target/i386: convert MOV from/to CR and DR to new decoder

Complete implementation of C and D operand types, then the operations
are just MOVs.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386/tcg/decode-new.c.inc

@@ -151,6 +151,8 @@
     X86_OP_GROUP3(op, op0, s0, 2op, s0, op1, s1, ## __VA_ARGS__)
 #define X86_OP_GROUPw(op, op0, s0, ...)                           \
     X86_OP_GROUP3(op, op0, s0, None, None, None, None, ## __VA_ARGS__)
+#define X86_OP_GROUPwr(op, op0, s0, op1, s1, ...)                 \
+    X86_OP_GROUP3(op, op0, s0, op1, s1, None, None, ## __VA_ARGS__)
 #define X86_OP_GROUP0(op, ...)                                    \
     X86_OP_GROUP3(op, None, None, None, None, None, None, ## __VA_ARGS__)
 
@@ -985,6 +987,24 @@ static void decode_0FE6(DisasContext *s, CPUX86State *env, X86OpEntry *entry, ui
     *entry = *decode_by_prefix(s, opcodes_0FE6);
 }
 
+/*
+ * These ignore the mod bits (assume (modrm&0xc0)==0xc0), so group the
+ * pre-decode tweak here for all MOVs from/to CR and DR.
+ *
+ * AMD documentation (24594.pdf) and testing of Intel 386 and 486
+ * processors all show that the mod bits are assumed to be 1's,
+ * regardless of actual values.
+ */
+static void decode_MOV_CR_DR(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b)
+{
+    /*
+     */
+    get_modrm(s, env);
+    s->modrm |= 0xC0;
+
+    entry->gen = gen_MOV;
+}
+
 static const X86OpEntry opcodes_0F[256] = {
     [0x0E] = X86_OP_ENTRY0(EMMS,                              cpuid(3DNOW)), /* femms */
     /*
@@ -1004,6 +1024,15 @@ static const X86OpEntry opcodes_0F[256] = {
     /* Incorrectly listed as Mq,Vq in the manual */
     [0x17] = X86_OP_ENTRY3(VMOVHPx_st,  M,q, None,None, V,dq, vex5 p_00_66),
 
+    /*
+     * Incorrectly listed as using "d" operand type in the manual.  In reality
+     * there's no 16-bit version (like y) and it does not use REX.W (like d64).
+     */
+    [0x20] = X86_OP_GROUPwr(MOV_CR_DR,   R,y_d64, C,y_d64, chk(cpl0) svm(READ_CR0)),
+    [0x21] = X86_OP_GROUPwr(MOV_CR_DR,   R,y_d64, D,y_d64, chk(cpl0) svm(READ_DR0)),
+    [0x22] = X86_OP_GROUPwr(MOV_CR_DR,   C,y_d64, R,y_d64, zextT0 chk(cpl0) svm(WRITE_CR0)),
+    [0x23] = X86_OP_GROUPwr(MOV_CR_DR,   D,y_d64, R,y_d64, zextT0 chk(cpl0) svm(WRITE_DR0)),
+
     [0x40] = X86_OP_ENTRY2(CMOVcc,     G,v, E,v, cpuid(CMOV)),
     [0x41] = X86_OP_ENTRY2(CMOVcc,     G,v, E,v, cpuid(CMOV)),
     [0x42] = X86_OP_ENTRY2(CMOVcc,     G,v, E,v, cpuid(CMOV)),
@@ -1725,6 +1754,10 @@ static bool decode_op_size(DisasContext *s, X86OpEntry *e, X86OpSize size, MemOp
         *ot = s->dflag == MO_16 ? MO_32 : s->dflag;
         return true;
 
+    case X86_SIZE_y_d64:  /* Full (not 16-bit) register access */
+        *ot = CODE64(s) ? MO_64 : MO_32;
+        return true;
+
     case X86_SIZE_z:  /* 16-bit for 16-bit operand size, else 32-bit */
         *ot = s->dflag == MO_16 ? MO_16 : MO_32;
         return true;
@@ -1802,11 +1835,34 @@ static bool decode_op(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
 
