Inserting code from userspace to kernel space

Inserting code from userspace to kernel space

[Bosko Radivojevic]

Hi!

I'm looking for a way to insert code from user space to the kernel
space, but without using kernel module. Actually, we are working on a
system that has to allow end user to change part of code (one or two
functions) that is executed from the kernel space.

For now we have two different kernel modules, one exporting a function
which another module uses. Process of compiling kernel module is quite
ineligible for standard end user (along with the requirement to
support multi platform cross compiling) compared to just cross
compiling a simple code that doesn't use any libraries than libc.

I know this sounds quite weird, but maybe there is a project that can help us ;)

Sincerely,
Bosko

Re: Inserting code from userspace to kernel space

[Jan Engelhardt]

On Thursday 2008-05-22 14:41, Bosko Radivojevic wrote:

>Hi!
>
>I'm looking for a way to insert code from user space to the kernel
>space, but without using kernel module. Actually, we are working on a
>system that has to allow end user to change part of code (one or two
>functions) that is executed from the kernel space.
>
>For now we have two different kernel modules, one exporting a function
>which another module uses. Process of compiling kernel module is quite
>ineligible for standard end user (along with the requirement to
>support multi platform cross compiling) compared to just cross
>compiling a simple code that doesn't use any libraries than libc.
>
>I know this sounds quite weird, but maybe there is a project that can help us ;)

I'm impressed you have yet refrained from coming up with the
idea of embedding Perl into the kernel.

Re: Inserting code from userspace to kernel space

[Helge Hafting]

Bosko Radivojevic wrote:
> Hi!
>
> I'm looking for a way to insert code from user space to the kernel
> space, but without using kernel module. Actually, we are working on a
> system that has to allow end user to change part of code (one or two
> functions) that is executed from the kernel space.
>
> For now we have two different kernel modules, one exporting a function
> which another module uses. Process of compiling kernel module is quite
> ineligible for standard end user (along with the requirement to
> support multi platform cross compiling) compared to just cross
> compiling a simple code that doesn't use any libraries than libc.
>
> I know this sounds quite weird, but maybe there is a project that can help us ;)
>   
Consider telling exactly what you want to do, what is this for?
Perhaps there is a better/more standard way of doing it.

Looks like your end users are capable of compiling ordinary code
but not kernel modules? If this is so, consider just automating
the kernel module part for them. I.e. provide a module that call
a function, they can then make that function and you can make a script
that link their function with your kernel module. Then they won't need
to know much about kernel modules.

But the kernel module can't use libc (and what would you want to
use libc for anyway? libc mostly helps userspace communicate
with the kernel, no _need_ for that when your code is in the
kernel already.)

Or do you merely need your kernel code to run a user supplied program
now and then?  The kernel code can't call userspace code directly,
but you could make a device driver for this purpose.

The userspace code can read from the device, and block. When the kernel
need to run userspace code, it unblocks the userspace program.  If you
need to transfer data to userspace, simply let the userpace program
read the data from your device. and write stuff back, if needed.

Re: Inserting code from userspace to kernel space

[Shuduo Sang]

On Thu, May 22, 2008 at 8:41 PM, Bosko Radivojevic
<bosko.radivojevic@gmail.com> wrote:
> Hi!
>
> I'm looking for a way to insert code from user space to the kernel
> space, but without using kernel module. Actually, we are working on a
> system that has to allow end user to change part of code (one or two
> functions) that is executed from the kernel space.
>
> For now we have two different kernel modules, one exporting a function
> which another module uses. Process of compiling kernel module is quite
> ineligible for standard end user (along with the requirement to
> support multi platform cross compiling) compared to just cross
> compiling a simple code that doesn't use any libraries than libc.
>
> I know this sounds quite weird, but maybe there is a project that can help us ;)
>

actually you can implement a parser/interpretor, whatever you name it,
in kernel space. And you can send the
function code from user space to kernel space by device driver
interface. the parser can parse that and do something
practical function in kernel.

I have worked on a similar task in a product powered by linux. It works great.

> Sincerely,
> Bosko
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

Re: Inserting code from userspace to kernel space

[devzero]

>> I know this sounds quite weird, but maybe there is a project that can help us ;)
>>
>
>actually you can implement a parser/interpretor, whatever you name it,
>in kernel space. And you can send the
>function code from user space to kernel space by device driver
>interface. the parser can parse that and do something
>practical function in kernel.
>
>I have worked on a similar task in a product powered by linux. It works great.

any pointer to the sourcecode so bosko (and others) can profit from that ?




