-/*===- X86DisassemblerDecoder.c - Disassembler decoder -------------*- C -*-==*
+/*===-- X86DisassemblerDecoder.c - Disassembler decoder ------------*- C -*-===*
*
* The LLVM Compiler Infrastructure
*
*
*===----------------------------------------------------------------------===*/
-#include <assert.h> /* for assert() */
#include <stdarg.h> /* for va_*() */
#include <stdio.h> /* for vsnprintf() */
#include <stdlib.h> /* for exit() */
#define TRUE 1
#define FALSE 0
-#ifdef __GNUC__
-#define NORETURN __attribute__((noreturn))
+typedef int8_t bool;
+
+#ifndef NDEBUG
+#define debug(s) do { x86DisassemblerDebug(__FILE__, __LINE__, s); } while (0)
#else
-#define NORETURN
+#define debug(s) do { } while (0)
#endif
-#define unreachable(s) \
- do { \
- fprintf(stderr, "%s:%d: %s\n", __FILE__, __LINE__, s); \
- exit(-1); \
- } while (0);
/*
* contextForAttrs - Client for the instruction context table. Takes a set of
* @return - TRUE if the ModR/M byte is required, FALSE otherwise.
*/
static int modRMRequired(OpcodeType type,
- InstructionContext insnContext,
- uint8_t opcode) {
+ InstructionContext insnContext,
+ uint8_t opcode) {
const struct ContextDecision* decision = 0;
switch (type) {
case THREEBYTE_3A:
decision = &THREEBYTE3A_SYM;
break;
+ case THREEBYTE_A6:
+ decision = &THREEBYTEA6_SYM;
+ break;
+ case THREEBYTE_A7:
+ decision = &THREEBYTEA7_SYM;
+ break;
}
-
+
return decision->opcodeDecisions[insnContext].modRMDecisions[opcode].
modrm_type != MODRM_ONEENTRY;
-
- unreachable("Unknown opcode type");
- return 0;
}
/*
* @param insnContext - See modRMRequired().
* @param opcode - See modRMRequired().
* @param modRM - The ModR/M byte if required, or any value if not.
+ * @return - The UID of the instruction, or 0 on failure.
*/
static InstrUID decode(OpcodeType type,
- InstructionContext insnContext,
- uint8_t opcode,
- uint8_t modRM) {
- struct ModRMDecision* dec;
+ InstructionContext insnContext,
+ uint8_t opcode,
+ uint8_t modRM) {
+ const struct ModRMDecision* dec = 0;
switch (type) {
- default:
- unreachable("Unknown opcode type");
case ONEBYTE:
dec = &ONEBYTE_SYM.opcodeDecisions[insnContext].modRMDecisions[opcode];
break;
case THREEBYTE_3A:
dec = &THREEBYTE3A_SYM.opcodeDecisions[insnContext].modRMDecisions[opcode];
break;
+ case THREEBYTE_A6:
+ dec = &THREEBYTEA6_SYM.opcodeDecisions[insnContext].modRMDecisions[opcode];
+ break;
+ case THREEBYTE_A7:
+ dec = &THREEBYTEA7_SYM.opcodeDecisions[insnContext].modRMDecisions[opcode];
+ break;
}
switch (dec->modrm_type) {
default:
- unreachable("Corrupt table! Unknown modrm_type");
+ debug("Corrupt table! Unknown modrm_type");
+ return 0;
case MODRM_ONEENTRY:
- return dec->instructionIDs[0];
+ return modRMTable[dec->instructionIDs];
case MODRM_SPLITRM:
if (modFromModRM(modRM) == 0x3)
- return dec->instructionIDs[1];
- else
- return dec->instructionIDs[0];
+ return modRMTable[dec->instructionIDs+1];
+ return modRMTable[dec->instructionIDs];
+ case MODRM_SPLITREG:
+ if (modFromModRM(modRM) == 0x3)
+ return modRMTable[dec->instructionIDs+((modRM & 0x38) >> 3)+8];
+ return modRMTable[dec->instructionIDs+((modRM & 0x38) >> 3)];
case MODRM_FULL:
- return dec->instructionIDs[modRM];
+ return modRMTable[dec->instructionIDs+modRM];
}
-
- return 0;
}
/*
* decode(); specifierForUID will not check bounds.
