Merge remote-tracking branch 'remotes/amarkovic/tags/mips-queue-may-19-2019-v3' into staging

MIPS queue for May 19th, 2019 - v3

# gpg: Signature made Sun 26 May 2019 17:07:07 BST
# gpg:                using RSA key D4972A8967F75A65
# gpg: Good signature from "Aleksandar Markovic <amarkovic@wavecomp.com>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 8526 FBF1 5DA3 811F 4A01  DD75 D497 2A89 67F7 5A65

* remotes/amarkovic/tags/mips-queue-may-19-2019-v3:
  BootLinuxSshTest: Test some userspace commands on Malta
  target/mips: realign comments to fix checkpatch warnings
  target/mips: add or remove space to fix checkpatch errors
  linux-user: fix __NR_semtimedop undeclared error
  mips: Decide to map PAGE_EXEC in map_address
  target/mips: Refactor and fix INSERT.<B|H|W|D> instructions
  target/mips: Refactor and fix COPY_U.<B|H|W> instructions
  target/mips: Refactor and fix COPY_S.<B|H|W|D> instructions
  target/mips: Fix MSA instructions ST.<B|H|W|D> on big endian host
  target/mips: Fix MSA instructions LD.<B|H|W|D> on big endian host
  target/mips: Make the results of MOD_<U|S>.<B|H|W|D> the same as on hardware
  target/mips: Make the results of DIV_<U|S>.<B|H|W|D> the same as on hardware

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

6 files changed, 674 insertions, 202 deletions

diff --git a/target/mips/cpu.h b/target/mips/cpu.h
index 1f41cf6..06a8ed4 100644
--- a/target/mips/cpu.h
+++ b/target/mips/cpu.h
@@ -22,10 +22,10 @@ typedef struct CPUMIPSTLBContext CPUMIPSTLBContext;
 
 typedef union wr_t wr_t;
 union wr_t {
-    int8_t  b[MSA_WRLEN/8];
-    int16_t h[MSA_WRLEN/16];
-    int32_t w[MSA_WRLEN/32];
-    int64_t d[MSA_WRLEN/64];
+    int8_t  b[MSA_WRLEN / 8];
+    int16_t h[MSA_WRLEN / 16];
+    int32_t w[MSA_WRLEN / 32];
+    int64_t d[MSA_WRLEN / 64];
 };
 
 typedef union fpr_t fpr_t;
@@ -37,7 +37,8 @@ union fpr_t {
 /* FPU/MSA register mapping is not tested on big-endian hosts. */
     wr_t     wr;   /* vector data */
 };
-/* define FP_ENDIAN_IDX to access the same location
+/*
+ *define FP_ENDIAN_IDX to access the same location
  * in the fpr_t union regardless of the host endianness
  */
 #if defined(HOST_WORDS_BIGENDIAN)
@@ -71,16 +72,29 @@ struct CPUMIPSFPUContext {
 #define FCR31_FS 24
 #define FCR31_ABS2008 19
 #define FCR31_NAN2008 18
-#define SET_FP_COND(num,env)     do { ((env).fcr31) |= ((num) ? (1 << ((num) + 24)) : (1 << 23)); } while(0)
-#define CLEAR_FP_COND(num,env)   do { ((env).fcr31) &= ~((num) ? (1 << ((num) + 24)) : (1 << 23)); } while(0)
-#define GET_FP_COND(env)         ((((env).fcr31 >> 24) & 0xfe) | (((env).fcr31 >> 23) & 0x1))
+#define SET_FP_COND(num, env)     do { ((env).fcr31) |=                 \
+                                       ((num) ? (1 << ((num) + 24)) :   \
+                                                (1 << 23));             \
+                                     } while (0)
+#define CLEAR_FP_COND(num, env)   do { ((env).fcr31) &=                 \
+                                       ~((num) ? (1 << ((num) + 24)) :  \
+                                                 (1 << 23));            \
+                                     } while (0)
+#define GET_FP_COND(env)         ((((env).fcr31 >> 24) & 0xfe) |        \
+                                 (((env).fcr31 >> 23) & 0x1))
 #define GET_FP_CAUSE(reg)        (((reg) >> 12) & 0x3f)
 #define GET_FP_ENABLE(reg)       (((reg) >>  7) & 0x1f)
 #define GET_FP_FLAGS(reg)        (((reg) >>  2) & 0x1f)
-#define SET_FP_CAUSE(reg,v)      do { (reg) = ((reg) & ~(0x3f << 12)) | ((v & 0x3f) << 12); } while(0)
-#define SET_FP_ENABLE(reg,v)     do { (reg) = ((reg) & ~(0x1f <<  7)) | ((v & 0x1f) << 7); } while(0)
-#define SET_FP_FLAGS(reg,v)      do { (reg) = ((reg) & ~(0x1f <<  2)) | ((v & 0x1f) << 2); } while(0)
-#define UPDATE_FP_FLAGS(reg,v)   do { (reg) |= ((v & 0x1f) << 2); } while(0)
+#define SET_FP_CAUSE(reg, v)      do { (reg) = ((reg) & ~(0x3f << 12)) | \
+                                               ((v & 0x3f) << 12);       \
+                                     } while (0)
+#define SET_FP_ENABLE(reg, v)     do { (reg) = ((reg) & ~(0x1f <<  7)) | \
+                                               ((v & 0x1f) << 7);        \
+                                     } while (0)
+#define SET_FP_FLAGS(reg, v)      do { (reg) = ((reg) & ~(0x1f <<  2)) | \
+                                               ((v & 0x1f) << 2);        \
+                                     } while (0)
+#define UPDATE_FP_FLAGS(reg, v)   do { (reg) |= ((v & 0x1f) << 2); } while (0)
 #define FP_INEXACT        1
 #define FP_UNDERFLOW      2
 #define FP_OVERFLOW       4
@@ -95,25 +109,25 @@ struct CPUMIPSFPUContext {
 typedef struct CPUMIPSMVPContext CPUMIPSMVPContext;
 struct CPUMIPSMVPContext {
     int32_t CP0_MVPControl;
-#define CP0MVPCo_CPA	3
-#define CP0MVPCo_STLB	2
-#define CP0MVPCo_VPC	1
-#define CP0MVPCo_EVP	0
+#define CP0MVPCo_CPA    3
+#define CP0MVPCo_STLB   2
+#define CP0MVPCo_VPC    1
+#define CP0MVPCo_EVP    0
     int32_t CP0_MVPConf0;
-#define CP0MVPC0_M	31
-#define CP0MVPC0_TLBS	29
-#define CP0MVPC0_GS	28
-#define CP0MVPC0_PCP	27
-#define CP0MVPC0_PTLBE	16
-#define CP0MVPC0_TCA	15
-#define CP0MVPC0_PVPE	10
-#define CP0MVPC0_PTC	0
+#define CP0MVPC0_M      31
+#define CP0MVPC0_TLBS   29
+#define CP0MVPC0_GS     28
+#define CP0MVPC0_PCP    27
+#define CP0MVPC0_PTLBE  16
+#define CP0MVPC0_TCA    15
+#define CP0MVPC0_PVPE   10
+#define CP0MVPC0_PTC    0
     int32_t CP0_MVPConf1;
-#define CP0MVPC1_CIM	31
-#define CP0MVPC1_CIF	30
-#define CP0MVPC1_PCX	20
-#define CP0MVPC1_PCP2	10
-#define CP0MVPC1_PCP1	0
+#define CP0MVPC1_CIM    31
+#define CP0MVPC1_CIF    30
+#define CP0MVPC1_PCX    20
+#define CP0MVPC1_PCP2   10
+#define CP0MVPC1_PCP1   0
 };
 