List:       linux-kernel
Subject:    Re: Inserting code from userspace to kernel space
From:       "Shuduo Sang" <sangshuduo () gmail ! com>
Date:       2008-05-22 16:04:35
Message-ID: ee101c360805220904h1c69db12q3048d0347d27c461 () mail ! gmail ! com
[Download message RAW]

On Thu, May 22, 2008 at 8:41 PM, Bosko Radivojevic
<bosko.radivojevic@gmail.com> wrote:
> Hi!
>
> I'm looking for a way to insert code from user space to the kernel
> space, but without using kernel module. Actually, we are working on a
> system that has to allow end user to change part of code (one or two
> functions) that is executed from the kernel space.
>
> For now we have two different kernel modules, one exporting a function
> which another module uses. Process of compiling kernel module is quite
> ineligible for standard end user (along with the requirement to
> support multi platform cross compiling) compared to just cross
> compiling a simple code that doesn't use any libraries than libc.
>
> I know this sounds quite weird, but maybe there is a project that can help us ;)
>

actually you can implement a parser/interpretor, whatever you name it,
in kernel space. And you can send the
function code from user space to kernel space by device driver
interface. the parser can parse that and do something
practical function in kernel.

I have worked on a similar task in a product powered by linux. It works great.

> Sincerely,
> Bosko
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>
______________________________________________________________
Jeden Monat 1 hochkarätiger maxdome-Blockbuster GRATIS!
Exklusiv für alle WEB.DE Nutzer. http://www.blockbuster.web.de

Re: Inserting code from userspace to kernel space

[Shuduo Sang]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=UTF-8, Size: 2716 bytes --]

On Fri, May 23, 2008 at 10:30 PM,  <devzero@web.de> wrote:>>> I know this sounds quite weird, but maybe there is a project that can help us ;)>>>>>>>actually you can implement a parser/interpretor, whatever you name it,>>in kernel space. And you can send the>>function code from user space to kernel space by device driver>>interface. the parser can parse that and do something>>practical function in kernel.>>>>I have worked on a similar task in a product powered by linux. It works great.>> any pointer to the sourcecode so bosko (and others) can profit from that ?>
someone does not want to leak propertied algorithm out with GPL. Thiscan be a way to protect.
>>>> List:       linux-kernel> Subject:    Re: Inserting code from userspace to kernel space> From:       "Shuduo Sang" <sangshuduo () gmail ! com>> Date:       2008-05-22 16:04:35> Message-ID: ee101c360805220904h1c69db12q3048d0347d27c461 () mail ! gmail ! com> [Download message RAW]>> On Thu, May 22, 2008 at 8:41 PM, Bosko Radivojevic> <bosko.radivojevic@gmail.com> wrote:>> Hi!>>>> I'm looking for a way to insert code from user space to the kernel>> space, but without using kernel module. Actually, we are working on a>> system that has to allow end user to change part of code (one or two>> functions) that is executed from the kernel space.>>>> For now we have two different kernel modules, one exporting a function>> which another module uses. Process of compiling kernel module is quite>> ineligible for standard end user (along with the requirement to>> support multi platform cross compiling) compared to just cross>> compiling a simple code that doesn't use any libraries than libc.>>>> I know this sounds quite weird, but maybe there is a project that can help us ;)>>>> actually you can implement a parser/interpretor, whatever you name it,> in kernel space. And you can send the> function code from user space to kernel space by device driver> interface. the parser can parse that and do something> practical function in kernel.>> I have worked on a similar task in a product powered by linux. It works great.>>> Sincerely,>> Bosko>> -->> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in>> the body of a message to majordomo@vger.kernel.org>> More majordomo info at  http://vger.kernel.org/majordomo-info.html>> Please read the FAQ at  http://www.tux.org/lkml/>>> ______________________________________________________________> Jeden Monat 1 hochkarätiger maxdome-Blockbuster GRATIS!> Exklusiv für alle WEB.DE Nutzer. http://www.blockbuster.web.de>>ÿôèº{.nÇ+‰·Ÿ®‰­†+%ŠËÿ±éݶ\x17¥Šwÿº{.nÇ+‰·¥Š{±þG«éÿŠ{ayº\x1dʇڙë,j\a­¢f£¢·hšïêÿ‘êçz_è®\x03(­éšŽŠÝ¢j"ú\x1a¶^[m§ÿÿ¾\a«þG«éÿ¢¸?™¨è­Ú&£ø§~á¶iO•æ¬z·švØ^\x14\x04\x1a¶^[m§ÿÿÃ\fÿ¶ìÿ¢¸?–I¥

Re: Inserting code from userspace to kernel space

[David Newall]

Shuduo Sang wrote:
> someone does not want to leak propertied algorithm out with GPL. This
> can be a way to protect.