* @return - A pointer to the specification for that instruction.
*/
-static struct InstructionSpecifier* specifierForUID(InstrUID uid) {
+static const struct InstructionSpecifier *specifierForUID(InstrUID uid) {
return &INSTRUCTIONS_SYM[uid];
}
BOOL isPrefix = TRUE;
BOOL prefixGroups[4] = { FALSE };
uint64_t prefixLocation;
- uint8_t byte;
+ uint8_t byte = 0;
BOOL hasAdSize = FALSE;
BOOL hasOpSize = FALSE;
insn->segmentOverride = SEG_OVERRIDE_GS;
break;
default:
- unreachable("Unhandled override");
+ debug("Unhandled override");
+ return -1;
}
if (prefixGroups[1])
dbgprintf(insn, "Redundant Group 2 prefix");
if (isPrefix)
dbgprintf(insn, "Found prefix 0x%hhx", byte);
}
+
+ insn->vexSize = 0;
- if (insn->mode == MODE_64BIT) {
- if ((byte & 0xf0) == 0x40) {
- uint8_t opcodeByte;
+ if (byte == 0xc4) {
+ uint8_t byte1;
- if(lookAtByte(insn, &opcodeByte) || ((opcodeByte & 0xf0) == 0x40)) {
- dbgprintf(insn, "Redundant REX prefix");
- return -1;
+ if (lookAtByte(insn, &byte1)) {
+ dbgprintf(insn, "Couldn't read second byte of VEX");
+ return -1;
+ }
+
+ if (insn->mode == MODE_64BIT || (byte1 & 0xc0) == 0xc0) {
+ insn->vexSize = 3;
+ insn->necessaryPrefixLocation = insn->readerCursor - 1;
+ }
+ else {
+ unconsumeByte(insn);
+ insn->necessaryPrefixLocation = insn->readerCursor - 1;
+ }
+
+ if (insn->vexSize == 3) {
+ insn->vexPrefix[0] = byte;
+ consumeByte(insn, &insn->vexPrefix[1]);
+ consumeByte(insn, &insn->vexPrefix[2]);
+
+ /* We simulate the REX prefix for simplicity's sake */
+
+ if (insn->mode == MODE_64BIT) {
+ insn->rexPrefix = 0x40
+ | (wFromVEX3of3(insn->vexPrefix[2]) << 3)
+ | (rFromVEX2of3(insn->vexPrefix[1]) << 2)
+ | (xFromVEX2of3(insn->vexPrefix[1]) << 1)
+ | (bFromVEX2of3(insn->vexPrefix[1]) << 0);
}
+
+ switch (ppFromVEX3of3(insn->vexPrefix[2]))
+ {
+ default:
+ break;
+ case VEX_PREFIX_66:
+ hasOpSize = TRUE;
+ break;
+ }
+
+ dbgprintf(insn, "Found VEX prefix 0x%hhx 0x%hhx 0x%hhx", insn->vexPrefix[0], insn->vexPrefix[1], insn->vexPrefix[2]);
+ }
+ }
+ else if (byte == 0xc5) {
+ uint8_t byte1;
+
+ if (lookAtByte(insn, &byte1)) {
+ dbgprintf(insn, "Couldn't read second byte of VEX");
+ return -1;
+ }
- insn->rexPrefix = byte;
- insn->necessaryPrefixLocation = insn->readerCursor - 2;
-
- dbgprintf(insn, "Found REX prefix 0x%hhx", byte);
- } else {
+ if (insn->mode == MODE_64BIT || (byte1 & 0xc0) == 0xc0) {
+ insn->vexSize = 2;
+ }
+ else {
+ unconsumeByte(insn);
+ }
+
+ if (insn->vexSize == 2) {
+ insn->vexPrefix[0] = byte;
+ consumeByte(insn, &insn->vexPrefix[1]);
+
+ if (insn->mode == MODE_64BIT) {
+ insn->rexPrefix = 0x40
+ | (rFromVEX2of2(insn->vexPrefix[1]) << 2);
+ }
+
+ switch (ppFromVEX2of2(insn->vexPrefix[1]))