 typedef struct mips_def_t mips_def_t;
@@ -481,44 +495,44 @@ struct CPUMIPSState {
  */
     int32_t CP0_Random;
     int32_t CP0_VPEControl;
-#define CP0VPECo_YSI	21
-#define CP0VPECo_GSI	20
-#define CP0VPECo_EXCPT	16
-#define CP0VPECo_TE	15
-#define CP0VPECo_TargTC	0
+#define CP0VPECo_YSI    21
+#define CP0VPECo_GSI    20
+#define CP0VPECo_EXCPT  16
+#define CP0VPECo_TE     15
+#define CP0VPECo_TargTC 0
     int32_t CP0_VPEConf0;
-#define CP0VPEC0_M	31
-#define CP0VPEC0_XTC	21
-#define CP0VPEC0_TCS	19
-#define CP0VPEC0_SCS	18
-#define CP0VPEC0_DSC	17
-#define CP0VPEC0_ICS	16
-#define CP0VPEC0_MVP	1
-#define CP0VPEC0_VPA	0
+#define CP0VPEC0_M      31
+#define CP0VPEC0_XTC    21
+#define CP0VPEC0_TCS    19
+#define CP0VPEC0_SCS    18
+#define CP0VPEC0_DSC    17
+#define CP0VPEC0_ICS    16
+#define CP0VPEC0_MVP    1
+#define CP0VPEC0_VPA    0
     int32_t CP0_VPEConf1;
-#define CP0VPEC1_NCX	20
-#define CP0VPEC1_NCP2	10
-#define CP0VPEC1_NCP1	0
+#define CP0VPEC1_NCX    20
+#define CP0VPEC1_NCP2   10
+#define CP0VPEC1_NCP1   0
     target_ulong CP0_YQMask;
     target_ulong CP0_VPESchedule;
     target_ulong CP0_VPEScheFBack;
     int32_t CP0_VPEOpt;
-#define CP0VPEOpt_IWX7	15
-#define CP0VPEOpt_IWX6	14
-#define CP0VPEOpt_IWX5	13
-#define CP0VPEOpt_IWX4	12
-#define CP0VPEOpt_IWX3	11
-#define CP0VPEOpt_IWX2	10
-#define CP0VPEOpt_IWX1	9
-#define CP0VPEOpt_IWX0	8
-#define CP0VPEOpt_DWX7	7
-#define CP0VPEOpt_DWX6	6
-#define CP0VPEOpt_DWX5	5
-#define CP0VPEOpt_DWX4	4
-#define CP0VPEOpt_DWX3	3
-#define CP0VPEOpt_DWX2	2
-#define CP0VPEOpt_DWX1	1
-#define CP0VPEOpt_DWX0	0
+#define CP0VPEOpt_IWX7  15
+#define CP0VPEOpt_IWX6  14
+#define CP0VPEOpt_IWX5  13
+#define CP0VPEOpt_IWX4  12
+#define CP0VPEOpt_IWX3  11
+#define CP0VPEOpt_IWX2  10
+#define CP0VPEOpt_IWX1  9
+#define CP0VPEOpt_IWX0  8
+#define CP0VPEOpt_DWX7  7
+#define CP0VPEOpt_DWX6  6
+#define CP0VPEOpt_DWX5  5
+#define CP0VPEOpt_DWX4  4
+#define CP0VPEOpt_DWX3  3
+#define CP0VPEOpt_DWX2  2
+#define CP0VPEOpt_DWX1  1
+#define CP0VPEOpt_DWX0  0
 /*
  * CP0 Register 2
  */
@@ -625,33 +639,33 @@ struct CPUMIPSState {
 #define CP0PC_PSN       0     /*  5..0  */
     int32_t CP0_SRSConf0_rw_bitmask;
     int32_t CP0_SRSConf0;
-#define CP0SRSC0_M	31
-#define CP0SRSC0_SRS3	20
-#define CP0SRSC0_SRS2	10
-#define CP0SRSC0_SRS1	0
+#define CP0SRSC0_M      31
+#define CP0SRSC0_SRS3   20
+#define CP0SRSC0_SRS2   10
+#define CP0SRSC0_SRS1   0
     int32_t CP0_SRSConf1_rw_bitmask;
     int32_t CP0_SRSConf1;
-#define CP0SRSC1_M	31
-#define CP0SRSC1_SRS6	20
-#define CP0SRSC1_SRS5	10
-#define CP0SRSC1_SRS4	0
+#define CP0SRSC1_M      31
+#define CP0SRSC1_SRS6   20
+#define CP0SRSC1_SRS5   10
+#define CP0SRSC1_SRS4   0
     int32_t CP0_SRSConf2_rw_bitmask;
     int32_t CP0_SRSConf2;
-#define CP0SRSC2_M	31
-#define CP0SRSC2_SRS9	20
-#define CP0SRSC2_SRS8	10
-#define CP0SRSC2_SRS7	0
+#define CP0SRSC2_M      31
+#define CP0SRSC2_SRS9   20
+#define CP0SRSC2_SRS8   10
+#define CP0SRSC2_SRS7   0
     int32_t CP0_SRSConf3_rw_bitmask;
     int32_t CP0_SRSConf3;
-#define CP0SRSC3_M	31
-#define CP0SRSC3_SRS12	20
-#define CP0SRSC3_SRS11	10
-#define CP0SRSC3_SRS10	0
+#define CP0SRSC3_M      31
+#define CP0SRSC3_SRS12  20
+#define CP0SRSC3_SRS11  10
+#define CP0SRSC3_SRS10  0
     int32_t CP0_SRSConf4_rw_bitmask;
     int32_t CP0_SRSConf4;
-#define CP0SRSC4_SRS15	20
-#define CP0SRSC4_SRS14	10
-#define CP0SRSC4_SRS13	0
+#define CP0SRSC4_SRS15  20
+#define CP0SRSC4_SRS14  10
+#define CP0SRSC4_SRS13  0
 /*
  * CP0 Register 7
  */
@@ -963,9 +977,11 @@ struct CPUMIPSState {
     /* TMASK defines different execution modes */
 #define MIPS_HFLAG_TMASK  0x1F5807FF
 #define MIPS_HFLAG_MODE   0x00007 /* execution modes                    */
-    /* The KSU flags must be the lowest bits in hflags. The flag order
-       must be the same as defined for CP0 Status. This allows to use
-       the bits as the value of mmu_idx. */
+    /*
+     * The KSU flags must be the lowest bits in hflags. The flag order
+     * must be the same as defined for CP0 Status. This allows to use
+     * the bits as the value of mmu_idx.
+     */
 #define MIPS_HFLAG_KSU    0x00003 /* kernel/supervisor/user mode mask   */
 #define MIPS_HFLAG_UM     0x00002 /* user mode flag                     */
 #define MIPS_HFLAG_SM     0x00001 /* supervisor mode flag               */
@@ -975,18 +991,22 @@ struct CPUMIPSState {
 #define MIPS_HFLAG_CP0    0x00010 /* CP0 enabled                        */
 #define MIPS_HFLAG_FPU    0x00020 /* FPU enabled                        */
 #define MIPS_HFLAG_F64    0x00040 /* 64-bit FPU enabled                 */
-    /* True if the MIPS IV COP1X instructions can be used.  This also
-       controls the non-COP1X instructions RECIP.S, RECIP.D, RSQRT.S
-       and RSQRT.D.  */
+    /*
+     * True if the MIPS IV COP1X instructions can be used.  This also
+     * controls the non-COP1X instructions RECIP.S, RECIP.D, RSQRT.S
+     * and RSQRT.D.
+     */
 #define MIPS_HFLAG_COP1X  0x00080 /* COP1X instructions enabled         */
 #define MIPS_HFLAG_RE     0x00100 /* Reversed endianness                */
 #define MIPS_HFLAG_AWRAP  0x00200 /* 32-bit compatibility address wrapping */
 #define MIPS_HFLAG_M16    0x00400 /* MIPS16 mode flag                   */
 #define MIPS_HFLAG_M16_SHIFT 10
-    /* If translation is interrupted between the branch instruction and
+    /*
+     * If translation is interrupted between the branch instruction and
      * the delay slot, record what type of branch it is so that we can
      * resume translation properly.  It might be possible to reduce
-     * this from three bits to two.  */
+     * this from three bits to two.
+     */
 #define MIPS_HFLAG_BMASK_BASE  0x803800
 #define MIPS_HFLAG_B      0x00800 /* Unconditional branch               */
 #define MIPS_HFLAG_BC     0x01000 /* Conditional branch                 */
@@ -1073,8 +1093,10 @@ void mips_cpu_list(void);
 extern void cpu_wrdsp(uint32_t rs, uint32_t mask_num, CPUMIPSState *env);
 extern uint32_t cpu_rddsp(uint32_t mask_num, CPUMIPSState *env);
 
