]> Git Repo - qemu.git/blobdiff - target-arm/translate.c
target-arm: Remove remaining old cp15 infrastructure
[qemu.git] / target-arm / translate.c
index f5507ec3b695c7812c85d25ce310d1be09858e51..a2a0ecddad9a662f280a3259269dea5f2aacd7b2 100644 (file)
@@ -25,7 +25,6 @@
 #include <inttypes.h>
 
 #include "cpu.h"
-#include "exec-all.h"
 #include "disas.h"
 #include "tcg-op.h"
 #include "qemu-log.h"
@@ -60,6 +59,7 @@ typedef struct DisasContext {
     struct TranslationBlock *tb;
     int singlestep_enabled;
     int thumb;
+    int bswap_code;
 #if !defined(CONFIG_USER_ONLY)
     int user;
 #endif
@@ -112,20 +112,20 @@ void arm_translate_init(void)
 
     for (i = 0; i < 16; i++) {
         cpu_R[i] = tcg_global_mem_new_i32(TCG_AREG0,
-                                          offsetof(CPUState, regs[i]),
+                                          offsetof(CPUARMState, regs[i]),
                                           regnames[i]);
     }
     cpu_exclusive_addr = tcg_global_mem_new_i32(TCG_AREG0,
-        offsetof(CPUState, exclusive_addr), "exclusive_addr");
+        offsetof(CPUARMState, exclusive_addr), "exclusive_addr");
     cpu_exclusive_val = tcg_global_mem_new_i32(TCG_AREG0,
-        offsetof(CPUState, exclusive_val), "exclusive_val");
+        offsetof(CPUARMState, exclusive_val), "exclusive_val");
     cpu_exclusive_high = tcg_global_mem_new_i32(TCG_AREG0,
-        offsetof(CPUState, exclusive_high), "exclusive_high");
+        offsetof(CPUARMState, exclusive_high), "exclusive_high");
 #ifdef CONFIG_USER_ONLY
     cpu_exclusive_test = tcg_global_mem_new_i32(TCG_AREG0,
-        offsetof(CPUState, exclusive_test), "exclusive_test");
+        offsetof(CPUARMState, exclusive_test), "exclusive_test");
     cpu_exclusive_info = tcg_global_mem_new_i32(TCG_AREG0,
-        offsetof(CPUState, exclusive_info), "exclusive_info");
+        offsetof(CPUARMState, exclusive_info), "exclusive_info");
 #endif
 
 #define GEN_HELPER 2
@@ -139,7 +139,7 @@ static inline TCGv load_cpu_offset(int offset)
     return tmp;
 }
 
-#define load_cpu_field(name) load_cpu_offset(offsetof(CPUState, name))
+#define load_cpu_field(name) load_cpu_offset(offsetof(CPUARMState, name))
 
 static inline void store_cpu_offset(TCGv var, int offset)
 {
@@ -148,7 +148,7 @@ static inline void store_cpu_offset(TCGv var, int offset)
 }
 
 #define store_cpu_field(var, name) \
-    store_cpu_offset(var, offsetof(CPUState, name))
+    store_cpu_offset(var, offsetof(CPUARMState, name))
 
 /* Set a variable to the value of a CPU register.  */
 static void load_reg_var(DisasContext *s, TCGv var, int reg)
@@ -369,7 +369,7 @@ static void gen_add16(TCGv t0, TCGv t1)
     tcg_temp_free_i32(t1);
 }
 
-#define gen_set_CF(var) tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, CF))
+#define gen_set_CF(var) tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, CF))
 
 /* Set CF to the top bit of var.  */
 static void gen_set_CF_bit31(TCGv var)
@@ -383,8 +383,8 @@ static void gen_set_CF_bit31(TCGv var)
 /* Set N and Z flags from var.  */
 static inline void gen_logic_CC(TCGv var)
 {
-    tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, NF));
-    tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, ZF));
+    tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, NF));
+    tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, ZF));
 }
 
 /* T0 += T1 + CF.  */
@@ -524,13 +524,13 @@ static void gen_arm_parallel_addsub(int op1, int op2, TCGv a, TCGv b)
 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
     case 1:
         tmp = tcg_temp_new_ptr();
-        tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
+        tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
         PAS_OP(s)
         tcg_temp_free_ptr(tmp);
         break;
     case 5:
         tmp = tcg_temp_new_ptr();
-        tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
+        tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
         PAS_OP(u)
         tcg_temp_free_ptr(tmp);
         break;
@@ -571,13 +571,13 @@ static void gen_thumb2_parallel_addsub(int op1, int op2, TCGv a, TCGv b)
 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
     case 0:
         tmp = tcg_temp_new_ptr();
-        tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
+        tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
         PAS_OP(s)
         tcg_temp_free_ptr(tmp);
         break;
     case 4:
         tmp = tcg_temp_new_ptr();
-        tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
+        tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
         PAS_OP(u)
         tcg_temp_free_ptr(tmp);
         break;
@@ -726,7 +726,7 @@ static inline void gen_bx_im(DisasContext *s, uint32_t addr)
     if (s->thumb != (addr & 1)) {
         tmp = tcg_temp_new_i32();
         tcg_gen_movi_i32(tmp, addr & 1);
-        tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUState, thumb));
+        tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUARMState, thumb));
         tcg_temp_free_i32(tmp);
     }
     tcg_gen_movi_i32(cpu_R[15], addr & ~1);
@@ -744,7 +744,7 @@ static inline void gen_bx(DisasContext *s, TCGv var)
 /* Variant of store_reg which uses branch&exchange logic when storing
    to r15 in ARM architecture v7 and above. The source must be a temporary
    and will be marked as dead. */
-static inline void store_reg_bx(CPUState *env, DisasContext *s,
+static inline void store_reg_bx(CPUARMState *env, DisasContext *s,
                                 int reg, TCGv var)
 {
     if (reg == 15 && ENABLE_ARCH_7) {
@@ -758,7 +758,7 @@ static inline void store_reg_bx(CPUState *env, DisasContext *s,
  * to r15 in ARM architecture v5T and above. This is used for storing
  * the results of a LDR/LDM/POP into r15, and corresponds to the cases
  * in the ARM ARM which use the LoadWritePC() pseudocode function. */
-static inline void store_reg_from_load(CPUState *env, DisasContext *s,
+static inline void store_reg_from_load(CPUARMState *env, DisasContext *s,
                                 int reg, TCGv var)
 {
     if (reg == 15 && ENABLE_ARCH_5) {
@@ -893,13 +893,29 @@ static inline void gen_add_datah_offset(DisasContext *s, unsigned int insn,
     }
 }
 
+static TCGv_ptr get_fpstatus_ptr(int neon)
+{
+    TCGv_ptr statusptr = tcg_temp_new_ptr();
+    int offset;
+    if (neon) {
+        offset = offsetof(CPUARMState, vfp.standard_fp_status);
+    } else {
+        offset = offsetof(CPUARMState, vfp.fp_status);
+    }
+    tcg_gen_addi_ptr(statusptr, cpu_env, offset);
+    return statusptr;
+}
+
 #define VFP_OP2(name)                                                 \
 static inline void gen_vfp_##name(int dp)                             \
 {                                                                     \
-    if (dp)                                                           \
-        gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, cpu_env); \
-    else                                                              \
-        gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, cpu_env); \
+    TCGv_ptr fpst = get_fpstatus_ptr(0);                              \
+    if (dp) {                                                         \
+        gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, fpst);    \
+    } else {                                                          \
+        gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, fpst);    \
+    }                                                                 \
+    tcg_temp_free_ptr(fpst);                                          \
 }
 
 VFP_OP2(add)
@@ -912,11 +928,13 @@ VFP_OP2(div)
 static inline void gen_vfp_F1_mul(int dp)
 {
     /* Like gen_vfp_mul() but put result in F1 */
+    TCGv_ptr fpst = get_fpstatus_ptr(0);
     if (dp) {
-        gen_helper_vfp_muld(cpu_F1d, cpu_F0d, cpu_F1d, cpu_env);
+        gen_helper_vfp_muld(cpu_F1d, cpu_F0d, cpu_F1d, fpst);
     } else {
-        gen_helper_vfp_muls(cpu_F1s, cpu_F0s, cpu_F1s, cpu_env);
+        gen_helper_vfp_muls(cpu_F1s, cpu_F0s, cpu_F1s, fpst);
     }
+    tcg_temp_free_ptr(fpst);
 }
 
 static inline void gen_vfp_F1_neg(int dp)
@@ -980,14 +998,7 @@ static inline void gen_vfp_F1_ld0(int dp)
 #define VFP_GEN_ITOF(name) \
 static inline void gen_vfp_##name(int dp, int neon) \
 { \
-    TCGv_ptr statusptr = tcg_temp_new_ptr(); \
-    int offset; \
-    if (neon) { \
-        offset = offsetof(CPUState, vfp.standard_fp_status); \
-    } else { \
-        offset = offsetof(CPUState, vfp.fp_status); \
-    } \
-    tcg_gen_addi_ptr(statusptr, cpu_env, offset); \
+    TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
     if (dp) { \
         gen_helper_vfp_##name##d(cpu_F0d, cpu_F0s, statusptr); \
     } else { \
@@ -1003,14 +1014,7 @@ VFP_GEN_ITOF(sito)
 #define VFP_GEN_FTOI(name) \
 static inline void gen_vfp_##name(int dp, int neon) \
 { \
-    TCGv_ptr statusptr = tcg_temp_new_ptr(); \
-    int offset; \
-    if (neon) { \
-        offset = offsetof(CPUState, vfp.standard_fp_status); \
-    } else { \
-        offset = offsetof(CPUState, vfp.fp_status); \
-    } \
-    tcg_gen_addi_ptr(statusptr, cpu_env, offset); \
+    TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
     if (dp) { \
         gen_helper_vfp_##name##d(cpu_F0s, cpu_F0d, statusptr); \
     } else { \
@@ -1029,14 +1033,7 @@ VFP_GEN_FTOI(tosiz)
 static inline void gen_vfp_##name(int dp, int shift, int neon) \
 { \
     TCGv tmp_shift = tcg_const_i32(shift); \
-    TCGv_ptr statusptr = tcg_temp_new_ptr(); \
-    int offset; \
-    if (neon) { \
-        offset = offsetof(CPUState, vfp.standard_fp_status); \
-    } else { \
-        offset = offsetof(CPUState, vfp.fp_status); \
-    } \
-    tcg_gen_addi_ptr(statusptr, cpu_env, offset); \
+    TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
     if (dp) { \
         gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, tmp_shift, statusptr); \
     } else { \
@@ -1151,24 +1148,24 @@ static inline void gen_mov_vreg_F0(int dp, int reg)
 
 static inline void iwmmxt_load_reg(TCGv_i64 var, int reg)
 {
-    tcg_gen_ld_i64(var, cpu_env, offsetof(CPUState, iwmmxt.regs[reg]));
+    tcg_gen_ld_i64(var, cpu_env, offsetof(CPUARMState, iwmmxt.regs[reg]));
 }
 
 static inline void iwmmxt_store_reg(TCGv_i64 var, int reg)
 {
-    tcg_gen_st_i64(var, cpu_env, offsetof(CPUState, iwmmxt.regs[reg]));
+    tcg_gen_st_i64(var, cpu_env, offsetof(CPUARMState, iwmmxt.regs[reg]));
 }
 
 static inline TCGv iwmmxt_load_creg(int reg)
 {
     TCGv var = tcg_temp_new_i32();
-    tcg_gen_ld_i32(var, cpu_env, offsetof(CPUState, iwmmxt.cregs[reg]));
+    tcg_gen_ld_i32(var, cpu_env, offsetof(CPUARMState, iwmmxt.cregs[reg]));
     return var;
 }
 
 static inline void iwmmxt_store_creg(int reg, TCGv var)
 {
-    tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, iwmmxt.cregs[reg]));
+    tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, iwmmxt.cregs[reg]));
     tcg_temp_free_i32(var);
 }
 
