1 /*===- X86DisassemblerDecoderInternal.h - Disassembler decoder -----*- C -*-==* 2 * 3 * The LLVM Compiler Infrastructure 4 * 5 * This file is distributed under the University of Illinois Open Source 6 * License. See LICENSE.TXT for details. 7 * 8 *===----------------------------------------------------------------------===* 9 * 10 * This file is part of the X86 Disassembler. 11 * It contains the public interface of the instruction decoder. 12 * Documentation for the disassembler can be found in X86Disassembler.h. 13 * 14 *===----------------------------------------------------------------------===*/ 15 16 #ifndef X86DISASSEMBLERDECODER_H 17 #define X86DISASSEMBLERDECODER_H 18 19 #ifdef __cplusplus 20 extern "C" { 21 #endif 22 23 #define INSTRUCTION_SPECIFIER_FIELDS \ 24 const char* name; 25 26 #define INSTRUCTION_IDS \ 27 const InstrUID *instructionIDs; 28 29 #include "X86DisassemblerDecoderCommon.h" 30 31 #undef INSTRUCTION_SPECIFIER_FIELDS 32 #undef INSTRUCTION_IDS 33 34 /* 35 * Accessor functions for various fields of an Intel instruction 36 */ 37 #define modFromModRM(modRM) (((modRM) & 0xc0) >> 6) 38 #define regFromModRM(modRM) (((modRM) & 0x38) >> 3) 39 #define rmFromModRM(modRM) ((modRM) & 0x7) 40 #define scaleFromSIB(sib) (((sib) & 0xc0) >> 6) 41 #define indexFromSIB(sib) (((sib) & 0x38) >> 3) 42 #define baseFromSIB(sib) ((sib) & 0x7) 43 #define wFromREX(rex) (((rex) & 0x8) >> 3) 44 #define rFromREX(rex) (((rex) & 0x4) >> 2) 45 #define xFromREX(rex) (((rex) & 0x2) >> 1) 46 #define bFromREX(rex) ((rex) & 0x1) 47 48 #define rFromVEX2of3(vex) (((~(vex)) & 0x80) >> 7) 49 #define xFromVEX2of3(vex) (((~(vex)) & 0x40) >> 6) 50 #define bFromVEX2of3(vex) (((~(vex)) & 0x20) >> 5) 51 #define mmmmmFromVEX2of3(vex) ((vex) & 0x1f) 52 #define wFromVEX3of3(vex) (((vex) & 0x80) >> 7) 53 #define vvvvFromVEX3of3(vex) (((~(vex)) & 0x78) >> 3) 54 #define lFromVEX3of3(vex) (((vex) & 0x4) >> 2) 55 #define ppFromVEX3of3(vex) ((vex) & 0x3) 56 57 #define rFromVEX2of2(vex) (((~(vex)) & 0x80) >> 7) 58 #define vvvvFromVEX2of2(vex) (((~(vex)) & 0x78) >> 3) 59 #define lFromVEX2of2(vex) (((vex) & 0x4) >> 2) 60 #define ppFromVEX2of2(vex) ((vex) & 0x3) 61 62 /* 63 * These enums represent Intel registers for use by the decoder. 64 */ 65 66 #define REGS_8BIT \ 67 ENTRY(AL) \ 68 ENTRY(CL) \ 69 ENTRY(DL) \ 70 ENTRY(BL) \ 71 ENTRY(AH) \ 72 ENTRY(CH) \ 73 ENTRY(DH) \ 74 ENTRY(BH) \ 75 ENTRY(R8B) \ 76 ENTRY(R9B) \ 77 ENTRY(R10B) \ 78 ENTRY(R11B) \ 79 ENTRY(R12B) \ 80 ENTRY(R13B) \ 81 ENTRY(R14B) \ 82 ENTRY(R15B) \ 83 ENTRY(SPL) \ 84 ENTRY(BPL) \ 85 ENTRY(SIL) \ 86 ENTRY(DIL) 87 88 #define EA_BASES_16BIT \ 89 ENTRY(BX_SI) \ 90 ENTRY(BX_DI) \ 91 ENTRY(BP_SI) \ 92 ENTRY(BP_DI) \ 93 ENTRY(SI) \ 94 ENTRY(DI) \ 95 ENTRY(BP) \ 96 ENTRY(BX) \ 97 ENTRY(R8W) \ 98 ENTRY(R9W) \ 99 ENTRY(R10W) \ 100 ENTRY(R11W) \ 101 ENTRY(R12W) \ 102 ENTRY(R13W) \ 103 ENTRY(R14W) \ 104 ENTRY(R15W) 105 106 #define REGS_16BIT \ 107 ENTRY(AX) \ 108 ENTRY(CX) \ 109 ENTRY(DX) \ 110 