1
0
mirror of https://github.com/bitdefender/bddisasm.git synced 2024-11-29 19:08:06 +00:00
bddisasm/inc/bddisasm.h

1553 lines
70 KiB
C
Raw Normal View History

2020-07-21 08:19:18 +00:00
/*
* Copyright (c) 2020 Bitdefender
* SPDX-License-Identifier: Apache-2.0
*/
#ifndef BDDISASM_H
#define BDDISASM_H
2020-07-21 08:19:18 +00:00
#include "disasmstatus.h"
#include "registers.h"
#include "constants.h"
#include "cpuidflags.h"
#include "version.h"
#ifdef _MSC_VER
#pragma warning(push)
#pragma warning(disable: 4214) // Bitfield in type other than int.
#pragma warning(disable: 4201) // Nonstandard extension used: nameless struct/union.
#endif
//
// Preferred vendor; the disassembler will try to be smart and disassemble as much as it can, but if there are
// encoding conflicts, than an alternate vendor can be selected. Note that this has effect only on conflicting
// encodings.
//
#define ND_VEND_ANY 0 // Generic decode, include any vendor.
#define ND_VEND_INTEL 1 // Prefer Intel.
#define ND_VEND_AMD 2 // Prefer AMD.
#define ND_VEND_GEODE 3 // Prefer Geode.
#define ND_VEND_CYRIX 4 // Prefer Cyrix.
//
// These control what instructions should be decoded if they map onto the wide NOP space (0F 1A and 0F 1B). Those are
// tricky, because they might be NOP if the feature is disabled, but might be something else (even #UD) if the feature
// is enabled. Ergo, we allow the user to select whether said feature is on or off, so that he controls whether he
// sees the NOPs or the MPX/CET/CLDEMOTE/etc. instructions instead.
//
#define ND_FEAT_NONE 0x00 // No feature/mode enabled.
#define ND_FEAT_MPX 0x01 // MPX support enabled.
#define ND_FEAT_CET 0x02 // CET support enabled.
#define ND_FEAT_CLDEMOTE 0x04 // CLDEMOTE support enabled.
#define ND_FEAT_ALL 0xFF // Decode as if all features are enabled. This is default.
2020-07-21 08:19:18 +00:00
//
// Code type
//
#define ND_CODE_16 0 // 16 bit decode mode.
#define ND_CODE_32 1 // 32 bit decode mode.
#define ND_CODE_64 2 // 64 bit decode mode.
//
// Data type
//
#define ND_DATA_16 0 // 16 bit data size.
#define ND_DATA_32 1 // 32 bit data size.
#define ND_DATA_64 2 // 64 bit data size.
//
// Stack type
//
#define ND_STACK_16 0 // 16 bit stack size.
#define ND_STACK_32 1 // 32 bit stack size.
#define ND_STACK_64 2 // 64 bit stack size.
//
// Addressing mode
//
#define ND_ADDR_16 0 // 16 bit addressing.
#define ND_ADDR_32 1 // 32 bit addressing.
#define ND_ADDR_64 2 // 64 bit addressing.
//
// Operand mode/size
//
#define ND_OPSZ_16 0 // 16 bit operand size.
#define ND_OPSZ_32 1 // 32 bit operand size.
#define ND_OPSZ_64 2 // 64 bit operand size.
//
// Vector mode/size
//
#define ND_VECM_128 0 // 128 bit vector size.
#define ND_VECM_256 1 // 256 bit vector size.
#define ND_VECM_512 2 // 512 bit vector size.
//
// Encoding mode
//
#define ND_ENCM_LEGACY 0 // Legacy encoded instruction.
#define ND_ENCM_XOP 1 // XOP encoded instruction.
#define ND_ENCM_VEX 2 // VEX (bot 2B or 3B) encoded instruction.
#define ND_ENCM_EVEX 3 // EVEX encoded instruction.
//
// VEX prefixes
//
#define ND_VEXM_2B 0 // 2B VEX prefix (0xC5).
#define ND_VEXM_3B 1 // 3B VEX prefix (0xC4).
//
// Size definitions
//
#define ND_SIZE_8BIT 1 // 1 byte.
#define ND_SIZE_16BIT 2 // 1 word or 2 bytes.
#define ND_SIZE_32BIT 4 // 1 double word or 4 bytes.
#define ND_SIZE_48BIT 6 // 1 fword or 6 bytes.
#define ND_SIZE_64BIT 8 // 1 qword or 8 bytes.
#define ND_SIZE_80BIT 10 // 1 fpu word or 10 bytes.
#define ND_SIZE_112BIT 14 // FPU environment, 14 bytes.
#define ND_SIZE_128BIT 16 // 1 xmm word or 16 bytes.
#define ND_SIZE_224BIT 28 // FPU environment, 28 bytes.
#define ND_SIZE_256BIT 32 // 1 ymm word or 32 bytes.
#define ND_SIZE_512BIT 64 // 1 zmm word or 64 bytes.
#define ND_SIZE_752BIT 94 // FPU state, 94 bytes.
#define ND_SIZE_864BIT 108 // FPU state, 108 bytes.
#define ND_SIZE_4096BIT 512 // Extended state, 512 bytes.
#define ND_SIZE_1KB 1024 // Tile register, 1KB.
#define ND_SIZE_CACHE_LINE 0xFFFFFFFE // The size of a cache line.
#define ND_SIZE_UNKNOWN 0xFFFFFFFF // Unknown/invalid size.
typedef uint32_t ND_OPERAND_SIZE;
typedef uint32_t ND_REG_SIZE;
//
// Prefix definitions
//
#define ND_PREFIX_G0_LOCK 0xF0 // LOCK prefix.
#define ND_PREFIX_G1_REPNE_REPNZ 0xF2 // REPNE/REPNZ prefix.
#define ND_PREFIX_G1_XACQUIRE 0xF2 // XACQUIRE prefix.
#define ND_PREFIX_G1_REPE_REPZ 0xF3 // REPE/REPZ prefix.
#define ND_PREFIX_G1_XRELEASE 0xF3 // XRELEASE prefix.
#define ND_PREFIX_G1_BND 0xF2 // BND prefix.
#define ND_PREFIX_G2_SEG_CS 0x2E // CS segment override.
#define ND_PREFIX_G2_SEG_SS 0x36 // SS segment override.
#define ND_PREFIX_G2_SEG_DS 0x3E // DS segment override.
#define ND_PREFIX_G2_SEG_ES 0x26 // ES segment override.
#define ND_PREFIX_G2_SEG_FS 0x64 // FS segment override.
#define ND_PREFIX_G2_SEG_GS 0x65 // GS segment override.
#define ND_PREFIX_G2_BR_NOT_TAKEN 0x2E // Branch not taken hint.
#define ND_PREFIX_G2_BR_TAKEN 0x3E // Branch taken hint.
#define ND_PREFIX_G2_BR_ALT 0x64 // Alternating branch hint.
#define ND_PREFIX_G2_NO_TRACK 0x3E // Do not track prefix.
#define ND_PREFIX_G3_OPERAND_SIZE 0x66 // Operand size override.
#define ND_PREFIX_G4_ADDR_SIZE 0x67 // Address size override.
#define ND_PREFIX_REX_MIN 0x40 // First REX prefix.
#define ND_PREFIX_REX_MAX 0x4F // Last REX prefix.
#define ND_PREFIX_VEX_2B 0xC5 // 2B VEX prefix.
#define ND_PREFIX_VEX_3B 0xC4 // 3B VEX prefix.
#define ND_PREFIX_XOP 0x8F // XOP prefix.
#define ND_PREFIX_EVEX 0x62 // EVEX prefix.
//
// Instruction flags
//
#define ND_FLAG_MODRM 0x00000001 // The instruction has modrm.
#define ND_FLAG_F64 0x00000002 // The operand is forced to 64 bit. Size changing prefix 0x66 is ignored.
#define ND_FLAG_D64 0x00000004 // The default operand size is 64 bit. Size changing prefix 0x66
// changes the size to 16 bit. No 32 bit version can be encoded.
#define ND_FLAG_O64 0x00000008 // The instruction is available only in 64 bit mode.
#define ND_FLAG_I64 0x00000010 // The instruction is invalid in 64 bit mode.
#define ND_FLAG_COND 0x00000020 // The instruction has a condition code encoded in low 4 bits of the opcode.
#define ND_FLAG_SSE_CONDB 0x00000040 // The instruction has a SSE condition byte.
#define ND_FLAG_VSIB 0x00000080 // The instruction uses VSIB addressing mode.
#define ND_FLAG_MIB 0x00000100 // The instruction used MIB addressing mode.
#define ND_FLAG_LIG 0x00000200 // L flag inside XOP/VEX/EVEX is ignored.
#define ND_FLAG_WIG 0x00000400 // W flag inside XOP/VEX/EVEX is ignored.
#define ND_FLAG_3DNOW 0x00000800 // The instruction is 3DNow!. The opcode is the last byte.
#define ND_FLAG_LOCK_SPECIAL 0x00001000 // MOV to/from CR in 16/32 bit, on AMD, extend the access to
// high 8 CRs via the LOCK prefix.
#define ND_FLAG_MMASK 0x00002000 // Mandatory mask. The instruction does not accept k0 as mask.
