1 /*===-- X86DisassemblerDecoderInternal.h - Disassembler decoder ---*- C -*-===*
3 * The LLVM Compiler Infrastructure
5 * This file is distributed under the University of Illinois Open Source
6 * License. See LICENSE.TXT for details.
8 *===----------------------------------------------------------------------===*
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.
14 *===----------------------------------------------------------------------===*/
16 #ifndef X86DISASSEMBLERDECODER_H
17 #define X86DISASSEMBLERDECODER_H
23 #define INSTRUCTION_SPECIFIER_FIELDS \
26 #define INSTRUCTION_IDS \
27 uint16_t instructionIDs;
29 #include "X86DisassemblerDecoderCommon.h"
31 #undef INSTRUCTION_SPECIFIER_FIELDS
32 #undef INSTRUCTION_IDS
35 * Accessor functions for various fields of an Intel instruction
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)
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)
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)
62 #define rFromXOP2of3(xop) (((~(xop)) & 0x80) >> 7)
63 #define xFromXOP2of3(xop) (((~(xop)) & 0x40) >> 6)
64 #define bFromXOP2of3(xop) (((~(xop)) & 0x20) >> 5)
65 #define mmmmmFromXOP2of3(xop) ((xop) & 0x1f)
66 #define wFromXOP3of3(xop) (((xop) & 0x80) >> 7)
67 #define vvvvFromXOP3of3(vex) (((~(vex)) & 0x78) >> 3)
68 #define lFromXOP3of3(xop) (((xop) & 0x4) >> 2)
69 #define ppFromXOP3of3(xop) ((xop) & 0x3)
72 * These enums represent Intel registers for use by the decoder.
97 #define EA_BASES_16BIT \
133 #define EA_BASES_32BIT \
169 #define EA_BASES_64BIT \
317 #define REGS_SEGMENT \
335 #define REGS_CONTROL \
346 #define ALL_EA_BASES \
351 #define ALL_SIB_BASES \
370 * EABase - All possible values of the base field for effective-address
371 * computations, a.k.a. the Mod and R/M fields of the ModR/M byte. We
372 * distinguish between bases (EA_BASE_*) and registers that just happen to be
373 * referred to when Mod == 0b11 (EA_REG_*).
377 #define ENTRY(x) EA_BASE_##x,
380 #define ENTRY(x) EA_REG_##x,
387 * SIBIndex - All possible values of the SIB index field.
388 * Borrows entries from ALL_EA_BASES with the special case that
389 * sib is synonymous with NONE.
390 * Vector SIB: index can be XMM or YMM.
394 #define ENTRY(x) SIB_INDEX_##x,
404 * SIBBase - All possible values of the SIB base field.
408 #define ENTRY(x) SIB_BASE_##x,
415 * EADisplacement - Possible displacement types for effective-address
426 * Reg - All possible values of the reg field in the ModR/M byte.
429 #define ENTRY(x) MODRM_REG_##x,
436 * SegmentOverride - All possible segment overrides.
450 * VEXLeadingOpcodeByte - Possible values for the VEX.m-mmmm field
457 } VEXLeadingOpcodeByte;
460 XOP_MAP_SELECT_8 = 0x8,
461 XOP_MAP_SELECT_9 = 0x9,
462 XOP_MAP_SELECT_A = 0xA
466 * VEXPrefixCode - Possible values for the VEX.pp field
470 VEX_PREFIX_NONE = 0x0,
477 TYPE_NO_VEX_XOP = 0x0,
483 typedef uint8_t BOOL;
486 * byteReader_t - Type for the byte reader that the consumer must provide to
487 * the decoder. Reads a single byte from the instruction's address space.
488 * @param arg - A baton that the consumer can associate with any internal
489 * state that it needs.
490 * @param byte - A pointer to a single byte in memory that should be set to
491 * contain the value at address.
492 * @param address - The address in the instruction's address space that should
494 * @return - -1 if the byte cannot be read for any reason; 0 otherwise.
496 typedef int (*byteReader_t)(const void* arg, uint8_t* byte, uint64_t address);
499 * dlog_t - Type for the logging function that the consumer can provide to
500 * get debugging output from the decoder.
501 * @param arg - A baton that the consumer can associate with any internal
502 * state that it needs.
503 * @param log - A string that contains the message. Will be reused after
504 * the logger returns.
506 typedef void (*dlog_t)(void* arg, const char *log);
509 * The x86 internal instruction, which is produced by the decoder.
