Skip to contentSkip to conversation

ipw3945 status

DISCUSSION Pavel Machek 11 participants 29 messages Jul 30, 2006 – Jul 31, 2006

lore
Share & tools
lore
#0 lore raw

Hi!

I'm unfortunate enough to have x60 with ipw card. Card works okay, but driver is 16K LoC and needs binary daemon (ugh). Plus, it lives as an external module...

Are there any plans to merge it into mainline? Is there any way I could help?
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#1 lore raw
Quoted 3 lines · Pavel Machek

I'm unfortunate enough to have x60 with ipw card. Card works okay, but driver is 16K LoC and needs binary daemon (ugh). Plus, it lives as an external module...

I have a mostly working replacement for the binary daemon, but it causes the firmware to crash at the point where it triggers an association. If anyone's keen on trying to figure out what's up, I'll clean up the code and stick it somewhere.

Signature · 1 line
Matthew Garrett | [email protected]
#2 lore raw

Hi!

Quoted 8 lines
Quoted 3 lines

I'm unfortunate enough to have x60 with ipw card. Card works okay, but driver is 16K LoC and needs binary daemon (ugh). Plus, it lives as an external module...

I have a mostly working replacement for the binary daemon, but it causes the firmware to crash at the point where it triggers an association. If anyone's keen on trying to figure out what's up, I'll clean up the code and stick it somewhere.

I can't promise anything, but yes, seeing that code would be great.
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#3 lore raw
In reply to Matthew Garrett
Quoted 9 lines · Matthew Garrett
Quoted 3 lines · Pavel Machek

I'm unfortunate enough to have x60 with ipw card. Card works okay, but driver is 16K LoC and needs binary daemon (ugh). Plus, it lives as an external module...

I have a mostly working replacement for the binary daemon, but it causes the firmware to crash at the point where it triggers an association. If anyone's keen on trying to figure out what's up, I'll clean up the code and stick it somewhere.

Why not get rid of the daemon like bsd did [0]? Otherwise in 5 years you'll have 42 daemons running which communicate with the firmware of various devices, each having a different inter- face.

Jan

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

#4 lore raw
Quoted 4 lines · Jan Dittmer

Why not get rid of the daemon like bsd did [0]? Otherwise in 5 years you'll have 42 daemons running which communicate with the firmware of various devices, each having a different inter- face.

Because it would involve a moderate rewriting of the driver, and we'd have to carry a delta against Intel's code forever.

Signature · 1 line
Matthew Garrett | [email protected]
#5 lore raw
Quoted 9 lines · Matthew Garrett
Quoted 4 lines · Jan Dittmer

Why not get rid of the daemon like bsd did [0]? Otherwise in 5 years you'll have 42 daemons running which communicate with the firmware of various devices, each having a different inter- face.

Because it would involve a moderate rewriting of the driver, and we'd have to carry a delta against Intel's code forever.

without knowing this for sure, dont you think that if a largely changed version of the driver appeared in the tree, intel may start developing on that? cause then they wouldnt be the ones that "broke" compliance with FCC(hah) by not doing binaryonly.

Quoted 1 line
#6 lore raw
Quoted 7 lines · Kasper Sandberg
Quoted 2 lines

Because it would involve a moderate rewriting of the driver, and we'd have to carry a delta against Intel's code forever.

without knowing this for sure, dont you think that if a largely changed version of the driver appeared in the tree, intel may start developing on that? cause then they wouldnt be the ones that "broke" compliance with FCC(hah) by not doing binaryonly.

It's just as likely that their lawyers would tell them that they would have to pretend that the modifications don't exist at all, and not release any changes for any driver (like OpenBSD's) that bypassed the regulatory daemon. The bigger worry would be if they decided that they couldn't risk supporting their current out-of-tree driver, and couldn't release Linux drivers for their softmac wireless devices in the future.