ENTRY(BX) \ 111 ENTRY(SP) \ 112 ENTRY(BP) \ 113 ENTRY(SI) \ 114 ENTRY(DI) \ 115 ENTRY(R8W) \ 116 ENTRY(R9W) \ 117 ENTRY(R10W) \ 118 ENTRY(R11W) \ 119 ENTRY(R12W) \ 120 ENTRY(R13W) \ 121 ENTRY(R14W) \ 122 ENTRY(R15W) 123 124 #define EA_BASES_32BIT \ 125 ENTRY(EAX) \ 126 ENTRY(ECX) \ 127 ENTRY(EDX) \ 128 ENTRY(EBX) \ 129 ENTRY(sib) \ 130 ENTRY(EBP) \ 131 ENTRY(ESI) \ 132 ENTRY(EDI) \ 133 ENTRY(R8D) \ 134 ENTRY(R9D) \ 135 ENTRY(R10D) \ 136 ENTRY(R11D) \ 137 ENTRY(R12D) \ 138 ENTRY(R13D) \ 139 ENTRY(R14D) \ 140 ENTRY(R15D) 141 142 #define REGS_32BIT \ 143 ENTRY(EAX) \ 144 ENTRY(ECX) \ 145 ENTRY(EDX) \ 146 ENTRY(EBX) \ 147 ENTRY(ESP) \ 148 ENTRY(EBP) \ 149 ENTRY(ESI) \ 150 ENTRY(EDI) \ 151 ENTRY(R8D) \ 152 ENTRY(R9D) \ 153 ENTRY(R10D) \ 154 ENTRY(R11D) \ 155 ENTRY(R12D) \ 156 ENTRY(R13D) \ 157 ENTRY(R14D) \ 158 ENTRY(R15D) 159 160 #define EA_BASES_64BIT \ 161 ENTRY(RAX) \ 162 ENTRY(RCX) \ 163 ENTRY(RDX) \ 164 ENTRY(RBX) \ 165 ENTRY(sib64) \ 166 ENTRY(RBP) \ 167 ENTRY(RSI) \ 168 ENTRY(RDI) \ 169 ENTRY(R8) \ 170 ENTRY(R9) \ 171 ENTRY(R10) \ 172 ENTRY(R11) \ 173 ENTRY(R12) \ 174 ENTRY(R13) \ 175 ENTRY(R14) \ 176 ENTRY(R15) 177 178 #define REGS_64BIT \ 179 ENTRY(RAX) \ 180 ENTRY(RCX) \ 181 ENTRY(RDX) \ 182 ENTRY(RBX) \ 183 ENTRY(RSP) \ 184 ENTRY(RBP) \ 185 ENTRY(RSI) \ 186 ENTRY(RDI) \ 187 ENTRY(R8) \ 188 ENTRY(R9) \ 189 ENTRY(R10) \ 190 ENTRY(R11) \ 191 ENTRY(R12) \ 192 ENTRY(R13) \ 193 ENTRY(R14) \ 194 ENTRY(R15) 195 196 #define REGS_MMX \ 197 ENTRY(MM0) \ 198 ENTRY(MM1) \ 199 ENTRY(MM2) \ 200 ENTRY(MM3) \ 201 ENTRY(MM4) \ 202 ENTRY(MM5) \ 203 ENTRY(MM6) \ 204 ENTRY(MM7) 205 206 #define REGS_XMM \ 207 ENTRY(XMM0) \ 208 ENTRY(XMM1) \ 209 ENTRY(XMM2) \ 210 ENTRY(XMM3) \ 211 ENTRY(XMM4) \ 212 ENTRY(XMM5) \ 213 ENTRY(XMM6) \ 214 ENTRY(XMM7) \ 215 ENTRY(XMM8) \ 216 ENTRY(XMM9) \ 217 ENTRY(XMM10) \ 218 ENTRY(XMM11) \ 219 ENTRY(XMM12) \ 220 ENTRY(XMM13) \ 221 ENTRY(XMM14) \ 222 ENTRY(XMM15) 223 224 #define REGS_YMM \ 225 ENTRY(YMM0) \ 226 ENTRY(YMM1) \ 227 ENTRY(YMM2) \ 228 ENTRY(YMM3) \ 229 ENTRY(YMM4) \ 230 ENTRY(YMM5) \ 231 ENTRY(YMM6) \ 232 ENTRY(YMM7) \ 233 ENTRY(YMM8) \ 234 ENTRY(YMM9) \ 235 ENTRY(YMM10) \ 236 ENTRY(YMM11) \ 237 ENTRY(YMM12) \ 238 ENTRY(YMM13) \ 239 ENTRY(YMM14) \ 240 ENTRY(YMM15) 241 242 #define REGS_SEGMENT \ 243 ENTRY(ES) \ 244 ENTRY(CS) \ 245 ENTRY(SS) \ 246 ENTRY(DS) \ 247 ENTRY(FS) \ 248 ENTRY(GS) 249 250 #define REGS_DEBUG \ 251 ENTRY(DR0) \ 252 ENTRY(DR1) \ 253 ENTRY(DR2) \ 254 ENTRY(DR3) \ 255 ENTRY(DR4) \ 256 ENTRY(DR5) \ 257 ENTRY(DR6) \ 258 ENTRY(DR7) 259 260 #define REGS_CONTROL \ 261 ENTRY(CR0) \ 262 ENTRY(CR1) \ 263 ENTRY(CR2) \ 264 ENTRY(CR3) \ 265 ENTRY(CR4) \ 266 ENTRY(CR5) \ 267 ENTRY(CR6) \ 268 ENTRY(CR7) \ 269 ENTRY(CR8) 270 271 #define ALL_EA_BASES \ 272 EA_BASES_16BIT \ 273 EA_BASES_32BIT \ 274 EA_BASES_64BIT 275 276 #define ALL_SIB_BASES \ 277 REGS_32BIT \ 278 REGS_64BIT 279 280 #define ALL_REGS \ 281 REGS_8BIT \ 282 REGS_16BIT \ 283 REGS_32BIT \ 284 REGS_64BIT \ 285 REGS_MMX \ 286 REGS_XMM \ 287 REGS_YMM \ 288 REGS_SEGMENT \ 289 REGS_DEBUG \ 290 REGS_CONTROL \ 291 ENTRY(RIP) 292 293 /* 294 * EABase - All possible values of the base field for effective-address 295 * computations, a.k.a. the Mod and R/M fields of the ModR/M byte. We 296 * distinguish between bases (EA_BASE_*) and registers that just happen to be 297 * referred to when Mod == 0b11 (EA_REG_*). 