511 struct InternalInstruction {
512 /* Reader interface (C) */
514 /* Opaque value passed to the reader */
515 const void* readerArg;
516 /* The address of the next byte to read via the reader */
517 uint64_t readerCursor;
519 /* Logger interface (C) */
521 /* Opaque value passed to the logger */
524 /* General instruction information */
526 /* The mode to disassemble for (64-bit, protected, real) */
527 DisassemblerMode mode;
528 /* The start of the instruction, usable with the reader */
529 uint64_t startLocation;
530 /* The length of the instruction, in bytes */
535 /* 1 if the prefix byte corresponding to the entry is present; 0 if not */
536 uint8_t prefixPresent[0x100];
537 /* contains the location (for use with the reader) of the prefix byte */
538 uint64_t prefixLocations[0x100];
539 /* The value of the VEX/XOP prefix, if present */
540 uint8_t vexXopPrefix[3];
541 /* The length of the VEX prefix (0 if not present) */
542 VEXXOPType vexXopType;
543 /* The value of the REX prefix, if present */
545 /* The location where a mandatory prefix would have to be (i.e., right before
546 the opcode, or right before the REX prefix if one is present) */
547 uint64_t necessaryPrefixLocation;
548 /* The segment override type */
549 SegmentOverride segmentOverride;
550 /* 1 if the prefix byte, 0xf2 or 0xf3 is xacquire or xrelease */
551 BOOL xAcquireRelease;
553 /* Sizes of various critical pieces of data, in bytes */
554 uint8_t registerSize;
556 uint8_t displacementSize;
557 uint8_t immediateSize;
559 /* Offsets from the start of the instruction to the pieces of data, which is
560 needed to find relocation entries for adding symbolic operands */
561 uint8_t displacementOffset;
562 uint8_t immediateOffset;
566 /* The last byte of the opcode, not counting any ModR/M extension */
568 /* The ModR/M byte of the instruction, if it is an opcode extension */
569 uint8_t modRMExtension;
573 /* The type of opcode, used for indexing into the array of decode tables */
574 OpcodeType opcodeType;
575 /* The instruction ID, extracted from the decode table */
576 uint16_t instructionID;
577 /* The specifier for the instruction, from the instruction info table */
578 const struct InstructionSpecifier *spec;
580 /* state for additional bytes, consumed during operand decode. Pattern:
581 consumed___ indicates that the byte was already consumed and does not
582 need to be consumed again */
584 /* The VEX.vvvv field, which contains a third register operand for some AVX
588 /* The ModR/M byte, which contains most register operands and some portion of
589 all memory operands */
593 /* The SIB byte, used for more complex 32- or 64-bit memory operands */
597 /* The displacement, used for memory operands */
598 BOOL consumedDisplacement;
599 int32_t displacement;
601 /* Immediates. There can be two in some cases */
602 uint8_t numImmediatesConsumed;
603 uint8_t numImmediatesTranslated;
604 uint64_t immediates[2];
606 /* A register or immediate operand encoded into the opcode */
607 BOOL consumedOpcodeModifier;
608 uint8_t opcodeModifier;
611 /* Portions of the ModR/M byte */
613 /* These fields determine the allowable values for the ModR/M fields, which
614 depend on operand and address widths */
619 /* The Mod and R/M fields can encode a base for an effective address, or a
620 register. These are separated into two fields here */
622 EADisplacement eaDisplacement;
623 /* The reg field always encodes a register */
631 const struct OperandSpecifier *operands;
634 /* decodeInstruction - Decode one instruction and store the decoding results in
635 * a buffer provided by the consumer.
636 * @param insn - The buffer to store the instruction in. Allocated by the
638 * @param reader - The byteReader_t for the bytes to be read.
639 * @param readerArg - An argument to pass to the reader for storing context
640 * specific to the consumer. May be NULL.
641 * @param logger - The dlog_t to be used in printing status messages from the
642 * disassembler. May be NULL.
643 * @param loggerArg - An argument to pass to the logger for storing context
644 * specific to the logger. May be NULL.
645 * @param startLoc - The address (in the reader's address space) of the first
646 * byte in the instruction.
647 * @param mode - The mode (16-bit, 32-bit, 64-bit) to decode in.
648 * @return - Nonzero if there was an error during decode, 0 otherwise.
650 int decodeInstruction(struct InternalInstruction* insn,
652 const void* readerArg,
657 DisassemblerMode mode);
659 /* x86DisassemblerDebug - C-accessible function for printing a message to
661 * @param file - The name of the file printing the debug message.
662 * @param line - The line number that printed the debug message.
663 * @param s - The message to print.
666 void x86DisassemblerDebug(const char *file,
670 const char *x86DisassemblerGetInstrName(unsigned Opcode, const void *mii);