IPC between the kernel and a user space program confuses me. Hopefully some of you can clear things up.

I want to be able to communicate with a kernel module from a user space prorgam in a fast and convinient (? :-)) way.

In my world I can imagine, in general, two basic ways how this can be performed:

1. Writing to some kind of file, where the parties can write/read to communicate.
2. Implement some kind of funtion call.

On the web I have found a few alternatives:

1. Implement a system call in the kernel module
2. Implement an ioctl in the kernel module
3. Use netlink/rtnetlink sockets
4. Register a device (e.g /dev)
5. Use the /proc file system thing

The problem is that I can't weigh pros and cons and then decide which one to use. I would appreciate it very much, if some one could say a few words on this and/or point me to some documentation (A link to a good document can be enough since I'm will to learn these things myself, but a few words from an experienced one will do no harm).

If there is a 2.4 vs 2.6 collision, I'm mostly interested in Linux 2.6.

Best Regards
Rickard

Look at the way that, say, udev is done. Or, look at the module loader.

Perhaps you could describe more about what your kernel-module is to be designed to do...?

Thank youy for your reply, I'll dig into that...

I'll try to create a firewall that can be mastered by passing rules from user space down to a the kernel module that make sure the rules are applied using the netfilter framework. The main purpose is to learn how stuff works... (I dont want to queue packets for userspace using the premade lib since this will be too slow I guess, and too easy...)

BR,
Rick