dasm_x86.lua

采用C语言写的Lua的解释器的代码！Lua不用介绍了吧

      wputmodrm(t.xsc, t.xreg, 5)
    else
      -- Pure displacement.
      wputmodrm(0, s, 5) -- [disp] -> (0, s, ebp)
    end
    wputdarg(disp)
    return
  end

  local m
  if tdisp == "number" then -- Check displacement size at assembly time.
    if disp == 0 and t.reg ~= 5 then m = 0  -- [ebp] -> [ebp+0] (in SIB, too)
    elseif disp >= -128 and disp <= 127 then m = 1
    else m = 2 end
  elseif tdisp == "table" then
    m = 2
  end

  -- Index register present or esp as base register: need SIB encoding.
  if t.xreg or t.reg == 4 then
    wputmodrm(m or 2, s, 4) -- ModRM.
    if (m == nil or imark) and tdisp ~= "table" then waction("MARK") end
    wputmodrm(t.xsc or 0, t.xreg or 4, t.reg) -- SIB.
  else
    wputmodrm(m or 2, s, t.reg) -- ModRM.
    if imark and (m == 1 or m == 2) then waction("MARK") end
  end

  -- Put displacement.
  if m == 1 then wputsbarg(disp)
  elseif m == 2 then wputdarg(disp)
  elseif not m then waction("DISP", disp) end
end

------------------------------------------------------------------------------

-- Return human-readable operand mode string.
local function opmodestr(op, args)
  local m = {}
  for i=1,#args do
    local a = args[i]
    m[#m+1] = sub(a.mode, 1, 1)..(a.opsize or "?")
  end
  return op.." "..concat(m, ",")
end

-- Convert number to valid integer or nil.
local function toint(expr)
  local n = tonumber(expr)
  if n then
    if n % 1 ~= 0 or n < -2147483648 or n > 4294967295 then
      werror("bad integer number `"..expr.."'")
    end
    return n
  end
end

-- Parse immediate expression.
local function immexpr(expr)
  -- &expr (pointer)
  if sub(expr, 1, 1) == "&" then
    return "iPJ", format("(ptrdiff_t)(%s)", sub(expr,2))
  end

  local prefix = sub(expr, 1, 2)
  -- =>expr (pc label reference)
  if prefix == "=>" then
    return "iJ", sub(expr, 3)
  end
  -- ->name (global label reference)
  if prefix == "->" then
    return "iJ", map_global[sub(expr, 3)]
  end

  -- [<>][1-9] (local label reference)
  local dir, lnum = match(expr, "^([<>])([1-9])$")
  if dir then -- Fwd: 247-255, Bkwd: 1-9.
    return "iJ", lnum + (dir == ">" and 246 or 0)
  end

  -- expr (interpreted as immediate)
  return "iI", expr
end

-- Parse displacement expression: +-num, +-expr, +-opsize*num
local function dispexpr(expr)
  local disp = expr == "" and 0 or toint(expr)
  if disp then return disp end
  local c, dispt = match(expr, "^([+-])%s*(.+)$")
  if c == "+" then
    expr = dispt
  elseif not c then
    werror("bad displacement expression `"..expr.."'")
  end
  local opsize, tailops = match(dispt, "^(%w+)%s*%*%s*(.+)$")
  local ops, imm = map_opsize[opsize], toint(tailops)
  if ops and imm then
    if c == "-" then imm = -imm end
    return imm*map_opsizenum[ops]
  end
  local mode, iexpr = immexpr(dispt)
  if mode == "iJ" then
    if c == "-" then werror("cannot invert label reference") end
    return { iexpr }
  end
  return expr -- Need to return original signed expression.
end

-- Parse register or type expression.
local function rtexpr(expr)
  if not expr then return end
  local tname, ovreg = match(expr, "^([%w_]+):(@[%w_]+)$")
  local tp = map_type[tname or expr]
  if tp then
    local reg = ovreg or tp.reg
    local rnum = map_reg_num[reg]
    if not rnum then
      werror("type `"..(tname or expr).."' needs a register override")
    end
    if not map_reg_valid_base[reg] then
      werror("bad base register override `"..(map_reg_rev[reg] or reg).."'")
    end
    return reg, rnum, tp
  end
  return expr, map_reg_num[expr]
end

-- Parse operand and return { mode, opsize, reg, xreg, xsc, disp, imm }.
local function parseoperand(param)
  local t = {}

  local expr = param
  local opsize, tailops = match(param, "^(%w+)%s*(.+)$")
  if opsize then
    t.opsize = map_opsize[opsize]
    if t.opsize then expr = tailops end
  end

  local br = match(expr, "^%[%s*(.-)%s*%]$")
  repeat
    if br then
      t.mode = "xm"

      -- [disp]
      t.disp = toint(br)
      if t.disp then
	t.mode = "xmO"
	break
      end

