Hi Sandeep, probably you want to look at how your program is loaded in memory. For example an ELF binary is understood by ELF format handler inside kernel. Format handler supply their load_binary methods to load a program image im memory and initial its different virtual memory areas (stack, heap etc.). exec system call searches for correct format handler for you based on file header. Please go thoroghly through Understanding the Linux Kernel, 3rd Edition, Chapter 20. Program ExZecution. -Rajat On Fri, May 27, 2011 at 12:14 PM, sandeep kumar <coolsandyforyou@gmail.com> wrote:

hi all, I am new to the linux kernel internals. I know there is a memory management subsystem which handles all the memory related things.

But Now i want to know a bit deeper how things work.

I want to start with the following question, How a program could generate the memory addresses for its variables, when it is about to run?

Can please somebody give pointers how to learn this kind of things like, "in the early stages (when our program is about to be executed..about to become a process) what are the things that will be done by the kernel?"

Please help me in this regard,

Thanking you, Sandeep Kumar A.

_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Hi... On Fri, May 27, 2011 at 13:44, sandeep kumar <coolsandyforyou@gmail.com> wrote:

I want to start with the following question, How a program could generate the memory addresses for its variables, when it is about to run?

first, linker will determine in which section and which offset (from the start of the program) the variables will be positioned. If it is a dynamic variable (one that ends up in stack), it will determined dynamically later according to stack position. later, loader will reposition those variables based on real memory address positioning. This is based on the information placed inside the binary (ELF header). Usually, it is followed strictly, but if it is PIE (Position Independent Executable), further flexible reposition is allowed based on certain policy. those address (for static variables) are then substituted into GOT (Global Offset Table), some kind of lookup table to determined address of variables (and function address too actually). for dynamic variable, usually simply using compiler generated code, it is sufficient to reference them. This is because they are referenced using relative offset against base address of frame (which is usually saved in esp or ebp in x86 arch) I hope I correctly describe it.... -- regards, Mulyadi Santosa Freelance Linux trainer and consultant blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com