#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>

linux/module.h — The most important header for any kernel module. It defines module_init(), module_exit(), MODULE_LICENSE(), and the core structures the kernel uses to manage modules. You cannot write a module without this.

linux/kernel.h — Provides pr_info(), pr_err(), KERN_INFO, and many common macros and type definitions used throughout kernel code.

linux/init.h — Defines the __init and __exit section markers (explained below).

static int __init hello_init(void)   { ... }
static void __exit hello_exit(void)  { ... }

__init tells the compiler to place the init function in a special memory section called .init.text. After the module’s init function has run and the module is loaded, the kernel can free the memory occupied by that section because the init function will never be called again. This saves RAM — important for embedded systems.

__exit places the exit function in a .exit.text section. If the module is compiled as built-in (Y, not M), the exit function is never needed (you cannot unload built-in code), so the kernel can simply discard it at compile time.

static means these functions are private to this translation unit. They are not exported and cannot be called directly by other kernel code — the kernel only calls them through the registered function pointers.

pr_info("Hello World! Module loaded successfully.\n");

pr_info() is a convenience wrapper around printk() that automatically prepends the KERN_INFO log level. It is the preferred way to print informational messages in modern kernel code.

The output does not go to your terminal — it goes to the kernel ring buffer, which you read with dmesg. Part 5 of this series covers printk and log levels in full detail.

return 0;  /* 0 = success, negative = failure */

The init function must return 0 to signal success. Any negative value signals failure and the kernel will not load the module. Standard error codes from <linux/errno.h> are used: -ENOMEM for memory allocation failure, -ENODEV for device not found, etc. Returning positive values is undefined behaviour in older kernels and should be avoided.

cd ~/hello_module
make

# Expected output (abbreviated):
# make -C /lib/modules/6.8.0-45-generic/build M=/home/user/hello_module modules
# make[1]: Entering directory '/usr/src/linux-headers-6.8.0-45-generic'
#   CC [M]  /home/user/hello_module/hello.o
#   MODPOST /home/user/hello_module/Module.symvers
#   CC [M]  /home/user/hello_module/hello.mod.o
#   LD [M]  /home/user/hello_module/hello.ko
# make[1]: Leaving directory '/usr/src/linux-headers-6.8.0-45-generic'

After a successful build, your directory will contain hello.ko — your kernel module file, ready to load.

sudo insmod hello.ko

# insmod gives no output on success.
# If it fails, check dmesg for the reason.

insmod (insert module) is the low-level command to load a .ko file. It does not resolve dependencies — it just loads the exact file you give it.

# Check if the module is listed in the running kernel:
lsmod | grep hello
# Output:
# hello                  16384  0

# See the printk output from our init function:
dmesg | tail -5
# Output:
# [12345.678901] Hello World! Module loaded successfully.
# [12345.678910] I am now running inside the kernel!

sudo rmmod hello

# Check it is gone:
lsmod | grep hello
# (no output — module removed)

# See the exit function's message:
dmesg | tail -3
# Output:
# [12390.123456] Goodbye! Module unloaded cleanly.

make clean
# Removes: hello.o, hello.ko, hello.mod.c, hello.mod.o,
#          Module.symvers, modules.order, .tmp_versions/