1---------------------------------------------------------------------------- 2-- LuaJIT x86/x64 disassembler module. 3-- 4-- Copyright (C) 2005-2017 Mike Pall. All rights reserved. 5-- Released under the MIT license. See Copyright Notice in luajit.h 6---------------------------------------------------------------------------- 7-- This is a helper module used by the LuaJIT machine code dumper module. 8-- 9-- Sending small code snippets to an external disassembler and mixing the 10-- output with our own stuff was too fragile. So I had to bite the bullet 11-- and write yet another x86 disassembler. Oh well ... 12-- 13-- The output format is very similar to what ndisasm generates. But it has 14-- been developed independently by looking at the opcode tables from the 15-- Intel and AMD manuals. The supported instruction set is quite extensive 16-- and reflects what a current generation Intel or AMD CPU implements in 17-- 32 bit and 64 bit mode. Yes, this includes MMX, SSE, SSE2, SSE3, SSSE3, 18-- SSE4.1, SSE4.2, SSE4a, AVX, AVX2 and even privileged and hypervisor 19-- (VMX/SVM) instructions. 20-- 21-- Notes: 22-- * The (useless) a16 prefix, 3DNow and pre-586 opcodes are unsupported. 23-- * No attempt at optimization has been made -- it's fast enough for my needs. 24------------------------------------------------------------------------------ 25 26local type = type 27local sub, byte, format = string.sub, string.byte, string.format 28local match, gmatch, gsub = string.match, string.gmatch, string.gsub 29local lower, rep = string.lower, string.rep 30local bit = require("bit") 31local tohex = bit.tohex 32 33-- Map for 1st opcode byte in 32 bit mode. Ugly? Well ... read on. 34local map_opc1_32 = { 35--0x 36[0]="addBmr","addVmr","addBrm","addVrm","addBai","addVai","push es","pop es", 37"orBmr","orVmr","orBrm","orVrm","orBai","orVai","push cs","opc2*", 38--1x 39"adcBmr","adcVmr","adcBrm","adcVrm","adcBai","adcVai","push ss","pop ss", 40"sbbBmr","sbbVmr","sbbBrm","sbbVrm","sbbBai","sbbVai","push ds","pop ds", 41--2x 42"andBmr","andVmr","andBrm","andVrm","andBai","andVai","es:seg","daa", 43"subBmr","subVmr","subBrm","subVrm","subBai","subVai","cs:seg","das", 44--3x 45"xorBmr","xorVmr","xorBrm","xorVrm","xorBai","xorVai","ss:seg","aaa", 46"cmpBmr","cmpVmr","cmpBrm","cmpVrm","cmpBai","cmpVai","ds:seg","aas", 47--4x 48"incVR","incVR","incVR","incVR","incVR","incVR","incVR","incVR", 49"decVR","decVR","decVR","decVR","decVR","decVR","decVR","decVR", 50--5x 51"pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR","pushUR", 52"popUR","popUR","popUR","popUR","popUR","popUR","popUR","popUR", 53--6x 54"sz*pushaw,pusha","sz*popaw,popa","boundVrm","arplWmr", 55"fs:seg","gs:seg","o16:","a16", 56"pushUi","imulVrmi","pushBs","imulVrms", 57"insb","insVS","outsb","outsVS", 58--7x 59"joBj","jnoBj","jbBj","jnbBj","jzBj","jnzBj","jbeBj","jaBj", 60"jsBj","jnsBj","jpeBj","jpoBj","jlBj","jgeBj","jleBj","jgBj", 61--8x 62"arith!Bmi","arith!Vmi","arith!Bmi","arith!Vms", 63"testBmr","testVmr","xchgBrm","xchgVrm", 64"movBmr","movVmr","movBrm","movVrm", 65"movVmg","leaVrm","movWgm","popUm", 66--9x 67"nop*xchgVaR|pause|xchgWaR|repne nop","xchgVaR","xchgVaR","xchgVaR", 68"xchgVaR","xchgVaR","xchgVaR","xchgVaR", 69"sz*cbw,cwde,cdqe","sz*cwd,cdq,cqo","call farViw","wait", 70"sz*pushfw,pushf","sz*popfw,popf","sahf","lahf", 71--Ax 72"movBao","movVao","movBoa","movVoa", 73"movsb","movsVS","cmpsb","cmpsVS", 74"testBai","testVai","stosb","stosVS", 75"lodsb","lodsVS","scasb","scasVS", 76--Bx 77"movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi","movBRi", 78"movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI","movVRI", 79--Cx 80"shift!Bmu","shift!Vmu","retBw","ret","vex*3$lesVrm","vex*2$ldsVrm","movBmi","movVmi", 81"enterBwu","leave","retfBw","retf","int3","intBu","into","iretVS", 82--Dx 83"shift!Bm1","shift!Vm1","shift!Bmc","shift!Vmc","aamBu","aadBu","salc","xlatb", 84"fp*0","fp*1","fp*2","fp*3","fp*4","fp*5","fp*6","fp*7", 85--Ex 86"loopneBj","loopeBj","loopBj","sz*jcxzBj,jecxzBj,jrcxzBj", 87"inBau","inVau","outBua","outVua", 88"callVj","jmpVj","jmp farViw","jmpBj","inBad","inVad","outBda","outVda", 89--Fx 90"lock:","int1","repne:rep","rep:","hlt","cmc","testb!Bm","testv!Vm", 91"clc","stc","cli","sti","cld","std","incb!Bm","incd!Vm", 92} 93assert(#map_opc1_32 == 255) 94 95-- Map for 1st opcode byte in 64 bit mode (overrides only). 