Perhaps you want something similar to the firmware files that some
devices need loaded?  (By the way, I think the word is "proprietary".)

Re: Inserting code from userspace to kernel space

[Alex Belits]

On Thu, 22 May 2008, Bosko Radivojevic wrote:

> I'm looking for a way to insert code from user space to the kernel
> space, but without using kernel module. Actually, we are working on a
> system that has to allow end user to change part of code (one or two
> functions) that is executed from the kernel space.
>
> For now we have two different kernel modules, one exporting a function
> which another module uses. Process of compiling kernel module is quite
> ineligible for standard end user (along with the requirement to
> support multi platform cross compiling) compared to just cross
> compiling a simple code that doesn't use any libraries than libc.

Why? You don't have to recompile everything, just link a module before 
loaing it -- a part containing the function can be one object file, the 
rest of the module (a wrapper) is compiled once for a particular kernel 
version, and both are linked together, producing a module valid for a 
particular kenel version.

This is how most of proprietary drivers are distributed, and Ubuntu has a 
whole infrastructure around that mechanism that simplifies transition 
between versions.

-- 
Alex

Re: Inserting code from userspace to kernel space

[Bart Van Assche]

On Thu, May 22, 2008 at 2:41 PM, Bosko Radivojevic
<bosko.radivojevic@gmail.com> wrote:
> I'm looking for a way to insert code from user space to the kernel
> space, but without using kernel module.

That would create a huge security risk. The number of Linux-specific
viruses is currently very limited, and I'd like to keep it that way.

Are you familiar with user-space device drivers ?

Bart.

Re: Inserting code from userspace to kernel space

[devzero]

> someone does not want to leak propertied algorithm out with GPL. This
> can be a way to protect.

this can be illegal, too.

http://www.gnu.org/licenses/gpl-faq.html#ModifiedJustBinary

>Can I release a modified version of a GPL-covered program in binary form only?
>No. The whole point of the GPL is that all modified versions must be free software — which means, 
>in particular, that the source code of the modified version is available to the users




> On Fri, May 23, 2008 at 10:30 PM,  <devzero@web.de> wrote:
> >>> I know this sounds quite weird, but maybe there is a project that can help us ;)
> >>>
> >>
> >>actually you can implement a parser/interpretor, whatever you name it,
> >>in kernel space. And you can send the
> >>function code from user space to kernel space by device driver
> >>interface. the parser can parse that and do something
> >>practical function in kernel.
> >>
> >>I have worked on a similar task in a product powered by linux. It works great.
> >
> > any pointer to the sourcecode so bosko (and others) can profit from that ?
> >
> 
> someone does not want to leak propertied algorithm out with GPL. This
> can be a way to protect.
> 
> >
> >
> >
> > List:       linux-kernel
> > Subject:    Re: Inserting code from userspace to kernel space
> > From:       "Shuduo Sang" <sangshuduo () gmail ! com>
> > Date:       2008-05-22 16:04:35
> > Message-ID: ee101c360805220904h1c69db12q3048d0347d27c461 () mail ! gmail ! com
> > [Download message RAW]
> >
> > On Thu, May 22, 2008 at 8:41 PM, Bosko Radivojevic
> > <bosko.radivojevic@gmail.com> wrote:
> >> Hi!
> >>
> >> I'm looking for a way to insert code from user space to the kernel
> >> space, but without using kernel module. Actually, we are working on a
> >> system that has to allow end user to change part of code (one or two
> >> functions) that is executed from the kernel space.
> >>
> >> For now we have two different kernel modules, one exporting a function
> >> which another module uses. Process of compiling kernel module is quite
> >> ineligible for standard end user (along with the requirement to
> >> support multi platform cross compiling) compared to just cross
> >> compiling a simple code that doesn't use any libraries than libc.
> >>
> >> I know this sounds quite weird, but maybe there is a project that can help us ;)
> >>
> >
> > actually you can implement a parser/interpretor, whatever you name it,
> > in kernel space. And you can send the
> > function code from user space to kernel space by device driver
> > interface. the parser can parse that and do something
> > practical function in kernel.
> >
> > I have worked on a similar task in a product powered by linux. It works great.
> >
> >> Sincerely,
> >> Bosko
> >> --
> >> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> >> the body of a message to majordomo@vger.kernel.org
> >> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> >> Please read the FAQ at  http://www.tux.org/lkml/
> >>
> > ______________________________________________________________
> > Jeden Monat 1 hochkarätiger maxdome-Blockbuster GRATIS!
> > Exklusiv für alle WEB.DE Nutzer. http://www.blockbuster.web.de
> >
> >
> 