Personally, I don't see why the requirement of an external daemon is really considered that evil. We allow drivers that depend on firmware loaders, don't we? I could imagine a device that required a digitally signed message (using RSA) with a challenge/response protocol embedded inside that was necessary to configure said device, which is calculated in userspace and then passed down into the kernel to be installed into the device so that it could function. Do we really want to consider that to be objectionable?

- Ted

#7 lore raw
Quoted 9 lines · Theodore Tso
Quoted 7 lines · Kasper Sandberg
Quoted 2 lines

Because it would involve a moderate rewriting of the driver, and we'd have to carry a delta against Intel's code forever.

without knowing this for sure, dont you think that if a largely changed version of the driver appeared in the tree, intel may start developing on that? cause then they wouldnt be the ones that "broke" compliance with FCC(hah) by not doing binaryonly.

Quoted 7 lines

It's just as likely that their lawyers would tell them that they would have to pretend that the modifications don't exist at all, and not release any changes for any driver (like OpenBSD's) that bypassed the regulatory daemon. The bigger worry would be if they decided that they couldn't risk supporting their current out-of-tree driver, and couldn't release Linux drivers for their softmac wireless devices in the future.

i think, that if no driver exists, there would be further incentive for people to reverse engineer, as i also believe that if nvidia didnt release their closed driver, there would be a project that would have created a working driver for it(also supporting 3d)

Quoted 4 lines

Personally, I don't see why the requirement of an external daemon is really considered that evil. We allow drivers that depend on firmware loaders, don't we? I could imagine a device that required a digitally

thats entirely different, if some firmware image is loaded into a card, thats that, but running a userspace daemon is just entirely different, what would happen if intel for some reason stopped supporting earlier cards(as hardware manufactureres do after some time), and linux kernel/userspace gets some change, preventing the binary daemon from running? then what? we have lost. but i do not believe any change can really be made, that would mean the existing binary firmware images could not be loaded into the hardware.

Quoted 8 lines

signed message (using RSA) with a challenge/response protocol embedded inside that was necessary to configure said device, which is calculated in userspace and then passed down into the kernel to be installed into the device so that it could function. Do we really want to consider that to be objectionable?

- Ted

#8 lore raw
Quoted 6 lines · Kasper Sandberg

thats entirely different, if some firmware image is loaded into a card, thats that, but running a userspace daemon is just entirely different, what would happen if intel for some reason stopped supporting earlier cards(as hardware manufactureres do after some time), and linux kernel/userspace gets some change, preventing the binary daemon from running? then what? we have lost.

Um, last time I checked we could still run some *minix* binaries from before Linux was born, and we still can run statically linked a.out programs created over a decade ago. I don't think this is a serious objection, given that historically the Linux kernel/userspace syscall interface has been quite stable.

Of course, I'd recomend against said driver using sysfs, but Greg K-H tells us that all breakagaes are the fault of buggy device drivers (just as supposedly all swsuspend problems are also about buggy device drivers), so I guess we're OK. :-)

- Ted

#9 lore raw
Quoted 8 lines · Theodore Tso
Quoted 6 lines · Kasper Sandberg

thats entirely different, if some firmware image is loaded into a card, thats that, but running a userspace daemon is just entirely different, what would happen if intel for some reason stopped supporting earlier cards(as hardware manufactureres do after some time), and linux kernel/userspace gets some change, preventing the binary daemon from running? then what? we have lost.

Quoted 5 lines

Um, last time I checked we could still run some *minix* binaries from before Linux was born, and we still can run statically linked a.out programs created over a decade ago. I don't think this is a serious objection, given that historically the Linux kernel/userspace syscall interface has been quite stable.

thats besides the point, i was arguing the difference between loading a firmware image and running a binary daemon.

Quoted 8 lines

Of course, I'd recomend against said driver using sysfs, but Greg K-H tells us that all breakagaes are the fault of buggy device drivers (just as supposedly all swsuspend problems are also about buggy device drivers), so I guess we're OK. :-)

- Ted

#10 lore raw
In reply to Theodore Tso
Quoted 2 lines · Theodore Tso

Personally, I don't see why the requirement of an external daemon is really considered that evil. We allow drivers that depend on firmware

Well the problem is more the fear that when one vendor gets this in the tree all others will follow. And there'll be several incompatible userspace daemons for every possible wireless card which you need to get the system to work (think boot cd, install over wlan).