     case X86_TYPE_C:  /* REG in the modrm byte selects a control register */
         op->unit = X86_OP_CR;
-        goto get_reg;
+        op->n = ((get_modrm(s, env) >> 3) & 7) | REX_R(s);
+        if (op->n == 0 && (s->prefix & PREFIX_LOCK) &&
+            (s->cpuid_ext3_features & CPUID_EXT3_CR8LEG)) {
+            op->n = 8;
+            s->prefix &= ~PREFIX_LOCK;
+        }
+        if (op->n != 0 && op->n != 2 && op->n != 3 && op->n != 4 && op->n != 8) {
+            return false;
+        }
+        if (decode->e.intercept) {
+            decode->e.intercept += op->n;
+        }
+        break;
 
     case X86_TYPE_D:  /* REG in the modrm byte selects a debug register */
         op->unit = X86_OP_DR;
-        goto get_reg;
+        op->n = ((get_modrm(s, env) >> 3) & 7) | REX_R(s);
+        if (op->n >= 8) {
+            /*
+             * illegal opcode.  The DR4 and DR5 case is checked in the generated
+             * code instead, to save on hflags bits.
+             */
+            return false;
+        }
+        if (decode->e.intercept) {
+            decode->e.intercept += op->n;
+        }
+        break;
 
     case X86_TYPE_G:  /* REG in the modrm byte selects a GPR */
         op->unit = X86_OP_INT;
@@ -2431,7 +2487,6 @@ static void disas_insn(DisasContext *s, CPUState *cpu)
             case 0x00 ... 0x03: /* mostly privileged instructions */
             case 0x05 ... 0x09:
             case 0x1a ... 0x1b: /* MPX */
-            case 0x20 ... 0x23: /* mov from/to CR and DR */
             case 0x30 ... 0x35: /* more privileged instructions */
             case 0xa2 ... 0xa5: /* CPUID, BT, SHLD */
             case 0xaa ... 0xae: /* RSM, SHRD, grp15 */

target/i386/tcg/decode-new.h

@@ -90,6 +90,7 @@ typedef enum X86OpSize {
     X86_SIZE_w,  /* 16-bit */
     X86_SIZE_x,  /* 128/256-bit, based on operand size */
     X86_SIZE_y,  /* 32/64-bit, based on operand size */
+    X86_SIZE_y_d64,  /* 32/64-bit, based on 64-bit mode */
     X86_SIZE_z,  /* 16-bit for 16-bit operand size, else 32-bit */
     X86_SIZE_z_f64,  /* 32-bit for 32-bit operand size or 64-bit mode, else 16-bit */
 

target/i386/tcg/emit.c.inc

@@ -242,12 +242,20 @@ static void gen_load(DisasContext *s, X86DecodedInsn *decode, int opn, TCGv v)
         tcg_gen_ld32u_tl(v, tcg_env,
                          offsetof(CPUX86State,segs[op->n].selector));
         break;
+#ifndef CONFIG_USER_ONLY
     case X86_OP_CR:
+        if (op->n == 8) {
+            translator_io_start(&s->base);
+            gen_helper_read_crN(v, tcg_env, tcg_constant_i32(op->n));
+        } else {
             tcg_gen_ld_tl(v, tcg_env, offsetof(CPUX86State, cr[op->n]));
+        }
         break;
     case X86_OP_DR:
-        tcg_gen_ld_tl(v, tcg_env, offsetof(CPUX86State, dr[op->n]));
+        /* CR4.DE tested in the helper.  */
+        gen_helper_get_dr(v, tcg_env, tcg_constant_i32(op->n));
         break;
+#endif
     case X86_OP_INT:
         if (op->has_ea) {
             if (v == s->T0 && decode->e.special == X86_SPECIAL_SExtT0) {
@@ -343,8 +351,20 @@ static void gen_writeback(DisasContext *s, X86DecodedInsn *decode, int opn, TCGv
                                  16, 16, 0);
         }
         break;
+#ifndef CONFIG_USER_ONLY
     case X86_OP_CR:
+        if (op->n == 8) {
+            translator_io_start(&s->base);
+        }
+        gen_helper_write_crN(tcg_env, tcg_constant_i32(op->n), v);
+        s->base.is_jmp = DISAS_EOB_NEXT;
+        break;
     case X86_OP_DR:
+        /* CR4.DE tested in the helper.  */
+        gen_helper_set_dr(tcg_env, tcg_constant_i32(op->n), v);
+        s->base.is_jmp = DISAS_EOB_NEXT;
+        break;
+#endif
     default:
         g_assert_not_reached();
     }