@@ -1207,15 +1204,22 @@ static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
     gen_helper_iwmmxt_##name(cpu_M0, cpu_M0, cpu_V1); \
 }
 
-#define IWMMXT_OP_SIZE(name) \
-IWMMXT_OP(name##b) \
-IWMMXT_OP(name##w) \
-IWMMXT_OP(name##l)
+#define IWMMXT_OP_ENV(name) \
+static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
+{ \
+    iwmmxt_load_reg(cpu_V1, rn); \
+    gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0, cpu_V1); \
+}
+
+#define IWMMXT_OP_ENV_SIZE(name) \
+IWMMXT_OP_ENV(name##b) \
+IWMMXT_OP_ENV(name##w) \
+IWMMXT_OP_ENV(name##l)
 
-#define IWMMXT_OP_1(name) \
+#define IWMMXT_OP_ENV1(name) \
 static inline void gen_op_iwmmxt_##name##_M0(void) \
 { \
-    gen_helper_iwmmxt_##name(cpu_M0, cpu_M0); \
+    gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0); \
 }
 
 IWMMXT_OP(maddsq)
@@ -1229,51 +1233,51 @@ IWMMXT_OP(muluhw)
 IWMMXT_OP(macsw)
 IWMMXT_OP(macuw)
 
-IWMMXT_OP_SIZE(unpackl)
-IWMMXT_OP_SIZE(unpackh)
-
-IWMMXT_OP_1(unpacklub)
-IWMMXT_OP_1(unpackluw)
-IWMMXT_OP_1(unpacklul)
-IWMMXT_OP_1(unpackhub)
-IWMMXT_OP_1(unpackhuw)
-IWMMXT_OP_1(unpackhul)
-IWMMXT_OP_1(unpacklsb)
-IWMMXT_OP_1(unpacklsw)
-IWMMXT_OP_1(unpacklsl)
-IWMMXT_OP_1(unpackhsb)
-IWMMXT_OP_1(unpackhsw)
-IWMMXT_OP_1(unpackhsl)
-
-IWMMXT_OP_SIZE(cmpeq)
-IWMMXT_OP_SIZE(cmpgtu)
-IWMMXT_OP_SIZE(cmpgts)
-
-IWMMXT_OP_SIZE(mins)
-IWMMXT_OP_SIZE(minu)
-IWMMXT_OP_SIZE(maxs)
-IWMMXT_OP_SIZE(maxu)
-
-IWMMXT_OP_SIZE(subn)
-IWMMXT_OP_SIZE(addn)
-IWMMXT_OP_SIZE(subu)
-IWMMXT_OP_SIZE(addu)
-IWMMXT_OP_SIZE(subs)
-IWMMXT_OP_SIZE(adds)
-
-IWMMXT_OP(avgb0)
-IWMMXT_OP(avgb1)
-IWMMXT_OP(avgw0)
-IWMMXT_OP(avgw1)
+IWMMXT_OP_ENV_SIZE(unpackl)
+IWMMXT_OP_ENV_SIZE(unpackh)
+
+IWMMXT_OP_ENV1(unpacklub)
+IWMMXT_OP_ENV1(unpackluw)
+IWMMXT_OP_ENV1(unpacklul)
+IWMMXT_OP_ENV1(unpackhub)
+IWMMXT_OP_ENV1(unpackhuw)
+IWMMXT_OP_ENV1(unpackhul)
+IWMMXT_OP_ENV1(unpacklsb)
+IWMMXT_OP_ENV1(unpacklsw)
+IWMMXT_OP_ENV1(unpacklsl)
+IWMMXT_OP_ENV1(unpackhsb)
+IWMMXT_OP_ENV1(unpackhsw)
+IWMMXT_OP_ENV1(unpackhsl)
+
+IWMMXT_OP_ENV_SIZE(cmpeq)
+IWMMXT_OP_ENV_SIZE(cmpgtu)
+IWMMXT_OP_ENV_SIZE(cmpgts)
+
+IWMMXT_OP_ENV_SIZE(mins)
+IWMMXT_OP_ENV_SIZE(minu)
+IWMMXT_OP_ENV_SIZE(maxs)
+IWMMXT_OP_ENV_SIZE(maxu)
+
+IWMMXT_OP_ENV_SIZE(subn)
+IWMMXT_OP_ENV_SIZE(addn)
+IWMMXT_OP_ENV_SIZE(subu)
+IWMMXT_OP_ENV_SIZE(addu)
+IWMMXT_OP_ENV_SIZE(subs)
+IWMMXT_OP_ENV_SIZE(adds)
+
+IWMMXT_OP_ENV(avgb0)
+IWMMXT_OP_ENV(avgb1)
+IWMMXT_OP_ENV(avgw0)
+IWMMXT_OP_ENV(avgw1)
 
 IWMMXT_OP(msadb)
 
-IWMMXT_OP(packuw)
-IWMMXT_OP(packul)
-IWMMXT_OP(packuq)
-IWMMXT_OP(packsw)
-IWMMXT_OP(packsl)
-IWMMXT_OP(packsq)
+IWMMXT_OP_ENV(packuw)
+IWMMXT_OP_ENV(packul)
+IWMMXT_OP_ENV(packuq)
+IWMMXT_OP_ENV(packsw)
+IWMMXT_OP_ENV(packsl)
+IWMMXT_OP_ENV(packsq)
 
 static void gen_op_iwmmxt_set_mup(void)
 {
@@ -1363,7 +1367,7 @@ static inline int gen_iwmmxt_shift(uint32_t insn, uint32_t mask, TCGv dest)
 
 /* Disassemble an iwMMXt instruction.  Returns nonzero if an error occurred
    (ie. an undefined instruction).  */
-static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
+static int disas_iwmmxt_insn(CPUARMState *env, DisasContext *s, uint32_t insn)
 {
     int rd, wrd;
     int rdhi, rdlo, rd0, rd1, i;
@@ -2007,13 +2011,13 @@ static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
         }
         switch ((insn >> 22) & 3) {
         case 1:
-            gen_helper_iwmmxt_srlw(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_srlw(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         case 2:
-            gen_helper_iwmmxt_srll(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_srll(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         case 3:
-            gen_helper_iwmmxt_srlq(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_srlq(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         }
         tcg_temp_free_i32(tmp);
@@ -2035,13 +2039,13 @@ static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
         }
         switch ((insn >> 22) & 3) {
         case 1:
-            gen_helper_iwmmxt_sraw(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_sraw(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         case 2:
-            gen_helper_iwmmxt_sral(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_sral(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         case 3:
-            gen_helper_iwmmxt_sraq(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_sraq(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         }
         tcg_temp_free_i32(tmp);
@@ -2063,13 +2067,13 @@ static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
         }
         switch ((insn >> 22) & 3) {
         case 1:
-            gen_helper_iwmmxt_sllw(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_sllw(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         case 2:
-            gen_helper_iwmmxt_slll(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_slll(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         case 3:
-            gen_helper_iwmmxt_sllq(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_sllq(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         }
         tcg_temp_free_i32(tmp);
@@ -2091,21 +2095,21 @@ static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
                 tcg_temp_free_i32(tmp);
                 return 1;
             }
-            gen_helper_iwmmxt_rorw(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_rorw(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         case 2:
             if (gen_iwmmxt_shift(insn, 0x1f, tmp)) {
                 tcg_temp_free_i32(tmp);
                 return 1;
             }
-            gen_helper_iwmmxt_rorl(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_rorl(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         case 3:
             if (gen_iwmmxt_shift(insn, 0x3f, tmp)) {
                 tcg_temp_free_i32(tmp);
                 return 1;
             }
-            gen_helper_iwmmxt_rorq(cpu_M0, cpu_M0, tmp);
+            gen_helper_iwmmxt_rorq(cpu_M0, cpu_env, cpu_M0, tmp);
             break;
         }
         tcg_temp_free_i32(tmp);
@@ -2239,7 +2243,7 @@ static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
         rd0 = (insn >> 16) & 0xf;
         gen_op_iwmmxt_movq_M0_wRn(rd0);
         tmp = tcg_const_i32(((insn >> 16) & 0xf0) | (insn & 0x0f));
-        gen_helper_iwmmxt_shufh(cpu_M0, cpu_M0, tmp);
+        gen_helper_iwmmxt_shufh(cpu_M0, cpu_env, cpu_M0, tmp);
         tcg_temp_free(tmp);
         gen_op_iwmmxt_movq_wRn_M0(wrd);
         gen_op_iwmmxt_set_mup();
@@ -2367,7 +2371,7 @@ static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
 
 /* Disassemble an XScale DSP instruction.  Returns nonzero if an error occurred
    (ie. an undefined instruction).  */
-static int disas_dsp_insn(CPUState *env, DisasContext *s, uint32_t insn)
+static int disas_dsp_insn(CPUARMState *env, DisasContext *s, uint32_t insn)
 {
     int acc, rd0, rd1, rdhi, rdlo;
     TCGv tmp, tmp2;
@@ -2435,195 +2439,6 @@ static int disas_dsp_insn(CPUState *env, DisasContext *s, uint32_t insn)
     return 1;
 }
 