+ {
+ default:
+ break;
+ case VEX_PREFIX_66:
+ hasOpSize = TRUE;
+ break;
+ }
+
+ dbgprintf(insn, "Found VEX prefix 0x%hhx 0x%hhx", insn->vexPrefix[0], insn->vexPrefix[1]);
+ }
+ }
+ else {
+ if (insn->mode == MODE_64BIT) {
+ if ((byte & 0xf0) == 0x40) {
+ uint8_t opcodeByte;
+
+ if (lookAtByte(insn, &opcodeByte) || ((opcodeByte & 0xf0) == 0x40)) {
+ dbgprintf(insn, "Redundant REX prefix");
+ return -1;
+ }
+
+ insn->rexPrefix = byte;
+ insn->necessaryPrefixLocation = insn->readerCursor - 2;
+
+ dbgprintf(insn, "Found REX prefix 0x%hhx", byte);
+ } else {
+ unconsumeByte(insn);
+ insn->necessaryPrefixLocation = insn->readerCursor - 1;
+ }
+ } else {
unconsumeByte(insn);
insn->necessaryPrefixLocation = insn->readerCursor - 1;
}
- } else {
- unconsumeByte(insn);
}
-
+
if (insn->mode == MODE_16BIT) {
insn->registerSize = (hasOpSize ? 4 : 2);
insn->addressSize = (hasAdSize ? 4 : 2);
insn->registerSize = (hasOpSize ? 2 : 4);
insn->addressSize = (hasAdSize ? 2 : 4);
insn->displacementSize = (hasAdSize ? 2 : 4);
- insn->immediateSize = (hasAdSize ? 2 : 4);
+ insn->immediateSize = (hasOpSize ? 2 : 4);
} else if (insn->mode == MODE_64BIT) {
if (insn->rexPrefix && wFromREX(insn->rexPrefix)) {
insn->registerSize = 8;
dbgprintf(insn, "readOpcode()");
insn->opcodeType = ONEBYTE;
+
+ if (insn->vexSize == 3)
+ {
+ switch (mmmmmFromVEX2of3(insn->vexPrefix[1]))
+ {
+ default:
+ dbgprintf(insn, "Unhandled m-mmmm field for instruction (0x%hhx)", mmmmmFromVEX2of3(insn->vexPrefix[1]));
+ return -1;
+ case 0:
+ break;
+ case VEX_LOB_0F:
+ insn->twoByteEscape = 0x0f;
+ insn->opcodeType = TWOBYTE;
+ return consumeByte(insn, &insn->opcode);
+ case VEX_LOB_0F38:
+ insn->twoByteEscape = 0x0f;
+ insn->threeByteEscape = 0x38;
+ insn->opcodeType = THREEBYTE_38;
+ return consumeByte(insn, &insn->opcode);
+ case VEX_LOB_0F3A:
+ insn->twoByteEscape = 0x0f;
+ insn->threeByteEscape = 0x3a;
+ insn->opcodeType = THREEBYTE_3A;
+ return consumeByte(insn, &insn->opcode);
+ }
+ }
+ else if (insn->vexSize == 2)
+ {
+ insn->twoByteEscape = 0x0f;
+ insn->opcodeType = TWOBYTE;
+ return consumeByte(insn, &insn->opcode);
+ }
+
if (consumeByte(insn, ¤t))
return -1;
return -1;
insn->opcodeType = THREEBYTE_3A;
+ } else if (current == 0xa6) {
+ dbgprintf(insn, "Found a three-byte escape prefix (0x%hhx)", current);
+
+ insn->threeByteEscape = current;
+
+ if (consumeByte(insn, ¤t))
+ return -1;
+
+ insn->opcodeType = THREEBYTE_A6;
+ } else if (current == 0xa7) {
+ dbgprintf(insn, "Found a three-byte escape prefix (0x%hhx)", current);
+
+ insn->threeByteEscape = current;
+
+ if (consumeByte(insn, ¤t))
+ return -1;
+
+ insn->opcodeType = THREEBYTE_A7;
} else {