      -- [reg...]
      local tp
      local reg, tailr = match(br, "^([@%w_:]+)%s*(.*)$")
      reg, t.reg, tp = rtexpr(reg)
      if not t.reg then
	-- [expr]
	t.mode = "xmO"
	t.disp = dispexpr("+"..br)
	break
      end

      -- [xreg*xsc] or [xreg*xsc+-disp] or [xreg*xsc+-expr]
      local xsc, tailsc = match(tailr, "^%*%s*([1248])%s*(.*)$")
      if xsc then
	if not map_reg_valid_index[reg] then
	  werror("bad index register `"..map_reg_rev[reg].."'")
	end
	t.xsc = map_xsc[xsc]
	t.xreg = t.reg
	t.reg = nil
	t.disp = dispexpr(tailsc)
	break
      end
      if not map_reg_valid_base[reg] then
	werror("bad base register `"..map_reg_rev[reg].."'")
      end

      -- [reg] or [reg+-disp]
      t.disp = toint(tailr) or (tailr == "" and 0)
      if t.disp then break end

      -- [reg+xreg...]
      local xreg, tailx = match(tailr, "^+%s*([@%w_:]+)%s*(.*)$")
      xreg, t.xreg, tp = rtexpr(xreg)
      if not t.xreg then
	-- [reg+-expr]
	t.disp = dispexpr(tailr)
	break
      end
      if not map_reg_valid_index[xreg] then
	werror("bad index register `"..map_reg_rev[xreg].."'")
      end

      -- [reg+xreg*xsc...]
      local xsc, tailsc = match(tailx, "^%*%s*([1248])%s*(.*)$")
      if xsc then
	t.xsc = map_xsc[xsc]
	tailx = tailsc
      end

      -- [...] or [...+-disp] or [...+-expr]
      t.disp = dispexpr(tailx)
    else
      -- imm or opsize*imm
      local imm = toint(expr)
      if not imm and sub(expr, 1, 1) == "*" and t.opsize then
	imm = toint(sub(expr, 2))
	if imm then
	  imm = imm * map_opsizenum[t.opsize]
	  t.opsize = nil
	end
      end
      if imm then
	if t.opsize then werror("bad operand size override") end
	local m = "i"
	if imm == 1 then m = m.."1" end
	if imm >= 4294967168 and imm <= 4294967295 then imm = imm-4294967296 end
	if imm >= -128 and imm <= 127 then m = m.."S" end
	t.imm = imm
	t.mode = m
	break
      end

      local tp
      local reg, tailr = match(expr, "^([@%w_:]+)%s*(.*)$")
      reg, t.reg, tp = rtexpr(reg)
      if t.reg then
	-- reg
	if tailr == "" then
	  if t.opsize then werror("bad operand size override") end
	  t.opsize = map_reg_opsize[reg]
	  if t.opsize == "f" then
	    t.mode = t.reg == 0 and "fF" or "f"
	  else
	    if reg == "@w4" then wwarn("bad idea, try again with `esp'") end
	    t.mode = t.reg == 0 and "rmR" or (reg == "@b1" and "rmC" or "rm")
	  end
	  break
	end

	-- type[idx], type[idx].field, type->field -> [reg+offset_expr]
	if not tp then werror("bad operand `"..param.."'") end
	t.mode = "xm"
	t.disp = format(tp.ctypefmt, tailr)
      else
	t.mode, t.imm = immexpr(expr)
	if sub(t.mode, -1) == "J" then
	  if t.opsize and t.opsize ~= addrsize then
	    werror("bad operand size override")
	  end
	  t.opsize = addrsize
	end
      end
    end
  until true
  return t
end

