arm64: handle alias insn in a better way, and add support for MNEG. bug reported by Patroklos Argyroudis

MCInst.c

@@ -35,11 +35,21 @@ void MCInst_setOpcode(MCInst *inst, unsigned Op)
 	inst->Opcode = Op;
 }
 
+void MCInst_setOpcodePub(MCInst *inst, unsigned Op)
+{
+	inst->OpcodePub = Op;
+}
+
 unsigned MCInst_getOpcode(const MCInst *inst)
 {
 	return inst->Opcode;
 }
 
+unsigned MCInst_getOpcodePub(const MCInst *inst)
+{
+	return inst->OpcodePub;
+}
+
 MCOperand *MCInst_getOperand(MCInst *inst, unsigned i)
 {
 	return &inst->Operands[i];

MCInst.h

@@ -89,6 +89,7 @@ struct MCInst {
 	unsigned size;	// number of operands
 	cs_insn pub_insn;	// insn to be exposed to public
 	cs_mode mode;	// to be referenced by internal code
+	unsigned OpcodePub;
 };
 
 void MCInst_Init(MCInst *inst);
@@ -101,6 +102,10 @@ void MCInst_setOpcode(MCInst *inst, unsigned Op);
 
 unsigned MCInst_getOpcode(const MCInst*);
 
+void MCInst_setOpcodePub(MCInst *inst, unsigned Op);
+
+unsigned MCInst_getOpcodePub(const MCInst*);
+
 MCOperand *MCInst_getOperand(MCInst *inst, unsigned i);
 
 unsigned MCInst_getNumOperands(const MCInst *inst);

arch/AArch64/AArch64InstPrinter.c

@@ -636,10 +636,11 @@ void AArch64_printInst(MCInst *MI, SStream *O, void *Info)
 	if (printAliasInstr(MI, O, Info)) {
 		char *mnem = strdup(O->buffer);
 		char *tab = strchr(mnem, '\t');
-		if (tab)
+		if (tab) {
 			*tab = '\0';
+		}
 		// reflect the new insn name (alias) in the opcode
-		MCInst_setOpcode(MI, AArch64_get_insn_id2(AArch64_map_insn(mnem)));
+		MCInst_setOpcodePub(MI, AArch64_map_insn(mnem));
 		free(mnem);
 	} else
 		AArch64InstPrinter_printInstruction(MI, O);

arch/AArch64/mapping.c

@@ -1854,6 +1854,7 @@ void AArch64_get_insn_id(cs_insn *insn, unsigned int id)
 	}
 }
 
+// given public insn id, return internal instruction ID
 unsigned int AArch64_get_insn_id2(unsigned int id)
 {
 	return insn_reverse_id(insns, ARR_SIZE(insns), id);
@@ -2224,6 +2225,7 @@ char *AArch64_insn_name(unsigned int id)
 	return insn_name_maps[id].name;
 }
 
+// map instruction name to public instruction ID
 arm64_reg AArch64_map_insn(char *name)
 {
 	// map *S instructions back to original id
@@ -2234,6 +2236,8 @@ arm64_reg AArch64_map_insn(char *name)
 		{ ARM64_INS_BIC, "BICS" },
 		{ ARM64_INS_SBC, "SBCS" },
 		{ ARM64_INS_SUB, "SUBS" },
+		// alias insn
+		{ ARM64_INS_MNEG, "MNEG" },
 	};
 
 	// NOTE: skip first NULL name in insn_name_maps

arch/AArch64/mapping.h

@@ -18,7 +18,7 @@ unsigned int AArch64_get_insn_id2(unsigned int id);
 
 char *AArch64_insn_name(unsigned int id);
 
-// map instruction name to instruction ID
+// map instruction name to public instruction ID
 arm64_reg AArch64_map_insn(char *name);
 
 #endif

cs.c

@@ -170,8 +170,12 @@ static void fill_insn(cs_struct *handle, cs_insn *insn, char *buffer, MCInst *mc
 	memcpy(insn, &mci->pub_insn, sizeof(*insn));
 
 	// map internal instruction opcode to public insn ID
-	if (handle->insn_id)
-		handle->insn_id(insn, MCInst_getOpcode(mci));
+	if (MCInst_getOpcodePub(mci))
+		MCInst_setOpcode(mci, MCInst_getOpcodePub(mci));
+	else {
+		if (handle->insn_id)
+			handle->insn_id(insn, MCInst_getOpcode(mci));
+	}
 
 	if (printer)
 		printer(insn->id, insn, buffer);

include/arm64.h

@@ -691,6 +691,10 @@ typedef enum arm64_insn {
 	ARM64_INS_USUBW,
 	ARM64_INS_UXTB,
 	ARM64_INS_UXTH,
+
+	// alias insn
+	ARM64_INS_MNEG,
+
 	ARM64_INS_MAX,
 } arm64_insn;
 

tests/test_arm64.c

@@ -131,6 +131,8 @@ static void test()
 //#define ARM64_CODE "\x5f\x3f\x03\xd5"	// clrex
 //#define ARM64_CODE "\x5f\x3e\x03\xd5"	// clrex #14
 //#define ARM64_CODE "\x20\x00\x02\xab"	// adds	 x0, x1, x2 (alias of adds x0, x1, x2, lsl #0)
+//#define ARM64_CODE "\x20\xf4\x18\x9e"	// fcvtzs	x0, s1, #3
+//#define ARM64_CODE "\x20\xfc\x02\x9b"	// mneg	x0, x1, x2
 #define ARM64_CODE "\x21\x7c\x02\x9b\x21\x7c\x00\x53\x00\x40\x21\x4b\xe1\x0b\x40\xb9\x20\x04\x81\xda\x20\x08\x02\x8b"
 
 	struct platform platforms[] = {