298 */ 299 typedef enum { 300 EA_BASE_NONE, 301 #define ENTRY(x) EA_BASE_##x, 302 ALL_EA_BASES 303 #undef ENTRY 304 #define ENTRY(x) EA_REG_##x, 305 ALL_REGS 306 #undef ENTRY 307 EA_max 308 } EABase; 309 310 /* 311 * SIBIndex - All possible values of the SIB index field. 312 * Borrows entries from ALL_EA_BASES with the special case that 313 * sib is synonymous with NONE. 314 */ 315 typedef enum { 316 SIB_INDEX_NONE, 317 #define ENTRY(x) SIB_INDEX_##x, 318 ALL_EA_BASES 319 #undef ENTRY 320 SIB_INDEX_max 321 } SIBIndex; 322 323 /* 324 * SIBBase - All possible values of the SIB base field. 325 */ 326 typedef enum { 327 SIB_BASE_NONE, 328 #define ENTRY(x) SIB_BASE_##x, 329 ALL_SIB_BASES 330 #undef ENTRY 331 SIB_BASE_max 332 } SIBBase; 333 334 /* 335 * EADisplacement - Possible displacement types for effective-address 336 * computations. 337 */ 338 typedef enum { 339 EA_DISP_NONE, 340 EA_DISP_8, 341 EA_DISP_16, 342 EA_DISP_32 343 } EADisplacement; 344 345 /* 346 * Reg - All possible values of the reg field in the ModR/M byte. 347 */ 348 typedef enum { 349 #define ENTRY(x) MODRM_REG_##x, 350 ALL_REGS 351 #undef ENTRY 352 MODRM_REG_max 353 } Reg; 354 355 /* 356 * SegmentOverride - All possible segment overrides. 357 */ 358 typedef enum { 359 SEG_OVERRIDE_NONE, 360 SEG_OVERRIDE_CS, 361 SEG_OVERRIDE_SS, 362 SEG_OVERRIDE_DS, 363 SEG_OVERRIDE_ES, 364 SEG_OVERRIDE_FS, 365 SEG_OVERRIDE_GS, 366 SEG_OVERRIDE_max 367 } SegmentOverride; 368 369 /* 370 * VEXLeadingOpcodeByte - Possible values for the VEX.m-mmmm field 371 */ 372 373 typedef enum { 374 VEX_LOB_0F = 0x1, 375 VEX_LOB_0F38 = 0x2, 376 VEX_LOB_0F3A = 0x3 377 } VEXLeadingOpcodeByte; 378 379 /* 380 * VEXPrefixCode - Possible values for the VEX.pp field 381 */ 382 383 typedef enum { 384 VEX_PREFIX_NONE = 0x0, 385 VEX_PREFIX_66 = 0x1, 386 VEX_PREFIX_F3 = 0x2, 387 VEX_PREFIX_F2 = 0x3 388 } VEXPrefixCode; 389 390 typedef uint8_t BOOL; 391 392 /* 393 * byteReader_t - Type for the byte reader that the consumer must provide to 394 * the decoder. Reads a single byte from the instruction's address space. 395 * @param arg - A baton that the consumer can associate with any internal 396 * state that it needs. 397 * @param byte - A pointer to a single byte in memory that should be set to 398 * contain the value at address. 399 * @param address - The address in the instruction's address space that should 400 * be read from. 401 * @return - -1 if the byte cannot be read for any reason; 0 otherwise. 402 */ 403 typedef int (*byteReader_t)(void* arg, uint8_t* byte, uint64_t address); 404 405 /* 406 * dlog_t - Type for the logging function that the consumer can provide to 407 * get debugging output from the decoder. 