_______________________________________________________________
Schon gehört? Der neue WEB.DE MultiMessenger kann`s mit allen: 
http://www.produkte.web.de/messenger/?did=3016

Re: Inserting code from userspace to kernel space

[David Newall]

devzero@web.de wrote:
> this can be illegal, too.
>
> http://www.gnu.org/licenses/gpl-faq.html#ModifiedJustBinary
>   

I think what he's trying to achieve is an unmodified or modified but
fully GPL kernel plus a separate, proprietary binary to be loaded at
runtime.

Re: Inserting code from userspace to kernel space

[Tom Spink]

Hi David,

2008/5/25 David Newall <davidn@davidnewall.com>:
> devzero@web.de wrote:
>> this can be illegal, too.
>>
>> http://www.gnu.org/licenses/gpl-faq.html#ModifiedJustBinary
>>
>
> I think what he's trying to achieve is an unmodified or modified but
> fully GPL kernel plus a separate, proprietary binary to be loaded at
> runtime.

I think so to, but I think Shuduo misunderstood what question DevZero
was asking.  DevZero was asking for the source-code to the
infrastructure Shuduo used to implement this mechanism, i.e. the
loader and parser/interpreter that lives in the kernel - not the
application code.

-- 
Tom Spink

Re: Inserting code from userspace to kernel space

[Shuduo Sang]

On Sun, May 25, 2008 at 11:47 AM, David Newall <davidn@davidnewall.com> wrote:
> Shuduo Sang wrote:
>> someone does not want to leak propertied algorithm out with GPL. This
>> can be a way to protect.
>
> Perhaps you want something similar to the firmware files that some
> devices need loaded?  (By the way, I think the word is "proprietary".)
>

sorry for my poor english and thanks for your correction. :)

Re: Inserting code from userspace to kernel space

[Geert Uytterhoeven]

On Thu, 22 May 2008, Bosko Radivojevic wrote:
> For now we have two different kernel modules, one exporting a function
> which another module uses. Process of compiling kernel module is quite
> ineligible for standard end user (along with the requirement to
> support multi platform cross compiling) compared to just cross
> compiling a simple code that doesn't use any libraries than libc.

For cross-compiling a kernel module, you need:
  - a cross-compiler

For cross-compiling `a simple code that doesn't use any libraries than
libc', you need:
  - a cross-compiler
  - a cross-libc

Hence the kernel module approach needs less ;-)

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: Inserting code from userspace to kernel space

[Jiri Kosina]

On Sun, 25 May 2008, David Newall wrote:

> > someone does not want to leak propertied algorithm out with GPL. This
> > can be a way to protect.
> Perhaps you want something similar to the firmware files that some 
> devices need loaded?  (By the way, I think the word is "proprietary".) 

This is different, because the firmware that is loaded by 
request_firmware() is not meant to be interpreted/executed by the kernel, 
drivers usually just pass it over to the device.

-- 
Jiri Kosina
SUSE Labs

Re: Inserting code from userspace to kernel space

[Jike Song]

Did you try systemtap?  It's available at:

http://sourceware.org/systemtap/

--
Thanks and regards,
Jike

On Thu, May 22, 2008 at 8:41 PM, Bosko Radivojevic
<bosko.radivojevic@gmail.com> wrote:
> Hi!
>
> I'm looking for a way to insert code from user space to the kernel
> space, but without using kernel module. Actually, we are working on a
> system that has to allow end user to change part of code (one or two
> functions) that is executed from the kernel space.
>
> For now we have two different kernel modules, one exporting a function
> which another module uses. Process of compiling kernel module is quite
> ineligible for standard end user (along with the requirement to
> support multi platform cross compiling) compared to just cross
> compiling a simple code that doesn't use any libraries than libc.
>
> I know this sounds quite weird, but maybe there is a project that can help us ;)
>
> Sincerely,
> Bosko
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

Re: Inserting code from userspace to kernel space

[David Newall]

Jiri Kosina wrote:
> This is different, because the firmware that is loaded by 
> request_firmware() is not meant to be interpreted/executed by the kernel, 
> drivers usually just pass it over to the device.

Usually, but not obligatory.  It seems like it might be useful to
overload the mechanism for the specified requirement.  It's just an idea.