96local map_opc1_64 = setmetatable({ 97 [0x06]=false, [0x07]=false, [0x0e]=false, 98 [0x16]=false, [0x17]=false, [0x1e]=false, [0x1f]=false, 99 [0x27]=false, [0x2f]=false, [0x37]=false, [0x3f]=false, 100 [0x60]=false, [0x61]=false, [0x62]=false, [0x63]="movsxdVrDmt", [0x67]="a32:", 101 [0x40]="rex*", [0x41]="rex*b", [0x42]="rex*x", [0x43]="rex*xb", 102 [0x44]="rex*r", [0x45]="rex*rb", [0x46]="rex*rx", [0x47]="rex*rxb", 103 [0x48]="rex*w", [0x49]="rex*wb", [0x4a]="rex*wx", [0x4b]="rex*wxb", 104 [0x4c]="rex*wr", [0x4d]="rex*wrb", [0x4e]="rex*wrx", [0x4f]="rex*wrxb", 105 [0x82]=false, [0x9a]=false, [0xc4]="vex*3", [0xc5]="vex*2", [0xce]=false, 106 [0xd4]=false, [0xd5]=false, [0xd6]=false, [0xea]=false, 107}, { __index = map_opc1_32 }) 108 109-- Map for 2nd opcode byte (0F xx). True CISC hell. Hey, I told you. 110-- Prefix dependent MMX/SSE opcodes: (none)|rep|o16|repne, -|F3|66|F2 111local map_opc2 = { 112--0x 113[0]="sldt!Dmp","sgdt!Ump","larVrm","lslVrm",nil,"syscall","clts","sysret", 114"invd","wbinvd",nil,"ud1",nil,"$prefetch!Bm","femms","3dnowMrmu", 115--1x 116"movupsXrm|movssXrvm|movupdXrm|movsdXrvm", 117"movupsXmr|movssXmvr|movupdXmr|movsdXmvr", 118"movhlpsXrm$movlpsXrm|movsldupXrm|movlpdXrm|movddupXrm", 119"movlpsXmr||movlpdXmr", 120"unpcklpsXrvm||unpcklpdXrvm", 121"unpckhpsXrvm||unpckhpdXrvm", 122"movlhpsXrm$movhpsXrm|movshdupXrm|movhpdXrm", 123"movhpsXmr||movhpdXmr", 124"$prefetcht!Bm","hintnopVm","hintnopVm","hintnopVm", 125"hintnopVm","hintnopVm","hintnopVm","hintnopVm", 126--2x 127"movUmx$","movUmy$","movUxm$","movUym$","movUmz$",nil,"movUzm$",nil, 128"movapsXrm||movapdXrm", 129"movapsXmr||movapdXmr", 130"cvtpi2psXrMm|cvtsi2ssXrvVmt|cvtpi2pdXrMm|cvtsi2sdXrvVmt", 131"movntpsXmr|movntssXmr|movntpdXmr|movntsdXmr", 132"cvttps2piMrXm|cvttss2siVrXm|cvttpd2piMrXm|cvttsd2siVrXm", 133"cvtps2piMrXm|cvtss2siVrXm|cvtpd2piMrXm|cvtsd2siVrXm", 134"ucomissXrm||ucomisdXrm", 135"comissXrm||comisdXrm", 136--3x 137"wrmsr","rdtsc","rdmsr","rdpmc","sysenter","sysexit",nil,"getsec", 138"opc3*38",nil,"opc3*3a",nil,nil,nil,nil,nil, 139--4x 140"cmovoVrm","cmovnoVrm","cmovbVrm","cmovnbVrm", 141"cmovzVrm","cmovnzVrm","cmovbeVrm","cmovaVrm", 142"cmovsVrm","cmovnsVrm","cmovpeVrm","cmovpoVrm", 143"cmovlVrm","cmovgeVrm","cmovleVrm","cmovgVrm", 144--5x 145"movmskpsVrXm$||movmskpdVrXm$","sqrtpsXrm|sqrtssXrm|sqrtpdXrm|sqrtsdXrm", 146"rsqrtpsXrm|rsqrtssXrvm","rcppsXrm|rcpssXrvm", 147"andpsXrvm||andpdXrvm","andnpsXrvm||andnpdXrvm", 148"orpsXrvm||orpdXrvm","xorpsXrvm||xorpdXrvm", 149"addpsXrvm|addssXrvm|addpdXrvm|addsdXrvm","mulpsXrvm|mulssXrvm|mulpdXrvm|mulsdXrvm", 150"cvtps2pdXrm|cvtss2sdXrvm|cvtpd2psXrm|cvtsd2ssXrvm", 151"cvtdq2psXrm|cvttps2dqXrm|cvtps2dqXrm", 152"subpsXrvm|subssXrvm|subpdXrvm|subsdXrvm","minpsXrvm|minssXrvm|minpdXrvm|minsdXrvm", 153"divpsXrvm|divssXrvm|divpdXrvm|divsdXrvm","maxpsXrvm|maxssXrvm|maxpdXrvm|maxsdXrvm", 154--6x 155"punpcklbwPrvm","punpcklwdPrvm","punpckldqPrvm","packsswbPrvm", 156"pcmpgtbPrvm","pcmpgtwPrvm","pcmpgtdPrvm","packuswbPrvm", 157"punpckhbwPrvm","punpckhwdPrvm","punpckhdqPrvm","packssdwPrvm", 158"||punpcklqdqXrvm","||punpckhqdqXrvm", 159"movPrVSm","movqMrm|movdquXrm|movdqaXrm", 160--7x 161"pshufwMrmu|pshufhwXrmu|pshufdXrmu|pshuflwXrmu","pshiftw!Pvmu", 162"pshiftd!Pvmu","pshiftq!Mvmu||pshiftdq!Xvmu", 163"pcmpeqbPrvm","pcmpeqwPrvm","pcmpeqdPrvm","emms*|", 164"vmreadUmr||extrqXmuu$|insertqXrmuu$","vmwriteUrm||extrqXrm$|insertqXrm$", 165nil,nil, 166"||haddpdXrvm|haddpsXrvm","||hsubpdXrvm|hsubpsXrvm", 167"movVSmMr|movqXrm|movVSmXr","movqMmr|movdquXmr|movdqaXmr", 168--8x 169"joVj","jnoVj","jbVj","jnbVj","jzVj","jnzVj","jbeVj","jaVj", 170"jsVj","jnsVj","jpeVj","jpoVj","jlVj","jgeVj","jleVj","jgVj", 171--9x 172"setoBm","setnoBm","setbBm","setnbBm","setzBm","setnzBm","setbeBm","setaBm", 173"setsBm","setnsBm","setpeBm","setpoBm","setlBm","setgeBm","setleBm","setgBm", 174--Ax 175"push fs","pop fs","cpuid","btVmr","shldVmru","shldVmrc",nil,nil, 176"push gs","pop gs","rsm","btsVmr","shrdVmru","shrdVmrc","fxsave!Dmp","imulVrm", 177--Bx 178"cmpxchgBmr","cmpxchgVmr","$lssVrm","btrVmr", 179"$lfsVrm","$lgsVrm","movzxVrBmt","movzxVrWmt", 180"|popcntVrm","ud2Dp","bt!Vmu","btcVmr", 