408 * @param arg - A baton that the consumer can associate with any internal 409 * state that it needs. 410 * @param log - A string that contains the message. Will be reused after 411 * the logger returns. 412 */ 413 typedef void (*dlog_t)(void* arg, const char *log); 414 415 /* 416 * The x86 internal instruction, which is produced by the decoder. 417 */ 418 struct InternalInstruction { 419 /* Reader interface (C) */ 420 byteReader_t reader; 421 /* Opaque value passed to the reader */ 422 void* readerArg; 423 /* The address of the next byte to read via the reader */ 424 uint64_t readerCursor; 425 426 /* Logger interface (C) */ 427 dlog_t dlog; 428 /* Opaque value passed to the logger */ 429 void* dlogArg; 430 431 /* General instruction information */ 432 433 /* The mode to disassemble for (64-bit, protected, real) */ 434 DisassemblerMode mode; 435 /* The start of the instruction, usable with the reader */ 436 uint64_t startLocation; 437 /* The length of the instruction, in bytes */ 438 size_t length; 439 440 /* Prefix state */ 441 442 /* 1 if the prefix byte corresponding to the entry is present; 0 if not */ 443 uint8_t prefixPresent[0x100]; 444 /* contains the location (for use with the reader) of the prefix byte */ 445 uint64_t prefixLocations[0x100]; 446 /* The value of the VEX prefix, if present */ 447 uint8_t vexPrefix[3]; 448 /* The length of the VEX prefix (0 if not present) */ 449 uint8_t vexSize; 450 /* The value of the REX prefix, if present */ 451 uint8_t rexPrefix; 452 /* The location where a mandatory prefix would have to be (i.e., right before 453 the opcode, or right before the REX prefix if one is present) */ 454 uint64_t necessaryPrefixLocation; 455 /* The segment override type */ 456 SegmentOverride segmentOverride; 457 458 /* Sizes of various critical pieces of data, in bytes */ 459 uint8_t registerSize; 460 uint8_t addressSize; 461 uint8_t displacementSize; 462 uint8_t immediateSize; 463 464 /* opcode state */ 465 466 /* The value of the two-byte escape prefix (usually 0x0f) */ 467 uint8_t twoByteEscape; 468 /* The value of the three-byte escape prefix (usually 0x38 or 0x3a) */ 469 uint8_t threeByteEscape; 470 /* The last byte of the opcode, not counting any ModR/M extension */ 471 uint8_t opcode; 472 /* The ModR/M byte of the instruction, if it is an opcode extension */ 473 uint8_t modRMExtension; 474 475 /* decode state */ 476 477 /* The type of opcode, used for indexing into the array of decode tables */ 478 OpcodeType opcodeType; 479 /* The instruction ID, extracted from the decode table */ 480 uint16_t instructionID; 481 /* The specifier for the instruction, from the instruction info table */ 482 const struct InstructionSpecifier *spec; 483 484 /* state for additional bytes, consumed during operand decode. Pattern: 485 consumed___ indicates that the byte was already consumed and does not 486 need