What is inside an Object File?

An Object File contains information required to execute a program (not only code).

• May not include all implementations, as this can be dynamically loaded.

Information is kept in sections, which are processed differently. Some are:

• .rodata: read-only data, containing strings.

• .got: Global Offset Table - maps symbols to memory locations (offsets).

• .plt: Procedure Linkage Table – uses the PLT to transfer execution to the correct location of a symbol, dealing with external symbols and fixing the GOT.

• .bss: Block Starting Symbol – contains uninitialized variables.

• .dynsym: List of symbols in allocatable memory.

• … many others:

  • To read sections: readelf -S hello

  • To dump all code: objdump -M intel -d hello

How are objects loaded?

The file is split according to existing sections. Each is loaded at a different location (with different access attributes).

Libraries are also mapped in the program address space. All code from libraries is present.

The stack grows downwards, heap grows upwards. On modern OS, growth may be limited, not on microcontrollers.

An interpreter is required to set up the binary in memory.

• ld-Linux.so or ntdll.dll

  • readelf -p .interp filename

• Will handle relocations, and resolve required symbols.

• If lazy-loading is used, relocation is done when the symbol is first used