#define ND_FLAG_NOMZ 0x00004000 // Zeroing is not permitted when accessing memory.
#define ND_FLAG_NOL0 0x00008000 // SSE 128 encoding is not permitted for the instruction.
#define ND_FLAG_NOA16 0x00010000 // The instruction does not support 16 bit addressing.
#define ND_FLAG_MFR 0x00020000 // The Mod inside Mod R/M is forced to reg. No SIB/disp present.
#define ND_FLAG_VECTOR 0x00040000 // The instruction is a SIMD instruction that operates on vector regs.
#define ND_FLAG_S66 0x00080000 // Special flag for mandatory 0x66 prefix that actually changes
// the default op length.
#define ND_FLAG_BITBASE 0x00100000 // The instruction uses bitbase addressing mode.
#define ND_FLAG_AG 0x00200000 // The instruction is an address generator; no actual memory access.
#define ND_FLAG_SHS 0x00400000 // The instruction does a shadow stack access.
#define ND_FLAG_CETT 0x00800000 // The instruction is CET tracked.
#define ND_FLAG_SERIAL 0x01000000 // The instruction is serializing.
#define ND_FLAG_NO_RIP_REL 0x02000000 // The instruction doesn't work with RIP relative addressing.
#define ND_FLAG_NO66 0x04000000 // The 0x66 prefix is not accepted by the instruction.
#define ND_FLAG_SIBMEM 0x08000000 // sibmem addressing is used (Intel AMX instructions).
#define ND_FLAG_I67 0x10000000 // Ignore the 0x67 prefix in 64 bit mode (Intel MPX instructions).
#define ND_FLAG_IER 0x20000000 // Ignore EVEX embedded rounding.
2020-07-21 08:19:18 +00:00
//
// Accepted prefixes map
//
#define ND_PREF_REP 0x0001 // The instruction supports REP prefix.
#define ND_PREF_REPC 0x0002 // The instruction supports REPZ/REPNZ prefixes.
#define ND_PREF_LOCK 0x0004 // The instruction supports LOCK prefix.
#define ND_PREF_HLE 0x0008 // The instruction supports XACQUIRE/XRELEASE prefixes.
#define ND_PREF_XACQUIRE 0x0010 // The instruction supports only XACQUIRE.
#define ND_PREF_XRELEASE 0x0020 // The instruction supports only XRELEASE.
#define ND_PREF_BND 0x0040 // The instruction supports BND prefix.
#define ND_PREF_BHINT 0x0080 // The instruction supports branch hints.
#define ND_PREF_HLE_WO_LOCK 0x0100 // HLE prefix is accepted without LOCK.
#define ND_PREF_DNT 0x0200 // The instruction supports the DNT (Do Not Track) CET prefix.
//
// Accepted decorators map. These are stored per-instruction. There are also per-operand indicators for
// each decorator.
//
#define ND_DECO_ER 0x01 // Embedded rounding is accepted.
#define ND_DECO_SAE 0x02 // Suppress all Exceptions is accepted.
#define ND_DECO_ZERO 0x04 // Zeroing is accepted.
#define ND_DECO_MASK 0x08 // Masking is accepted.
#define ND_DECO_BROADCAST 0x10 // Memory broadcast is accepted.
//
// Operand access flags.
//
#define ND_ACCESS_NONE 0x00 // The operand is not accessed.
#define ND_ACCESS_READ 0x01 // The operand is read.
#define ND_ACCESS_WRITE 0x02 // The operand is written.
#define ND_ACCESS_COND_READ 0x04 // The operand is read only if some conditions are met.
#define ND_ACCESS_COND_WRITE 0x08 // The operand is written only if some conditions are met (ie: CMOVcc).
2020-07-21 08:19:18 +00:00
#define ND_ACCESS_ANY_READ (ND_ACCESS_READ | ND_ACCESS_COND_READ) // Any read mask.
#define ND_ACCESS_ANY_WRITE (ND_ACCESS_WRITE | ND_ACCESS_COND_WRITE) // Any write mask.
#define ND_ACCESS_PREFETCH 0x10 // The operand is prefetched.
2020-07-21 08:19:18 +00:00
//
// Predicate/condition definitions.
//
#define ND_COND_OVERFLOW 0x0 // OF
#define ND_COND_CARRY 0x2 // CF
#define ND_COND_BELOW 0x2 // CF
#define ND_COND_NOT_ABOVE_OR_EQUAL 0x2 // CF
#define ND_COND_ZERO 0x4 // ZF
#define ND_COND_EQUAL 0x4 // ZF
#define ND_COND_BELOW_OR_EQUAL 0x6 // CF | ZF
#define ND_COND_NOT_ABOVE 0x6 // CF | ZF
#define ND_COND_SIGN 0x8 // SF
#define ND_COND_PARITY 0xA // PF
#define ND_COND_LESS 0xC // SF ^ OF
#define ND_COND_LESS_OR_EQUAL 0xE // (SF ^ OF) | ZF
#define ND_COND_NOT(p) ((p) | 0x1) // Negates the predicate.
// Kept for backwards compatibility - renamed to ND_COND*.
#define ND_PRED_OVERFLOW 0x0 // OF
#define ND_PRED_CARRY 0x2 // CF
#define ND_PRED_BELOW 0x2 // CF
#define ND_PRED_NOT_ABOVE_OR_EQUAL 0x2 // CF
#define ND_PRED_ZERO 0x4 // ZF
#define ND_PRED_EQUAL 0x4 // ZF
#define ND_PRED_BELOW_OR_EQUAL 0x6 // CF | ZF
#define ND_PRED_NOT_ABOVE 0x6 // CF | ZF
#define ND_PRED_SIGN 0x8 // SF
#define ND_PRED_PARITY 0xA // PF
#define ND_PRED_LESS 0xC // SF ^ OF
#define ND_PRED_LESS_OR_EQUAL 0xE // (SF ^ OF) | ZF
#define ND_PRED_NOT(p) ((p) | 0x1) // Negates the predicate.
//
// Condition code definitions. These apply to condition codes encoded in instruction bytes as used by some SSE/AVX
// instructions.
//
#define ND_SSE_COND_EQ 0x00 // Equal.
#define ND_SSE_COND_LT 0x01 // Less-than.
#define ND_SSE_COND_LE 0x02 // Less-than or equal.
#define ND_SSE_COND_UNORD 0x03 // Unordered.
#define ND_SSE_COND_FALSE1 0x03 // False.
#define ND_SSE_COND_NEQ 0x04 // Not equal.
#define ND_SSE_COND_NLT 0x05 // Not less-than.
#define ND_SSE_COND_NLE 0x06 // Not less-than or equal.
#define ND_SSE_COND_ORD 0x07 // Ordered.
#define ND_SSE_COND_TRUE1 0x07 // True.
#define ND_SSE_COND_EQ_UQ 0x08 // Equal.
#define ND_SSE_COND_NGE 0x09 // Not greater-than or equal.
#define ND_SSE_COND_NGT 0x0A // Not greater-than.
#define ND_SSE_COND_FALSE 0x0B // False.
#define ND_SSE_COND_NEQ_OQ 0x0C // Not equal.
#define ND_SSE_COND_GE 0x0D // Greater-than or equal.
#define ND_SSE_COND_GT 0x0E // Greater-than.
#define ND_SSE_COND_TRUE 0x0F // True.
#define ND_SSE_COND_EQ_OS 0x10 // Equal.
#define ND_SSE_COND_LT_OQ 0x11 // Less-than.
#define ND_SSE_COND_LE_OQ 0x12 // Less-than or equal.
#define ND_SSE_COND_UNORD_S 0x13 // Unordered.
#define ND_SSE_COND_NEQ_US 0x14 // Not equal.
#define ND_SSE_COND_NLT_UQ 0x15 // Not less-than.
#define ND_SSE_COND_NLE_UQ 0x16 // Not less-than or equal.
#define ND_SSE_COND_ORD_S 0x17 // Ordered.
#define ND_SSE_COND_EQ_US 0x18 // Equal.
#define ND_SSE_COND_NGE_UQ 0x19 // Not greater-than or equal.
#define ND_SSE_COND_NGT_UQ 0x1A // Not greater-than.
#define ND_SSE_COND_FALSE_OS 0x1B // False.
#define ND_SSE_COND_NEQ_OS 0x1C // Not equal.
#define ND_SSE_COND_GE_OQ 0x1D // Greater-than or equal.
#define ND_SSE_COND_GT_OQ 0x1E // Greater-than.
#define ND_SSE_COND_TRUE_US 0x1F // True.
//
// Valid CPU modes.
//
// Group 1: ring
#define ND_MOD_R0 0x0001 // Instruction valid in ring 0.
#define ND_MOD_R1 0x0002 // Instruction valid in ring 1.
#define ND_MOD_R2 0x0004 // Instruction valid in ring 2.
#define ND_MOD_R3 0x0008 // Instruction valid in ring 3.
// Group 2: operating mode.
#define ND_MOD_REAL 0x0010 // Instruction valid in real mode.
#define ND_MOD_V8086 0x0020 // Instruction valid in virtual 8086 mode.
#define ND_MOD_PROT 0x0040 // Instruction valid in protected mode.
#define ND_MOD_COMPAT 0x0080 // Instruction valid in compatibility mode.