So if this is done, there should be a clearly abstracted interface for such a daemon. I don't see what the daemon is doing more than echo 1 4 7 8 > /sys/.../allowed_channels and a control circuit for tx/rx power.

Quoted 6 lines

loaders, don't we? I could imagine a device that required a digitally signed message (using RSA) with a challenge/response protocol embedded inside that was necessary to configure said device, which is calculated in userspace and then passed down into the kernel to be installed into the device so that it could function. Do we really want to consider that to be objectionable?

If it's done via a standard interface as the firmware loading is, no objection.

Jan

#11 lore raw
Quoted 4 lines · Jan Dittmer

So if this is done, there should be a clearly abstracted interface for such a daemon. I don't see what the daemon is doing more than echo 1 4 7 8 > /sys/.../allowed_channels and a control circuit for tx/rx power.

The daemon sets the list of acceptable channels, the transmission powers for each of them and modifies this based on things like the temperature.

Oh, yes, and it writes 802.11 scan frames that get passed straight to the firmware and broadcast. Insane crack. For reasons I don't as yet understand, various received frames also get passed back up to the daemon. I probably really don't want to know why.

Signature · 1 line
Matthew Garrett | [email protected]
#12 lore raw
In reply to Theodore Tso

Ar Sul, 2006-07-30 am 10:53 -0400, ysgrifennodd Theodore Tso:

Quoted 2 lines

Personally, I don't see why the requirement of an external daemon is really considered that evil. We allow drivers that depend on firmware

An open source daemon doing the supervising seems fine. We do that for IPv4 even (dhcpd, dhcpcd, routed, gated, zebra, ....) ;)

Quoted 6 lines

loaders, don't we? I could imagine a device that required a digitally signed message (using RSA) with a challenge/response protocol embedded inside that was necessary to configure said device, which is calculated in userspace and then passed down into the kernel to be installed into the device so that it could function. Do we really want to consider that to be objectionable?

If it controls the use of the device inappropriately then yes we should. Suppose it passes down an RSA signed message that is computed by verifying you are running a Red Hat Enterprise Linux 4 official shipped kernel on an approved IBM platform and things like that - that would annoy me.

Using GPG keys to verify code matches approved code is fine. The ISDN layer has done this for many years and if its not the signed code version it didn't have the old silly approvals stuff (Fortunately nowdays even politicians have realised that if you need approval for something to stop it doing bad stuff to the phone network then bad people can do bad stuff anyway and this is bad for national security)

I think from a vendor perspective being able to ship a source set for such a daemon which Intel has signed as "this source meets the rules" is great. As an end user I want the flexibility to do other stuff except where prohibited by law (not by Intel, not by paranoid lawyers and not by FCC lack of clarity)

I also want to violate the power and other policy rules Intel wishes to enforce on their hardware for legal and legitimate reasons. As a radio amateur I am permitted in law to transmit certain classes of signals in that frequency space at higher power than in the hands of a random UK user of license exempt technology.

Alan

#13 lore raw
In reply to Theodore Tso

Hi!

Quoted 16 lines
Quoted 7 lines
Quoted 2 lines

Because it would involve a moderate rewriting of the driver, and we'd have to carry a delta against Intel's code forever.

without knowing this for sure, dont you think that if a largely changed version of the driver appeared in the tree, intel may start developing on that? cause then they wouldnt be the ones that "broke" compliance with FCC(hah) by not doing binaryonly.

It's just as likely that their lawyers would tell them that they would have to pretend that the modifications don't exist at all, and not release any changes for any driver (like OpenBSD's) that bypassed the regulatory daemon. The bigger worry would be if they decided that they couldn't risk supporting their current out-of-tree driver, and couldn't release Linux drivers for their softmac wireless devices in the future.