-/* MMU modes definitions. We carefully match the indices with our
-   hflags layout. */
+/*
+ * MMU modes definitions. We carefully match the indices with our
+ * hflags layout.
+ */
 #define MMU_MODE0_SUFFIX _kernel
 #define MMU_MODE1_SUFFIX _super
 #define MMU_MODE2_SUFFIX _user
@@ -1090,14 +1112,15 @@ static inline int hflags_mmu_index(uint32_t hflags)
     }
 }
 
-static inline int cpu_mmu_index (CPUMIPSState *env, bool ifetch)
+static inline int cpu_mmu_index(CPUMIPSState *env, bool ifetch)
 {
     return hflags_mmu_index(env->hflags);
 }
 
 #include "exec/cpu-all.h"
 
-/* Memory access type :
+/*
+ * Memory access type :
  * may be needed for precise access rights control and precise exceptions.
  */
 enum {
@@ -1182,7 +1205,7 @@ void cpu_mips_soft_irq(CPUMIPSState *env, int irq, int level);
 void itc_reconfigure(struct MIPSITUState *tag);
 
 /* helper.c */
-target_ulong exception_resume_pc (CPUMIPSState *env);
+target_ulong exception_resume_pc(CPUMIPSState *env);
 
 static inline void restore_snan_bit_mode(CPUMIPSState *env)
 {
diff --git a/target/mips/helper.c b/target/mips/helper.c
index 9799f2e..68e44df 100644
--- a/target/mips/helper.c
+++ b/target/mips/helper.c
@@ -43,7 +43,7 @@ int no_mmu_map_address (CPUMIPSState *env, hwaddr *physical, int *prot,
                         target_ulong address, int rw, int access_type)
 {
     *physical = address;
-    *prot = PAGE_READ | PAGE_WRITE;
+    *prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
     return TLBRET_MATCH;
 }
 
@@ -61,7 +61,7 @@ int fixed_mmu_map_address (CPUMIPSState *env, hwaddr *physical, int *prot,
     else
         *physical = address;
 
-    *prot = PAGE_READ | PAGE_WRITE;
+    *prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
     return TLBRET_MATCH;
 }
 
@@ -101,6 +101,9 @@ int r4k_map_address (CPUMIPSState *env, hwaddr *physical, int *prot,
                 *prot = PAGE_READ;
                 if (n ? tlb->D1 : tlb->D0)
                     *prot |= PAGE_WRITE;
+                if (!(n ? tlb->XI1 : tlb->XI0)) {
+                    *prot |= PAGE_EXEC;
+                }
                 return TLBRET_MATCH;
             }
             return TLBRET_DIRTY;
@@ -182,7 +185,7 @@ static int get_seg_physical_address(CPUMIPSState *env, hwaddr *physical,
     } else {
         /* The segment is unmapped */
         *physical = physical_base | (real_address & segmask);
-        *prot = PAGE_READ | PAGE_WRITE;
+        *prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
         return TLBRET_MATCH;
     }
 }
@@ -907,7 +910,7 @@ bool mips_cpu_tlb_fill(CPUState *cs, vaddr address, int size,
     }
     if (ret == TLBRET_MATCH) {
         tlb_set_page(cs, address & TARGET_PAGE_MASK,
-                     physical & TARGET_PAGE_MASK, prot | PAGE_EXEC,
+                     physical & TARGET_PAGE_MASK, prot,
                      mmu_idx, TARGET_PAGE_SIZE);
         return true;
     }
@@ -927,7 +930,7 @@ bool mips_cpu_tlb_fill(CPUState *cs, vaddr address, int size,
                                        access_type, mips_access_type, mmu_idx);
             if (ret == TLBRET_MATCH) {
                 tlb_set_page(cs, address & TARGET_PAGE_MASK,
-                             physical & TARGET_PAGE_MASK, prot | PAGE_EXEC,
+                             physical & TARGET_PAGE_MASK, prot,
                              mmu_idx, TARGET_PAGE_SIZE);
                 return true;
             }
diff --git a/target/mips/helper.h b/target/mips/helper.h
index a6d687e..2863f60 100644
--- a/target/mips/helper.h
+++ b/target/mips/helper.h
@@ -876,9 +876,7 @@ DEF_HELPER_5(msa_hsub_u_df, void, env, i32, i32, i32, i32)
 
 DEF_HELPER_5(msa_sldi_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_splati_df, void, env, i32, i32, i32, i32)
-DEF_HELPER_5(msa_copy_s_df, void, env, i32, i32, i32, i32)
-DEF_HELPER_5(msa_copy_u_df, void, env, i32, i32, i32, i32)
-DEF_HELPER_5(msa_insert_df, void, env, i32, i32, i32, i32)
+
 DEF_HELPER_5(msa_insve_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_3(msa_ctcmsa, void, env, tl, i32)
 DEF_HELPER_2(msa_cfcmsa, tl, env, i32)
@@ -938,6 +936,18 @@ DEF_HELPER_4(msa_pcnt_df, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_nloc_df, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_nlzc_df, void, env, i32, i32, i32)
 
+DEF_HELPER_4(msa_copy_s_b, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_copy_s_h, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_copy_s_w, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_copy_s_d, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_copy_u_b, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_copy_u_h, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_copy_u_w, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_insert_b, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_insert_h, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_insert_w, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_insert_d, void, env, i32, i32, i32)
+
 DEF_HELPER_4(msa_fclass_df, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ftrunc_s_df, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ftrunc_u_df, void, env, i32, i32, i32)
diff --git a/target/mips/msa_helper.c b/target/mips/msa_helper.c
index c74e3cd..ee1b1fa 100644
--- a/target/mips/msa_helper.c
+++ b/target/mips/msa_helper.c
@@ -641,14 +641,15 @@ static inline int64_t msa_div_s_df(uint32_t df, int64_t arg1, int64_t arg2)
     if (arg1 == DF_MIN_INT(df) && arg2 == -1) {
         return DF_MIN_INT(df);
     }
-    return arg2 ? arg1 / arg2 : 0;
+    return arg2 ? arg1 / arg2
+                : arg1 >= 0 ? -1 : 1;
 }
 
 static inline int64_t msa_div_u_df(uint32_t df, int64_t arg1, int64_t arg2)
 {
     uint64_t u_arg1 = UNSIGNED(arg1, df);
     uint64_t u_arg2 = UNSIGNED(arg2, df);
-    return u_arg2 ? u_arg1 / u_arg2 : 0;
+    return arg2 ? u_arg1 / u_arg2 : -1;
 }
 
 static inline int64_t msa_mod_s_df(uint32_t df, int64_t arg1, int64_t arg2)
@@ -656,14 +657,14 @@ static inline int64_t msa_mod_s_df(uint32_t df, int64_t arg1, int64_t arg2)
     if (arg1 == DF_MIN_INT(df) && arg2 == -1) {
         return 0;
     }
-    return arg2 ? arg1 % arg2 : 0;
+    return arg2 ? arg1 % arg2 : arg1;
 }
 
 static inline int64_t msa_mod_u_df(uint32_t df, int64_t arg1, int64_t arg2)
 {
     uint64_t u_arg1 = UNSIGNED(arg1, df);
     uint64_t u_arg2 = UNSIGNED(arg2, df);
-    return u_arg2 ? u_arg1 % u_arg2 : 0;
+    return u_arg2 ? u_arg1 % u_arg2 : u_arg1;
 }
 