#define ND_MOD_LONG 0x0100 // Instruction valid in long mode.
#define ND_MOD_SMM 0x0200 // Instruction valid in System-Management Mode.
// Group 3: SGX and TSX
#define ND_MOD_SGX 0x0400 // Instruction valid in SGX enclaves.
#define ND_MOD_TSX 0x0800 // Instruction valid in TSX transactional regions.
// Group 4: VMX
#define ND_MOD_VMXR 0x1000 // Instruction valid in VMX Root mode.
#define ND_MOD_VMXN 0x2000 // Instruction valid in VMX non-root mode.
#define ND_MOD_VMXO 0x4000 // Instruction valid outside VMX operation.
#define ND_MOD_RING_MASK 0x000F // Valid ring mask.
#define ND_MOD_MODE_MASK 0x03F0 // Valid mode mask.
#define ND_MOD_OTHER_MASK 0x0C00 // Misc mask.
#define ND_MOD_VMX_MASK 0x7000 // VMX mask.
// For instructions valid in any operating mode.
#define ND_MOD_ANY 0xFFFF // Instruction valid in any mode.
//
// Misc constants
//
#define ND_MAX_INSTRUCTION_LENGTH 15 // 15 bytes is the maximum instruction length supported by the x86 arch.
#define ND_MAX_MNEMONIC_LENGTH 32 // Should do for now.
#define ND_MIN_BUF_SIZE 128 // Textual disassembly minimal buffer size.
#define ND_MAX_OPERAND 10 // No more than 10 operands/instruction, but I'm generous.
#define ND_MAX_REGISTER_SIZE 64 // Maximum register size - 64 bytes.
#define ND_MAX_GPR_REGS 16 // Max number of GPRs.
#define ND_MAX_SEG_REGS 8 // Max number of segment registers.
#define ND_MAX_FPU_REGS 8 // Max number of FPU registers.
#define ND_MAX_MMX_REGS 8 // Max number of MMX registers.
#define ND_MAX_SSE_REGS 32 // Max number of SSE registers.
#define ND_MAX_CR_REGS 16 // Max number of control registers.
#define ND_MAX_DR_REGS 16 // Max number of debug registers.
#define ND_MAX_TR_REGS 16 // Max number of test registers.
#define ND_MAX_MSK_REGS 8 // Max number of mask registers.
#define ND_MAX_BND_REGS 4 // Max number of bound registers.
#define ND_MAX_SYS_REGS 8 // Max number of system registers.
#define ND_MAX_X87_REGS 8 // Max number of x87 state/control registers registers.
#define ND_MAX_TILE_REGS 8 // Max number of tile registers.
//
// Misc macros.
//
// Sign extend 8 bit to 64 bit.
#define ND_SIGN_EX_8(x) ((x) & 0x00000080 ? 0xFFFFFFFFFFFFFF00 | (x) : (x))
// Sign extend 16 bit to 64 bit.
#define ND_SIGN_EX_16(x) ((x) & 0x00008000 ? 0xFFFFFFFFFFFF0000 | (x) : (x))
// Sign extend 32 bit to 64 bit.
#define ND_SIGN_EX_32(x) ((x) & 0x80000000 ? 0xFFFFFFFF00000000 | (x) : (x))
// Wrapper for for ND_SIGN_EX_8/ND_SIGN_EX_16/ND_SIGN_EX_32. Sign extend sz bytes to 64 bits.
#define ND_SIGN_EX(sz, x) ((sz) == 1 ? ND_SIGN_EX_8(x) : (sz) == 2 ? ND_SIGN_EX_16(x) : \
(sz) == 4 ? ND_SIGN_EX_32(x) : (x))
// Trim 64 bits to sz bytes.
#define ND_TRIM(sz, x) ((sz) == 1 ? (x) & 0xFF : (sz) == 2 ? (x) & 0xFFFF : \
(sz) == 4 ? (x) & 0xFFFFFFFF : (x))
// Returns most significant bit, given size in bytes sz.
#define ND_MSB(sz, x) ((sz) == 1 ? ((x) >> 7) & 1 : (sz) == 2 ? ((x) >> 15) & 1 : \
(sz) == 4 ? ((x) >> 31) & 1 : ((x) >> 63) & 1)
// Returns least significant bit.
#define ND_LSB(sz, x) ((x) & 1)
// Convert a size in bytes to a bitmask.
#define ND_SIZE_TO_MASK(sz) (((sz) < 8) ? ((1ULL << ((sz) * 8)) - 1) : (0xFFFFFFFFFFFFFFFF))
// Get bit at position bit from x.
#define ND_GET_BIT(bit, x) (((x) >> (bit)) & 1)
// Return the sign of sz bytes long value x.
#define ND_GET_SIGN(sz, x) ND_MSB(sz, x)
// Sets the sign of the sz bytes long value x.
#define ND_SET_SIGN(sz, x) ND_SIGN_EX(sz, x)
#ifdef BIG_ENDIAN
#define ND_FETCH_64(b) ((uint64_t)FETCH_uint32_t((char *)b) | ((uint64_t)FETCH_uint32_t((char *)b + 4) << 32))
#define ND_FETCH_32(b) ((uint32_t)FETCH_WORD((char *)b) | ((uint32_t)FETCH_WORD((char *)b + 2) << 16))
#define ND_FETCH_16(b) ((((char *)b)[0]) | (((char *)b)[1] << 8))
#define ND_FETCH_8(b) (*((char *)b))
#else
#define ND_FETCH_64(b) (*((uint64_t *)(b)))
#define ND_FETCH_32(b) (*((uint32_t *)(b)))
#define ND_FETCH_16(b) (*((uint16_t *)(b)))
#define ND_FETCH_8(b) (*((uint8_t *)(b)))
#endif
//
// Helper macros which simply test the presence of various ND_FLAG_* in the instruction attributes.
//
#define ND_IS_3DNOW(ix) (!!((ix)->Attributes & ND_FLAG_3DNOW))
#define ND_HAS_PREDICATE(ix) (!!((ix)->Attributes & ND_FLAG_COND))
#define ND_HAS_CONDITION(ix) (!!((ix)->Attributes & ND_FLAG_COND))
#define ND_HAS_SSE_CONDITION(ix) (!!((ix)->Attributes & ND_FLAG_SSE_CONDB))
#define ND_HAS_MODRM(ix) (!!((ix)->Attributes & ND_FLAG_MODRM))
#define ND_HAS_VSIB(ix) (!!((ix)->Attributes & ND_FLAG_VSIB))
#define ND_HAS_MIB(ix) (!!((ix)->Attributes & ND_FLAG_MIB))
#define ND_HAS_VECTOR(ix) (!!((ix)->Attributes & ND_FLAG_VECTOR))
#define ND_HAS_BITBASE(ix) (!!((ix)->Attributes & ND_FLAG_BITBASE))
#define ND_HAS_AG(ix) (!!((ix)->Attributes & ND_FLAG_AG))
#define ND_HAS_SIBMEM(ix) (!!((ix)->Attributes & ND_FLAG_SIBMEM))
#define ND_HAS_SHS(ix) (!!((ix)->Attributes & ND_FLAG_SHS))
#define ND_HAS_CETT(ix) (!!((ix)->Attributes & ND_FLAG_CETT))
//
// Supported prefixes macros.
//
#define ND_REP_SUPPORT(ix) (!!((ix)->ValidPrefixes.Rep))
#define ND_REPC_SUPPORT(ix) (!!((ix)->ValidPrefixes.RepCond))
#define ND_LOCK_SUPPORT(ix) (!!((ix)->ValidPrefixes.Lock))
#define ND_HLE_SUPPORT(ix) (!!((ix)->ValidPrefixes.Hle))
#define ND_XACQUIRE_SUPPORT(ix) (!!((ix)->ValidPrefixes.Xacquire))
#define ND_XRELEASE_SUPPORT(ix) (!!((ix)->ValidPrefixes.Xrelease))
#define ND_BND_SUPPORT(ix) (!!((ix)->ValidPrefixes.Bnd))
#define ND_BHINT_SUPPORT(ix) (!!((ix)->ValidPrefixes.Bhint))
#define ND_DNT_SUPPORT(ix) (!!((ix)->ValidPrefixes.Dnt))
//
// Decorators map macros.
//
#define ND_DECORATOR_SUPPORT(ix) ((ix)->ValidDecorators.Raw != 0)
#define ND_MASK_SUPPORT(ix) (!!((ix)->ValidDecorators.Mask))
#define ND_ZERO_SUPPORT(ix) (!!((ix)->ValidDecorators.Zero))
#define ND_ER_SUPPORT(ix) (!!((ix)->ValidDecorators.Er))
#define ND_SAE_SUPPORT(ix) (!!((ix)->ValidDecorators.Sae))
#define ND_BROADCAST_SUPPORT(ix) (!!((ix)->ValidDecorators.Broadcast))
// Generates a unique ID per register type, size and reg.
#define ND_OP_REG_ID(op) (((op)->Type << 24) | ((op)->Info.Register.Type << 16) | \
((op)->Info.Register.Size << 8) | ((op)->Info.Register.Reg))
// Example: ND_IS_OP_REG(op, ND_REG_GPR, 4, REG_ESP)
// Example: ND_IS_OP_REG(op, ND_REG_CR, 8, REG_CR3)
// Example: ND_IS_OP_REG(op, ND_REG_RIP, 8, 0)
// Checks if the indicated operand op is a register of type t, with size s and index r.