Quoted 4 lines

Personally, I don't see why the requirement of an external daemon is really considered that evil. We allow drivers that depend on firmware loaders, don't we? I could imagine a device that required a digitally

Well, drivers that depend on external, non-redistributable firmware (aka ipw2200/ipw2100) already are evil, but at least I do not run untrusted code on my CPU. (And yes, firmware in ipw2200 crashes a _lot_).

Plus, that regulatory daemon probably contains security hole (or two) inside, and I can't properly audit it. (Yes, firmware in ipw2200 probably contains security holes, too, but at least those are more "interesting" to exploit).

And... Intel will not even tell you WTF that daemon does. They claim it is for FCC, but it seems to be doing more than that. So maybe I'm not _that_ paranoid.

(Of course, binary-only kernel module would be even uglier).

Ouch and binary-only daemon will conviently prevent me from using that wireless card in ppc machine.
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#14 lore raw

On Monday 31 July 2006 00:12, Pavel Machek wrote:
[snip]

Quoted 3 lines

And... Intel will not even tell you WTF that daemon does. They claim it is for FCC, but it seems to be doing more than that. So maybe I'm not _that_ paranoid.

Agreed, from what Matthew's said it seems like the daemon is being used to hide intellectual property, not something we should really be encouraging.

I think the title "regulatory daemon" has multiple meanings, it REGULATES your frequencies to FCC specs, it REGULATES your wireless card's power and temperature levels, and it REGULATES your right to use the hardware ;-)

Ultimately the question remains, will we open this can of worms by accepting drivers that depend on proprietary software (i.e. they will not function at all without it). I'm fairly sure the answer should be "No".

Signature · 5 lines
Cheers,
Alistair.

Final year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.
#15 lore raw
Quoted 9 lines · Alistair John Strachan

On Monday 31 July 2006 00:12, Pavel Machek wrote:
[snip]

Quoted 3 lines

And... Intel will not even tell you WTF that daemon does. They claim it is for FCC, but it seems to be doing more than that. So maybe I'm not _that_ paranoid.

Agreed, from what Matthew's said it seems like the daemon is being used to hide intellectual property, not something we should really be encouraging.

Quoted 7 lines

I think the title "regulatory daemon" has multiple meanings, it REGULATES your frequencies to FCC specs, it REGULATES your wireless card's power and temperature levels, and it REGULATES your right to use the hardware ;-)

Ultimately the question remains, will we open this can of worms by accepting drivers that depend on proprietary software (i.e. they will not function at all without it). I'm fairly sure the answer should be "No".

I entirely agree that this should not be merged, those will accept these kindof things, can use intels out of tree driver.

i sincerely hope for a forked/rewritten driver which does not depend on closed userspace daemons.

Quoted 1 line
#16 lore raw
Quoted 5 lines · Kasper Sandberg

I entirely agree that this should not be merged, those will accept these kindof things, can use intels out of tree driver.

i sincerely hope for a forked/rewritten driver which does not depend on closed userspace daemons.

I personally think that this also violates the GPL license... The GPL part cannot stand by it-self and require none GPLed component in order to make it work.

The fact that the regularity daemon work using external sysfs interface without linkage requirements does not escape derived work in this case.

Best Regards,
Alon Bar-Lev

#17 lore raw
Quoted 7 lines · Alon Bar-Lev
Quoted 5 lines · Kasper Sandberg

I entirely agree that this should not be merged, those will accept these kindof things, can use intels out of tree driver.

i sincerely hope for a forked/rewritten driver which does not depend on closed userspace daemons.

Quoted 7 lines

I personally think that this also violates the GPL license... The GPL part cannot stand by it-self and require none GPLed component in order to make it work.

The fact that the regularity daemon work using external sysfs interface without linkage requirements does not escape derived work in this case.

i tend to agree, however if a generic interface was created for "regulatory daemons" this may be different, but if what i hear is true, the intel daemon does more than that, and then, well...

