MODULE_LICENSE and EXPORT_SYMBOL_GPL
'Twas brillig when Arjan van de Ven scrobe:
Quoted 4 lines
I think you're missing one thing: binary only modules are only allowed because of an exception license grant Linus made for functions that are marked
EXPORT_SYMBOL().EXPORT_SYMBOL_GPL()just says "not part of this exception grant"....
With all respect to Linus, I don't believe that module insertion is an exclusive right granted to authors that is within Linus' legal power as holder of the Copyright to the kernel to grant or to restrict. Does Microsoft have a legal right to disallow any third-party drivers from registering themselves with the OS? Does Linus?
Quoted 11 lines · Roy Murphy
'Twas brillig when Arjan van de Ven scrobe:
Quoted 4 lines
I think you're missing one thing: binary only modules are only allowed because of an exception license grant Linus made for functions that are marked
EXPORT_SYMBOL().EXPORT_SYMBOL_GPL()just says "not part of this exception grant"....With all respect to Linus, I don't believe that module insertion is an exclusive right granted to authors that is within Linus' legal power as holder of the Copyright to the kernel to grant or to restrict. Does Microsoft have a legal right to disallow any third-party drivers from registering themselves with the OS? Does Linus?
I'm sorry. "module inserting" is LINKING. A kernel module does, in my oppinion, NOT fall under the gpl stated "mere aggregation" boundary of the GPL, it is compiled with kernel headers, contains kernel _code_ from these headers etc etc, and is for all intents and purposes part of the GPL program "kernel" once loaded. It uses normal function calls etc etc, symbols are resolved using normal linking mechanisms etc etc.
On Thu, 2001-10-18 11:43:15 -0500, Roy Murphy <[email protected]> wrote in message <[email protected]>:
Quoted 9 lines
'Twas brillig when Arjan van de Ven scrobe:
Quoted 4 lines
I think you're missing one thing: binary only modules are only allowed because of an exception license grant Linus made for functions that are marked
EXPORT_SYMBOL().EXPORT_SYMBOL_GPL()just says "not part of this exception grant"....of the Copyright to the kernel to grant or to restrict. Does Microsoft have a legal right to disallow any third-party drivers from registering themselves with the OS? Does Linus?
They do, but they won't use it. They want to *sell* windows and they're (more or less) willing to decode their blue screens produced by 3rd vendor's drivers. However, GPL people may (or may not) be willing to spend time in searching bugs in other company's drivers. However, *I* am not willig to do other people's job, especially if *they* earn money therefor...
MfG, JBG
Signature · 1 line
Jan-Benedict Glaw . [email protected] . +49-172-7608481
Quoted 8 lines · Arjan van de Ven
Quoted 5 lines · Roy Murphy
With all respect to Linus, I don't believe that module insertion is an exclusive right granted to authors that is within Linus' legal power as holder of the Copyright to the kernel to grant or to restrict. Does Microsoft have a legal right to disallow any third-party drivers from registering themselves with the OS? Does Linus?
Quoted 6 lines
I'm sorry. "module inserting" is LINKING. A kernel module does, in my oppinion, NOT fall under the gpl stated "mere aggregation" boundary of the GPL, it is compiled with kernel headers, contains kernel _code_ from these headers etc etc, and is for all intents and purposes part of the GPL program "kernel" once loaded. It uses normal function calls etc etc, symbols are resolved using normal linking mechanisms etc etc.
"module inserting is linking" is true for almost
any other operating system which supports loadable modules (including Windows), and really has no bearing on Mr. Murphy's argument. The stronger argument is with the kernel header files being GPL, to the degree that they contain macros containing actual code, which expand into binary code actually in the compiled module.
However, Mr. Murphy's point is correct. The law supports the idea that certain INTERFACES may create a legitimate boundary between different code bases to distinguish them (as not being derivative works of each other). This is true no matter what the FSF thinks of static linking vs. dyamic linking, or what Linus has said about permissibility of non-GPL loadable modules.
I don't think even the interface creator can decide whether a particular interface meets the criteria for isolating or not the calling program from being a derivative work. It's really up to a judge.
However, the EXPORT_SYMBOL_GPL facility may play a role in indicating to a judge or jury the intent of the original developer with regard to the interface. This is unlikely to have as big an impact on the legality of using the symbol from a non-GPL module as other factors that come into play here.
Please don't misinterpret me. I'm glad Linus made his exception (even though it may not exactly jibe with copyright law and the GPL license), and I see value in having developer's express their licensing intent for their published interfaces. It's just dangerous to assume this holds much legal weight.
