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 uint16_t operands; 25 26 #define INSTRUCTION_IDS \ 27 unsigned 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 * Vector SIB: index can be XMM or YMM. 315 */ 316 typedef enum { 317 SIB_INDEX_NONE, 318 #define ENTRY(x) SIB_INDEX_##x, 319 ALL_EA_BASES 320 REGS_XMM 321 REGS_YMM 322 #undef ENTRY 323 SIB_INDEX_max 324 } SIBIndex; 325 326 /* 327 * SIBBase - All possible values of the SIB base field. 328 */ 329 typedef enum { 330 SIB_BASE_NONE, 331 #define ENTRY(x) SIB_BASE_##x, 332 ALL_SIB_BASES 333 #undef ENTRY 334 SIB_BASE_max 335 } SIBBase; 336 337 /* 338 * EADisplacement - Possible displacement types for effective-address 339 * computations. 340 */ 341 typedef enum { 342 EA_DISP_NONE, 343 EA_DISP_8, 344 EA_DISP_16, 345 EA_DISP_32 346 } EADisplacement; 347 348 /* 349 * Reg - All possible values of the reg field in the ModR/M byte. 350 */ 351 typedef enum { 352 #define ENTRY(x) MODRM_REG_##x, 353 ALL_REGS 354 #undef ENTRY 355 MODRM_REG_max 356 } Reg; 357 358 /* 359 * SegmentOverride - All possible segment overrides. 360 */ 361 typedef enum { 362 SEG_OVERRIDE_NONE, 363 SEG_OVERRIDE_CS, 364 SEG_OVERRIDE_SS, 365 SEG_OVERRIDE_DS, 366 SEG_OVERRIDE_ES, 367 SEG_OVERRIDE_FS, 368 SEG_OVERRIDE_GS, 369 SEG_OVERRIDE_max 370 } SegmentOverride; 371 372 /* 373 * VEXLeadingOpcodeByte - Possible values for the VEX.m-mmmm field 374 */ 375 376 typedef enum { 377 VEX_LOB_0F = 0x1, 378 VEX_LOB_0F38 = 0x2, 379 VEX_LOB_0F3A = 0x3 380 } VEXLeadingOpcodeByte; 381 382 /* 383 * VEXPrefixCode - Possible values for the VEX.pp field 384 */ 385 386 typedef enum { 387 VEX_PREFIX_NONE = 0x0, 388 VEX_PREFIX_66 = 0x1, 389 VEX_PREFIX_F3 = 0x2, 390 VEX_PREFIX_F2 = 0x3 391 } VEXPrefixCode; 392 393 typedef uint8_t BOOL; 394 395 /* 396 * byteReader_t - Type for the byte reader that the consumer must provide to 397 * the decoder. Reads a single byte from the instruction's address space. 398 * @param arg - A baton that the consumer can associate with any internal 399 * state that it needs. 400 * @param byte - A pointer to a single byte in memory that should be set to 401 * contain the value at address. 402 * @param address - The address in the instruction's address space that should 403 * be read from. 404 * @return - -1 if the byte cannot be read for any reason; 0 otherwise. 405 */ 406 typedef int (*byteReader_t)(const void* arg, uint8_t* byte, uint64_t address); 407 408 /* 409 * dlog_t - Type for the logging function that the consumer can provide to 410 * get debugging output from the decoder. 411 * @param arg - A baton that the consumer can associate with any internal 412 * state that it needs. 413 * @param log - A string that contains the message. Will be reused after 414 * the logger returns. 415 */ 416 typedef void (*dlog_t)(void* arg, const char *log); 417 418 /* 419 * The x86 internal instruction, which is produced by the decoder. 420 */ 421 struct InternalInstruction { 422 /* Reader interface (C) */ 423 byteReader_t reader; 424 /* Opaque value passed to the reader */ 425 const void* readerArg; 426 /* The address of the next byte to read via the reader */ 427 uint64_t readerCursor; 428 429 /* Logger interface (C) */ 430 dlog_t dlog; 431 /* Opaque value passed to the logger */ 432 void* dlogArg; 433 434 /* General instruction information */ 435 436 /* The mode to disassemble for (64-bit, protected, real) */ 437 DisassemblerMode mode; 438 /* The start of the instruction, usable with the reader */ 439 uint64_t startLocation; 440 /* The length of the instruction, in bytes */ 441 size_t length; 442 443 /* Prefix state */ 444 445 /* 1 if the prefix byte corresponding to the entry is present; 0 if not */ 446 uint8_t prefixPresent[0x100]; 447 /* contains the location (for use with the reader) of the prefix byte */ 448 uint64_t prefixLocations[0x100]; 449 /* The value of the VEX prefix, if present */ 450 uint8_t vexPrefix[3]; 451 /* The length of the VEX prefix (0 if not present) */ 452 uint8_t vexSize; 453 /* The value of the REX prefix, if present */ 454 uint8_t rexPrefix; 455 /* The location where a mandatory prefix would have to be (i.e., right before 456 the opcode, or right before the REX prefix if one is present) */ 457 uint64_t necessaryPrefixLocation; 458 /* The segment override type */ 459 SegmentOverride segmentOverride; 460 461 /* Sizes of various critical pieces of data, in bytes */ 