Quoted 9 lines

Best Regards,
Alon Bar-Lev
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#18 lore raw
In reply to Kasper Sandberg
Quoted 6 lines · Kasper Sandberg
Quoted 1 line

Because it would involve a moderate rewriting of the driver, and we'd have to carry a delta against Intel's code forever.

without knowing this for sure, dont you think that if a largely changed version of the driver appeared in the tree, intel may start developing on that? cause then they wouldnt be the ones that "broke" compliance with FCC(hah) by not doing binaryonly.

Where can I find this FCC law that seems to be crippling open source wlan development?

I am not from the USA, so I don't have to comply with the FCC. Could we make a non-crippled totally open source driver for use by people outside the USA?

For example, here in the UK one can own radios that can transmit on any frequency one likes, but if you actually press the TX button without a the appropriate License, you break the law.

James

#19 lore raw
Quoted 9 lines · James Courtier-Dutton
Quoted 7 lines · Kasper Sandberg
Quoted 2 lines

Because it would involve a moderate rewriting of the driver, and we'd have to carry a delta against Intel's code forever.

without knowing this for sure, dont you think that if a largely changed version of the driver appeared in the tree, intel may start developing on that? cause then they wouldnt be the ones that "broke" compliance with FCC(hah) by not doing binaryonly.

Quoted 6 lines

Where can I find this FCC law that seems to be crippling open source wlan development?

I am not from the USA, so I don't have to comply with the FCC. Could we make a non-crippled totally open source driver for use by people outside the USA?

as with encryption, arent some of the encryption stuff widely used in the opensource community also illegal in the united states?

Quoted 4 lines

For example, here in the UK one can own radios that can transmit on any frequency one likes, but if you actually press the TX button without a the appropriate License, you break the law.

im quite certain this is also the case in for example Denmark.

Quoted 3 lines

James

#20 lore raw

Hi,

it would be totally ok if the kernel had a country= command line switch and the driver limitting functionality due that.

People that want to violate the local regulations would require to lie to the kernel as they could install other country windows and drivers as well.

Sure, people could modify the source as they could specify a wrong country argument. But you can heedit the windows binary as well.

Yours,

Signature · 3 lines
  René Rebe - ExactCODE - Berlin (Europe / Germany)
  http://exactcode.de | http://t2-project.org | http://rene.rebe.name
  +49 (0)30 / 255 897 45
#21 lore raw
Quoted 4 lines · Rene Rebe

Hi,

it would be totally ok if the kernel had a country= command line switch and the driver limitting functionality due that.

or simply state this in the help in Kconfig?

Quoted 4 lines

People that want to violate the local regulations would require to lie to the kernel as they could install other country windows and drivers as well.

besides, im not even sure that specifying in Kconfig is necessary, wouldnt it only be illegal in countries, if people actually modified the source?

Quoted 6 lines

Sure, people could modify the source as they could specify a wrong country argument. But you can heedit the windows binary as well.

Yours,

#22 lore raw

HI,

Quoted 10 lines · Kasper Sandberg
Quoted 2 lines

it would be totally ok if the kernel had a country= command line switch and the driver limitting functionality due that.

or simply state this in the help in Kconfig?

Quoted 4 lines

People that want to violate the local regulations would require to lie to the kernel as they could install other country windows and drivers as well.

besides, im not even sure that specifying in Kconfig is necessary, wouldnt it only be illegal in countries, if people actually modified the source?

I proposed a kernel command so distributors have a way to run-time change this.

However now that I think about it a bit more, a simple sysfs attribute would be way more useful so the gui tool of the the distribution can switch the country immediatly and do not require a windows-al-like reboot.

Yours,

Signature · 3 lines
  René Rebe - ExactCODE - Berlin (Europe / Germany)
  http://exactcode.de | http://t2-project.org | http://rene.rebe.name
  +49 (0)30 / 255 897 45
#23 lore raw
Quoted 15 lines · Rene Rebe

HI,

Quoted 10 lines · Kasper Sandberg
Quoted 2 lines

it would be totally ok if the kernel had a country= command line switch and the driver limitting functionality due that.

or simply state this in the help in Kconfig?