181"bsfVrm","bsrVrm|lzcntVrm|bsrWrm","movsxVrBmt","movsxVrWmt", 182--Cx 183"xaddBmr","xaddVmr", 184"cmppsXrvmu|cmpssXrvmu|cmppdXrvmu|cmpsdXrvmu","$movntiVmr|", 185"pinsrwPrvWmu","pextrwDrPmu", 186"shufpsXrvmu||shufpdXrvmu","$cmpxchg!Qmp", 187"bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR","bswapVR", 188--Dx 189"||addsubpdXrvm|addsubpsXrvm","psrlwPrvm","psrldPrvm","psrlqPrvm", 190"paddqPrvm","pmullwPrvm", 191"|movq2dqXrMm|movqXmr|movdq2qMrXm$","pmovmskbVrMm||pmovmskbVrXm", 192"psubusbPrvm","psubuswPrvm","pminubPrvm","pandPrvm", 193"paddusbPrvm","padduswPrvm","pmaxubPrvm","pandnPrvm", 194--Ex 195"pavgbPrvm","psrawPrvm","psradPrvm","pavgwPrvm", 196"pmulhuwPrvm","pmulhwPrvm", 197"|cvtdq2pdXrm|cvttpd2dqXrm|cvtpd2dqXrm","$movntqMmr||$movntdqXmr", 198"psubsbPrvm","psubswPrvm","pminswPrvm","porPrvm", 199"paddsbPrvm","paddswPrvm","pmaxswPrvm","pxorPrvm", 200--Fx 201"|||lddquXrm","psllwPrvm","pslldPrvm","psllqPrvm", 202"pmuludqPrvm","pmaddwdPrvm","psadbwPrvm","maskmovqMrm||maskmovdquXrm$", 203"psubbPrvm","psubwPrvm","psubdPrvm","psubqPrvm", 204"paddbPrvm","paddwPrvm","padddPrvm","ud", 205} 206assert(map_opc2[255] == "ud") 207 208-- Map for three-byte opcodes. Can't wait for their next invention. 209local map_opc3 = { 210["38"] = { -- [66] 0f 38 xx 211--0x 212[0]="pshufbPrvm","phaddwPrvm","phadddPrvm","phaddswPrvm", 213"pmaddubswPrvm","phsubwPrvm","phsubdPrvm","phsubswPrvm", 214"psignbPrvm","psignwPrvm","psigndPrvm","pmulhrswPrvm", 215"||permilpsXrvm","||permilpdXrvm",nil,nil, 216--1x 217"||pblendvbXrma",nil,nil,nil, 218"||blendvpsXrma","||blendvpdXrma","||permpsXrvm","||ptestXrm", 219"||broadcastssXrm","||broadcastsdXrm","||broadcastf128XrlXm",nil, 220"pabsbPrm","pabswPrm","pabsdPrm",nil, 221--2x 222"||pmovsxbwXrm","||pmovsxbdXrm","||pmovsxbqXrm","||pmovsxwdXrm", 223"||pmovsxwqXrm","||pmovsxdqXrm",nil,nil, 224"||pmuldqXrvm","||pcmpeqqXrvm","||$movntdqaXrm","||packusdwXrvm", 225"||maskmovpsXrvm","||maskmovpdXrvm","||maskmovpsXmvr","||maskmovpdXmvr", 226--3x 227"||pmovzxbwXrm","||pmovzxbdXrm","||pmovzxbqXrm","||pmovzxwdXrm", 228"||pmovzxwqXrm","||pmovzxdqXrm","||permdXrvm","||pcmpgtqXrvm", 229"||pminsbXrvm","||pminsdXrvm","||pminuwXrvm","||pminudXrvm", 230"||pmaxsbXrvm","||pmaxsdXrvm","||pmaxuwXrvm","||pmaxudXrvm", 231--4x 232"||pmulddXrvm","||phminposuwXrm",nil,nil, 233nil,"||psrlvVSXrvm","||psravdXrvm","||psllvVSXrvm", 234--5x 235[0x58] = "||pbroadcastdXrlXm",[0x59] = "||pbroadcastqXrlXm", 236[0x5a] = "||broadcasti128XrlXm", 237--7x 238[0x78] = "||pbroadcastbXrlXm",[0x79] = "||pbroadcastwXrlXm", 239--8x 240[0x8c] = "||pmaskmovXrvVSm", 241[0x8e] = "||pmaskmovVSmXvr", 242--Dx 243[0xdc] = "||aesencXrvm", [0xdd] = "||aesenclastXrvm", 244[0xde] = "||aesdecXrvm", [0xdf] = "||aesdeclastXrvm", 245--Fx 246[0xf0] = "|||crc32TrBmt",[0xf1] = "|||crc32TrVmt", 247[0xf7] = "| sarxVrmv| shlxVrmv| shrxVrmv", 248}, 249 250["3a"] = { -- [66] 0f 3a xx 251--0x 252[0x00]="||permqXrmu","||permpdXrmu","||pblenddXrvmu",nil, 253"||permilpsXrmu","||permilpdXrmu","||perm2f128Xrvmu",nil, 254"||roundpsXrmu","||roundpdXrmu","||roundssXrvmu","||roundsdXrvmu", 255"||blendpsXrvmu","||blendpdXrvmu","||pblendwXrvmu","palignrPrvmu", 256--1x 257nil,nil,nil,nil, 258"||pextrbVmXru","||pextrwVmXru","||pextrVmSXru","||extractpsVmXru", 259"||insertf128XrvlXmu","||extractf128XlXmYru",nil,nil, 260nil,nil,nil,nil, 261--2x 262"||pinsrbXrvVmu","||insertpsXrvmu","||pinsrXrvVmuS",nil, 263--3x 264[0x38] = "||inserti128Xrvmu",[0x39] = "||extracti128XlXmYru", 265--4x 266[0x40] = "||dppsXrvmu", 267[0x41] = "||dppdXrvmu", 268[0x42] = "||mpsadbwXrvmu", 269[0x44] = "||pclmulqdqXrvmu", 270[0x46] = "||perm2i128Xrvmu", 271[0x4a] = "||blendvpsXrvmb",[0x4b] = "||blendvpdXrvmb", 272[0x4c] = "||pblendvbXrvmb", 273--6x 274[0x60] = "||pcmpestrmXrmu",[0x61] = "||pcmpestriXrmu", 275[0x62] = "||pcmpistrmXrmu",[0x63] = "||pcmpistriXrmu", 276[0xdf] = "||aeskeygenassistXrmu", 277--Fx 278[0xf0] = "||| rorxVrmu", 279}, 280} 281 282-- Map for VMX/SVM opcodes 0F 01 C0-FF (sgdt group with register operands). 283local map_opcvm = { 284[0xc1]="vmcall",[0xc2]="vmlaunch",[0xc3]="vmresume",[0xc4]="vmxoff", 285[0xc8]="monitor",[0xc9]="mwait", 286[0xd8]="vmrun",[0xd9]="vmmcall",[0xda]="vmload",[0xdb]="vmsave", 287[0xdc]="stgi",[0xdd]="clgi",[0xde]="skinit",[0xdf]="invlpga", 288[0xf8]="swapgs",[0xf9]="rdtscp", 289} 290 291-- Map for FP opcodes. And you thought stack machines are simple? 