cfenollosa_os-tutorial/12-kernel-c/README.md

Concepts you may want to Google beforehand: C, object code, linker, disassemble

Goal: Learn to write the same low-level code as we did with assembler, but in C

Compile

Let's see how the C compiler compiles our code and compare it to the machine code generated with the assembler.

We will start writing a simple program which contains a function, function.c. Open the file and examine it.

To compile system-independent code, we need the flag -ffreestanding, so compile function.c in this fashion:

i386-elf-gcc -ffreestanding -c function.c -o function.o

Let's examine the machine code generated by the compiler:

i386-elf-objdump -d function.o

Now that is something we recognize, isn't it?

Link

Finally, to produce a binary file, we will use the linker. An important part of this step is to learn how high level languages call function labels. Which is the offset where our function will be placed in memory? We don't actually know. For this example, we'll place the offset at 0x0 and use the binary format which generates machine code without any labels and/or metadata

i386-elf-ld -o function.bin -Ttext 0x0 --oformat binary function.o

Note: a warning may appear when linking, disregard it

Now examine both "binary" files, function.o and function.bin using xxd. You will see that the .bin file is machine code, while the .o file has a lot of debugging information, labels, etc.

Decompile

As a curiosity, we will examine the machine code.

ndisasm -b 32 function.bin

More

I encourage you to write more small programs, which feature:

• Local variables localvars.c
• Function calls functioncalls.c
• Pointers pointers.c

Then compile and disassemble them, and examine the resulting machine code. Follow the os-guide.pdf for explanations. Try to answer this question: why does the disassemblement of pointers.c not resemble what you would expect? Where is the ASCII 0x48656c6c6f for "Hello"?