-/* Disassemble system coprocessor instruction.  Return nonzero if
-   instruction is not defined.  */
-static int disas_cp_insn(CPUState *env, DisasContext *s, uint32_t insn)
-{
-    TCGv tmp, tmp2;
-    uint32_t rd = (insn >> 12) & 0xf;
-    uint32_t cp = (insn >> 8) & 0xf;
-    if (IS_USER(s)) {
-        return 1;
-    }
-
-    if (insn & ARM_CP_RW_BIT) {
-        if (!env->cp[cp].cp_read)
-            return 1;
-        gen_set_pc_im(s->pc);
-        tmp = tcg_temp_new_i32();
-        tmp2 = tcg_const_i32(insn);
-        gen_helper_get_cp(tmp, cpu_env, tmp2);
-        tcg_temp_free(tmp2);
-        store_reg(s, rd, tmp);
-    } else {
-        if (!env->cp[cp].cp_write)
-            return 1;
-        gen_set_pc_im(s->pc);
-        tmp = load_reg(s, rd);
-        tmp2 = tcg_const_i32(insn);
-        gen_helper_set_cp(cpu_env, tmp2, tmp);
-        tcg_temp_free(tmp2);
-        tcg_temp_free_i32(tmp);
-    }
-    return 0;
-}
-
-static int cp15_user_ok(uint32_t insn)
-{
-    int cpn = (insn >> 16) & 0xf;
-    int cpm = insn & 0xf;
-    int op = ((insn >> 5) & 7) | ((insn >> 18) & 0x38);
-
-    if (cpn == 13 && cpm == 0) {
-        /* TLS register.  */
-        if (op == 2 || (op == 3 && (insn & ARM_CP_RW_BIT)))
-            return 1;
-    }
-    if (cpn == 7) {
-        /* ISB, DSB, DMB.  */
-        if ((cpm == 5 && op == 4)
-                || (cpm == 10 && (op == 4 || op == 5)))
-            return 1;
-    }
-    return 0;
-}
-
-static int cp15_tls_load_store(CPUState *env, DisasContext *s, uint32_t insn, uint32_t rd)
-{
-    TCGv tmp;
-    int cpn = (insn >> 16) & 0xf;
-    int cpm = insn & 0xf;
-    int op = ((insn >> 5) & 7) | ((insn >> 18) & 0x38);
-
-    if (!arm_feature(env, ARM_FEATURE_V6K))
-        return 0;
-
-    if (!(cpn == 13 && cpm == 0))
-        return 0;
-
-    if (insn & ARM_CP_RW_BIT) {
-        switch (op) {
-        case 2:
-            tmp = load_cpu_field(cp15.c13_tls1);
-            break;
-        case 3:
-            tmp = load_cpu_field(cp15.c13_tls2);
-            break;
-        case 4:
-            tmp = load_cpu_field(cp15.c13_tls3);
-            break;
-        default:
-            return 0;
-        }
-        store_reg(s, rd, tmp);
-
-    } else {
-        tmp = load_reg(s, rd);
-        switch (op) {
-        case 2:
-            store_cpu_field(tmp, cp15.c13_tls1);
-            break;
-        case 3:
-            store_cpu_field(tmp, cp15.c13_tls2);
-            break;
-        case 4:
-            store_cpu_field(tmp, cp15.c13_tls3);
-            break;
-        default:
-            tcg_temp_free_i32(tmp);
-            return 0;
-        }
-    }
-    return 1;
-}
-
-/* Disassemble system coprocessor (cp15) instruction.  Return nonzero if
-   instruction is not defined.  */
-static int disas_cp15_insn(CPUState *env, DisasContext *s, uint32_t insn)
-{
-    uint32_t rd;
-    TCGv tmp, tmp2;
-
-    /* M profile cores use memory mapped registers instead of cp15.  */
-    if (arm_feature(env, ARM_FEATURE_M))
-       return 1;
-
-    if ((insn & (1 << 25)) == 0) {
-        if (insn & (1 << 20)) {
-            /* mrrc */
-            return 1;
-        }
-        /* mcrr.  Used for block cache operations, so implement as no-op.  */
-        return 0;
-    }
-    if ((insn & (1 << 4)) == 0) {
-        /* cdp */
-        return 1;
-    }
-    if (IS_USER(s) && !cp15_user_ok(insn)) {
-        return 1;
-    }
-
-    /* Pre-v7 versions of the architecture implemented WFI via coprocessor
-     * instructions rather than a separate instruction.
-     */
-    if ((insn & 0x0fff0fff) == 0x0e070f90) {
-        /* 0,c7,c0,4: Standard v6 WFI (also used in some pre-v6 cores).
-         * In v7, this must NOP.
-         */
-        if (!arm_feature(env, ARM_FEATURE_V7)) {
-            /* Wait for interrupt.  */
-            gen_set_pc_im(s->pc);
-            s->is_jmp = DISAS_WFI;
-        }
-        return 0;
-    }
-
-    if ((insn & 0x0fff0fff) == 0x0e070f58) {
-        /* 0,c7,c8,2: Not all pre-v6 cores implemented this WFI,
-         * so this is slightly over-broad.
-         */
-        if (!arm_feature(env, ARM_FEATURE_V6)) {
-            /* Wait for interrupt.  */
-            gen_set_pc_im(s->pc);
-            s->is_jmp = DISAS_WFI;
-            return 0;
-        }
-        /* Otherwise fall through to handle via helper function.
-         * In particular, on v7 and some v6 cores this is one of
-         * the VA-PA registers.
-         */
-    }
-
-    rd = (insn >> 12) & 0xf;
-
-    if (cp15_tls_load_store(env, s, insn, rd))
-        return 0;
-
-    tmp2 = tcg_const_i32(insn);
-    if (insn & ARM_CP_RW_BIT) {
-        tmp = tcg_temp_new_i32();
-        gen_helper_get_cp15(tmp, cpu_env, tmp2);
-        /* If the destination register is r15 then sets condition codes.  */
-        if (rd != 15)
-            store_reg(s, rd, tmp);
-        else
-            tcg_temp_free_i32(tmp);
-    } else {
-        tmp = load_reg(s, rd);
-        gen_helper_set_cp15(cpu_env, tmp2, tmp);
-        tcg_temp_free_i32(tmp);
-        /* Normally we would always end the TB here, but Linux
-         * arch/arm/mach-pxa/sleep.S expects two instructions following
-         * an MMU enable to execute from cache.  Imitate this behaviour.  */
-        if (!arm_feature(env, ARM_FEATURE_XSCALE) ||
-                (insn & 0x0fff0fff) != 0x0e010f10)
-            gen_lookup_tb(s);
-    }
-    tcg_temp_free_i32(tmp2);
-    return 0;
-}
-
 #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n))
 #define VFP_SREG(insn, bigbit, smallbit) \
   ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1))
@@ -2713,7 +2528,7 @@ static TCGv gen_load_and_replicate(DisasContext *s, TCGv addr, int size)
 
 /* Disassemble a VFP instruction.  Returns nonzero if an error occurred
    (ie. an undefined instruction).  */
-static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
+static int disas_vfp_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
 {
     uint32_t rd, rn, rm, op, i, n, offset, delta_d, delta_m, bank_mask;
     int dp, veclen;
@@ -2872,7 +2687,7 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                         case ARM_VFP_MVFR0:
                         case ARM_VFP_MVFR1:
                             if (IS_USER(s)
-                                || !arm_feature(env, ARM_FEATURE_VFP3))
+                                || !arm_feature(env, ARM_FEATURE_MVFR))
                                 return 1;
                             tmp = load_cpu_field(vfp.xregs[rn]);
                             break;
@@ -3037,6 +2852,17 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     /* Source and destination the same.  */
                     gen_mov_F0_vreg(dp, rd);
                     break;
+                case 4:
+                case 5:
+                case 6:
+                case 7:
+                    /* VCVTB, VCVTT: only present with the halfprec extension,
+                     * UNPREDICTABLE if bit 8 is set (we choose to UNDEF)
+                     */
+                    if (dp || !arm_feature(env, ARM_FEATURE_VFP_FP16)) {
+                        return 1;
+                    }
+                    /* Otherwise fall through */
                 default:
                     /* One source operand.  */
                     gen_mov_F0_vreg(dp, rm);
@@ -3096,6 +2922,57 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                 case 8: /* div: fn / fm */
                     gen_vfp_div(dp);
                     break;
+                case 10: /* VFNMA : fd = muladd(-fd,  fn, fm) */
+                case 11: /* VFNMS : fd = muladd(-fd, -fn, fm) */
+                case 12: /* VFMA  : fd = muladd( fd,  fn, fm) */
+                case 13: /* VFMS  : fd = muladd( fd, -fn, fm) */
+                    /* These are fused multiply-add, and must be done as one
+                     * floating point operation with no rounding between the
+                     * multiplication and addition steps.
+                     * NB that doing the negations here as separate steps is
+                     * correct : an input NaN should come out with its sign bit
+                     * flipped if it is a negated-input.
+                     */
+                    if (!arm_feature(env, ARM_FEATURE_VFP4)) {
+                        return 1;
+                    }
+                    if (dp) {
+                        TCGv_ptr fpst;
+                        TCGv_i64 frd;
+                        if (op & 1) {
+                            /* VFNMS, VFMS */
+                            gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
+                        }
+                        frd = tcg_temp_new_i64();
+                        tcg_gen_ld_f64(frd, cpu_env, vfp_reg_offset(dp, rd));
+                        if (op & 2) {
+                            /* VFNMA, VFNMS */
+                            gen_helper_vfp_negd(frd, frd);
+                        }
+                        fpst = get_fpstatus_ptr(0);
+                        gen_helper_vfp_muladdd(cpu_F0d, cpu_F0d,
+                                               cpu_F1d, frd, fpst);
+                        tcg_temp_free_ptr(fpst);
+                        tcg_temp_free_i64(frd);
+                    } else {
+                        TCGv_ptr fpst;
+                        TCGv_i32 frd;
+                        if (op & 1) {
+                            /* VFNMS, VFMS */
+                            gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
+                        }
+                        frd = tcg_temp_new_i32();
+                        tcg_gen_ld_f32(frd, cpu_env, vfp_reg_offset(dp, rd));
+                        if (op & 2) {
+                            gen_helper_vfp_negs(frd, frd);
+                        }
+                        fpst = get_fpstatus_ptr(0);
+                        gen_helper_vfp_muladds(cpu_F0s, cpu_F0s,
+                                               cpu_F1s, frd, fpst);
+                        tcg_temp_free_ptr(fpst);
+                        tcg_temp_free_i32(frd);
+                    }
+                    break;
                 case 14: /* fconst */
                     if (!arm_feature(env, ARM_FEATURE_VFP3))
                       return 1;
@@ -3133,24 +3010,18 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                         gen_vfp_sqrt(dp);
                         break;
                     case 4: /* vcvtb.f32.f16 */
-                        if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
-                          return 1;
                         tmp = gen_vfp_mrs();
                         tcg_gen_ext16u_i32(tmp, tmp);
                         gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp, cpu_env);
                         tcg_temp_free_i32(tmp);
                         break;
                     case 5: /* vcvtt.f32.f16 */
-                        if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
-                          return 1;
                         tmp = gen_vfp_mrs();
                         tcg_gen_shri_i32(tmp, tmp, 16);
                         gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp, cpu_env);
                         tcg_temp_free_i32(tmp);
                         break;
                     case 6: /* vcvtb.f16.f32 */
-                        if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
-                          return 1;
                         tmp = tcg_temp_new_i32();
                         gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
                         gen_mov_F0_vreg(0, rd);
@@ -3161,8 +3032,6 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                         gen_vfp_msr(tmp);
                         break;
                     case 7: /* vcvtt.f16.f32 */
-                        if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
-                          return 1;
                         tmp = tcg_temp_new_i32();
                         gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
                         tcg_gen_shli_i32(tmp, tmp, 16);
@@ -3251,12 +3120,10 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                         gen_vfp_toul(dp, 32 - rm, 0);
                         break;
                     default: /* undefined */
-                        printf ("rn:%d\n", rn);
                         return 1;
                     }
                     break;
                 default: /* undefined */
-                    printf ("op:%d\n", op);
                     return 1;
                 }
 
