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 LLVM_LIB_TARGET_X86_DISASSEMBLER_X86DISASSEMBLERDECODER_H
17 #define LLVM_LIB_TARGET_X86_DISASSEMBLER_X86DISASSEMBLERDECODER_H
19 #include "X86DisassemblerDecoderCommon.h"
20 #include "llvm/ADT/ArrayRef.h"
23 namespace X86Disassembler {
25 // Accessor functions for various fields of an Intel instruction
26 #define modFromModRM(modRM) (((modRM) & 0xc0) >> 6)
27 #define regFromModRM(modRM) (((modRM) & 0x38) >> 3)
28 #define rmFromModRM(modRM) ((modRM) & 0x7)
29 #define scaleFromSIB(sib) (((sib) & 0xc0) >> 6)
30 #define indexFromSIB(sib) (((sib) & 0x38) >> 3)
31 #define baseFromSIB(sib) ((sib) & 0x7)
32 #define wFromREX(rex) (((rex) & 0x8) >> 3)
33 #define rFromREX(rex) (((rex) & 0x4) >> 2)
34 #define xFromREX(rex) (((rex) & 0x2) >> 1)
35 #define bFromREX(rex) ((rex) & 0x1)
37 #define rFromEVEX2of4(evex) (((~(evex)) & 0x80) >> 7)
38 #define xFromEVEX2of4(evex) (((~(evex)) & 0x40) >> 6)
39 #define bFromEVEX2of4(evex) (((~(evex)) & 0x20) >> 5)
40 #define r2FromEVEX2of4(evex) (((~(evex)) & 0x10) >> 4)
41 #define mmFromEVEX2of4(evex) ((evex) & 0x3)
42 #define wFromEVEX3of4(evex) (((evex) & 0x80) >> 7)
43 #define vvvvFromEVEX3of4(evex) (((~(evex)) & 0x78) >> 3)
44 #define ppFromEVEX3of4(evex) ((evex) & 0x3)
45 #define zFromEVEX4of4(evex) (((evex) & 0x80) >> 7)
46 #define l2FromEVEX4of4(evex) (((evex) & 0x40) >> 6)
47 #define lFromEVEX4of4(evex) (((evex) & 0x20) >> 5)
48 #define bFromEVEX4of4(evex) (((evex) & 0x10) >> 4)
49 #define v2FromEVEX4of4(evex) (((~evex) & 0x8) >> 3)
50 #define aaaFromEVEX4of4(evex) ((evex) & 0x7)
52 #define rFromVEX2of3(vex) (((~(vex)) & 0x80) >> 7)
53 #define xFromVEX2of3(vex) (((~(vex)) & 0x40) >> 6)
54 #define bFromVEX2of3(vex) (((~(vex)) & 0x20) >> 5)
55 #define mmmmmFromVEX2of3(vex) ((vex) & 0x1f)
56 #define wFromVEX3of3(vex) (((vex) & 0x80) >> 7)
57 #define vvvvFromVEX3of3(vex) (((~(vex)) & 0x78) >> 3)
58 #define lFromVEX3of3(vex) (((vex) & 0x4) >> 2)
59 #define ppFromVEX3of3(vex) ((vex) & 0x3)
61 #define rFromVEX2of2(vex) (((~(vex)) & 0x80) >> 7)
62 #define vvvvFromVEX2of2(vex) (((~(vex)) & 0x78) >> 3)
63 #define lFromVEX2of2(vex) (((vex) & 0x4) >> 2)
64 #define ppFromVEX2of2(vex) ((vex) & 0x3)
66 #define rFromXOP2of3(xop) (((~(xop)) & 0x80) >> 7)
67 #define xFromXOP2of3(xop) (((~(xop)) & 0x40) >> 6)
68 #define bFromXOP2of3(xop) (((~(xop)) & 0x20) >> 5)
69 #define mmmmmFromXOP2of3(xop) ((xop) & 0x1f)
70 #define wFromXOP3of3(xop) (((xop) & 0x80) >> 7)
71 #define vvvvFromXOP3of3(vex) (((~(vex)) & 0x78) >> 3)
72 #define lFromXOP3of3(xop) (((xop) & 0x4) >> 2)
73 #define ppFromXOP3of3(xop) ((xop) & 0x3)
75 // These enums represent Intel registers for use by the decoder.
98 #define EA_BASES_16BIT \
134 #define EA_BASES_32BIT \
170 #define EA_BASES_64BIT \
328 #define REGS_SEGMENT \
346 #define REGS_CONTROL \
364 #define ALL_EA_BASES \
369 #define ALL_SIB_BASES \
388 /// \brief All possible values of the base field for effective-address
389 /// computations, a.k.a. the Mod and R/M fields of the ModR/M byte.
390 /// We distinguish between bases (EA_BASE_*) and registers that just happen
391 /// to be referred to when Mod == 0b11 (EA_REG_*).
394 #define ENTRY(x) EA_BASE_##x,
397 #define ENTRY(x) EA_REG_##x,
403 /// \brief All possible values of the SIB index field.
404 /// borrows entries from ALL_EA_BASES with the special case that
405 /// sib is synonymous with NONE.
406 /// Vector SIB: index can be XMM or YMM.
409 #define ENTRY(x) SIB_INDEX_##x,
418 /// \brief All possible values of the SIB base field.