to be consumed again */ 487 488 /* The VEX.vvvv field, which contains a third register operand for some AVX 489 instructions */ 490 Reg vvvv; 491 492 /* The ModR/M byte, which contains most register operands and some portion of 493 all memory operands */ 494 BOOL consumedModRM; 495 uint8_t modRM; 496 497 /* The SIB byte, used for more complex 32- or 64-bit memory operands */ 498 BOOL consumedSIB; 499 uint8_t sib; 500 501 /* The displacement, used for memory operands */ 502 BOOL consumedDisplacement; 503 int32_t displacement; 504 505 /* Immediates. There can be two in some cases */ 506 uint8_t numImmediatesConsumed; 507 uint8_t numImmediatesTranslated; 508 uint64_t immediates[2]; 509 510 /* A register or immediate operand encoded into the opcode */ 511 BOOL consumedOpcodeModifier; 512 uint8_t opcodeModifier; 513 Reg opcodeRegister; 514 515 /* Portions of the ModR/M byte */ 516 517 /* These fields determine the allowable values for the ModR/M fields, which 518 depend on operand and address widths */ 519 EABase eaBaseBase; 520 EABase eaRegBase; 521 Reg regBase; 522 523 /* The Mod and R/M fields can encode a base for an effective address, or a 524 register. These are separated into two fields here */ 525 EABase eaBase; 526 EADisplacement eaDisplacement; 527 /* The reg field always encodes a register */ 528 Reg reg; 529 530 /* SIB state */ 531 SIBIndex sibIndex; 532 uint8_t sibScale; 533 SIBBase sibBase; 534 }; 535 536 /* decodeInstruction - Decode one instruction and store the decoding results in 537 * a buffer provided by the consumer. 538 * @param insn - The buffer to store the instruction in. Allocated by the 539 * consumer. 540 * @param reader - The byteReader_t for the bytes to be read. 541 * @param readerArg - An argument to pass to the reader for storing context 542 * specific to the consumer. May be NULL. 543 * @param logger - The dlog_t to be used in printing status messages from the 544 * disassembler. May be NULL. 545 * @param loggerArg - An argument to pass to the logger for storing context 546 * specific to the logger. May be NULL. 547 * @param startLoc - The address (in the reader's address space) of the first 548 * byte in the instruction. 549 * @param mode - The mode (16-bit, 32-bit, 64-bit) to decode in. 550 * @return - Nonzero if there was an error during decode, 0 otherwise. 551 */ 552 int decodeInstruction(struct InternalInstruction* insn, 553 byteReader_t reader, 554 void* readerArg, 555 dlog_t logger, 556 void* loggerArg, 557 uint64_t startLoc, 558 DisassemblerMode mode); 559 560 /* x86DisassemblerDebug - C-accessible function for printing a message to 561 * debugs() 562 * @param file - The name of the file printing the debug message. 563 * @param line - The line number that printed the debug message. 564 * @param s - The message to print. 565 */ 566 567 void x86DisassemblerDebug(const char *file, 568 unsigned line, 569 const char *s); 570 571 #ifdef __cplusplus 572 } 573 #endif 574 575 #endif 576