____________________________________________________________
Tim Bird Lineo, Inc. Senior VP, Research 390 South 400 West [email protected] Lindon, UT 84042
so what will the export_symbol_gpl stuff do with the BSD license? it may or may not have source avilable so is it allowed to use the exported symbols or not?
for the tainting module process there is the same problem.
knowing the license the code was released under does not tell you if the source is available or not.
David Lang
Quoted 34 lines · Jan-Benedict Glaw
Date: Thu, 18 Oct 2001 18:07:06 +0200
From: Jan-Benedict Glaw <[email protected]>
To: [email protected]
Subject: Re:MODULE_LICENSEandEXPORT_SYMBOL_GPLOn Thu, 2001-10-18 11:43:15 -0500, Roy Murphy <[email protected]> wrote in message <[email protected]>:
Quoted 9 lines
'Twas brillig when Arjan van de Ven scrobe:
Quoted 4 lines
I think you're missing one thing: binary only modules are only allowed because of an exception license grant Linus made for functions that are marked
EXPORT_SYMBOL().EXPORT_SYMBOL_GPL()just says "not part of this exception grant"....of the Copyright to the kernel to grant or to restrict. Does Microsoft have a legal right to disallow any third-party drivers from registering themselves with the OS? Does Linus?
They do, but they won't use it. They want to *sell* windows and they're (more or less) willing to decode their blue screens produced by 3rd vendor's drivers. However, GPL people may (or may not) be willing to spend time in searching bugs in other company's drivers. However, *I* am not willig to do other people's job, especially if *they* earn money therefor...
MfG, JBG
--
Jan-Benedict Glaw . [email protected] . +49-172-7608481
-
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/
Quoted 8 lines · David Lang
so what will the
export_symbol_gplstuff do with the BSD license? it may or may not have source avilable so is it allowed to use the exported symbols or not?for the tainting module process there is the same problem.
knowing the license the code was released under does not tell you if the source is available or not.
Linus said that all existing entry points would remain untagged. Thus existing modules would not be affected.
john
so are you saying that BSD licened modules are not going to be allowed to use any future entry points (assuming they are all put under the the export_symbol_gpl limits)?
saying that existing stuff will not change doesn't answer the problem about licences that may have source avilable (tainting) but may not. if you assume either way you will cause a bunch of problems.
David Lang
Quoted 22 lines · John Alvord
Date: Thu, 18 Oct 2001 17:46:18 -0700
From: John Alvord <[email protected]>
To: David Lang <[email protected]>
Cc: [email protected]
Subject: Re:MODULE_LICENSEandEXPORT_SYMBOL_GPLQuoted 8 lines · David Lang
so what will the
export_symbol_gplstuff do with the BSD license? it may or may not have source avilable so is it allowed to use the exported symbols or not?for the tainting module process there is the same problem.
knowing the license the code was released under does not tell you if the source is available or not.
Linus said that all existing entry points would remain untagged. Thus existing modules would not be affected.
john
Quoted 10 lines · David Lang
so are you saying that BSD licened modules are not going to be allowed to use any future entry points (assuming they are all put under the the
export_symbol_gpllimits)?saying that existing stuff will not change doesn't answer the problem about licences that may have source avilable (tainting) but may not. if you assume either way you will cause a bunch of problems.
David Lang
That's nonsense. OK, with a strict interpretation on the use of the MODULE_LICENSE and depending on the implementation of license strings, old BSD licensed code could taint a kernel, but according to http://www.gnu.org/licenses/license-list.html modified BSD licensed code or code falling under the similar X11 license is GPL compatible. Therefore if you distribute or contribute source for a module that is BSD licensed, and MODULE_LICENSE is implemented in a remotely sane way you will be able to install a BSD licensed module without tainting the kernel.
Quoted 6 lines · John Alvord
Quoted 5 lines
Linus said that all existing entry points would remain untagged. Thus existing modules would not be affected.
john
The GPL includes as derivative works any code, when compiled or executed, that is linked with GPL covered code, calls functions or shares data structures. Simply fork()ing or exec()ing is not a derivative work. FSF states that linking and simply calling main() and returning is a border case(1). From that we can extrapolate a number of things. Simply doing a syscall is not a derivative work. As I have read in the past, Linus, for practical reasons, has stated that modules using published interfaces need not qualify as derivative works. Also, "modules" in the generic sense, say plugins for example, when written to link with GPL code need not be GPL themselves. So long as they are GPL compatible, they can be linked without violating the license. Hence, MODULE_LICENSE strings may include GPL compatible source distributing licenses, eg. newBSD or X11.