 #define SIGNED_EVEN(a, df) \
@@ -1248,78 +1249,152 @@ void helper_msa_splati_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
     msa_splat_df(df, pwd, pws, n);
 }
 
-void helper_msa_copy_s_df(CPUMIPSState *env, uint32_t df, uint32_t rd,
-                          uint32_t ws, uint32_t n)
+void helper_msa_copy_s_b(CPUMIPSState *env, uint32_t rd,
+                         uint32_t ws, uint32_t n)
 {
-    n %= DF_ELEMENTS(df);
+    n %= 16;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 8) {
+        n = 8 - n - 1;
+    } else {
+        n = 24 - n - 1;
+    }
+#endif
+    env->active_tc.gpr[rd] = (int8_t)env->active_fpu.fpr[ws].wr.b[n];
+}
 
-    switch (df) {
-    case DF_BYTE:
-        env->active_tc.gpr[rd] = (int8_t)env->active_fpu.fpr[ws].wr.b[n];
-        break;
-    case DF_HALF:
-        env->active_tc.gpr[rd] = (int16_t)env->active_fpu.fpr[ws].wr.h[n];
-        break;
-    case DF_WORD:
-        env->active_tc.gpr[rd] = (int32_t)env->active_fpu.fpr[ws].wr.w[n];
-        break;
-#ifdef TARGET_MIPS64
-    case DF_DOUBLE:
-        env->active_tc.gpr[rd] = (int64_t)env->active_fpu.fpr[ws].wr.d[n];
-        break;
+void helper_msa_copy_s_h(CPUMIPSState *env, uint32_t rd,
+                         uint32_t ws, uint32_t n)
+{
+    n %= 8;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 4) {
+        n = 4 - n - 1;
+    } else {
+        n = 12 - n - 1;
+    }
 #endif
-    default:
-        assert(0);
+    env->active_tc.gpr[rd] = (int16_t)env->active_fpu.fpr[ws].wr.h[n];
+}
+
+void helper_msa_copy_s_w(CPUMIPSState *env, uint32_t rd,
+                         uint32_t ws, uint32_t n)
+{
+    n %= 4;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 2) {
+        n = 2 - n - 1;
+    } else {
+        n = 6 - n - 1;
     }
+#endif
+    env->active_tc.gpr[rd] = (int32_t)env->active_fpu.fpr[ws].wr.w[n];
 }
 
-void helper_msa_copy_u_df(CPUMIPSState *env, uint32_t df, uint32_t rd,
-                          uint32_t ws, uint32_t n)
+void helper_msa_copy_s_d(CPUMIPSState *env, uint32_t rd,
+                         uint32_t ws, uint32_t n)
 {
-    n %= DF_ELEMENTS(df);
+    n %= 2;
+    env->active_tc.gpr[rd] = (int64_t)env->active_fpu.fpr[ws].wr.d[n];
+}
 
-    switch (df) {
-    case DF_BYTE:
-        env->active_tc.gpr[rd] = (uint8_t)env->active_fpu.fpr[ws].wr.b[n];
-        break;
-    case DF_HALF:
-        env->active_tc.gpr[rd] = (uint16_t)env->active_fpu.fpr[ws].wr.h[n];
-        break;
-    case DF_WORD:
-        env->active_tc.gpr[rd] = (uint32_t)env->active_fpu.fpr[ws].wr.w[n];
-        break;
-#ifdef TARGET_MIPS64
-    case DF_DOUBLE:
-        env->active_tc.gpr[rd] = (uint64_t)env->active_fpu.fpr[ws].wr.d[n];
-        break;
+void helper_msa_copy_u_b(CPUMIPSState *env, uint32_t rd,
+                         uint32_t ws, uint32_t n)
+{
+    n %= 16;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 8) {
+        n = 8 - n - 1;
+    } else {
+        n = 24 - n - 1;
+    }
 #endif
-    default:
-        assert(0);
+    env->active_tc.gpr[rd] = (uint8_t)env->active_fpu.fpr[ws].wr.b[n];
+}
+
+void helper_msa_copy_u_h(CPUMIPSState *env, uint32_t rd,
+                         uint32_t ws, uint32_t n)
+{
+    n %= 8;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 4) {
+        n = 4 - n - 1;
+    } else {
+        n = 12 - n - 1;
+    }
+#endif
+    env->active_tc.gpr[rd] = (uint16_t)env->active_fpu.fpr[ws].wr.h[n];
+}
+
+void helper_msa_copy_u_w(CPUMIPSState *env, uint32_t rd,
+                         uint32_t ws, uint32_t n)
+{
+    n %= 4;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 2) {
+        n = 2 - n - 1;
+    } else {
+        n = 6 - n - 1;
     }
+#endif
+    env->active_tc.gpr[rd] = (uint32_t)env->active_fpu.fpr[ws].wr.w[n];
 }
 
-void helper_msa_insert_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
+void helper_msa_insert_b(CPUMIPSState *env, uint32_t wd,
                           uint32_t rs_num, uint32_t n)
 {
     wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
     target_ulong rs = env->active_tc.gpr[rs_num];
+    n %= 16;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 8) {
+        n = 8 - n - 1;
+    } else {
+        n = 24 - n - 1;
+    }
+#endif
+    pwd->b[n] = (int8_t)rs;
+}
 
-    switch (df) {
-    case DF_BYTE:
-        pwd->b[n] = (int8_t)rs;
-        break;
-    case DF_HALF:
-        pwd->h[n] = (int16_t)rs;
-        break;
-    case DF_WORD:
-        pwd->w[n] = (int32_t)rs;
-        break;
-    case DF_DOUBLE:
-        pwd->d[n] = (int64_t)rs;
-        break;
-    default:
-        assert(0);
+void helper_msa_insert_h(CPUMIPSState *env, uint32_t wd,
+                          uint32_t rs_num, uint32_t n)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    target_ulong rs = env->active_tc.gpr[rs_num];
+    n %= 8;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 4) {
+        n = 4 - n - 1;
+    } else {
+        n = 12 - n - 1;
     }
+#endif
+    pwd->h[n] = (int16_t)rs;
+}
+
+void helper_msa_insert_w(CPUMIPSState *env, uint32_t wd,
+                          uint32_t rs_num, uint32_t n)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    target_ulong rs = env->active_tc.gpr[rs_num];
+    n %= 4;
+#if defined(HOST_WORDS_BIGENDIAN)
+    if (n < 2) {
+        n = 2 - n - 1;
+    } else {
+        n = 6 - n - 1;
+    }
+#endif
+    pwd->w[n] = (int32_t)rs;
+}
+
+void helper_msa_insert_d(CPUMIPSState *env, uint32_t wd,
+                          uint32_t rs_num, uint32_t n)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    target_ulong rs = env->active_tc.gpr[rs_num];
+    n %= 2;
+    pwd->d[n] = (int64_t)rs;
 }
 
 void helper_msa_insve_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
diff --git a/target/mips/op_helper.c b/target/mips/op_helper.c
index 6d86912..3918027 100644
--- a/target/mips/op_helper.c
+++ b/target/mips/op_helper.c
@@ -4356,31 +4356,179 @@ FOP_CONDN_S(sne,  (float32_lt(fst1, fst0, &env->active_fpu.fp_status)
 #define MEMOP_IDX(DF)
 #endif
 
