关于init_module的一个邪恶想法

sales are assholes, engineers are dicks. assholes always ask dicks to do dirty things, when it's done, assholes are happy, while engineers are covered with shit all over them.

sale: "please modify this driver, make it not work on other OS."
me: "why?"
after 0.1 second...
me: "i see."
me: "where is the source code?"
sale: "no source code"
me: "i cannot modify the binary without the source code."
sale: "can you just hide the driver somewhere?"
me: "they will find it anyway."
sale: "anyway, i just want it tomorrow."
me: "OK, let me think what i can do."

OK, basic knowledge of linking and ELF is not totally useless. here is how to do that:

support the original driver is foo.ko,

1. rename the __this_module to something else (e.g. __THIS_module)

2. write some code to replace the original __this_module symbol:

struct module __this_module
__attribute__((section(".gnu.linkonce.this_module"))) = {
.name = __stringify(KBUILD_MODNAME),
.init = new_module_init,
.exit = cleanup_module,
};

3. in new_module_init(), check if OS is ours, if NOT, puke. otherwise, call the original module_init():

static int is_ax(void)
{
        if (strstr(system_utsname.release, "AX") == NULL)
                return 0;
        else
                return 1;
}

static int new_module_init(void)
{
        int err = -1;
        if (is_ax())
                err = original_init();
        return err;
}

4. link foo.ko with the code generated in step 2 and 3. build a new driver module.

that's it. simple and evil.

when you dive deep into technical issues, the notion of ethic simply disappears, so you wont go crazy.