Question

Apart from global and static data what else are allocated on Data segment ?

I remember reading somewhere that constant strings are also allocated on Data segment and same memory is used when a reference to same string constant is made

ex:

char* returnPointer()
{
   char *p = "hello world"

   //some code

   return p;
}

void foo()
{
   char *s = "hello world"
  //some code
}

In the above code,

1. Does the memory for constant “hello world” allocated on Data segement or stack (just like any other local variable) ?

2. If allocated on Data segment, does both p and s point to same location ?

Answer 1

Does the memory for constant “hello world” allocated on Data segement or stack (just like any other local variable) ?

This is compiler-specific, but typically string literals live on the initialized data segment (often a special read-only portion of the initialized data segment). I don’t know of any architecture where string literals could plausibly live on the stack.

If allocated on Data segment, does both p and s point to same location ?

That’s up to the compiler.

When I compile your code using gcc 4.4.6, I see the following:

        .section        .rodata
.LC0:
        .string "hello world"
        .text
.globl returnPointer
        .type   returnPointer, @function
returnPointer:
.LFB0:
        .cfi_startproc
        ...
        movq    $.LC0, -8(%rbp)
        ...
        ret
        .cfi_endproc
.LFE0:
        .size   returnPointer, .-returnPointer
.globl foo
        .type   foo, @function
foo:
.LFB1:
        .cfi_startproc
        ...
        movq    $.LC0, -8(%rbp)
        ...
        ret
        .cfi_endproc
.LFE1:
        .size   foo, .-foo
        .ident  "GCC: (GNU) 4.4.6 20110731 (Red Hat 4.4.6-3)"

Here, the string literal is stored in a read-only data (.rodata) section. The compiler is smart enough to realize that exactly the same literal is being used twice in the compilation unit, and it puts only one copy of it in .rodata.