying this daemon as closed source... All they wish is people don't mess with their frequencies, and sooner or later someone will...

Don't misinterpret me, that's exactly my feeling too. Ignoring the politics and questionable legality entirely, as a matter of practicality, when Intel inevitably abandon these cards 5 years from now and their daemon rots, will I still be able to use my wireless card? I think not.

Also, I can't see any reason why a mini-pci version of this couldn't work in a system other than x86/x86-64, even if Intel don't design such machines.

In my view, it would be unacceptable to merge.

Signature · 5 lines
Cheers,
Alistair.

Final year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.
#17 lore raw
In reply to Alon Bar-Lev
Quoted 5 lines · Alon Bar-Lev

Also there is no good reason why supplying this daemon as closed source... All they
wish is people don't mess with their frequencies, and sooner or later someone will...

Using interesting frequencies or output power would be fun for radio amateurs (like me). 2.4GHz is one of our playgrounds after all :-)

Just because Joe Average isn't allowed to use such features doesn't mean that there aren't any legitimate users for it.

Preventing the accidental use of unauthorized features would be enough, I think (warnings that force you to look up the manual to find out the correct --force option or similar)
I expect developers to be sensible enough to only offer 'public legal' values in the default options list.

just my 2 cents,
Thorsten

Signature · 3 lines
| Thorsten Kranzkowski        Internet: [email protected]                      |
| Mobile: ++49 170 1876134       Snail: Kiebitzstr. 14, 49324 Melle, Germany  |
| Ampr: dl8bcu@db0lj.#rpl.deu.eu, [email protected] [44.130.8.19] |
#18 lore raw

On July 11 at 16:16 EDT, Thorsten Kranzkowski hastily scribbled:

Quoted 9 lines · Alon Bar-Lev
Quoted 5 lines

Also there is no good reason why supplying this daemon as closed source... All they
wish is people don't mess with their frequencies, and sooner or later someone will...

Using interesting frequencies or output power would be fun for radio amateurs (like me). 2.4GHz is one of our playgrounds after all :-)

Hear, hear!

Quoted 8 lines

Just because Joe Average isn't allowed to use such features doesn't mean that there aren't any legitimate users for it.

Preventing the accidental use of unauthorized features would be enough, I think (warnings that force you to look up the manual to find out the correct --force option or similar)
I expect developers to be sensible enough to only offer 'public legal' values in the default options list.

Frankly, I think Intel is misinterpreting how strict the FCC is being (or maybe the FCC is being too strict). I would interpret their mandates as meaning that, as purchased, equipment can't transmit on unauthorized frequencies, and that it's not "user-modifiable". User modification doesn't include things like opening the case of a toy walkie-talkie up and swapping out a crystal, nor does it include things like opening up the firmware or driver for something and messing with it.

Frankly, I'm annoyed that, if Intel understood the full extent of the problem, that they didn't take a better approach and simply give the card a set of legal values. It doesn't need to understand the subtleties of what they mean. It just needs to know frequencies 1, 2, and 3 are okay, but not 4, 5, and 6, and that the max power is xx dBm.

OTOH, I'd love to get one of these cards and play with it. I'd love a WiFi card that can run up to 5 or 10 watts. (1500 is a little much, and my laptop's battery can't supply it :-)

-- Thomas Tuttle

#19 lore raw
Quoted 10 lines · Thomas Tuttle

Frankly, I think Intel is misinterpreting how strict the FCC is being (or maybe the FCC is being too strict). I would interpret their mandates as meaning that, as purchased, equipment can't transmit on unauthorized frequencies, and that it's not "user-modifiable". User modification doesn't include things like opening the case of a toy walkie-talkie up and swapping out a crystal, nor does it include things like opening up the firmware or driver for something and messing with it.

Unfortunately you're wrong. Some manufacturers have gotten rapped for marketing equipment that can be modded by stuff like desoldering diodes.

The FCC has some incredibly heavy-weight regulations, like:

- Problem: unlicensed high-power CB operation
- Solution: ban amplifiers that work anywhere near the CB band, including several amateur radio bands (even for sale to users with valid amateur licenses)

- Problem: unencrypted cell phones
- Solutions: ban scanners that work on the cell phone bands

... etc ...

-hpa

#20 lore raw
In reply to Thomas Tuttle

From: Thomas Tuttle <[email protected]>
Date: Tue, 11 Jul 2006 20:42:12 -0400

Quoted 5 lines

Frankly, I'm annoyed that, if Intel understood the full extent of the problem, that they didn't take a better approach and simply give the card a set of legal values. It doesn't need to understand the subtleties of what they mean. It just needs to know frequencies 1, 2, and 3 are okay, but not 4, 5, and 6, and that the max power is xx dBm.

You miss many important issues in your diatribe. I don't like the situation either, but I hold this position understanding the conditions (both technical and legal) under which companies such as Atheros and Intel must operate.

First off, the reason these radios are fully programmable, not fixed in on-board firmware or likewise, is so that people doing "special stuff" outside the normal operating frequencies and power levels, and have a license to do so, can use these wireless chips out of the box. Otherwise custom boards would need to be produced and that is prohibitively expensive and restrictive for what some of these folks want to do.

Such companies can thus provide firmware or drivers that operate within a customer's specially licensed frequency or power range once that customer proves they do indeed have a license from the FCC to use it.

Secondarily, it is up to lawyers, not you, to decide what is a safe manner for the maker of a wireless chipset to abide by the FCC regulations. And across the board, lawyers representing these companies and other entities seem to agree that providing the full source code to a wireless chip driver's radio programming makes it "user-modifiable", whereas hiding the radio programming behind a binary-only blob or firmware satisfies the FCC requirements.