#define ND_IS_OP_REG(op, t, s, r) ((uint32_t)(ND_OP_REG_ID(op)) == (uint32_t)((ND_OP_REG << 24)|(t << 16)|(s << 8)|(r)))
// Checjs if the indicated operand is the stack.
#define ND_IS_OP_STACK(op) ((op)->Type == ND_OP_MEM && (op)->Info.Memory.IsStack)
//
// Operand types.
//
typedef enum _ND_OPERAND_TYPE
{
ND_OP_NOT_PRESENT, // Indicates the absence of any operand.
ND_OP_REG, // The operand is a register.
ND_OP_MEM, // The operand is located in memory.
ND_OP_IMM, // The operand is an immediate.
ND_OP_OFFS, // The operand is a relative offset.
ND_OP_ADDR, // The operand is an absolute address, in the form seg:offset.
ND_OP_CONST, // The operand is an implicit constant.
ND_OP_BANK, // An entire bank/set of registers are being accessed. Used in PUSHA/POPA/XSAVE/LOADALL.
} ND_OPERAND_TYPE;
//
// Register types.
//
typedef enum _ND_REG_TYPE
{
ND_REG_NOT_PRESENT,
ND_REG_GPR, // The register is a 8/16/32/64 bit general purpose register.
ND_REG_SEG, // The register is a segment register.
ND_REG_FPU, // The register is a 80-bit FPU register.
ND_REG_MMX, // The register is a 64-bit MMX register.
ND_REG_SSE, // The register is a 128/256/512 bit SSE vector register.
ND_REG_CR, // The register is a control register.
ND_REG_DR, // The register is a debug register.
ND_REG_TR, // The register is a test register.
ND_REG_BND, // The register is a bound register.
ND_REG_MSK, // The register is a mask register.
ND_REG_TILE, // The register is a tile register.
ND_REG_MSR, // The register is a model specific register.
ND_REG_XCR, // The register is a extended control register.
ND_REG_SYS, // The register is a system register.
ND_REG_X87, // The register is a x87 status/control register.
ND_REG_MXCSR, // The register is the MXCSR register.
ND_REG_PKRU, // The register is the PKRU register.
ND_REG_SSP, // The register is the SSP (Shadow Stack Pointer) register.
ND_REG_FLG, // The register is the FLAGS register.
ND_REG_RIP, // The register is the instruction pointer register.
} ND_REG_TYPE;
//
// Operand encoding types.
//
typedef enum _ND_OPERAND_ENCODING
{
ND_OPE_NP, // No encoding present.
ND_OPE_R, // Operand encoded in modrm.reg.
ND_OPE_M, // Operand encoded in modrm.rm.
ND_OPE_V, // Operand encoded in Xop/Vex/Evex/Mvex.(v')vvvv
ND_OPE_D, // Operand is encoded inside subsequent instruction bytes.
ND_OPE_O, // Operand is encoded in low 3 bit of the opcode.
ND_OPE_I, // Operand is an immediate.
ND_OPE_C, // Operand is CL.
ND_OPE_1, // Operand is 1.
ND_OPE_L, // Operand is reg encoded in immediate.
ND_OPE_A, // Operand is encoded in Evex.aaa.
ND_OPE_E, // Operand is a MSR or XCR encoded in ECX register.
ND_OPE_S, // Operand is implicit/suppressed. Not encoded anywhere.
} ND_OPERAND_ENCODING;
//
// Instruction tuple type; used to determine compressed displacement size for disp8 EVEX instructions. Note that
// most of the EVEX encoded instructions use the compressed displacement addressing scheme.
//
typedef enum _ND_TUPLE
{
ND_TUPLE_None,
ND_TUPLE_FV, // Full Vector
ND_TUPLE_HV, // Half Vector
ND_TUPLE_T1S8, // Tuple1 scalar, size 8 bit
ND_TUPLE_T1S16, // Tuple1 scalar, size 16 bit
ND_TUPLE_T1S, // Tuple1 scalar, size 32/64 bit
ND_TUPLE_T1F, // Tuple1 float, size 32/64 bit
ND_TUPLE_T2, // Tuple2, 64/128 bit
ND_TUPLE_T4, // Tuple4, 128/256 bit
ND_TUPLE_T8, // Tuple8, 256 bit
ND_TUPLE_FVM, // Full Vector Memory
ND_TUPLE_HVM, // Half Vector Memory
ND_TUPLE_QVM, // Quarter Vector Memory
ND_TUPLE_OVM, // Oct Vector Memory
ND_TUPLE_M128, // M128, 128 bit
ND_TUPLE_DUP, // DUP (VMOVDDUP)
ND_TUPLE_T1_4X, // 4 x 32 bit Memory Elements are referenced
} ND_TUPLE;
//
// EVEX rounding control.
//
typedef enum _ND_ROUNDING
{
ND_RND_RNE, // Round to nearest equal.
ND_RND_RD, // Round down.
ND_RND_RU, // Round up.
ND_RND_RZ, // round to zero.
} ND_ROUNDING;
//
// Exception classes. Different instruction sets or encodings are covered by different exception classes.
//
typedef enum _ND_EX_CLASS
{
ND_EXC_None,
ND_EXC_SSE_AVX, // SSE/AVX exception class (for legacy encoded SSE instructions and VEX instructions).
ND_EXC_EVEX, // EVEX exception class (for EVEX encoded AVX* instructions).
ND_EXC_OPMASK, // Opmask instructions exception class.
ND_EXC_AMX, // AMX exception class type (for VEX encoded AMX instructions).
} ND_EX_CLASS;
//
// SSE/AVX exception types. Covers VEX instructions as well, except for AMX instructions.
//
typedef enum _ND_EX_TYPE_SSE_AVX
{
ND_EXT_SSE_AVX_None,
ND_EXT_1,
ND_EXT_2,
ND_EXT_3,
ND_EXT_4,
ND_EXT_5,
ND_EXT_6,
ND_EXT_7,
ND_EXT_8,
ND_EXT_9,
ND_EXT_10,
ND_EXT_11,
ND_EXT_12,
ND_EXT_13,
} ND_EX_TYPE_SSE_AVX;
//
// EVEX exception types.
//
typedef enum _ND_EX_TYPE_EVEX
{
ND_EXT_EVEX_None,
ND_EXT_E1,
ND_EXT_E1NF,
ND_EXT_E2,
ND_EXT_E3,
ND_EXT_E3NF,
ND_EXT_E4,
ND_EXT_E4nb,
ND_EXT_E4NF,
ND_EXT_E4NFnb,
ND_EXT_E5,
ND_EXT_E5NF,
ND_EXT_E6,
ND_EXT_E6NF,
ND_EXT_E7NM,
ND_EXT_E9,
ND_EXT_E9NF,
ND_EXT_E10,
ND_EXT_E10NF,
ND_EXT_E11,
ND_EXT_E12,
ND_EXT_E12NP,
} ND_EX_TYPE_EVEX;
//
// Opmask exception types.
//
typedef enum _ND_EX_TYPE_OPMASK
{
ND_EXT_OPMASK_None,
ND_EXT_K20,
ND_EXT_K21,
} ND_EX_TYPE_OPMASK;
//
// AMX exception types.
//
typedef enum _ND_EX_TYPE_AMX
{
ND_EXT_AMX_None,
ND_EXT_AMX_E1,
ND_EXT_AMX_E2,
ND_EXT_AMX_E3,
ND_EXT_AMX_E4,
ND_EXT_AMX_E5,
ND_EXT_AMX_E6,
} ND_EX_TYPE_AMX;
//
// Operands access map. Contains every register except for MSR & XCR, includes memory, flags, RIP, stack.
// Use NdGetFullAccessMap to populate this structure.
//
typedef struct _ND_ACCESS_MAP
{
uint8_t RipAccess;
uint8_t FlagsAccess;
uint8_t StackAccess;
uint8_t MemAccess;
uint8_t MxcsrAccess;
uint8_t PkruAccess;
uint8_t SspAccess;
uint8_t GprAccess[ND_MAX_GPR_REGS];
uint8_t SegAccess[ND_MAX_SEG_REGS];
uint8_t FpuAccess[ND_MAX_FPU_REGS];
uint8_t MmxAccess[ND_MAX_MMX_REGS];
uint8_t SseAccess[ND_MAX_SSE_REGS];
uint8_t CrAccess [ND_MAX_CR_REGS ];
uint8_t DrAccess [ND_MAX_DR_REGS ];
uint8_t TrAccess [ND_MAX_TR_REGS ];
uint8_t BndAccess[ND_MAX_BND_REGS];
uint8_t MskAccess[ND_MAX_MSK_REGS];
uint8_t SysAccess[ND_MAX_SYS_REGS];
uint8_t X87Access[ND_MAX_X87_REGS];
} ND_ACCESS_MAP, *PND_ACCESS_MAP;
//
// Operand access mode.