292local map_opcfp = { 293-- D8-DF 00-BF: opcodes with a memory operand. 294-- D8 295[0]="faddFm","fmulFm","fcomFm","fcompFm","fsubFm","fsubrFm","fdivFm","fdivrFm", 296"fldFm",nil,"fstFm","fstpFm","fldenvVm","fldcwWm","fnstenvVm","fnstcwWm", 297-- DA 298"fiaddDm","fimulDm","ficomDm","ficompDm", 299"fisubDm","fisubrDm","fidivDm","fidivrDm", 300-- DB 301"fildDm","fisttpDm","fistDm","fistpDm",nil,"fld twordFmp",nil,"fstp twordFmp", 302-- DC 303"faddGm","fmulGm","fcomGm","fcompGm","fsubGm","fsubrGm","fdivGm","fdivrGm", 304-- DD 305"fldGm","fisttpQm","fstGm","fstpGm","frstorDmp",nil,"fnsaveDmp","fnstswWm", 306-- DE 307"fiaddWm","fimulWm","ficomWm","ficompWm", 308"fisubWm","fisubrWm","fidivWm","fidivrWm", 309-- DF 310"fildWm","fisttpWm","fistWm","fistpWm", 311"fbld twordFmp","fildQm","fbstp twordFmp","fistpQm", 312-- xx C0-FF: opcodes with a pseudo-register operand. 313-- D8 314"faddFf","fmulFf","fcomFf","fcompFf","fsubFf","fsubrFf","fdivFf","fdivrFf", 315-- D9 316"fldFf","fxchFf",{"fnop"},nil, 317{"fchs","fabs",nil,nil,"ftst","fxam"}, 318{"fld1","fldl2t","fldl2e","fldpi","fldlg2","fldln2","fldz"}, 319{"f2xm1","fyl2x","fptan","fpatan","fxtract","fprem1","fdecstp","fincstp"}, 320{"fprem","fyl2xp1","fsqrt","fsincos","frndint","fscale","fsin","fcos"}, 321-- DA 322"fcmovbFf","fcmoveFf","fcmovbeFf","fcmovuFf",nil,{nil,"fucompp"},nil,nil, 323-- DB 324"fcmovnbFf","fcmovneFf","fcmovnbeFf","fcmovnuFf", 325{nil,nil,"fnclex","fninit"},"fucomiFf","fcomiFf",nil, 326-- DC 327"fadd toFf","fmul toFf",nil,nil, 328"fsub toFf","fsubr toFf","fdivr toFf","fdiv toFf", 329-- DD 330"ffreeFf",nil,"fstFf","fstpFf","fucomFf","fucompFf",nil,nil, 331-- DE 332"faddpFf","fmulpFf",nil,{nil,"fcompp"}, 333"fsubrpFf","fsubpFf","fdivrpFf","fdivpFf", 334-- DF 335nil,nil,nil,nil,{"fnstsw ax"},"fucomipFf","fcomipFf",nil, 336} 337assert(map_opcfp[126] == "fcomipFf") 338 339-- Map for opcode groups. The subkey is sp from the ModRM byte. 340local map_opcgroup = { 341 arith = { "add", "or", "adc", "sbb", "and", "sub", "xor", "cmp" }, 342 shift = { "rol", "ror", "rcl", "rcr", "shl", "shr", "sal", "sar" }, 343 testb = { "testBmi", "testBmi", "not", "neg", "mul", "imul", "div", "idiv" }, 344 testv = { "testVmi", "testVmi", "not", "neg", "mul", "imul", "div", "idiv" }, 345 incb = { "inc", "dec" }, 346 incd = { "inc", "dec", "callUmp", "$call farDmp", 347 "jmpUmp", "$jmp farDmp", "pushUm" }, 348 sldt = { "sldt", "str", "lldt", "ltr", "verr", "verw" }, 349 sgdt = { "vm*$sgdt", "vm*$sidt", "$lgdt", "vm*$lidt", 350 "smsw", nil, "lmsw", "vm*$invlpg" }, 351 bt = { nil, nil, nil, nil, "bt", "bts", "btr", "btc" }, 352 cmpxchg = { nil, "sz*,cmpxchg8bQmp,cmpxchg16bXmp", nil, nil, 353 nil, nil, "vmptrld|vmxon|vmclear", "vmptrst" }, 354 pshiftw = { nil, nil, "psrlw", nil, "psraw", nil, "psllw" }, 355 pshiftd = { nil, nil, "psrld", nil, "psrad", nil, "pslld" }, 356 pshiftq = { nil, nil, "psrlq", nil, nil, nil, "psllq" }, 357 pshiftdq = { nil, nil, "psrlq", "psrldq", nil, nil, "psllq", "pslldq" }, 358 fxsave = { "$fxsave", "$fxrstor", "$ldmxcsr", "$stmxcsr", 359 nil, "lfenceDp$", "mfenceDp$", "sfenceDp$clflush" }, 360 prefetch = { "prefetch", "prefetchw" }, 361 prefetcht = { "prefetchnta", "prefetcht0", "prefetcht1", "prefetcht2" }, 362} 363 364------------------------------------------------------------------------------ 365 366-- Maps for register names. 367local map_regs = { 368 B = { "al", "cl", "dl", "bl", "ah", "ch", "dh", "bh", 369 "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, 370 B64 = { "al", "cl", "dl", "bl", "spl", "bpl", "sil", "dil", 371 "r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b" }, 372 W = { "ax", "cx", "dx", "bx", "sp", "bp", "si", "di", 373 "r8w", "r9w", "r10w", "r11w", "r12w", "r13w", "r14w", "r15w" }, 374 D = { "eax", "ecx", "edx", "ebx", "esp", "ebp", "esi", "edi", 375 "r8d", "r9d", "r10d", "r11d", "r12d", "r13d", "r14d", "r15d" }, 376 Q = { "rax", "rcx", "rdx", "rbx", "rsp", "rbp", "rsi", "rdi", 377 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15" }, 378 M = { "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7", 379 "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7" }, -- No x64 ext! 380 X = { "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7", 381 "xmm8", "xmm9", "xmm10", "xmm11", "xmm12", "xmm13", "xmm14", "xmm15" }, 382 Y = { "ymm0", "ymm1", "ymm2", "ymm3", "ymm4", "ymm5", "ymm6", "ymm7", 383 "ymm8", "ymm9", "ymm10", "ymm11", "ymm12", "ymm13", "ymm14", "ymm15" }, 384} 385local map_segregs = { "es", "cs", "ss", "ds", "fs", "gs", "segr6", "segr7" } 386 387-- Maps for size names. 