[PATCH v2 03/13] x86/kvm/emulate: Introduce COP2

[Peter Zijlstra <peterz@infradead.org>]

Replace the FASTOP2 instructions.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/x86/kvm/emulate.c |   87 +++++++++++++++++++++++++++----------------------
 1 file changed, 48 insertions(+), 39 deletions(-)

--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -300,7 +300,7 @@ static int em_##op(struct x86_emulate_ct
 			_ASM_EXTABLE_TYPE_REG(10b, 11f, EX_TYPE_ZERO_REG, %%esi))
 
 #define COP_ASM2(op, dst, src) \
-		COP_ASM(#op " %" #src ", %" #dst " \n\t")
+		COP_ASM(#op " %%" #src ", %%" #dst " \n\t")
 
 #define COP_END \
 	} \
@@ -317,6 +317,15 @@ static int em_##op(struct x86_emulate_ct
 	ON64(case 8: COP_ASM1(op##q, rax); break;) \
 	COP_END
 
+/* 2-operand, using "a" (dst), "d" (src) */
+#define COP2(op) \
+	COP_START(op) \
+	case 1: COP_ASM2(op##b, al, dl); break; \
+	case 2: COP_ASM2(op##w, ax, dx); break; \
+	case 4: COP_ASM2(op##l, eax, edx); break; \
+	ON64(case 8: COP_ASM2(op##q, rax, rdx); break;) \
+	COP_END
+
 /*
  * fastop functions have a special calling convention:
  *
@@ -1027,15 +1036,16 @@ static int read_descriptor(struct x86_em
 	return rc;
 }
 
-FASTOP2(add);
-FASTOP2(or);
-FASTOP2(adc);
-FASTOP2(sbb);
-FASTOP2(and);
-FASTOP2(sub);
-FASTOP2(xor);
-FASTOP2(cmp);
-FASTOP2(test);
+COP2(add);
+COP2(or);
+COP2(adc);
+COP2(sbb);
+COP2(and);
+COP2(sub);
+COP2(xor);
+COP2(cmp);
+COP2(test);
+COP2(xadd);
 
 FASTOP1SRC2(mul, mul_ex);
 FASTOP1SRC2(imul, imul_ex);
@@ -1067,7 +1077,6 @@ FASTOP2W(bts);
 FASTOP2W(btr);
 FASTOP2W(btc);
 
-FASTOP2(xadd);
 
 FASTOP2R(cmp, cmp_r);
 
@@ -2304,7 +2313,7 @@ static int em_cmpxchg(struct x86_emulate
 	ctxt->dst.val = reg_read(ctxt, VCPU_REGS_RAX);
 	ctxt->src.orig_val = ctxt->src.val;
 	ctxt->src.val = ctxt->dst.orig_val;
-	fastop(ctxt, em_cmp);
+	em_cmp(ctxt);
 
 	if (ctxt->eflags & X86_EFLAGS_ZF) {
 		/* Success: write back to memory; no update of EAX */
@@ -3069,7 +3078,7 @@ static int em_das(struct x86_emulate_ctx
 	ctxt->src.type = OP_IMM;
 	ctxt->src.val = 0;
 	ctxt->src.bytes = 1;
-	fastop(ctxt, em_or);
+	em_or(ctxt);
 	ctxt->eflags &= ~(X86_EFLAGS_AF | X86_EFLAGS_CF);
 	if (cf)
 		ctxt->eflags |= X86_EFLAGS_CF;
@@ -3095,7 +3104,7 @@ static int em_aam(struct x86_emulate_ctx
 	ctxt->src.type = OP_IMM;
 	ctxt->src.val = 0;
 	ctxt->src.bytes = 1;
-	fastop(ctxt, em_or);
+	em_or(ctxt);
 
 	return X86EMUL_CONTINUE;
 }
@@ -3113,7 +3122,7 @@ static int em_aad(struct x86_emulate_ctx
 	ctxt->src.type = OP_IMM;
 	ctxt->src.val = 0;
 	ctxt->src.bytes = 1;
-	fastop(ctxt, em_or);
+	em_or(ctxt);
 
 	return X86EMUL_CONTINUE;
 }
@@ -3998,9 +4007,9 @@ static int check_perm_out(struct x86_emu
 #define I2bvIP(_f, _e, _i, _p) \
 	IIP((_f) | ByteOp, _e, _i, _p), IIP(_f, _e, _i, _p)
 
-#define F6ALU(_f, _e) F2bv((_f) | DstMem | SrcReg | ModRM, _e),		\
-		F2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock, _e),	\
-		F2bv(((_f) & ~Lock) | DstAcc | SrcImm, _e)
+#define I6ALU(_f, _e) I2bv((_f) | DstMem | SrcReg | ModRM, _e),		\
+		I2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock, _e),	\
+		I2bv(((_f) & ~Lock) | DstAcc | SrcImm, _e)
 
 static const struct opcode group7_rm0[] = {
 	N,
@@ -4038,14 +4047,14 @@ static const struct opcode group7_rm7[]
 };
 
