1 //===-- X86DisassemblerDecoderInternal.h - Disassembler decoder -*- C++ -*-===//
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //===----------------------------------------------------------------------===//
9 // This file is part of the X86 Disassembler.
10 // It contains the public interface of the instruction decoder.
11 // Documentation for the disassembler can be found in X86Disassembler.h.
13 //===----------------------------------------------------------------------===//
15 #ifndef LLVM_LIB_TARGET_X86_DISASSEMBLER_X86DISASSEMBLERDECODER_H
16 #define LLVM_LIB_TARGET_X86_DISASSEMBLER_X86DISASSEMBLERDECODER_H
18 #include "llvm/ADT/ArrayRef.h"
19 #include "llvm/Support/X86DisassemblerDecoderCommon.h"
22 namespace X86Disassembler
{
24 // Accessor functions for various fields of an Intel instruction
25 #define modFromModRM(modRM) (((modRM) & 0xc0) >> 6)
26 #define regFromModRM(modRM) (((modRM) & 0x38) >> 3)
27 #define rmFromModRM(modRM) ((modRM) & 0x7)
28 #define scaleFromSIB(sib) (((sib) & 0xc0) >> 6)
29 #define indexFromSIB(sib) (((sib) & 0x38) >> 3)
30 #define baseFromSIB(sib) ((sib) & 0x7)
31 #define wFromREX(rex) (((rex) & 0x8) >> 3)
32 #define rFromREX(rex) (((rex) & 0x4) >> 2)
33 #define xFromREX(rex) (((rex) & 0x2) >> 1)
34 #define bFromREX(rex) ((rex) & 0x1)
36 #define rFromEVEX2of4(evex) (((~(evex)) & 0x80) >> 7)
37 #define xFromEVEX2of4(evex) (((~(evex)) & 0x40) >> 6)
38 #define bFromEVEX2of4(evex) (((~(evex)) & 0x20) >> 5)
39 #define r2FromEVEX2of4(evex) (((~(evex)) & 0x10) >> 4)
40 #define mmFromEVEX2of4(evex) ((evex) & 0x3)
41 #define wFromEVEX3of4(evex) (((evex) & 0x80) >> 7)
42 #define vvvvFromEVEX3of4(evex) (((~(evex)) & 0x78) >> 3)
43 #define ppFromEVEX3of4(evex) ((evex) & 0x3)
44 #define zFromEVEX4of4(evex) (((evex) & 0x80) >> 7)
45 #define l2FromEVEX4of4(evex) (((evex) & 0x40) >> 6)
46 #define lFromEVEX4of4(evex) (((evex) & 0x20) >> 5)
47 #define bFromEVEX4of4(evex) (((evex) & 0x10) >> 4)
48 #define v2FromEVEX4of4(evex) (((~evex) & 0x8) >> 3)
49 #define aaaFromEVEX4of4(evex) ((evex) & 0x7)
51 #define rFromVEX2of3(vex) (((~(vex)) & 0x80) >> 7)
52 #define xFromVEX2of3(vex) (((~(vex)) & 0x40) >> 6)
53 #define bFromVEX2of3(vex) (((~(vex)) & 0x20) >> 5)
54 #define mmmmmFromVEX2of3(vex) ((vex) & 0x1f)
55 #define wFromVEX3of3(vex) (((vex) & 0x80) >> 7)
56 #define vvvvFromVEX3of3(vex) (((~(vex)) & 0x78) >> 3)
57 #define lFromVEX3of3(vex) (((vex) & 0x4) >> 2)
58 #define ppFromVEX3of3(vex) ((vex) & 0x3)
60 #define rFromVEX2of2(vex) (((~(vex)) & 0x80) >> 7)
61 #define vvvvFromVEX2of2(vex) (((~(vex)) & 0x78) >> 3)
62 #define lFromVEX2of2(vex) (((vex) & 0x4) >> 2)
63 #define ppFromVEX2of2(vex) ((vex) & 0x3)
65 #define rFromXOP2of3(xop) (((~(xop)) & 0x80) >> 7)
66 #define xFromXOP2of3(xop) (((~(xop)) & 0x40) >> 6)
67 #define bFromXOP2of3(xop) (((~(xop)) & 0x20) >> 5)
68 #define mmmmmFromXOP2of3(xop) ((xop) & 0x1f)
69 #define wFromXOP3of3(xop) (((xop) & 0x80) >> 7)
70 #define vvvvFromXOP3of3(vex) (((~(vex)) & 0x78) >> 3)
71 #define lFromXOP3of3(xop) (((xop) & 0x4) >> 2)
72 #define ppFromXOP3of3(xop) ((xop) & 0x3)
74 // 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 \
327 #define REGS_SEGMENT \
353 #define REGS_CONTROL \
377 #define ALL_EA_BASES \
382 #define ALL_SIB_BASES \
402 /// All possible values of the base field for effective-address
403 /// computations, a.k.a. the Mod and R/M fields of the ModR/M byte.