388local map_sz2n = { 389 B = 1, W = 2, D = 4, Q = 8, M = 8, X = 16, Y = 32, 390} 391local map_sz2prefix = { 392 B = "byte", W = "word", D = "dword", 393 Q = "qword", 394 M = "qword", X = "xword", Y = "yword", 395 F = "dword", G = "qword", -- No need for sizes/register names for these two. 396} 397 398------------------------------------------------------------------------------ 399 400-- Output a nicely formatted line with an opcode and operands. 401local function putop(ctx, text, operands) 402 local code, pos, hex = ctx.code, ctx.pos, "" 403 local hmax = ctx.hexdump 404 if hmax > 0 then 405 for i=ctx.start,pos-1 do 406 hex = hex..format("%02X", byte(code, i, i)) 407 end 408 if #hex > hmax then hex = sub(hex, 1, hmax)..". " 409 else hex = hex..rep(" ", hmax-#hex+2) end 410 end 411 if operands then text = text.." "..operands end 412 if ctx.o16 then text = "o16 "..text; ctx.o16 = false end 413 if ctx.a32 then text = "a32 "..text; ctx.a32 = false end 414 if ctx.rep then text = ctx.rep.." "..text; ctx.rep = false end 415 if ctx.rex then 416 local t = (ctx.rexw and "w" or "")..(ctx.rexr and "r" or "").. 417 (ctx.rexx and "x" or "")..(ctx.rexb and "b" or "").. 418 (ctx.vexl and "l" or "") 419 if ctx.vexv and ctx.vexv ~= 0 then t = t.."v"..ctx.vexv end 420 if t ~= "" then text = ctx.rex.."."..t.." "..gsub(text, "^ ", "") 421 elseif ctx.rex == "vex" then text = gsub("v"..text, "^v ", "") end 422 ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false 423 ctx.rex = false; ctx.vexl = false; ctx.vexv = false 424 end 425 if ctx.seg then 426 local text2, n = gsub(text, "%[", "["..ctx.seg..":") 427 if n == 0 then text = ctx.seg.." "..text else text = text2 end 428 ctx.seg = false 429 end 430 if ctx.lock then text = "lock "..text; ctx.lock = false end 431 local imm = ctx.imm 432 if imm then 433 local sym = ctx.symtab[imm] 434 if sym then text = text.."\t->"..sym end 435 end 436 ctx.out(format("%08x %s%s\n", ctx.addr+ctx.start, hex, text)) 437 ctx.mrm = false 438 ctx.vexv = false 439 ctx.start = pos 440 ctx.imm = nil 441end 442 443-- Clear all prefix flags. 444local function clearprefixes(ctx) 445 ctx.o16 = false; ctx.seg = false; ctx.lock = false; ctx.rep = false 446 ctx.rexw = false; ctx.rexr = false; ctx.rexx = false; ctx.rexb = false 447 ctx.rex = false; ctx.a32 = false; ctx.vexl = false 448end 449 450-- Fallback for incomplete opcodes at the end. 451local function incomplete(ctx) 452 ctx.pos = ctx.stop+1 453 clearprefixes(ctx) 454 return putop(ctx, "(incomplete)") 455end 456 457-- Fallback for unknown opcodes. 458local function unknown(ctx) 459 clearprefixes(ctx) 460 return putop(ctx, "(unknown)") 461end 462 463-- Return an immediate of the specified size. 464local function getimm(ctx, pos, n) 465 if pos+n-1 > ctx.stop then return incomplete(ctx) end 466 local code = ctx.code 467 if n == 1 then 468 local b1 = byte(code, pos, pos) 469 return b1 470 elseif n == 2 then 471 local b1, b2 = byte(code, pos, pos+1) 472 return b1+b2*256 473 else 474 local b1, b2, b3, b4 = byte(code, pos, pos+3) 475 local imm = b1+b2*256+b3*65536+b4*16777216 476 ctx.imm = imm 477 return imm 478 end 479end 480 481-- Process pattern string and generate the operands. 482local function putpat(ctx, name, pat) 483 local operands, regs, sz, mode, sp, rm, sc, rx, sdisp 484 local code, pos, stop, vexl = ctx.code, ctx.pos, ctx.stop, ctx.vexl 485 486 -- Chars used: 1DFGIMPQRSTUVWXYabcdfgijlmoprstuvwxyz 487 for p in gmatch(pat, ".") do 488 local x = nil 489 if p == "V" or p == "U" then 490 if ctx.rexw then sz = "Q"; ctx.rexw = false 491 elseif ctx.o16 then sz = "W"; ctx.o16 = false 492 elseif p == "U" and ctx.x64 then sz = "Q" 493 else sz = "D" end 494 regs = map_regs[sz] 495 elseif p == "T" then 496 if ctx.rexw then sz = "Q"; ctx.rexw = false else sz = "D" end 497 regs = map_regs[sz] 498 elseif p == "B" then 499 sz = "B" 500 regs = ctx.rex and map_regs.B64 or map_regs.B 501 elseif