Practical Advice for Assembly Programming

• The .section directive specifies which “section” of the executable program assembled code or data will be placed in

Put things in the right place!

• Code goes in .text
• Read-only data such as string constants go in .rodata
• Uninitialized (zero-filled) variables and buffers go in .bss
  • Use the .space directive to indicate how large these are
• Initialized (non-zero-filled) variables and buffers go in .data
  • There are various directives such as .byte, .2byte, .4byte, etc. to specify initialized data values

Labels:

• Labels are names representing addresses of code or data in memory

• For functions and global variables, use appropriate names

  • Functions and data exported to other modules must be marked with .globl

• For control-flow targets within a function, use local labels

  • These are labels which start with .L (dot, followed by upper case L)
  • The assembler will not add these to the module’s symbol table
  • Using “normal” labels for control flow makes debugging difficult because gdb thinks they are functions!

Using GDB

• You can debug assembly programs using gdb!

• “Debugging by adding print statements” is much less practical for assembly programs than programs in a high level language

  • Which isn’t to say it’s not possible or (occasionally) useful

• Being able to use gdb confidently will greatly enhance your ability to develop working assembly language programs

• Set breakpoints (break main, break myProg.S:123)

• where: see current call stack

• If you compiled your code with debugging symbols (i.e., using -g flag to gcc), next and step commands work as expected!

• If code is compiled without debug symbols, it’s more difficult:

  • disassemble (or just disas): display assembly code of currentfunction
  • stepi: step to next instruction
  • nexti: step to next instruction (stepping over call instructions)

• Set breakpoints (break main, break myProg.S:123)

• where: see current call stack

• If you compiled your code with debugging symbols (i.e., using -g flag to gcc), next and step commands work as expected!

• If code is compiled without debug symbols, it’s more difficult:

  • disassemble (or just disas): display assembly code of current function
  • stepi: step to next instruction
  • nexti: step to next instruction (stepping over call instructions)

• Use $ prefix to refer to registers (e.g., $rax, $edi, etc.)

• Use print and casts to C data types when inspecting data:

• Print 64 bit value %rsp points to: print *(unsigned long *)$rsp

• Print character string %rdi points to: print (char *)$rdi

• Print fourth element of array of int elements that %r12 points to: print ((int *)$r12)[3]

• Print contents of %rcx is hexadecimal: print/x $rcx

Assembly Program Outline

.section .rodata
	// read only strings

.section .bss
	// zero-initialized global variables

.section .text

	.global main
main:
	subq $8, %rsp

.LinputLoop:
	// do stuff