And if you think they haven't invested any effort to look into alternatives that will satisfy both the FCC and the open source crowd, think again. You can be sure they've spent a lot of time thinking about how to deal with this. It is absurd to say things which suggest that these guys are sitting around twiddling their thumbs about the issue, and think the current state of affairs is ok.

It's not a matter of "impossible" vs. "possible" to modify the frequencies and power levels outside of the allowed range, rather it's a matter of making it "difficult enough" for an end user to modify these restrictions.

As long as it's Intel's or Atheros's ass that gets reamed by the FCC for running afoul of the radio frequency regulations, they will not be posting the source code to program their radios. On the other hand, if it happens to get legally reverse engineered, then unless these companies assisted in that reverse engineering effort, the FCC really couldn't go after them. Such companies would also not be able to participate in maintainence of a driver for their chips containing the reverse engineered components. However, we've dealt with that kind of situation just fine in the past :)

So we will be in this endless loop finding ways to legally reverse engineer binary blobs to get fully free wireless drivers, until the FCC regulation situation is rectified.

#21 lore raw
In reply to Thomas Tuttle
Quoted 2 lines · Thomas Tuttle

Frankly, I think Intel is misinterpreting how strict the FCC is being (or maybe the FCC is being too strict).

The FCC has in the past managed to come up with the most astoundingly brain-dead regulations (for instance, "fixing" the lack of any actual encryption on some cell and portable phone by banning scanners...)

Quoted 5 lines

Frankly, I'm annoyed that, if Intel understood the full extent of the problem, that they didn't take a better approach and simply give the card a set of legal values. It doesn't need to understand the subtleties of what they mean. It just needs to know frequencies 1, 2, and 3 are okay, but not 4, 5, and 6, and that the max power is xx dBm.

The problem is that in the US, 1,2, and 3 are OK, in Sweden 1,4,6 are OK, in Germany it's 2,3, and 7, and China insists on 1.5,2.5, and 3.5....

#22 lore raw
In reply to Thomas Tuttle
Quoted 21 lines · Thomas Tuttle

On July 11 at 16:16 EDT, Thorsten Kranzkowski hastily scribbled:

Quoted 8 lines · Alon Bar-Lev
Quoted 4 lines

Also there is no good reason why supplying this daemon as closed source... All they
wish is people don't mess with their frequencies, and sooner or later someone will...

Using interesting frequencies or output power would be fun for radio amateurs (like me). 2.4GHz is one of our playgrounds after all :-)

Hear, hear!

Quoted 8 lines

Just because Joe Average isn't allowed to use such features doesn't mean that there aren't any legitimate users for it.

Preventing the accidental use of unauthorized features would be enough, I think (warnings that force you to look up the manual to find out the correct --force option or similar)
I expect developers to be sensible enough to only offer 'public legal' values in the default options list.

Quoted 8 lines

Frankly, I think Intel is misinterpreting how strict the FCC is being (or maybe the FCC is being too strict). I would interpret their mandates as meaning that, as purchased, equipment can't transmit on unauthorized frequencies, and that it's not "user-modifiable". User modification doesn't include things like opening the case of a toy walkie-talkie up and swapping out a crystal, nor does it include things like opening up the firmware or driver for something and messing with it.

If you give Matthieu's link[1] a quick read, the OpenBSD developer that reverse engineered the regulatory blob seems to indicate that the FCC regulations are just an excuse, so Intel can hide their IP inside the blob.

I'm not sure to what extent this is accurate, but it would seem that you would leave the driver functionally impaired if you simply removed the regulatory checks.

[1] http://kerneltrap.org/node/6650

Signature · 5 lines
Cheers,
Alistair.

Final year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.
#23 lore raw
Quoted 32 lines · Alistair John Strachan
Quoted 22 lines · Thomas Tuttle

On July 11 at 16:16 EDT, Thorsten Kranzkowski hastily scribbled:

Quoted 9 lines · Alon Bar-Lev
Quoted 4 lines

Also there is no good reason why supplying this daemon as closed source... All they
wish is people don't mess with their frequencies, and sooner or later someone will...

Using interesting frequencies or output power would be fun for radio amateurs (like me). 2.4GHz is one of our playgrounds after all :-)

Hear, hear!

Quoted 8 lines

Just because Joe Average isn't allowed to use such features doesn't mean that there aren't any legitimate users for it.

Preventing the accidental use of unauthorized features would be enough, I think (warnings that force you to look up the manual to find out the correct --force option or similar)
I expect developers to be sensible enough to only offer 'public legal' values in the default options list.

Quoted 8 lines

Frankly, I think Intel is misinterpreting how strict the FCC is being (or maybe the FCC is being too strict). I would interpret their mandates as meaning that, as purchased, equipment can't transmit on unauthorized frequencies, and that it's not "user-modifiable". User modification doesn't include things like opening the case of a toy walkie-talkie up and swapping out a crystal, nor does it include things like opening up the firmware or driver for something and messing with it.

Quoted 3 lines

If you give Matthieu's link[1] a quick read, the OpenBSD developer that reverse engineered the regulatory blob seems to indicate that the FCC regulations are just an excuse, so Intel can hide their IP inside the blob.

Sorry, correction, reverse engineered the interface between the blob and driver, not the regulatory blob itself.

Signature · 5 lines
Cheers,
Alistair.

Final year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.