match(p, "[WDQMXYFG]") then 502 sz = p 503 if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end 504 regs = map_regs[sz] 505 elseif p == "P" then 506 sz = ctx.o16 and "X" or "M"; ctx.o16 = false 507 if sz == "X" and vexl then sz = "Y"; ctx.vexl = false end 508 regs = map_regs[sz] 509 elseif p == "S" then 510 name = name..lower(sz) 511 elseif p == "s" then 512 local imm = getimm(ctx, pos, 1); if not imm then return end 513 x = imm <= 127 and format("+0x%02x", imm) 514 or format("-0x%02x", 256-imm) 515 pos = pos+1 516 elseif p == "u" then 517 local imm = getimm(ctx, pos, 1); if not imm then return end 518 x = format("0x%02x", imm) 519 pos = pos+1 520 elseif p == "b" then 521 local imm = getimm(ctx, pos, 1); if not imm then return end 522 x = regs[imm/16+1] 523 pos = pos+1 524 elseif p == "w" then 525 local imm = getimm(ctx, pos, 2); if not imm then return end 526 x = format("0x%x", imm) 527 pos = pos+2 528 elseif p == "o" then -- [offset] 529 if ctx.x64 then 530 local imm1 = getimm(ctx, pos, 4); if not imm1 then return end 531 local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end 532 x = format("[0x%08x%08x]", imm2, imm1) 533 pos = pos+8 534 else 535 local imm = getimm(ctx, pos, 4); if not imm then return end 536 x = format("[0x%08x]", imm) 537 pos = pos+4 538 end 539 elseif p == "i" or p == "I" then 540 local n = map_sz2n[sz] 541 if n == 8 and ctx.x64 and p == "I" then 542 local imm1 = getimm(ctx, pos, 4); if not imm1 then return end 543 local imm2 = getimm(ctx, pos+4, 4); if not imm2 then return end 544 x = format("0x%08x%08x", imm2, imm1) 545 else 546 if n == 8 then n = 4 end 547 local imm = getimm(ctx, pos, n); if not imm then return end 548 if sz == "Q" and (imm < 0 or imm > 0x7fffffff) then 549 imm = (0xffffffff+1)-imm 550 x = format(imm > 65535 and "-0x%08x" or "-0x%x", imm) 551 else 552 x = format(imm > 65535 and "0x%08x" or "0x%x", imm) 553 end 554 end 555 pos = pos+n 556 elseif p == "j" then 557 local n = map_sz2n[sz] 558 if n == 8 then n = 4 end 559 local imm = getimm(ctx, pos, n); if not imm then return end 560 if sz == "B" and imm > 127 then imm = imm-256 561 elseif imm > 2147483647 then imm = imm-4294967296 end 562 pos = pos+n 563 imm = imm + pos + ctx.addr 564 if imm > 4294967295 and not ctx.x64 then imm = imm-4294967296 end 565 ctx.imm = imm 566 if sz == "W" then 567 x = format("word 0x%04x", imm%65536) 568 elseif ctx.x64 then 569 local lo = imm % 0x1000000 570 x = format("0x%02x%06x", (imm-lo) / 0x1000000, lo) 571 else 572 x = "0x"..tohex(imm) 573 end 574 elseif p == "R" then 575 local r = byte(code, pos-1, pos-1)%8 576 if ctx.rexb then r = r + 8; ctx.rexb = false end 577 x = regs[r+1] 578 elseif p == "a" then x = regs[1] 579 elseif p == "c" then x = "cl" 580 elseif p == "d" then x = "dx" 581 elseif p == "1" then x = "1" 582 else 583 if not mode then 584 mode = ctx.mrm 585 if not mode then 586 if pos > stop then return incomplete(ctx) end 587 mode = byte(code, pos, pos) 588 pos = pos+1 589 end 590 rm = mode%8; mode = (mode-rm)/8 591 sp = mode%8; mode = (mode-sp)/8 592 sdisp = "" 593 if mode < 3 then 594 if rm == 4 then 595 if pos > stop then return incomplete(ctx) end 596 sc = byte(code, pos, pos) 597 pos = pos+1 598 rm = sc%8; sc = (sc-rm)/8 599 rx = sc%8; sc = (sc-rx)/8 600 if ctx.rexx then rx = rx + 8; ctx.rexx = false end 601 if rx == 4 then rx = nil end 602 end 603 if mode > 0 or rm == 5 then 604 local dsz = mode 605 if dsz ~= 1 then dsz = 4 end 606 local disp = getimm(ctx, pos, dsz); if not disp then return end 607 if mode == 0 then rm = nil end 608 if rm or rx or (not sc and ctx.x64 and not ctx.a32) then 609 if dsz == 1 and disp > 127 then 610 sdisp = format("-0x%x", 256-disp) 611 elseif disp >= 0 and disp <= 0x7fffffff then 612 sdisp = format("+0x%x", disp) 613 else 614 sdisp = format("-0x%x", (0xffffffff+1)-disp) 615 end 616 else 617 sdisp = format(ctx.x64 and not ctx.a32 and 618 not (disp >= 0 and disp <= 0x7fffffff) 619 and "0xffffffff%08x" or "0x%08x", disp) 620 end 621 pos = pos+dsz 622 end 623 end 624 if rm and ctx.rexb then rm = rm + 8; ctx.rexb = false end 625 if ctx.rexr then sp = sp + 8; ctx.rexr = false end 626 end 627 if p == "m" then 628 if mode == 3 then x = regs[rm+1] 629 else 630 local aregs = ctx.a32 and map_regs.D or ctx.aregs 631 local srm, srx = "", "" 632 if rm then srm = aregs[rm+1] 633 elseif not sc and ctx.x64 and not ctx.a32 then srm = "rip" end 634 ctx.a32 = false 635 if rx then 636 if rm then srm = srm.."