------------------------------------------------------------------------------
-- x86 Template String Description
-- ===============================
--
-- Each template string is a list of [match:]pattern pairs,
-- separated by "|". The first match wins. No match means a
-- bad or unsupported combination of operand modes or sizes.
--
-- The match part and the ":" is omitted if the operation has
-- no operands. Otherwise the first N characters are matched
-- against the mode strings of each of the N operands.
--
-- The mode string for each operand type is (see parseoperand()):
--   Integer register: "rm", +"R" for eax, ax, al, +"C" for cl
--   FP register:      "f",  +"F" for st0
--   Index operand:    "xm", +"O" for [disp] (pure offset)
--   Immediate:        "i",  +"S" for signed 8 bit, +"1" for 1,
--                     +"I" for arg, +"P" for pointer
--   Any:              +"J" for valid jump targets
--
-- So a match character "m" (mixed) matches both an integer register
-- and an index operand (to be encoded with the ModRM/SIB scheme).
-- But "r" matches only a register and "x" only an index operand
-- (e.g. for FP memory access operations).
--
-- The operand size match string starts right after the mode match
-- characters and ends before the ":". "dwb" is assumed, if empty.
-- The effective data size of the operation is matched against this list.
--
-- If only the regular "b", "w", "d", "q", "t" operand sizes are
-- present, then all operands must be the same size. Unspecified sizes
-- are ignored, but at least one operand must have a size or the pattern
-- won't match (use the "byte", "word", "dword", "qword", "tword"
-- operand size overrides. E.g.: mov dword [eax], 1).
--
-- If the list has a "1" or "2" prefix, the operand size is taken
-- from the respective operand and any other operand sizes are ignored.
-- If the list contains only ".", all operand sizes are ignored.
-- If the list has a "/" prefix, the concatenated (mixed) operand sizes
-- are compared to the match.
--
-- E.g. "rrdw" matches for either two dword registers or two word
-- registers. "Fx2dq" matches an st0 operand plus an index operand
-- pointing to a dword (float) or qword (double).
--
-- Every character after the ":" is part of the pattern string:
--   Hex chars are accumulated to form the opcode (left to right).
--   "n"       disables the standard opcode mods
--             (otherwise: -1 for "b", o16 prefix for "w")
--   "r"/"R"   adds the reg. number from the 1st/2nd operand to the opcode.
--   "m"/"M"   generates ModRM/SIB from the 1st/2nd operand.
--             The spare 3 bits are either filled with the last hex digit or
--             the result from a previous "r"/"R". The opcode is restored.
--
-- All of the following characters force a flush of the opcode:
--   "o"/"O"   stores a pure 32 bit disp (offset) from the 1st/2nd operand.
--   "S"       stores a signed 8 bit immediate from the last operand.
--   "U"       stores an unsigned 8 bit immediate from the last operand.
--   "W"       stores an unsigned 16 bit immediate from the last operand.
--   "i"       stores an operand sized immediate from the last operand.
--   "I"       dito, but generates an action code to optionally modify
--             the opcode (+2) for a signed 8 bit immediate.
--   "J"       generates one of the REL action codes from the last operand.
--
------------------------------------------------------------------------------

-- Template strings for x86 instructions. Ordered by first opcode byte.
-- Unimplemented opcodes (deliberate omissions) are marked with *.
local map_op = {
  -- 00-05: add...
  -- 06: *push es
  -- 07: *pop es
  -- 08-0D: or...
  -- 0E: *push cs
  -- 0F: two byte opcode prefix
  -- 10-15: adc...
  -- 16: *push ss
  -- 17: *pop ss
  -- 18-1D: sbb...
  -- 1E: *push ds
  -- 1F: *pop ds
  -- 20-25: and...
  es_0 =	"26",
  -- 27: *daa
  -- 28-2D: sub...
  cs_0 =	"2E",
  -- 2F: *das
  -- 30-35: xor...
  ss_0 =	"36",
  -- 37: *aaa
  -- 38-3D: cmp...
  ds_0 =	"3E",
  -- 3F: *aas
  inc_1 =	"rdw:40r|m:FF0m",
  dec_1 =	"rdw:48r|m:FF1m",
  push_1 =	"rdw:50r|mdw:FF6m|S.:6AS|ib:n6Ai|i.:68i",
  pop_1 =	"rdw:58r|mdw:8F0m",
  -- 60: *pusha, *pushad, *pushaw
  -- 61: *popa, *popad, *popaw
  -- 62: *bound rdw,x
  -- 63: *arpl mw,rw
  fs_0 =	"64",
  gs_0 =	"65",
  o16_0 =	"66",
  a16_0 =	"67",
  -- 68: push idw
  -- 69: imul rdw,mdw,idw
  -- 6A: push ib
  -- 6B: imul rdw,mdw,S
  -- 6C: *insb
  -- 6D: *insd, *insw
  -- 6E: *outsb
  -- 6F: *outsd, *outsw
  -- 70-7F: jcc lb
  -- 80: add... mb,i
  -- 81: add... mdw,i
  -- 82: *undefined
  -- 83: add... mdw,S
  test_2 =	"mr:85Rm|rm:85rM|Ri:A9i|mi:F70mi",
  -- 86: xchg rb,mb
  -- 87: xchg rdw,mdw
  -- 88: mov mb,r
  -- 89: mov mdw,r
  -- 8A: mov r,mb
  -- 8B: mov r,mdw
  -- 8C: *mov mdw,seg
  lea_2 =	"rxd:8DrM",
  -- 8E: *mov seg,mdw
  -- 8F: pop mdw
  nop_0 =	"90",
  xchg_2 =	"Rrdw:90R|rRdw:90r|rm:87rM|mr:87Rm",
  cbw_0 =	"6698",
  cwde_0 =	"98",
  cwd_0 =	"6699",
  cdq_0 =	"99",
  -- 9A: *call iw:idw
  wait_0 =	"9B",
  fwait_0 =	"9B",