-#define MSA_LD_DF(DF, TYPE, LD_INSN, ...)                               \
-void helper_msa_ld_ ## TYPE(CPUMIPSState *env, uint32_t wd,             \
-                            target_ulong addr)                          \
-{                                                                       \
-    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);                          \
-    wr_t wx;                                                            \
-    int i;                                                              \
-    MEMOP_IDX(DF)                                                       \
-    for (i = 0; i < DF_ELEMENTS(DF); i++) {                             \
-        wx.TYPE[i] = LD_INSN(env, addr + (i << DF), ##__VA_ARGS__);     \
-    }                                                                   \
-    memcpy(pwd, &wx, sizeof(wr_t));                                     \
+void helper_msa_ld_b(CPUMIPSState *env, uint32_t wd,
+                     target_ulong addr)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    MEMOP_IDX(DF_BYTE)
+#if !defined(CONFIG_USER_ONLY)
+#if !defined(HOST_WORDS_BIGENDIAN)
+    pwd->b[0]  = helper_ret_ldub_mmu(env, addr + (0  << DF_BYTE), oi, GETPC());
+    pwd->b[1]  = helper_ret_ldub_mmu(env, addr + (1  << DF_BYTE), oi, GETPC());
+    pwd->b[2]  = helper_ret_ldub_mmu(env, addr + (2  << DF_BYTE), oi, GETPC());
+    pwd->b[3]  = helper_ret_ldub_mmu(env, addr + (3  << DF_BYTE), oi, GETPC());
+    pwd->b[4]  = helper_ret_ldub_mmu(env, addr + (4  << DF_BYTE), oi, GETPC());
+    pwd->b[5]  = helper_ret_ldub_mmu(env, addr + (5  << DF_BYTE), oi, GETPC());
+    pwd->b[6]  = helper_ret_ldub_mmu(env, addr + (6  << DF_BYTE), oi, GETPC());
+    pwd->b[7]  = helper_ret_ldub_mmu(env, addr + (7  << DF_BYTE), oi, GETPC());
+    pwd->b[8]  = helper_ret_ldub_mmu(env, addr + (8  << DF_BYTE), oi, GETPC());
+    pwd->b[9]  = helper_ret_ldub_mmu(env, addr + (9  << DF_BYTE), oi, GETPC());
+    pwd->b[10] = helper_ret_ldub_mmu(env, addr + (10 << DF_BYTE), oi, GETPC());
+    pwd->b[11] = helper_ret_ldub_mmu(env, addr + (11 << DF_BYTE), oi, GETPC());
+    pwd->b[12] = helper_ret_ldub_mmu(env, addr + (12 << DF_BYTE), oi, GETPC());
+    pwd->b[13] = helper_ret_ldub_mmu(env, addr + (13 << DF_BYTE), oi, GETPC());
+    pwd->b[14] = helper_ret_ldub_mmu(env, addr + (14 << DF_BYTE), oi, GETPC());
+    pwd->b[15] = helper_ret_ldub_mmu(env, addr + (15 << DF_BYTE), oi, GETPC());
+#else
+    pwd->b[0]  = helper_ret_ldub_mmu(env, addr + (7  << DF_BYTE), oi, GETPC());
+    pwd->b[1]  = helper_ret_ldub_mmu(env, addr + (6  << DF_BYTE), oi, GETPC());
+    pwd->b[2]  = helper_ret_ldub_mmu(env, addr + (5  << DF_BYTE), oi, GETPC());
+    pwd->b[3]  = helper_ret_ldub_mmu(env, addr + (4  << DF_BYTE), oi, GETPC());
+    pwd->b[4]  = helper_ret_ldub_mmu(env, addr + (3  << DF_BYTE), oi, GETPC());
+    pwd->b[5]  = helper_ret_ldub_mmu(env, addr + (2  << DF_BYTE), oi, GETPC());
+    pwd->b[6]  = helper_ret_ldub_mmu(env, addr + (1  << DF_BYTE), oi, GETPC());
+    pwd->b[7]  = helper_ret_ldub_mmu(env, addr + (0  << DF_BYTE), oi, GETPC());
+    pwd->b[8]  = helper_ret_ldub_mmu(env, addr + (15 << DF_BYTE), oi, GETPC());
+    pwd->b[9]  = helper_ret_ldub_mmu(env, addr + (14 << DF_BYTE), oi, GETPC());
+    pwd->b[10] = helper_ret_ldub_mmu(env, addr + (13 << DF_BYTE), oi, GETPC());
+    pwd->b[11] = helper_ret_ldub_mmu(env, addr + (12 << DF_BYTE), oi, GETPC());
+    pwd->b[12] = helper_ret_ldub_mmu(env, addr + (11 << DF_BYTE), oi, GETPC());
+    pwd->b[13] = helper_ret_ldub_mmu(env, addr + (10 << DF_BYTE), oi, GETPC());
+    pwd->b[14] = helper_ret_ldub_mmu(env, addr + (9  << DF_BYTE), oi, GETPC());
+    pwd->b[15] = helper_ret_ldub_mmu(env, addr + (8  << DF_BYTE), oi, GETPC());
+#endif
+#else
+#if !defined(HOST_WORDS_BIGENDIAN)
+    pwd->b[0]  = cpu_ldub_data(env, addr + (0  << DF_BYTE));
+    pwd->b[1]  = cpu_ldub_data(env, addr + (1  << DF_BYTE));
+    pwd->b[2]  = cpu_ldub_data(env, addr + (2  << DF_BYTE));
+    pwd->b[3]  = cpu_ldub_data(env, addr + (3  << DF_BYTE));
+    pwd->b[4]  = cpu_ldub_data(env, addr + (4  << DF_BYTE));
+    pwd->b[5]  = cpu_ldub_data(env, addr + (5  << DF_BYTE));
+    pwd->b[6]  = cpu_ldub_data(env, addr + (6  << DF_BYTE));
+    pwd->b[7]  = cpu_ldub_data(env, addr + (7  << DF_BYTE));
+    pwd->b[8]  = cpu_ldub_data(env, addr + (8  << DF_BYTE));
+    pwd->b[9]  = cpu_ldub_data(env, addr + (9  << DF_BYTE));
+    pwd->b[10] = cpu_ldub_data(env, addr + (10 << DF_BYTE));
+    pwd->b[11] = cpu_ldub_data(env, addr + (11 << DF_BYTE));
+    pwd->b[12] = cpu_ldub_data(env, addr + (12 << DF_BYTE));
+    pwd->b[13] = cpu_ldub_data(env, addr + (13 << DF_BYTE));
+    pwd->b[14] = cpu_ldub_data(env, addr + (14 << DF_BYTE));
+    pwd->b[15] = cpu_ldub_data(env, addr + (15 << DF_BYTE));
+#else
+    pwd->b[0]  = cpu_ldub_data(env, addr + (7  << DF_BYTE));
+    pwd->b[1]  = cpu_ldub_data(env, addr + (6  << DF_BYTE));
+    pwd->b[2]  = cpu_ldub_data(env, addr + (5  << DF_BYTE));
+    pwd->b[3]  = cpu_ldub_data(env, addr + (4  << DF_BYTE));
+    pwd->b[4]  = cpu_ldub_data(env, addr + (3  << DF_BYTE));
+    pwd->b[5]  = cpu_ldub_data(env, addr + (2  << DF_BYTE));
+    pwd->b[6]  = cpu_ldub_data(env, addr + (1  << DF_BYTE));
+    pwd->b[7]  = cpu_ldub_data(env, addr + (0  << DF_BYTE));
+    pwd->b[8]  = cpu_ldub_data(env, addr + (15 << DF_BYTE));
+    pwd->b[9]  = cpu_ldub_data(env, addr + (14 << DF_BYTE));
+    pwd->b[10] = cpu_ldub_data(env, addr + (13 << DF_BYTE));
+    pwd->b[11] = cpu_ldub_data(env, addr + (12 << DF_BYTE));
+    pwd->b[12] = cpu_ldub_data(env, addr + (11 << DF_BYTE));
+    pwd->b[13] = cpu_ldub_data(env, addr + (10 << DF_BYTE));
+    pwd->b[14] = cpu_ldub_data(env, addr + (9 << DF_BYTE));
+    pwd->b[15] = cpu_ldub_data(env, addr + (8 << DF_BYTE));
+#endif
+#endif
+}
+
+void helper_msa_ld_h(CPUMIPSState *env, uint32_t wd,
+                     target_ulong addr)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    MEMOP_IDX(DF_HALF)
+#if !defined(CONFIG_USER_ONLY)
+#if !defined(HOST_WORDS_BIGENDIAN)
+    pwd->h[0] = helper_ret_lduw_mmu(env, addr + (0 << DF_HALF), oi, GETPC());
+    pwd->h[1] = helper_ret_lduw_mmu(env, addr + (1 << DF_HALF), oi, GETPC());
+    pwd->h[2] = helper_ret_lduw_mmu(env, addr + (2 << DF_HALF), oi, GETPC());
+    pwd->h[3] = helper_ret_lduw_mmu(env, addr + (3 << DF_HALF), oi, GETPC());
+    pwd->h[4] = helper_ret_lduw_mmu(env, addr + (4 << DF_HALF), oi, GETPC());
+    pwd->h[5] = helper_ret_lduw_mmu(env, addr + (5 << DF_HALF), oi, GETPC());
+    pwd->h[6] = helper_ret_lduw_mmu(env, addr + (6 << DF_HALF), oi, GETPC());
+    pwd->h[7] = helper_ret_lduw_mmu(env, addr + (7 << DF_HALF), oi, GETPC());
+#else
+    pwd->h[0] = helper_ret_lduw_mmu(env, addr + (3 << DF_HALF), oi, GETPC());
+    pwd->h[1] = helper_ret_lduw_mmu(env, addr + (2 << DF_HALF), oi, GETPC());
+    pwd->h[2] = helper_ret_lduw_mmu(env, addr + (1 << DF_HALF), oi, GETPC());
+    pwd->h[3] = helper_ret_lduw_mmu(env, addr + (0 << DF_HALF), oi, GETPC());
+    pwd->h[4] = helper_ret_lduw_mmu(env, addr + (7 << DF_HALF), oi, GETPC());
+    pwd->h[5] = helper_ret_lduw_mmu(env, addr + (6 << DF_HALF), oi, GETPC());
+    pwd->h[6] = helper_ret_lduw_mmu(env, addr + (5 << DF_HALF), oi, GETPC());
+    pwd->h[7] = helper_ret_lduw_mmu(env, addr + (4 << DF_HALF), oi, GETPC());
+#endif
+#else
+#if !defined(HOST_WORDS_BIGENDIAN)
+    pwd->h[0] = cpu_lduw_data(env, addr + (0 << DF_HALF));
+    pwd->h[1] = cpu_lduw_data(env, addr + (1 << DF_HALF));
+    pwd->h[2] = cpu_lduw_data(env, addr + (2 << DF_HALF));
+    pwd->h[3] = cpu_lduw_data(env, addr + (3 << DF_HALF));
+    pwd->h[4] = cpu_lduw_data(env, addr + (4 << DF_HALF));
+    pwd->h[5] = cpu_lduw_data(env, addr + (5 << DF_HALF));
+    pwd->h[6] = cpu_lduw_data(env, addr + (6 << DF_HALF));
+    pwd->h[7] = cpu_lduw_data(env, addr + (7 << DF_HALF));
+#else
+    pwd->h[0] = cpu_lduw_data(env, addr + (3 << DF_HALF));
+    pwd->h[1] = cpu_lduw_data(env, addr + (2 << DF_HALF));
+    pwd->h[2] = cpu_lduw_data(env, addr + (1 << DF_HALF));
+    pwd->h[3] = cpu_lduw_data(env, addr + (0 << DF_HALF));
+    pwd->h[4] = cpu_lduw_data(env, addr + (7 << DF_HALF));
+    pwd->h[5] = cpu_lduw_data(env, addr + (6 << DF_HALF));
+    pwd->h[6] = cpu_lduw_data(env, addr + (5 << DF_HALF));
+    pwd->h[7] = cpu_lduw_data(env, addr + (4 << DF_HALF));
+#endif
+#endif
 }
 