//
typedef union _ND_OPERAND_ACCESS
{
uint8_t Access;
struct
{
uint8_t Read : 1; // The operand is read.
uint8_t Write : 1; // The operand is written.
uint8_t CondRead : 1; // The operand is read only under some conditions.
uint8_t CondWrite : 1; // The operand is written only under some conditions.
uint8_t Prefetch : 1; // The operand is prefetched.
2020-07-21 08:19:18 +00:00
};
} ND_OPERAND_ACCESS;
//
// Operand flags.
//
typedef union _ND_OPERAND_FLAGS
{
uint8_t Flags;
struct
{
uint8_t IsDefault : 1; // 1 if the operand is default. This also applies to implicit ops.
uint8_t SignExtendedOp1 : 1; // 1 if the operand is sign extended to the first operands' size.
uint8_t SignExtendedDws : 1; // 1 if the operand is sign extended to the default word size.
};
} ND_OPERAND_FLAGS;
//
// Constant operand.
//
typedef struct _ND_OPDESC_CONSTANT
{
uint64_t Const; // Instruction constant, ie ROL reg, 1.
} ND_OPDESC_CONSTANT;
//
// Immediate operand.
//
typedef struct _ND_OPDESC_IMMEDIATE
{
uint64_t Imm; // Immediate. Only Size bytes are valid. The rest are undefined.
} ND_OPDESC_IMMEDIATE;
//
// Relative offset operand.
//
typedef struct _ND_OPDESC_REL_OFFSET
{
uint64_t Rel; // Relative offset (relative to the current RIP). Sign extended.
} ND_OPDESC_RELOFFSET;
//
// Describes a register operand. Count registers are used starting with Reg.
//
typedef struct _ND_OPDESC_REGISTER
{
ND_REG_TYPE Type; // The register type. See enum ND_REG_TYPE.
ND_REG_SIZE Size; // Indicates the register size. This may not be equal to the Size
// field, as a smaller amount of data may be processed from a
// register (especially if we have a SSE register or a mask register).
// Also note that as of now, 64 bytes is the maximum register size.
uint32_t Reg; // The register number/ID.
uint32_t Count; // The number of registers accessed, starting with Reg.
bool IsHigh8:1; // TRUE if this is AH, CH, DH or BH register.
bool IsBlock:1; // TRUE if this is a block register addressing.
} ND_OPDESC_REGISTER;
//
// Describes an address operand.
//
typedef struct _ND_OPDESC_ADDRESS
{
// Size is the size of the address. Usually 4 (16 bit mode) or 6 (32 bit mode) or 10 (64 bit).
uint16_t BaseSeg; // Base segment selector of the address.
uint64_t Offset; // Offset inside the segment.
} ND_OPDESC_ADDRESS;
//
// Shadow stack access types.
//
typedef enum _ND_SHSTK_ACCESS
{
ND_SHSTK_NONE = 0,
ND_SHSTK_EXPLICIT, // Explicit memory operand accessed as shadow stack.
ND_SHSTK_SSP_LD_ST, // Shadow Stack Pointer (SSP) used as base for addressing using conventional load/store.
ND_SHSTK_SSP_PUSH_POP, // Shadow Stack Pointer (SSP) used as base for addressing using push/pop.
ND_SHSTK_PL0_SSP, // Privilege 0 SSP (IA32_PL0_SSP) used (SETSSBSY).
} ND_SHSTK_ACCESS;
2020-07-21 08:19:18 +00:00
//
// Describes a memory operand.
//
typedef struct _ND_OPDESC_MEMORY
{
bool HasSeg:1; // TRUE if segment is present & used.
bool HasBase:1; // TRUE if base register is present.
bool HasIndex:1; // TRUE if index & scale are present.
bool HasDisp:1; // TRUE if displacement is present.
bool HasCompDisp:1; // TRUE if compressed disp8 is used (EVEX instructions).
bool HasBroadcast:1; // TRUE if the memory operand is a broadcast operand.
bool IsRipRel:1; // TRUE if this is a rip-relative addressing. Base, Index, Scale are
// all ignored.
bool IsStack:1; // TRUE if this is a stack op. Note that explicit stack accesses are not
// included (eg: mov eax, [rsp] will NOT set IsStack).
bool IsString:1; // TRUE for [RSI] and [RDI] operands inside string operations.
bool IsShadowStack:1; // TRUE if this is a shadow stack access. Check out ShStkType for more info.
2020-07-21 08:19:18 +00:00
bool IsDirect:1; // TRUE if direct addressing (MOV [...], EAX, 0xA3).
bool IsBitbase:1; // TRUE if this is a bit base. Used for BT* instructions. The bitbase
// stored in the second operand must be added to the linear address.
bool IsAG:1; // TRUE if the memory operand is address generation and no mem access is
// made.
bool IsMib:1; // TRUE if MIB addressing is used (MPX instructions).
bool IsVsib:1; // TRUE if the index register selects a vector register.
bool IsSibMem:1; // TRUE if the addressing uses sibmem (AMX instructions).
ND_REG_SIZE BaseSize; // Base size, in bytes. Max 8 bytes.
ND_REG_SIZE IndexSize; // Ditto for index size. Max 8 bytes.
uint8_t DispSize; // Displacement size. Max 4 bytes.
uint8_t CompDispSize; // Compressed displacement size - 1, 2, 4, 8, 16, 32, 64.
uint8_t ShStkType; // Shadow stack access type. Check out ND_SHSTK_ACCESS.
2020-07-21 08:19:18 +00:00
struct
{
uint8_t IndexSize; // VSIB index size.
uint8_t ElemSize; // VSIB element size.
uint8_t ElemCount; // Number of elements scattered/gathered/prefetched.
} Vsib;
uint8_t Seg; // Base segment used to address the memory. 0 = es, 1 = cs, etc.
uint8_t Base; // Base register. Can only be a GPR.
uint8_t Index; // Index register. Can be a vector reg (ZMM0-ZMM31).
uint8_t Scale; // Scale: 1, 2, 4 or 8. Always present if Index is present.
uint64_t Disp; // Sign extended displacement.
} ND_OPDESC_MEMORY;
//
// Describes a decorator.
//
typedef struct _ND_OPERAND_DECORATOR
{
bool HasMask:1; // TRUE if mask is present, 0 otherwise.
bool HasZero:1; // TRUE if zeroing will be made, 0 if merging will be made.
bool HasBroadcast:1; // TRUE if broadcasting is being made. Valid only for memory operands.
// These are used only to indicate where the SAE and ER decorators should be placed in the disassembly.
// Otherwise, SAE and ER are global, per instruction, and don't apply to a single operand.
bool HasSae:1; // TRUE if SAE is present.
bool HasEr:1; // TRUE if ER is present.
// Mask register specifier.
struct
{
uint8_t Msk; // Mask register used. Only k0-k7 can be encoded.
} Mask;
// Broadcast specifier
struct
{
uint8_t Count; // Number of times to broadcast the element.
uint8_t Size; // Size of one element.
} Broadcast;
} ND_OPERAND_DECORATOR;
//
// Extended operand information.
//
typedef struct _ND_OPERAND
{
ND_OPERAND_TYPE Type; // Operand type. One of ND_OPERAND_TYPE enumerations.
ND_OPERAND_ENCODING Encoding; // Where is the operand encoded - modrm.reg, modrm.rm, etc.
ND_OPERAND_SIZE Size; // Operand size in bytes. This should be used when operating with
// the operand. It includes sign-extension or zero-extensions.
// Note that the Size field indicates the actual amount of data
// used for processing. If the operand type is a register, it MAY NOT
// indicate the register size. Use the Info.Register.Size
// field to get the actual register size.
ND_OPERAND_SIZE RawSize; // Raw size inside the instruction. This will usually be identical
// to Size; however, some instructions force the actual size of their
// operands to 64 bit; PUSH/POP or branches are good examples; although
// the raw size of the relative offset or the immediate will be RawSize,
// internally, the CPU will use Size (usually sign-extended).
ND_OPERAND_ACCESS Access; // Access mode (read, write, read-write, etc.)
ND_OPERAND_FLAGS Flags; // Misc operand flags.
// Depending in the Type field, one of these subfields contains information about the operand.
union
{
ND_OPDESC_CONSTANT Constant; // Constant, if operand type is ND_OP_CONST.
ND_OPDESC_IMMEDIATE Immediate; // Immediate, if operand type is ND_OP_IMM.
ND_OPDESC_REGISTER Register; // Register, if operand type if ND_OP_REG.
ND_OPDESC_RELOFFSET RelativeOffset; // Relative offset, if operand type is ND_OP_REL_OFFS.
ND_OPDESC_ADDRESS Address; // Address, if operand type is ND_OP_ADDR.
ND_OPDESC_MEMORY Memory; // Memory, if operand type is ND_OP_MEM.
} Info;
ND_OPERAND_DECORATOR Decorator; // Decorator information.
} ND_OPERAND, *PND_OPERAND;
//
// REX prefix.
//
typedef union _ND_REX
{
uint8_t Rex;
struct
{
uint8_t b : 1; // b (rm or low opcode) extension field.
uint8_t x : 1; // x (index) extension field.
uint8_t r : 1; // r (reg) extension field.
uint8_t w : 1; // w (size) extension field. Promotes to 64 bit.
};
} ND_REX;
//
// Mod R/M byte.