@@ -3363,17 +3230,18 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                 VFP_DREG_D(rd, insn);
             else
                 rd = VFP_SREG_D(insn);
-            if (s->thumb && rn == 15) {
-                addr = tcg_temp_new_i32();
-                tcg_gen_movi_i32(addr, s->pc & ~2);
-            } else {
-                addr = load_reg(s, rn);
-            }
             if ((insn & 0x01200000) == 0x01000000) {
                 /* Single load/store */
                 offset = (insn & 0xff) << 2;
                 if ((insn & (1 << 23)) == 0)
                     offset = -offset;
+                if (s->thumb && rn == 15) {
+                    /* This is actually UNPREDICTABLE */
+                    addr = tcg_temp_new_i32();
+                    tcg_gen_movi_i32(addr, s->pc & ~2);
+                } else {
+                    addr = load_reg(s, rn);
+                }
                 tcg_gen_addi_i32(addr, addr, offset);
                 if (insn & (1 << 20)) {
                     gen_vfp_ld(s, dp, addr);
@@ -3385,11 +3253,34 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                 tcg_temp_free_i32(addr);
             } else {
                 /* load/store multiple */
+                int w = insn & (1 << 21);
                 if (dp)
                     n = (insn >> 1) & 0x7f;
                 else
                     n = insn & 0xff;
 
+                if (w && !(((insn >> 23) ^ (insn >> 24)) & 1)) {
+                    /* P == U , W == 1  => UNDEF */
+                    return 1;
+                }
+                if (n == 0 || (rd + n) > 32 || (dp && n > 16)) {
+                    /* UNPREDICTABLE cases for bad immediates: we choose to
+                     * UNDEF to avoid generating huge numbers of TCG ops
+                     */
+                    return 1;
+                }
+                if (rn == 15 && w) {
+                    /* writeback to PC is UNPREDICTABLE, we choose to UNDEF */
+                    return 1;
+                }
+
+                if (s->thumb && rn == 15) {
+                    /* This is actually UNPREDICTABLE */
+                    addr = tcg_temp_new_i32();
+                    tcg_gen_movi_i32(addr, s->pc & ~2);
+                } else {
+                    addr = load_reg(s, rn);
+                }
                 if (insn & (1 << 24)) /* pre-decrement */
                     tcg_gen_addi_i32(addr, addr, -((insn & 0xff) << 2));
 
@@ -3409,7 +3300,7 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     }
                     tcg_gen_addi_i32(addr, addr, offset);
                 }
-                if (insn & (1 << 21)) {
+                if (w) {
                     /* writeback */
                     if (insn & (1 << 24))
                         offset = -offset * n;
@@ -3476,7 +3367,7 @@ static inline void gen_mulxy(TCGv t0, TCGv t1, int x, int y)
 }
 
 /* Return the mask of PSR bits set by a MSR instruction.  */
-static uint32_t msr_mask(CPUState *env, DisasContext *s, int flags, int spsr) {
+static uint32_t msr_mask(CPUARMState *env, DisasContext *s, int flags, int spsr) {
     uint32_t mask;
 
     mask = 0;
@@ -3707,13 +3598,13 @@ static int gen_neon_unzip(int rd, int rm, int size, int q)
     if (q) {
         switch (size) {
         case 0:
-            gen_helper_neon_qunzip8(tmp, tmp2);
+            gen_helper_neon_qunzip8(cpu_env, tmp, tmp2);
             break;
         case 1:
-            gen_helper_neon_qunzip16(tmp, tmp2);
+            gen_helper_neon_qunzip16(cpu_env, tmp, tmp2);
             break;
         case 2:
-            gen_helper_neon_qunzip32(tmp, tmp2);
+            gen_helper_neon_qunzip32(cpu_env, tmp, tmp2);
             break;
         default:
             abort();
@@ -3721,10 +3612,10 @@ static int gen_neon_unzip(int rd, int rm, int size, int q)
     } else {
         switch (size) {
         case 0:
-            gen_helper_neon_unzip8(tmp, tmp2);
+            gen_helper_neon_unzip8(cpu_env, tmp, tmp2);
             break;
         case 1:
-            gen_helper_neon_unzip16(tmp, tmp2);
+            gen_helper_neon_unzip16(cpu_env, tmp, tmp2);
             break;
         default:
             abort();
@@ -3746,13 +3637,13 @@ static int gen_neon_zip(int rd, int rm, int size, int q)
     if (q) {
         switch (size) {
         case 0:
-            gen_helper_neon_qzip8(tmp, tmp2);
+            gen_helper_neon_qzip8(cpu_env, tmp, tmp2);
             break;
         case 1:
-            gen_helper_neon_qzip16(tmp, tmp2);
+            gen_helper_neon_qzip16(cpu_env, tmp, tmp2);
             break;
         case 2:
-            gen_helper_neon_qzip32(tmp, tmp2);
+            gen_helper_neon_qzip32(cpu_env, tmp, tmp2);
             break;
         default:
             abort();
@@ -3760,10 +3651,10 @@ static int gen_neon_zip(int rd, int rm, int size, int q)
     } else {
         switch (size) {
         case 0:
-            gen_helper_neon_zip8(tmp, tmp2);
+            gen_helper_neon_zip8(cpu_env, tmp, tmp2);
             break;
         case 1:
-            gen_helper_neon_zip16(tmp, tmp2);
+            gen_helper_neon_zip16(cpu_env, tmp, tmp2);
             break;
         default:
             abort();
@@ -3836,7 +3727,7 @@ static struct {
 
 /* Translate a NEON load/store element instruction.  Return nonzero if the
    instruction is invalid.  */
-static int disas_neon_ls_insn(CPUState * env, DisasContext *s, uint32_t insn)
+static int disas_neon_ls_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
 {
     int rd, rn, rm;
     int op;
@@ -4163,9 +4054,9 @@ static inline void gen_neon_narrow(int size, TCGv dest, TCGv_i64 src)
 static inline void gen_neon_narrow_sats(int size, TCGv dest, TCGv_i64 src)
 {
     switch (size) {
-    case 0: gen_helper_neon_narrow_sat_s8(dest, src); break;
-    case 1: gen_helper_neon_narrow_sat_s16(dest, src); break;
-    case 2: gen_helper_neon_narrow_sat_s32(dest, src); break;
+    case 0: gen_helper_neon_narrow_sat_s8(dest, cpu_env, src); break;
+    case 1: gen_helper_neon_narrow_sat_s16(dest, cpu_env, src); break;
+    case 2: gen_helper_neon_narrow_sat_s32(dest, cpu_env, src); break;
     default: abort();
     }
 }
@@ -4173,9 +4064,9 @@ static inline void gen_neon_narrow_sats(int size, TCGv dest, TCGv_i64 src)
 static inline void gen_neon_narrow_satu(int size, TCGv dest, TCGv_i64 src)
 {
     switch (size) {
-    case 0: gen_helper_neon_narrow_sat_u8(dest, src); break;
-    case 1: gen_helper_neon_narrow_sat_u16(dest, src); break;
-    case 2: gen_helper_neon_narrow_sat_u32(dest, src); break;
+    case 0: gen_helper_neon_narrow_sat_u8(dest, cpu_env, src); break;
+    case 1: gen_helper_neon_narrow_sat_u16(dest, cpu_env, src); break;
+    case 2: gen_helper_neon_narrow_sat_u32(dest, cpu_env, src); break;
     default: abort();
     }
 }
@@ -4183,9 +4074,9 @@ static inline void gen_neon_narrow_satu(int size, TCGv dest, TCGv_i64 src)
 static inline void gen_neon_unarrow_sats(int size, TCGv dest, TCGv_i64 src)
 {
     switch (size) {
-    case 0: gen_helper_neon_unarrow_sat8(dest, src); break;
-    case 1: gen_helper_neon_unarrow_sat16(dest, src); break;
-    case 2: gen_helper_neon_unarrow_sat32(dest, src); break;
+    case 0: gen_helper_neon_unarrow_sat8(dest, cpu_env, src); break;
+    case 1: gen_helper_neon_unarrow_sat16(dest, cpu_env, src); break;
+    case 2: gen_helper_neon_unarrow_sat32(dest, cpu_env, src); break;
     default: abort();
     }
 }
@@ -4277,8 +4168,8 @@ static inline void gen_neon_negl(TCGv_i64 var, int size)
 static inline void gen_neon_addl_saturate(TCGv_i64 op0, TCGv_i64 op1, int size)
 {
     switch (size) {
-    case 1: gen_helper_neon_addl_saturate_s32(op0, op0, op1); break;
-    case 2: gen_helper_neon_addl_saturate_s64(op0, op0, op1); break;
+    case 1: gen_helper_neon_addl_saturate_s32(op0, cpu_env, op0, op1); break;
+    case 2: gen_helper_neon_addl_saturate_s64(op0, cpu_env, op0, op1); break;
     default: abort();
     }
 }
@@ -4358,6 +4249,7 @@ static void gen_neon_narrow_op(int op, int u, int size, TCGv dest, TCGv_i64 src)
 #define NEON_3R_VPMIN 21
 #define NEON_3R_VQDMULH_VQRDMULH 22
 #define NEON_3R_VPADD 23
+#define NEON_3R_VFM 25 /* VFMA, VFMS : float fused multiply-add */
 #define NEON_3R_FLOAT_ARITH 26 /* float VADD, VSUB, VPADD, VABD */
 #define NEON_3R_FLOAT_MULTIPLY 27 /* float VMLA, VMLS, VMUL */
 #define NEON_3R_FLOAT_CMP 28 /* float VCEQ, VCGE, VCGT */
@@ -4390,6 +4282,7 @@ static const uint8_t neon_3r_sizes[] = {
     [NEON_3R_VPMIN] = 0x7,
     [NEON_3R_VQDMULH_VQRDMULH] = 0x6,
     [NEON_3R_VPADD] = 0x7,
+    [NEON_3R_VFM] = 0x5, /* size bit 1 encodes op */
     [NEON_3R_FLOAT_ARITH] = 0x5, /* size bit 1 encodes op */
     [NEON_3R_FLOAT_MULTIPLY] = 0x5, /* size bit 1 encodes op */
     [NEON_3R_FLOAT_CMP] = 0x5, /* size bit 1 encodes op */
@@ -4510,7 +4403,7 @@ static const uint8_t neon_2rm_sizes[] = {
    We process data in a mixture of 32-bit and 64-bit chunks.
    Mostly we use 32-bit chunks so we can use normal scalar instructions.  */
 