+void helper_msa_ld_w(CPUMIPSState *env, uint32_t wd,
+                     target_ulong addr)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    MEMOP_IDX(DF_WORD)
 #if !defined(CONFIG_USER_ONLY)
-MSA_LD_DF(DF_BYTE,   b, helper_ret_ldub_mmu, oi, GETPC())
-MSA_LD_DF(DF_HALF,   h, helper_ret_lduw_mmu, oi, GETPC())
-MSA_LD_DF(DF_WORD,   w, helper_ret_ldul_mmu, oi, GETPC())
-MSA_LD_DF(DF_DOUBLE, d, helper_ret_ldq_mmu,  oi, GETPC())
+#if !defined(HOST_WORDS_BIGENDIAN)
+    pwd->w[0] = helper_ret_ldul_mmu(env, addr + (0 << DF_WORD), oi, GETPC());
+    pwd->w[1] = helper_ret_ldul_mmu(env, addr + (1 << DF_WORD), oi, GETPC());
+    pwd->w[2] = helper_ret_ldul_mmu(env, addr + (2 << DF_WORD), oi, GETPC());
+    pwd->w[3] = helper_ret_ldul_mmu(env, addr + (3 << DF_WORD), oi, GETPC());
 #else
-MSA_LD_DF(DF_BYTE,   b, cpu_ldub_data)
-MSA_LD_DF(DF_HALF,   h, cpu_lduw_data)
-MSA_LD_DF(DF_WORD,   w, cpu_ldl_data)
-MSA_LD_DF(DF_DOUBLE, d, cpu_ldq_data)
+    pwd->w[0] = helper_ret_ldul_mmu(env, addr + (1 << DF_WORD), oi, GETPC());
+    pwd->w[1] = helper_ret_ldul_mmu(env, addr + (0 << DF_WORD), oi, GETPC());
+    pwd->w[2] = helper_ret_ldul_mmu(env, addr + (3 << DF_WORD), oi, GETPC());
+    pwd->w[3] = helper_ret_ldul_mmu(env, addr + (2 << DF_WORD), oi, GETPC());
 #endif
+#else
+#if !defined(HOST_WORDS_BIGENDIAN)
+    pwd->w[0] = cpu_ldl_data(env, addr + (0 << DF_WORD));
+    pwd->w[1] = cpu_ldl_data(env, addr + (1 << DF_WORD));
+    pwd->w[2] = cpu_ldl_data(env, addr + (2 << DF_WORD));
+    pwd->w[3] = cpu_ldl_data(env, addr + (3 << DF_WORD));
+#else
+    pwd->w[0] = cpu_ldl_data(env, addr + (1 << DF_WORD));
+    pwd->w[1] = cpu_ldl_data(env, addr + (0 << DF_WORD));
+    pwd->w[2] = cpu_ldl_data(env, addr + (3 << DF_WORD));
+    pwd->w[3] = cpu_ldl_data(env, addr + (2 << DF_WORD));
+#endif
+#endif
+}
+
+void helper_msa_ld_d(CPUMIPSState *env, uint32_t wd,
+                     target_ulong addr)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    MEMOP_IDX(DF_DOUBLE)
+#if !defined(CONFIG_USER_ONLY)
+    pwd->d[0] = helper_ret_ldq_mmu(env, addr + (0 << DF_DOUBLE), oi, GETPC());
+    pwd->d[1] = helper_ret_ldq_mmu(env, addr + (1 << DF_DOUBLE), oi, GETPC());
+#else
+    pwd->d[0] = cpu_ldq_data(env, addr + (0 << DF_DOUBLE));
+    pwd->d[1] = cpu_ldq_data(env, addr + (1 << DF_DOUBLE));
+#endif
+}
 
 #define MSA_PAGESPAN(x) \
         ((((x) & ~TARGET_PAGE_MASK) + MSA_WRLEN/8 - 1) >= TARGET_PAGE_SIZE)
@@ -4402,31 +4550,191 @@ static inline void ensure_writable_pages(CPUMIPSState *env,
 #endif
 }
 
