Program Sections

26 March, 2017 | santosh

Every program is split into multiple sections. In general, these are:

  • The code segment containing executable code
  • The stack segment
  • Data segment, which constitute the following:
BSS
Uninitialized data
Heap
Memory allocated during run time
Data
Initialized data

They are just the defaults, there can be many more sections in a binary. For example our sample code contains the following different sections.

$ objdump -h a.out

Most of the section compiled into the binary is either empty or just debugging information. The most important of the sections are the .text, .bss, .data. The .text section contains the executable instructions. This data can be viewed using objdump

$ # disassemble only the .text section
$ objdump -d -j .text a.out

The next section .bss will contain uninitialized data. The C standard say that all uninitialized global variables much be set to zero. So instead of wasting space in the binary with zeroes, it contains only the size of the .bss section, as the kernel provided memory will always be initialized to zero. The other sections too are not present in the binary as they are available during the run time, like the stack and the heap sections. The symbols present in various sections can be found using the nm command.

$ nm a.out

The type of symbol is printed in the second column. b or B means BSS and d or D refers to the symbols in .data section.

In this section we have seen one more new term - the binary loader. This is the dynamic loader which links the executable with the shared libraries(.so) at run time. This is the topic of the next section.

Continue reading other posts in the series, "Life of a process"

Site design and logo and content © fossix.org