MODULE_LICENSE & EXPORT_SYMBOL_GPL are good things. MODULE_LICENSE provides an easier way to screen binary only modules from submitted bugs. It doesn't prevent someone from running a tainted kernel and it also doesn't prevent someone on linux-kernel from investigating the bug report if they want to. MODULE_LICENSE also has the potential to educate users. EXPORT_SYMBOL_GPL throws a speed bump in front of potentially dangerous subsystem modules. The Linux Security Module interface is the most recent example of this(2). If modules are allowed to insert hooks willy nilly into the kernel, eventually large proprietary functionality could result, unfairly taking advantage of Linus' practical foresight. EXPORT_SYMBOL_GPL allows us a simple way to define "public" interfaces that binary only module writers can use and which symbols are off limits. There is no restriction on fair use, as you can still remove the
disabling GPL compatible exports for internal, non-released code.
Also, we are afforded more leeway in the future. If general consesus is that we need to let a proprietary vendor use a GPL symbol, we can always remove the restriction. Alternatively, if a vendor realizes he could write a better module if only he could use GPL-compatible-only symbols he may think twice and release the source. In the end, thats a solution I can live with.
Regards,
Reid
1. From GNU GPL FAQ: http://www.gnu.org/licenses/gpl-faq.html#GPLModuleLicense
Quoted 6 lines · Arjan van de Ven
I'm sorry. "module inserting" is LINKING. A kernel module does, in my oppinion, NOT fall under the gpl stated "mere aggregation" boundary of the GPL, it is compiled with kernel headers, contains kernel _code_ from these headers etc etc, and is for all intents and purposes part of the GPL program "kernel" once loaded. It uses normal function calls etc etc, symbols are resolved using normal linking mechanisms etc etc.
You're quite right. Module insertion is linking. And distributing a kernel with binary-only modules already inserted would be a GPL violation. What modules do is let people do the link at the last stage -- the end user. The GPL does not restrict what end-users do with your code if it doesn't involve redistribution.
In short: Copyright holders have the right to attempt to restrict their interfaces. But end-users also have the right to ignore those attempts. That is, unless the DMCA comes into play.
Signature · 4 lines
Taral <[email protected]> This message is digitally signed. Please PGP encrypt mail to me. "Any technology, no matter how primitive, is magic to those who don't understand it." -- Florence Ambrose
I use this example becouse the question came up a week or so ago and I don't remember seeing the issue resolved.
the problem posted is that a BSD no adv licence is GPL compatable and source may be available and therefor BSD no adv modules should not taint the kernel, on the other hand it's possible to have a BSD no adv licensed module that the source is not available for and therefor it should taint the kernel.
just knowing the license doesn't tell you if the source is available (which is the stated goal of the taint stuff)
if instead of MODULE_LICENSE there was a MODULE_SOURCE tag this would not be an issue, and the EXPORT_SYMBOL_GPL would obviously be a seperate issue. as it is they both use the MODULE_LICENSE tag which confuses the intent of both of them.
if BSD no adv is considered a non-tainting license then what's to stop all the binary-module vendors from useing it in their modules? it doesn't requrire that they give the source to anyone so it's no risk to their IP but it avaids the 'bad press' of the kernel announcing that useing their stuff taints the kernel.
on the other hand if BSD no adv is considered a taining license then you are going against the statement that it's compatable with the GPL and you are telling module programmers who release the source that BSD isn't good enough they can only work with the linux kernel if they change to the GPL (or one of the other approved licenses)
again my point is that knowing the license is not enough by itself to know if the kernel should be considered tainted, so let's not try to do it this way.
or if the intent really is to force everything to be GPL then just say so rather then claiming that that's not your intent.
David Lang
Quoted 81 lines · Reid Hekman
Date: Fri, 19 Oct 2001 07:44:43 -0500
From: Reid Hekman <[email protected]>
To: David Lang <[email protected]>
Cc: [email protected]
Subject: Re:MODULE_LICENSEandEXPORT_SYMBOL_GPLQuoted 10 lines · David Lang
so are you saying that BSD licened modules are not going to be allowed to use any future entry points (assuming they are all put under the the
export_symbol_gpllimits)?saying that existing stuff will not change doesn't answer the problem about licences that may have source avilable (tainting) but may not. if you assume either way you will cause a bunch of problems.