//
typedef union _ND_MODRM
{
uint8_t ModRm;
struct
{
uint8_t rm : 3; // rm field.
uint8_t reg : 3; // reg field.
uint8_t mod : 2; // mod field. Indicates memory access (0, 1 or 2), or register access (3).
};
} ND_MODRM;
//
// SIB byte.
//
typedef union _ND_SIB
{
uint8_t Sib;
struct
{
uint8_t base : 3; // Base register.
uint8_t index : 3; // Index register.
uint8_t scale : 2; // Scale.
};
} ND_SIB;
//
// DREX byte. Exactly as AMD defined it. No actual instructions exist that use DREX encoding.
//
typedef union _ND_DREX
{
uint8_t Drex;
struct
{
uint8_t b : 1;
uint8_t x : 1;
uint8_t r : 1;
uint8_t oc0 : 1;
uint8_t vd : 3;
uint8_t d : 1;
};
} ND_DREX;
//
// 2-bytes VEX. Exactly as Intel defined it.
//
typedef union _ND_VEX2
{
uint8_t Vex[2];
struct
{
uint8_t op; // 0xC5
uint8_t p : 2; // p0, p1
uint8_t l : 1; // L
uint8_t v : 4; // ~v0, ~v1, ~v2, ~v3
uint8_t r : 1; // ~R
};
} ND_VEX2;
//
// 3-bytes VEX. Exactly as Intel defined it.
//
typedef union _ND_VEX3
{
uint8_t Vex[3];
struct
{
uint8_t op; // 0xC4
uint8_t m : 5; // m0, m1, m2, m3, m4
uint8_t b : 1; // ~B
uint8_t x : 1; // ~X
uint8_t r : 1; // ~R
uint8_t p : 2; // p0, p1
uint8_t l : 1; // L
uint8_t v : 4; // ~v0, ~v1, ~v2, ~v3
uint8_t w : 1; // W
};
} ND_VEX3;
//
// XOP. Exactly as AMD defined it.
//
typedef union _ND_XOP
{
uint8_t Xop[3];
struct
{
uint8_t op; // 0x8F
uint8_t m : 5; // m0, m1, m2, m3, m4
uint8_t b : 1; // ~B
uint8_t x : 1; // ~X
uint8_t r : 1; // ~R
uint8_t p : 2; // p0, p1
uint8_t l : 1; // L
uint8_t v : 4; // ~v0, ~v1, ~v2, ~v3
uint8_t w : 1; // W
};
} ND_XOP;
//
// EVEX prefix. Exactly as Intel defined it.
//
typedef union _ND_EVEX
{
uint8_t Evex[4];
struct
{
uint8_t op; // 0x62
uint8_t m : 2; // m0, m1
uint8_t zero : 2; // 00
uint8_t rp : 1; // ~R'
uint8_t b : 1; // ~B
uint8_t x : 1; // ~X
uint8_t r : 1; // ~R
uint8_t p : 2; // p0, p1
uint8_t one : 1; // 1
uint8_t v : 4; // ~v0, ~v1, ~v2, ~v3
uint8_t w : 1; // W
uint8_t a : 3; // a0, a1, a2
uint8_t vp : 1; // ~V'
uint8_t bm : 1; // b
uint8_t l : 2; // L'L
uint8_t z : 1; // z
};
} ND_EVEX;
//
// Describes the CPUID leaf, sub-leaf, register & bit that indicate whether an instruction is supported or not.
// If Leaf == ND_CFF_NO_LEAF, the instruction is supported on any CPU, and no CPUID flag exists.
// If SubLeaf == ND_CFF_NO_SUBLEAF, there is no subleaf to check.
//
typedef union _ND_CPUID_FLAG
{
uint64_t Flag;
struct
{
uint32_t Leaf; // CPUID leaf. ND_CFF_NO_LEAF if not applicable.
uint32_t SubLeaf : 24; // CPUID sub-leaf. ND_CFF_NO_SUBLEAF if not applicable.
uint32_t Reg : 3; // The register that contains info regarding the instruction.
uint32_t Bit : 5; // Bit inside the register that indicates whether the instruction is present.
};
} ND_CPUID_FLAG;
//
// Each instruction may accept one or more prefixes. This structure indicates which prefixes are valid for the
// given instruction.
//
typedef union _ND_VALID_PREFIXES
{
uint16_t Raw;
struct
{
uint16_t Rep : 1; // The instruction supports REP prefix.
uint16_t RepCond : 1; // The instruction supports REPZ/REPNZ prefixes.
uint16_t Lock : 1; // The instruction supports LOCK prefix.
uint16_t Hle : 1; // The instruction supports XACQUIRE/XRELEASE prefixes.
uint16_t Xacquire : 1; // The instruction supports only XACQUIRE.
uint16_t Xrelease : 1; // The instruction supports only XRELEASE.
uint16_t Bnd : 1; // The instruction supports BND prefix.
uint16_t Bhint : 1; // The instruction supports branch hints.
uint16_t HleNoLock : 1; // HLE prefix is accepted without LOCK.
uint16_t Dnt : 1; // The instruction supports the DNT (Do Not Track) CET prefix.
};
} ND_VALID_PREFIXES, *PND_VALID_PREFIXES;
//
// Each instruction may accept several decorators. This instruction indicates which decorators are valid for the
// given instruction.
//
typedef union _ND_VALID_DECORATORS
{
uint8_t Raw;
struct
{
uint8_t Er : 1; // The instruction supports embedded rounding mode.
uint8_t Sae : 1; // The instruction supports suppress all exceptions mode.
uint8_t Zero : 1; // The instruction supports zeroing.
uint8_t Mask : 1; // The instruction supports mask registers.
uint8_t Broadcast : 1; // The instruction supports broadcast.
};
} ND_VALID_DECORATORS, *PND_VALID_DECORATORS;
//
// Each instruction is valid or invalid in any certain mode. This indicates which modes the instruction is valid in.
// If the bit is set, the isntruction is valid in that given mode.
//
typedef union _ND_VALID_MODES
{
uint16_t Raw;
struct
{
// Group 1: privilege level.
uint16_t Ring0 : 1; // The instruction is valid in ring 0.
uint16_t Ring1 : 1; // The instruction is valid in ring 1.
uint16_t Ring2 : 1; // The instruction is valid in ring 2.
uint16_t Ring3 : 1; // The instruction is valid in ring 3.
// Group 2: operating mode - the CPU can be on only one of these modes at any moment.
uint16_t Real : 1; // The instruction is valid in real mode.
uint16_t V8086 : 1; // The instruction is valid in Virtual 8086 mode.
uint16_t Protected : 1; // The instruction is valid in protected mode (32 bit).
uint16_t Compat : 1; // The instruction is valid in compatibility mode (32 bit in 64 bit).
uint16_t Long : 1; // The instruction is valid in long mode.
uint16_t Smm : 1; // The instruction is valid in System Management Mode.
// Group 3: special modes - these may be active inside other modes (example: TSX in Long mode).
uint16_t Sgx : 1; // The instruction is valid in SGX mode.
uint16_t Tsx : 1; // The instruction is valid in transactional regions.
// Group 4: VMX mode - they engulf all the other modes.
uint16_t VmxRoot : 1; // The instruction is valid in VMX root mode.
uint16_t VmxNonRoot : 1;// The instruction is valid in VMX non root mode.
uint16_t VmxOff : 1; // The instruction is valid outside VMX operation.
};
} ND_VALID_MODES, *PND_VALID_MODES;
//
// RFLAGS register. This structure reflects the actual position of each flag insdide the RFLAGS register, so it can
// be used for direct processing.
//
typedef union _ND_RFLAGS
{
uint32_t Raw;
struct
{
uint32_t CF : 1; // Carry flag.
uint32_t Reserved1 : 1; // Reserved, must be 1.
uint32_t PF : 1; // Parity flag.
uint32_t Reserved2 : 1; // Reserved.
uint32_t AF : 1; // Auxiliary flag.
uint32_t Reserved3 : 1; // Reserved.
uint32_t ZF : 1; // Zero flag.
uint32_t SF : 1; // Sign flag.
uint32_t TF : 1; // Trap flag.
uint32_t IF : 1; // Interrupt flag.
uint32_t DF : 1; // Direction flag.
uint32_t OF : 1; // Overflow flag.
uint32_t IOPL : 2; // I/O privilege level flag.
uint32_t NT : 1; // Nested task flag.
uint32_t Reserved4 : 1; // Reserved.
uint32_t RF : 1; // Resume flag.
uint32_t VM : 1; // Virtual mode flag.
uint32_t AC : 1; // Alignment check flag.
uint32_t VIF : 1; // Virtual interrupts flag.
uint32_t VIP : 1; // Virtual interrupt pending flag.
uint32_t ID : 1; // CPUID identification flag.
};
} ND_RFLAGS, *PND_RFLAGS;
#define ND_FPU_FLAG_SET_0 0 // The FPU status flag is cleared to 0.
#define ND_FPU_FLAG_SET_1 1 // The FPU status flag is set to 1.
#define ND_FPU_FLAG_MODIFIED 2 // The FPU status flag is modified according to a result.
#define ND_FPU_FLAG_UNDEFINED 3 // The FPU status flag is undefined or unaffected.
//
// FPU status flags. Each status flag can be one of ND_FPU_FLAG*.
