I have written a device driver in linux-2.6.15. There are certain parts of code that involve conditional execution using
#ifdef LINUX statements.(LINUX being my userdefined variable)

Now when i compile my code using a make file, i want to define/undefile linux through that makefile.
But unfortunately make -C /lib/modules/linux/build modules does not allow the declaration of variables that can be communicated to my module..

Please help me..

I am looking for a statement like the following:

make -LINUX -C /lib/modules/2.6.15/build M=`pwd` modules
(obviously this statement doesnt work, so please suggest how to make my module recognize the definition of LINUX variable??)

Usually, the options and definitions are given in the CFLAGS definition in the Makefile. You can define a variable like this: -DLINUX.

You can also define variables on the line before or after the make command. The manual for make will much more indepth information on the order of preference for variables defined before, after the make command, and in the Makefile.
example:
make CFLAGS='-DLINUX'

the CFLAGS and LDFLAGS etc. work only while making a user level file when i'm using gcc command. But thats not the case here.

I am relying on the parent makefile in /lib/modules/2.6.15/build for the gcc command. It is to this file i want to pass my own flag (called LINUX). How do i do that???

Please help someone. My situation is dire....

Does the Makefile have a $(CFLAGS) or $(CXFLAGS) reference in the definition of $(CC). How about the Makefile for your module? I would think that an environment variable would still be defined when your Makefile is called.

I had an earlier message which hasn't shown up. Perhaps it wasn't posted.

You could have the source located outside of the kernel source. Then you can write your own ./configuration file and Makefile.

The nvidia driver and ndiswrapper driver are installed this way. The kernal module is compiled, modprobe scripts are modified as needed, libraries are copied if needed, and depmod -a and ldconfig are run as part of the "make install" target.

Official patches and drivers have configuration items defined in the .config file for the kernel. The #ifdefs in module source will reference these values.
You could also have tests based on whether standard linux definitions exists.
In other words, perform an indirect test to make the same determination. An indirect test for a definition, that only a Linux kernel would have, could be placed in one of your header files and result in defining LINUX then.

Also, if you study several of the text and docbook docs in /usr/source/kernel-<version>/Documentation/ you should get a good grasp on how kernel module developement is done. You may need to install a kernel-doc package, if the kernel source you have was installed from a package.