+" end 637 srx = aregs[rx+1] 638 if sc > 0 then srx = srx.."*"..(2^sc) end 639 end 640 x = format("[%s%s%s]", srm, srx, sdisp) 641 end 642 if mode < 3 and 643 (not match(pat, "[aRrgp]") or match(pat, "t")) then -- Yuck. 644 x = map_sz2prefix[sz].." "..x 645 end 646 elseif p == "r" then x = regs[sp+1] 647 elseif p == "g" then x = map_segregs[sp+1] 648 elseif p == "p" then -- Suppress prefix. 649 elseif p == "f" then x = "st"..rm 650 elseif p == "x" then 651 if sp == 0 and ctx.lock and not ctx.x64 then 652 x = "CR8"; ctx.lock = false 653 else 654 x = "CR"..sp 655 end 656 elseif p == "v" then 657 if ctx.vexv then 658 x = regs[ctx.vexv+1]; ctx.vexv = false 659 end 660 elseif p == "y" then x = "DR"..sp 661 elseif p == "z" then x = "TR"..sp 662 elseif p == "l" then vexl = false 663 elseif p == "t" then 664 else 665 error("bad pattern `"..pat.."'") 666 end 667 end 668 if x then operands = operands and operands..", "..x or x end 669 end 670 ctx.pos = pos 671 return putop(ctx, name, operands) 672end 673 674-- Forward declaration. 675local map_act 676 677-- Fetch and cache MRM byte. 678local function getmrm(ctx) 679 local mrm = ctx.mrm 680 if not mrm then 681 local pos = ctx.pos 682 if pos > ctx.stop then return nil end 683 mrm = byte(ctx.code, pos, pos) 684 ctx.pos = pos+1 685 ctx.mrm = mrm 686 end 687 return mrm 688end 689 690-- Dispatch to handler depending on pattern. 691local function dispatch(ctx, opat, patgrp) 692 if not opat then return unknown(ctx) end 693 if match(opat, "%|") then -- MMX/SSE variants depending on prefix. 694 local p 695 if ctx.rep then 696 p = ctx.rep=="rep" and "%|([^%|]*)" or "%|[^%|]*%|[^%|]*%|([^%|]*)" 697 ctx.rep = false 698 elseif ctx.o16 then p = "%|[^%|]*%|([^%|]*)"; ctx.o16 = false 699 else p = "^[^%|]*" end 700 opat = match(opat, p) 701 if not opat then return unknown(ctx) end 702-- ctx.rep = false; ctx.o16 = false 703 --XXX fails for 66 f2 0f 38 f1 06 crc32 eax,WORD PTR [esi] 704 --XXX remove in branches? 705 end 706 if match(opat, "%$") then -- reg$mem variants. 707 local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end 708 opat = match(opat, mrm >= 192 and "^[^%$]*" or "%$(.*)") 709 if opat == "" then return unknown(ctx) end 710 end 711 if opat == "" then return unknown(ctx) end 712 local name, pat = match(opat, "^([a-z0-9 ]*)(.*)") 713 if pat == "" and patgrp then pat = patgrp end 714 return map_act[sub(pat, 1, 1)](ctx, name, pat) 715end 716 717-- Get a pattern from an opcode map and dispatch to handler. 718local function dispatchmap(ctx, opcmap) 719 local pos = ctx.pos 720 local opat = opcmap[byte(ctx.code, pos, pos)] 721 pos = pos + 1 722 ctx.pos = pos 723 return dispatch(ctx, opat) 724end 725 726-- Map for action codes. The key is the first char after the name. 727map_act = { 728 -- Simple opcodes without operands. 729 [""] = function(ctx, name, pat) 730 return putop(ctx, name) 731 end, 732 733 -- Operand size chars fall right through. 734 B = putpat, W = putpat, D = putpat, Q = putpat, 735 V = putpat, U = putpat, T = putpat, 736 M = putpat, X = putpat, P = putpat, 737 F = putpat, G = putpat, Y = putpat, 738 739 -- Collect prefixes. 740 [":"] = function(ctx, name, pat) 741 ctx[pat == ":" and name or sub(pat, 2)] = name 742 if ctx.pos - ctx.start > 5 then return unknown(ctx) end -- Limit #prefixes. 743 end, 744 745 -- Chain to special handler specified by name. 746 ["*"] = function(ctx, name, pat) 747 return map_act[name](ctx, name, sub(pat, 2)) 748 end, 749 750 -- Use named subtable for opcode group. 751 ["!"] = function(ctx, name, pat) 752 local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end 753 return dispatch(ctx, map_opcgroup[name][((mrm-(mrm%8))/8)%8+1], sub(pat, 2)) 754 end, 755 756 -- o16,o32[,o64] variants. 757 sz = function(ctx, name, pat) 758 if ctx.o16 then ctx.o16 = false 759 else 760 pat = match(pat, ",(.*)") 761 if ctx.rexw then 762 local p = match(pat, ",(.*)") 763 if p then pat = p; ctx.rexw = false end 764 end 765 end 766 pat = match(pat, "^[^,]*") 767 return dispatch(ctx, pat) 768 end, 769 770 -- Two-byte opcode dispatch. 