-static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
+static int disas_neon_data_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
 {
     int op;
     int q;
@@ -4554,16 +4447,20 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                 switch (op) {
                 case NEON_3R_VQADD:
                     if (u) {
-                        gen_helper_neon_qadd_u64(cpu_V0, cpu_V0, cpu_V1);
+                        gen_helper_neon_qadd_u64(cpu_V0, cpu_env,
+                                                 cpu_V0, cpu_V1);
                     } else {
-                        gen_helper_neon_qadd_s64(cpu_V0, cpu_V0, cpu_V1);
+                        gen_helper_neon_qadd_s64(cpu_V0, cpu_env,
+                                                 cpu_V0, cpu_V1);
                     }
                     break;
                 case NEON_3R_VQSUB:
                     if (u) {
-                        gen_helper_neon_qsub_u64(cpu_V0, cpu_V0, cpu_V1);
+                        gen_helper_neon_qsub_u64(cpu_V0, cpu_env,
+                                                 cpu_V0, cpu_V1);
                     } else {
-                        gen_helper_neon_qsub_s64(cpu_V0, cpu_V0, cpu_V1);
+                        gen_helper_neon_qsub_s64(cpu_V0, cpu_env,
+                                                 cpu_V0, cpu_V1);
                     }
                     break;
                 case NEON_3R_VSHL:
@@ -4575,9 +4472,11 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     break;
                 case NEON_3R_VQSHL:
                     if (u) {
-                        gen_helper_neon_qshl_u64(cpu_V0, cpu_V1, cpu_V0);
+                        gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
+                                                 cpu_V1, cpu_V0);
                     } else {
-                        gen_helper_neon_qshl_s64(cpu_V0, cpu_V1, cpu_V0);
+                        gen_helper_neon_qshl_s64(cpu_V0, cpu_env,
+                                                 cpu_V1, cpu_V0);
                     }
                     break;
                 case NEON_3R_VRSHL:
@@ -4589,9 +4488,11 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                     break;
                 case NEON_3R_VQRSHL:
                     if (u) {
-                        gen_helper_neon_qrshl_u64(cpu_V0, cpu_V1, cpu_V0);
+                        gen_helper_neon_qrshl_u64(cpu_V0, cpu_env,
+                                                  cpu_V1, cpu_V0);
                     } else {
-                        gen_helper_neon_qrshl_s64(cpu_V0, cpu_V1, cpu_V0);
+                        gen_helper_neon_qrshl_s64(cpu_V0, cpu_env,
+                                                  cpu_V1, cpu_V0);
                     }
                     break;
                 case NEON_3R_VADD_VSUB:
@@ -4659,6 +4560,11 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                 return 1;
             }
             break;
+        case NEON_3R_VFM:
+            if (!arm_feature(env, ARM_FEATURE_VFP4) || u) {
+                return 1;
+            }
+            break;
         default:
             break;
         }
@@ -4689,7 +4595,7 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
             GEN_NEON_INTEGER_OP(hadd);
             break;
         case NEON_3R_VQADD:
-            GEN_NEON_INTEGER_OP(qadd);
+            GEN_NEON_INTEGER_OP_ENV(qadd);
             break;
         case NEON_3R_VRHADD:
             GEN_NEON_INTEGER_OP(rhadd);
@@ -4732,7 +4638,7 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
             GEN_NEON_INTEGER_OP(hsub);
             break;
         case NEON_3R_VQSUB:
-            GEN_NEON_INTEGER_OP(qsub);
+            GEN_NEON_INTEGER_OP_ENV(qsub);
             break;
         case NEON_3R_VCGT:
             GEN_NEON_INTEGER_OP(cgt);
@@ -4744,13 +4650,13 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
             GEN_NEON_INTEGER_OP(shl);
             break;
         case NEON_3R_VQSHL:
-            GEN_NEON_INTEGER_OP(qshl);
+            GEN_NEON_INTEGER_OP_ENV(qshl);
             break;
         case NEON_3R_VRSHL:
             GEN_NEON_INTEGER_OP(rshl);
             break;
         case NEON_3R_VQRSHL:
-            GEN_NEON_INTEGER_OP(qrshl);
+            GEN_NEON_INTEGER_OP_ENV(qrshl);
             break;
         case NEON_3R_VMAX:
             GEN_NEON_INTEGER_OP(max);
@@ -4832,14 +4738,22 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
         case NEON_3R_VQDMULH_VQRDMULH: /* Multiply high.  */
             if (!u) { /* VQDMULH */
                 switch (size) {
-                case 1: gen_helper_neon_qdmulh_s16(tmp, tmp, tmp2); break;
-                case 2: gen_helper_neon_qdmulh_s32(tmp, tmp, tmp2); break;
+                case 1:
+                    gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2);
+                    break;
+                case 2:
+                    gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2);
+                    break;
                 default: abort();
                 }
             } else { /* VQRDMULH */
                 switch (size) {
-                case 1: gen_helper_neon_qrdmulh_s16(tmp, tmp, tmp2); break;
-                case 2: gen_helper_neon_qrdmulh_s32(tmp, tmp, tmp2); break;
+                case 1:
+                    gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2);
+                    break;
+                case 2:
+                    gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2);
+                    break;
                 default: abort();
                 }
             }
@@ -4853,63 +4767,98 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
             }
             break;
         case NEON_3R_FLOAT_ARITH: /* Floating point arithmetic. */
+        {
+            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
             switch ((u << 2) | size) {
             case 0: /* VADD */
-                gen_helper_neon_add_f32(tmp, tmp, tmp2);
+            case 4: /* VPADD */
+                gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
                 break;
             case 2: /* VSUB */
-                gen_helper_neon_sub_f32(tmp, tmp, tmp2);
-                break;
-            case 4: /* VPADD */
-                gen_helper_neon_add_f32(tmp, tmp, tmp2);
+                gen_helper_vfp_subs(tmp, tmp, tmp2, fpstatus);
                 break;
             case 6: /* VABD */
-                gen_helper_neon_abd_f32(tmp, tmp, tmp2);
+                gen_helper_neon_abd_f32(tmp, tmp, tmp2, fpstatus);
                 break;
             default:
                 abort();
             }
+            tcg_temp_free_ptr(fpstatus);
             break;
+        }
         case NEON_3R_FLOAT_MULTIPLY:
-            gen_helper_neon_mul_f32(tmp, tmp, tmp2);
+        {
+            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
+            gen_helper_vfp_muls(tmp, tmp, tmp2, fpstatus);
             if (!u) {
                 tcg_temp_free_i32(tmp2);
                 tmp2 = neon_load_reg(rd, pass);
                 if (size == 0) {
-                    gen_helper_neon_add_f32(tmp, tmp, tmp2);
+                    gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
                 } else {
-                    gen_helper_neon_sub_f32(tmp, tmp2, tmp);
+                    gen_helper_vfp_subs(tmp, tmp2, tmp, fpstatus);
                 }
             }
+            tcg_temp_free_ptr(fpstatus);
             break;
+        }
         case NEON_3R_FLOAT_CMP:
+        {
+            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
             if (!u) {
-                gen_helper_neon_ceq_f32(tmp, tmp, tmp2);
+                gen_helper_neon_ceq_f32(tmp, tmp, tmp2, fpstatus);
             } else {
-                if (size == 0)
-                    gen_helper_neon_cge_f32(tmp, tmp, tmp2);
-                else
-                    gen_helper_neon_cgt_f32(tmp, tmp, tmp2);
+                if (size == 0) {
+                    gen_helper_neon_cge_f32(tmp, tmp, tmp2, fpstatus);
+                } else {
+                    gen_helper_neon_cgt_f32(tmp, tmp, tmp2, fpstatus);
+                }
             }
+            tcg_temp_free_ptr(fpstatus);
             break;
+        }
         case NEON_3R_FLOAT_ACMP:
-            if (size == 0)
-                gen_helper_neon_acge_f32(tmp, tmp, tmp2);
-            else
-                gen_helper_neon_acgt_f32(tmp, tmp, tmp2);
+        {
+            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
+            if (size == 0) {
+                gen_helper_neon_acge_f32(tmp, tmp, tmp2, fpstatus);
+            } else {
+                gen_helper_neon_acgt_f32(tmp, tmp, tmp2, fpstatus);
+            }
+            tcg_temp_free_ptr(fpstatus);
             break;
+        }
         case NEON_3R_FLOAT_MINMAX:
-            if (size == 0)
-                gen_helper_neon_max_f32(tmp, tmp, tmp2);
-            else
-                gen_helper_neon_min_f32(tmp, tmp, tmp2);
+        {
+            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
+            if (size == 0) {
+                gen_helper_neon_max_f32(tmp, tmp, tmp2, fpstatus);
+            } else {
+                gen_helper_neon_min_f32(tmp, tmp, tmp2, fpstatus);
+            }
+            tcg_temp_free_ptr(fpstatus);
             break;
+        }
         case NEON_3R_VRECPS_VRSQRTS:
             if (size == 0)
                 gen_helper_recps_f32(tmp, tmp, tmp2, cpu_env);
             else
                 gen_helper_rsqrts_f32(tmp, tmp, tmp2, cpu_env);
             break;
+        case NEON_3R_VFM:
+        {
+            /* VFMA, VFMS: fused multiply-add */
+            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
+            TCGv_i32 tmp3 = neon_load_reg(rd, pass);
+            if (size) {
+                /* VFMS */
+                gen_helper_vfp_negs(tmp, tmp);
+            }
+            gen_helper_vfp_muladds(tmp, tmp, tmp2, tmp3, fpstatus);
+            tcg_temp_free_i32(tmp3);
+            tcg_temp_free_ptr(fpstatus);
+            break;
+        }
         default:
             abort();
         }
@@ -5010,14 +4959,15 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                             gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
                             break;
                         case 6: /* VQSHLU */
-                            gen_helper_neon_qshlu_s64(cpu_V0, cpu_V0, cpu_V1);
+                            gen_helper_neon_qshlu_s64(cpu_V0, cpu_env,
+                                                      cpu_V0, cpu_V1);
                             break;
                         case 7: /* VQSHL */
                             if (u) {
-                                gen_helper_neon_qshl_u64(cpu_V0,
+                                gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
                                                          cpu_V0, cpu_V1);
                             } else {
-                                gen_helper_neon_qshl_s64(cpu_V0,
+                                gen_helper_neon_qshl_s64(cpu_V0, cpu_env,
                                                          cpu_V0, cpu_V1);
                             }
                             break;
@@ -5069,20 +5019,23 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                         case 6: /* VQSHLU */
                             switch (size) {
                             case 0:
-                                gen_helper_neon_qshlu_s8(tmp, tmp, tmp2);
+                                gen_helper_neon_qshlu_s8(tmp, cpu_env,
+                                                         tmp, tmp2);
                                 break;
                             case 1:
-                                gen_helper_neon_qshlu_s16(tmp, tmp, tmp2);
+                                gen_helper_neon_qshlu_s16(tmp, cpu_env,
+                                                          tmp, tmp2);
                                 break;
                             case 2:
-                                gen_helper_neon_qshlu_s32(tmp, tmp, tmp2);
+                                gen_helper_neon_qshlu_s32(tmp, cpu_env,
+                                                          tmp, tmp2);
                                 break;
                             default:
                                 abort();
                             }
                             break;
                         case 7: /* VQSHL */
-                            GEN_NEON_INTEGER_OP(qshl);
+                            GEN_NEON_INTEGER_OP_ENV(qshl);
                             break;
                         }
                         tcg_temp_free_i32(tmp2);
@@ -5591,18 +5544,20 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                         tmp2 = neon_load_reg(rn, pass);
                         if (op == 12) {
                             if (size == 1) {
-                                gen_helper_neon_qdmulh_s16(tmp, tmp, tmp2);
+                                gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2);
                             } else {
-                                gen_helper_neon_qdmulh_s32(tmp, tmp, tmp2);
+                                gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2);
                             }
                         } else if (op == 13) {
                             if (size == 1) {
-                                gen_helper_neon_qrdmulh_s16(tmp, tmp, tmp2);
+                                gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2);
                             } else {
-                                gen_helper_neon_qrdmulh_s32(tmp, tmp, tmp2);
+                                gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2);
                             }
                         } else if (op & 1) {
-                            gen_helper_neon_mul_f32(tmp, tmp, tmp2);
+                            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
+                            gen_helper_vfp_muls(tmp, tmp, tmp2, fpstatus);
+                            tcg_temp_free_ptr(fpstatus);
                         } else {
                             switch (size) {
                             case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
@@ -5620,14 +5575,22 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                                 gen_neon_add(size, tmp, tmp2);
                                 break;
                             case 1:
-                                gen_helper_neon_add_f32(tmp, tmp, tmp2);
+                            {
+                                TCGv_ptr fpstatus = get_fpstatus_ptr(1);
+                                gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
+                                tcg_temp_free_ptr(fpstatus);
                                 break;
+                            }
                             case 4:
                                 gen_neon_rsb(size, tmp, tmp2);
                                 break;
                             case 5:
-                                gen_helper_neon_sub_f32(tmp, tmp2, tmp);
+                            {
+                                TCGv_ptr fpstatus = get_fpstatus_ptr(1);
+                                gen_helper_vfp_subs(tmp, tmp2, tmp, fpstatus);
+                                tcg_temp_free_ptr(fpstatus);
                                 break;
+                            }
                             default:
                                 abort();
                             }
@@ -5961,17 +5924,29 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                             break;
                         case NEON_2RM_VQABS:
                             switch (size) {
-                            case 0: gen_helper_neon_qabs_s8(tmp, tmp); break;
-                            case 1: gen_helper_neon_qabs_s16(tmp, tmp); break;
-                            case 2: gen_helper_neon_qabs_s32(tmp, tmp); break;
+                            case 0:
+                                gen_helper_neon_qabs_s8(tmp, cpu_env, tmp);
+                                break;
+                            case 1:
+                                gen_helper_neon_qabs_s16(tmp, cpu_env, tmp);
+                                break;
+                            case 2:
+                                gen_helper_neon_qabs_s32(tmp, cpu_env, tmp);
+                                break;
                             default: abort();
                             }
                             break;
                         case NEON_2RM_VQNEG:
                             switch (size) {
-                            case 0: gen_helper_neon_qneg_s8(tmp, tmp); break;
-                            case 1: gen_helper_neon_qneg_s16(tmp, tmp); break;
-                            case 2: gen_helper_neon_qneg_s32(tmp, tmp); break;
+                            case 0:
+                                gen_helper_neon_qneg_s8(tmp, cpu_env, tmp);
+                                break;
+                            case 1:
+                                gen_helper_neon_qneg_s16(tmp, cpu_env, tmp);
+                                break;
+                            case 2:
+                                gen_helper_neon_qneg_s32(tmp, cpu_env, tmp);
+                                break;
                             default: abort();
                             }
                             break;
@@ -6025,30 +6000,50 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                             tcg_temp_free(tmp2);
                             break;
                         case NEON_2RM_VCGT0_F:
+                        {
+                            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
                             tmp2 = tcg_const_i32(0);
-                            gen_helper_neon_cgt_f32(tmp, tmp, tmp2);
+                            gen_helper_neon_cgt_f32(tmp, tmp, tmp2, fpstatus);
                             tcg_temp_free(tmp2);
+                            tcg_temp_free_ptr(fpstatus);
                             break;
+                        }
                         case NEON_2RM_VCGE0_F:
+                        {
+                            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
                             tmp2 = tcg_const_i32(0);
-                            gen_helper_neon_cge_f32(tmp, tmp, tmp2);
+                            gen_helper_neon_cge_f32(tmp, tmp, tmp2, fpstatus);
                             tcg_temp_free(tmp2);
+                            tcg_temp_free_ptr(fpstatus);
                             break;
+                        }
                         case NEON_2RM_VCEQ0_F:
+                        {
+                            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
                             tmp2 = tcg_const_i32(0);
-                            gen_helper_neon_ceq_f32(tmp, tmp, tmp2);
+                            gen_helper_neon_ceq_f32(tmp, tmp, tmp2, fpstatus);
                             tcg_temp_free(tmp2);
+                            tcg_temp_free_ptr(fpstatus);
                             break;
+                        }
                         case NEON_2RM_VCLE0_F:
+                        {
+                            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
                             tmp2 = tcg_const_i32(0);
-                            gen_helper_neon_cge_f32(tmp, tmp2, tmp);
+                            gen_helper_neon_cge_f32(tmp, tmp2, tmp, fpstatus);
                             tcg_temp_free(tmp2);
+                            tcg_temp_free_ptr(fpstatus);
                             break;
+                        }
                         case NEON_2RM_VCLT0_F:
+                        {
+                            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
                             tmp2 = tcg_const_i32(0);
-                            gen_helper_neon_cgt_f32(tmp, tmp2, tmp);
+                            gen_helper_neon_cgt_f32(tmp, tmp2, tmp, fpstatus);
                             tcg_temp_free(tmp2);
+                            tcg_temp_free_ptr(fpstatus);
                             break;
+                        }
                         case NEON_2RM_VABS_F:
                             gen_vfp_abs(0);
                             break;
@@ -6173,108 +6168,18 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
     return 0;
 }
 