404 /// We distinguish between bases (EA_BASE_*) and registers that just happen
405 /// to be referred to when Mod == 0b11 (EA_REG_*).
408 #define ENTRY(x) EA_BASE_##x,
411 #define ENTRY(x) EA_REG_##x,
417 /// All possible values of the SIB index field.
418 /// borrows entries from ALL_EA_BASES with the special case that
419 /// sib is synonymous with NONE.
420 /// Vector SIB: index can be XMM or YMM.
423 #define ENTRY(x) SIB_INDEX_##x,
432 /// All possible values of the SIB base field.
435 #define ENTRY(x) SIB_BASE_##x,
441 /// Possible displacement types for effective-address computations.
449 /// All possible values of the reg field in the ModR/M byte.
451 #define ENTRY(x) MODRM_REG_##x,
457 /// All possible segment overrides.
458 enum SegmentOverride
{
469 /// Possible values for the VEX.m-mmmm field
470 enum VEXLeadingOpcodeByte
{
477 XOP_MAP_SELECT_8
= 0x8,
478 XOP_MAP_SELECT_9
= 0x9,
479 XOP_MAP_SELECT_A
= 0xA
482 /// Possible values for the VEX.pp/EVEX.pp field
484 VEX_PREFIX_NONE
= 0x0,
490 enum VectorExtensionType
{
491 TYPE_NO_VEX_XOP
= 0x0,
498 /// Type for the byte reader that the consumer must provide to
499 /// the decoder. Reads a single byte from the instruction's address space.
500 /// \param arg A baton that the consumer can associate with any internal
501 /// state that it needs.
502 /// \param byte A pointer to a single byte in memory that should be set to
503 /// contain the value at address.
504 /// \param address The address in the instruction's address space that should
506 /// \return -1 if the byte cannot be read for any reason; 0 otherwise.
507 typedef int (*byteReader_t
)(const void *arg
, uint8_t *byte
, uint64_t address
);
509 /// Type for the logging function that the consumer can provide to
510 /// get debugging output from the decoder.
511 /// \param arg A baton that the consumer can associate with any internal
512 /// state that it needs.
513 /// \param log A string that contains the message. Will be reused after
514 /// the logger returns.
515 typedef void (*dlog_t
)(void *arg
, const char *log
);
517 /// The specification for how to extract and interpret a full instruction and
519 struct InstructionSpecifier
{
523 /// The x86 internal instruction, which is produced by the decoder.