dbgprintf(insn, "Didn't find a three-byte escape prefix");
insn->opcode);
if (hasModRMExtension) {
- readModRM(insn);
+ if (readModRM(insn))
+ return -1;
*instructionID = decode(insn->opcodeType,
instructionClass,
static BOOL is16BitEquvalent(const char* orig, const char* equiv) {
off_t i;
- for(i = 0;; i++) {
- if(orig[i] == '\0' && equiv[i] == '\0')
- return TRUE;
- if(orig[i] == '\0' || equiv[i] == '\0')
- return FALSE;
- if(orig[i] != equiv[i]) {
- if((orig[i] == 'Q' || orig[i] == 'L') && equiv[i] == 'W')
- continue;
- if((orig[i] == '6' || orig[i] == '3') && equiv[i] == '1')
- continue;
- if((orig[i] == '4' || orig[i] == '2') && equiv[i] == '6')
- continue;
- return FALSE;
- }
- }
-}
-
-/*
- * is64BitEquivalent - Determines whether two instruction names refer to
- * equivalent instructions but one is 64-bit whereas the other is not.
- *
- * @param orig - The instruction that is not 64-bit
- * @param equiv - The instruction that is 64-bit
- */
-static BOOL is64BitEquivalent(const char* orig, const char* equiv) {
- off_t i;
-
- for(i = 0;; i++) {
- if(orig[i] == '\0' && equiv[i] == '\0')
+ for (i = 0;; i++) {
+ if (orig[i] == '\0' && equiv[i] == '\0')
return TRUE;
- if(orig[i] == '\0' || equiv[i] == '\0')
+ if (orig[i] == '\0' || equiv[i] == '\0')
return FALSE;
- if(orig[i] != equiv[i]) {
- if((orig[i] == 'W' || orig[i] == 'L') && equiv[i] == 'Q')
+ if (orig[i] != equiv[i]) {
+ if ((orig[i] == 'Q' || orig[i] == 'L') && equiv[i] == 'W')
continue;
- if((orig[i] == '1' || orig[i] == '3') && equiv[i] == '6')
+ if ((orig[i] == '6' || orig[i] == '3') && equiv[i] == '1')
continue;
- if((orig[i] == '6' || orig[i] == '2') && equiv[i] == '4')
+ if ((orig[i] == '4' || orig[i] == '2') && equiv[i] == '6')
continue;
return FALSE;
}
}
}
-
/*
* getID - Determines the ID of an instruction, consuming the ModR/M byte as
* appropriate for extended and escape opcodes. Determines the attributes and
* @return - 0 if the ModR/M could be read when needed or was not needed;
* nonzero otherwise.
*/
-static int getID(struct InternalInstruction* insn) {
+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, 0xf3, insn->necessaryPrefixLocation))
+ attrMask |= ATTR_XS;
+ else if (isPrefixAtLocation(insn, 0xf2, insn->necessaryPrefixLocation))
+ attrMask |= ATTR_XD;
+ }
+
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))
+
+ if (getIDWithAttrMask(&instructionID, insn, attrMask))
return -1;
-
+
/* The following clauses compensate for limitations of the tables. */
-
- if ((attrMask & ATTR_XD) && (attrMask & ATTR_REXW)) {
+
+ if ((attrMask & ATTR_VEXL) && (attrMask & ATTR_REXW) &&
+ !(attrMask & ATTR_OPSIZE)) {
/*
- * 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.