//
typedef struct _ND_FPU_FLAGS
{
uint8_t C0 : 2; // C0 flag access mode. See ND_FPU_FLAG_*.
uint8_t C1 : 2; // C1 flag access mode. See ND_FPU_FLAG_*.
uint8_t C2 : 2; // C2 flag access mode. See ND_FPU_FLAG_*.
uint8_t C3 : 2; // C3 flag access mode. See ND_FPU_FLAG_*.
} ND_FPU_FLAGS, *PND_FPU_FLAGS;
//
// Describes a decoded instruction. All the possible information about the instruction is contained in this structure.
// You don't have to call any other APIs to gather any more info about it.
//
typedef struct _INSTRUX
{
uint8_t DefCode:4; // ND_CODE_*. Indicates disassembly mode.
uint8_t DefData:4; // ND_DATA_*. Indicates default data size.
uint8_t DefStack:4; // ND_STACK_*. Indicates default stack pointer width.
uint8_t VendMode:4; // ND_VEND_*. Indicates vendor mode.
uint8_t FeatMode; // ND_FEAT_*. Indicates which features are enabled.
2020-07-21 08:19:18 +00:00
uint8_t EncMode:4; // ND_ENCM_*. Indicates encoding mode.
uint8_t VexMode:4; // ND_VEX_*. Indicates the VEX mode, if any.
uint8_t AddrMode:4; // ND_ADDR_*. Indicates addressing mode.
uint8_t OpMode:4; // ND_OPSZ_*. Indicates operand mode/size.
uint8_t EfOpMode:4; // ND_OPSZ_*. Indicates effective operand mode/size.
uint8_t VecMode:4; // ND_VECM_*. Indicates vector length.
uint8_t EfVecMode:4; // ND_VECM_*. Indicates effective vector length.
2020-07-21 08:19:18 +00:00
// Prefixes.
bool HasRex:1; // TRUE - REX is present.
bool HasVex:1; // TRUE - VEX is present.
bool HasXop:1; // TRUE - XOP is present.
bool HasEvex:1; // TRUE - EVEX is present.
bool HasMvex:1; // TRUE - MVEX is present.
bool HasOpSize:1; // TRUE - 0x66 present.
bool HasAddrSize:1; // TRUE - 0x67 present.
bool HasLock:1; // TRUE - 0xF0 present.
bool HasRepnzXacquireBnd:1; // TRUE - 0xF2 present.
bool HasRepRepzXrelease:1; // TRUE - 0xF3 present.
bool HasSeg:1; // TRUE - segment override present.
// Indicators for prefix activation.
bool IsRepeated:1; // TRUE - the instruction is REPed up to RCX times.
bool IsXacquireEnabled:1; // TRUE - the instruction is XACQUIRE enabled.
bool IsXreleaseEnabled:1; // TRUE - the instruction is XRELEASE enabled.
bool IsRipRelative:1; // TRUE - the instruction uses RIP relative addressing.
bool IsCetTracked:1; // TRUE - this is an indirect CALL/JMP that is CET tracked.
// Instruction chunks.
bool HasModRm:1; // TRUE - we have valid MODRM.
bool HasSib:1; // TRUE - we have valid SIB.
bool HasDrex:1; // TRUE - we have valid DREX.
bool HasDisp:1; // TRUE - the instruction has displacement.
bool HasAddr:1; // TRUE - the instruction contains a direct address (ie, CALL far 0x9A)
bool HasMoffset:1; // TRUE - the instruction contains a moffset (ie, MOV al, [mem], 0xA0)
bool HasImm1:1; // TRUE - immediate present.
bool HasImm2:1; // TRUE - second immediate present.
bool HasImm3:1; // TRUE - third immediate present.
bool HasRelOffs:1; // TRUE - the instruction contains a relative offset (ie, Jcc 0x7x).
bool HasSseImm:1; // TRUE - SSE immediate that encodes additional registers is present.
bool HasCompDisp:1; // TRUE - the instruction uses compressed displacement
bool HasBroadcast:1; // TRUE - the instruction uses broadcast addressing
bool HasMask:1; // TRUE - the instruction has mask.
bool HasZero:1; // TRUE - the instruction uses zeroing.
bool HasEr:1; // TRUE - the instruction has embedded rounding.
bool HasSae:1; // TRUE - the instruction has SAE.
bool SignDisp:1; // Displacement sign. 0 is positive, 1 is negative.
// Encoding specifics.
bool HasMandatory66:1; // 0x66 is mandatory prefix. Does not behave as size-changing prefix.
bool HasMandatoryF2:1; // 0x66 is mandatory prefix. Does not behave as REP prefix.
bool HasMandatoryF3:1; // 0x66 is mandatory prefix. Does not behave as REP prefix.
// Instruction components lengths. Will be 0 if the given field is not present.
uint8_t Length; // 1-15 valid. Instructions longer than 15 bytes will cause #GP.
uint8_t WordLength:4; // The length of the instruction word. 2, 4 or 8.
uint8_t PrefLength:4; // The total number of bytes consumed by prefixes. This will also be
// the offset to the first opcode. The primary opcode will always be
// the last one, so at offset PrefixesLength + OpcodeLength - 1
uint8_t OpLength:4; // Number of opcode bytes. Max 3.
uint8_t DispLength:4; // Displacement length, in bytes. Maximum 4.
uint8_t AddrLength:4; // Absolute address length, in bytes. Maximum 8 bytes.
uint8_t MoffsetLength:4; // Memory offset length, in bytes. Maximum 8 bytes.
uint8_t Imm1Length:4; // First immediate length, in bytes. Maximum 8 bytes.
uint8_t Imm2Length:4; // Second immediate length, in bytes. Maximum 8 bytes.
uint8_t Imm3Length:4; // Third immediate length, in bytes. Maximum 8 bytes.
uint8_t RelOffsLength:4; // Relative offset length, in bytes. Maximum 4 bytes.
// Instruction components offsets. Will be 0 if the given field is not present. Prefixes ALWAYS start at offset 0.
uint8_t OpOffset:4; // The offset of the first opcode, inside the instruction.
uint8_t MainOpOffset:4; // The offset of the nominal opcode, inside the instruction.
uint8_t DispOffset:4; // The offset of the displacement, inside the instruction
uint8_t AddrOffset:4; // The offset of the hard-coded address.
uint8_t MoffsetOffset:4; // The offset of the absolute address, inside the instruction
uint8_t Imm1Offset:4; // The offset of the immediate, inside the instruction
uint8_t Imm2Offset:4; // The offset of the second immediate, if any, inside the instruction
uint8_t Imm3Offset:4; // The offset of the third immediate, if any, inside the instruction
uint8_t RelOffsOffset:4; // The offset of the relative offset used in instruction.
uint8_t SseImmOffset:4; // The offset of the SSE immediate, if any, inside the instruction.
uint8_t ModRmOffset:4; // The offset of the mod rm byte inside the instruction, if any.
// If SIB is also present, it will always be at ModRmOffset + 1.
uint8_t StackWords; // Number of words accessed on/from the stack.
uint8_t Rep; // The last rep/repz/repnz prefix. 0 if none.
uint8_t Seg; // The last segment override prefix. 0 if none. FS/GS if 64 bit.
uint8_t Bhint; // The last segment override indicating a branch hint.
ND_REX Rex; // REX prefix.
ND_MODRM ModRm; // ModRM byte.
ND_SIB Sib; // SIB byte.
ND_DREX Drex; // DREX byte. Part of AMD SSE 5. Never shipped.
union
{
ND_VEX2 Vex2; // VEX 2 prefix.
ND_VEX3 Vex3; // VEX 3 prefix.
ND_XOP Xop; // XOP prefix.
ND_EVEX Evex; // EVEX prefix.
};
// This structures contains the fields extracted from either REX, XOP, VEX, EVEX or MVEX fields. They're globally
// placed here, in order to avoid testing for each kind of prefix each time. Instead, one can use the different
// fields directly from here, regardless the actual encoding mode.
struct
{
uint32_t w:1; // REX/XOP/VEX/EVEX/MVEX.W field
uint32_t r:1; // REX/XOP/VEX/EVEX/MVEX.R field (reg extension)
uint32_t x:1; // REX/XOP/VEX/EVEX/MVEX.X field (index extension)
uint32_t b:1; // REX/XOP/VEX/EVEX/MVEX.B field (base extension)
uint32_t rp:1; // EVEX/MVEX.R' (reg extension)
uint32_t p:2; // XOP/VEX/EVEX/MVEX.pp (embedded prefix)
uint32_t m:5; // XOP/VEX/EVEX/MVEX.mmmmm (decoding table)
uint32_t l:2; // XOP/VEX.L or EVEX.L'L (vector length)
uint32_t v:4; // XOP/VEX.VVVV or EVEX/MVEX.VVVV (additional operand)
uint32_t vp:1; // EVEX/MVEX.V' (vvvv extension)
uint32_t bm:1; // EVEX.b (broadcast)
uint32_t e:1; // MVEX.e (eviction hint)
uint32_t z:1; // EVEX.z (zero)
uint32_t k:3; // EVEX.aaa/MVEX.kkk (mask registers)
uint32_t s:3; // MVEX.sss (swizzle)
} Exs;
union
{
struct
{
uint32_t Ip;
uint16_t Cs;
};
} Address; // seg:offset address.
uint64_t Moffset; // Offset. Used by 'O' operands. It's an absolute address.
uint32_t Displacement; // Displacement. Max 4 bytes. Used in ModRM instructions.
uint32_t RelativeOffset; // Relative offset, used for branches. Max 4 bytes.
uint64_t Immediate1; // Can be 8 bytes on x64
uint8_t Immediate2; // For enter, mainly. Can only be 1 byte.
uint8_t Immediate3; // Third additional immediate. Limited to 1 byte for now.
uint8_t SseImmediate; // This immediate actually selects a register.
uint8_t SseCondition; // Condition code encoded in additional byte.
uint8_t Condition; // Condition code encoded in low 4 bit of the opcode.
uint8_t Predicate; // Same as Condition, kept for backwards compatibility.
uint8_t OperandsCount; // Number of operands.
uint8_t ExpOperandsCount; // Number of explicit operands. Use this if you want to ignore
// implicit operands such as stack, flags, etc.
uint16_t OperandsEncodingMap; // What parts of the instruction encode operands.