-static int disas_cp14_read(CPUState * env, DisasContext *s, uint32_t insn)
-{
-    int crn = (insn >> 16) & 0xf;
-    int crm = insn & 0xf;
-    int op1 = (insn >> 21) & 7;
-    int op2 = (insn >> 5) & 7;
-    int rt = (insn >> 12) & 0xf;
-    TCGv tmp;
-
-    /* Minimal set of debug registers, since we don't support debug */
-    if (op1 == 0 && crn == 0 && op2 == 0) {
-        switch (crm) {
-        case 0:
-            /* DBGDIDR: just RAZ. In particular this means the
-             * "debug architecture version" bits will read as
-             * a reserved value, which should cause Linux to
-             * not try to use the debug hardware.
-             */
-            tmp = tcg_const_i32(0);
-            store_reg(s, rt, tmp);
-            return 0;
-        case 1:
-        case 2:
-            /* DBGDRAR and DBGDSAR: v7 only. Always RAZ since we
-             * don't implement memory mapped debug components
-             */
-            if (ENABLE_ARCH_7) {
-                tmp = tcg_const_i32(0);
-                store_reg(s, rt, tmp);
-                return 0;
-            }
-            break;
-        default:
-            break;
-        }
-    }
-
-    if (arm_feature(env, ARM_FEATURE_THUMB2EE)) {
-        if (op1 == 6 && crn == 0 && crm == 0 && op2 == 0) {
-            /* TEECR */
-            if (IS_USER(s))
-                return 1;
-            tmp = load_cpu_field(teecr);
-            store_reg(s, rt, tmp);
-            return 0;
-        }
-        if (op1 == 6 && crn == 1 && crm == 0 && op2 == 0) {
-            /* TEEHBR */
-            if (IS_USER(s) && (env->teecr & 1))
-                return 1;
-            tmp = load_cpu_field(teehbr);
-            store_reg(s, rt, tmp);
-            return 0;
-        }
-    }
-    fprintf(stderr, "Unknown cp14 read op1:%d crn:%d crm:%d op2:%d\n",
-            op1, crn, crm, op2);
-    return 1;
-}
-
-static int disas_cp14_write(CPUState * env, DisasContext *s, uint32_t insn)
+static int disas_coproc_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
 {
-    int crn = (insn >> 16) & 0xf;
-    int crm = insn & 0xf;
-    int op1 = (insn >> 21) & 7;
-    int op2 = (insn >> 5) & 7;
-    int rt = (insn >> 12) & 0xf;
-    TCGv tmp;
-
-    if (arm_feature(env, ARM_FEATURE_THUMB2EE)) {
-        if (op1 == 6 && crn == 0 && crm == 0 && op2 == 0) {
-            /* TEECR */
-            if (IS_USER(s))
-                return 1;
-            tmp = load_reg(s, rt);
-            gen_helper_set_teecr(cpu_env, tmp);
-            tcg_temp_free_i32(tmp);
-            return 0;
-        }
-        if (op1 == 6 && crn == 1 && crm == 0 && op2 == 0) {
-            /* TEEHBR */
-            if (IS_USER(s) && (env->teecr & 1))
-                return 1;
-            tmp = load_reg(s, rt);
-            store_cpu_field(tmp, teehbr);
-            return 0;
-        }
-    }
-    fprintf(stderr, "Unknown cp14 write op1:%d crn:%d crm:%d op2:%d\n",
-            op1, crn, crm, op2);
-    return 1;
-}
-
-static int disas_coproc_insn(CPUState * env, DisasContext *s, uint32_t insn)
-{
-    int cpnum;
+    int cpnum, is64, crn, crm, opc1, opc2, isread, rt, rt2;
+    const ARMCPRegInfo *ri;
+    ARMCPU *cpu = arm_env_get_cpu(env);
 
     cpnum = (insn >> 8) & 0xf;
     if (arm_feature(env, ARM_FEATURE_XSCALE)
            && ((env->cp15.c15_cpar ^ 0x3fff) & (1 << cpnum)))
        return 1;
 