target/i386/tcg/translate.c

@@ -247,9 +247,6 @@ STUB_HELPER(outb, TCGv_env env, TCGv_i32 port, TCGv_i32 val)
 STUB_HELPER(outw, TCGv_env env, TCGv_i32 port, TCGv_i32 val)
 STUB_HELPER(outl, TCGv_env env, TCGv_i32 port, TCGv_i32 val)
 STUB_HELPER(rdmsr, TCGv_env env)
-STUB_HELPER(read_crN, TCGv ret, TCGv_env env, TCGv_i32 reg)
-STUB_HELPER(get_dr, TCGv ret, TCGv_env env, TCGv_i32 reg)
-STUB_HELPER(set_dr, TCGv_env env, TCGv_i32 reg, TCGv val)
 STUB_HELPER(stgi, TCGv_env env)
 STUB_HELPER(svm_check_intercept, TCGv_env env, TCGv_i32 type)
 STUB_HELPER(vmload, TCGv_env env, TCGv_i32 aflag)
@@ -4192,82 +4189,6 @@ static void disas_insn_old(DisasContext *s, CPUState *cpu, int b)
         gen_nop_modrm(env, s, modrm);
         break;
 
-    case 0x120: /* mov reg, crN */
-    case 0x122: /* mov crN, reg */
-        if (!check_cpl0(s)) {
-            break;
-        }
-        modrm = x86_ldub_code(env, s);
-        /*
-         * Ignore the mod bits (assume (modrm&0xc0)==0xc0).
-         * AMD documentation (24594.pdf) and testing of Intel 386 and 486
-         * processors all show that the mod bits are assumed to be 1's,
-         * regardless of actual values.
-         */
-        rm = (modrm & 7) | REX_B(s);
-        reg = ((modrm >> 3) & 7) | REX_R(s);
-        switch (reg) {
-        case 0:
-            if ((prefixes & PREFIX_LOCK) &&
-                (s->cpuid_ext3_features & CPUID_EXT3_CR8LEG)) {
-                reg = 8;
-            }
-            break;
-        case 2:
-        case 3:
-        case 4:
-        case 8:
-            break;
-        default:
-            goto unknown_op;
-        }
-        ot  = (CODE64(s) ? MO_64 : MO_32);
-
-        translator_io_start(&s->base);
-        if (b & 2) {
-            gen_svm_check_intercept(s, SVM_EXIT_WRITE_CR0 + reg);
-            gen_op_mov_v_reg(s, ot, s->T0, rm);
-            gen_helper_write_crN(tcg_env, tcg_constant_i32(reg), s->T0);
-            s->base.is_jmp = DISAS_EOB_NEXT;
-        } else {
-            gen_svm_check_intercept(s, SVM_EXIT_READ_CR0 + reg);
-            gen_helper_read_crN(s->T0, tcg_env, tcg_constant_i32(reg));
-            gen_op_mov_reg_v(s, ot, rm, s->T0);
-        }
-        break;
-
-    case 0x121: /* mov reg, drN */
-    case 0x123: /* mov drN, reg */
-        if (check_cpl0(s)) {
-            modrm = x86_ldub_code(env, s);
-            /* Ignore the mod bits (assume (modrm&0xc0)==0xc0).
-             * AMD documentation (24594.pdf) and testing of
-             * intel 386 and 486 processors all show that the mod bits
-             * are assumed to be 1's, regardless of actual values.
-             */
-            rm = (modrm & 7) | REX_B(s);
-            reg = ((modrm >> 3) & 7) | REX_R(s);
-            if (CODE64(s))
-                ot = MO_64;
-            else
-                ot = MO_32;
-            if (reg >= 8) {
-                goto illegal_op;
-            }
-            if (b & 2) {
-                gen_svm_check_intercept(s, SVM_EXIT_WRITE_DR0 + reg);
-                gen_op_mov_v_reg(s, ot, s->T0, rm);
-                tcg_gen_movi_i32(s->tmp2_i32, reg);
-                gen_helper_set_dr(tcg_env, s->tmp2_i32, s->T0);
-                s->base.is_jmp = DISAS_EOB_NEXT;
-            } else {
-                gen_svm_check_intercept(s, SVM_EXIT_READ_DR0 + reg);
-                tcg_gen_movi_i32(s->tmp2_i32, reg);
-                gen_helper_get_dr(s->T0, tcg_env, s->tmp2_i32);
-                gen_op_mov_reg_v(s, ot, rm, s->T0);
-            }
-        }
-        break;
     case 0x106: /* clts */
         if (check_cpl0(s)) {
             gen_svm_check_intercept(s, SVM_EXIT_WRITE_CR0);