+ * 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.
*/
-
- struct InstructionSpecifier* spec;
- uint16_t instructionIDWithREXw;
- struct InstructionSpecifier* specWithREXw;
-
+
+ const struct InstructionSpecifier *spec;
+ uint16_t instructionIDWithWBit;
+ const struct InstructionSpecifier *specWithWBit;
+
spec = specifierForUID(instructionID);
-
- if (getIDWithAttrMask(&instructionIDWithREXw,
+
+ if (getIDWithAttrMask(&instructionIDWithWBit,
insn,
- attrMask & (~ATTR_XD))) {
- /*
- * Decoding with REX.w would yield nothing; give up and return original
- * decode.
- */
-
+ (attrMask & (~ATTR_VEXL)) | ATTR_REXW)) {
insn->instructionID = instructionID;
insn->spec = spec;
return 0;
}
-
- specWithREXw = specifierForUID(instructionIDWithREXw);
-
- if (is64BitEquivalent(spec->name, specWithREXw->name)) {
- insn->instructionID = instructionIDWithREXw;
- insn->spec = specWithREXw;
+
+ 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
* in the right place we check if there's a 16-bit operation.
*/
- struct InstructionSpecifier* spec;
+ const struct InstructionSpecifier *spec;
uint16_t instructionIDWithOpsize;
- struct InstructionSpecifier* specWithOpsize;
+ const char *specName, *specWithOpSizeName;
spec = specifierForUID(instructionID);
return 0;
}
- specWithOpsize = specifierForUID(instructionIDWithOpsize);
-
- if (is16BitEquvalent(spec->name, specWithOpsize->name)) {
+ specName = x86DisassemblerGetInstrName(instructionID, miiArg);
+ specWithOpSizeName =
+ x86DisassemblerGetInstrName(instructionIDWithOpsize, miiArg);
+
+ if (is16BitEquvalent(specName, specWithOpSizeName)) {
insn->instructionID = instructionIDWithOpsize;
- insn->spec = specWithOpsize;
+ 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);
insn->sibIndex = SIB_INDEX_NONE;
break;
default:
- insn->sibIndex = (EABase)(sibIndexBase + index);
+ insn->sibIndex = (SIBIndex)(sibIndexBase + index);
if (insn->sibIndex == SIB_INDEX_sib ||
insn->sibIndex == SIB_INDEX_sib64)
insn->sibIndex = SIB_INDEX_NONE;
SIB_BASE_EBP : SIB_BASE_RBP);
break;
case 0x3:
- unreachable("Cannot have Mod = 0b11 and a SIB byte");
+ debug("Cannot have Mod = 0b11 and a SIB byte");
+ return -1;
}
break;
default:
- insn->sibBase = (EABase)(sibBaseBase + base);
+ insn->sibBase = (SIBBase)(sibBaseBase + base);
break;
}
if (insn->consumedModRM)
return 0;
- consumeByte(insn, &insn->modRM);
+ if (consumeByte(insn, &insn->modRM))
+ return -1;
insn->consumedModRM = TRUE;
mod = modFromModRM(insn->modRM);
if (rm == 0x6) {
insn->eaBase = EA_BASE_NONE;
insn->eaDisplacement = EA_DISP_16;
- if(readDisplacement(insn))
+ if (readDisplacement(insn))
return -1;
} else {
insn->eaBase = (EABase)(insn->eaBaseBase + rm);
case 0x1:
insn->eaBase = (EABase)(insn->eaBaseBase + rm);
insn->eaDisplacement = EA_DISP_8;
- if(readDisplacement(insn))
+ if (readDisplacement(insn))
return -1;
break;
case 0x2:
insn->eaBase = (EABase)(insn->eaBaseBase + rm);
insn->eaDisplacement = EA_DISP_16;
- if(readDisplacement(insn))
+ if (readDisplacement(insn))
return -1;
break;
case 0x3:
insn->eaBase = (EABase)(insn->eaRegBase + rm);
- if(readDisplacement(insn))
+ if (readDisplacement(insn))
return -1;
break;
}
insn->eaBase = (insn->addressSize == 4 ?