Quoted 4 lines

People that want to violate the local regulations would require to lie to the kernel as they could install other country windows and drivers as well.

besides, im not even sure that specifying in Kconfig is necessary, wouldnt it only be illegal in countries, if people actually modified the source?

Quoted 7 lines

I proposed a kernel command so distributors have a way to run-time change this.

However now that I think about it a bit more, a simple sysfs attribute would be way more useful so the gui tool of the the distribution can switch the country immediatly and do not require a windows-al-like reboot.

or perhaps people should just not install/use stuff illegal in their country.

Quoted 2 lines

Yours,

#24 lore raw

Ar Sul, 2006-07-30 am 23:02 +0200, ysgrifennodd Kasper Sandberg:

Quoted 2 lines

or perhaps people should just not install/use stuff illegal in their country.

Most users really don't understand the issues around wireless and country specific rules. Some wireless implementations also don't deal with moving between countries live (as happens all the time today in Europe).

That means as a distribution vendor its really important to ship people something that by default does the right thing and the legal thing here. If people want to recompile kernels or hack firmware thats their business, but out of the box it should behave.

#25 lore raw
Quoted 9 lines · Alan Cox

Ar Sul, 2006-07-30 am 23:02 +0200, ysgrifennodd Kasper Sandberg:

Quoted 2 lines

or perhaps people should just not install/use stuff illegal in their country.

Most users really don't understand the issues around wireless and country specific rules. Some wireless implementations also don't deal with moving between countries live (as happens all the time today in Europe).

Quoted 4 lines

That means as a distribution vendor its really important to ship people something that by default does the right thing and the legal thing here. If people want to recompile kernels or hack firmware thats their business, but out of the box it should behave.

as it will never do properly requiring a binary daemon, distributions are having a hard enough time to try and redistribute those firmwares where its legal, some even wont, but a userspace daemon is out of the question for most.

Quoted 2 lines
#26 lore raw

Hi,

Quoted 10 lines · Kasper Sandberg
Quoted 9 lines · Alan Cox

Ar Sul, 2006-07-30 am 23:02 +0200, ysgrifennodd Kasper Sandberg:

Quoted 2 lines

or perhaps people should just not install/use stuff illegal in their country.

Most users really don't understand the issues around wireless and country specific rules. Some wireless implementations also don't deal with moving between countries live (as happens all the time today in Europe).

Quoted 8 lines
Quoted 4 lines

That means as a distribution vendor its really important to ship people something that by default does the right thing and the legal thing here. If people want to recompile kernels or hack firmware thats their business, but out of the box it should behave.

as it will never do properly requiring a binary daemon, distributions are having a hard enough time to try and redistribute those firmwares where its legal, some even wont, but a userspace daemon is out of the question for most.

Sure not. That is why I wanted to direct the discussion in what country enforcing scheme we should go.

Signature · 3 lines
  René Rebe - ExactCODE - Berlin (Europe / Germany)
  http://exactcode.de | http://t2-project.org | http://rene.rebe.name
  +49 (0)30 / 255 897 45
#27 lore raw
In reply to Matthew Garrett
Quoted 10 lines · Matthew Garrett
Quoted 3 lines · Pavel Machek

I'm unfortunate enough to have x60 with ipw card. Card works okay, but driver is 16K LoC and needs binary daemon (ugh). Plus, it lives as an external module...

I have a mostly working replacement for the binary daemon, but it causes the firmware to crash at the point where it triggers an association. If anyone's keen on trying to figure out what's up, I'll clean up the code and stick it somewhere.

Is your daemon somehow releated to wifi frequencies goegraphy database
proposed some time ago on netdev?

Signature · 2 lines
Tomasz Torcz                 "God, root, what's the difference?"
[email protected]         "God is more forgiving."
#28 lore raw
Quoted 2 lines · Tomasz Torcz

Is your daemon somehow releated to wifi frequencies goegraphy database
proposed some time ago on netdev?

No. So far, I haven't implemented that at all.

Signature · 1 line
Matthew Garrett | [email protected]