462 uint8_t registerSize; 463 uint8_t addressSize; 464 uint8_t displacementSize; 465 uint8_t immediateSize; 466 467 /* Offsets from the start of the instruction to the pieces of data, which is 468 needed to find relocation entries for adding symbolic operands */ 469 uint8_t displacementOffset; 470 uint8_t immediateOffset; 471 472 /* opcode state */ 473 474 /* The value of the two-byte escape prefix (usually 0x0f) */ 475 uint8_t twoByteEscape; 476 /* The value of the three-byte escape prefix (usually 0x38 or 0x3a) */ 477 uint8_t threeByteEscape; 478 /* The last byte of the opcode, not counting any ModR/M extension */ 479 uint8_t opcode; 480 /* The ModR/M byte of the instruction, if it is an opcode extension */ 481 uint8_t modRMExtension; 482 483 /* decode state */ 484 485 /* The type of opcode, used for indexing into the array of decode tables */ 486 OpcodeType opcodeType; 487 /* The instruction ID, extracted from the decode table */ 488 uint16_t instructionID; 489 /* The specifier for the instruction, from the instruction info table */ 490 const struct InstructionSpecifier *spec; 491 492 /* state for additional bytes, consumed during operand decode. Pattern: 493 consumed___ indicates that the byte was already consumed and does not 494 need to be consumed again */ 495 496 /* The VEX.vvvv field, which contains a third register operand for some AVX 497 instructions */ 498 Reg vvvv; 499 500 /* The ModR/M byte, which contains most register operands and some portion of 501 all memory operands */ 502 BOOL consumedModRM; 503 uint8_t modRM; 504 505 /* The SIB byte, used for more complex 32- or 64-bit memory operands */ 506 BOOL consumedSIB; 507 uint8_t sib; 508 509 /* The displacement, used for memory operands */ 510 BOOL consumedDisplacement; 511 int32_t displacement; 512 513 /* Immediates. There can be two in some cases */ 514 uint8_t numImmediatesConsumed; 515 uint8_t numImmediatesTranslated; 516 uint64_t immediates[2]; 517 518 /* A register or immediate operand encoded into the opcode */ 519 BOOL consumedOpcodeModifier; 520 uint8_t opcodeModifier; 521 Reg opcodeRegister; 522 523 /* Portions of the ModR/M byte */ 524 525 /* These fields determine the allowable values for the ModR/M fields, which 526 depend on operand and address widths */ 527 EABase eaBaseBase; 528 EABase eaRegBase; 529 Reg regBase; 530 531 /* The Mod and R/M fields can encode a base for an effective address, or a 532 register. These are separated into two fields here */ 533 EABase eaBase; 534 EADisplacement eaDisplacement; 535 /* The reg field always encodes a register */ 536 Reg reg; 537 538 /* SIB state */ 539 SIBIndex sibIndex; 540 uint8_t sibScale; 541 SIBBase sibBase; 542 543 const struct OperandSpecifier *operands; 544 }; 545 546 /* decodeInstruction - Decode one instruction and store the decoding results in 547 * a buffer provided by the consumer. 548 * @param insn - The buffer to store the instruction in. Allocated by the 549 * consumer. 550 * @param reader - The byteReader_t for the bytes to be read. 551 * @param readerArg - An argument to pass to the reader for storing context 552 * specific to the consumer. May be NULL. 553 * @param logger - The dlog_t to be used in printing status messages from the 554 * disassembler. May be NULL. 555 * @param loggerArg - An argument to pass to the logger for storing context 556 * specific to the logger. May be NULL. 557 * @param startLoc - The address (in the reader's address space) of the first 558 * byte in the instruction. 559 * @param mode - The mode (16-bit, 32-bit, 64-bit) to decode in. 560 * @return - Nonzero if there was an error during decode, 0 otherwise. 561 */ 562 int decodeInstruction(struct InternalInstruction* insn, 563 byteReader_t reader, 564 const void* readerArg, 565 dlog_t logger, 566 void* loggerArg, 567 const void* miiArg, 568 uint64_t startLoc, 569 DisassemblerMode mode); 570 571 /* x86DisassemblerDebug - C-accessible function for printing a message to 572 * debugs() 573 * @param file - The name of the file printing the debug message. 574 * @param line - The line number that printed the debug message. 575 * @param s - The message to print. 576 */ 577 578 void x86DisassemblerDebug(const char *file, 579 unsigned line, 580 const char *s); 581 582 const char *x86DisassemblerGetInstrName(unsigned Opcode, const void *mii); 583 584 #ifdef __cplusplus 585 } 586 #endif 587 588 #endif 589