524 struct InternalInstruction
{
525 // Reader interface (C)
527 // Opaque value passed to the reader
528 const void* readerArg
;
529 // The address of the next byte to read via the reader
530 uint64_t readerCursor
;
532 // Logger interface (C)
534 // Opaque value passed to the logger
537 // General instruction information
539 // The mode to disassemble for (64-bit, protected, real)
540 DisassemblerMode mode
;
541 // The start of the instruction, usable with the reader
542 uint64_t startLocation
;
543 // The length of the instruction, in bytes
548 // The possible mandatory prefix
549 uint8_t mandatoryPrefix
;
550 // The value of the vector extension prefix(EVEX/VEX/XOP), if present
551 uint8_t vectorExtensionPrefix
[4];
552 // The type of the vector extension prefix
553 VectorExtensionType vectorExtensionType
;
554 // The value of the REX prefix, if present
556 // The segment override type
557 SegmentOverride segmentOverride
;
558 // 1 if the prefix byte, 0xf2 or 0xf3 is xacquire or xrelease
559 bool xAcquireRelease
;
561 // Address-size override
563 // Operand-size override
567 // The repeat prefix if any
568 uint8_t repeatPrefix
;
570 // Sizes of various critical pieces of data, in bytes
571 uint8_t registerSize
;
573 uint8_t displacementSize
;
574 uint8_t immediateSize
;
576 // Offsets from the start of the instruction to the pieces of data, which is
577 // needed to find relocation entries for adding symbolic operands.
578 uint8_t displacementOffset
;
579 uint8_t immediateOffset
;
583 // The last byte of the opcode, not counting any ModR/M extension
588 // The type of opcode, used for indexing into the array of decode tables
589 OpcodeType opcodeType
;
590 // The instruction ID, extracted from the decode table
591 uint16_t instructionID
;
592 // The specifier for the instruction, from the instruction info table
593 const InstructionSpecifier
*spec
;
595 // state for additional bytes, consumed during operand decode. Pattern:
596 // consumed___ indicates that the byte was already consumed and does not
597 // need to be consumed again.
599 // The VEX.vvvv field, which contains a third register operand for some AVX
603 // The writemask for AVX-512 instructions which is contained in EVEX.aaa
606 // The ModR/M byte, which contains most register operands and some portion of
607 // all memory operands.
611 // The SIB byte, used for more complex 32- or 64-bit memory operands
615 // The displacement, used for memory operands
616 bool consumedDisplacement
;
617 int32_t displacement
;
619 // Immediates. There can be two in some cases
620 uint8_t numImmediatesConsumed
;
621 uint8_t numImmediatesTranslated
;
622 uint64_t immediates
[2];
624 // A register or immediate operand encoded into the opcode
627 // Portions of the ModR/M byte
629 // These fields determine the allowable values for the ModR/M fields, which
630 // depend on operand and address widths.
634 // The Mod and R/M fields can encode a base for an effective address, or a
635 // register. These are separated into two fields here.
637 EADisplacement eaDisplacement
;
638 // The reg field always encodes a register
642 SIBIndex sibIndexBase
;
647 // Embedded rounding control.
650 ArrayRef
<OperandSpecifier
> operands
;
653 /// Decode one instruction and store the decoding results in
654 /// a buffer provided by the consumer.
655 /// \param insn The buffer to store the instruction in. Allocated by the
657 /// \param reader The byteReader_t for the bytes to be read.
658 /// \param readerArg An argument to pass to the reader for storing context
659 /// specific to the consumer. May be NULL.
660 /// \param logger The dlog_t to be used in printing status messages from the
661 /// disassembler. May be NULL.
662 /// \param loggerArg An argument to pass to the logger for storing context
663 /// specific to the logger. May be NULL.
664 /// \param startLoc The address (in the reader's address space) of the first
665 /// byte in the instruction.
666 /// \param mode The mode (16-bit, 32-bit, 64-bit) to decode in.
667 /// \return Nonzero if there was an error during decode, 0 otherwise.
668 int decodeInstruction(InternalInstruction
*insn
,
670 const void *readerArg
,
675 DisassemblerMode mode
);
677 /// Print a message to debugs()
678 /// \param file The name of the file printing the debug message.
679 /// \param line The line number that printed the debug message.
680 /// \param s The message to print.
681 void Debug(const char *file
, unsigned line
, const char *s
);
683 StringRef
GetInstrName(unsigned Opcode
, const void *mii
);
685 } // namespace X86Disassembler