771 opc2 = function(ctx, name, pat) 772 return dispatchmap(ctx, map_opc2) 773 end, 774 775 -- Three-byte opcode dispatch. 776 opc3 = function(ctx, name, pat) 777 return dispatchmap(ctx, map_opc3[pat]) 778 end, 779 780 -- VMX/SVM dispatch. 781 vm = function(ctx, name, pat) 782 return dispatch(ctx, map_opcvm[ctx.mrm]) 783 end, 784 785 -- Floating point opcode dispatch. 786 fp = function(ctx, name, pat) 787 local mrm = getmrm(ctx); if not mrm then return incomplete(ctx) end 788 local rm = mrm%8 789 local idx = pat*8 + ((mrm-rm)/8)%8 790 if mrm >= 192 then idx = idx + 64 end 791 local opat = map_opcfp[idx] 792 if type(opat) == "table" then opat = opat[rm+1] end 793 return dispatch(ctx, opat) 794 end, 795 796 -- REX prefix. 797 rex = function(ctx, name, pat) 798 if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. 799 for p in gmatch(pat, ".") do ctx["rex"..p] = true end 800 ctx.rex = "rex" 801 end, 802 803 -- VEX prefix. 804 vex = function(ctx, name, pat) 805 if ctx.rex then return unknown(ctx) end -- Only 1 REX or VEX prefix allowed. 806 ctx.rex = "vex" 807 local pos = ctx.pos 808 if ctx.mrm then 809 ctx.mrm = nil 810 pos = pos-1 811 end 812 local b = byte(ctx.code, pos, pos) 813 if not b then return incomplete(ctx) end 814 pos = pos+1 815 if b < 128 then ctx.rexr = true end 816 local m = 1 817 if pat == "3" then 818 m = b%32; b = (b-m)/32 819 local nb = b%2; b = (b-nb)/2 820 if nb == 0 then ctx.rexb = true end 821 local nx = b%2 822 if nx == 0 then ctx.rexx = true end 823 b = byte(ctx.code, pos, pos) 824 if not b then return incomplete(ctx) end 825 pos = pos+1 826 if b >= 128 then ctx.rexw = true end 827 end 828 ctx.pos = pos 829 local map 830 if m == 1 then map = map_opc2 831 elseif m == 2 then map = map_opc3["38"] 832 elseif m == 3 then map = map_opc3["3a"] 833 else return unknown(ctx) end 834 local p = b%4; b = (b-p)/4 835 if p == 1 then ctx.o16 = "o16" 836 elseif p == 2 then ctx.rep = "rep" 837 elseif p == 3 then ctx.rep = "repne" end 838 local l = b%2; b = (b-l)/2 839 if l ~= 0 then ctx.vexl = true end 840 ctx.vexv = (-1-b)%16 841 return dispatchmap(ctx, map) 842 end, 843 844 -- Special case for nop with REX prefix. 845 nop = function(ctx, name, pat) 846 return dispatch(ctx, ctx.rex and pat or "nop") 847 end, 848 849 -- Special case for 0F 77. 850 emms = function(ctx, name, pat) 851 if ctx.rex ~= "vex" then 852 return putop(ctx, "emms") 853 elseif ctx.vexl then 854 ctx.vexl = false 855 return putop(ctx, "zeroall") 856 else 857 return putop(ctx, "zeroupper") 858 end 859 end, 860} 861 862------------------------------------------------------------------------------ 863 864-- Disassemble a block of code. 865local function disass_block(ctx, ofs, len) 866 if not ofs then ofs = 0 end 867 local stop = len and ofs+len or #ctx.code 868 ofs = ofs + 1 869 ctx.start = ofs 870 ctx.pos = ofs 871 ctx.stop = stop 872 ctx.imm = nil 873 ctx.mrm = false 874 clearprefixes(ctx) 875 while ctx.pos <= stop do dispatchmap(ctx, ctx.map1) end 876 if ctx.pos ~= ctx.start then incomplete(ctx) end 877end 878 879-- Extended API: create a disassembler context. Then call ctx:disass(ofs, len). 880local function create(code, addr, out) 881 local ctx = {} 882 ctx.code = code 883 ctx.addr = (addr or 0) - 1 884 ctx.out = out or io.write 885 ctx.symtab = {} 886 ctx.disass = disass_block 887 ctx.hexdump = 16 888 ctx.x64 = false 889 ctx.map1 = map_opc1_32 890 ctx.aregs = map_regs.D 891 return ctx 892end 893 894local function create64(code, addr, out) 895 local ctx = create(code, addr, out) 896 ctx.x64 = true 897 ctx.map1 = map_opc1_64 898 ctx.aregs = map_regs.Q 899 return ctx 900end 901 902-- Simple API: disassemble code (a string) at address and output via out. 903local function disass(code, addr, out) 904 create(code, addr, out):disass() 905end 906 907local function disass64(code, addr, out) 908 create64(code, addr, out):disass() 909end 910 911-- Return register name for RID. 912local function regname(r) 913 if r < 8 then return map_regs.D[r+1] end 914 return map_regs.X[r-7] 915end 916 917local function regname64(r) 918 if r < 16 then return map_regs.Q[r+1] end 919 return map_regs.X[r-15] 920end 921 922-- Public module functions. 923return { 924 create = create, 925 create64 = create64, 926 disass = disass, 927 disass64 = disass64, 928 regname = regname, 929 regname64 = regname64 930} 931 932