-#define MSA_ST_DF(DF, TYPE, ST_INSN, ...)                               \
-void helper_msa_st_ ## TYPE(CPUMIPSState *env, uint32_t wd,             \
-                            target_ulong addr)                          \
-{                                                                       \
-    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);                          \
-    int mmu_idx = cpu_mmu_index(env, false);				\
-    int i;                                                              \
-    MEMOP_IDX(DF)                                                       \
-    ensure_writable_pages(env, addr, mmu_idx, GETPC());                 \
-    for (i = 0; i < DF_ELEMENTS(DF); i++) {                             \
-        ST_INSN(env, addr + (i << DF), pwd->TYPE[i], ##__VA_ARGS__);    \
-    }                                                                   \
+void helper_msa_st_b(CPUMIPSState *env, uint32_t wd,
+                     target_ulong addr)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    int mmu_idx = cpu_mmu_index(env, false);
+
+    MEMOP_IDX(DF_BYTE)
+    ensure_writable_pages(env, addr, mmu_idx, GETPC());
+#if !defined(CONFIG_USER_ONLY)
+#if !defined(HOST_WORDS_BIGENDIAN)
+    helper_ret_stb_mmu(env, addr + (0  << DF_BYTE), pwd->b[0],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (1  << DF_BYTE), pwd->b[1],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (2  << DF_BYTE), pwd->b[2],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (3  << DF_BYTE), pwd->b[3],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (4  << DF_BYTE), pwd->b[4],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (5  << DF_BYTE), pwd->b[5],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (6  << DF_BYTE), pwd->b[6],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (7  << DF_BYTE), pwd->b[7],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (8  << DF_BYTE), pwd->b[8],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (9  << DF_BYTE), pwd->b[9],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (10 << DF_BYTE), pwd->b[10], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (11 << DF_BYTE), pwd->b[11], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (12 << DF_BYTE), pwd->b[12], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (13 << DF_BYTE), pwd->b[13], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (14 << DF_BYTE), pwd->b[14], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (15 << DF_BYTE), pwd->b[15], oi, GETPC());
+#else
+    helper_ret_stb_mmu(env, addr + (7  << DF_BYTE), pwd->b[0],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (6  << DF_BYTE), pwd->b[1],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (5  << DF_BYTE), pwd->b[2],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (4  << DF_BYTE), pwd->b[3],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (3  << DF_BYTE), pwd->b[4],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (2  << DF_BYTE), pwd->b[5],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (1  << DF_BYTE), pwd->b[6],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (0  << DF_BYTE), pwd->b[7],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (15 << DF_BYTE), pwd->b[8],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (14 << DF_BYTE), pwd->b[9],  oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (13 << DF_BYTE), pwd->b[10], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (12 << DF_BYTE), pwd->b[11], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (11 << DF_BYTE), pwd->b[12], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (10 << DF_BYTE), pwd->b[13], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (9  << DF_BYTE), pwd->b[14], oi, GETPC());
+    helper_ret_stb_mmu(env, addr + (8  << DF_BYTE), pwd->b[15], oi, GETPC());
+#endif
+#else
+#if !defined(HOST_WORDS_BIGENDIAN)
+    cpu_stb_data(env, addr + (0  << DF_BYTE), pwd->b[0]);
+    cpu_stb_data(env, addr + (1  << DF_BYTE), pwd->b[1]);
+    cpu_stb_data(env, addr + (2  << DF_BYTE), pwd->b[2]);
+    cpu_stb_data(env, addr + (3  << DF_BYTE), pwd->b[3]);
+    cpu_stb_data(env, addr + (4  << DF_BYTE), pwd->b[4]);
+    cpu_stb_data(env, addr + (5  << DF_BYTE), pwd->b[5]);
+    cpu_stb_data(env, addr + (6  << DF_BYTE), pwd->b[6]);
+    cpu_stb_data(env, addr + (7  << DF_BYTE), pwd->b[7]);
+    cpu_stb_data(env, addr + (8  << DF_BYTE), pwd->b[8]);
+    cpu_stb_data(env, addr + (9  << DF_BYTE), pwd->b[9]);
+    cpu_stb_data(env, addr + (10 << DF_BYTE), pwd->b[10]);
+    cpu_stb_data(env, addr + (11 << DF_BYTE), pwd->b[11]);
+    cpu_stb_data(env, addr + (12 << DF_BYTE), pwd->b[12]);
+    cpu_stb_data(env, addr + (13 << DF_BYTE), pwd->b[13]);
+    cpu_stb_data(env, addr + (14 << DF_BYTE), pwd->b[14]);
+    cpu_stb_data(env, addr + (15 << DF_BYTE), pwd->b[15]);
+#else
+    cpu_stb_data(env, addr + (7  << DF_BYTE), pwd->b[0]);
+    cpu_stb_data(env, addr + (6  << DF_BYTE), pwd->b[1]);
+    cpu_stb_data(env, addr + (5  << DF_BYTE), pwd->b[2]);
+    cpu_stb_data(env, addr + (4  << DF_BYTE), pwd->b[3]);
+    cpu_stb_data(env, addr + (3  << DF_BYTE), pwd->b[4]);
+    cpu_stb_data(env, addr + (2  << DF_BYTE), pwd->b[5]);
+    cpu_stb_data(env, addr + (1  << DF_BYTE), pwd->b[6]);
+    cpu_stb_data(env, addr + (0  << DF_BYTE), pwd->b[7]);
+    cpu_stb_data(env, addr + (15 << DF_BYTE), pwd->b[8]);
+    cpu_stb_data(env, addr + (14 << DF_BYTE), pwd->b[9]);
+    cpu_stb_data(env, addr + (13 << DF_BYTE), pwd->b[10]);
+    cpu_stb_data(env, addr + (12 << DF_BYTE), pwd->b[11]);
+    cpu_stb_data(env, addr + (11 << DF_BYTE), pwd->b[12]);
+    cpu_stb_data(env, addr + (10 << DF_BYTE), pwd->b[13]);
+    cpu_stb_data(env, addr + (9  << DF_BYTE), pwd->b[14]);
+    cpu_stb_data(env, addr + (8  << DF_BYTE), pwd->b[15]);
+#endif
+#endif
+}
+
+void helper_msa_st_h(CPUMIPSState *env, uint32_t wd,
+                     target_ulong addr)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    int mmu_idx = cpu_mmu_index(env, false);
+
+    MEMOP_IDX(DF_HALF)
+    ensure_writable_pages(env, addr, mmu_idx, GETPC());
+#if !defined(CONFIG_USER_ONLY)
+#if !defined(HOST_WORDS_BIGENDIAN)
+    helper_ret_stw_mmu(env, addr + (0 << DF_HALF), pwd->h[0], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (1 << DF_HALF), pwd->h[1], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (2 << DF_HALF), pwd->h[2], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (3 << DF_HALF), pwd->h[3], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (4 << DF_HALF), pwd->h[4], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (5 << DF_HALF), pwd->h[5], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (6 << DF_HALF), pwd->h[6], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (7 << DF_HALF), pwd->h[7], oi, GETPC());
+#else
+    helper_ret_stw_mmu(env, addr + (3 << DF_HALF), pwd->h[0], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (2 << DF_HALF), pwd->h[1], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (1 << DF_HALF), pwd->h[2], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (0 << DF_HALF), pwd->h[3], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (7 << DF_HALF), pwd->h[4], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (6 << DF_HALF), pwd->h[5], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (5 << DF_HALF), pwd->h[6], oi, GETPC());
+    helper_ret_stw_mmu(env, addr + (4 << DF_HALF), pwd->h[7], oi, GETPC());
+#endif
+#else
+#if !defined(HOST_WORDS_BIGENDIAN)
+    cpu_stw_data(env, addr + (0 << DF_HALF), pwd->h[0]);
+    cpu_stw_data(env, addr + (1 << DF_HALF), pwd->h[1]);
+    cpu_stw_data(env, addr + (2 << DF_HALF), pwd->h[2]);
+    cpu_stw_data(env, addr + (3 << DF_HALF), pwd->h[3]);
+    cpu_stw_data(env, addr + (4 << DF_HALF), pwd->h[4]);
+    cpu_stw_data(env, addr + (5 << DF_HALF), pwd->h[5]);
+    cpu_stw_data(env, addr + (6 << DF_HALF), pwd->h[6]);
+    cpu_stw_data(env, addr + (7 << DF_HALF), pwd->h[7]);
+#else
+    cpu_stw_data(env, addr + (3 << DF_HALF), pwd->h[0]);
+    cpu_stw_data(env, addr + (2 << DF_HALF), pwd->h[1]);
+    cpu_stw_data(env, addr + (1 << DF_HALF), pwd->h[2]);
+    cpu_stw_data(env, addr + (0 << DF_HALF), pwd->h[3]);
+    cpu_stw_data(env, addr + (7 << DF_HALF), pwd->h[4]);
+    cpu_stw_data(env, addr + (6 << DF_HALF), pwd->h[5]);
+    cpu_stw_data(env, addr + (5 << DF_HALF), pwd->h[6]);
+    cpu_stw_data(env, addr + (4 << DF_HALF), pwd->h[7]);
+#endif
+#endif
 }
 