EA_BASE_sib : EA_BASE_sib64);
readSIB(insn);
- if(readDisplacement(insn))
+ if (readDisplacement(insn))
return -1;
break;
case 0x5:
insn->eaBase = EA_BASE_NONE;
insn->eaDisplacement = EA_DISP_32;
- if(readDisplacement(insn))
+ if (readDisplacement(insn))
return -1;
break;
default:
case 0xc: /* in case REXW.b is set */
insn->eaBase = EA_BASE_sib;
readSIB(insn);
- if(readDisplacement(insn))
+ if (readDisplacement(insn))
return -1;
break;
default:
insn->eaBase = (EABase)(insn->eaBaseBase + rm);
- if(readDisplacement(insn))
+ if (readDisplacement(insn))
return -1;
break;
}
*valid = 1; \
switch (type) { \
default: \
- unreachable("Unhandled register type"); \
+ debug("Unhandled register type"); \
+ *valid = 0; \
+ return 0; \
case TYPE_Rv: \
return base + index; \
case TYPE_R8: \
- if(insn->rexPrefix && \
+ if (insn->rexPrefix && \
index >= 4 && index <= 7) { \
return prefix##_SPL + (index - 4); \
} else { \
return prefix##_EAX + index; \
case TYPE_R64: \
return prefix##_RAX + index; \
+ case TYPE_XMM256: \
+ return prefix##_YMM0 + index; \
case TYPE_XMM128: \
case TYPE_XMM64: \
case TYPE_XMM32: \
case TYPE_MM64: \
case TYPE_MM32: \
case TYPE_MM: \
- if(index > 7) \
+ if (index > 7) \
*valid = 0; \
return prefix##_MM0 + index; \
case TYPE_SEGMENTREG: \
- if(index > 5) \
+ if (index > 5) \
*valid = 0; \
return prefix##_ES + index; \
case TYPE_DEBUGREG: \
- if(index > 7) \
+ if (index > 7) \
*valid = 0; \
return prefix##_DR0 + index; \
- case TYPE_CR32: \
- if(index > 7) \
- *valid = 0; \
- return prefix##_ECR0 + index; \
- case TYPE_CR64: \
- if(index > 8) \
+ case TYPE_CONTROLREG: \
+ if (index > 8) \
*valid = 0; \
- return prefix##_RCR0 + index; \
+ return prefix##_CR0 + index; \
} \
}
* @param index - The existing value of the field as reported by readModRM().
* @param valid - The address of a uint8_t. The target is set to 1 if the
* field is valid for the register class; 0 if not.
+ * @return - The proper value.
*/
GENERIC_FIXUP_FUNC(fixupRegValue, insn->regBase, MODRM_REG)
GENERIC_FIXUP_FUNC(fixupRMValue, insn->eaRegBase, EA_REG)
* invalid for its class.
*/
static int fixupReg(struct InternalInstruction *insn,
- struct OperandSpecifier *op) {
+ const struct OperandSpecifier *op) {
uint8_t valid;
dbgprintf(insn, "fixupReg()");
switch ((OperandEncoding)op->encoding) {
default:
- unreachable("Expected a REG or R/M encoding in fixupReg");
+ debug("Expected a REG or R/M encoding in fixupReg");
+ return -1;
+ case ENCODING_VVVV:
+ insn->vvvv = (Reg)fixupRegValue(insn,
+ (OperandType)op->type,
+ insn->vvvv,
+ &valid);
+ if (!valid)
+ return -1;
+ break;
case ENCODING_REG:
insn->reg = (Reg)fixupRegValue(insn,
(OperandType)op->type,
* @param insn - The instruction whose opcode field is to be read.
