/*
* getID - Determines the ID of an instruction, consuming the ModR/M byte as
* appropriate for extended and escape opcodes. Determines the attributes and
* context for the instruction before doing so.
*
* @param insn - The instruction whose ID is to be determined.
* @return - 0 if the ModR/M could be read when needed or was not needed;
* nonzero otherwise.
*/
static int getID(struct InternalInstruction* insn, void *miiArg) {
uint8_t attrMask;
uint16_t instructionID;
dbgprintf(insn, "getID()");
attrMask = ATTR_NONE;
if (insn->mode == MODE_64BIT)
attrMask |= ATTR_64BIT;
if (insn->vexSize) {
attrMask |= ATTR_VEX;
if (insn->vexSize == 3) {
switch (ppFromVEX3of3(insn->vexPrefix[2])) {
case VEX_PREFIX_66:
attrMask |= ATTR_OPSIZE;
break;
case VEX_PREFIX_F3:
attrMask |= ATTR_XS;
break;
case VEX_PREFIX_F2:
attrMask |= ATTR_XD;
break;
}
if (lFromVEX3of3(insn->vexPrefix[2]))
attrMask |= ATTR_VEXL;
}
else if (insn->vexSize == 2) {
switch (ppFromVEX2of2(insn->vexPrefix[1])) {
case VEX_PREFIX_66:
attrMask |= ATTR_OPSIZE;
break;
case VEX_PREFIX_F3:
attrMask |= ATTR_XS;
break;
case VEX_PREFIX_F2:
attrMask |= ATTR_XD;
break;
}
if (lFromVEX2of2(insn->vexPrefix[1]))
attrMask |= ATTR_VEXL;
}
else {
return -1;
}
}
else {
if (isPrefixAtLocation(insn, 0x66, insn->necessaryPrefixLocation))
attrMask |= ATTR_OPSIZE;
else if (isPrefixAtLocation(insn, 0x67, insn->necessaryPrefixLocation))
attrMask |= ATTR_ADSIZE;
else if (isPrefixAtLocation(insn, 0xf3, insn->necessaryPrefixLocation))
attrMask |= ATTR_XS;
else if (isPrefixAtLocation(insn, 0xf2, insn->necessaryPrefixLocation))
attrMask |= ATTR_XD;
}
if (insn->rexPrefix & 0x08)
attrMask |= ATTR_REXW;
if (getIDWithAttrMask(&instructionID, insn, attrMask))
return -1;
/* The following clauses compensate for limitations of the tables. */
if ((attrMask & ATTR_VEXL) && (attrMask & ATTR_REXW) &&
!(attrMask & ATTR_OPSIZE)) {
/*
* Some VEX instructions ignore the L-bit, but use the W-bit. Normally L-bit
* has precedence since there are no L-bit with W-bit entries in the tables.
* So if the L-bit isn't significant we should use the W-bit instead.
* We only need to do this if the instruction doesn't specify OpSize since
* there is a VEX_L_W_OPSIZE table.
*/
const struct InstructionSpecifier *spec;
uint16_t instructionIDWithWBit;
const struct InstructionSpecifier *specWithWBit;
spec = specifierForUID(instructionID);
if (getIDWithAttrMask(&instructionIDWithWBit,
insn,
(attrMask & (~ATTR_VEXL)) | ATTR_REXW)) {
insn->instructionID = instructionID;
insn->spec = spec;
return 0;
}
specWithWBit = specifierForUID(instructionIDWithWBit);
if (instructionID != instructionIDWithWBit) {
insn->instructionID = instructionIDWithWBit;
insn->spec = specWithWBit;
} else {
insn->instructionID = instructionID;
insn->spec = spec;
}
return 0;
}
if (insn->prefixPresent[0x66] && !(attrMask & ATTR_OPSIZE)) {
/*
* The instruction tables make no distinction between instructions that
* allow OpSize anywhere (i.e., 16-bit operations) and that need it in a
* particular spot (i.e., many MMX operations). In general we're
* conservative, but in the specific case where OpSize is present but not
* in the right place we check if there's a 16-bit operation.
*/
const struct InstructionSpecifier *spec;
uint16_t instructionIDWithOpsize;
const char *specName, *specWithOpSizeName;
spec = specifierForUID(instructionID);
if (getIDWithAttrMask(&instructionIDWithOpsize,
insn,
attrMask | ATTR_OPSIZE)) {
/*
* ModRM required with OpSize but not present; give up and return version
* without OpSize set
*/
insn->instructionID = instructionID;
insn->spec = spec;
return 0;
}
specName = x86DisassemblerGetInstrName(instructionID, miiArg);
specWithOpSizeName =
x86DisassemblerGetInstrName(instructionIDWithOpsize, miiArg);
if (is16BitEquvalent(specName, specWithOpSizeName)) {
insn->instructionID = instructionIDWithOpsize;
insn->spec = specifierForUID(instructionIDWithOpsize);
} else {
insn->instructionID = instructionID;
insn->spec = spec;
}
return 0;
}
if (insn->opcodeType == ONEBYTE && insn->opcode == 0x90 &&
insn->rexPrefix & 0x01) {
/*
* NOOP shouldn't decode as NOOP if REX.b is set. Instead
* it should decode as XCHG %r8, %eax.