David Lang
That's nonsense. OK, with a strict interpretation on the use of the
MODULE_LICENSEand depending on the implementation of license strings, old BSD licensed code could taint a kernel, but according to
http://www.gnu.org/licenses/license-list.html modified BSD licensed code or code falling under the similar X11 license is GPL compatible. Therefore if you distribute or contribute source for a module that is BSD licensed, andMODULE_LICENSEis implemented in a remotely sane way you will be able to install a BSD licensed module without tainting the kernel.Quoted 6 lines · John Alvord
Quoted 5 lines
Linus said that all existing entry points would remain untagged. Thus existing modules would not be affected.
john
The GPL includes as derivative works any code, when compiled or executed, that is linked with GPL covered code, calls functions or shares data structures. Simply fork()ing or exec()ing is not a derivative work. FSF states that linking and simply calling
main()and returning is a border case(1). From that we can extrapolate a number of things. Simply doing a syscall is not a derivative work. As I have read in the past, Linus, for practical reasons, has stated that modules using published interfaces need not qualify as derivative works. Also, "modules" in the generic sense, say plugins for example, when written to link with GPL code need not be GPL themselves. So long as they are GPL compatible, they can be linked without violating the license. Hence,MODULE_LICENSEstrings may include GPL compatible source distributing licenses, eg. newBSD or X11.
MODULE_LICENSE&EXPORT_SYMBOL_GPLare good things.MODULE_LICENSEprovides an easier way to screen binary only modules from submitted bugs. It doesn't prevent someone from running a tainted kernel and it also doesn't prevent someone on linux-kernel from investigating the bug report if they want to.MODULE_LICENSEalso has the potential to educate users.EXPORT_SYMBOL_GPLthrows a speed bump in front of potentially dangerous subsystem modules. The Linux Security Module interface is the most recent example of this(2). If modules are allowed to insert hooks willy nilly into the kernel, eventually large proprietary functionality could result, unfairly taking advantage of Linus' practical foresight.EXPORT_SYMBOL_GPLallows us a simple way to define "public" interfaces that binary only module writers can use and which symbols are off limits. There is no restriction on fair use, as you can still remove the
disabling GPL compatible exports for internal, non-released code.
Also, we are afforded more leeway in the future. If general consesus is that we need to let a proprietary vendor use a GPL symbol, we can always remove the restriction. Alternatively, if a vendor realizes he could write a better module if only he could use GPL-compatible-only symbols he may think twice and release the source. In the end, thats a solution I can live with.Regards,
Reid1. From GNU GPL FAQ:
http://www.gnu.org/licenses/gpl-faq.html#GPLModuleLicense
Quoted 2 lines · David Lang
I use this example becouse the question came up a week or so ago and I don't remember seeing the issue resolved.
I wish it was easy, but legality/copyright/property rights are not always easy topics
Quoted 8 lines
the problem posted is that a BSD no adv licence is GPL compatable and source may be available and therefor BSD no adv modules should not taint the kernel, on the other hand it's possible to have a BSD no adv licensed module that the source is not available for and therefor it should taint the kernel.
just knowing the license doesn't tell you if the source is available (which is the stated goal of the taint stuff)
Good point. Anybody else want to chime in here? For linux-kernel's purposes at least, a BSD licensed piece of code with no source is about as useful as ice-cream to an eskimo. Of course creating BSD licensed binaries from closed source is kind of silly. Say we see an external module (one not distributed with the kernel source) that identifies itself as BSD licensed loaded into somebodies' kernel then ask the provider for source and he refuses. Rightfully that kernel should have been marked tainted. Is their anything in the licensing that would allow us to say that specifying a BSD license in the MODULE_LICENSE tag refers to source distribution? Or is that implied?
Quoted 4 lines
if instead of
MODULE_LICENSEthere was aMODULE_SOURCEtag this would not be an issue, and theEXPORT_SYMBOL_GPLwould obviously be a seperate issue. as it is they both use theMODULE_LICENSEtag which confuses the intent of both of them.
What's in a name? If it's called MODULE_LICENSE or MODULE_SOURCE if it's defined to mean availability and licensing of module source, that's the important distinction. Further, EXPORT_SYMBOL_GPL is a separate issue regardless, it allows us a definite way to specify "private" interfaces.
Quoted 5 lines
if BSD no adv is considered a non-tainting license then what's to stop all the binary-module vendors from useing it in their modules? it doesn't requrire that they give the source to anyone so it's no risk to their IP but it avaids the 'bad press' of the kernel announcing that useing their stuff taints the kernel.