* @param inModRM - Indicates that the opcode field is to be read from the
* ModR/M extension; useful for escape opcodes
+ * @return - 0 on success; nonzero otherwise.
*/
-static void readOpcodeModifier(struct InternalInstruction* insn) {
+static int readOpcodeModifier(struct InternalInstruction* insn) {
dbgprintf(insn, "readOpcodeModifier()");
if (insn->consumedOpcodeModifier)
- return;
+ return 0;
insn->consumedOpcodeModifier = TRUE;
- switch(insn->spec->modifierType) {
+ switch (insn->spec->modifierType) {
default:
- unreachable("Unknown modifier type.");
+ debug("Unknown modifier type.");
+ return -1;
case MODIFIER_NONE:
- unreachable("No modifier but an operand expects one.");
+ debug("No modifier but an operand expects one.");
+ return -1;
case MODIFIER_OPCODE:
insn->opcodeModifier = insn->opcode - insn->spec->modifierBase;
- break;
+ return 0;
case MODIFIER_MODRM:
insn->opcodeModifier = insn->modRM - insn->spec->modifierBase;
- break;
+ return 0;
}
}
* @param size - The width (in bytes) of the register being specified.
* 1 means AL and friends, 2 means AX, 4 means EAX, and 8 means
* RAX.
+ * @return - 0 on success; nonzero otherwise.
*/
-static void readOpcodeRegister(struct InternalInstruction* insn, uint8_t size) {
+static int readOpcodeRegister(struct InternalInstruction* insn, uint8_t size) {
dbgprintf(insn, "readOpcodeRegister()");
- readOpcodeModifier(insn);
+ if (readOpcodeModifier(insn))
+ return -1;
if (size == 0)
size = insn->registerSize;
case 1:
insn->opcodeRegister = (Reg)(MODRM_REG_AL + ((bFromREX(insn->rexPrefix) << 3)
| insn->opcodeModifier));
- if(insn->rexPrefix &&
- insn->opcodeRegister >= MODRM_REG_AL + 0x4 &&
- insn->opcodeRegister < MODRM_REG_AL + 0x8) {
+ if (insn->rexPrefix &&
+ insn->opcodeRegister >= MODRM_REG_AL + 0x4 &&
+ insn->opcodeRegister < MODRM_REG_AL + 0x8) {
insn->opcodeRegister = (Reg)(MODRM_REG_SPL
+ (insn->opcodeRegister - MODRM_REG_AL - 4));
}
| insn->opcodeModifier));
break;
case 4:
- insn->opcodeRegister = (Reg)(MODRM_REG_EAX +
+ insn->opcodeRegister = (Reg)(MODRM_REG_EAX
+ ((bFromREX(insn->rexPrefix) << 3)
| insn->opcodeModifier));
break;
| insn->opcodeModifier));
break;
}
+
+ return 0;
}
/*
dbgprintf(insn, "readImmediate()");
- if (insn->numImmediatesConsumed == 2)
- unreachable("Already consumed two immediates");
+ if (insn->numImmediatesConsumed == 2) {
+ debug("Already consumed two immediates");
+ return -1;
+ }
if (size == 0)
size = insn->immediateSize;
return 0;
}
+/*
+ * readVVVV - Consumes vvvv from an instruction if it has a VEX prefix.
+ *
+ * @param insn - The instruction whose operand is to be read.
+ * @return - 0 if the vvvv was successfully consumed; nonzero
+ * otherwise.
+ */
+static int readVVVV(struct InternalInstruction* insn) {
+ dbgprintf(insn, "readVVVV()");
+
+ if (insn->vexSize == 3)
+ insn->vvvv = vvvvFromVEX3of3(insn->vexPrefix[2]);
+ else if (insn->vexSize == 2)
+ insn->vvvv = vvvvFromVEX2of2(insn->vexPrefix[1]);
+ else
+ return -1;
+
+ if (insn->mode != MODE_64BIT)
+ insn->vvvv &= 0x7;
+
+ return 0;
+}
+
/*
* readOperands - Consults the specifier for an instruction and consumes all
* operands for that instruction, interpreting them as it goes.
