CODEWITHSUNDEEP
assemblyarm
Muhammad Irshad
Muhammad Irshad

Reputation: 53

How to classify binary arm instruction during dis-assembly

I am writing a simple arm emulator. arm has different formats/Class of instruction as given on this link After reading instruction in binary format How can I determine to which class/format instruction belongs.

Upvotes: 3

Views: 194

Answers (1)

Johan
Johan

Reputation: 76723

I've recently written a ARM emulator myself.

Here's the strategy I've used:

Use a mask and a value to test against after the mask has been applied.

Here's a snippet from the GNU disassembler for ARM:

  (* V7 instructions.  *)
  (arch:ARM_EXT_V7; value:$f910f000; mask:$ff70f000; i:_und; assembly:'pli'#9'%a'),
  (arch:ARM_EXT_V7; value:$f3af80f0; mask:$fffffff0; i:_und; assembly:'dbg'#9'#%0-3d'),
  (arch:ARM_EXT_V7; value:$f3bf8f50; mask:$fffffff0; i:_und; assembly:'dmb'#9'%U'),
  (arch:ARM_EXT_V7; value:$f3bf8f40; mask:$fffffff0; i:_und; assembly:'dsb'#9'%U'),
  (arch:ARM_EXT_V7; value:$f3bf8f60; mask:$fffffff0; i:_und; assembly:'isb'#9'%U'),

And here's how you apply it:

  1. Read next instruction bytes.
  2. Correct for big/little endian mode if needed.
  3. if (InstructionBytes and Mask) = value then we have a match
  4. Interpreted the remaining bits according to the instructions in the assembly string.
  5. If ARM interpreted the condition code etc...

You need to make sure you apply the masks in the correct order.
Instructions introduced in a newer revision should always be checked first.

There is no need to worry whether you're in ARM or Thumb mode.
Just disassemble for both and store both disassemblies side by side. If a mode change occurs switch the table you're reading your disassembled instructions from.

Upvotes: 1

Related Questions