If it becomes public knowledge that a binary-module vendor distributes modules marked as BSD licensed but the source is unavailable that would be bad PR for them. First they would probably be ignored by linux-kernel. They should also be branded as liars for disingenuosly marking their products.
Quoted 5 lines
on the other hand if BSD no adv is considered a taining license then you are going against the statement that it's compatable with the GPL and you are telling module programmers who release the source that BSD isn't good enough they can only work with the linux kernel if they change to the GPL (or one of the other approved licenses)
I don't think anybody wants to disallow linking of GPL compatible code with Linux. Given that the FSF thinks that BSD-no-adv is compatible, I'd tend to agree with them. If MODULE_LICENSE is to be used as your "MODULE_SOURCE" idea is (which I think it is), it's not an issue. As far as programmers releasing BSD source, they seem to be OK with it, check the source yourself: http://lxr.linux.no/search?string=MODULE_LICENSE.
Quoted 3 lines
again my point is that knowing the license is not enough by itself to know if the kernel should be considered tainted, so let's not try to do it this way.
Making access control mandatory with source code availability is hard. Nobody said this was a be-all, end-all solution, it simply helps screen those that would take advantage of people's limited time and effort. It depends partly on respect and good-will, so if a vendor wants to violate that trust and flaunt arrogance to the folks of the linux community, so be it.
Quoted 4 lines
or if the intent really is to force everything to be GPL then just say so rather then claiming that that's not your intent.
David Lang
The intent is to more clearly define the gracious exception that has been made for binary-only module vendors. If we decide to disallow the use of BSD licensed code (when everybody else in the world can use it) that's our loss. In the end we want to respect other people's copyrights and have our's respected in turn.
Regards,
Reid
Quoted 5 lines · David Lang
the problem posted is that a BSD no adv licence is GPL compatable and source may be available and therefor BSD no adv modules should not taint the kernel, on the other hand it's possible to have a BSD no adv licensed module that the source is not available for and therefor it should taint the kernel.
Do not cc: me on this thread.
Check module.h and insmod.c before making statements like that. The BSD no advert license is not considered GPL compatible in the kernel. Only Dual BSD/GPL is non-tainting, because the code is also available under the GPL. Anybody taking advantage of the BSD no adv license to create a binary only module will be tainted.
That causes a problem for some modules in the kernel tree that have the BSD no advert license. Ideally these should be changed to dual BSD/GPL but that may not be possible. Since their source code is always available, we might change them to
"BSD without advertisement clause, source is in the kernel tree"
and add that string to the non-tainting list. BSD no advert code that is not in the kernel tree can either switch to dual BSD/GPL or be tainted, at the choice of the supplier.
The affected code in the kernel tree is
drivers/net/bsd_comp.cdrivers/net/pcmcia/wavelan_cs.cdrivers/net/ppp_deflate.cdrivers/net/slhc.cdrivers/net/strip.cdrivers/scsi/advansys.cdrivers/scsi/dpt_i2o.cdrivers/scsi/pci2000.cdrivers/scsi/pci2220i.cdrivers/scsi/u14-34f.cfs/nls/nls_cp1251.cfs/nls/nls_cp1255.cfs/nls/nls_cp437.cfs/nls/nls_cp737.cfs/nls/nls_cp775.cfs/nls/nls_cp850.cfs/nls/nls_cp852.cfs/nls/nls_cp855.cfs/nls/nls_cp857.cfs/nls/nls_cp860.cfs/nls/nls_cp861.cfs/nls/nls_cp862.cfs/nls/nls_cp863.cfs/nls/nls_cp864.cfs/nls/nls_cp865.cfs/nls/nls_cp866.cfs/nls/nls_cp869.cfs/nls/nls_cp874.cnet/ipv4/netfilter/ipchains_core.c
Quoted 3 lines
and add that string to the non-tainting list. BSD no advert code that is not in the kernel tree can either switch to dual BSD/GPL or be tainted, at the choice of the supplier.
BSD no advertising with no patent issues (and therefore compliant) linked with GPL code ends up as GPL anyway, so I don't see the problem in using the dual BSD/GPL tag
Alan
Quoted 4 lines
exclusive right granted to authors that is within Linus' legal power as holder of the Copyright to the kernel to grant or to restrict. Does Microsoft have a legal right to disallow any third-party drivers from registering themselves with the OS? Does Linus?
If they are derivative works - yes.