+    /* First check for coprocessor space used for actual instructions */
     switch (cpnum) {
       case 0:
       case 1:
@@ -6287,22 +6192,154 @@ static int disas_coproc_insn(CPUState * env, DisasContext *s, uint32_t insn)
     case 10:
     case 11:
        return disas_vfp_insn (env, s, insn);
-    case 14:
-        /* Coprocessors 7-15 are architecturally reserved by ARM.
-           Unfortunately Intel decided to ignore this.  */
-        if (arm_feature(env, ARM_FEATURE_XSCALE))
-            goto board;
-        if (insn & (1 << 20))
-            return disas_cp14_read(env, s, insn);
-        else
-            return disas_cp14_write(env, s, insn);
-    case 15:
-       return disas_cp15_insn (env, s, insn);
     default:
-    board:
-       /* Unknown coprocessor.  See if the board has hooked it.  */
-       return disas_cp_insn (env, s, insn);
+        break;
+    }
+
+    /* Otherwise treat as a generic register access */
+    is64 = (insn & (1 << 25)) == 0;
+    if (!is64 && ((insn & (1 << 4)) == 0)) {
+        /* cdp */
+        return 1;
+    }
+
+    crm = insn & 0xf;
+    if (is64) {
+        crn = 0;
+        opc1 = (insn >> 4) & 0xf;
+        opc2 = 0;
+        rt2 = (insn >> 16) & 0xf;
+    } else {
+        crn = (insn >> 16) & 0xf;
+        opc1 = (insn >> 21) & 7;
+        opc2 = (insn >> 5) & 7;
+        rt2 = 0;
     }
+    isread = (insn >> 20) & 1;
+    rt = (insn >> 12) & 0xf;
+
+    ri = get_arm_cp_reginfo(cpu,
+                            ENCODE_CP_REG(cpnum, is64, crn, crm, opc1, opc2));
+    if (ri) {
+        /* Check access permissions */
+        if (!cp_access_ok(env, ri, isread)) {
+            return 1;
+        }
+
+        /* Handle special cases first */
+        switch (ri->type & ~(ARM_CP_FLAG_MASK & ~ARM_CP_SPECIAL)) {
+        case ARM_CP_NOP:
+            return 0;
+        case ARM_CP_WFI:
+            if (isread) {
+                return 1;
+            }
+            gen_set_pc_im(s->pc);
+            s->is_jmp = DISAS_WFI;
+            break;
+        default:
+            break;
+        }
+
+        if (isread) {
+            /* Read */
+            if (is64) {
+                TCGv_i64 tmp64;
+                TCGv_i32 tmp;
+                if (ri->type & ARM_CP_CONST) {
+                    tmp64 = tcg_const_i64(ri->resetvalue);
+                } else if (ri->readfn) {
+                    TCGv_ptr tmpptr;
+                    gen_set_pc_im(s->pc);
+                    tmp64 = tcg_temp_new_i64();
+                    tmpptr = tcg_const_ptr(ri);
+                    gen_helper_get_cp_reg64(tmp64, cpu_env, tmpptr);
+                    tcg_temp_free_ptr(tmpptr);
+                } else {
+                    tmp64 = tcg_temp_new_i64();
+                    tcg_gen_ld_i64(tmp64, cpu_env, ri->fieldoffset);
+                }
+                tmp = tcg_temp_new_i32();
+                tcg_gen_trunc_i64_i32(tmp, tmp64);
+                store_reg(s, rt, tmp);
+                tcg_gen_shri_i64(tmp64, tmp64, 32);
+                tcg_gen_trunc_i64_i32(tmp, tmp64);
+                store_reg(s, rt2, tmp);
+            } else {
+                TCGv tmp;
+                if (ri->type & ARM_CP_CONST) {
+                    tmp = tcg_const_i32(ri->resetvalue);
+                } else if (ri->readfn) {
+                    TCGv_ptr tmpptr;
+                    gen_set_pc_im(s->pc);
+                    tmp = tcg_temp_new_i32();
+                    tmpptr = tcg_const_ptr(ri);
+                    gen_helper_get_cp_reg(tmp, cpu_env, tmpptr);
+                    tcg_temp_free_ptr(tmpptr);
+                } else {
+                    tmp = load_cpu_offset(ri->fieldoffset);
+                }
+                if (rt == 15) {
+                    /* Destination register of r15 for 32 bit loads sets
+                     * the condition codes from the high 4 bits of the value
+                     */
+                    gen_set_nzcv(tmp);
+                    tcg_temp_free_i32(tmp);
+                } else {
+                    store_reg(s, rt, tmp);
+                }
+            }
+        } else {
+            /* Write */
+            if (ri->type & ARM_CP_CONST) {
+                /* If not forbidden by access permissions, treat as WI */
+                return 0;
+            }
+
+            if (is64) {
+                TCGv tmplo, tmphi;
+                TCGv_i64 tmp64 = tcg_temp_new_i64();
+                tmplo = load_reg(s, rt);
+                tmphi = load_reg(s, rt2);
+                tcg_gen_concat_i32_i64(tmp64, tmplo, tmphi);
+                tcg_temp_free_i32(tmplo);
+                tcg_temp_free_i32(tmphi);
+                if (ri->writefn) {
+                    TCGv_ptr tmpptr = tcg_const_ptr(ri);
+                    gen_set_pc_im(s->pc);
+                    gen_helper_set_cp_reg64(cpu_env, tmpptr, tmp64);
+                    tcg_temp_free_ptr(tmpptr);
+                } else {
+                    tcg_gen_st_i64(tmp64, cpu_env, ri->fieldoffset);
+                }
+                tcg_temp_free_i64(tmp64);
+            } else {
+                if (ri->writefn) {
+                    TCGv tmp;
+                    TCGv_ptr tmpptr;
+                    gen_set_pc_im(s->pc);
+                    tmp = load_reg(s, rt);
+                    tmpptr = tcg_const_ptr(ri);
+                    gen_helper_set_cp_reg(cpu_env, tmpptr, tmp);
+                    tcg_temp_free_ptr(tmpptr);
+                    tcg_temp_free_i32(tmp);
+                } else {
+                    TCGv tmp = load_reg(s, rt);
+                    store_cpu_offset(tmp, ri->fieldoffset);
+                }
+            }
+            /* We default to ending the TB on a coprocessor register write,
+             * but allow this to be suppressed by the register definition
+             * (usually only necessary to work around guest bugs).
+             */
+            if (!(ri->type & ARM_CP_SUPPRESS_TB_END)) {
+                gen_lookup_tb(s);
+            }
+        }
+        return 0;
+    }
+
+    return 1;
 }
 
 
@@ -6486,7 +6523,7 @@ static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
 }
 #endif
 
-static void disas_arm_insn(CPUState * env, DisasContext *s)
+static void disas_arm_insn(CPUARMState * env, DisasContext *s)
 {
     unsigned int cond, insn, val, op1, i, shift, rm, rs, rn, rd, sh;
     TCGv tmp;
@@ -6495,7 +6532,7 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
     TCGv addr;
     TCGv_i64 tmp64;
 
-    insn = ldl_code(s->pc);
+    insn = arm_ldl_code(s->pc, s->bswap_code);
     s->pc += 4;
 
     /* M variants do not implement ARM mode.  */
@@ -6556,8 +6593,8 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
         if ((insn & 0x0ffffdff) == 0x01010000) {
             ARCH(6);
             /* setend */
-            if (insn & (1 << 9)) {
-                /* BE8 mode not implemented.  */
+            if (((insn >> 9) & 1) != s->bswap_code) {
+                /* Dynamic endianness switching not implemented. */
                 goto illegal_op;
             }
             return;
@@ -7377,7 +7414,7 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
                         tmp = load_reg(s, rn);
                         tmp2 = load_reg(s, rm);
                         tmp3 = tcg_temp_new_i32();
-                        tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUState, GE));
+                        tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUARMState, GE));
                         gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
                         tcg_temp_free_i32(tmp3);
                         tcg_temp_free_i32(tmp2);
@@ -7431,11 +7468,16 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
                     }
                     break;
                 case 2: /* Multiplies (Type 3).  */
-                    tmp = load_reg(s, rm);
-                    tmp2 = load_reg(s, rs);
-                    if (insn & (1 << 20)) {
+                    switch ((insn >> 20) & 0x7) {
+                    case 5:
+                        if (((insn >> 6) ^ (insn >> 7)) & 1) {
+                            /* op2 not 00x or 11x : UNDEF */
+                            goto illegal_op;
+                        }
                         /* Signed multiply most significant [accumulate].
                            (SMMUL, SMMLA, SMMLS) */
+                        tmp = load_reg(s, rm);
+                        tmp2 = load_reg(s, rs);
                         tmp64 = gen_muls_i64_i32(tmp, tmp2);
 
                         if (rd != 15) {
@@ -7454,7 +7496,15 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
                         tcg_gen_trunc_i64_i32(tmp, tmp64);
                         tcg_temp_free_i64(tmp64);
                         store_reg(s, rn, tmp);
-                    } else {
+                        break;
+                    case 0:
+                    case 4:
+                        /* SMLAD, SMUAD, SMLSD, SMUSD, SMLALD, SMLSLD */
+                        if (insn & (1 << 7)) {
+                            goto illegal_op;
+                        }
+                        tmp = load_reg(s, rm);
+                        tmp2 = load_reg(s, rs);
                         if (insn & (1 << 5))
                             gen_swap_half(tmp2);
                         gen_smul_dual(tmp, tmp2);
@@ -7487,6 +7537,28 @@ static void disas_arm_insn(CPUState * env, DisasContext *s)
                               }
                             store_reg(s, rn, tmp);
                         }
+                        break;
+                    case 1:
+                    case 3:
+                        /* SDIV, UDIV */
+                        if (!arm_feature(env, ARM_FEATURE_ARM_DIV)) {
+                            goto illegal_op;
+                        }
+                        if (((insn >> 5) & 7) || (rd != 15)) {
+                            goto illegal_op;
+                        }
+                        tmp = load_reg(s, rm);
+                        tmp2 = load_reg(s, rs);
+                        if (insn & (1 << 21)) {
+                            gen_helper_udiv(tmp, tmp, tmp2);
+                        } else {
+                            gen_helper_sdiv(tmp, tmp, tmp2);
+                        }
+                        tcg_temp_free_i32(tmp2);
+                        store_reg(s, rn, tmp);
+                        break;
+                    default:
+                        goto illegal_op;
                     }
                     break;
                 case 3:
@@ -7832,7 +7904,7 @@ gen_thumb2_data_op(DisasContext *s, int op, int conds, uint32_t shifter_out, TCG
 