ND_OPERAND Operands[ND_MAX_OPERAND]; // Instruction operands.
// As extracted from the operands themselves.
uint8_t RipAccess; // RIP access mode (read/write).
uint8_t StackAccess; // Stack access mode (push/pop).
uint8_t MemoryAccess; // Memory access mode (read/write, including stack or shadow stack).
struct
{
uint8_t RegAccess; // RFLAGS access mode (read/write), as per the entire register.
ND_RFLAGS Tested; // Tested flags.
ND_RFLAGS Modified; // Modified (according to the result) flags.
ND_RFLAGS Set; // Flags that are always set to 1.
ND_RFLAGS Cleared; // Flags that are always cleared to 0.
ND_RFLAGS Undefined; // Undefined flags.
} FlagsAccess;
ND_FPU_FLAGS FpuFlagsAccess; // FPU status word C0-C3 bits access. Valid only for FPU instructions!
uint8_t ExceptionClass; // ND_EX_CLASS_TYPE, indicates the exception class type.
uint8_t ExceptionType; // Exception type. Depends on ExceptionClass.
uint8_t TupleType; // EVEX tuple type, if EVEX. Check out ND_TUPLE.
uint8_t RoundingMode; // EVEX rounding mode, if present. Check out ND_ROUNDING.
// Stored inside the instruction entry as well. These are specific for an instruction and do not depend on
// encoding. Use the flags definitions (ND_FLAG_*, ND_PREF_*, ND_DECO_*, ND_EXOP_*) to access specific bits.
uint32_t Attributes; // Instruction attributes/flags. A collection of ND_FLAG_*.
union
{
ND_INS_CLASS Instruction; // One of the ND_INS_*
ND_INS_CLASS Iclass; // Same as Instruction.
};
ND_INS_CATEGORY Category; // One of the ND_CAT_*
ND_INS_SET IsaSet; // One of the ND_SET_*
ND_CPUID_FLAG CpuidFlag; // CPUID support flag.
ND_VALID_MODES ValidModes; // Valid CPU modes for the instruction.
ND_VALID_PREFIXES ValidPrefixes; // Indicates which prefixes are valid for this instruction.
ND_VALID_DECORATORS ValidDecorators; // What decorators are accepted by the instruction.
uint8_t Reserved1[3]; // Padding purpose. Aligns the mnemonic to 8 bytes.
2020-07-21 08:19:18 +00:00
char Mnemonic[ND_MAX_MNEMONIC_LENGTH]; // Instruction mnemonic.
uint8_t OpCodeBytes[3]; // Opcode bytes - escape codes and main op code
uint8_t PrimaryOpCode; // Main/nominal opcode
uint32_t Reserved2; // Padding purpose. Aligns the InstructionBytes to 16 bytes.
uint8_t InstructionBytes[16]; // The entire instruction.
} INSTRUX, *PINSTRUX;
//
// Decoder context. Such a structure must be passed to the NdDecodeWithContext API. This structure must be initialized
// only once, and then it can be re-used across NdDecodeWithContext calls.
//
typedef struct _ND_CONTEXT
{
uint64_t DefCode : 4; // Decode mode - one of the ND_CODE_* values.
uint64_t DefData : 4; // Data mode - one of the ND_DATA_* values.
uint64_t DefStack : 4; // Stack mode - one of the ND_STACK_* values.
uint64_t VendMode : 4; // Prefered vendor - one of the ND_VEND_* values.
uint64_t FeatMode : 8; // Supported features mask. A combination of ND_FEAT_* values.
uint64_t Reserved : 40; // Reserved for future use.
} ND_CONTEXT;
//
// Returns the bddisasm version.
2020-07-21 08:19:18 +00:00
//
void
NdGetVersion(
uint32_t *Major,
uint32_t *Minor,
uint32_t *Revision,
char **BuildDate,
char **BuildTime
);
//
// Decode one instruction. Note that this is equivalent to:
// NdDecodeEx(Instrux, Code, ND_MAX_INSTRUCTION_LEN, DefCode, DefData).
// This version should be used if the caller doesn't care about the length of the buffer. Otherwise, use the other
// decode API.
//
2020-07-21 08:19:18 +00:00
NDSTATUS
NdDecode(
INSTRUX *Instrux, // Output decoded instruction.
const uint8_t *Code, // Buffer containing the instruction bytes.
uint8_t DefCode, // Decode mode - one of the ND_CODE_* values.
uint8_t DefData // Data mode - one of the ND_DATA_* value.
2020-07-21 08:19:18 +00:00
);
//
// Decode one instruction. Note that this is equivalent to:
// NdDecodeEx2(Instrux, Code, Size, DefCode, DefData, DefCode, ND_VEND_ANY).
// By default, the used vendor will be ND_VEND_ANY, so all instructions will be decoded.
// By default, the feature mode will be ND_FEAT_ALL, so all instructions will be decoded (but may yield error where
// otherwise a NOP would be encoded - use ND_FEAT_NONE in that case).
//
2020-07-21 08:19:18 +00:00
NDSTATUS
NdDecodeEx(
INSTRUX *Instrux, // Output decoded instruction.
const uint8_t *Code, // Buffer containing the instruction bytes.
size_t Size, // Maximum size of the Code buffer.
uint8_t DefCode, // Decode mode - one of the ND_CODE_* values.
uint8_t DefData // Data mode - one of the ND_DATA_* value.
2020-07-21 08:19:18 +00:00
);
//
// Fills a ND_CONTEXT structure, and calls NdDecodeWithContext. The feature mode will be ND_FEAT_ALL by default.
//
2020-07-21 08:19:18 +00:00
NDSTATUS
NdDecodeEx2(
INSTRUX *Instrux, // Output decoded instruction.
const uint8_t *Code, // Buffer containing the instruction bytes.
size_t Size, // Maximum size of the Code buffer.
uint8_t DefCode, // Decode mode - one of the ND_CODE_* values.
uint8_t DefData, // Data mode - one of the ND_DATA_* value.
uint8_t DefStack, // Stack mode - one of the ND_STACK_* values.
uint8_t PreferedVendor // Preferred vendor - one of the ND_VEND_* values.
2020-07-21 08:19:18 +00:00
);
//
// This API received a decode context, where it expects DefCode, DefData, DefStack, VendMode and FeatMode to be
// already initialized. The Context will not be modified by the decoder, so it can be reused across decode calls.
// The Context should initially be initialized using NdInitContext. This will ensure backwards compatibility
// by setting new fields to default values.
// Note that this is the base decoding API, and this ends up being called by all the other decoding APIs, after
// providing default arguments and filling them in the Context structure. For maximum speed, use this instead of
// the others.
//
NDSTATUS
NdDecodeWithContext(
INSTRUX *Instrux, // Output decoded instruction.
const uint8_t *Code, // Buffer containing the instruction bytes.
size_t Size, // Maximum size of the Code buffer.
ND_CONTEXT *Context // Context describing decode mode, vendor mode and supported features.
);
//
// Convert the given instruction into textual representation (Intel syntax).
//
2020-07-21 08:19:18 +00:00
NDSTATUS
NdToText(
const INSTRUX *Instrux,
uint64_t Rip,
uint32_t BufferSize,
char *Buffer
);
//
// Returns true if the instruction is RIP relative. Note that this function is kept for backwards compatibility, since
// there already is a IsRipRelative field inside INSTRUX.
//
2020-07-21 08:19:18 +00:00
bool
NdIsInstruxRipRelative(
const INSTRUX *Instrux
);
//
// Returns an access map that contains the access for each register.
//
2020-07-21 08:19:18 +00:00
NDSTATUS
NdGetFullAccessMap(
const INSTRUX *Instrux,
ND_ACCESS_MAP *AccessMap
);
//
// Initialize the decoder context.
//
void
NdInitContext(
ND_CONTEXT *Context
);
2020-07-21 08:19:18 +00:00
// #pragma warning(default: 4214) // Bitfield in type other than int.
// #pragma warning(default: 4201) // Nonstandard extension used: nameless struct/union.
#ifdef _MSC_VER
#pragma warning(pop)
#endif
#endif // BDDISASM_H