+void helper_msa_st_w(CPUMIPSState *env, uint32_t wd,
+                     target_ulong addr)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    int mmu_idx = cpu_mmu_index(env, false);
+
+    MEMOP_IDX(DF_WORD)
+    ensure_writable_pages(env, addr, mmu_idx, GETPC());
 #if !defined(CONFIG_USER_ONLY)
-MSA_ST_DF(DF_BYTE,   b, helper_ret_stb_mmu, oi, GETPC())
-MSA_ST_DF(DF_HALF,   h, helper_ret_stw_mmu, oi, GETPC())
-MSA_ST_DF(DF_WORD,   w, helper_ret_stl_mmu, oi, GETPC())
-MSA_ST_DF(DF_DOUBLE, d, helper_ret_stq_mmu, oi, GETPC())
+#if !defined(HOST_WORDS_BIGENDIAN)
+    helper_ret_stl_mmu(env, addr + (0 << DF_WORD), oi, GETPC(), pwd->w[0]);
+    helper_ret_stl_mmu(env, addr + (1 << DF_WORD), oi, GETPC(), pwd->w[1]);
+    helper_ret_stl_mmu(env, addr + (2 << DF_WORD), oi, GETPC(), pwd->w[2]);
+    helper_ret_stl_mmu(env, addr + (3 << DF_WORD), oi, GETPC(), pwd->w[3]);
 #else
-MSA_ST_DF(DF_BYTE,   b, cpu_stb_data)
-MSA_ST_DF(DF_HALF,   h, cpu_stw_data)
-MSA_ST_DF(DF_WORD,   w, cpu_stl_data)
-MSA_ST_DF(DF_DOUBLE, d, cpu_stq_data)
+    helper_ret_stl_mmu(env, addr + (1 << DF_WORD), oi, GETPC(), pwd->w[0]);
+    helper_ret_stl_mmu(env, addr + (0 << DF_WORD), oi, GETPC(), pwd->w[1]);
+    helper_ret_stl_mmu(env, addr + (3 << DF_WORD), oi, GETPC(), pwd->w[2]);
+    helper_ret_stl_mmu(env, addr + (2 << DF_WORD), oi, GETPC(), pwd->w[3]);
+#endif
+#else
+#if !defined(HOST_WORDS_BIGENDIAN)
+    cpu_stl_data(env, addr + (0 << DF_WORD), pwd->w[0]);
+    cpu_stl_data(env, addr + (1 << DF_WORD), pwd->w[1]);
+    cpu_stl_data(env, addr + (2 << DF_WORD), pwd->w[2]);
+    cpu_stl_data(env, addr + (3 << DF_WORD), pwd->w[3]);
+#else
+    cpu_stl_data(env, addr + (1 << DF_WORD), pwd->w[0]);
+    cpu_stl_data(env, addr + (0 << DF_WORD), pwd->w[1]);
+    cpu_stl_data(env, addr + (3 << DF_WORD), pwd->w[2]);
+    cpu_stl_data(env, addr + (2 << DF_WORD), pwd->w[3]);
+#endif
 #endif
+}
+
+void helper_msa_st_d(CPUMIPSState *env, uint32_t wd,
+                     target_ulong addr)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    int mmu_idx = cpu_mmu_index(env, false);
+
+    MEMOP_IDX(DF_DOUBLE)
+    ensure_writable_pages(env, addr, mmu_idx, GETPC());
+#if !defined(CONFIG_USER_ONLY)
+    helper_ret_stq_mmu(env, addr + (0 << DF_DOUBLE), pwd->d[0], oi, GETPC());
+    helper_ret_stq_mmu(env, addr + (1 << DF_DOUBLE), pwd->d[1], oi, GETPC());
+#else
+    cpu_stq_data(env, addr + (0 << DF_DOUBLE), pwd->d[0]);
+    cpu_stq_data(env, addr + (1 << DF_DOUBLE), pwd->d[1]);
+#endif
+}
 
 void helper_cache(CPUMIPSState *env, target_ulong addr, uint32_t op)
 {
diff --git a/target/mips/translate.c b/target/mips/translate.c
index f96c0d0..dd706ad 100644
--- a/target/mips/translate.c
+++ b/target/mips/translate.c
@@ -28297,20 +28297,73 @@ static void gen_msa_elm_df(CPUMIPSState *env, DisasContext *ctx, uint32_t df,
             generate_exception_end(ctx, EXCP_RI);
             break;
         }
+        if ((MASK_MSA_ELM(ctx->opcode) == OPC_COPY_U_df) &&
+              (df == DF_WORD)) {
+            generate_exception_end(ctx, EXCP_RI);
+            break;
+        }
 #endif
         switch (MASK_MSA_ELM(ctx->opcode)) {
         case OPC_COPY_S_df:
             if (likely(wd != 0)) {
-                gen_helper_msa_copy_s_df(cpu_env, tdf, twd, tws, tn);
+                switch (df) {
+                case DF_BYTE:
+                    gen_helper_msa_copy_s_b(cpu_env, twd, tws, tn);
+                    break;
+                case DF_HALF:
+                    gen_helper_msa_copy_s_h(cpu_env, twd, tws, tn);
+                    break;
+                case DF_WORD:
+                    gen_helper_msa_copy_s_w(cpu_env, twd, tws, tn);
+                    break;
+#if defined(TARGET_MIPS64)
+                case DF_DOUBLE:
+                    gen_helper_msa_copy_s_d(cpu_env, twd, tws, tn);
+                    break;
+#endif
+                default:
+                    assert(0);
+                }
             }
             break;
         case OPC_COPY_U_df:
             if (likely(wd != 0)) {
-                gen_helper_msa_copy_u_df(cpu_env, tdf, twd, tws, tn);
+                switch (df) {
+                case DF_BYTE:
+                    gen_helper_msa_copy_u_b(cpu_env, twd, tws, tn);
+                    break;
+                case DF_HALF:
+                    gen_helper_msa_copy_u_h(cpu_env, twd, tws, tn);
+                    break;
+#if defined(TARGET_MIPS64)
+                case DF_WORD:
+                    gen_helper_msa_copy_u_w(cpu_env, twd, tws, tn);
+                    break;
+#endif
+                default:
+                    assert(0);
+                }
             }
             break;
         case OPC_INSERT_df:
-            gen_helper_msa_insert_df(cpu_env, tdf, twd, tws, tn);
+            switch (df) {
+            case DF_BYTE:
+                gen_helper_msa_insert_b(cpu_env, twd, tws, tn);
+                break;
+            case DF_HALF:
+                gen_helper_msa_insert_h(cpu_env, twd, tws, tn);
+                break;
+            case DF_WORD:
+                gen_helper_msa_insert_w(cpu_env, twd, tws, tn);
+                break;
+#if defined(TARGET_MIPS64)
+            case DF_DOUBLE:
+                gen_helper_msa_insert_d(cpu_env, twd, tws, tn);
+                break;
+#endif
+            default:
+                assert(0);
+            }
             break;
         }
         break;