Search code examples
cassemblygdbreverse-engineeringobjdump

Why addresses of instructions from GDB and Objdump are the same?

I am diving into Reverse Engineering, which is really fun.

I have a question however: why the addresses of instructions I get from GDB and Objdump are the same?

Shouldn't the binary be loaded at a different address every time?

Thank you. Julien

Solution

  • GDB disables ASLR by default. If you set disable-randomization off, then a PIE executable (Position Indepdent) will load at a randomized address even when you run it from inside GDB.

    See 32-bit absolute addresses no longer allowed in x86-64 Linux? for more about PIE.

    Position-dependent executables are always loaded at the same address, and only their stack address can be randomized. The code+data can hard-code addresses as 32-bit absolute, and they don't contain relocation info for every place where that was done. (e.g. like mov $string, %edi ; call puts).

    Look at gcc's code-gen for Hello World with/without -fPIE on the Godbolt compiler explorer.

    .LC0:
    .string "Hello World!"
main:
    lea     rdi, .LC0[rip]     # RIP-relative with -fPIE
    sub     rsp, 8
    call    puts@PLT
    xor     eax, eax
    add     rsp, 8
    ret

    but with -fno-PIE (the default on Godbolt, often not the default on modern Linux distros), you get mov edi, OFFSET FLAT:.LC0, a 32-bit absolute address.

    (The rest of the code is the same, except it emits call puts and lets the linker convert that to call puts@PLT. Use -fno-plt to inline an indirect call through the GOT address.)