*/
static int readOperands(struct InternalInstruction* insn) {
int index;
+ int hasVVVV, needVVVV;
+ int sawRegImm = 0;
dbgprintf(insn, "readOperands()");
+
+ /* If non-zero vvvv specified, need to make sure one of the operands
+ uses it. */
+ hasVVVV = !readVVVV(insn);
+ needVVVV = hasVVVV && (insn->vvvv != 0);
for (index = 0; index < X86_MAX_OPERANDS; ++index) {
switch (insn->spec->operands[index].encoding) {
dbgprintf(insn, "We currently don't hande code-offset encodings");
return -1;
case ENCODING_IB:
+ if (sawRegImm) {
+ /* Saw a register immediate so don't read again and instead split the
+ previous immediate. FIXME: This is a hack. */
+ insn->immediates[insn->numImmediatesConsumed] =
+ insn->immediates[insn->numImmediatesConsumed - 1] & 0xf;
+ ++insn->numImmediatesConsumed;
+ break;
+ }
if (readImmediate(insn, 1))
return -1;
+ if (insn->spec->operands[index].type == TYPE_IMM3 &&
+ insn->immediates[insn->numImmediatesConsumed - 1] > 7)
+ return -1;
+ if (insn->spec->operands[index].type == TYPE_XMM128 ||
+ insn->spec->operands[index].type == TYPE_XMM256)
+ sawRegImm = 1;
break;
case ENCODING_IW:
if (readImmediate(insn, 2))
return -1;
break;
case ENCODING_Iv:
- readImmediate(insn, insn->immediateSize);
+ if (readImmediate(insn, insn->immediateSize))
+ return -1;
break;
case ENCODING_Ia:
- readImmediate(insn, insn->addressSize);
+ if (readImmediate(insn, insn->addressSize))
+ return -1;
break;
case ENCODING_RB:
- readOpcodeRegister(insn, 1);
+ if (readOpcodeRegister(insn, 1))
+ return -1;
break;
case ENCODING_RW:
- readOpcodeRegister(insn, 2);
+ if (readOpcodeRegister(insn, 2))
+ return -1;
break;
case ENCODING_RD:
- readOpcodeRegister(insn, 4);
+ if (readOpcodeRegister(insn, 4))
+ return -1;
break;
case ENCODING_RO:
- readOpcodeRegister(insn, 8);
+ if (readOpcodeRegister(insn, 8))
+ return -1;
break;
case ENCODING_Rv:
- readOpcodeRegister(insn, 0);
+ if (readOpcodeRegister(insn, 0))
+ return -1;
break;
case ENCODING_I:
- readOpcodeModifier(insn);
+ if (readOpcodeModifier(insn))
+ return -1;
+ break;
+ case ENCODING_VVVV:
+ needVVVV = 0; /* Mark that we have found a VVVV operand. */
+ if (!hasVVVV)
+ return -1;
+ if (fixupReg(insn, &insn->spec->operands[index]))
+ return -1;
break;
case ENCODING_DUP:
break;
return -1;
}
}
+
+ /* If we didn't find ENCODING_VVVV operand, but non-zero vvvv present, fail */
+ if (needVVVV) return -1;
return 0;
}
void* readerArg,
dlog_t logger,
void* loggerArg,
+ void* miiArg,
uint64_t startLoc,
DisassemblerMode mode) {
memset(insn, 0, sizeof(struct InternalInstruction));
if (readPrefixes(insn) ||
readOpcode(insn) ||
- getID(insn) ||
+ getID(insn, miiArg) ||
insn->instructionID == 0 ||
readOperands(insn))
return -1;