*/
const struct InstructionSpecifier *spec;
uint16_t instructionIDWithNewOpcode;
const struct InstructionSpecifier *specWithNewOpcode;
spec = specifierForUID(instructionID);
/* Borrow opcode from one of the other XCHGar opcodes */
insn->opcode = 0x91;
if (getIDWithAttrMask(&instructionIDWithNewOpcode,
insn,
attrMask)) {
insn->opcode = 0x90;
insn->instructionID = instructionID;
insn->spec = spec;
return 0;
}
specWithNewOpcode = specifierForUID(instructionIDWithNewOpcode);
/* Change back */
insn->opcode = 0x90;
insn->instructionID = instructionIDWithNewOpcode;
insn->spec = specWithNewOpcode;
return 0;
}
insn->instructionID = instructionID;
insn->spec = specifierForUID(insn->instructionID);
return 0;
}
/*
* getID - Determines the ID of an instruction, consuming the ModR/M byte as
* appropriate for extended and escape opcodes. Determines the attributes and
* context for the instruction before doing so.
*
* @param insn - The instruction whose ID is to be determined.
* @return - 0 if the ModR/M could be read when needed or was not needed;
* nonzero otherwise.
*/
static int getID(struct InternalInstruction* insn) {
uint8_t attrMask;
uint16_t instructionID;
dbgprintf(insn, "getID()");
attrMask = ATTR_NONE;
if (insn->mode == MODE_64BIT)
attrMask |= ATTR_64BIT;
if (insn->vexSize) {
attrMask |= ATTR_VEX;
if (insn->vexSize == 3) {
switch (ppFromVEX3of3(insn->vexPrefix[2])) {
case VEX_PREFIX_66:
attrMask |= ATTR_OPSIZE;
break;
case VEX_PREFIX_F3:
attrMask |= ATTR_XS;
break;
case VEX_PREFIX_F2:
attrMask |= ATTR_XD;
break;
}
if (wFromVEX3of3(insn->vexPrefix[2]))
attrMask |= ATTR_REXW;
if (lFromVEX3of3(insn->vexPrefix[2]))
attrMask |= ATTR_VEXL;
}
else if (insn->vexSize == 2) {
switch (ppFromVEX2of2(insn->vexPrefix[1])) {
case VEX_PREFIX_66:
attrMask |= ATTR_OPSIZE;
break;
case VEX_PREFIX_F3:
attrMask |= ATTR_XS;
break;
case VEX_PREFIX_F2:
attrMask |= ATTR_XD;
break;
}
if (lFromVEX2of2(insn->vexPrefix[1]))
attrMask |= ATTR_VEXL;
}
else {
return -1;
}
}
else {
if (insn->rexPrefix & 0x08)
attrMask |= ATTR_REXW;
if (isPrefixAtLocation(insn, 0x66, insn->necessaryPrefixLocation))
attrMask |= ATTR_OPSIZE;
else if (isPrefixAtLocation(insn, 0xf3, insn->necessaryPrefixLocation))
attrMask |= ATTR_XS;
else if (isPrefixAtLocation(insn, 0xf2, insn->necessaryPrefixLocation))
attrMask |= ATTR_XD;
}
if (getIDWithAttrMask(&instructionID, insn, attrMask))
return -1;
/* The following clauses compensate for limitations of the tables. */
if ((attrMask & ATTR_XD) && (attrMask & ATTR_REXW)) {
/*
* Although for SSE instructions it is usually necessary to treat REX.W+F2
* as F2 for decode (in the absence of a 64BIT_REXW_XD category) there is
* an occasional instruction where F2 is incidental and REX.W is the more
* significant. If the decoded instruction is 32-bit and adding REX.W
* instead of F2 changes a 32 to a 64, we adopt the new encoding.
*/
const struct InstructionSpecifier *spec;
uint16_t instructionIDWithREXw;
const struct InstructionSpecifier *specWithREXw;
spec = specifierForUID(instructionID);
if (getIDWithAttrMask(&instructionIDWithREXw,
insn,
attrMask & (~ATTR_XD))) {
/*
* Decoding with REX.w would yield nothing; give up and return original
* decode.
*/
insn->instructionID = instructionID;
insn->spec = spec;
return 0;
}
specWithREXw = specifierForUID(instructionIDWithREXw);
if (is64BitEquivalent(spec->name, specWithREXw->name)) {
insn->instructionID = instructionIDWithREXw;
insn->spec = specWithREXw;
} else {
insn->instructionID = instructionID;
insn->spec = spec;
}
return 0;
}
if (insn->prefixPresent[0x66] && !(attrMask & ATTR_OPSIZE)) {
/*
* The instruction tables make no distinction between instructions that
* allow OpSize anywhere (i.e., 16-bit operations) and that need it in a
* particular spot (i.e., many MMX operations). In general we're
* conservative, but in the specific case where OpSize is present but not
* in the right place we check if there's a 16-bit operation.
*/
const struct InstructionSpecifier *spec;
uint16_t instructionIDWithOpsize;
const struct InstructionSpecifier *specWithOpsize;
spec = specifierForUID(instructionID);
if (getIDWithAttrMask(&instructionIDWithOpsize,
insn,
attrMask | ATTR_OPSIZE)) {
/*
* ModRM required with OpSize but not present; give up and return version
* without OpSize set
*/
insn->instructionID = instructionID;
insn->spec = spec;
return 0;
}
specWithOpsize = specifierForUID(instructionIDWithOpsize);
if (is16BitEquvalent(spec->name, specWithOpsize->name)) {
insn->instructionID = instructionIDWithOpsize;
insn->spec = specWithOpsize;
} else {
insn->instructionID = instructionID;
insn->spec = spec;
}
return 0;
}
insn->instructionID = instructionID;
insn->spec = specifierForUID(insn->instructionID);
return 0;
}