 /* Translate a 32-bit thumb instruction.  Returns nonzero if the instruction
    is not legal.  */
-static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)
+static int disas_thumb2_insn(CPUARMState *env, DisasContext *s, uint16_t insn_hw1)
 {
     uint32_t insn, imm, shift, offset;
     uint32_t rd, rn, rm, rs;
@@ -7888,7 +7960,7 @@ static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)
         /* Fall through to 32-bit decode.  */
     }
 
-    insn = lduw_code(s->pc);
+    insn = arm_lduw_code(s->pc, s->bswap_code);
     s->pc += 2;
     insn |= (uint32_t)insn_hw1 << 16;
 
@@ -8243,7 +8315,7 @@ static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)
                 case 0x10: /* sel */
                     tmp2 = load_reg(s, rm);
                     tmp3 = tcg_temp_new_i32();
-                    tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUState, GE));
+                    tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUARMState, GE));
                     gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
                     tcg_temp_free_i32(tmp3);
                     tcg_temp_free_i32(tmp2);
@@ -8359,8 +8431,9 @@ static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)
             tmp2 = load_reg(s, rm);
             if ((op & 0x50) == 0x10) {
                 /* sdiv, udiv */
-                if (!arm_feature(env, ARM_FEATURE_DIV))
+                if (!arm_feature(env, ARM_FEATURE_THUMB_DIV)) {
                     goto illegal_op;
+                }
                 if (op & 0x20)
                     gen_helper_udiv(tmp, tmp, tmp2);
                 else
@@ -8779,14 +8852,16 @@ static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)
                     goto illegal_op;
                 }
                 if (rn == 15) {
-                    /* UNPREDICTABLE or unallocated hint */
+                    /* UNPREDICTABLE, unallocated hint or
+                     * PLD/PLDW/PLI (literal)
+                     */
                     return 0;
                 }
                 if (op1 & 1) {
-                    return 0; /* PLD* or unallocated hint */
+                    return 0; /* PLD/PLDW/PLI or unallocated hint */
                 }
                 if ((op2 == 0) || ((op2 & 0x3c) == 0x30)) {
-                    return 0; /* PLD* or unallocated hint */
+                    return 0; /* PLD/PLDW/PLI or unallocated hint */
                 }
                 /* UNDEF space, or an UNPREDICTABLE */
                 return 1;
@@ -8897,7 +8972,7 @@ illegal_op:
     return 1;
 }
 
-static void disas_thumb_insn(CPUState *env, DisasContext *s)
+static void disas_thumb_insn(CPUARMState *env, DisasContext *s)
 {
     uint32_t val, insn, op, rm, rn, rd, shift, cond;
     int32_t offset;
@@ -8915,7 +8990,7 @@ static void disas_thumb_insn(CPUState *env, DisasContext *s)
         }
     }
 
-    insn = lduw_code(s->pc);
+    insn = arm_lduw_code(s->pc, s->bswap_code);
     s->pc += 2;
 
     switch (insn >> 12) {
@@ -9456,32 +9531,49 @@ static void disas_thumb_insn(CPUState *env, DisasContext *s)
             store_reg(s, rd, tmp);
             break;
 
-        case 6: /* cps */
-            ARCH(6);
-            if (IS_USER(s))
+        case 6:
+            switch ((insn >> 5) & 7) {
+            case 2:
+                /* setend */
+                ARCH(6);
+                if (((insn >> 3) & 1) != s->bswap_code) {
+                    /* Dynamic endianness switching not implemented. */
+                    goto illegal_op;
+                }
                 break;
-            if (IS_M(env)) {
-                tmp = tcg_const_i32((insn & (1 << 4)) != 0);
-                /* PRIMASK */
-                if (insn & 1) {
-                    addr = tcg_const_i32(16);
-                    gen_helper_v7m_msr(cpu_env, addr, tmp);
-                    tcg_temp_free_i32(addr);
+            case 3:
+                /* cps */
+                ARCH(6);
+                if (IS_USER(s)) {
+                    break;
                 }
-                /* FAULTMASK */
-                if (insn & 2) {
-                    addr = tcg_const_i32(17);
-                    gen_helper_v7m_msr(cpu_env, addr, tmp);
-                    tcg_temp_free_i32(addr);
+                if (IS_M(env)) {
+                    tmp = tcg_const_i32((insn & (1 << 4)) != 0);
+                    /* FAULTMASK */
+                    if (insn & 1) {
+                        addr = tcg_const_i32(19);
+                        gen_helper_v7m_msr(cpu_env, addr, tmp);
+                        tcg_temp_free_i32(addr);
+                    }
+                    /* PRIMASK */
+                    if (insn & 2) {
+                        addr = tcg_const_i32(16);
+                        gen_helper_v7m_msr(cpu_env, addr, tmp);
+                        tcg_temp_free_i32(addr);
+                    }
+                    tcg_temp_free_i32(tmp);
+                    gen_lookup_tb(s);
+                } else {
+                    if (insn & (1 << 4)) {
+                        shift = CPSR_A | CPSR_I | CPSR_F;
+                    } else {
+                        shift = 0;
+                    }
+                    gen_set_psr_im(s, ((insn & 7) << 6), 0, shift);
                 }
-                tcg_temp_free_i32(tmp);
-                gen_lookup_tb(s);
-            } else {
-                if (insn & (1 << 4))
-                    shift = CPSR_A | CPSR_I | CPSR_F;
-                else
-                    shift = 0;
-                gen_set_psr_im(s, ((insn & 7) << 6), 0, shift);
+                break;
+            default:
+                goto undef;
             }
             break;
 
@@ -9582,7 +9674,7 @@ undef:
 /* generate intermediate code in gen_opc_buf and gen_opparam_buf for
    basic block 'tb'. If search_pc is TRUE, also generate PC
    information for each intermediate instruction. */
-static inline void gen_intermediate_code_internal(CPUState *env,
+static inline void gen_intermediate_code_internal(CPUARMState *env,
                                                   TranslationBlock *tb,
                                                   int search_pc)
 {
@@ -9607,6 +9699,7 @@ static inline void gen_intermediate_code_internal(CPUState *env,
     dc->singlestep_enabled = env->singlestep_enabled;
     dc->condjmp = 0;
     dc->thumb = ARM_TBFLAG_THUMB(tb->flags);
+    dc->bswap_code = ARM_TBFLAG_BSWAP_CODE(tb->flags);
     dc->condexec_mask = (ARM_TBFLAG_CONDEXEC(tb->flags) & 0xf) << 1;
     dc->condexec_cond = ARM_TBFLAG_CONDEXEC(tb->flags) >> 4;
 #if !defined(CONFIG_USER_ONLY)
@@ -9637,22 +9730,22 @@ static inline void gen_intermediate_code_internal(CPUState *env,
     /* A note on handling of the condexec (IT) bits:
      *
      * We want to avoid the overhead of having to write the updated condexec
-     * bits back to the CPUState for every instruction in an IT block. So:
+     * bits back to the CPUARMState for every instruction in an IT block. So:
      * (1) if the condexec bits are not already zero then we write
-     * zero back into the CPUState now. This avoids complications trying
+     * zero back into the CPUARMState now. This avoids complications trying
      * to do it at the end of the block. (For example if we don't do this
      * it's hard to identify whether we can safely skip writing condexec
      * at the end of the TB, which we definitely want to do for the case
      * where a TB doesn't do anything with the IT state at all.)
      * (2) if we are going to leave the TB then we call gen_set_condexec()
-     * which will write the correct value into CPUState if zero is wrong.
+     * which will write the correct value into CPUARMState if zero is wrong.
      * This is done both for leaving the TB at the end, and for leaving
      * it because of an exception we know will happen, which is done in
      * gen_exception_insn(). The latter is necessary because we need to
      * leave the TB with the PC/IT state just prior to execution of the
      * instruction which caused the exception.
      * (3) if we leave the TB unexpectedly (eg a data abort on a load)
-     * then the CPUState will be wrong and we need to reset it.
+     * then the CPUARMState will be wrong and we need to reset it.
      * This is handled in the same way as restoration of the
      * PC in these situations: we will be called again with search_pc=1
      * and generate a mapping of the condexec bits for each PC in
@@ -9661,7 +9754,7 @@ static inline void gen_intermediate_code_internal(CPUState *env,
      *
      * Note that there are no instructions which can read the condexec
      * bits, and none which can write non-static values to them, so
-     * we don't need to care about whether CPUState is correct in the
+     * we don't need to care about whether CPUARMState is correct in the
      * middle of a TB.
      */
 
@@ -9840,7 +9933,8 @@ done_generating:
     if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
         qemu_log("----------------\n");
         qemu_log("IN: %s\n", lookup_symbol(pc_start));
-        log_target_disas(pc_start, dc->pc - pc_start, dc->thumb);
+        log_target_disas(pc_start, dc->pc - pc_start,
+                         dc->thumb | (dc->bswap_code << 1));
         qemu_log("\n");
     }
 #endif
@@ -9855,12 +9949,12 @@ done_generating:
     }
 }
 
-void gen_intermediate_code(CPUState *env, TranslationBlock *tb)
+void gen_intermediate_code(CPUARMState *env, TranslationBlock *tb)
 {
     gen_intermediate_code_internal(env, tb, 0);
 }
 
-void gen_intermediate_code_pc(CPUState *env, TranslationBlock *tb)
+void gen_intermediate_code_pc(CPUARMState *env, TranslationBlock *tb)
 {
     gen_intermediate_code_internal(env, tb, 1);
 }
@@ -9870,7 +9964,7 @@ static const char *cpu_mode_names[16] = {
   "???", "???", "???", "und", "???", "???", "???", "sys"
 };
 
-void cpu_dump_state(CPUState *env, FILE *f, fprintf_function cpu_fprintf,
+void cpu_dump_state(CPUARMState *env, FILE *f, fprintf_function cpu_fprintf,
                     int flags)
 {
     int i;
@@ -9922,7 +10016,7 @@ void cpu_dump_state(CPUState *env, FILE *f, fprintf_function cpu_fprintf,
 #endif
 }
 
-void restore_state_to_opc(CPUState *env, TranslationBlock *tb, int pc_pos)
+void restore_state_to_opc(CPUARMState *env, TranslationBlock *tb, int pc_pos)
 {
     env->regs[15] = gen_opc_pc[pc_pos];
     env->condexec_bits = gen_opc_condexec_bits[pc_pos];
This page took 0.092609 seconds and 4 git commands to generate.