 static const struct opcode group1[] = {
-	F(Lock, em_add),
-	F(Lock | PageTable, em_or),
-	F(Lock, em_adc),
-	F(Lock, em_sbb),
-	F(Lock | PageTable, em_and),
-	F(Lock, em_sub),
-	F(Lock, em_xor),
-	F(NoWrite, em_cmp),
+	I(Lock, em_add),
+	I(Lock | PageTable, em_or),
+	I(Lock, em_adc),
+	I(Lock, em_sbb),
+	I(Lock | PageTable, em_and),
+	I(Lock, em_sub),
+	I(Lock, em_xor),
+	I(NoWrite, em_cmp),
 };
 
 static const struct opcode group1A[] = {
@@ -4064,8 +4073,8 @@ static const struct opcode group2[] = {
 };
 
 static const struct opcode group3[] = {
-	F(DstMem | SrcImm | NoWrite, em_test),
-	F(DstMem | SrcImm | NoWrite, em_test),
+	I(DstMem | SrcImm | NoWrite, em_test),
+	I(DstMem | SrcImm | NoWrite, em_test),
 	I(DstMem | SrcNone | Lock, em_not),
 	I(DstMem | SrcNone | Lock, em_neg),
 	F(DstXacc | Src2Mem, em_mul_ex),
@@ -4258,29 +4267,29 @@ static const struct instr_dual instr_dua
 
 static const struct opcode opcode_table[256] = {
 	/* 0x00 - 0x07 */
-	F6ALU(Lock, em_add),
+	I6ALU(Lock, em_add),
 	I(ImplicitOps | Stack | No64 | Src2ES, em_push_sreg),
 	I(ImplicitOps | Stack | No64 | Src2ES, em_pop_sreg),
 	/* 0x08 - 0x0F */
-	F6ALU(Lock | PageTable, em_or),
+	I6ALU(Lock | PageTable, em_or),
 	I(ImplicitOps | Stack | No64 | Src2CS, em_push_sreg),
 	N,
 	/* 0x10 - 0x17 */
-	F6ALU(Lock, em_adc),
+	I6ALU(Lock, em_adc),
 	I(ImplicitOps | Stack | No64 | Src2SS, em_push_sreg),
 	I(ImplicitOps | Stack | No64 | Src2SS, em_pop_sreg),
 	/* 0x18 - 0x1F */
-	F6ALU(Lock, em_sbb),
+	I6ALU(Lock, em_sbb),
 	I(ImplicitOps | Stack | No64 | Src2DS, em_push_sreg),
 	I(ImplicitOps | Stack | No64 | Src2DS, em_pop_sreg),
 	/* 0x20 - 0x27 */
-	F6ALU(Lock | PageTable, em_and), N, N,
+	I6ALU(Lock | PageTable, em_and), N, N,
 	/* 0x28 - 0x2F */
-	F6ALU(Lock, em_sub), N, I(ByteOp | DstAcc | No64, em_das),
+	I6ALU(Lock, em_sub), N, I(ByteOp | DstAcc | No64, em_das),
 	/* 0x30 - 0x37 */
-	F6ALU(Lock, em_xor), N, N,
+	I6ALU(Lock, em_xor), N, N,
 	/* 0x38 - 0x3F */
-	F6ALU(NoWrite, em_cmp), N, N,
+	I6ALU(NoWrite, em_cmp), N, N,
 	/* 0x40 - 0x4F */
 	X8(I(DstReg, em_inc)), X8(I(DstReg, em_dec)),
 	/* 0x50 - 0x57 */
@@ -4306,7 +4315,7 @@ static const struct opcode opcode_table[
 	G(DstMem | SrcImm, group1),
 	G(ByteOp | DstMem | SrcImm | No64, group1),
 	G(DstMem | SrcImmByte, group1),
-	F2bv(DstMem | SrcReg | ModRM | NoWrite, em_test),
+	I2bv(DstMem | SrcReg | ModRM | NoWrite, em_test),
 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_xchg),
 	/* 0x88 - 0x8F */
 	I2bv(DstMem | SrcReg | ModRM | Mov | PageTable, em_mov),
@@ -4329,7 +4338,7 @@ static const struct opcode opcode_table[
 	I2bv(SrcSI | DstDI | Mov | String | TwoMemOp, em_mov),
 	F2bv(SrcSI | DstDI | String | NoWrite | TwoMemOp, em_cmp_r),
 	/* 0xA8 - 0xAF */
-	F2bv(DstAcc | SrcImm | NoWrite, em_test),
+	I2bv(DstAcc | SrcImm | NoWrite, em_test),
 	I2bv(SrcAcc | DstDI | Mov | String, em_mov),
 	I2bv(SrcSI | DstAcc | Mov | String, em_mov),
 	F2bv(SrcAcc | DstDI | String | NoWrite, em_cmp_r),
@@ -4467,7 +4476,7 @@ static const struct opcode twobyte_table
 	I(DstReg | SrcMem | ModRM, em_bsr_c),
 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
 	/* 0xC0 - 0xC7 */
-	F2bv(DstMem | SrcReg | ModRM | SrcWrite | Lock, em_xadd),
+	I2bv(DstMem | SrcReg | ModRM | SrcWrite | Lock, em_xadd),
 	N, ID(0, &instr_dual_0f_c3),
 	N, N, N, GD(0, &group9),
 	/* 0xC8 - 0xCF */