421 #define ENTRY(x) SIB_BASE_##x,
427 /// \brief Possible displacement types for effective-address computations.
435 /// \brief All possible values of the reg field in the ModR/M byte.
437 #define ENTRY(x) MODRM_REG_##x,
443 /// \brief All possible segment overrides.
444 enum SegmentOverride {
455 /// \brief Possible values for the VEX.m-mmmm field
456 enum VEXLeadingOpcodeByte {
463 XOP_MAP_SELECT_8 = 0x8,
464 XOP_MAP_SELECT_9 = 0x9,
465 XOP_MAP_SELECT_A = 0xA
468 /// \brief Possible values for the VEX.pp/EVEX.pp field
470 VEX_PREFIX_NONE = 0x0,
476 enum VectorExtensionType {
477 TYPE_NO_VEX_XOP = 0x0,
484 /// \brief Type for the byte reader that the consumer must provide to
485 /// the decoder. Reads a single byte from the instruction's address space.
486 /// \param arg A baton that the consumer can associate with any internal
487 /// state that it needs.
488 /// \param byte A pointer to a single byte in memory that should be set to
489 /// contain the value at address.
490 /// \param address The address in the instruction's address space that should
492 /// \return -1 if the byte cannot be read for any reason; 0 otherwise.
493 typedef int (*byteReader_t)(const void *arg, uint8_t *byte, uint64_t address);
495 /// \brief Type for the logging function that the consumer can provide to
496 /// get debugging output from the decoder.
497 /// \param arg A baton that the consumer can associate with any internal
498 /// state that it needs.
499 /// \param log A string that contains the message. Will be reused after
500 /// the logger returns.
501 typedef void (*dlog_t)(void *arg, const char *log);
503 /// The specification for how to extract and interpret a full instruction and
505 struct InstructionSpecifier {
509 /// The x86 internal instruction, which is produced by the decoder.
510 struct InternalInstruction {
511 // Reader interface (C)
513 // Opaque value passed to the reader
514 const void* readerArg;
515 // The address of the next byte to read via the reader
516 uint64_t readerCursor;
518 // Logger interface (C)
520 // Opaque value passed to the logger
523 // General instruction information
525 // The mode to disassemble for (64-bit, protected, real)
526 DisassemblerMode mode;
527 // The start of the instruction, usable with the reader
528 uint64_t startLocation;
529 // The length of the instruction, in bytes
534 // 1 if the prefix byte corresponding to the entry is present; 0 if not
535 uint8_t prefixPresent[0x100];
536 // contains the location (for use with the reader) of the prefix byte
537 uint64_t prefixLocations[0x100];
538 // The value of the vector extension prefix(EVEX/VEX/XOP), if present
539 uint8_t vectorExtensionPrefix[4];
540 // The type of the vector extension prefix
541 VectorExtensionType vectorExtensionType;
542 // The value of the REX prefix, if present
544 // The location where a mandatory prefix would have to be (i.e., right before
545 // the opcode, or right before the REX prefix if one is present).
546 uint64_t necessaryPrefixLocation;
547 // The segment override type
548 SegmentOverride segmentOverride;
549 // 1 if the prefix byte, 0xf2 or 0xf3 is xacquire or xrelease
550 bool xAcquireRelease;
552 // Sizes of various critical pieces of data, in bytes
553 uint8_t registerSize;
555 uint8_t displacementSize;
556 uint8_t immediateSize;
558 // Offsets from the start of the instruction to the pieces of data, which is
559 // needed to find relocation entries for adding symbolic operands.
560 uint8_t displacementOffset;
561 uint8_t immediateOffset;
565 // The last byte of the opcode, not counting any ModR/M extension
567 // The ModR/M byte of the instruction, if it is an opcode extension
568 uint8_t modRMExtension;
572 // The type of opcode, used for indexing into the array of decode tables
573 OpcodeType opcodeType;
574 // The instruction ID, extracted from the decode table
575 uint16_t instructionID;
576 // The specifier for the instruction, from the instruction info table
577 const InstructionSpecifier *spec;
579 // state for additional bytes, consumed during operand decode. Pattern:
580 // consumed___ indicates that the byte was already consumed and does not
581 // need to be consumed again.
583 // The VEX.vvvv field, which contains a third register operand for some AVX
587 // The writemask for AVX-512 instructions which is contained in EVEX.aaa
590 // The ModR/M byte, which contains most register operands and some portion of
591 // all memory operands.
595 // The SIB byte, used for more complex 32- or 64-bit memory operands
599 // The displacement, used for memory operands
600 bool consumedDisplacement;
601 int32_t displacement;
603 // Immediates. There can be two in some cases
604 uint8_t numImmediatesConsumed;
605 uint8_t numImmediatesTranslated;
606 uint64_t immediates[2];
608 // A register or immediate operand encoded into the opcode
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 ArrayRef<OperandSpecifier> operands;
634 /// \brief 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.
649 int decodeInstruction(InternalInstruction *insn,
651 const void *readerArg,
656 DisassemblerMode mode);
658 /// \brief Print a message to debugs()
659 /// \param file The name of the file printing the debug message.
660 /// \param line The line number that printed the debug message.
661 /// \param s The message to print.
662 void Debug(const char *file, unsigned line, const char *s);
664 const char *GetInstrName(unsigned Opcode, const void *mii);
666 } // namespace X86Disassembler